AD9546/PCBZ [ADI]
Dual DPLL Digitized Clock Synchronizer;型号: | AD9546/PCBZ |
厂家: | ADI |
描述: | Dual DPLL Digitized Clock Synchronizer |
文件: | 总205页 (文件大小:3061K) |
中文: | 中文翻译 | 下载: | 下载PDF数据表文档文件 |
Dual DPLL Digitized Clock Synchronizer
Data Sheet
AD9546
FEATURES
APPLICATIONS
Digitized clock transport subsystem
5G timing transport high precision synchronization
9 independent UTS blocks (time stamp egress ports)
Global positioning system (GPS), precision time protocol
2 independent IUTS blocks (time stamp ingress ports)
Dual DPLL synchronizes 1 Hz to 750 MHz physical layer
clocks, providing frequency translation with jitter
cleaning of noisy references
(PTP) (IEEE 1588), and synchronous Ethernet (SyncE) jitter
cleanup and synchronization
Optical transport networks (OTN), synchronous digital
hierarchy (SDH), and macro and small cell base stations
Small base station clocking (baseband and radio)
Stratum 2, Stratum 3e, and Stratum 3 holdover, jitter
cleanup, and phase transient control
JESD204B support for analog-to-digital converter (ADC) and
digital-to-analog converter (DAC) clocking
Carrier Ethernet
Complies with ITU-T G.8262 and Telcordia GR-253
Supports Telcordia GR-1244, ITU-T G.812, ITU-T G.813,
ITU-T G.823, ITU-T G.824, ITU-T G.825, and ITU-T G.8273.2
Continuous frequency monitoring and reference validation
for frequency deviation as low as 50 ppb (5 × 10−8
)
Both DPLLs feature a 24-bit fractional divider with 24-bit
programmable modulus
Programmable digital loop filter bandwidth: 0.0001 Hz to
1850 Hz
2 independent, programmable auxiliary NCOs (1 Hz to
65,535 Hz, resolution < 1.37 pHz), suitable for IEEE 1588
Version 2 servo feedback in PTP applications
Automatic and manual holdover and reference switchover,
providing zero delay, hitless, or phase buildout operation
Programmable priority-based reference switching with
manual, automatic revertive, and automatic nonrevertive
modes supported
GENERAL DESCRIPTION
The AD9546 incorporates digitized clocking technology that
efficiently transports and distributes clock signals in systems.
Digitized clocking allows the design of flexible and scalable
clock transport systems with well controlled phase (time)
alignment. These characteristics make the AD9546 a leading
choice for the design of network equipment that must meet the
synchronization requirements for IEEE® 1588™ boundary clocks
per ITU-T G.8273.2 Class D. Digitized clocking is also relevant
in applications requiring the accurate transport of frequency
and phase to multiple usage endpoints (for example,
5 pairs of clock output pins with each pair useable as
differential LVDS/HCSL/CML or as 2 single-ended outputs
(1 Hz to 500 MHz)
distributing synchronized system reference (SYSREF) clocks to
an array of ADC channels).
2 differential or 8 single-ended input references
Crosspoint mux interconnects reference inputs to PLLs
Supports embedded (modulated) input/output clock signals
Fast DPLL locking modes
Provides internal capability to combine the low phase noise
of a crystal resonator or crystal oscillator with the
frequency stability and accuracy of a TCXO or OCXO
External EEPROM support for autonomous initialization
Single 1.8 V power supply operation with internal regulation
Built in temperature monitor and alarm and temperature
compensation for enhanced zero delay performance
The AD9546 supports existing and emerging International
Telecommunications Union (ITU) standards for the delivery of
frequency, phase, and time of day over service provider packet
networks (ITU-T G.8262, ITU-T G.812, ITU-T G.813, ITU-T
G.823, ITU-T G.824, ITU-T G.825, and ITU-T G.8273.2).
The 10 clock outputs of the AD9546 synchronize to any one of
up to eight input references. The digital phase-locked loops
(DPLLs) reduce timing jitter associated with the external
references, and the analog phase-locked loops (APLLs) provide
frequency translation with low jitter output clocks. The digitally
controlled loop and holdover circuitry continuously generate a
low jitter output signal, even when all reference inputs fail.
The AD9546 is available in a 48-lead LFCSP (7 mm × 7 mm)
package and operates over the −40°C to +85°C temperature range.
Throughout this data sheet, a single function of a multifunction
pin name may be referenced when only that function is relevant
(for example, M5 for SDO/M5).
Rev. 0
Document Feedback
Information furnished by Analog Devices is believed to be accurate and reliable. However, no
responsibility is assumed by Analog Devices for its use, nor for any infringements of patents or other
rights of third parties that may result from its use. Specifications subject to change without notice.
No license is granted by implication or otherwise under any patent or patent rights of Analog
Devices. Trademarks and registeredtrademarks are the property oftheir respective owners.
One Technology Way, P.O. Box 9106, Norwood, MA 02062-9106, U.S.A.
Tel: 781.329.4700
Technical Support
©2020 Analog Devices, Inc. All rights reserved.
www.analog.com
AD9546
Data Sheet
TABLE OF CONTENTS
Features.............................................................................................. 1
Logic Output Specifications (M0 to CSB/M6 Pins) .............. 31
Serial Port Specifications........................................................... 32
Jitter Generation (Random Jitter)............................................ 34
Phase Noise ................................................................................. 35
Absolute Maximum Ratings......................................................... 38
Thermal Resistance.................................................................... 38
ESD Caution ............................................................................... 38
Pin Configuration and Function Descriptions .......................... 39
Typical Performance Characteristics .......................................... 42
Terminology.................................................................................... 47
Theory of Operation ...................................................................... 48
Input/Output Termination Recommendations......................... 49
System Clock Inputs .................................................................. 49
Reference Clock Inputs ............................................................. 49
Clock Outputs............................................................................. 50
Digitized Clocking.......................................................................... 52
Overview...................................................................................... 52
System Clock PLL Component ................................................ 52
Common Clock DPLL Component ........................................ 52
Physical Clock Converter.......................................................... 52
Physical Clock Generator.......................................................... 52
Common Clock Synchronizer Component ........................... 54
Applications ...................................................................................... 1
General Description......................................................................... 1
Revision History ............................................................................... 5
Functional Block Diagram .............................................................. 6
Specifications .................................................................................... 7
Operating Temperature............................................................... 7
Supply Voltage.............................................................................. 7
Supply Current ............................................................................. 7
Power Dissipation ........................................................................ 8
System Clock Inputs, XOA and XOB........................................ 9
Reference Inputs......................................................................... 10
Reference to Reference Coupling............................................. 12
REFx to REFx Input Timing Skew........................................... 14
REFx to Auxiliary REFx Input Timing Skew ......................... 15
Reference Monitors.................................................................... 16
Distribution Clock Outputs...................................................... 17
Output to Output Timing Skew............................................... 19
Output Timing Skew Between Mx Pins and OUTxyP and/or
OUTxyN Pins ............................................................................. 20
Time Duration of Digital Functions........................................ 20
DPLL0 and DPLL1 Specifications............................................ 21
DPLL Lock Detection Specifications....................................... 21
DPLL Phase Characteristics...................................................... 22
DPLL Propagation Delay .......................................................... 23
DPLL Propagation Delay Variation ........................................ 24
Holdover Specifications............................................................. 25
Analog PLL (APLL0 and APLL1) Specifications................... 25
Output Channel Divider Specifications.................................. 25
Time to Digital Converter (TDC) Specifications .................. 26
Auxiliary NCO Specifications .................................................. 26
Common Clock DPLL Specifications...................................... 26
Common Clock Synchronizer (CCS) Specifications ............ 27
User Time Stampers (UTS) Specifications ............................. 27
Inverse User Time Stampers (IUTS) Specifications.............. 28
Analog Loopback (Round Trip Delay) Specifications .......... 29
Mx to Mx Pin Output Timing Skew........................................ 30
System Clock Compensation Specifications .......................... 30
Temperature Sensor Specifications ......................................... 30
Logic Input Specifications (RESETB, M0 to CSB/M6 Pins) 30
User Time Stamper (UTS) and Inverse User Timer Stamper
(IUTS) Components.................................................................. 54
A Digitized Clocking Node Example ...................................... 54
Common Clock DPLL (CCDPLL)............................................... 56
Overview...................................................................................... 56
Common Clock References ...................................................... 56
Common Clock Reference Monitor........................................ 57
CCDPLL Lock Detector ............................................................ 57
CCDPLL Loop Filter.................................................................. 57
Common Clock Reference (CCR) Period Declaration......... 57
Common Clock Reference Switchover ................................... 58
Active Status................................................................................ 58
Common Clock Synchronizer (CCS).......................................... 59
Overview...................................................................................... 59
Synchronization Source ............................................................ 60
Tagged Synchronization ........................................................... 60
Synchronization Time............................................................... 61
Adding Time Skew to the Synchronization Time................. 62
Rev. 0 | Page 2 of 205
Data Sheet
AD9546
Adding Time Offset to the Synchronization Time ................62
Synchronization Offset Refinement.........................................62
Maximum Magnitude Detection..............................................62
Latency Detection .......................................................................62
Synchronization Slew Limiter...................................................63
Restart...........................................................................................63
Synchronization Guard..............................................................63
Initial Synchronization...............................................................64
Ready Status.................................................................................64
User Time Stamper (UTS).............................................................65
Overview ......................................................................................65
UTS Channel Control Registers Addresses.............................65
UTS Channel Enable ..................................................................65
UTS Channel Time Stamp Source............................................66
UTS Channel Time Offset .........................................................66
UTS Channel Tag........................................................................66
UTS Channel Time Format.......................................................66
UTS Channel Status Flag Assignment .....................................66
UTS Readback FIFO...................................................................67
Inverse User Time Stamper (IUTS)..............................................69
Overview ......................................................................................69
Parameter Destination Selection ..............................................70
Time Code and Format..............................................................70
IUTS Time Offset........................................................................70
IUTS Status ..................................................................................70
IUTS Restart ................................................................................70
IUTS Bypass Lock Option .........................................................70
Declaring an IUTS Invalid.........................................................70
Analog Clock Loopback.................................................................71
Overview ......................................................................................71
Analog Loopback Feature..........................................................72
Measuring Round Trip Delay ...................................................72
Excessive Round Trip Delay......................................................73
System Clock PLL ...........................................................................74
System Clock PLL Overview .....................................................74
System Clock Input Frequency Declaration ...........................74
System Clock Source ..................................................................74
Prescale Divider...........................................................................75
Feedback Divider ........................................................................75
System Clock PLL Output Frequency......................................75
System Clock PLL Lock Detector .............................................75
System Clock Stability Timer....................................................75
System Clock Calibration ..........................................................75
System Clock Stability Compensation.....................................76
Reference Clock Input Resources.................................................77
Reference Receivers ........................................................................78
Reference Receivers Overview ..................................................78
Single-Ended Mode ....................................................................78
Differential Mode .......................................................................79
Reference Dividers (R Dividers)...................................................80
Reference Monitor..........................................................................81
Reference Monitor Overview....................................................81
Reference Monitor Base Period ................................................82
Reference Monitor Status Indicators .......................................82
Reference Monitor Controls .....................................................82
Discontinuity Detection.............................................................85
Reference Period Jitter Estimation...........................................85
Reference Monitor Decision Time...........................................85
Reference Validation..................................................................85
Reference Monitor Reset ...........................................................86
Reference Demodulator .................................................................87
Reference Demodulator Overview...........................................87
Interaction Between Demodulator and Reference Monitor.87
Modulated Cycles and Modulation Events.............................88
Balanced and Unbalanced Modulation ...................................88
Enabling a Demodulator............................................................88
Modulation Polarity Detection.................................................88
Demodulator Sensitivity ............................................................89
Demodulator Persistence...........................................................89
Demodulator Bandwidth...........................................................89
Demodulator Synchronization .................................................90
Distribution Clock Output Drivers..............................................91
Distribution Clock Output Drivers Overview........................91
Output Current Control ............................................................91
Output Mode Control................................................................91
Output Driver Configurations..................................................92
Output Driver Reset ...................................................................93
Output Muting............................................................................93
Distribution Dividers (Q Dividers)..............................................94
Distribution Dividers Overview ...............................................94
Q Divider Clock Source Selection ............................................95
Integer Division ..........................................................................95
Rev. 0 | Page 3 of 205
AD9546
Data Sheet
Half Integer Division ................................................................. 96
Q Divider Reset........................................................................... 96
Q Divider Constraints ............................................................... 96
Hitless/Zero Delay Feedback.................................................... 96
Distribution Phase Offset Control............................................... 97
Output Phase Offset Overview................................................. 97
Initial Phase Offset..................................................................... 97
Subsequent Phase Offsets.......................................................... 97
Distribution N Shot/PRBS Output Clocking ........................... 100
N Shot/PRBS Clocking Overview .......................................... 100
Randomized Clock (PRBS)..................................................... 101
N Shot (JESD204B and Gapped Clocking)........................... 101
Distribution Embedded Output Clock Modulation................ 105
Modulation Controller Overview .......................................... 105
Modulation Magnitude ........................................................... 105
Modulation Period................................................................... 107
Balanced and Unbalanced Modulation................................. 107
Modulation Synchronization.................................................. 109
Modulation Trigger.................................................................. 110
Distribution Output Clock Synchronization............................ 111
Synchronization Overview...................................................... 111
Output Signals After Power-Up or Reset.............................. 111
Manual Synchronization Trigger........................................... 111
Autoreconfiguration Synchronization Trigger.................... 111
Autosynchronization Trigger................................................. 112
Reference Synchronization ..................................................... 112
Frequency Translation Loops..................................................... 113
Frequency Translation Loops Overview............................... 113
Translation Profiles.................................................................. 113
Profile Enable............................................................................ 114
Profile Priority .......................................................................... 114
Input Reference Source Selection .......................................... 114
Active Reference Indication.................................................... 114
Translation Modes ................................................................... 115
Phase Buildout Mode............................................................... 116
Internal Zero Delay (Hitless) Mode ...................................... 117
External Zero Delay (Hitless) Mode...................................... 118
Source Profiles .............................................................................. 119
Source Profiles Overview ........................................................ 119
DPLL Phase/Frequency Lock Detector................................. 119
Phase Step Limit ....................................................................... 119
Skew Adjustment ..................................................................... 120
Initial Phase Skew Refinement Steps..................................... 120
Digital PLL (DPLL)...................................................................... 122
DPLL Overview ........................................................................ 122
DPLL Loop Controller ............................................................ 122
DPLL Feedback Divider (N Divider) .................................... 123
DPLL Loop Filter ..................................................................... 124
DPLL NCO................................................................................ 126
NCO Gain Tuning Word Filter Bandwidth......................... 127
DPLL Lock Detectors .............................................................. 128
Freerun Tuning Word............................................................. 129
DPLL Fast Acquisition Options............................................. 130
DPLL Phase Offset Control .................................................... 132
Tuning Word Offset Clamp ................................................... 133
Phase Slew Rate Limit.............................................................. 134
Tuning Word History.............................................................. 135
Delay Compensation ............................................................... 138
Time Stamp Tagging Options ................................................ 140
Cascaded DPLL Configuration .................................................. 141
Caveats of Cascaded DPLL Operation.................................. 143
Analog PLL (APLL) ..................................................................... 144
APLL Overview ........................................................................ 144
Voltage Controlled Oscillator (VCO)................................... 144
APLL Feedback Divider (M Divider).................................... 145
Phase Frequency Detector (PFD) .......................................... 145
Charge Pump............................................................................ 145
APLL Loop Filter...................................................................... 146
Reference Switching..................................................................... 147
Reference Switching Overview............................................... 147
Forced Freerun Mode.............................................................. 147
Forced Holdover Mode........................................................... 147
Manual/Automatic Translation Profile Selection ............... 147
Time to Digital Converter (TDC).............................................. 154
Time Stamps ................................................................................. 155
Time Stamps Overview ........................................................... 155
Digital Crosspoint Mux........................................................... 155
Tagged Time Stamps ............................................................... 156
User Time Stamp Processor (UTSP)......................................... 157
UTSP Overview ........................................................................ 157
Reading User Time Stamps .................................................... 157
Interpreting User Time Stamps ............................................. 158
Rev. 0 | Page 4 of 205
Data Sheet
AD9546
Tagged Time Stamps ............................................................... 159
UTSP with System Clock Compensation............................. 159
Timing Skew Measurements Using Two TDCs ...................... 160
Tagged Skew Measurement Time Stamps............................ 161
Auxiliary NCOs............................................................................ 162
Auxiliary NCO Overview ....................................................... 162
Auxiliary NCO Frequency...................................................... 162
Auxiliary NCO Phase Offset .................................................. 163
Auxiliary NCO Phase Slew Limit .......................................... 164
Elapsed Time Adjustment ...................................................... 164
Auxiliary NCO Time Stamps ................................................. 164
Auxiliary NCO Pulse Output................................................. 165
Temperature Sensor..................................................................... 167
Temperature Sensor Overview .............................................. 167
Temperature Source Selection ............................................... 167
Internal Temperature Sensor ................................................. 167
External Temperature Source ................................................ 168
System Clock Compensation...................................................... 169
System Clock Compensation Overview................................ 169
Compensation Method 1 ........................................................ 171
Compensation Method 2 ........................................................ 174
Compensation Method 3 ........................................................ 175
Integrated Compensation Subsystem ................................... 177
System Clock Compensation Programming Registers....... 178
Status and Control Pins............................................................... 180
Status and Control Pins Overview......................................... 180
Multifunction Pins at Reset or Power-Up............................ 180
Defining an Mx Pin for Status or Control............................ 181
Status Functionality..................................................................181
Control Functionality...............................................................182
Interrupt Request (IRQ) ..............................................................183
IRQ Overview............................................................................183
IRQ Status Bits ..........................................................................183
IRQ Mask Bits ...........................................................................183
IRQ Clear Bits ...........................................................................183
Watchdog Timer...........................................................................185
EEPROM Usage ............................................................................186
EEPROM Overview..................................................................186
EEPROM Controller General Operation ..............................186
EEPROM Instruction Set.........................................................187
Multidevice Support.................................................................189
Applications Information............................................................191
Digitized Clocking Application ..............................................191
Clock Propagation Delay Measurement................................193
Time Stamp Measurement ......................................................194
IEEE 1588 Servo........................................................................195
Initialization Sequence.................................................................196
Serial Control Port........................................................................199
Serial Control Port Overview..................................................199
SPI/I²C Port Selection..............................................................199
SPI Serial Port Operation ........................................................199
I²C Serial Port Operation.........................................................202
Outline Dimensions .....................................................................205
Ordering Guide.........................................................................205
REVISION HISTORY
10/2020—Revision 0: Initial Version
Rev. 0 | Page 5 of 205
AD9546
Data Sheet
FUNCTIONAL BLOCK DIAGRAM
REFERENCE
INPUTS (4)
DISTRIBUTION
OUTPUTS
PLL0
APLL0
DPLL0
OUT0AP/N
OUT0BP/N
OUT0CP/N
REFA
REFAA
÷R
TDC
DPLL NCO
PLL VCO
÷2
÷Q
REFB
REFBB
REFERENCE
MONITORS
INTERNAL ZERO DELAY
PLL1
REFERENCE
SWITCHING
CONTROL
DPLL1
APLL1
OUT1AP/N
OUT1BP/N
DPLL NCO
PLL VCO
÷2
÷Q
AUXILIARY
REFERENCE (4)
DISTRIBUTION
OUTPUTS
COMMON CLOCK
SYNCHRONIZER
÷R
TDC
SYSTEM
CLOCK PLL
XOA/
XOB
SYSTEM
CLOCK
USER TIME
STAMP (9)
AUXILIARY
NCO’S (2)
INVERSE USER
TIME STAMP (2)
SYSTEM CLOCK
COMPENSATION
DIGITAL
CROSS
POINT
MUX
2
SERIAL PORT (SPI/I C)
AND EEPROM
TEMPERATURE
SENSOR
STATUS AND
CONTROL PINS
CONTROLLER
AD9546
M0 TO M6
SERIAL PORT
APLL VCO FREQUENCY RANGE
APLL0: 2424MHz TO 3232MHz
APLL1: 3232MHz TO 4040MHz
EEPROM
(OPTIONAL)
Figure 1.
Rev. 0 | Page 6 of 205
Data Sheet
AD9546
SPECIFICATIONS
The minimum and maximum values apply for the full range of supply voltage and operating temperature variations. Typical values apply
for VDD = 1.8 V and TA= 25°C, unless otherwise noted.
OPERATING TEMPERATURE
Table 1.
Parameter
Min Typ Max
−40 +85
Unit
Test Conditions/Comments
OPERATING TEMPERATURE
Ambient Temperature1
Die Temperature2
°C
+105 °C
Internal temperature sensor readings exceeding 105°C indicate
excessive die temperature
1 See the Thermal Resistance section for additional information.
2 The maximum die temperature supports the performance stated in Table 2 to Table 39. The maximum operating die temperature takes precedence over the
maximum ambient operating temperature. Use the internal temperature sensor to measure the AD9546 die temperature (see the Temperature Sensor section).
SUPPLY VOLTAGE
Table 2.
Parameter
Min Typ Max
Unit
Test Conditions/Comments
SUPPLY VOLTAGE
VDDIOA and VDDIOB
VDD
1.71 1.8
1.71 1.8
3.465
1.89
V
V
1.8 V, 2.5 V, and 3.3 V operation supported
SUPPLY CURRENT
The minimum, typical, and maximum supply current values given in Table 3 are for the minimum, typical, and maximum supply voltage
specifications in Table 2.
Table 3.
Parameter
Min
Typ
Max
Unit
Test Conditions/Comments
SUPPLY CURRENT FOR TYPICAL
CONFIGURATION
Uses the Typical Configuration conditions in Table 4
VDDIOx Current (IVDDIOx
VDD Current (IVDD
)
5
325
8
330
mA
mA
Aggregate current for all VDDIOx pins (where x = A or B)
Aggregate current for all VDD pins
)
SUPPLY CURRENT FOR LINE CARD
CONFIGURATION
Uses the Line Card Configuration conditions in Table 4; see the
Digitized Clocking Application section for line card information
IVDDIOx
IVDD
5
325
8
330
mA
mA
Aggregate current for all VDDIOx pins (where x = A or B)
Aggregate current for all VDD pins
SUPPLY CURRENT FOR TIMING CARD
CONFIGURATION
Uses the Timing Card Configuration conditions in Table 4; see
the Digitized Clocking Application section for timing card
information
IVDDIOx
IVDD
5
325
8
330
mA
mA
Aggregate current for all VDDIOx pins (where x = A or B)
Aggregate current for all VDD pins
SUPPLY CURRENT FOR ALL BLOCKS
RUNNING CONFIGURATION
Uses the All Blocks Running Configuration conditions in Table 4
IVDDIOx
IVDD
5
410
8
440
mA
mA
Aggregate current for all VDDIOx pins (where x = A or B)
Aggregate current for all VDD pins
Rev. 0 | Page 7 of 205
AD9546
Data Sheet
POWER DISSIPATION
The typical values apply for VDD = 1.8 V, and the maximum values apply for VDD = 1.89 V.
Table 4.
Parameter
Min
Typ
585
Max Unit
Test Conditions/Comments
POWER DISSIPATION
Typical Configuration
fOSC is the frequency at the XOA/XOB pins
fOSC = 49.152 MHz crystal; two active DPLLs;
760
mW
two 19.44 MHz input references in differential mode;
two ac-coupled PLL Channel 0 (PLL0) current mode logic
(CML) output drivers at 245.76 MHz; and two PLL Channel 1
(PLL1) CML output drivers at 156.25 MHz
Line Card Configuration
585
585
760
mW
fOSC = 52 MHz crystal; two active DPLLs (hitless mode); five
single-ended input references with two operating at 10 MHz
and three with active 1 Hz reference demodulation operating
at 50 MHz; five HCSL (15 mA) output drivers operating at
10 MHz; one auxiliary NCO operating at 500 Hz; two inverse
user timer stampers (IUTSs) operating at 500 Hz; auxiliary
DPLL enabled; see the Digitized Clocking Application section
for line card information
Timing Card Configuration
760
mW
fOSC = 52 MHz crystal; two active DPLLs (one hitless mode, one
phase build-out mode); eight single-ended input references
with four operating at 10 MHz and four with active 1 Hz
reference demodulation operating at 50 MHz; five HCSL
(15 mA) output drivers with two operating at 10 MHz and
three with active 1 Hz embedded clock modulation operating
at 50 MHz; two auxiliary NCOs with one operating at 500 Hz
and one operating at 1 Hz; one IUTS operating at 500 Hz;
auxiliary DPLL enabled; see the Digitized Clocking Application
section for timing card information
All Blocks Running Configuration
745
125
795
mW
mW
fOSC = 49.152 MHz crystal; two active DPLLs;
two 19.44 MHz input references in differential mode;
three ac-coupled PLL0 HCSL output drivers at 400 MHz; and
two PLL1 HCSL output drivers at 400 MHz
Based on the Typical Configuration specification, but with
Register 0x2000, Bit 0 = 1
Full Power-Down
Incremental Power Dissipation
Based on the Typical Configuration specification; the values in
this section indicate the change in power due to the indicated
operation relative to the Typical Configuration specification
Complete DPLL/APLL On vs. Off
160
mW
Change in dissipated power relative to the Typical Configuration
specification; the powered down blocks consist of one reference
input, one DPLL, one APLL, two channel dividers, and two
output drivers
Input Reference On vs. Off
Differential (AC-Coupled Mode)
Differential (DC-Coupled Mode)
Single-Ended
20
21
13
mW
mW
mW
Reference frequency (fREF) = 19.44 MHz (see Figure 36)
fREF = 19.44 MHz (see Figure 37)
fREF = 19.44 MHz
Output Distribution Driver On vs. Off
15 mA Mode
12.5 mA Mode
7.5 mA Mode
Auxiliary DPLL On vs. Off
At 156.25 MHz
28
23
15
1
mW
mW
mW
mW
mW
Auxiliary Numerically Controlled
Oscillator (NCO) to Mx Pin On vs. Off
1
Fundamental set to 50 kHz
Auxiliary Time to Digital Converter
(TDC) Input from Mx Pin On vs. Off
1
mW
Input frequency (fIN) = 10 MHz, auxiliary TDC rate = 200 kHz
Rev. 0 | Page 8 of 205
Data Sheet
AD9546
SYSTEM CLOCK INPUTS, XOA AND XOB
Table 5.
Parameter
Min
Typ
Max
Unit
Test Conditions/Comments
SYSTEM CLOCK MULTIPLIER
Output Frequency Range
2250
2415 MHz
The frequency range of the internal voltage controlled
oscillator (VCO) places limits on the choice of the system
clock input frequency
Phase Frequency Detector (PFD) Rate
SYSTEM CLOCK INPUT DIRECT PATH
20
20
300
300
MHz
MHz
The direct path requires ac coupling to the XOA and XOB
pins (see Figure 31). Do not enable the frequency doubler
when using the direct path
Input Frequency Range
Degraded phase noise performance typically occurs for
frequencies below 50 MHz appearing at the input of the
phase frequency detector of the system clock PLL; for
optimal phase noise performance, the recommendation is
to use the quartz crystal resonator path (instead of the
direct path), a crystal resonator with a resonant frequency
≥50 MHz, the frequency doubler enabled, and the device
configured to use the system clock Compensation Method 3
(see the System Clock Compensation section)
Input Duty Cycle
Self Biased Common-Mode Voltage
40
60
%
V
0.75
Internally generated
Input Voltage
High
For dc-coupled, single-ended operation
0.9
V
Low
0.5
V
Differential Input Voltage Sensitivity 250
mV p-p
Minimum voltage swing required (as measured with a
differential probe) across the XOA/XOB pins to ensure
switching between logic states; the instantaneous voltage on
either pin must not exceed 1.2 V; accommodate the single-
ended input by ac grounding the complementary input;
800 mV p-p recommended for optimal jitter performance
Slew Rate for Sinusoidal Input
Functional
Operational
System Clock Input Divider
(J Divider) Frequency
Input Resistance
>6
>31
V/µs
V/µs
MHz
Required to provide a locked and stable system clock
Required for <1 dBc phase noise degradation
100
25
5
kΩ
SYSTEM CLOCK INPUT QUARTZ
CRYSTAL RESONATOR PATH
Resonator Frequency Range
Crystal Motional Resistance
The crystal resonator path includes an optional frequency
doubler
Fundamental mode, AT cut crystal
For crystal resonant frequency > 52 MHz, the motional
resistance must not exceed 50 Ω and the crystal load
capacitance (CL) must not exceed 8 pF
80
100
MHz
Ω
Rev. 0 | Page 9 of 205
AD9546
Data Sheet
REFERENCE INPUTS
Table 6.
Parameter
Min
Typ Max
Unit
Test Conditions/Comments
DIFFERENTIAL MODE
Differential mode specifications assume ac coupling of the
input signal to the reference input pins
Frequency Range
Sinusoidal Input
Low Voltage Positive Emitter
Coupled Logic (LVPECL)
Input
750
MHz
Lower limit dependent on input slew rate
Lower limit dependent on ac coupling
1
1
750 × 106 Hz
LVDS Input
500 × 106 Hz
V/µs
Assumes an LVDS minimum of 494 mV p-p differential
amplitude; lower limit dependent on ac coupling
Required for proper reference monitor operation; output
jitter degradation can occur for slew rates < 35 V/µs
Slew Rate
>4.1
0.64
Common-Mode Input Voltage
Differential Input Amplitude
V
Internally generated self bias voltage
Peak-to-peak differential voltage swing across pins required
to ensure switching between logic levels as measured with a
differential probe; instantaneous voltage on either pin must
not exceed 1.3 V
fIN < 500 MHz
fIN = 500 MHz to 750 MHz
350
500
2100
2100
100
mV p-p
mV p-p
mV
Differential Input Voltage
Hysteresis
Input Resistance
Input Pulse Width
LVPECL
55
16
kΩ
Equivalent differential input resistance
Applies for dc coupling to an LVDS source
600
900
ps
ps
LVDS
DC-COUPLED, LVDS-COMPATIBLE
MODE
Frequency Range
Slew Rate
1
450 × 106 Hz
V/µs
>1.2
Required for proper reference monitor operation; output
jitter degradation can occur for slew rates < 35 V/µs
Common-Mode Input Voltage
Differential Input Amplitude
1.125
400
1.375
1200
V
mV p-p Differential voltage across pins required to ensure switching
between logic levels; instantaneous voltage on either pin
must not exceed the supply rails
Differential Input Voltage
Hysteresis
Input Resistance
Input Pulse Width
SINGLE-ENDED MODE
55
16
100
mV
kΩ
ns
1
Single-ended mode specifications assume dc coupling of the
input signal to the reference input pins
1.2 V CMOS
Frequency Range
Input Voltage
High, VIH
1
500 × 106 Hz
Assumes dc-coupled input signal to the reference input pins
0.78
1.38
0.42
V
V
Low, VIL
Input Resistance
Slew Rate
30
>8
kΩ
V/µs
Required for proper reference monitor operation; output
jitter degradation can occur for slew rates < 35 V/µs
Input Pulse Width
1.8 V CMOS
Frequency Range
900
1
ps
500 × 106 Hz
Assumes dc-coupled input signal to the reference input pins
Rev. 0 | Page 10 of 205
Data Sheet
AD9546
Parameter
Input Voltage
High, VIH
Min
Typ Max
Unit
Test Conditions/Comments
1.17
2.07
V
Low, VIL
0.63
V
Input Resistance
Slew Rate
30
>8
kΩ
V/µs
Required for proper reference monitor operation; output
jitter degradation can occur for slew rates <35 V/µs
Input Pulse Width
AC-Coupled
Frequency Range
Input Amplitude
900
ps
1
360
500 × 106 Hz
Applies to 1.2 V CMOS ac-coupled
Peak-to-peak single-ended voltage swing; instantaneous
voltage must not exceed 1.3 V
1200
mV
Input Impedance
Input Pulse Width
Slew Rate
15
kΩ
ps
V/µs
900
>8
Required for proper reference monitor operation; output
jitter degradation can occur for slew rates < 35 V/µs
Common-Mode Voltage
AUXILIARY REFERENCES
610
mV
Internally generated self bias voltage; applies to 1.2 V CMOS
ac-coupled
Applies to the auxiliary Reference 0 (REF0) to auxiliary
Reference 3 (REF3) inputs via appropriately configured Mx
control pins (see the Reference Clock Input Resources
section)
Frequency Range
Input Voltage
High, VIH
225
0.4
MHz
V
VDDIOx
− 0.4
Low, VIL
V
Slew rate
>8
V/µs
Required for proper reference monitor operation; output
jitter degradation can occur for slew rates < 35 V/µs; applies
for VDDIOB = 1.8 V, 2.5 V, or 3.3 V
REFERENCE DEMODULATOR
f
OUT is the Q divider output frequency; fSYS is the system clock
frequency and must be in the 2250 MHz to 2415 MHz range;
SYS = fOSC × K/J, where fOSC is the frequency of the system
f
clock oscillator connected to the XOA/XOB pins, K is the
feedback divider ratio, and J is the system clock (SYSCLK)
input scale factor; when the SYSCLK doubler is disabled, J is
the value of the J divider (1, 2, 4, or 8); when the SYSCLK
doubler is enabled, J = ½
Carrier Frequency
(Synchronization Edge =
1, 2, 3)
Band 0 and Band 1 (see the Demodulator Bandwidth
section)
Band 0
DC Balanced Modulation
Unbalanced Modulation
Band 1
0.5
0.5
40
60
MHz
MHz
DC Balanced Modulation
Unbalanced Modulation
Embedded Clock Rate
Duty Cycle Deviation
1
1
1
120
180
fCAR/8
MHz
MHz
Hz
fCAR is the unmodulated input frequency
S is the decimal value of Bits[1:0] in Register 0x0302,
Register 0x0303, Register 0x0306, or Register 0x0307 (see the
Demodulator Sensitivity section); B depends on the selected
band, where B is fSYS/3 for Band 0 or fSYS for Band 1 (see the
Demodulator Bandwidth section)
DC Balanced Modulation
Unbalanced Modulation
Polarity Detection Enabled
(5 + S)/3B
(5 + S)/2B
(5 + S)/B
1/(4fCAR
1/(4fCAR
1/(4fCAR
)
)
)
sec
sec
sec
Rev. 0 | Page 11 of 205
AD9546
Data Sheet
REFERENCE TO REFERENCE COUPLING
The reference to reference coupling measurement always involves two reference inputs: a target reference input and an aggressor
reference input. The device configuration comprises a frequency translation from the target reference input to an appropriate
OUTxyP/OUTxyN output pair. The target reference frequency is 10 MHz, whereas the aggressor reference frequency deviates from the
target reference frequency over a range of 100 ppm in steps of 0.5 ppm. The values shown in Table 7 are the levels of the worst spur
appearing at the OUTxyP/OUTxyN output pair, measured at the appropriate offset frequency relative to the aggressor reference input
frequency. To prevent the DPLL from suppressing the coupling spur, the DPLL loop bandwidth is always twice the aggressor reference
offset frequency.
Table 7.
Parameter
Min
Typ
Max
Unit
TARGET REFERENCE AND INPUT MODE
REFA 1.2 V CMOS
Aggressor Reference and Input Mode
REFAA 1.2 V CMOS
REFAA 1.8 V CMOS
REFAA Single-Ended AC-Coupled
REFA 1.8 V CMOS
−62
−59
−69
dBc
dBc
dBc
Aggressor Reference and Input Mode
REFAA 1.2 V CMOS
REFAA 1.8 V CMOS
REFAA Single-Ended AC-Coupled
REFA Single-Ended AC-Coupled
Aggressor Reference and Input Mode
REFAA 1.2 V CMOS
−65
−66
−67
dBc
dBc
dBc
−60
−58
−71
dBc
dBc
dBc
REFAA 1.8 V CMOS
REFAA Single-Ended AC-Coupled
REFA Differential AC-Coupled
Aggressor Reference and Input Mode
REFB Differential AC-Coupled
REFB Differential LVDS
REFBB 1.2 V CMOS
REFBB 1.8 V CMOS
REFBB Single-Ended AC-Coupled
REFA Differential LVDS
−76
−76
−75
−74
−76
dBc
dBc
dBc
dBc
dBc
Aggressor Reference and Input Mode
REFB Differential AC-Coupled
REFB Differential LVDS
REFBB 1.2 V CMOS
REFBB 1.8 V CMOS
−76
−76
−76
−75
−68
dBc
dBc
dBc
dBc
dBc
REFBB Single-Ended AC-Coupled
REFAA 1.2 V CMOS
Aggressor Reference and Input Mode
REFB Differential AC-Coupled
REFB Differential LVDS
REFBB 1.2 V CMOS
REFBB 1.8 V CMOS
−76
−75
−74
−76
−76
dBc
dBc
dBc
dBc
dBc
REFBB Single-Ended AC-Coupled
Rev. 0 | Page 12 of 205
Data Sheet
AD9546
Parameter
Min
Typ
Max
Unit
REFAA 1.8 V CMOS
Aggressor Reference and Input Mode
REFB Differential AC-Coupled
REFB Differential LVDS
REFBB 1.2 V CMOS
REFBB 1.8 V CMOS
REFBB Single-Ended AC-Coupled
REFAA Single-Ended AC-Coupled
Aggressor Reference and Input Mode
REFB Differential AC-Coupled
REFB Differential LVDS
REFBB 1.2 V CMOS
REFBB 1.8 V CMOS
REFBB Single-Ended AC-Coupled
REFB or REFBB Using Any Single-Ended Input Mode
Aggressor Reference (Not REFA or REFAA)
−76
−76
−76
−76
−76
dBc
dBc
dBc
dBc
dBc
−75
−75
−76
−76
−76
dBc
dBc
dBc
dBc
dBc
M0, M1, M2, or M3 with VDDIOx = 1.8 V or 3.3 V; and Connected to Any
Auxiliary Reference
REFBB Using Any Single-Ended Input Mode
Aggressor Reference (Not REFA or REFAA)
M0, M1, or M2 (Not M3) with VDDIOx = 1.8 V or 3.3 V; and Connected to Any
Auxiliary Reference
REFB and REFBB Using Any Differential Input Mode
Aggressor Reference (Not REFA or REFAA)
M0, M1, M2, or M3 with VDDIOx = 1.8 V or 3.3 V; and Connected to Any
Auxiliary Reference
M0 with VDDIOx = 1.8 V or 3.3 V
Aggressor Reference (Not REFA or REFAA)
REFB or REFBB Using Any Input Mode; or M1, M2, or M3 with VDDIOx = 1.8 V
or 3.3 V; and Connected to Any Auxiliary Reference
M0 with VDDIOx = 1.8 V or 3.3 V
Aggressor Reference (Not REFA or REFAA)
REFB or REFBB Using Any Input Mode; or M2 or M3 (Not M1) with VDDIOx =
1.8 V or 3.3 V; and Connected to Any Auxiliary Reference
M1 with VDDIOx = 1.8 V or 3.3 V
Aggressor Reference (Not REFA or REFAA)
REFB or REFBB Using Any Input Mode; or M0, M2, or M3 with VDDIOx = 1.8 V
or 3.3 V; and Connected to Any Auxiliary Reference
M2 with VDDIOx = 1.8 V or 3.3 V
Aggressor Reference (Not REFA or REFAA)
−66
−74
−70
−65
−70
−68
−67
−69
−80
dBc
dBc
dBc
dBc
dBc
dBc
dBc
dBc
dBc
REFB or REFBB Using Any Input Mode; or M1, M3, or M4 with VDDIOx = 1.8 V
or 3.3 V; and Connected to Any Auxiliary Reference
M3 with VDDIOx = 1.8 V or 3.3 V
Aggressor Reference (Not REFA or REFAA)
REFB or REFBB Using Any Input Mode; or M0, M1, or M2 with VDDIOx = 1.8 V
or 3.3 V; and Connected to Any Auxiliary Reference
M3 with VDDI0x = 3.3 V (Not 1.8 V)
Aggressor Reference (Not REFA, REFAA, REFB, or REFBB)
M0, M1, or M2 with VDDIOx = 3.3 V (Not 1.8 V); and Connected to Any
Auxiliary Reference
Rev. 0 | Page 13 of 205
AD9546
Data Sheet
REFx TO REFx INPUT TIMING SKEW
The values shown in Table 8 indicate the internal timing skew between a clock signal applied to the REFA input (base input) and the
same clock signal applied to the REFAA, REFB, or REFBB input (target input) with both clock signals time aligned at the device pins.
The top section of the table indicates timing skew for REFA configured as the CMOS 1.2 V base input against REFx (x = AA, B, BB) as
the target input with various target input configurations. The bottom section of the table captures timing skew for REFA as the base
input with input configurations other than CMOS 1.2 V against REFB CMOS 1.2 V as the target input, which establishes REFB CMOS
1.2 V as a common target for different REFA input configurations. The common target allows the user to extrapolate the skew to other
target inputs when REFA is configured for other than CMOS 1.2 V.
Table 8.
Parameter
Min
Typ
Max
Unit
TIMING SKEW—REFA CMOS 1.2 V BASE REFERENCE INPUT
Target Reference Input
REFAA CMOS 1.2 V
REFAA CMOS 1.8 V
REFAA AC-Coupled Single-Ended
REFB CMOS 1.2 V
REFBB CMOS 1.2 V
−190
−95
−70
−155
−170
−250
−190
−70
+130
+55
−15
−30
−100
+30
+45
ps
ps
ps
ps
ps
ps
ps
+330
+215
+125
+110
+50
REFB Differential, LVDS
REFB Differential, DC-Coupled
TIMING SKEW—OTHER BASE REFERENCE INPUT CONFIGURATIONS
REFA CMOS 1.8 V Base Reference Input
Target Reference Input
+270
REFB CMOS 1.2 V
REFA AC-Coupled, Single-Ended Base Reference Input
Target Reference Input
REFB CMOS 1.2 V
REFA Differential, LVDS Base Reference Input
Target Reference Input
REFB CMOS 1.2 V
REFA Differential, DC-Coupled Base Reference Input
Target Reference Input
−180
−200
40
−5
+155
+95
ps
ps
ps
ps
−55
195
+85
340
REFB CMOS 1.2 V
−145
+285
Rev. 0 | Page 14 of 205
Data Sheet
AD9546
REFx TO AUXILIARY REFx INPUT TIMING SKEW
The values shown in Table 9 indicate the internal timing skew between a clock signal applied to the REFA input (configured for CMOS
1.2 V) and the same clock signal applied to an auxiliary REFx via an Mx pin with both clock signals time aligned at the device pins. Skew
data for the various VDDIOB voltages is also captured in Table 9 because VDDIOB is the power supply for the Mx pins. Note that REFA
CMOS 1.2 V is always the base input for a target Mx and auxiliary REFx combination. Skew relative to a different input configuration for
REFA or to another REFx input (with various input configurations) can be inferred using Table 8 in conjunction with Table 9. Skew data
appears only for M0, M1, M2, and M3 because these are the only Mx pins recommended for use with the auxiliary REFx references. The
data reveals that operating with VDDIOB = 3.3 V yields lower REFx to auxiliary REFx timing skew than operating with VDDIOB = 1.8 V
or 2.5 V.
Table 9.
Parameter
Min
Typ
Max
Unit
INTERNAL TIMING SKEW
M0
VDDIOB = 1.8 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
VDDIOB = 2.5 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
VDDIOB = 3.3 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
M1
950
845
715
720
1465
1355
1240
1250
2040
1930
1820
1835
ps
ps
ps
ps
495
380
250
255
870
760
640
655
1255
1150
1045
1075
ps
ps
ps
ps
330
215
95
610
505
385
400
925
820
715
740
ps
ps
ps
ps
95
VDDIOB = 1.8 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
VDDIOB = 2.5 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
VDDIOB = 3.3 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
M2
1000
865
800
700
1485
1355
1300
1200
2050
1925
1885
1785
ps
ps
ps
ps
455
325
255
150
830
700
640
540
1265
1135
1085
990
ps
ps
ps
ps
315
185
115
35
585
460
400
310
915
795
735
635
ps
ps
ps
ps
VDDIOB = 1.8 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
1020
870
825
725
1515
1390
1325
1235
2075
1970
1890
1815
ps
ps
ps
ps
Rev. 0 | Page 15 of 205
AD9546
Data Sheet
Parameter
Min
Typ
Max
Unit
VDDIOB = 2.5 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
VDDIOB = 3.3 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
485
365
295
190
860
735
690
580
1290
1160
1135
1025
ps
ps
ps
ps
340
215
130
35
620
500
455
340
940
815
820
680
ps
ps
ps
ps
M3
VDDIOB = 1.8 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
VDDIOB = 2.5 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
VDDIOB = 3.3 V
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
1095
955
860
790
1585
1460
1380
1310
2135
2035
1970
1905
ps
ps
ps
ps
550
415
340
275
945
820
750
685
1410
1300
1235
1170
ps
ps
ps
ps
370
285
210
145
705
585
540
485
1105
960
930
890
ps
ps
ps
ps
REFERENCE MONITORS
Table 10.
Parameter
Min
Typ
Max
Unit
Test Conditions/Comments
REFERENCE MONITORS
Reference Monitor
t
PFD is the nominal phase detector period; tPFD = R/fREF, where R is
the frequency division factor determined by the R divider, and
REF is the frequency of the active reference
tPFD = R/fREF with fREF in units of MHz
f
Loss of Reference
Detection Time
Frequency Out of
Range Limit
4.9 + 0.13 ×
tPFD
µs
5 × 10−8
0.015
Δf/fREF Minimum deviation from fREF for an unfaulted reference to
become faulted; programmable with the lower limit subject to
quality of the system clock (or the source of system clock
compensation); Δf = frequency deviation from fREF; to prevent
false reference invalidation due to the inherent peak jitter of the
TDC (TDCJpk), requires fTDC < Threshold × 10−9/TDCJpk, where fTDC is
the frequency applied at the TDC input (see the Reference
Period Threshold section for the threshold parameter);
minimum = 5 × 10−8 requires fTDC < 200 Hz
Frequency In
Range Limit
(5 × 10−8) ×
(1 – SF)
0.015 ×
(1 – SF)
Δf/fREF Maximum deviation from fREF for a faulted reference to become
unfaulted; lower limit subject to quality of the system clock (or
the source of system clock compensation); SF = programmable
hysteresis scale factor (see Table 54); Δf = frequency deviation
from fREF; to prevent false reference revalidation due to TDCJpk
requires fTDC < Threshold × 10−9 × SF/TDCJpk, where fTDC is the
frequency applied at the TDC input (see the Hysteresis section
for SF); minimum = (5 × 10−8) × (1 – SF) requires fTDC < SF × 200 Hz
Validation Timer
Excessive Jitter Alarm
Threshold
0.001
1
1048
65,535
sec
ns
Programmable in 1 ms steps; maximum is (220 – 1) ms
Programmable in 1 ns steps
Rev. 0 | Page 16 of 205
Data Sheet
AD9546
DISTRIBUTION CLOCK OUTPUTS
Table 11.
Parameter
Min
Typ
Max
Unit Test Conditions/Comments
DIFFERENTIAL MODE
Output Frequency
All testing is both ac-coupled and dc-coupled
Frequency range determined by driver functionality; actual
frequency synthesis may be limited by the APLL VCO
frequency range
CML
1
1
500 × 106
500 × 106
Hz
Hz
Terminated per Figure 40
Terminated per Figure 39
High Speed Current Steering
Logic (HCSL)
Differential Output Voltage
Swing
Voltage between output pins measured with output driver
static; peak-to-peak differential output amplitude is twice
that shown when driver is toggling and measured using a
differential probe
Output Current = 7.5 mA
HCSL
CML
312
257
368
348
402
408
mV
mV
Terminated per Figure 39
Terminated to VDD (nominal 1.8 V) per Figure 40
Output Current = 15 mA
HCSL
CML
631
578
745
729
809
818
mV
mV
Terminated per Figure 39
Terminated to VDD (nominal 1.8 V) per Figure 40
Common-Mode Output Voltage
Output Current = 7.5 mA
HCSL
CML
155
184
201
mV
Terminated per Figure 39
VDD − 208 VDD − 188 VDD − 169 mV
Terminated to VDD (nominal 1.8 V) per Figure 40 (maximum
common-mode voltage case occurs at the minimum
amplitude)
Output Current = 15 mA
HCSL
CML
316
372
405
mV
Terminated per Figure 39
Terminated to VDD (nominal 1.8 V) per Figure 40 (maximum
common-mode voltage case occurs at the minimum amplitude)
VDD − 416 VDD − 371 VDD − 327 mV
Rise/Fall Time
Rise/fall times measured on a 50 MHz output signal (parasitic
load ≈ 5 pF)
HCSL
7.5 mA Drive Current
15 mA Drive Current
CML
228
218
342
340
ps
ps
7.5 mA Drive Current
15 mA Drive Current
Duty Cycle
205
185
323
303
ps
ps
HCSL
Frequency with 7.5 mA
Drive Current
≤100 MHz
500 MHz
47.5
43.8
50
50
52.5
55
%
%
Frequency with 15 mA
Drive Current
≤100 MHz
500 MHz
CML
47.5
46.5
50
50
52.5
52.5
%
%
Frequency with 7.5 mA
Drive Current
≤100 MHz
500 MHz
47.5
45
50
50
52.5
55
%
%
Frequency with 15 mA
Drive Current
≤100 MHz
500 MHz
47.5
45
50
50
52.5
54
%
%
Rev. 0 | Page 17 of 205
AD9546
Data Sheet
Parameter
Min
Typ
Max
Unit Test Conditions/Comments
SINGLE-ENDED MODE
Output Frequency
1
500 × 106
Hz
Frequency range determined by driver functionality; actual
frequency synthesis may be limited by the APLL VCO
frequency range
Output Current = 12.5 mA
Voltage Swing (Peak to Peak)
HCSL Driver Mode
509
456
584
565
634
644
mV
mV
Each output terminated per Figure 44 with load resistor
(RL) = 50 Ω
Each output terminated per Figure 44 with RL = 50 Ω
connected to VDD (nominal 1.8 V) instead of ground
CML Driver Mode
Voltage Swing Midpoint
HCSL Driver Mode
CML Driver Mode
255
292
317
mV
Each output terminated per Figure 44 with RL = 50 Ω
Each output terminated per Figure 44 with RL = 50 Ω
connected to VDD (nominal 1.8 V) instead of ground
VDD − 325 VDD − 291 VDD − 266 mV
Output Current = 15 mA
Voltage Swing (Peak to Peak)
HCSL Driver Mode
645
589
734
721
796
815
mV
mV
Each output terminated per Figure 44 with RL = 50 Ω
Each output terminated per Figure 44 with RL = 50 Ω
connected to VDD (nominal 1.8 V) instead of ground
CML Driver Mode
Voltage Swing Midpoint
HCSL Driver Mode
CML Driver Mode
322
367
398
mV
Each output terminated per Figure 44 with RL = 50 Ω
Each output terminated per Figure 44 with RL = 50 Ω
connected to VDD (nominal 1.8 V) instead of ground
VDD − 411 VDD − 367 VDD − 334 mV
Rise/Fall Time
Rise/fall times based on 50 MHz output signal (parasitic load
≈ 5 pF)
HCSL
12.5 mA Drive Current
15 mA Drive Current
CML
183
181
367
348
ps
ps
12.5 mA Drive Current
15 mA Drive Current
Duty Cycle
178
183
353
393
ps
ps
HCSL
Frequency with 12.5 mA
Drive Current
50 MHz
100 MHz
500 MHz
46.5
43.8
36.2
49
48.5
42
51.5
53.8
47.8
%
%
%
Frequency with 15 mA
Drive Current
50 MHz
100 MHz
500 MHz
CML
46.5
46.5
38
49
49
43.5
51.5
51.3
50
%
%
%
Frequency with 12.5 mA
Drive Current
50 MHz
100 MHz
500 MHz
48.5
49
50
51
51.5
58
53.5
54
63
%
%
%
Frequency with 15 mA
Drive Current
50 MHz
100 MHz
500 MHz
48.5
49
50
51
51.5
57.5
53.5
54
62.5
%
%
%
Rev. 0 | Page 18 of 205
Data Sheet
AD9546
OUTPUT TO OUTPUT TIMING SKEW
The values in Table 12 indicate the time offset of the target output clock edge relative to the base output clock edge. fREF = 25 MHz; fREF
applied to a single reference input with that reference assigned to Digital Phase-Locked Loop 0 (DPLL0) and Digital Phase-Locked Loop
1 (DPLL1); DPLL0 and DPLL1 operate in zero delay (hitless mode); PLL0 and PLL1 frequency translation is such that the frequency of
the OUTxyP and OUTxyN pins (fOUTxy) = fREF for all outputs.
Table 12.
Parameter
Min
Typ
Max
Unit
TARGET OUTPUT1
OUT1A Differential CML
Base Output OUT0A
Differential CML
−21
−89
+53
−36
+123
+14
ps
ps
Differential HCSL
OUT1A Differential HCSL
Base Output OUT0A
Differential CML
3
−61
83
−5
174
+50
ps
ps
Differential HCSL
OUT1A Single-Ended CML
Base Output OUT0A
Differential CML
−24
−120
−18
+48
−43
+63
−56
+109
+25
+123
+8
ps
ps
ps
ps
Differential HCSL
Single-Ended CML
Single-Ended HCSL
OUT1A Single-Ended HCSL
Base Output OUT0A
Differential CML
−132
93
16
92
0
175
84
191
77
254
190
282
134
ps
ps
ps
ps
Differential HCSL
Single-Ended CML
Single-Ended HCSL
OUT0B or OUT0C Differential CML
Base Output OUT0A
Differential CML
37
−67
60
−30
97
+2
ps
ps
Differential HCSL
OUT0B or OUT0C Differential HCSL
Base Output OUT0A
Differential CML
53
−25
87
−3
118
+22
ps
ps
Differential HCSL
OUT0B or OUT0C Single-Ended CML
Base Output OUT0A
Differential CML
15
−84
37
44
−48
67
66
−21
98
ps
ps
ps
ps
Differential HCSL
Single-Ended CML
Single-Ended HCSL
OUT0B or OUT0C Single-Ended HCSL
Base Output OUT0A
Differential CML
−100
−53
−24
135
53
162
46
175
80
198
79
212
112
273
104
ps
ps
ps
ps
Differential HCSL
Single-Ended CML
Single-Ended HCSL
1 OUTxy refers to Output xy where x is 0 or 1, and y is A, B, or C. See the Distribution Clock Output Drivers Overview section for more information about the output
drivers.
Rev. 0 | Page 19 of 205
AD9546
Data Sheet
OUTPUT TIMING SKEW BETWEEN Mx PINS AND OUTxyP AND/OR OUTxyN PINS
The timing skew between the Mx pins and the OUTxyP and/or OUTxyN pins is important when using an auxiliary NCO as a reference
input to one of the DPLLs and simultaneously using an Mx pin as an output for the auxiliary NCO (see the Auxiliary NCOs section and
the Status and Control Pins section). To obtain the data given in Table 13, the device configuration uses an auxiliary NCO as the
reference input to DPLL0 and DPLL1, with both phase-locked loop (PLL) channels employing a 1:1 frequency translation from the
auxiliary NCO to the distribution output. The auxiliary NCO output routes to one of the Mx pins that is configured as a status output.
Because of the 1:1 frequency translation, the Mx pin output signal and the OUTxyP/OUTxyN output signal have the same frequency.
The data in Table 13 is for the timing skew between the OUT0AP, OUT0AN, OUT1AP, or OUT1AN pin and the Mx pin carrying the
auxiliary NCO output signal. Note that the auxiliary NCO exhibits ~1.25 ns quantization, and the tabulated data in Table 13 is an average
of the quantization variation.
Table 13.
Parameter
Min
Typ
Max
Unit
TIMING SKEW BETWEEN Mx AND OUT0AP/OUT0AN PINS
VDDIOA and VDDIOB Supply Voltage
1.8 V
2.5 V
3.3 V
−5.5
−5.15
−5.0
ns
ns
ns
TIME DURATION OF DIGITAL FUNCTIONS
Table 14.
Parameter
Min Typ
Max
Unit
Test Conditions/Comments
TIME DURATION OF DIGITAL
FUNCTIONS
EEPROM to Register Download
Time
10
ms
Using the Typical Configuration conditions from Table 4
Power-On Reset (POR)
Mx Pin to RESETB Rising Edge
Setup Time
25
1
ms
ns
Time from power supplies > 80% to release of internal reset
Mx refers to Pin M0 through Pin CSB/M6
Mx Pin to RESETB Rising Edge
Hold Time
2
ns
Multiple Mx Pin Timing Skew
RESETB Falling Edge to Mx Pin
High-Z Time
39
14
ns
ns
Applies only to multibit Mx pin functions
Time from Release of Power-
Down to Completion of
System Clock PLL Calibration
Time from Release of Power-
Down to System Clock PLL
Locked and Calibrated
Time for APLLx to Calibrate and
Lock
170
172
5
µs
µs
ms
Excludes time delay associated with the user programmable
system clock stability timer (50 ms default)
Excludes time delay associated with the user programmable
system clock stability timer (50 ms default)
Time from assertion of IO update (with Bit 1 = 1 in Register 0x2100
and/or Register 0x2200) to APLLx indicating calibrated and locked
status; uses the Typical Configuration conditions in Table 4;
prerequisites = system clock PLL locked and stable, and DPLLx
phase locked
TIME FROM START OF DPLL
ACTIVATION TO ACTIVE PHASE
DETECTOR OUTPUT
Untagged Operation
10
10
tPFD
tPFD = R/fREF, where R is the frequency division factor determined by
the R divider, and fREF is the frequency of the active reference
Tag period = (tag ratio/fTAG), where fTAG is either fREF (for tagged
Tagged Operation
Tag
period reference mode) or the feedback frequency of DPLLx, fFEEDBACK (for all
other tagged modes); the tag ratio corresponds to the selection of fTAG
(as fREF or fFEEDBACK
)
Rev. 0 | Page 20 of 205
Data Sheet
AD9546
DPLL0 AND DPLL1 SPECIFICATIONS
Table 15.
Parameter
Min
Typ Max
2 × 105
Unit
Test Conditions/Comments
DPLL
Digital Phase Detector (DPD)
Input Frequency Range
1
Hz
Loop Filter
Profile 0
Bandwidth
0.0001
0.0001
1850
Hz
Degrees
dB
Programmable design parameter; (fPFD/bandwidth) ≥ 20
Phase Margin
Closed-Loop Peaking
Profile 1
70
1.1
Bandwidth
305
0.1
Hz
Degrees
dB
Programmable design parameter; (fPFD/bandwidth) ≥ 20
Phase Margin
Closed-Loop Peaking
DPLL NCO DIVISION RATIO
88.5
In accordance with Telcordia GR-253 jitter transfer specifications
These specifications cover limitations on the freerun tuning word
(FTW0) of DPLLx; the AD9546 evaluation software frequency
planning wizard sets these values automatically for the user, and
the AD9546 evaluation software is available for download from
the AD9546 product page at www.analog.com/AD9546; NCO
division = 248/FTW0, which takes the form of INT.FRAC, where INT
is the integer portion, and FRAC is the fractional portion
NCO Integer
NCO Fraction
7
0.05
13
0.95
This is the integer portion of NCO division
This is the fractional portion of NCO division
DPLL LOCK DETECTION SPECIFICATIONS
Table 16.
Parameter
Min
Typ
1
Max
Unit
Test Conditions/Comments
PHASE LOCK DETECTOR
Threshold Programming Range
Threshold Resolution
10
224 − 1
224 − 1
232 − 1
ps
ps
FREQUENCY LOCK DETECTOR
Threshold Programming Range
Threshold Resolution
10
ps
ps
1
PHASE STEP DETECTOR
Threshold Programming Range
Threshold Resolution
100
ps
ps
Setting this value too low causes false triggers
1
Rev. 0 | Page 21 of 205
AD9546
Data Sheet
DPLL PHASE CHARACTERISTICS
Table 17.
Parameter
Min
Typ
Max
Unit
Test Conditions/Comments
MAXIMUM OUTPUT PHASE
PERTURBATION
Assumes a jitter free reference; satisfies Telcordia
GR-1244 requirements; 0 ppm frequency difference
between references; reference switch initiated via
register map (see UG-1793) by faulting the active
reference input
Phase Refinement Disabled
50 Hz DPLL loop bandwidth; normal phase margin
mode; frequency translation = 19.44 MHz to
155.52 MHz; 49.152 MHz signal generator used for
system clock source
Peak
Steady State
Phase Buildout Operation
Hitless Operation
Phase Refinement Enabled
20
175
160
ps
18
0
ps
ps
50 Hz DPLL loop bandwidth; high phase margin mode;
phase refinement iterations = 4; frequency translation
= 19.44 MHz to 155.52 MHz; 49.152 MHz signal
generator used for system clock source
Peak
Steady State
Phase Buildout Operation
Hitless Operation
PHASE SLEW LIMITER
5
44
43
ps
4
0
ps
ps
0.001
250
µs/sec
See the AN-1420 Application Note, Phase Buildout and
Hitless Switchover with Digital Phase-Locked Loops
(DPLLs)
Rev. 0 | Page 22 of 205
Data Sheet
AD9546
DPLL PROPAGATION DELAY
DPLL propagation delay is the difference in clock edge timing at the output (OUTxyP/OUTxyN) relative to the REFA input (1.2 V
CMOS mode). To accommodate edge timing measurements, the DPLL is configured for internal zero delay (hitless) operation and with
a 1:1 frequency translation ratio: fREF = fOUT = 10 MHz. Delay compensation is not active (see the Delay Compensation section).
Absolute delay values use Q Divider QxA as feedback for DPLLx, where x is 0 or 1. The data shows propagation delay for various output
driver configurations (single-ended, differential), driver modes (HCSL, CML), and driver output current. The data also shows the
propagation delay to outputs associated with a Q divider that is not the Q divider used for DPLL feedback. Also captured is the delay
when using DPLL1 instead of DPLL0.
Relative delay values show the impact of using other Q dividers for DPLL feedback (relative to QxA for DPLLx, where x is 0 or 1).
For the propagation delay variation due to using a reference input mode other than 1.2 V CMOS, use Table 8.
Table 18.
Parameter
Min
Typ
Max
Unit
Test Conditions/Comments
ABSOLUTE DELAY
REFA to OUT0AP
CML at 12.5 mA
CML at 15 mA
HCSL at 12.5 mA
HCSL at 15 mA
REFA to OUT0AN
REFA to OUT0A1
CML at 7.5 mA
CML at 15 mA
HCSL at 7.5 mA
HCSL at 15 mA
REFA to OUT0B1
REFA to OUT0C1
REFA to OUT1AP
REFA to OUT1A1
REFA to OUT1B1
RELATIVE DELAY2
QxAA
Q0A = DPLL0 feedback; driver output = single-ended
1055
1065
1180
1145
1185
1265
1270
1385
1345
1390
1430
1430
1550
1515
1565
ps
ps
ps
ps
ps
Driver output = single-ended HCSL at 15 mA
Driver output = differential
805
760
845
785
770
785
1150
810
740
1000
955
1030
975
960
970
1370
985
970
1180
1130
1200
1155
1140
1145
1565
1140
1180
ps
ps
ps
ps
ps
ps
ps
ps
ps
Driver output = differential HCSL at 15 mA
Driver output = differential HCSL at 15 mA
Q1A = DPLL1 feedback; driver output = single-ended HCSL at 15 mA
Driver output = differential HCSL at 15 mA
Driver output = differential HCSL at 15 mA
With respect to Q0A (PLL0) or Q1A (PLL1)
14
9
23
36
50
ps
ps
ps
ps
ps
QxB
QxBB
Q0C
Q0CC
1 OUTxy refers to Output xy where x is 0 or 1, and y is A, B, or C. See the Distribution Clock Output Drivers Overview section for more information about the output
drivers.
2 Qxy denotes a specific Q divider in the clock distribution section, where x is 0 or 1, and y is AA, B, BB, C, or CC.
Rev. 0 | Page 23 of 205
AD9546
Data Sheet
DPLL PROPAGATION DELAY VARIATION
The PLL channels of the AD9546 have an inherent propagation delay (see Table 18) that can be nominally offset using the skew
adjustment feature in the source profiles of the device. Dynamic variations that are proportional to temperature change can be mitigated
with the delay compensation mechanism of the AD9546.
A polynomial with programmable coefficients of temperature (see the Delay Compensation section) determines the response of the delay
compensation circuit. An additive inverse of the polynomial function derives from a best fit algorithm applied to the uncompensated
delay variation measurements. Active delay compensation uses the additive inverse polynomial in an attempt to nullify the delay
variation over temperature. For a second-order polynomial, the measured data without delay compensation in Table 19 is the first-order
representation for the various device configurations. A repeat of the same measurement with the inverse coefficient programmed yields
the slopes of the residual delay in the specifications with delay compensation in Table 19. See Figure 25 and Figure 26 for the related
plots.
The tabulated data with delay compensation in Table 19 reflects an average of several devices that are programmed with the same delay
compensation coefficients. The tabulated residual errors are greater than expected for coefficients that are optimized for an individual
device.
The DPLL propagation delay variation results depend on the device being configured with a 1:1 frequency translation ratio, where fREF
OUT = 10 MHz, the reference input configured to 1.2V CMOS, and DPLL configured to zero delay (hitless). A die temperature range of
=
f
−40°C to +105°C applies for both the uncompensated and compensated results.
Table 19.
Parameter
Min
Typ
Max
Unit
OUTPUT DRIVER MODE
Differential
CML
Without Delay Compensation
With Delay Compensation
HCSL
0.517
−0.0155
ps/°C
ps/°C
Without Delay Compensation
With Delay Compensation
0.624
−0.0253
ps/°C
ps/°C
Single-Ended
CML
Without Delay Compensation
With Delay Compensation
HCSL
0.681
0.0208
ps/°C
ps/°C
Without Delay Compensation
With Delay Compensation
0.861
0.0288
ps/°C
ps/°C
Rev. 0 | Page 24 of 205
Data Sheet
AD9546
HOLDOVER SPECIFICATIONS
Table 20.
Parameter
Min
Typ
0.01
Max
Unit Test Conditions/Comments
HOLDOVER SPECIFICATIONS
Initial Frequency
Accuracy
0.1
ppb AD9546 is configured using Device Configuration 1 from Table 37; excludes
frequency drift of SYSCLK source; excludes frequency drift of input
reference prior to entering holdover; 160 ms history timer; history
holdoff setting = 8 decimal in Register 0x1010 and Register 0x1410; three
features (bits) are Logic 1: Bit 4 in Register 0x100E and Register 0x140E,
Bit 3 in Register 0x100E and Register 0x140E, and Bit 5 in Register 0x100E
and Register 0x140E
Relative Frequency
Accuracy Between
Channels
Cascaded DPLL
Operation
0
ppb
Noncascaded DPLL
Operation
<1
ppb fOSC = 52 MHz (stable external oscillator); fOUT = 155.52 MHz; fREF = 38.88
MHz; DPLL loop bandwidth = 50 Hz; DPLL history accumulation timer = 1
ms; DPLL history holdoff value = 1
History Averaging
Window
0.001
26,8435 sec
ANALOG PLL (APLL0 AND APLL1) SPECIFICATIONS
Table 21.
Parameter
Min
Typ
Max
Unit
VCO FREQUENCY RANGE
Analog PLL0 (APLL0)
Analog PLL1 (APLL1)
PFD INPUT FREQUENCY RANGE
LOOP BANDWIDTH
PHASE MARGIN
2424
3232
162
3232
4040
350
MHz
MHz
MHz
260
68
kHz
Degrees
OUTPUT CHANNEL DIVIDER SPECIFICATIONS
Table 22.
Parameter
Min
Typ Max
Unit Test Conditions/Comments
OUTPUT PHASE ADJUST
STEP SIZE
1
tVCO
VCO time period (tVCO) = 1/(APLLx VCO frequency), where x = 0 or 1
MODULATOR
Carrier Frequency
fVCO/16 Hz
216 − 1 tVCO
The maximum value is the APLLx VCO frequency divided by 16, where x = 0 or 1;
fVCO is the VCO frequency
tVCO = 1/(APLLx VCO frequency), where x = 0 or 1; the maximum value is limited
to the Qxy divide ratio − 1; Qxy refers to the distribution dividers on each
output, where x is either 0 (for PLL0) or 1 (for PLL1), and y is A, B, or C
Time Deviation (from
Nominal Duty Cycle
of Carrier Clock)
0
Embedded Frequency
fOUT
/
fOUT/6
Hz
fOUT is the output frequency
(228 − 1)
Rev. 0 | Page 25 of 205
AD9546
Data Sheet
TIME TO DIGITAL CONVERTER (TDC) SPECIFICATIONS
Table 23.
Parameter
Min
Typ
Max
Unit
Test Conditions/Comments
TDCs
Frequency Range
Pulse Width
200
kHz
ns
5
Time Stamp Jitter
Root Mean Square (TDCJrms
)
5
6
11
250
ps
ps
µs
ms
Peak (TDCJpk
)
Retrigger Blackout Period
Start-Up Time
4.9
Interval between successive rising edges at the TDC input
The time required, after the system clock PLL locks, before
the TDCs can generate time stamps
AUXILIARY NCO SPECIFICATIONS
Table 24.
Parameter
Min
Typ
Max
Unit
Test Conditions/Comments
NCOs
Auxiliary NCO 0 and auxiliary NCO 1
Fundamental Frequency
Range
1
1.27
5
65,535
1.37
232
Hz
pHz
ppb
Quantization
Phase Slew Limiter
Output Signal
Pulse Width
Duty Cycle
pHz is picohertz
Actual units are fractional part (ideal)/actual unit interval (UI)
38
45
ns
%
55
Assumes the device is programmed to produce a nominal
pulse width of 50%
Quantization
1.4
ns
COMMON CLOCK DPLL SPECIFICATIONS
Table 25.
Parameter
Min
0
Typ
1
Max
Unit
Test Conditions/Comments
PHASE LOCK DETECTOR
Threshold Programming Range
Threshold Resolution
Post Detection Delay Range
REFERENCE MONITOR
65,535
ps
ps
ms
Programmed values < 10 ps not recommended
0
65,535
Loss of Reference Detection Time
2 × tPFD
µs
%
tPFD = R/fREF, where R is the reference input divider
value and fREF is the reference input frequency in MHz
Input Period Deviation Limit
Reference Validation Delay
3.125
Relative to the programmed value of the reference
period, tREF
Applies only to the common clock DPLL secondary
frequency source (see the Common Clock Reference
Switchover section)
2
ms
Rev. 0 | Page 26 of 205
Data Sheet
AD9546
COMMON CLOCK SYNCHRONIZER (CCS) SPECIFICATIONS
Table 26.
Parameter
Min
Typ Max
Unit
Test Conditions/Comments
ABSOLUTE LATENCY
Time from asserting an input/output (IO) update (Bit 0 in
Register 0x000F) to the CSS indicating ready status (asserting IO
update constitutes insertion of a time value; see the synchronization
time input in Figure 51)
Fractional Seconds Format
PTP Format
1
3
μs
μs
μs
BLACKOUT PERIOD
10
The time interval following assertion of an input synchronization
event (see the synchronization input in Figure 51) before asserting an
IO update to insert a time value (see the synchronization time input in
Figure 51)
GUARD PARAMETERS
Maximum Magnitude Guard
Range
0
0
0
953
909
ns
fs
Minimum value applies when maximum magnitude guard is disabled;
maximum value is (220 − 1) × 2−40 sec
Resolution
Latency Guard
Range
2−40 sec
0.99998 sec
Minimum value applies when latency guard is disabled; maximum
value is 1 − 2−16 sec
2−16 sec
Resolution
SLEW LIMITER
Range
15.3
μs
244
μs/sec Minimum value applies when slew limiter is disabled; Max value is
(224 − 1) × 2−36 sec
ps/sec 2−36 sec/sec
Resolution
TIME OFFSET
Range
Resolution
TIME SKEW
Range
14.6
3.55
3.55
−7.63
−29.8
+7.63
+29.8
μs
fs
Minimum value is −2−17 sec; maximum value is (231 – 1) × 2−48 sec
2−48 sec
ns
fs
Minimum value is −2−25s; maximum value is (223 – 1) × 2−48 sec
2−48 sec
Resolution
USER TIME STAMPERS (UTS) SPECIFICATIONS
Table 27.
Parameter
Min
Typ Max
Unit
Test Conditions/Comments
LATENCY
Fractional Seconds Format
PTP Format
4
5
6
7
μs
μs
THROUGHPUT RATE
The sample rate is shared among all UTSs that are connected to a time
stamp source; data capture involves asserting an IO update and reading
the UTS first in, first out (FIFO) registers as a contiguous 14-byte block;
the indicated rate is the maximum continuous UTS traffic possible
without overloading the UTS FIFO
SCLK = 50 MHz with
Address Looping
446
kSPS
Dedicated, 100% serial port interface (SPI) utilization using address
looping (loop length = 14 decimal in Register 0x0010), which eliminates
the overhead of asserting an IO update prior to reading the UTS FIFO
registers
SCLK = 50 MHz
SCLK = 25 MHz
TIME OFFSET
Range
324
16
kSPS
kSPS
Dedicated, 100% SPI utilization
10% SPI utilization
−29.8
+29.8 ns
fs
Signed 24-bit range scaled by 2−48 sec
2−48 sec
Resolution
3.55
FIFO DEPTH
18
Samples Each UTS has a one-sample buffer, which can extend the apparent FIFO
depth by a few samples beyond 18 samples
Rev. 0 | Page 27 of 205
AD9546
Data Sheet
INVERSE USER TIME STAMPERS (IUTS) SPECIFICATIONS
Table 28.
Parameter
Min
Typ
Max
Unit
kHz
%
Test Conditions/Comments
FREQUENCY RANGE
PERIOD MONITOR THRESHOLD
0.001
2
200
IUTS frequency transport range for digitized clocking
2−6 relative to the average period IUTS transport
frequency
1.56
THROUGHPUT RATE
Data transfer involves a 14-byte block write to the IUTS
control registers followed by assertion of an IO update;
the indicated rate is the maximum continuous IUTS
traffic possible without causing IUTS invalidation; SPI
utilization other than for IUTS operation reduces
throughput
SCLK = 50 MHz
SCLK = 25 MHz
TIME OFFSET
Range
64
16
kSPS
kSPS
20% SPI utilization allocation for IUTS operations
10% SPI utilization allocation for IUTS operations
−29.8
+29.8
ns
fs
Signed 24-bit range scaled by 2−48 sec
2−48 sec
Resolution
3.55
Rev. 0 | Page 28 of 205
Data Sheet
AD9546
ANALOG LOOPBACK (ROUND TRIP DELAY) SPECIFICATIONS
The data in Table 29 reflects REFB configured for dc-coupled 1.2 V CMOS. The data for Path A captures the delay associated with the
path from the REFB receiver through the loopback selection mux to the M4 pin driver and the effect of operating the M4 driver with
nominal supply voltages of 1.8 V, 2.5 V, and 3.3 V. The data for Path B captures the delay associated with the reference divider. The data
for Path C captures the delay associated with the reference demodulator. See Figure 56 for a depiction of the three paths (A, B, and C).
In the upper propagation delay section of Table 29, the tabulated values include variation of supply voltage from nominal and
temperature variation.
In the lower propagation delay section of Table 29, the tabulated values include only variation of supply voltage from nominal. This data
is useful when the user has the means to correct for delay variation due to temperature by using the tabulated temperature coefficients to
apply an offset to round trip delay measurements.
The M4 output section of Table 29 specifies the output drive capability of the M4 pin driver when operating in analog loopback mode.
Table 29.
Parameter
Min
Typ
Max
Unit
Test Conditions/Comments
PROPAGATION DELAY
See Figure 56 for path details
REFB TO M4 Accounting for Supply
Variation from Nominal and
Temperature Variation
Path A
VDDIOB = 1.8 V
VDDIOB = 2.5 V
VDDIOB = 3.3 V
Path B
2.09
1.79
1.63
2.54
2.09
1.91
3.00
2.45
2.22
ns
ns
ns
VDDIOB = 1.8 V
R Divider > 1
R Divider = 1
Path C
2.95
2.76
4.14
3.57
3.33
5.11
4.20
3.90
6.06
ns
ns
ns
VDDIOB = 1.8 V
REFB to M4 Accounting for Supply
Variation from Nominal
See Figure 56 for path details
Path A
VDDIOB = 1.8 V
Temperature Coefficient
VDDIOB = 2.5 V
Temperature Coefficient
VDDIOB = 3.3 V
Temperature Coefficient
Path B
2.41
1.98
1.81
2.54
4.12
2.09
2.92
1.91
2.33
2.66
2.20
2.02
ns
ps/°C
ns
ps/°C
ns
ps/°C
VDDIOB = 1.8 V
VDDIOB = 1.8 V
R Divider > 1
Temperature Coefficient
R Divider = 1
3.32
3.12
4.75
3.57
4.82
3.33
4.73
5.11
7.60
3.80
3.52
5.45
ns
ps/°C
ns
ps/°C
ns
ps/°C
Temperature Coefficient
Path C
Temperature Coefficient
M4 OUTPUT SPECIFICATION
Applies only for analog loopback operation;
5 pF capacitive load
Frequency range
Pulse Width High
0
3
125
MHz
ns
Maximum limit applies for full output voltage swing
with VDDIOB = 1.8 V, 2.5 V, or 3.3 V; output voltage
swing degrades for frequencies above 125 MHz
Minimum duration of positive half cycle that supports
full output voltage swing; positive output voltage
swing degrades for frequencies above 125 MHz or for
pulse width < 3 ns
Pulse Width Low
3
ns
Minimum duration of negative half cycle that supports
full output voltage swing; negative output voltage
swing degrades for frequencies above 125 MHz or for
pulse width < 3 ns
Rev. 0 | Page 29 of 205
AD9546
Data Sheet
Mx TO Mx PIN OUTPUT TIMING SKEW
The data in Table 30 indicates the time offset of the target Mx pin clock edge relative to the M0 pin clock edge. Table 30 indicates the
difference in output timing between Mx pins that are configured as outputs. The M0 pin and target Mx pins use the same internal status
source. To determine the skew between any two Mx pins (for example, M2 to SDO/M5), subtract the values in the appropriate rows
(skew from M2 to SDO/M5 is 190 − 760 = −570 ps). Note that the supply voltage on VDDIOA and VDDIOB has minimal effect (when
the nominal supply voltage of each is similar) on the output timing skew.
Table 30.
Parameter
TIMING SKEW
M0 to M1
M0 to M2
M0 to M3
M0 to M4
M0 to SDO/M5
M0 to CSB/M6
Min
Typ
Max
Unit
160
190
630
1250
760
150
ps
ps
ps
ps
ps
ps
SYSTEM CLOCK COMPENSATION SPECIFICATIONS
Table 31.
Parameter
Min
Typ
Max
Unit
Test Conditions/Comments
OPEN-LOOP COMPENSATION
Resolution
See the Compensation Method 1 section
ppt is parts per trillion (10−12
0.028
ppt
)
CLOSED-LOOP COMPENSATION
(AUXILIARY DPLL)
See the Compensation Method 3 section
Phase Detector Frequency
Loop Bandwidth
Reference Monitor Threshold
2
0.1
200
2 × 103
kHz
Hz
%
5
TEMPERATURE SENSOR SPECIFICATIONS
Table 32.
Parameter
TEMPERATURE
Accuracy
Min
Typ
Max
Unit
Test Conditions/Comments
TA = −50°C to +110°C
Absolute Die Temperature
Relative Die Temperature
Resolution
Conversion time
REPEATABILITY
DRIFT
5
1.7
0.0078
0.18
0.02
0.1
°C
%
°C
ms
°C
°C
16-bit (signed) resolution
TA = 25°C
500-hour stress test at 100°C
LOGIC INPUT SPECIFICATIONS (RESETB, M0 TO CSB/M6 PINS)
Table 33.
Parameter
Min
Typ Max
Unit
Test Conditions/Comments
Valid for 3.3 V ≥ VDDIOA ≥ 1.8 V; internal 100 kΩ pull-up resistor
RESETB
Input Voltage
High, VIH
Low, VIL
VDDIOA − 0.4
V
V
0.4
Input Current
High, IINH
Low, IINL
1
µA
µA
15
125
Rev. 0 | Page 30 of 205
Data Sheet
AD9546
Parameter
Min
Typ Max
Unit
Test Conditions/Comments
LOGIC INPUTS (M0 to
CSB/M6)
Valid for 3.3 V ≥ VDDIOx ≥ 1.8 V; VDDIOA applies to the SDO/M5
pin and the SDA/M6 pin; VDDIOB applies to the M0, M1, M2, M3,
and M4 pins; the M3 and M4 pins have internal 100 kΩ pull-down
resistors
Input Voltage
High, VIH
VDDIOx − 0.4
V
Low, VIL
0.4
V
Input Current, IINH, IINL
15
125
µA
LOGIC OUTPUT SPECIFICATIONS (M0 TO CSB/M6 PINS)
Table 34.
Parameter
Min
Typ
Max Unit
Test Conditions/Comments
LOGIC OUTPUTS (M0 to CSB/M6)
Valid for 3.3 V ≥ VDDIOx ≥ 1.8 V; VDDIOA applies for the
SDO/M5 and CSB/M6 pins; VDDIOB applies for the M0
to M4 pins; normal (default) output drive current
setting for the M0 to CSB/M6 pins
Frequency Range
26
MHz
M4 capable of 125 MHz when configured for analog
loopback (see Table 29)
Output Voltage
High, VOH
VDDIOx − 0.6
VDDIOx – 0.2
V
V
V
V
Load current = 10 mA
Load current = 1 mA
Load current = 10 mA
Load current = 1 mA
Low, VOL
0.6
0.2
Rise/Fall Time
Normal Drive Strength (Default)
5 pF Capacitive Load
VDDIOx = 1.8 V
880
740
580
ps
ps
ps
VDDIOx = 2.5 V
VDDIOx = 3.3 V
10 pF Capacitive Load
VDDIOx = 1.8 V
VDDIOx = 2.5 V
1210
1020
900
ps
ps
ps
VDDIOx = 3.3 V
20 pF Capacitive Load
VDDIOx = 1.8 V
VDDIOx = 2.5 V
1900
1230
970
ps
ps
ps
VDDIOx = 3.3 V
Weak Drive Strength
5 pF Capacitive Load
VDDIOx = 1.8 V
1900
1220
970
ps
ps
ps
VDDIOx = 2.5 V
VDDIOx = 3.3 V
10 pF Capacitive Load
VDDIOx = 1.8 V
VDDIOx = 2.5 V
2730
1810
1420
ps
ps
ps
VDDIOx = 3.3 V
20 pF Capacitive Load
VDDIOx = 1.8 V
VDDIOx = 2.5 V
4040
2890
2340
ps
ps
ps
VDDIOx = 3.3 V
Rev. 0 | Page 31 of 205
AD9546
Data Sheet
SERIAL PORT SPECIFICATIONS
SPI Mode
Table 35.
Parameter
Min
Typ
Max
Unit
Test Conditions/Comments
CSB
Valid for VDDIOA = 1.8 V, 2.5 V, and 3.3 V
Input Voltage
Logic 1
Logic 0
VDDIOA − 0.4
V
V
0.4
Input Current
Logic 1
Logic 0
1
1
µA
µA
SCLK
Input Voltage
Logic 1
Logic 0
VDDIOA − 0.4
V
V
0.4
Input Current
Logic 1
Logic 0
1
1
µA
µA
SDIO
As an Input
Input Voltage
Logic 1
Logic 0
Input Current
Logic 1
VDDIOA − 0.4
V
V
0.4
0.2
1
1
µA
µA
Logic 0
As an Output
Output Voltage
Logic 1
1 mA load current
1 mA load current
VDDIOA − 0.2
VDDIOA − 0.2
V
V
Logic 0
SDO
Output Voltage
Logic 1
V
Logic 0
Leakage Current
TIMING
0.2
1
V
µA
SDO inactive (high impedance)
Valid for VDDIOA = 1.8 V, 2.5 V, and 3.3 V
SCLK
Clock Rate, 1/tCLK
Pulse Width High, tHIGH
Pulse Width Low, tLOW
SDIO to SCLK Setup, tDS
SCLK to SDIO Hold, tDH
SCLK to Valid SDIO and SDO, tDV
CSB to SCLK Setup, tS
CSB to SCLK Hold, tC
CSB Minimum Pulse Width High
50
9
MHz
ns
ns
ns
ns
ns
ns
ns
tCLK
5
9
2.2
0
1.5
0
1
Rev. 0 | Page 32 of 205
Data Sheet
AD9546
I2C Mode
Table 36.
Parameter
SDA, SCL (AS INPUTS)
Input Voltage
Logic 1
Min
Typ Max
Unit
Test Conditions/Comments
Valid for VDDIOA = 1.8 V, 2.5 V, and 3.3 V
70
% of VDDIOA
Logic 0
0.3 × VDDIOA
V
Input Current
Hysteresis of Schmitt Trigger
Inputs
−10
1.5
+10
µA
For VIN = 10% to 90% of VDDIOA
Output current (IOUT) = 3 mA
% of VDDIOA
SDA (AS OUTPUT)
Output Logic 0 Voltage
0.2
V
Output Fall Time from VIH
Minimum to VIL Maximum
20 + 0.1 × CB
250
ns
10 pF ≤ CB ≤ 400 pF; CB is the bus
capacitance
TIMING
SCL Clock Rate
Bus Free Time Between a Stop and
Start Condition, tBUF
Repeated Start Condition Setup
Time, tSU; STA
Repeated Hold Time Start
Condition, tHD; STA
400
kHz
µs
1.3
0.6
0.6
µs
µs
After this period, the first clock pulse is
generated
Stop Condition Setup Time, tSU; STO
Low Period of the SCL Clock, tLOW
High Period of the SCL Clock, tHIGH
SCL/SDA Rise Time, tR
SCL/SDA Fall Time, tF
Data Setup Time, tSU; DAT
0.6
1.3
0.6
µs
µs
µs
ns
ns
ns
ns
pF
20 + 0.1 × CB
20 + 0.1 × CB
100
300
300
Data Hold Time, tHD; DAT
Capacitive Load, CB
100
400
Applies to individual bus lines
Rev. 0 | Page 33 of 205
AD9546
Data Sheet
JITTER GENERATION (RANDOM JITTER)
System clock frequency doubler enabled. DPLL configured with the high phase margin loop filter. The integration bandwidth for rms
jitter values is 12 kHz to 20 MHz. The effect on jitter values due to using different output driver modes is insignificant.
Device Configuration 1:
f
OSC = 52 MHz crystal, fREF = 38.88 MHz, fOUT = 155.52 MHz, DPLL bandwidth (BWDPLL) = 50 Hz, phase buildout operation;
Channel 0: fVCO = 2488.32 MHz; Channel 1: VCO frequency (fVCO) = 3265.92 MHz, Q divider half divide enabled
Device Configuration 2:
f
OSC = 52 MHz crystal, fREF = 30.72 MHz, fOUT = 245.76 MHz, BWDPLL = 50 Hz, internal zero delay operation;
Channel 0: fVCO = 2457.6 MHz; Channel 1: fVCO = 3686.4 MHz, Q divider half divide enabled
Device Configuration 3:
f
OSC = 52 MHz crystal, fCOMP = 19.2 MHz temperature compensated crystal oscillator (TCXO), loop bandwidth of the auxiliary DPLL
servo loop (BWCOMP) = 50 Hz (see the Auxiliary DPLL Loop Bandwidth section for more details), fREF = 1 Hz, fOUT = 491.52 MHz,
BWDPLL = 0.05 Hz, phase buildout operation; Channel 0: fVCO = 2949.12 MHz; Channel 1: fVCO = 3932.16 MHz
Device Configuration 4:
f
OSC = 52 MHz crystal, fCOMP = 19.2 MHz TCXO, BWCOMP = 50 Hz, fREF = 125 MHz, fOUT = 125 MHz, BWDPLL = 0.1 Hz, phase buildout
operation; Channel 0: fVCO = 2500 MHz; Channel 1: fVCO = 3250 MHz
Device Configuration 5:
f
OSC = 52 MHz crystal, fREF = 25 MHz, fOUT = 312.5 MHz, BWDPLL = 50 Hz , phase buildout operation; Channel 0: fVCO = 2500 MHz;
Channel 1: fVCO = 3750 MHz
Device Configuration 6:
f
OSC = 52 MHz crystal, fREF = 155.52 MHz, MHz, fOUT = (155.52 × 255/227) MHz, BWDPLL = 50 Hz; Channel 0: fVCO = 2620.5463 MHz;
Channel 1: fVCO = 3319.3586 MHz
Table 37.
Parameter
Min Typ
Max Unit Test Conditions/Comments
JITTER GENERATION
Channel 0—DPLL0, APLL0
fOUT = 155.52 MHz
fOUT = 245.76 MHz
fOUT = 491.52 MHz
fOUT = 125.0 MHz
Channel 1 powered down
223
220
235
213
217
230
fs
fs
fs
fs
fs
fs
Device Configuration 1
Device Configuration 2
Device Configuration 3
Device Configuration 4
Device Configuration 5
Device Configuration 6
Channel 0 powered down
Device Configuration 1
Device Configuration 2
Device Configuration 3
Device Configuration 4
Device Configuration 5
Device Configuration 6
fOUT = 312.5 MHz
fOUT = 174.7030837 MHz
Channel 1—DPLL1, APLL1
fOUT = 155.52 MHz
fOUT = 245.76 MHz
fOUT = 491.52 MHz
fOUT = 125.0 MHz
247
280
323
243
266
264
fs
fs
fs
fs
fs
fs
fOUT = 312.5 MHz
fOUT = 174.7030837 MHz
Rev. 0 | Page 34 of 205
Data Sheet
AD9546
PHASE NOISE
System clock frequency doubler enabled. DPLL configured with the high phase margin loop filter.
Device Configuration 1:
f
OSC = 52 MHz crystal, fREF = 38.88 MHz, fOUT = 155.52 MHz, BWDPLL = 50 Hz, phase buildout operation; Channel 0: fVCO = 2488.32 MHz;
Channel 1: VCO frequency (fVCO) = 3265.92 MHz, Q divider half divide enabled
Device Configuration 2:
f
OSC = 52 MHz crystal, fREF = 30.72 MHz, fOUT = 245.76 MHz, BWDPLL = 50 Hz, internal zero delay operation; Channel 0: fVCO =
2457.6 MHz; Channel 1: fVCO = 3686.4 MHz, Q divider half divide enabled
Device Configuration 3:
f
f
OSC = 52 MHz crystal, fCOMP = 19.2 MHz TCXO, BWCOMP = 50 Hz (see the Auxiliary DPLL Loop Bandwidth section for more details),
REF = 1 Hz, fOUT = 491.52 MHz, BWDPLL = 0.05 Hz, phase buildout operation; Channel 0: fVCO = 2949.12 MHz; Channel 1: fVCO = 3932.16 MHz
Device Configuration 4:
OSC = 52 MHz crystal, fCOMP = 19.2 MHz TCXO, BWCOMP = 50 Hz, fREF = 125 MHz, fOUT = 125 MHz, BWDPLL = 0.1 Hz, phase buildout
f
operation; Channel 0: fVCO = 2500 MHz; Channel 1: fVCO = 3250 MHz
Device Configuration 5:
f
OSC = 52 MHz crystal, fREF = 25 MHz, fOUT = 312.5 MHz, BWDPLL = 50 Hz , phase buildout operation; Channel 0: fVCO = 2500 MHz;
Channel 1: fVCO = 3750 MHz
Device Configuration 6:
f
OSC = 52 MHz crystal, fREF = 155.52 MHz, MHz, fOUT = (155.52 × 255/227) MHz, BWDPLL = 50 Hz; Channel 0: fVCO = 2620.5463 MHz;
Channel 1: fVCO = 3319.3586 MHz
Table 38.
Parameter
Min Typ
Max Unit
Test Conditions/Comments
PHASE NOISE
Channel 0—DPLL0, APLL0
fOUT = 155.52 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
Channel 1 powered down
Device Configuration 1
−81
−98
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−118
−128
−134
−144
−158
−161
fOUT = 245.76 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
Device Configuration 2
−77
−93
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−114
−125
−130
−140
−156
−161
fOUT = 491.52 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz offset
Floor
Device Configuration 3
−74
−89
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−108
−119
−123
−134
−152
−159
Rev. 0 | Page 35 of 205
AD9546
Data Sheet
Parameter
Min Typ
Max Unit
dBc/Hz
Test Conditions/Comments
fOUT = 125 MHz
Device Configuration 4
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
−84
−106
−120
−131
−136
−147
−160
−163
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
fOUT = 312.5 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
Device Configuration 5
−74
−91
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−112
−123
−128
−138
−154
−161
fOUT = 174.7030837 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
Device Configuration 6
−82
−99
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−117
−127
−133
−143
−157
−160
Channel 1—DPLL1, APLL1
fOUT = 155.52 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
Channel 0 powered down
Device Configuration 1
−81
−98
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−118
−128
−132
−144
−158
−162
fOUT = 245.76 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
Device Configuration 2
−76
−93
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−114
−124
−127
−138
−156
−161
Rev. 0 | Page 36 of 205
Data Sheet
AD9546
Parameter
Min Typ
Max Unit
dBc/Hz
Test Conditions/Comments
fOUT = 491.52 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
Device Configuration 3
−74
−90
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−108
−118
−120
−131
−150
−160
fOUT = 125 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
fOUT = 312.5 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
fOUT = 174.7030837 MHz
10 Hz Offset
100 Hz Offset
1 kHz Offset
10 kHz Offset
100 kHz Offset
1 MHz Offset
10 MHz Offset
Floor
Device Configuration 4
Device configuration 5
Device Configuration 6
−83
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−106
−120
−131
−135
−145
−160
−163
−73
−91
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−112
−122
−125
−137
−154
−161
−77
−99
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
dBc/Hz
−117
−127
−131
−142
−158
−161
Rev. 0 | Page 37 of 205
AD9546
Data Sheet
ABSOLUTE MAXIMUM RATINGS
THERMAL RESISTANCE
Table 39.
Thermal performance is directly linked to printed circuit board
(PCB) design and operating environment. Careful attention to
PCB thermal design is required.
Parameter
Rating
2 V
2 V
1.8 V Supply Voltage (VDD)
Output Drivers (OUT0AP, OUT0AN,
OUT0BP, OUT0BN, OUT0CP,
OUT0CN, OUT1AP, OUT1AN,
OUT1BP, and OUT1BN Pins)
Input/Output Supply Voltage
(VDDIOA and VDDIOB)
Input Voltage Range (XOA, XOB, REFA,
REFAA, REFB, and REFBB Pins)
Digital Input Voltage Range
θ
JA is the junction to ambient thermal resistance, 0.0 m/sec
airflow per JEDEC JESD51-2 (still air).
JMA is the junction to ambient thermal resistance, 1.0 m/sec
airflow or 2.5 m/sec airflow per JEDEC JESD51-6 (moving air).
JC is the junction to case thermal resistance (die to heat sink)
3.6 V
θ
−0.5 V to VDD + 0.5 V
θ
per MIL-STD 883, Method 1012.1.
SDO/M5, SCLK/SCL, SDIO/SDA,
and CSB/M6 Pins
M0, M1, M2, M3, and M4 Pins
Storage Temperature Range
−0.5 V to VDDIOA + 0.5 V
Values of θJA are for package comparison and PCB design
considerations. θJA provides for a first-order approximation
of TJ per the following equation:
−0.5 V to VDDIOB + 0.5 V
−65°C to +150°C
300°C
TJ = TA + (θJA × PD)
Lead Temperature (Soldering 10 sec)
where:
TA is the ambient temperature (°C).
PD is the power dissipation in watts.
Stresses at or above those listed under Absolute Maximum
Ratings may cause permanent damage to the product. This is a
stress rating only; functional operation of the product at these
or any other conditions above those indicated in the
operational section of this specification is not implied.
Operation beyond the maximum operating conditions for
extended periods may affect product reliability.
Values of θJC are for package comparison and PCB design
considerations when an external heat sink is required.
Table 40. Thermal Resistance
1
Package Type
CP-48-132, 3
θJA
θJMA
θJC
Unit
23.1
18.4, 17.1
0.7
°C/W
1 θJMA is 18.4°C/W at 1.0 m/sec airflow and 17.1°C/W at 2.5 m/sec airflow.
2 Thermal characteristics derived using a JEDEC51-7 plus JEDEC51-5 2S2P test
board. The exposed pad on the bottom of the package must be soldered to
ground to achieve the specified thermal performance.
3 Results are from simulations. The PCB is a JEDEC multilayer type. Thermal
performance for actual applications requires careful inspection of the
conditions in the application to determine if they are similar to those
assumed in these calculations.
ESD CAUTION
Rev. 0 | Page 38 of 205
Data Sheet
AD9546
PIN CONFIGURATION AND FUNCTION DESCRIPTIONS
1
2
3
4
5
6
7
8
9
SDO/M5
SCLK/SCL
VDDIOA
SDIO/SDA
CSB/M6
VDD
36 M3
35 M2
34 VDDIOB
33 M1
PIN 1
INDICATOR
32
M0
31 VDD
LDO1
AD9546
TOP VIEW
LDO0
LF0
VDD
VDD
30
(Not to Scale)
29 LF1
28 VDD
27 VDD
10
OUT0AP 11
OUT0AN 12
OUT1AN
25 OUT1AP
26
NOTES
1. EXPOSED PAD. THE EXPOSED PAD IS THE GROUND CONNECTION
ON THE CHIP. THE EXPOSED PAD MUST BE SOLDERED TO THE
ANALOG GROUND OF THE PCB TO ENSURE PROPER
FUNCTIONALITY AND FOR HEAT DISSIPATION, NOISE, AND
MECHANICAL STRENGTH BENEFITS.
2. DNC = DO NOT CONNECT. DO NOT CONNECT TO THIS PIN.
Figure 2. Pin Configuration
Table 41. Pin Function Descriptions
Input/
Mnemonic Output Pin Type
Pin No.
Description
1
SDO/M5
Output CMOS
Serial Data Output (SDO). This pin is for reading serial data in 4-wire SPI mode.
Changes to the VDDIOA supply voltage affect the VIH and VOH values for this pin.
Configurable Input/Output (M5). This pin is a status and control pin when the
device is not in 4-wire SPI mode.
2
SCLK/SCL
Input
Input
CMOS
Serial Programming Clock Pin in SPI Mode (SCLK). Changes to the VDDIOA
supply voltage affect the VIH and VOH values for this pin.
Serial Clock Pin in I2C Mode (SCL). Changes to the VDDIOA supply voltage affect
the VIH and VOH values for this pin.
Serial Port Power Supply. The valid supply voltage is 1.8 V, 2.5 V, or 3.3 V. The
VDDIOA pin can be connected to the VDD supply bus if 1.8 V operation is
desired.
Serial Data Input/Output in SPI Mode (SDIO). Write data to this pin in 4-wire SPI
mode. This pin has no internal pull-up or pull-down resistor. Changes to the
VDDIOA supply voltage affect the VIH and VOH values for this pin.
Serial Data Pin in I2C Mode (SDA).
3
4
VDDIOA
Power
CMOS
SDIO/SDA
Input/
output
5
CSB/M6
Input/
output
CMOS
Chip Select in SPI Mode (CSB). Active low input. Maintain a Logic 0 level on this
pin when programming the device in SPI mode. This pin has an internal 10 kΩ
pull-up resistor. Changes to the VDDIOA supply voltage affect the VIH and VOH
values for this pin.
Configurable Input/Output (M6). This pin is a status and control pin when the
device is not in SPI mode.
6, 9, 10, 13, 19,
20, 24, 27, 28,
31, 40, 41, 45
VDD
Input
Input
Power
1.8 V Power Supply.
7
LDO0
LDO bypass
APLL0 Loop Filter Voltage Regulator. Connect a 0.22 μF capacitor from the
LDO0 pin to ground. LDO0 is the ac ground reference for the integrated APLL0
loop filter.
8
LF0
Input/
output
Loop filter
for APLL0
Loop Filter Node for APLL0. Connect a 3.9 nF capacitor from the LF0 pin to Pin 7
(LDO0).
11
OUT0AP
Output HCSL, LVDS, PLL0 Output 0A.
CML
Rev. 0 | Page 39 of 205
AD9546
Data Sheet
Input/
Pin No.
Mnemonic Output Pin Type
Description
12
OUT0AN
OUT0BP
OUT0BN
Output HCSL, LVDS, PLL0 Complementary Output 0A.
CML
Output HCSL, LVDS, PLL0 Output 0B.
CML
Output HCSL, LVDS, PLL0 Complementary Output 0B.
CML
14
15
16, 21, 44
17
DNC
OUT0CP
DNC
No connect
Do Not Connect. Leave the DNC pins floating.
Output HCSL, LVDS, PLL0 Output 0C.
CML
18
22
23
25
26
29
30
OUT0CN
OUT1BP
OUT1BN
OUT1AP
OUT1AN
LF1
Output HCSL, LVDS, PLL0 Complementary Output 0C.
CML
Output HCSL, LVDS, PLL1 Output 1B.
CML
Output HCSL, LVDS, PLL1 Complementary Output 1B.
CML
Output HCSL, LVDS, PLL1 Output 1A.
CML
Input/
HCSL, LVDS, PLL1 Complementary Output 1A.
Output CML
Input/
output
Loop filter
for APLL1
Loop Filter Node for APLL1. Connect a 3.9 nF capacitor from the LF1 pin to Pin 30
(LDO1).
LDO1
Input
LDO bypass
APLL1 Loop Filter Voltage Regulator. Connect a 0.22 μF capacitor from the
LDO1 pin to ground. LDO1 is the ac ground reference for the integrated APLL1
loop filter.
32, 33, 35, 36, 37 M0, M1,
Input/
M2, M3, M4 output
CMOS
Power
Configurable Input/Output Pins. Mx are status and control pins. Changes to the
VDDIOB supply voltage affect the VIH and VOH values for these Mx pins. M3 and
M4 have internal 100 kΩ pull-down resistors. M0, M1, and M2 do not have
internal resistors.
Mx Pin Power Supply. The VDDIOB power supply powers the digital section
that controls the M0 to M4 pins. Valid supply voltages are 1.8 V, 2.5 V, or 3.3 V.
The VDDIOB pin can be connected to the VDD supply bus if 1.8 V operation is
desired.
Reference B Input. This internally biased REFB input is typically ac-coupled.
When configured in this manner, REFB can accept any differential signal with a
single-ended swing up to the VDD power supply. If dc-coupled, the REFB input
can be LVDS or single-ended 1.8 V CMOS.
34
38
39
42
VDDIOB
REFB
Input
Input
Input
Input
1.8 V single-
ended or
differential
input
REFBB
XOA
1.8 V single- Reference BB Input or Complementary Reference B Input. If REFB is in
ended or
differential
input
Differential
input
differential mode, the REFB complementary signal is on the REFBB pin. No
connection is necessary to the REFBB pin if REFB is a single-ended input and
REFBB is not used.
System Clock Input. XOA contains internal dc biasing and must be ac-coupled with
a 0.1 μF capacitor except when using a crystal. When a crystal is used, connect
the crystal across XOA and XOB. A single-ended CMOS input is also an option,
but it can produce spurious spectral content when the duty cycle is not 50%.
When using XOA as a single-ended input, connect a 0.1 μF capacitor from XOB
to ground.
43
46
47
XOB
Input
Input
Input
Differential
input
Complementary System Clock Input. XOB is the complementary signal to XOA.
XOB contains internal dc biasing and must be ac-coupled with a 0.1 μF capacitor
except when using a crystal. When a crystal is used, connect the crystal across
XOA and XOB.
Reference AA input or Complementary REFA Input. If REFA is in differential
mode, the REFA complementary signal is on the REFAA pin. No connection is
necessary to the REFAA pin if REFA is a single-ended input and REFAA is not
used. If dc-coupled, the REFAA input is single-ended 1.8 V CMOS.
Reference A Input. The internally biased REFA input is typically ac-coupled.
When REFA is configured in this manner, REFA can accept any differential
signal with a single-ended swing up to the VDD power supply. If dc-coupled,
the input can be LVDS or single-ended 1.8 V CMOS.
REFAA
REFA
1.8 V single-
ended or
differential
input
1.8 V single-
ended or
differential
input
Rev. 0 | Page 40 of 205
Data Sheet
AD9546
Input/
Pin No.
Mnemonic Output Pin Type
Description
48
RESETB
Input
CMOS logic
Active Low Chip Reset. The RESETB pin has an internal 100 kΩ pull-up resistor.
When asserted, the chip goes into reset. Changes to the VDDIOA supply voltage
affect the VIH values for RESETB.
EPAD
Output Exposed
pad
Exposed Pad. The exposed pad is the ground connection on the chip. The
exposed pad must be soldered to the analog ground of the PCB to ensure proper
functionality and for heat dissipation, noise, and mechanical strength benefits.
Rev. 0 | Page 41 of 205
AD9546
Data Sheet
TYPICAL PERFORMANCE CHARACTERISTICS
–30
–30
–40
INTEGRATED RMS JITTER
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 213fs
–40
(12kHz TO 20MHz): 224fs
f
OUT = 155.52MHz
f
OUT = 125.0MHz
–50
–50
PHASE NOISE (dBc/Hz):
PHASE NOISE (dBc/Hz):
–60
–60
10Hz
–81
10Hz
–84
100Hz
1kHz
–98
100Hz
1kHz
–106
–120
–131
–136
–147
–160
–160
–163
–70
–70
–118
–128
–134
–144
–158
–159
–161
10kHz
100kHz
1MHz
10MHz
>30MHz
FLOOR
10kHz
100kHz
1MHz
10MHz
>30MHz
FLOOR
–80
–80
–90
–90
–100
–110
–120
–130
–140
–150
–160
–170
–100
–110
–120
–130
–140
–150
–160
–170
10
100
1k
10k
100k
1M
100M
10
100
1k
10k
100k
1M
100M
10M
10M
FREQUENCY (Hz)
FREQUENCY (Hz)
Figure 3. Absolute Phase Noise (PLL0, Device Configuration 1, HCSL Mode,
fREF = 38.88 MHz, fOUT = 155.52 MHz, fOSC = 52 MHz Crystal,
50 Hz DPLL Bandwidth)
Figure 6. Absolute Phase Noise (PLL0, Device Configuration 4, HCSL Mode,
fREF = 125 MHz, fOUT = 125.0 MHz, fCOMP = 19.2 MHz TCXO,
fOSC = 52 MHz Crystal, 0.1 Hz DPLL Bandwidth, Phase Buildout Mode)
–30
–30
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 217fs
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 220fs
–40
–40
f
OUT = 312.5MHz
f
OUT = 245.76MHz
–50
–60
–50
–60
PHASE NOISE (dBc/Hz):
PHASE NOISE (dBc/Hz):
10Hz
100Hz
1kHz
10kHz
100kHz
1MHz
10MHz
>30MHz
FLOOR
–74
10Hz
100Hz
1kHz
10kHz
100kHz
1MHz
10MHz
>30MHz
FLOOR
–77
–91
–93
–70
–70
–112
–123
–128
–138
–154
–157
–161
–114
–125
–130
–140
–156
–158
–161
–80
–80
–90
–90
–100
–110
–120
–130
–140
–150
–160
–170
–100
–110
–120
–130
–140
–150
–160
–170
10
100
1k
10k
100k
1M
100M
10M
10
100
1k
10k
100k
1M
100M
10M
FREQUENCY (Hz)
FREQUENCY (Hz)
Figure 7. Absolute Phase Noise (PLL0, Device Configuration 5, HCSL Mode,
fREF = 25 MHz, fOUT = 312.5 MHz, fOSC = 52 MHz Crystal,
50 Hz DPLL Bandwidth, Phase Buildout Mode)
Figure 4. Absolute Phase Noise (PLL0, Device Configuration 2, HCSL Mode,
fREF = 30.72 MHz, fOUT = 245.76 MHz, fOSC = 52 MHz Crystal,
50 Hz DPLL Bandwidth)
–30
–30
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 230fs
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 234.5fs
–40
–40
f
OUT = 174.7MHz
fOUT = 491.52MHz
PHASE NOISE (dBc/Hz):
–50
–60
–50
–60
PHASE NOISE (dBc/Hz):
10Hz
100Hz
1kHz
10kHz
100kHz
1MHz
10MHz
>30MHz
FLOOR
–82
10Hz
100Hz
1kHz
–74
–99
–89
–70
–70
–80
–117
–127
–133
–143
–157
–158
–160
–108
–119
–123
–134
–152
–155
–159
10kHz
100kHz
1MHz
–80
–90
–90
10MHz
>30MHz
FLOOR
–100
–110
–120
–130
–140
–150
–160
–170
–100
–110
–120
–130
–140
–150
–160
–170
10
100
1k
10k
100k
1M
100M
10M
10
100
1k
10k
100k
1M
100M
10M
FREQUENCY (Hz)
FREQUENCY (Hz)
Figure 5. Absolute Phase Noise (PLL0, Device Configuration 3, HCSL Mode,
Figure 8. Absolute Phase Noise (PLL0, Device Configuration 6, HCSL Mode,
fREF = 155.52 MHz, fOUT = 174.7 MHz, fOSC = 52 MHz Crystal,
50 Hz DPLL Bandwidth, Phase Buildout Mode)
f
REF = 1 Hz, fOUT = 491.52 MHz, fCOMP = 19.2 MHz TCXO,
fOSC = 52 MHz Crystal, 0.05 Hz DPLL Bandwidth)
Rev. 0 | Page 42 of 205
Data Sheet
AD9546
–30
–40
–30
–40
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 255fs
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 243fs
fOUT = 155.52MHz
PHASE NOISE (dBc/Hz):
f
OUT = 125.0MHz
–50
–50
PHASE NOISE (dBc/Hz):
–60
–60
10Hz
–81
10Hz
–83
100Hz
1kHz
–98
100Hz
1kHz
–106
–120
–131
–135
–145
–160
–160
–163
–70
–70
–118
–128
–132
–143
–158
–160
–162
10kHz
100kHz
1MHz
10MHz
>30MHz
FLOOR
10kHz
100kHz
1MHz
10MHz
>30MHz
FLOOR
–80
–80
–90
–90
–100
–110
–120
–130
–140
–150
–160
–170
–100
–110
–120
–130
–140
–150
–160
–170
10
100
1k
10k
100k
1M
100M
10
100
1k
10k
100k
1M
100M
10M
10M
FREQUENCY (Hz)
FREQUENCY (Hz)
Figure 9. Absolute Phase Noise (PLL1, Device Configuration 1, HCSL Mode,
fREF = 38.88 MHz, fOUT = 155.52 MHz, fOSC = 52 MHz Crystal,
50 Hz DPLL Bandwidth)
Figure 12. Absolute Phase Noise (PLL1, Device Configuration 4, HCSL Mode,
fREF = 125 MHz, fOUT = 125 MHz, fCOMP = 19.2 MHz TCXO,
fOSC = 52 MHz Crystal, 0.1 Hz DPLL Bandwidth, Phase Buildout Mode)
–30
–30
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 266fs
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 280fs
–40
–40
f
OUT = 312.5MHz
fOUT = 245.76MHz
PHASE NOISE (dBc/Hz):
–50
–60
–50
–60
PHASE NOISE (dBc/Hz):
10Hz
100Hz
1kHz
10kHz
100kHz
1MHz
10MHz
>30MHz
FLOOR
–73
10Hz
100Hz
1kHz
–76
–91
–93
–70
–70
–80
–112
–122
–125
–137
–154
–158
–161
–114
–124
–127
–138
–156
–159
–161
10kHz
100kHz
1MHz
–80
–90
–90
10MHz
>30MHz
FLOOR
–100
–110
–120
–130
–140
–150
–160
–170
–100
–110
–120
–130
–140
–150
–160
–170
10
100
1k
10k
100k
1M
100M
10M
10
100
1k
10k
100k
1M
100M
10M
FREQUENCY (Hz)
FREQUENCY (Hz)
Figure 13. Absolute Phase Noise (PLL1, Device Configuration 5, HCSL Mode,
Figure 10. Absolute Phase Noise (PLL1, Device Configuration 2, HCSL Mode,
fREF = 30.72 MHz, fOUT = 245.76 MHz, fOSC = 52 MHz Crystal,
50 Hz DPLL Bandwidth)
f
REF = 25 MHz, fOUT = 312.5 MHz, fOSC = 52 MHz Crystal, 50 Hz DPLL
Bandwidth, Phase Buildout Mode)
–30
–40
–30
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 264fs
INTEGRATED RMS JITTER
(12kHz TO 20MHz): 322.7fs
–40
f
OUT = 174.7MHz
fOUT = 491.52MHz
PHASE NOISE (dBc/Hz):
–50
–50
–60
PHASE NOISE (dBc/Hz):
–60
10Hz
100Hz
1kHz
10kHz
100kHz
1MHz
10MHz
>30MHz
FLOOR
–77
10Hz
100Hz
1kHz
–74
–99
–90
–70
–70
–80
–117
–127
–131
–142
–158
–159
–161
–108
–118
–120
–131
–150
–154
–160
10kHz
100kHz
1MHz
–80
–90
–90
10MHz
>30MHz
FLOOR
–100
–110
–120
–130
–140
–150
–160
–170
–100
–110
–120
–130
–140
–150
–160
–170
10
100
1k
10k
100k
1M
100M
10M
10
100
1k
10k
100k
1M
100M
10M
FREQUENCY (Hz)
FREQUENCY (Hz)
Figure 14. Absolute Phase Noise (PLL1, Device Configuration 6, HCSL Mode,
fREF = 155.52 MHz, fOUT = 174.7 MHz, fOSC = 52 MHz Crystal,
50 Hz DPLL Bandwidth, Phase Buildout Mode)
Figure 11. Absolute Phase Noise (PLL1, Device Configuration 3, HCSL Mode,
fREF = 1 Hz, fOUT = 491.52 MHz, fCOMP = 19.2 MHz TCXO,
fOSC = 52 MHz Crystal, 0.05 Hz DPLL Bandwidth)
Rev. 0 | Page 43 of 205
AD9546
Data Sheet
1.0
0.8
800
700
600
500
400
300
200
100
0.6
0.6
0.2
0
–0.2
–0.4
–0.6
–0.8
–1.0
CML
HCSL
0
0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
0
1
2
3
4
5
6
7
8
9
10 11 12
TIME (Seconds)
TIME (ns)
Figure 15. DC-Coupled, Single-Ended, 1 Hz Output Waveforms Using HCSL
7.5 mA and 15 mA Mode Terminated 50 Ω to Ground per Figure 46;
Slew Rate ≈ 7 V/ns for 15 mA Mode; ~3.5 V/ns for 7.5 mA Mode
Figure 18. 156.25 MHz Output Waveform (15 mA Driver Setting);
HCSL Drivers Terminated 50 Ω to Ground per Figure 39;
CML Drivers Terminated 50 Ω to 1.8 V per Figure 40
1.0
0.8
1.0
CML
HCSL
0.8
0.6
0.6
0.4
0.6
0.2
0.2
0
0
–0.2
–0.4
–0.6
–0.2
–0.4
–0.6
–0.8
–1.0
–0.8
HCSL, 7.5mA (SLEW RATE ~2.4V/ns)
CML, 7.5mA (SLEW RATE ~2.7V/ns)
–1.0
HCSL, 15mA (SLEW RATE ~5.4V/ns)
CML, 15mA (SLEW RATE ~6V/ns)
–1.2
0
50
100
150
200
250
300
350
400
0
1
2
3
4
5
6
7
8
9
10 11 12
TIME (μs)
TIME (ns)
Figure 16. 8 kHz Output Waveforms for Various Driver Settings;
HCSL Drivers Terminated 50 Ω to Ground per Figure 39;
CML Drivers Terminated 50 Ω to 1.8 V per Figure 40
Figure 19. 312.5 × 66 MHz/64 MHz Output Waveform (15 mA Driver
Setting); HCSL Drivers Terminated 50 Ω to Ground per Figure 39;
CML Drivers Terminated 50 Ω to 1.8 V per Figure 40
1.0
1.0
0.8
HCSL, 7.5mA, 10MHz
CML, 7.5mA, 10MHz
HCSL, 15mA, 10MHz
CML, 15mA, 10MHz
0.8
0.6
0.6
0.6
0.4
0.2
0.2
0
0
–0.2
–0.4
–0.6
–0.2
–0.4
–0.6
–0.8
–1.0
CML
HCSL
–0.8
–1.0
0
20
40
60
80
100 120 140 160 180 200
0
1
2
3
4
5
TIME (ns)
TIME (ns)
Figure 20. 491.52 MHz Output Waveform (15 mA Driver Setting);
HCSL Drivers Terminated 50 Ω to Ground per Figure 39;
CML Drivers Terminated 50 Ω to 1.8 V per Figure 40
Figure 17. 10 MHz Output Waveforms for Various Driver Settings;
HCSL Drivers Terminated 50 Ω to Ground per Figure 39;
CML Drivers Terminated 50 Ω to 1.8 V per Figure 40
Rev. 0 | Page 44 of 205
Data Sheet
AD9546
5
0
2000
1800
1600
1400
1200
1000
800
BW = 0.01Hz
BW = 0.1Hz
BW = 1Hz
BW = 10Hz
BW = 100Hz
–5
–10
–15
–20
–25
–30
600
CML, DIFFERENTIAL, 15mA
PHASE MARGIN: 88.5°
PEAKING: <0.1dB
ATTENUATION:
CML, DIFFERENTIAL, 12.5mA
CML, DIFFERENTIAL, 7.5mA
HCSL, DIFFERENTIAL, 15mA
HCSL, DIFFERENTIAL, 12.5mA
HCSL, DIFFERENTIAL, 7.5mA
400
200
0
AT LOOP BW: 5.5dB
2× LOOP BW: 18dB/DECADE
100× LOOP BW: 41dB/DECADE
1000× LOOP BW: 60dB/DECADE
1
10
100
1k
10k 100k
1M
10M 100M 1G
0.001
0.01
0.1
1
10
100
1k
FREQUENCY (Hz)
OFFSET FREQUENCY (Hz)
Figure 21. Differential Peak-to-Peak Voltage Swing vs. Frequency;
HCSL Drivers Terminated 50 Ω to Ground per Figure 39;
CML Drivers Terminated 50 Ω to 1.8 V per Figure 40
Figure 24. DPLL Closed-Loop Transfer Function High Phase Margin Loop
Filter Setting
150
1.9
1.8
1.7
1.6
DIFFERENTIAL CML OUTPUTS- UNCOMPENSATED
125
100
75
SINGLE- ENDED CML OUTPUTS- UNCOMPENSATED
DIFFERENTIAL CML OUTPUTS- COMPENSATED
SINGLE- ENDED CML OUTPUTS- COMPENSATED
C
C
C
= 5pF
LOAD
LOAD
LOAD
1.5
1.4
1.3
1.2
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
= 10pF
= 20pF
50
25
0
–25
–50
–75
–50 –40 –30 –20 –10
0
10
20
30
40
50
–50 –40 –30 –20 –10
0
10 20 30 40 50 60 70 80 90 100 110
TIME (ns)
AD9546 DIE TEMPERATURE (°C)
Figure 25. Propagation Delay Variation vs. AD9546 Die Temperature; CML
Output Mode; 1.2 V CMOS Reference Input Mode; fSYS = 2392 MHz; Internal
Zero Delay (Hitless) Mode
Figure 22. Mx Pin Waveforms for Various Load Conditions; Normal Output
Drive Strength (Default); VDDIOx = 1.8 V
150
5
0
DIFFERENTIAL CML OUTPUTS- UNCOMPENSATED
125
100
75
SINGLE- ENDED CML OUTPUTS- UNCOMPENSATED
DIFFERENTIAL CML OUTPUTS- COMPENSATED
SINGLE- ENDED CML OUTPUTS- COMPENSATED
–5
–10
50
BW = 0.01Hz
25
BW = 0.1Hz
–15
BW = 1Hz
BW = 10Hz
BW = 100Hz
0
–20
PHASE MARGIN:70°
PEAKING: 1.07dB
ATTENUATION:
–25
–50
–75
AT LOOP BW: 2.5dB
2× LOOP BW: 21dB/
DECADE
100× LOOP BW: 60dB/
DECADE
–25
–30
–40 –30 –20 –10
0
10 20 30 40 50 60 70 80 90 100 110
0.001
0.01
0.1
1
10
100
1000
OFFSET FREQUENCY (Hz)
AD9546 DIE TEMPERATURE (°C)
Figure 23. DPLL Closed-Loop Transfer Function Nominal Phase Margin
Loop Filter Setting
Figure 26. Propagation Delay Variation vs. AD9546 Die Temperature; HCSL
Output Mode; 1.2 V CMOS Reference Input Mode; fSYS = 2392 MHz; Internal
Zero Delay (Hitless) Mode
Rev. 0 | Page 45 of 205
AD9546
Data Sheet
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
HYSTERESIS = 25.0%
HYSTERESIS = 12.5%
HYSTERESIS = NONE
HYSTERESIS = 25.0%
HYSTERESIS = 12.5%
HYSTERESIS = NONE
REFERENCE FREQUENCY INPUT OFFSET FROM
PROGRAMMED PART VALUE (ppb)
REFERENCE FREQUENCY INPUT OFFSET FROM
PROGRAMMED PART VALUE (ppb)
Figure 27. Reference Valid Probability vs. Reference Frequency Input Offset
from Programmed Part Value; Tolerance = 4.6 ppm
Figure 29. Reference Valid Probability vs. Reference Frequency Input Offset
from Programmed Part Value; Tolerance = 100 ppm
1.1
1.0
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
HYSTERESIS = 25.0%
HYSTERESIS = 12.5%
0.1
HYSTERESIS = NONE
0
REFERENCE FREQUENCY INPUT OFFSET FROM
PROGRAMMED PART VALUE (ppb)
Figure 28. Reference Valid Probability vs. Reference Frequency Input Offset
from Programmed Part Value; Tolerance = 25 ppm
Rev. 0 | Page 46 of 205
Data Sheet
AD9546
TERMINOLOGY
Zero Delay
Phase Buildout Switchover
Zero delay is a PLL-based clock architecture that establishes
zero phase offset between the output clock signal and input
clock signal. A zero delay architecture requires an Integer N
type PLL, by definition, because there must be a harmonic
relationship between the input and output frequencies to
preserve a static phase relationship between input and output
edges. Note that zero delay provides a zero input to output phase
offset in the static (steady state) sense only. That is, transient
events may cause the PLL to slew its output phase, which tempo-
rarily breaks the zero delay condition (for example, when the
PLL is in the process of phase or frequency acquisition). An
example of a transient event is when a multiple input PLL
switches from one input reference signal to another.
Phase buildout switchover applies to PLLs that can switch
between multiple input clock signals (usually operating at the
same frequency). Generally, the multiple input clock signals are
not phase aligned. Therefore, switching from one to another
can create a substantial phase transient in the output clock
signal. Phase buildout capability enables the PLL to absorb the
phase difference between the original and the switched to
reference clock signals, thereby preventing a phase disturbance
in the output clock signal (unlike hitless switchover, which
gradually slews the output phase to the new input phase).
Because phase buildout operation prevents a disturbance in the
phase of the output clock signal, there is no guarantee of phase
alignment between the input and output clock signals. Unlike
hitless switchover, phase buildout places no restriction on the
output/input frequency ratio.
Hitless Switchover
Hitless switchover applies to PLLs that can switch between
multiple input clock signals (usually operating at the same
frequency). Generally, the multiple input clock signals are not
phase aligned. Therefore, switching from one to another can
create a substantial phase transient in the output clock signal.
Hitless switchover limits the phase transient at the output by
allowing the output clock signal to slew its phase in a prescribed
manner to accommodate the phase of the new reference signal
(typically by limiting the output frequency deviation to some
prescribed maximum value). A PLL employing hitless switchover
capability requires the output/input frequency ratio to be an
integer greater than or equal to one. Otherwise, the output
period constitutes phase in excess of 360° relative to the input
period, which creates unresolvable phase ambiguity.
For more information, see the AN-1420 Application Note.
Time Base
Time base is a frequency source, the period of which denotes
the passage of time.
Time Scale
Time scale is the accumulation of the periods of a time base. A
time scale allows identifying a particular instant as having a
particular time value. The starting point of a time scale (its
epoch) is arbitrary.
Epoch
Epoch is the starting time associated with a time scale. For
example, the epoch of the UTC time scale is midnight,
January 1, 1970.
Note that the output phase transient limitation imposed by hitless
switchover also applies any time the PLL is in the process of
phase or frequency acquisition (that is, it is not necessarily
limited to a reference clock switchover). A phase transient
limitation exists in hitless operation because the primary enabler
for hitless operation is the use of a narrow loop bandwidth,
which governs the transient characteristics of the PLL in
general.
Time Stamp
In the context of this data sheet, time stamp is a digital
(numeric) time value with an arbitrary epoch suitable for
internal device usage.
Time Code
In the context of this data sheet, time code is a time stamp with
a defined epoch suitable for general usage external to the
device.
Rev. 0 | Page 47 of 205
AD9546
Data Sheet
THEORY OF OPERATION
Figure 1 summarizes the fundamental building blocks of the
AD9546. The core of the device comprises two independent
PLLs, PLL0 and PLL1, as well as proprietary resources enabling
the digital transport of clock signals (digitized clocking)
between multiple AD9546 devices over a single digital link.
Digitized clocking streamlines the design of systems that rely
on the transport of time scales between spatially separated
devices. The digitized clocking functionality comprises a common
clock DPLL (CCDPLL), a common clock Synchronizer (CCS),
nine user time stampers (UTS 0 to UTS 8), and two inverse user
time stampers (IUTS 0 and IUTS 1).
The user also has access to two independent auxiliary NCOs.
These NCOs provide for internally generated frequencies up to
65,535 Hz with a tuning resolution of 2−40 Hz (approximately
1 pHz).
The DPLL portion of each PLL channel includes a programmable
digital loop filter with extremely low loop bandwidth capability.
Low loop bandwidth capability enables a significant reduction
of jitter transfer from the reference input to the output. Each
DPLL provides precise frequency translation by means of a 48-bit
NCO capable of generating clock signals in the range of 162 MHz
to 350 MHz. Furthermore, both DPLLs enable support of
manual or automatic transition to holdover operation.
The AD9546 provides a high degree of backward compatibility
(legacy) with the AD9545, but with enhancements to support
digitized clocking. The user activates most enhancements
individually via register programming. Activating certain
enhancements may affect legacy functionality and override
settings in the registers associated with legacy functionality. In
such cases, text describing a specific enhancement in this data
sheet expressly states any impact on legacy functionality or
register overrides.
Holdover operation constitutes the device generating an output
frequency synthesized from the system clock (as opposed to
translating a frequency from a reference input to an output).
During holdover operation, the AD9546 continuously provides
an output clock signal while the system clock is present. The
AD9546 is also capable of setting the holdover output
frequency based on a time average of the output frequency
history prior to the transition to holdover operation.
The AD9546 synthesizes its system clock via a system clock PLL
that upconverts the frequency applied to the system clock input
pins, XOA and XOB. The system clock input pins accept
frequencies from 20 MHz to 300 MHz from an external clock
source. Alternatively, the user can connect a crystal resonator
(25 MHz to 80 MHz) directly across the XOA and XOB pins.
The output of the system clock PLL is the primary time base for
time keeping functions within the AD9546.
The output signal from the DPLL routes to the input of the APLL,
which upconverts the signal to a range of 2.424 GHz to 3.232 GHz
(APLL0) or 3.232 GHz to 4.040 GHz (APLL1). A high frequency
divider (÷2) follows the APLL VCO output providing a factor of 2
frequency reduction prior to delivery to the clock distribution
section.
The clock distribution section consists of a bank of dividers (Q)
and output drivers dedicated to each output pin. The PLL0 channel
has six Q divider/driver units, whereas the PLL1 channel has
four Q divider/driver units. Each Q divider has 32-bit program-
mable division depth and programmable phase offset control.
The drivers operate up to 500 MHz and are configurable as a
single-ended or differential output with adjustable output
current (7.5 mA, 12.5 mA, or 15 mA).
The AD9546 supports up to eight reference input clock signals (via
the REFA, REFAA, REFB, and REFBB pins and four appropriately
configured Mx pins feeding auxiliary REF0, auxiliary REF1,
auxiliary REF2, and auxiliary REF3). Each reference input has a
dedicated reference divider as well as a reference monitor to detect
when the reference signal is lost or compromised. Furthermore,
when the AD9546 detects a faulty reference, it can automatically
switch to another (valid) reference with the user-programmable
reference selection priority.
The AD9546 includes an integrated temperature sensor and
system clock compensation circuitry, which allows the device to
adjust automatically for frequency drift errors associated with
the system clock source.
Rev. 0 | Page 48 of 205
Data Sheet
AD9546
INPUT/OUTPUT TERMINATION RECOMMENDATIONS
SYSTEM CLOCK INPUTS
C
TUNE
0.1µF
221Ω
XOA
3.3V
CMOS
TCXO
649Ω
XOA
25MHz TO 80MHz
FUNDAMENTAL
AT CUT CRYSTAL
AD9546
XOB
AD9546
XOB
0.1µF
C
TUNE
Figure 30. System Clock Input (XOA/XOB) Using the Crystal Path
(CTUNE = 2 × (CLOAD − CSTRAY), Where Typical CSTRAY = 2 pF to 5 pF); see the Crystal
Path Section for Capacitor Definitions)
Figure 31. System Clock Input (XOA, XOB) Using the Direct
Path with a TCXO or Oven Controlled Crystal Oscillator (OCXO) Having a
3.3 V CMOS Output
REFERENCE CLOCK INPUTS
REFx
AD9546
AD9546
OR
1.8V RECEIVER
REFxx
R1
REFx
CMOS
DRIVER
R2
+
High-Z
INPUT
100Ω
–
LVDS DRIVER
(1.2V COMMON MODE)
REFxx
LVDS
RECEIVER
INTERNAL DC THRESHOLD
FOR 1.2V OR 1.8V CMOS
Figure 35. Differential LVDS Input Mode
Figure 32. Single-Ended DC-Coupled Mode, 1.2 V or 1.8 V CMOS
REFx
1.2V
1.2V
OR
R
DIFFERENTIAL
L
AD9546
AD9546
47kΩ
REFxx
DRIVER
(OPTIONAL)
1.8V RECEIVER
DRIVER
47kΩ
0.1µF
47kΩ
47kΩ
0.1µF
0.1µF
REFx
0.6V
0.6V
R
L
0.6V
REFxx
47kΩ
47kΩ
Figure 36. Differential AC-Coupled Mode
(RL = 100 Ω Is Recommended, Except For HCSL)
Figure 33. Single-Ended AC-Coupled Interface Mode
REFx
1.2V
OR
REFxx
DIFFERENTIAL DRIVER
AD9546
AD9546
47kΩ
(0.6V COMMON MODE)
DRIVER
REFx
0.6V
R
L
REFxx
Figure 37. Differential DC-Coupled Mode
Figure 34. Single-Ended Internal Pull-Up Resistor Mode
Rev. 0 | Page 49 of 205
AD9546
Data Sheet
CLOCK OUTPUTS
AD9546
OUTxyP
50Ω
0.1µF
0.1µF
I
Qxy
LVDS OR
LVPECL
RECEIVER
+1.8V
I
100Ω
PLLx
50Ω
Qxyy
OUTxyN
I
MODE
SOURCE
Figure 38. LVDS-Compatible Output Swing, AC-Coupled (375 mV (Differential) Across 100 Ω Termination Resistor for I = 15 mA)
The output drivers of the AD9546 are capable of either
The Qxy divider has both normal and inverted logic outputs,
which means when the Qxy divider is connected to both output
drivers (as in Figure 38, for example), one current source is on
while the other is off, allowing a differential output signal. See
the Distribution Clock Output Drivers section for configuring
the clock output drivers.
sourcing or sinking current. Each pin of an output pin pair
(OUTxyP/OUTxyN) has an independent current source.
However, the source or sink function applies to both current
sources of an output pin pair (that is, either the source current
(ISOURCE) or sink current (ISINK) of both pin drivers). The current
source or sink current of the individual driver is either on or off,
depending on the output logic level of the associated Qxy or
Qxyy divider (0 = off, 1 = on).
AD9546
AD9546
OUTxyP
OUTxyP
63.4Ω 237Ω
I
I
I
50Ω
Qxy
Qxy
TO HCSL
RECEIVER
1.8V
R
+1.8V
I
L
PLLx
PLLx
63.4Ω 237Ω
50Ω
Qxyy
Qxyy
OUTxyN
OUTxyN
I
MODE
I
MODE
SOURCE
SINK
Figure 39. HCSL Output, Peak-to-Peak Voltage ≈ 750 mV per Section
(I = 15 mA)
Figure 42. LVDS-Compatible Output, 1.2 V Common-Mode, Thevenin Bias
Network (I = 7.5 mA; I = 15 mA with Extra 100 Ω Termination, RL)
AD9546
AD9546
OUTxyP
OUTxyP
56.2Ω 453Ω
I
I
50Ω
Qxy
Qxy
1.2V, 1.5V
OR 1.8V
1.8V
PLLx
PLLx
56.2Ω 453Ω
I
50Ω
I
Qxyy
Qxyy
OUTxyN
OUTxyN
I
MODE
I
MODE
SINK
SINK
Figure 40. CML Output (I = 7.5 mA; I = 15 mA Options for 1.5 V or 1.8 V
Supply)
Figure 43. 2.5 V LVPECL or Double Amplitude LVDS-Compatible Boost
Output, 1.5 V p-p, 1.24 V Common Mode (I = 15 mA)
AD9546
AD9546
OUTxyP
OUTxyP
I
I
50Ω
I
R
L
L
Qxy
Qxy
50Ω
50Ω
IN PHASE
OUTPUTS
1.8V
R
+1.8V
I
L
PLLx
PLLx
10nF
R
Qxyy
Qxyy
OUTxyN
OUTxyN
I
MODE
I
MODE
SOURCE
SINK
Figure 41. LVDS-Compatible Output, 1.24 V Common Mode, T Network
(I = 7.5 mA; I = 15 mA with Extra 100 Ω Termination, RL)
Figure 44. Single-Divider, Single-Ended Mode Providing In Phase Outputs
(Current Source Mode)
Rev. 0 | Page 50 of 205
Data Sheet
AD9546
AD9546
AD9546
OUTxyP
OUTxyP
I
I
I
I
R
R
L
L
L
L
Qxy
Qxy
IN PHASE
OUTPUTS
INDEPENDENT
OUTPUTS
+1.8V
+1.8V
PLLx
PLLx
R
R
Qxyy
Qxyy
OUTxyN
OUTxyN
I
MODE
I
MODE
SINK
SINK
DUAL-DIVIDER, SINGLE-ENDED MODE (CURRENT SINK)
Figure 45. Single-Divider, Single-Ended Mode Providing In Phase Outputs
(Current Sink Mode)
Figure 47. Dual-Divider, Single-Ended Mode Providing Independent Outputs
(Current Sink Mode)
AD9546
OUTxyP
I
R
L
L
Qxy
INDEPENDENT
OUTPUTS
+1.8V
I
PLLx
R
Qxyy
OUTxyN
I
MODE
SOURCE
Figure 46. Dual-Divider, Single-Ended Mode Providing Independent Outputs
(Current Source Mode)
Rev. 0 | Page 51 of 205
AD9546
Data Sheet
DIGITIZED CLOCKING
any one of the time stamp sources, which is the reason those
components include a time stamp selection mux at their input.
OVERVIEW
Digitized clocking enables precise timing transport (frequency
and phase) among spatially separated devices by the exchange
of numeric time codes to synchronize local clocks (assuming all
devices have access to a common reference clock signal).
SYSTEM CLOCK PLL COMPONENT
The system clock PLL typically uses an external crystal
resonator as a frequency source, which tends to offer optimal
overall phase noise performance. The system clock PLL
synthesizes a high frequency internal system clock signal
(~2.4 GHz) from the external frequency source, which provides
the basic internal timing for the device. The TDCs of the device
use the system clock to generate time stamps (see the Time to
Digital Converter (TDC) section).
A digitized clocking system comprises multiple timing nodes
with each node relying on a shared external time base
(common clock reference). Because each node shares the same
common clock reference, each node increments at the same
rate, resulting in all nodes having the same time scale. Although
all the nodes have the same time scale, there is no guarantee
they share the same epoch. Digitized clocking relies on a user
provided synchronization signal at each node to establish a
common epoch, ultimately leading to all nodes being time
aligned. A unique characteristic of digitized clocking is the
ability to exchange numeric time codes between nodes via a
digital bus, which provides a means to send multiple clocks
over one digital bus (clock aggregation) instead of having to
route multiple analog clock signals throughout a system. The
AD9546 contains all the necessary building blocks to
Note that the internal system clock signal is one of the two
frequency sources the common clock DPLL component uses to
generate the local time scale.
COMMON CLOCK DPLL COMPONENT
The purpose of the common clock DPLL is to produce a local
time scale. The common clock DPLL uses the system clock for
internal timing, but the phase locks to a reference time base, the
common clock reference. The output of the common clock
DPLL is the local time scale, which constitutes the accumulated
period of cycles of the common clock reference and provides an
internal sense of time for the digitized clocking components.
implement one digitized clocking node.
Figure 48 shows the individual components of a complete
digitized clock node as implemented in the AD9546. These
digitized clocking components include the following:
The inherent closed-loop architecture of the common clock
DPLL results in the generation of an error signal, Correction
Factor 3 (CF3), that effectively compensates for system clock
fluctuations. Note the CF3 signal is available to the system clock
compensation functional block of the AD9546 (see the
Compensation Method 3 section).
•
•
•
•
A system clock PLL for generating a system clock
frequency for internal digital timing functions (see the
System Clock PLL section)
A common clock DPLL to phase lock to a reference time
base (common clock reference) and maintain a local time
scale
A physical clock converter for converting the rising edges
of a physical clock to time stamps derived from the system
clock
A physical clock generator for converting time stamps
derived from the common time scale to a physical clock
output signal
PHYSICAL CLOCK CONVERTER
The physical clock converters provide an input to receive an
external clock signal, provide optional integer frequency
division, and convert rising clock edges to numeric time
stamps. The time stamps are then available for use by those
digitized clocking components that require a time stamp
source.
•
•
A CCS for creating a common time scale by assigning an
epoch to the local time scale
A UTS to convert internal time stamps to time codes based
on the common time scale and provide those time codes to
the user for external use
PHYSICAL CLOCK GENERATOR
The physical clock generator makes use of the AD9546 DPLL
channels and output distribution drivers to produce an analog
clock output signal from time stamps. In a digitized clocking
application, the time stamps derive from the common time
scale. Because the DPLL can accept those time stamps as its
reference input, the output of the physical clock generator
consists of analog clock signals.
•
An IUTS to convert a series of time codes based on the
common time scale (a digitized clock signal) provided by
the user to internal time stamps
Figure 48 shows a variety of time stamp sources. Certain
digitized clocking components require the ability to connect to
Rev. 0 | Page 52 of 205
Data Sheet
AD9546
SYSTEM CLOCK
COMMON CLOCK SYNCHRONIZER
SYSCLK (~2.4GHz)
COMMON
CLOCK
DPLL
XOA
SYSTEM
SKEW
REGISTERS
REGISTERS
REGISTERS
DIGCLK (SYSCLK/3)
CORECLK (DIGCLK/32)
TDC TIME SCALE
COMMON
CLOCK SYNC
CLOCK
TIMING
PLL
OFFSET
XOB
SLEW LIMIT
GUARD CONTROLS
CORECLK
REF
FREQ
ASSIGNED TIME
COMMON CLOCK DPLL
REGISTERS
REFERENCE TIME
PRIMARY
SOURCE
TIME
STAMPS
LOCAL
TIME SCALE
SYNC
TIME
COMMON
STAMPS
CLOCK
REFERENCE
MONITOR
CORRECTION
FACTOR
CF
3
STATUS
COMMON TIME SCALE
LOCK
R
DETECTOR
TARGET
PERIOD
PHYSICAL CLOCK GENERATOR
NUM/DEN
CNTL
LOCK
REGISTERS
REGISTERS
STATUS
DRIVERS
SECONDARY
SOURCE
OUTxy
PINS
TIME
COMMON
CLOCK
STAMPS
REFERENCE
MONITOR
DPLL
CHANNEL
TIME
STAMPS
STATUS
REGISTERS
LOOP BW
OUTxy
PINS
R
TARGET
PERIOD
NUM/DEN
REGISTERS
PHYSICAL CLOCK CONVERTER
INVERSE USER TIME STAMPER
TIME
STAMPS
TIME
REGISTERS
REGISTERS
IUTS
COMMON TIME SCALE
LOCAL TIME SCALE
OFFSET
REFx
PINS
TIME
STAMPS
TDC
÷R
REGISTERS
R
USER TIME STAMPER
DIGCLK
CORECLK
TDC TIME SCALE
TIME
STAMPS
UTS
Mx
PINS
TIME
STAMPS
TDC
÷R
TIME
REGISTERS
COMMON TIME SCALE
LOCAL TIME SCALE
OFFSET
REGISTERS
R
REGISTERS
NOTES
1. DIGCLK IS DIGITAL CLOCK.
2. NUM/DEN ARE NUMERATOR/DENOMINATOR.
Figure 48. Digitized Clock Components
Rev. 0 | Page 53 of 205
AD9546
Data Sheet
COMMON CLOCK SYNCHRONIZER COMPONENT
A DIGITIZED CLOCKING NODE EXAMPLE
The common clock synchronizer provides a precision method
for assigning an epoch to the local time scale yielding the
common time scale. Essentially, the user provides a
Figure 49 shows the interaction of the various digitized clocking
components as applied to a single node in a hypothetical
digitized clocking system. The analog input signals consist of
the common clock (CC), a synchronization source (SS), and a
local user clock (LUC). The local user clock is any analog clock
signal intended for transport to another node or for local use as
a local analog clock output signal.
synchronization trigger event, the rising edge of an external
analog input signal, which the device time stamps. The time
stamp serves as an internal trigger event. Immediately following
the trigger event (and before the next trigger event), the user
programs the device with a time code corresponding to the
time associated with the trigger event and asserts an IO update,
which aligns the common time scale with the programmed
time code. In this way, the common time scale carries the
proper time. By carrying out the synchronization process across
all nodes (with a distributed common clock as the time base for
every node) and properly accounting for time delay in the
system, the common time scale of every node increments at the
same rate and carries the same epoch. Thus, all nodes are time
aligned to a high degree of precision.
The three analog input clock signals route through their
respective physical clock converters, which convert the analog
clock signals to time stamps. The common clock time stamps
route to the common clock DPLL, which is responsible for
regulating the local time scale. The synchronization source time
stamps route to the common clock synchronizer, which allows
the user to translate the local time scale to a common time scale
with a known epoch. The local user clock time stamps route to
a UTS, allowing an external processor to capture local user
clock time codes related to the common time scale.
USER TIME STAMPER (UTS) AND INVERSE USER
TIMER STAMPER (IUTS) COMPONENTS
The local user clock time codes captured by the external
processor exist as a digitized clock signal. Thus, the processor
can transport the digitized clock signal in the form of data
packets to remote nodes.
A typical digitized clocking system can use one AD9546 as a
master time unit with various nodes synchronized to the
master. Thus, a means of transporting numeric time codes
between nodes is essential for transporting clocks and
correcting time errors throughout the system. Such is the role
of the UTS and IUTS. The AD9546 possesses nine UTS units
and two IUTS units.
Alternatively, the processor can route the local user clock time
codes from a local UTS to a local IUTS, which converts the
local user clock time codes back into time stamps. The time
stamps are then available to a physical clock generator for
conversion to an analog output signal. This basic example of
translating input clock edges to numeric values (time codes)
and reconstituting an analog clock signal from the time codes
within a single chip demonstrates the concept of digitized
clocking.
A UTS converts internal time stamps to time codes (numeric
time values based on the common time scale) and provides the
user with a means to read the corresponding time codes.
Because time codes relate to the common time scale, the UTS
provides a means to export a digitized clock signal to remote
digitized clocking nodes.
The local user clock time codes can be transported to a remote
node (via the processor) and then converted to an analog signal
at the far end. Assuming the remote node shares the same
common clock (a requirement for digitized clocking), and the
processor at the remote node properly synchronizes the remote
common time scale, then it is possible to replicate the local user
clock at remote nodes in the native time domain of the local
user clock.
An IUTS provides the user with a means to send a digital clock
signal to a node (a digital clock signal being a continuous series
of uniformly increasing numeric time codes). An IUTS
converts the digital clock signal to time stamps (based on the
common time scale), which can be distributed internally as
needed.
Rev. 0 | Page 54 of 205
Data Sheet
AD9546
SYSTEM CLOCK
COMMON CLOCK SYNCHRONIZER
SYSCLK (~2.4GHz)
COMMON
CLOCK
DPLL
XOA
XOB
SYSTEM
CLOCK
PLL
SKEW
REGISTERS
REGISTERS
REGISTERS
DIGCLK (SYSCLK/3)
CORECLK (DIGCLK/32)
TDC TIME SCALE
COMMON
CLOCK SYNC
TIMING
OFFSET
SLEW LIMIT
GUARD CONTROLS
CORECLK
REF
FREQ
ASSIGNED TIME
COMMON CLOCK DPLL
REGISTERS
REFERENCE TIME
PRIMARY
SOURCE
SS
TIME
STAMPS
LOCAL
TIME SCALE
SYNC
CC
TIME
STAMPS
COMMON
CLOCK
REFERENCE
MONITOR
CORRECTION
FACTOR
CF
3
TRANSLATE LOCAL
STATUS
COMMON TIME SCALE
USER CLOCK IN ITS
NATIVE TIME DOMAIN
TO A LOCAL ANALOG
CLOCK SIGNAL
LOCK
DETECTOR
R
TARGET
PERIOD
PHYSICAL CLOCK GENERATOR
NUM/DEN
CNTL
LOCK
REGISTERS
REGISTERS
STATUS
ANALOG
CLOCK SIGNALS
DRIVERS
SECONDARY
SOURCE
OUTXY
PINS
TIME
COMMON
CLOCK
STAMPS
REFERENCE
MONITOR
DPLL
CHANNEL
TIME
STAMPS
STATUS
REGISTERS
LOOP BW
OUTXY
PINS
R
TARGET
PERIOD
NUM/DEN
REGISTERS
PHYSICAL CLOCK CONVERTER
INVERSE USER TIME STAMPER
DIGITIZED
TIME
CLOCK SIGNALS
STAMPS
REGISTERS
REGISTERS
TRANSPORT TIME
TIME
REFx
PINS
IUTS
CODES OF LOCAL
USER CLOCK(S)
FROM REMOTE
NODE(S) TO
COMMON TIME SCALE
LOCAL TIME SCALE
OFFSET
TIME
STAMPS
TDC
÷R
TIME
CODES
ANALOG
CLOCK SIGNALS
LOCAL NODE
REGISTERS
R
COMMON
USER TIME STAMPER
CC
SS
LUC
CLOCK
DIGCLK
TIME
CODES
PROCESSOR
SYNC
SOURCE
CORECLK
TDC TIME SCALE
LOCAL
USER CLOCK
TRANSPORT TIME
CODES OF LOCAL
USER CLOCK(S)
FROM LOCAL
LUC
TIME
TIME
CODES
STAMPS
UTS
Mx
PINS
TIME
STAMPS
TDC
÷R
TIME
OFFSET
COMMON TIME SCALE
LOCAL TIME SCALE
REGISTERS
NODE(S) TO
REMOTE NODE
REGISTERS
R
REGISTERS
Figure 49. Digitized Clocking Node
Rev. 0 | Page 55 of 205
AD9546
Data Sheet
COMMON CLOCK DPLL (CCDPLL)
access to the analog loopback function (see the Analog Clock
Loopback section).
OVERVIEW
The CCDPLL embodies most of the hardware associated with
the auxiliary DPLL (see the Compensation Method 3
subsection of the System Clock Compensation section).
However, the CCDPLL incorporates additional enhancements
to the auxiliary DPLL hardware to meet the needs of a digitized
clocking system. Figure 50 shows a functional block diagram of
the CCDPLL.
CCR Enable and CCR Frequency Source
Each CCR has an enable bit (Bit 7 of Register 0x0D10 and/or
Register 0x0D20 for CCR0 and CCR1, respectively). Logic 0
(default) disables the CCR, whereas Logic 1 enables it. When
operating the device with both CCRs, the enable bit gives the
user a simple mechanism to force a reference switchover (see
the Common Clock Reference Switchover section) by disabling
the active CCR.
COMMON CLOCK REFERENCES
The CCDPLL supports two independent time stamp sources as
common clock reference (CCR) frequency inputs. CCR0 is the
primary frequency reference, whereas CCR1 is the secondary
frequency reference. The primary and secondary distinction
matters with respect to the handling of the CCDPLL of
automatic switching between references (see the Common
Clock Reference Switchover section).
The user can independently assign each CCR to a specific time
stamp source (via Bits[4:0] of Register 0x0D10 and
Register 0x0D20 for CCR0 and CCR1, respectively). The
default value, 0x31, is the null selection. That is, the associated
CCR has no time stamp source.
For CCR0, programming a nondefault value for the enable or
time stamp source selection settings results in the CCR0 time
stamp source selection overriding the reference TDC selection
of the auxiliary DPLL (see the Compensation Method 3
subsection of the System Clock Compensation section).
When employing only a single CCR, the user must use CCR0.
That is, assigning a reference to CCR1 with no reference
assigned to CCR0 is an invalid condition.
It is best practice to use REFB, REFBB, or both input pins for
the common clock DPLL reference frequency because the REFB
and REFBB inputs are the only reference inputs that provide
SYSCLK REF FREQ
SYSTEM CLOCK
LOCAL TIME
CORE
CLOCK
(~25MHz BITS)
REG 0x0206 TO 0x0202, BITS[39:0]
COMPENSATION
SCALE
GENERATOR
CCR0
VALID
SUBSYSTEM
INTERFACE
CCR0 SOURCE
REG 0x0D40, BIT 4
REG 0x0D10, BITS[4:0]
LOCAL
TIME
SCALE
COMP
X
CCDPLL
REF MON
CCR0 INDEX
ACTIVE
BANDWIDTH
REG 0x0286 TO 0x0285, BITS[15:0]
REG 0x0D40, BIT 0
CCR0
TIME STAMPS
FROM
TIME
TIME STAMPS
HOLDOVER
PERIOD ERROR
TIME STAMP
PROCESSOR
LOOP
HOLDOVER
REGISTER
STAMP
CF
3
FILTER
SOURCES
ENBL0
REG 0x0D10, BIT 7
COMMON
CLOCK
LOOP
REG 0x0D02, BITS[7:0] FILL
CONTROLLER
ENBL1
REG 0x0D20, BIT 7
CCR1
REG 0x0D03, BITS[7:0] DRAIN
REG 0x0D01 TO 0x0D00, BITS[15:0]
REG 0x0D05 TO 0x0D04, BITS[15:0]
LOCK
DETECT
THRESHOLD
STABILITY
REG 0x0D40, BIT 1
FROM
TIME
TIME STAMPS
REG 0x3002, BIT 1
LOCKED
STAMP
SOURCES
REG 0x0403 TO 0x0400, BITS[29:0]
TARGET
REFERENCE
PERIOD
REG 0x0423 TO 0x0420, BITS[29:0]
REG 0x0443 TO 0x0440, BITS[29:0]
REG 0x0463 TO 0x0460, BITS[29:0]
REG 0x0483 TO 0x0480, BITS[29:0]
REG 0x04A3 TO 0x04A0, BITS[29:0]
REG 0x04C3 TO 0x04C0, BITS[29:0]
REG 0x04E3 TO 0x04E0, BITS[29:0]
CCDPLL
REF MON
CCR1 INDEX
R
REG 0x0D20, BITS[4:0]
ACTIVE
CCR1
VALID
CCR1 SOURCE
REG 0x0D40, BIT 5
SOURCE
ACTIVE
SOURCE
REG 0x0D40, BIT 3
CCR0 INDEX
CCR1 INDEX
0
1
NUM0
DEN0
REG 0x0D15 TO 0x0D12, BITS[31:0]
REG 0x0D1A TO 0x0D16, BITS[39:0]
REG 0x040B TO 0x0404, BITS[59:0]
REG 0x042B TO 0x0424, BITS[59:0]
REG 0x044B TO 0x0444 ,BITS[59:0]
REG 0x046B TO 0x0464, BITS[59:0]
REG 0x048B TO 0x0484, BITS[59:0]
REG 0x04AB TO 0x04A4, BITS[59:0]
REG 0x04CB TO 0x04C4, BITS[59:0]
REG 0x04EB TO 0x04E4, BITS[59:0]
0
1
TARGET
NUM
DEN
REFERENCE
PERIOD
PROCESSOR
REG 0x0D25 TO 0x0D22, BITS[31:0]
REG 0x0D2A TO 0x0D26, BITS[39:0]
REF
NUM1
DEN1
UPDATES ON ARRIVAL
OF TIME STAMPS
FROM ACTIVE CCR
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
Figure 50. Common Clock DPLL Block Diagram
Rev. 0 | Page 56 of 205
Data Sheet
AD9546
lock threshold parameter. Otherwise, the phase lock fill and
drain parameters have no function, because they are
meaningless in the context of the legacy auxiliary DPLL.
COMMON CLOCK REFERENCE MONITOR
The reference monitor for CCR0 uses the reference monitor
hardware of the legacy auxiliary DPLL. Because the CCDPLL
supports reference redundancy, however, it requires a second
reference monitor for CCR1. Thus, the CCDPLL implements a
duplicate reference monitor for CCR1. The CCR1 reference
monitor includes a validation delay feature (see the Common
Clock Reference Switchover section).
When the user enables the CCDPLL lock detector, a stability
timer becomes available for use. The stability timer allows the
user to specify a fixed delay for indicating a locked status after
the phase lock detector makes a transition from an unlocked to
a locked state. The time delay provides for a more robust
indication of lock, because it allows additional time for lock and
unlock chatter to resolve. The stability timer uses an unsigned
16-bit value (units of milliseconds) via Register 0x0D04 to
Register 0x0D05. Programming a nonzero value enables the
stability timer. Otherwise, the CCDPLL lock detector bypasses
the stability timer.
Each CCR reference monitor continuously observes the post
R divider period (tCCR) of each CCR source, where tCCR = R × tREF
(see the Reference Monitor Controls section of the Reference
Monitor section for details on tREF). The reference monitor
considers a CCR valid when the period of the specified source
of the CCR is within 3.125% of tCCR. Conversely, a CCR is
invalid when the CCR reference monitor fails to detect the next
rising edge event within two tCCR periods, or the user clears the
enable bit associated with the CCR.
The user has access to the status of the CCDPLL lock detector
via Bit 1 of Register 0x0D40 or Bit 1 of Register 0x3002. Logic 1
indicates locked, and Logic 0 indicates unlocked.
Each CCR reference monitor provides a status bit to indicate
that the CCR is valid (Logic 1), via Bit 4 and Bit 5 of
CCDPLL LOOP FILTER
The CCDPLL loop filter and the auxiliary DPLL loop filter use
the same loop filter hardware. As such, the user can program
the bandwidth of the loop filter as described in the Auxiliary
DPLL Loop Bandwidth section of the System Clock
Compensation section.
Register 0x0D40, for CCR0 and CCR1, respectively. The status
of the CCR reference monitors (valid and invalid) are also part
of the IRQ mechanism (see Bits[7:4] of Register 0x301D and
the Interrupt Request (IRQ) section). The status bits (Bit 4 and
Bit 5 of Register 0x0D40) are undefined for legacy auxiliary
DPLL operation. Furthermore, when using the common clock
DPLL, the legacy auxiliary DPLL reference status bit (Bit 2 of
Register 0x3002) is undefined.
COMMON CLOCK REFERENCE (CCR) PERIOD
DECLARATION
Because a CCR is a time stamp source, the user programs the
expected reference period, tREF, in the reference monitor
controls for a specific source (see the Reference Monitor
Controls section of the Reference Monitor section for details on
The CCDPLL uses the status of the CCR reference monitors to
coordinate automatic reference switchover as described in the
Common Clock Reference Switchover section.
tREF). However, tREF is subject to rounding errors, which are
CCDPLL LOCK DETECTOR
typically small enough to ignore.
By default, the CCDPLL uses the same lock detector as the
legacy auxiliary DPLL (see the Auxiliary DPLL Reference
Monitor Status section of the System Clock Compensation
section). However, the auxiliary DPLL lock detector has fixed
parameters that are not adjustable by the user. Thus, the default
lock detector is not well suited to digitized clocking applications.
To resolve this issue, the CCDPLL has access to a specialized
lock detector with programmable features, the CCDPLL lock
detector. The general functionality of the CCDPLL lock
detector is identical to the phase lock detector employed in the
channel DPLLs (DPLL0 and DPLL1). Refer to the DPLL Phase
Lock Detector section of the Digital PLL (DPLL) section for
details.
Although tREF is sufficiently precise for use by the reference
monitors, it is generally not precise enough for use by the
CCDPLL. The reason is the CCDPLL generates a local time
scale based on integrating tREF. Integrating tREF, however, poses a
problem because even a minuscule error accumulates over
time. To remedy this problem, the CCDPLL provides a
mechanism for declaring the reference period rationally as a
numerator (NUM) and denominator (DEN).
For example, consider a 38.88 MHz reference frequency. The
period, in rational terms, is 1/38,880,000. For this case, the
numerator = 1 and the denominator = 38,880,000. The user
programs the numerator and denominator values in
Register 0x0D12 to Register 0x0D1A for CCR0 and
Register 0x0D22 to Register 0x0D2A for CCR1.
The phase lock threshold, phase lock fill, and phase lock drain
parameters of the CCDPLL lock detector appear in different
registers than those associated with DPLL0 and DPLL1:
Register 0x0D00 to Register 0x0D01, Register 0x0D02, and
Register 0x0D03 for threshold, fill, and drain, respectively.
Furthermore, the common clock DPLL uses the legacy auxiliary
DPLL lock detector by default. To enable the CCDPLL lock
detector, the user must program a nonzero value for the phase
The rational period functionality is inactive by default (that is,
denominator = 0), which means the common clock DPLL uses
tREF strictly as the period definition by default. As such, to make
use of the rational period functionality, the user must program
the denominator with a nonzero value.
Rev. 0 | Page 57 of 205
AD9546
Data Sheet
solves this potential problem by adding a validation delay to the
validation of CCR1. The delay ensures nonrevertive priority to
CCR0, but the delay is only in effect when CCR0 is invalid (the
delay is unnecessary otherwise).
COMMON CLOCK REFERENCE SWITCHOVER
The CCDPLL uses an integrated controller to handle the
activation and deactivation of the loop based on the status of
the CCRs. Because the CCDPLL supports redundant CCRs, the
controller requires a priority scheme to handle transitions from
one CCR to another.
The user can determine which reference is currently active via
Bit 3 of Register 0x0D40. Logic 0 or Logic 1 indicates CCR0 or
CCR1 is active, respectively. Bit 3 is undefined for legacy
auxiliary DPLL operation.
The controller gives nonrevertive priority to CCR0, meaning
that after the controller activates the loop using a specific CCR,
the controller continues using that CCR until the CCR
ACTIVE STATUS
invalidates. If both CCRs are operational at power-up, however,
there is the risk that CCR1 may validate before CCR0, causing
CCR1 to be the active reference, which is contrary to the
desired nonrevertive priority assigned to CCR0. The controller
The CCDPLL provides a status bit that indicates the CCDPLL
is actively tracking a reference input signal. To check the active
status of the CCDPLL, use Bit 0 of Register 0x0D40, where
Logic 1 indicates the common clock DPLL is active.
Rev. 0 | Page 58 of 205
Data Sheet
AD9546
COMMON CLOCK SYNCHRONIZER (CCS)
OVERVIEW
FROM
IUTS
CSS SYNC 0 RESTART
CSS SYNC 1 RESTART
0
1
ACTIVE SOURCE INDEX (0,1)
RESTART
FROM THE COMMON
CLOCK DPLL
TIME OF DAY
SOURCE
SYNC
TIME
96
CSS SYNC 0 TIME CODE
CSS SYNC 1 TIME CODE
0
1
0
1
0
1
REG 0x0D11, BITS[4:0]
REG 0x0D21, BITS[4:0]
REG 0x0D11, BIT 7 REG 0x0D1D TO 0x0D1B, BITS[23:0]
REG 0x0D21, BIT 7 REG 0x0D2D TO 0x0D2B, BITS[23:0]
0
1
SKEW
TIME
96
TIME OF DAY
SOURCE
TAG
TIME
STAMPS
TIME STAMP
TAG
DETECT
SYSTEM
CLOCK
SYNC
REG 0x0D3B TO 0x0D39, BITS[19:0]
GUARD
EVENT
STABILITY
STATUS
DETECT
INITIAL
SYNC
ADJUSTMENT
MAXIMUM
MAGNITUDE
DETECT
VALID
SYNC
INITIAL SYNC
DETECT
SYNC
EVENT
GUARD
BYPASS
LOCK
REG 0x0D3C, BIT 0
LATENCY
DETECT
NORMALLY
CLOSED
GUARD
STATUS
COMMON CLOCK
DPLL LOCK STATUS
GUARD
LATENCY
SYNC OFFSET VALUES
UPDATE ON EACH
VALID SYNC EVENT
REG 0x0D38 TO 0x0D37, BITS[15:0]
SYNC OFFSET
REFINEMENT
SYNC OFFSET
REFINEMENT
RESTART
SLEW
LIMITER
SLEW
LIMITER
STATUS
SLEW LIMIT
REG 0x0D36 TO 0D34, BITS[23:0]
LOCAL
TIME SCALE
COMMON
TIME SCALE
OFFSET
REG 0x0D33 TO 0x0D30, BITS[31:0]
SUMMING
NODE
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
Figure 51. Common Clock Synchronizer
The CCDPLL phase locks to a physical common clock source
(the common clock reference) and generates an internal local
time scale. The local time scale constitutes time as the
accumulation of the period of the system clock as defined by
the system clock input frequency programmed by the user. The
CCDPLL relies on a crystal resonator (via the XOA and XOB
pins) for internal timing, which is a relatively unstable
frequency source. However, the architecture of the CCDPLL
allows it to phase lock to the common clock reference even with
the relative instability of the crystal resonator. Thus, the local
time scale is immune to instability associated with the
frequency source of the system clock.
Because the CCDPLL derives its timing from the system clock
but the phase locks to the common clock reference, instabilities
associated with either source contribute to the loop stability of
the CCDPLL. Thus, the user must tailor the loop bandwidth of
the CCDPLL in accordance with the combined stability of the
system clock and common clock reference frequencies. The
frequency stability of the sources puts a limit on the minimum
allowable loop bandwidth that yields a stable loop and
minimizes the phase offset associated with the ability of the PLL
to track frequency variations.
Rev. 0 | Page 59 of 205
AD9546
Data Sheet
Although the local time scale is precise (frequency stable), it is
likely not accurate. That is, the time indicated by the local time
scale is with respect to an arbitrary starting time. Stated another
way, the epoch of the local time scale is arbitrary. The purpose
of the CCS is to provide a means to produce an internal
common time scale with a precise epoch determined by the
user. In practice, the user provides a physical trigger event (the
synchronization arrow in Figure 51) along with a numeric
entry of the time associated with the trigger event (the time
arrow in Figure 51), which effectively assigns an epoch. The
CCS uses the trigger event and numeric entry to produce a
precise time shifted version of the local time scale (the common
time scale) synchronized to the desired epoch. More
importantly, in the context of multiple AD9546 devices in a
digitized clocking system, the user can use the CCS of each
device to obtain precise alignment of the common time scale
across devices.
The local time scale appears as one of the synchronization
sources associated with the synchronization input in Figure 51.
The user can employ the local time scale for immediate
synchronization rather than using a time stamp source. Immediate
synchronization is useful for forcing a synchronization trigger
in applications where the user is not transporting the common
time scale across multiple devices or does not require the
common time scale to have a specific epoch, but wants to use
one (or more) of the digitized clocking resources (a UTS, for
example). In such cases, the immediate trigger is necessary to
make the CCS ready for use by the digitized clocking resources.
To select immediate synchronization, use 0x1E or 0x1F for
Bits[4:0] associated with the primary or secondary
synchronization source selection.
When the common clock DPLL experiences a reference
switchover, there is a direct impact on the CCS. First, the
synchronization source changes to primary or secondary in
keeping with the new reference source (primary or secondary).
Likewise, the synchronization time (see the Synchronization
Time section) and skew (see the Adding Time Skew to the
Synchronization Time section) change in keeping with the new
reference source (primary or secondary). In addition, the CCS
flushes any synchronization data previously captured.
However, the synchronization offset presently applied by the
CCS to the common time scale remains in effect. Likewise, the
filtering state of the synchronization offset refinement block
(see the Synchronization Offset Refinement section) and the
condition of the slew limiter (see the Synchronization Slew
Limiter section) also remain in effect.
The diagram in Figure 51 is a functional representation of the
CCS. The key points of interest are the time and synchronization
inputs, synchronization offset refinement, the slew limiter, and
the summing node with the common time scale (a time shifted
version of the local time scale) as its output.
SYNCHRONIZATION SOURCE
The synchronization source constitutes the time stamp source
associated with the synchronization input of Figure 51. The
arrival of each time stamp from the selected synchronization
source constitutes a synchronization event. In operation, when
a synchronization event occurs, the user updates the synchroniza-
tion time (see the Synchronization Time section) associated
with that synchronization event, but before the occurrence of
the next synchronization event.
Because the CCS flushes old synchronization data on a
reference switchover, the CCS has no synchronization event to
which it can compare the first synchronization time update.
This condition causes a synchronization error and prevents the
CCS from performing the intended synchronization. However,
the next complete synchronization event clears the
synchronization error with synchronization proceeding
normally from that point onward.
Because the common clock DPLL has access to a primary and
secondary (optional) reference source for producing the local
time scale, the CCS also has access to a primary and secondary
(optional) synchronization source. Whenever the common
clock DPLL switches between primary and secondary
references, the CCS automatically switches to its primary or
secondary synchronization source, commensurate with the
common clock DPLL (a one to one correspondence).
TAGGED SYNCHRONIZATION
Tagged synchronization only applies when the synchronization
source originates from a TDC generating tagged time stamps.
Thus, tagged synchronization does not apply to immediate
synchronization (that is, using the common time scale as the
synchronization source).
To assign the primary synchronization source, use Bits[4:0] of
Register 0x0D11. To assign the secondary synchronization
source, use Bits[4:0] of Register 0x0D21. In legacy auxiliary
DPLL operation, the CCS uses the source specified in Bits[4:0]
of Register 0x0D11.
For a source with tagged time stamps (see the Tagged Time
Stamps section), the user has the option of synchronizing to
only the tagged time stamps from that source. To enable the
tagged synchronization feature, program a Logic 1 in Bit 7 of
Register 0x0D11 for the primary synchronization source and/or
Register 0x0D21 for the secondary synchronization source.
If possible, the recommendation is to use REFB and/or REFBB
as the synchronization source to the CCS because REFB and
REFBB provide access to the analog loopback function (see the
Analog Clock Loopback section).
Rev. 0 | Page 60 of 205
Data Sheet
AD9546
For example, consider a synchronization time value of
148,264.05394857 sec. The integer seconds portion is 148,264,
which, in 48-bit hexadecimal format, is 0x 0000 0002 4328.
SYNCHRONIZATION TIME
The user programs the time associated with a forthcoming
trigger event via a 96-bit time value written to Register 0x0F0A
to Register 0x0F15. However, the user must also program
Register 0x0F09, Bits[4:0] = 0x30 or 0x31 to associate the time
value with the primary or secondary CCS source, respectively
(see Figure 54 and Table 45 in the Inverse User Time Stamper
(IUTS) section). Then, upon issuing an IO update (via
Register 0x000F, Bit 0), the CCS considers the programmed
value as the time associated with the previous synchronization
event (via the synchronization path in Figure 51).
In fractional seconds format, the fractional portion of the
seconds entry (to the right of the virtual decimal point)
constitutes a 48-bit number with an LSB weight of 2−48 sec.
Using the same synchronization time value as in the previous
example, the fractional seconds portion is 0.05394857, which
converts to units of 2−48 sec as follows:
0.05394857 × 248 = 15,185,172,484,323.19496192
Round the result to the nearest integer, which yields
15,185,172,484,323, and convert to 48-bit hexadecimal format:
0x 0DCF 92CF D0E3.
When the device initializes, the CCS has no synchronization
event to which it can compare the first synchronization time
update. Thus, the first synchronization event following
initialization causes a synchronization error (see the
Synchronization Guard and Ready Status sections). However, a
subsequent synchronization event clears the synchronization
error and synchronizes the CCS accordingly.
In PTP format, the fractional portion of the seconds entry (to
the right of the virtual decimal point) constitutes a 46-bit
number (thus, the exclusion of the two fractional nanoseconds
MSBs in the bottom half of Figure 52) with an LSB weight of
2−16 ns. Using the same fractional portion value (0.05394857),
convert to units of 2−16 ns as follows:
There are two options for the time format: fractional seconds
and fractional nanoseconds (PTP format). Figure 52 shows the
register arrangement for both formats. Regardless of the
format, the register value for the integer portion of the seconds
is the same.
0.05394857 × 109 × 216 = 3,535,573,483,520
This value in 46-bit hexadecimal format (rounded to the
nearest integer, if necessary) is 0x 0337 309A 0000.
FRACTIONAL SECONDS FORMAT
REGISTER
0x0F15
REGISTER
0x0F14
REGISTER
0x0F13
REGISTER
0x0F12
REGISTER
0x0F11
REGISTER
0x0F10
REGISTER
0x0F0F
REGISTER
0x0F0E
REGISTER
0x0F0D
REGISTER
0x0F0C
REGISTER
0x0F0B
REGISTER
0x0F0A
MSB
LSB
SECONDS
FRACTIONAL SECONDS
VIRTUAL
DECIMAL POINT
FRACTIONAL NANOSECONDS FORMAT (PTP)
REGISTER
0x0F15
REGISTER
0x0F14
REGISTER
0x0F13
REGISTER
0x0F12
REGISTER
0x0F11
REGISTER
0x0F10
REGISTER
0x0F0F
REGISTER
0x0F0E
REGISTER
0x0F0D
REGISTER
0x0F0C
REGISTER
0x0F0B
REGISTER
0x0F0A
MSB
LSB
MSB
LSB
BITS[7:6] NOT USED
FRACTIONAL NANOSECONDS
SECONDS
VIRTUAL
DECIMAL POINT
Figure 52. Time Formats
Rev. 0 | Page 61 of 205
AD9546
Data Sheet
The user can restart the synchronization offset refinement at
any time (see the Restart section).
ADDING TIME SKEW TO THE SYNCHRONIZATION
TIME
In a digitized clocking system with redundant common clock
references, a time skew specific to each reference path may
exist. The user can compensate the skew by programming the
appropriate time skew for each path.
MAXIMUM MAGNITUDE DETECTION
The CCS provides the user with the option to prevent excessive
synchronization offset values from propagating into the
synchronization offset refinement process. The user activates
the maximum magnitude detection block by programming a
nonzero guard adjustment value in Register 0x0D39 to Register
0x0D3B. A zero value (default) disables the maximum
magnitude detection feature.
The skew programs as a 24-bit signed value. Primary skew is via
Register 0x0D1B to Register 0x0D1D. Secondary skew is via
Register 0x0D2B to Register 0x0D2D.
The 24-bit programmed time skew is in units of 2−48 sec,
yielding a range of approximately 29.8 ns. For example, to
program a time skew of −5 ns (−5 × 10−9 sec), first convert to
the appropriate units and round the result to the nearest integer
as follows:
When activated, the maximum magnitude detector monitors
each new synchronization offset value and compares it to the
guard adjustment value. If a synchronization offset value
exceeds the guard adjustment value, the maximum magnitude
detector signals a guard event (see the Synchronization Guard
section), thereby preventing the offending synchronization
offset value from propagating into the synchronization offset
refinement process.
−5 × 10−9 × 248 = −1,407,375
This resulting value in 24-bit hexadecimal format is: 0x EA 8671.
ADDING TIME OFFSET TO THE
SYNCHRONIZATION TIME
The maximum magnitude detector ignores the first
synchronization event following a restart or after the initial
synchronization block indicates initial synchronization (see the
Initial Synchronization section).
In a digitized clocking system with redundant common clock
references, a time skew common to both reference paths may
exist. The user can compensate for the common time skew by
programming the appropriate time offset value.
The guard adjustment value comprises a 20-bit unsigned
number in units of 2−40 sec (~0.91 ps), yielding a maximum
guard value of up to 954 ns (approximately). For example, to
set the maximum magnitude detection value to 10 ns, first
convert to the appropriate units and round the result to the
nearest integer as follows:
The time offset programs as a 32-bit signed value in
Register 0x0D30 to Register 0x0D33 in units of 2−48 sec, yielding
a range of approximately 15.26 μs. For example, to program a
time skew of 50 ns (50 × 10−9 sec), first convert to the appropriate
units and round the result to the nearest integer as follows:
10 × 10−9 × 240 = 10,995
50 × 10−9 × 248 = 14,073,749
This value in 20-bit hexadecimal format is: 0x 0 2AF3.
LATENCY DETECTION
This value in 32-bit hexadecimal format is: 0x 00D6 BF95.
SYNCHRONIZATION OFFSET REFINEMENT
The normal synchronization process involves the occurrence of
a synchronization event followed by the application of the
synchronization time associated with the synchronization
event. When a synchronization event occurs, the common
clock synchronizer waits for the user to issue a synchronization
time value associated with the synchronization event. The
purpose of the latency detector is to allow the user to apply a
time limit on how long the CCS waits for a synchronization time
value before discarding the synchronization event. If a
synchronization time value does not arrive within the specified
period, the latency detector signals a guard event (see the
Synchronization Guard section).
In a digitized clocking application, the common clock
synchronization process is typically iterative (though not a
strict requirement). That is, the user applies a continuous
synchronization source and continuously writes
synchronization time values for each expected synchronization
event. As such, each synchronization and time pair results in a
specific synchronization offset value. However, a series of
synchronization offset values is likely to introduce jitter onto
the common time scale.
As such, the CCS employs a synchronization offset refinement
block that gradually softens the impact of each new
synchronization offset value. The goal is to yield an average
synchronization offset over time. That is, the synchronization
offset value produced by the first synchronization event
(following a restart condition) propagates through the
refinement block unaltered. As synchronization events
continue to occur, the synchronization offset refinement block
applies progressively more filtering (up to a limit) to
subsequent synchronization offset values.
The user activates latency detection by writing a nonzero value to
the 16-bit unsigned guard latency bit field in Register 0x0D37
to Register 0x0D38 in units of 2−16 sec, yielding a maximum
latency guard time of slightly less than 1 sec. A zero value
(default) disables the latency guard feature.
Rev. 0 | Page 62 of 205
Data Sheet
AD9546
For example, to set latency detection value to 1 ms (10−3 sec),
first convert to the appropriate units and round the result to the
nearest integer as follows:
The status of the slew limiter is available via several mechanisms.
Bit 6 of Register 0x0D40 is Logic 1 when the slew limiter is
actively slewing and Logic 0 when it is not slewing. The status
of the slew limiter is also available via an appropriately
configured Mx status pin (see the Status and Control Pins
section). In addition, slew limiter status is available as part of
the IRQ mechanism (see the Interrupt Request (IRQ) section)
via Bit 2 and Bit 3 of Register 0x301D, which indicate when the
slew limiter starts and stops slewing, respectively.
10−3 × 216 = 66
This value in 16-bit hexadecimal format is: 0x 0042.
SYNCHRONIZATION SLEW LIMITER
A typical digitized clocking system sometimes experiences the
injection of a relatively large phase adjustment as part of the
common clock synchronization process. An example is the
injection of a phase jump due to switching between two CCRs
having a static phase offset (see the Common Clock Reference
Switchover section). Another example is a phase jump due to
the injection of a phase adjustment to compensate for round
trip delay in a clock signal path (see the Analog Clock
Loopback section). Phase adjustments constitute instantaneous
phase jumps, which equate to frequency impulses. The
frequency impulses transfer to the common time scale via the
action of the CCS.
RESTART
When a synchronization restart occurs, the next generated
synchronization offset value propagates through the refinement
block unaltered, with subsequent synchronization offset values
receiving progressively more filtering over time. The user can
manually restart the synchronization offset refinement process
from the beginning at any time by programming Bit 0 of Register
0x0F08 to Logic 1.
The synchronization event that follows a synchronization
restart is not the same as an initial synchronization event (a
synchronization event following a device power-up, for example).
Thus, a synchronization restart does not cause the slew limiter
to bypass automatically on the first synchronization event
following a synchronization restart.
To mitigate the frequency impulses resulting from phase
adjustments, the CCS provides a slew limiter following the
synchronization offset refinement block (see Figure 51). The
slew limiter converts large phase jumps (frequency impulses) to
a constant phase slope (which equates to a constant frequency
offset). Thus, the action of the slew limiter puts a user defined
upper bound on the frequency deviation associated with a
phase adjustment. Because the slew limiter is physically situated
at the output of the CCS (see Figure 51), it effectively limits the
magnitude of the frequency offset injected on the common time
scale as the result of a phase adjustment.
SYNCHRONIZATION GUARD
Because common clock synchronization has a direct impact on
the common time scale, safeguards are in place to mitigate
corruption of the common time scale with invalid synchroniza-
tion events. Such is the function of the guard element shown in
Figure 51. Conceptually, the guard controls a normally closed
switch that opens when the guard detects a guard event, thereby
preventing the most recent synchronization offset value from
propagating into the synchronization offset refinement process.
The user activates the slew limiter by writing a nonzero value to
the 24-bit unsigned slew limit value in Register 0x0D34 to
Register 0x0D36 in units of 2−36 sec/sec. For example, to
constrain the frequency deviation on the common time scale to
a maximum of 0.15 ppm (1.5 × 10−7), first convert to the
appropriate units and round the result to the nearest integer as
follows:
Assuming the initial synchronization has occurred (see the
Initial Synchronization section), the following three conditions
can trigger the synchronization guard to open the guard switch:
•
The common clock DPLL unlocks (assuming the guard
bypass lock bit is not set)
1.5 × 10−7 × 236 = 10,308
•
•
A maximum magnitude detection event
A latency detection event
This value in 24-bit hexadecimal format is: 0x 00 2844.
When the user programs a new slew limit value, the slew
limiter stalls briefly while implementing the slew limit value
change. From an application perspective, the stalling behavior
is inconsequential.
Any one of these conditions trips the synchronization guard,
which causes a synchronization error. To check the
synchronization error status, use Bit 7 of Register 0x0D40,
where Logic 1 indicates a synchronization error. Synchronization
error status is also available via an appropriately configured Mx
status pin (see the Status and Control Pins section) and via Bit 1
of Register 0x301D as part of the IRQ mechanism (see the
Interrupt Request (IRQ) section).
The user can bypass the slew limiter at any time by
programming a slew limit value of zero. The slew limiter
automatically bypasses for the initial application of a
synchronization offset value. However, subsequent
synchronization offset values are subject to slew limiting
(assuming the user programs a nonzero slew limit value).
Rev. 0 | Page 63 of 205
AD9546
Data Sheet
synchronization occurs, the initial synchronization detect block
holds the maximum magnitude detection block, the
Guard Bypass Lock
The user has the option of preventing an unlock condition of
the common clock DPLL from triggering the synchronization
guard via the guard bypass lock bit, Bit 0 of Register 0x0D3C.
Logic 0 (default) allows an unlock condition to trigger the
synchronization guard. Logic 1 prevents an unlock condition
from triggering the synchronization guard.
synchronization offset refinement block, and the slew limiter in
a reset state. The initial synchronization detect block does not
indicate initial synchronization until the system clock PLL is
stable and a synchronization event occurs that does not trigger a
guard event (see the Synchronization Guard section).
READY STATUS
Setting the guard bypass lock bit is useful when using the
digitized clock resources of the AD9546 in a nondigitized clock
application, for example, using two UTS resources to measure
the phase offset between two clock signals.
The CCS provides a status indicating that the common time
scale is ready to support digitized clocking resources. To check
the ready status of the CCS, use Bit 2 of Register 0x0D40, where
Logic 1 indicates the CCS is ready.
INITIAL SYNCHRONIZATION
Ready status is also available via Bit 0 of Register 0x301D,
which is part of the IRQ mechanism (see the Interrupt Request
(IRQ) section) and through an appropriately configured Mx
status pin (see the Status and Control Pins section).
The initial synchronization detect block ensures the CCS starts
in a known state. The CCS is not operational until the
synchronization detect block indicates initial synchronization
(the falling edge shown in Figure 51). Until initial
Rev. 0 | Page 64 of 205
Data Sheet
AD9546
USER TIME STAMPER (UTS)
The UTS system is independent of the two user time stamp
processors (see the User Time Stamp Processor (UTSP)
section).
OVERVIEW
The AD9546 has nine UTSs comprising the UTS system (see
Figure 53). Each UTS in the UTS system converts internal
device time stamps (based on the local time scale) to time codes
based on the common time scale (see the Common Clock
Synchronizer (CCS) section) and makes those time codes
accessible to the user via the register map.
UTS CHANNEL CONTROL REGISTERS ADDRESSES
Each UTS channel has a dedicated UTS channel control
register, differentiated by an integer, x (where x is 0 to 8). Each
UTS Channel x control register occupies a range of register
addresses. UTS 0 occupies Register 0x0E00 to Register 0x0E04.
More generally, UTS x occupies the UTS 0 address range, but
with an offset of x × 0x0005.
Because the UTS system relies on the common time scale
(which originates with the CCS) the status of the CCS matters
to the UTS system. As such, the UTS system is inoperative
unless the CCS indicates a ready status.
UTS CHANNEL ENABLE
The UTS system consists of nine independent UTSs with each
UTS having a dedicated UTS channel control register. Time
codes output by each UTS route to a readback FIFO. The FIFO
contains time code samples originating from all active
(enabled) UTSs. The FIFO samples include supplemental
information allowing the user to discern the source of a time
code sample and the status of the UTS channel associated with
the time code sample. The FIFO also provides a status related
to its own condition.
Each UTS in the UTS system has an independent enable bit.
The enable bit is Bit 0 in Register 0x0E00 for UTS 0, or more
generally, for UTS x (where x = 0 to 8, integer), add x × 0x0005
to the register address given for UTS 0. Logic 1 enables the
corresponding UTS channel. Logic 0 (default) disables the
corresponding UTS, in which case it stops delivering new time
code samples to the FIFO with no effect on old time code
samples already delivered to the FIFO.
CCS READY
COMMON TIME
SCALE
UTS 0
TIME
TIME STAMPS
96
CODE
SOURCE
FLAGS
REG 0x0E3A TO 0x0E2F, BITS[95:0]
REG 0x0E2E, BITS[4:0]
REG 0x0E2E, BITS[7:5]
REG 0x0E2D, BITS[6:0]
REG 0x0E2D, BIT 7
SOURCE
TIME
5
SOURCE
5
3
REG 0x0E01, BITS[4:0]
SELECTOR
CODE
96
TAG
FORMAT
OFFSET
5
READ
BACK
FIFO
REG 0x0E01, BIT 5
REG 0x0E00, BIT 1
TIME STAMP
PROCESSOR
COUNT
7
UTS
CHANNEL 0
CONTROL
ENCODED
FLAGS
24
3
REG 0x0E04 TO 0x0E02, BITS[23:0]
OVERFILL
IRQ
ENABLE
FLAG
ENCODER
REG 0x301C, BITS[7:5]
REG 0x0E00, BIT 0
3
FLAGS
REG 0x0E00, BITS[4:2]
CCDPLL
UNLOCK
IO UPDATE
CCR INVALID
CCR FAULT
UTS 1
35
UTS CHANNEL 1
CONTROL
UTS 8
35
UTS CHANNEL
CONTROL
8
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
2. ADDRESSES SHOWN FOR UTS CHANNEL CONTROL ARE SPECIFIC TO UTS 0
Figure 53. UTS System
Rev. 0 | Page 65 of 205
AD9546
Data Sheet
UTS CHANNEL TIME STAMP SOURCE
UTS CHANNEL TAG
Each UTS channel has independent access to the various time
stamp sources within the AD9546. The user assigns a specific
time stamp source via Bits[4:0] of Register 0x0E01, or more
generally, for UTS x (where x = 0 to 8, integer), add x × 0x0005
to the register addresses given for UTS 0.
The user can have a UTS channel selectively process only
tagged time stamps (see the Tagged Time Stamps section) from
the specified time stamp source (rather than every time stamp
from that source). Tagged time stamp selection is via Bit 5 in
Register 0x0E01 for UTS 0, or more generally, for UTS x (where
x = 0 to 8, integer), add x × 0x0005 to the register addresses
given for UTS 0. Logic 1 selects tagged time stamp processing,
whereas Logic 0 (default) causes the UTS channel to process all
time stamps from the specified time stamp source.
The decimal value of the five source bits relates to the time
stamp source per Table 42.
Table 42. Time Stamp Source Assignment Codes
Source Bit Field Value (Decimal)
Time Stamp Source
REFA (default)
REFAA
REFB
REFBB
DPLL0 feedback
DPLL1 feedback
Auxiliary REF0
Auxiliary REF1
Auxiliary NCO 0
Auxiliary NCO 1
Reserved
Auxiliary REF2
Auxiliary REF3
IUTS 0
UTS CHANNEL TIME FORMAT
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
The user can select the output time format of each UTS channel
independently via a format bit. Format selection is via Bit 1 in
Register 0x0E00 for UTS 0, or more generally, for UTS x (where
x = 0 to 8, integer), add x × 0x0005 to the register address given
for UTS 0. Logic 0 (default) selects fractional seconds format,
and Logic 1 selects PTP format.
See the Synchronization Time subsection of the Common
Clock Synchronizer (CCS) section for details on fractional
seconds format and PTP format.
UTS CHANNEL STATUS FLAG ASSIGNMENT
Whenever a UTS channel converts an input time stamp to a
time code and delivers it to the FIFO, the UTS channel also
includes a 3-bit status value as part of the time code.
IUTS 1
The 3-bit status value is an encoding of the following five
available status conditions:
15 to 31
Reserved
The recommendation is to use the same time stamp source
while a UTS channel remains enabled. However, the user can
enable a UTS channel and assign a time stamp source
simultaneously with no negative implications.
•
•
•
•
Invalid (common clock reference invalid)
Fault (common clock reference fault)
Unlocked (common clock DPLL unlocked)
Tag (state of the tag bit associated with the UTS channel
that processed the time stamp)
UTS CHANNEL TIME OFFSET
The user has the option of having a UTS channel apply a time
offset to time stamps processed by the channel, with the time
offset being relative to the common time scale in units of
2−48 sec. Time offset is via Bits[23:0] (signed) in Register 0x0E02
to Register 0x0E04 for UTS 0, or more generally, for UTS x
(where x = 0 to 8, integer), add x × 0x0005 to the register
addresses given for UTS 0.
For example, to program a time offset of −25 ns (−25 × 10−9 sec),
first convert to the appropriate units and round the result to the
nearest integer as follows:
•
Format (state of the format bit associated with the UTS
channel that processed the time stamp)
The user can choose one of seven predefined 3-bit encoded
status patterns (see Table 43). The selection is via the 3-bit flags
bit field in Bits[4:2] of Register 0x0E00 for UTS 0, or more
generally, for UTS x (where x = 0 to 8, integer), add x × 0x0005
to the register address given for UTS 0.
Each bit of the flags bit field encodes to a unique status bit (or
logical combination of status bits). That is, each bit of the 3-bit
flags bit field encodes to a specific status bit pattern based on
the value of the 3-bit flags bit field.
−25 × 10−9 × 248 = −7,036,874
This value in 24-bit hexadecimal format is: 0x 94 A036.
Rev. 0 | Page 66 of 205
Data Sheet
AD9546
Table 43. UTS Channel Flag Assignments
Encoded Status
Flags Bit Field,
Bits[4:2]
Bit 4 Status
Encode
Bit 3 Status Bit 2 Status
A 3-bit encoded status flags bit field (Bits[7:5] of
Encode
Invalid
Fault
Invalid
Fault
Encode
Register 0x0E2E) accompanies each time code. The meaning of
each bit depends on the flags bit field of the UTS channel
associated with the time code sample in accordance with Table 44.
000 (Default)
Tag
Tag
Format
Format
Fault
Unlocked
Unlocked
Unlocked
Unlocked
Unlocked
001
010
011
100
101
To determine the meaning of each of the three bits, the user
must be aware of which UTS channel provided the time code
(see the Source Code section) and the state of the 3-bit flags bit
field associated with that UTS channel (see the UTS Channel
Status Flag Assignment section).
Invalid
Tag
Format
Unlocked or
invalid
110
111
Format
Unused
Tag
Unlocked or
fault
Table 44. UTS FIFO Readback Encoded Status Bits
Unused
Unused
Bit 7
Bit 6
Bit 5
Encoded
Status
Encoded
Status
Encoded
Status
UTS READBACK FIFO
Bits[4:2]1
000
001
010
011
100
101
The UTS system uses a FIFO to collect time code samples (up
to 18) from all enabled UTS channels. The FIFO allows queuing
of time codes arriving from the various UTS channels for
readback by the user via the register map. Each FIFO sample
consists of the following:
Tag
Tag
Format
Format
Fault
Invalid
Fault
Invalid
Fault
Invalid
Tag
Unlocked
Unlocked
Unlocked
Unlocked
Unlocked
Unlocked
or invalid
Unlocked
or fault
•
•
•
96-bit time code
3-bit encoded status code
5-bit source code
Format
110
111
Format
Unused
Tag
Time Code
Unused
Unused
The 96-bit time code appears in Register 0x0E2F to
Register 0x0E3A, which adheres to the same functionality as
Register 0x0F0A to Register 0x0F15 described in the
Synchronization Time subsection of the Common Clock
Synchronizer (CCS) section. Note that interpretation of the
time code requires knowledge of the underlying format
(fractional seconds or PTP).
1 Bits[4:2] in the UTS Channel x control register associated with the time code.
UTS FIFO Status
Because the FIFO effectively works in the background
collecting UTS time codes, the FIFO provides the user with a
continuous status of the current state. The status information is
essential to keep the user aware of how frequently to read FIFO
samples or if sample data is lost due to an overflow condition.
There are two categories of status information: FIFO overfill (a
single status bit) and FIFO count (7-bit field).
There are two ways to discern the format of a time code sample.
One way is to use the source code information (see the Source
Code section) to determine which UTS is the source of the time
code sample and then read the state of the format bit in the
appropriate UTS Channel x control register (where x is 0 to 8,
integer). Another way is to use one of the seven predefined
encoded status patterns that includes the format status in Table 43.
In this method, the encoded status code (see the Encoded
Status section) contains the format information for the time
code sample.
FIFO Overfill
The FIFO overfill bit in Bit 7 of Register 0x0E2D indicates that
the FIFO has exceeded the capacity, where Logic 1 indicates
overfill. When the FIFO asserts the overfill bit, any time codes
arriving at the FIFO input are lost.
The state of the FIFO overfill bit appears as part of the IRQ
mechanism (see the Interrupt Request (IRQ) section) via Bit 7
of Register 0x301C. The overfill status of the FIFO is also
available via an appropriately configured Mx status pin (see the
Status and Control Pins section).
Source Code
A 5-bit source code bit field accompanies each time code. The
source code appears in Bits[4:0] of Register 0x0E2E. The source
code value for a given time code sample indicates the time
stamp source shown in Table 42. The source code that
accompanies a time code is identical to the source code
assignment of the UTS channel associated with the time code
(see the UTS Channel Time Stamp Source section).
The FIFO overfill bit and Mx pin status (if configured) clear
when the user reads a sample from the FIFO.
The source code does not indicate the UTS channel associated
with the time code, but the original source of the time stamp.
Rev. 0 | Page 67 of 205
AD9546
Data Sheet
First, the user asserts an IO update (Bit 0 in Register 0x000F),
which latches the FIFO overfill and FIFO count status bits into
the register map. If the user reads the previous time code from
the register map prior to asserting an IO update, the FIFO shifts
the contents and the next available FIFO sample overwrites the
previous contents in the register map upon an IO update
assertion. The FIFO does not shift the contents unless the user
had previously read the time code from the register map prior
to the assertion of an IO update. Thus, the user is free to read
the current FIFO time code multiple times as long as the user
does not assert an IO update.
FIFO Count
The FIFO provides sample count information via the 7-bit
FIFO status count bit field in Bits[6:0] of Register 0x0E2D.
During normal operation (that is, overfill = 0), the FIFO status
count bit field value indicates the number of time code samples
currently in the FIFO. The user has access to the not empty
status of the FIFO (that is, when the value of Bits[6:0] of
Register 0x0E2D is greater than zero) via an appropriately
configured Mx status pin. However, when FIFO overfill = 1, the
FIFO status count bit field value indicates the number of lost
time code samples. That is, the meaning of the value in the
FIFO status count bit field changes based on the status of the
FIFO overfill bit.
Next, the user reads the FIFO status. If the FIFO count is zero,
the FIFO is empty and there is no need to continue. If the FIFO
count is nonzero and the FIFO overfill bit is Logic 0, the time
code value in the register map is ready for reading. However, if
the FIFO overfill bit is Logic 1, the FIFO is in an overflow
condition and time code samples have been lost. In an overflow
condition, the FIFO count value indicates how many samples
were lost. The FIFO count value has a limit of 127. Therefore,
there is no way to determine if more than 127 samples were
lost. Upon reaching an overflow condition, the FIFO does not
accept new time code samples. However, the contents remaining
in the FIFO are valid but may be too old to be useful.
A transition of the FIFO status count bit field from a value of
zero to a nonzero value is an indication that the FIFO is not
empty. That is, the FIFO was empty (FIFO status count = 0) but
now the FIFO has one (or more) time code samples (FIFO
status count ≠ 0) available for reading by the user. The zero to
nonzero transition appears as a status flag as part of the IRQ
mechanism (see the Interrupt Request (IRQ) section) via Bit 5
of Register 0x301C. The not empty status of the FIFO is also
available via an appropriately configured Mx status pin (see the
Status and Control Pins section).
Finally, assuming the FIFO status is not empty and not in an
overflow condition, the user may read the time code as well as
the source code and encoded status bits associated with the
current FIFO sample. The user can read the existing time code
and status information repeatedly, provided the IO update
remains unasserted. Assertion of an IO update overwrites the
register map contents but only if the user reads the time code in
the register map prior to asserting an IO update.
The Mx pin status indication of FIFO not empty remains
asserted until the FIFO becomes empty (FIFO status count = 0).
Reading Back FIFO Data
The FIFO captures time code samples (up to 18) and status
information associated with those time code samples from all
the enabled UTS channels. Because each FIFO sample consists
of a time code and status information, reading time code
samples from the FIFO requires a sequence of operations.
When the FIFO receives a new UTS sample, it flags an IRQ via
Bit 6 of Register 0x301C.
Rev. 0 | Page 68 of 205
Data Sheet
AD9546
INVERSE USER TIME STAMPER (IUTS)
code and asserting an IO update effectively constitutes a
digitized clock signal.
OVERVIEW
The AD9546 has two IUTSs: IUTS 0 and IUTS 1 (see Figure 54).
As the name implies, an IUTS performs the inverse function of
a UTS: accepting time codes based on the common time scale
provided by the user via the register map and converting those
time codes to internal device time stamps based on the local
time scale (see the Common Clock Synchronizer (CCS)
section). Because the IUTSs rely on the common time scale
(which originates with the CCS), the status of the CCS matters
for the proper function of an IUTS. As such, each IUTS is
inoperative unless the CCS indicates ready status.
The IUTS computes the period of successive time codes via the
period processor, where the current time code minus the
previous time code constitutes an instantaneous period sample.
Because period samples may exhibit jitter from sample to
sample, the period processor employs an averaging mechanism
to filter out period jitter from the underlying digitized clock
signal. The averaging mechanism allows the period processor to
provide a reliable estimate of the next time code in the
sequence, despite the presence of period jitter on the incoming
sequence of time codes.
The intended operation of an IUTS is to translate a digitized
clock signal (provided by the user in the form of a periodic
sequence of time codes written to the register map) to an
internal digitized clock signal for use by other resources within
the AD9546 (for example, as the input to one of the DPLL
channels). In a real-world, typical digitized clocking
application, the user reads time codes from a UTS and delivers
those time codes to an IUTS, which constitutes the transport of
a digitized clock signal (see Figure 49).
The IUTS does not produce an output time stamp from the
estimated time code until the time indicated by the common
time scale is greater than or equal to the value of the estimated
time code. That is, a causal gateway prevents estimated time
codes from appearing as output time stamps until they are
current (in terms of the common time scale).
The period processor of the IUTS includes a period monitor
that observes the period of each time code sample processed. If
a period sample deviates by more than about 1.5625%, the
IUTS that processed the sample indicates an invalid status (see
the IUTS Status section).
In operation, the user writes a time code to an IUTS and asserts
an IO update. The writing of a time code and subsequent
assertion of an IO update causes the IUTS to process the new
time code. Repeatedly carrying out the process of writing a time
REG 0x0F03 TO 0x0F01,
BITS[23:0]
REG 0x0F00, BIT 1
REG 0x0F00, BIT 0
BYPASS
LOCK
24 OFFSET
FORMAT
ADJUSTED
TIME CODE
ESTIMATED
REG 0x0F09, BIT 7
FORMATTED
TIME CODE
IUTS 0 TIME CODE
IUTS 1 TIME CODE
96
96
96
96
TIME CODE
TIME
PERIOD
96
96
STAMP
PROCESSOR
STATUS
REG 0x0F15 TO 0x0F0A,
CAUSAL
FORMATTER
BITS[95:0]
GATEWAY
REG 0x3023, BIT 0
REG 0x301E, BITS[1:0]
PERIOD
MONITOR
TIME CODE
DESTINATION
IUTS 0
RESTART
5
REG 0x0F09, BITS[4:0]
COMMON TIME SCALE
IUTS 0
RESTART
REG 0x0F07 TO 0x0F05,
BITS[23:0]
REG 0x0F08, BIT 0
REG 0x0F04, BIT 0
INVALIDATE
REG 0x0F04, BIT 1
TIME
STAMPS
BYPASS
LOCK
TO COMMON CLOCK
SYNCHRONIZER
24 OFFSET
IUTS 1
RESTART
CSS SYNC 1 RESTART
CSS SYNC 0 RESTART
ESTIMATED ADJUSTED
TIME CODE TIME CODE
PERIOD
TIME
CSS SYNC 1 TIME CODE
CSS SYNC 0 TIME CODE
PROCESSOR
STAMP
STATUS
CAUSAL
GATEWAY
REG 0x3023, BIT 1
PERIOD
MONITOR
REG 0x301E, BITS[3:2]
COMMON TIME SCALE
IUTS 1
COMMON TIME SCALE
CCS READY
CCDPLL UNLOCK
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
Figure 54. IUTS Block Diagram
Rev. 0 | Page 69 of 205
AD9546
Data Sheet
status. IUTS valid status appears in Bit 0 and Bit 2 (for IUTS 0
and IUTS 1, respectively). IUTS invalid status appears in Bit 1 and
Bit 3 (for IUTS 0 and IUTS 1, respectively).
PARAMETER DESTINATION SELECTION
IUTS 0 and IUTS 1 each require the following two parameters
from the register map as part of normal operation: time code
and restart.
The status bits exist as part of the IRQ mechanism (see the
Interrupt Request (IRQ) section). IUTS status indication is also
available via an appropriately configured Mx status pin (see the
Status and Control Pins section).
The register map has one instance of the two parameters which
the user designates for a specific IUTS via the 5-bit unsigned
destination bit field in Bits[4:0] of Register 0x0F09. The
destination bits select the desired path of the two parameters to
the target resource (via the two 1 to 4 demuxes shown in Figure 54)
according to Table 45. The target resources include both IUTSs
as well as the CCS. For destination bit field values of 30 and 31,
see the Synchronization Time section of the Common Clock
Synchronizer (CCS) section for details.
An IUTS indicates invalid status under the following conditions:
•
The user declares an IUTS as invalid (see the Declaring an
IUTS section)
•
The common clock DPLL indicates unlocked (unless the
user activates the bypass lock option (see the IUTS Bypass
Lock Option section)
•
A restart (see the IUTS Restart section) accompanies an
input time code
Table 45. Destination Selection
Destination Bit Field Value (Decimal) Target Resource
•
•
An input time code arrives when the IUTS is busy
The IUTS period monitor detects a period discontinuity
0 to 12
13
Reserved
IUTS 0
14
15 to 29
30
IUTS 1
Reserved
CCS Sync 0 time code
CCS Sync 1 time code
IUTS RESTART
The IUTS period processor uses an internal algorithm for filtering
period jitter that may be present on the incoming sequence of time
codes. The algorithm gradually reduces the filtering bandwidth
as it collects period samples over time. In some cases, there may
be a need to restart the filtering process (such as the user
switching the IUTS to a new time code source, for example).
31
TIME CODE AND FORMAT
The time code and format functionality is identical to that
described in the Synchronization Time section of the Common
Clock Synchronizer (CCS) section. The only difference is that
the user selects IUTS 0 or IUTS 1 as the target resource per
Table 45 via Bits[4:0] of Register 0x0F09.
To restart the IUTS, use the restart bit in Bit 0 of Register 0x0F08.
Logic 0 (default) is normal operation, whereas Logic 1 forces
the IUTS to restart.
IUTS TIME OFFSET
IUTS BYPASS LOCK OPTION
Each IUTS allow the user to add a constant time offset value to
each estimated time code generated by the IUTS period
processor. The time offset programs as a 24-bit signed value in
Register 0x0F01 to Register 0x0F03 for IUTS 0 and
Register 0x0F05 to Register 0x0F07 for IUTS 1.
The time offset carries units of 2−48 sec, yielding a range of
approximately 29.8 ns. For example, to program a time offset
of 5 ns (5 × 10−9 sec), first convert to the appropriate units and
round the result to the nearest integer as follows:
Normally, each IUTS requires the common clock DPLL to
indicate locked status to processes time codes. The user,
however, can remove this constraint for each IUTS
independently via Bit 1 of Register 0x0F00 for IUTS 0 and
Register 0x0F04 for IUTS 1. Logic 0 (default) means the IUTS is
inoperative when the common clock DPLL indicates an unlock
condition. Logic 1 causes the IUTS to disregard the lock status
of the common clock DPLL.
DECLARING AN IUTS INVALID
5 × 10−9 × 248 = 1,407,375
Because the output of an IUTS is a digitized clock signal
available to other digitized clocking resources, the user may
need to declare an IUTS as an invalid clock source. For
example, the user may be aware of a problem with the source of
input time codes to the IUTS that renders the IUTS output
unsuitable as a digitized clock source to downstream resources.
To handle such conditions, each IUTS allows manual
invalidation of an IUTS as a time stamp source.
This value in 24-bit hexadecimal format is: 0x 15 798F.
IUTS STATUS
When the IUTS period processor establishes the period
associated with successive time codes, it indicates a ready status
(Logic 1) via Bits[1:0] of Register 0x3023, with Bit 1 and Bit 0
indicating the status of IUTS 1 and IUTS 0, respectively.
However, the user can manually override the valid status of an
IUTS (see the Declaring an IUTS Invalid section).
To invalidate an IUTS, use Bit 0 of Register 0x0F00 for IUTS 0 and
Register 0x0F04 for IUTS 1. Logic 0 (default) invalidates the
IUTS, whereas Logic 1 allows normal IUTS operation. Note
that manual invalidation of an IUTS causes the IUTS to
indicate an invalid status (see the IUTS Status section).
The user also has access to two IRQ status bits for each IUTS
via Bits[3:0] of Register 0x301E. One bit of the pair indicates
IUTS valid status, and the other bit indicates IUTS invalid
Rev. 0 | Page 70 of 205
Data Sheet
AD9546
ANALOG CLOCK LOOPBACK
route to the slave nodes. The common clocks result from
OVERVIEW
routing the common clock reference through the master DPLL
channel, preferably operating in zero delay mode to minimize
phase (delay) offset. The slave nodes use the common clock to
generate their own common time scale and output clock
signals. The goal of the distributed clock system is to have all
output clock signals of the slave nodes synchronous to the
common clock of the slave node, which is synchronous to the
common clock reference of the master.
Analog loopback provides a means for designers of distributed
clock systems to loop an analog clock signal back to measure
the round trip delay. Knowledge of the round trip delay allows
the user to estimate the timing error associated with physical
connections (cables and circuit board traces, for example) and
compensate for those errors in downstream nodes. The net
effect is improved synchronization between nodes in a
distributed clock system.
A significant challenge in the implementation of a high
precision distributed clock system is accounting for the time
delay of the common clock as it propagates to each slave node.
The delay arises from the physical interconnects (cables, for
example) as well as active components (drivers and buffers, for
example) in the signal path. Without knowledge of the amount
of delay associated with each master to slave common clock
path, the delay constitutes a timing error that degrades the
precision of synchronization between nodes. To help remedy
delay errors, the AD9546 offers high resolution time
measurement capability (the UTS system) together with an
analog loopback feature to provide a mechanism to quantify
timing errors in the signal path. By quantifying the time errors,
the user can compensate for those errors by inserting corrective
time offsets via the various offset controls provided by the
AD9546.
The diagram of Figure 55 shows an example of a distributed
digitized clocking system comprising three AD9546 devices.
On the left is the master timing node, and on the right are the
slave timing nodes. Each node has a processor with a SPI
connection to the AD9546 associated with the node and a
digital bus interconnection between the processors of all nodes.
The main purpose of the digital bus is the transport of digitized
clocks between nodes. However, the digital bus enables the
transport of other data between nodes, allowing node
processors to share data (for example, time offset information
for correcting delays between nodes).
The primary timing signal for the system is a common clock
reference, which drives the REFB input of the master node. The
master node generates its common time scale from the
common clock reference. The master node also uses the
common clock reference to generate common clock signals that
DIGITIZED CLOCK TRANSPORT
PROCESSOR
ANALOG
CLOCK
SIGNALS
SPI
ANALOG
MASTER
M4
PROCESSOR
DIGITIZED
CLOCKING
NODE
LOOPBACK
ANALOG
CLOCK
SLAVE
DIGITIZED
CLOCKING
NODE
A
B
C
SPI
TC
SIGNALS
TC
TC
1
3
2
PHYSICAL
CLOCK
REFB
TIME
STAMPS
COMMON
TIME
SCALE
UTS
UTS
UTS
CONVERTER
PHYSICAL
CLOCK
TS
3
AD9546
CONVERTER
TIME
STAMPS
TS
COMMON CLOCK
DPLL AND
SYNCHRONIZER
PHYSICAL
CLOCK
CONVERTER
ANALOG
LOOPBACK
2
PROCESSOR
TIME
PHYSICAL
COMMON REFB
CLOCK
REFERENCE
TS
STAMPS
DPLL
1
CLOCK
CHANNEL
CONVERTER
SPI
ANALOG
M4
AD9546
LOOPBACK
ANALOG
CLOCK
SIGNALS
SLAVE
DIGITIZED
CLOCKING
NODE
A
B
C
ZERO DELAY
OPERATING
MODE
PHYSICAL
CLOCK
CONVERTER
REFB
TIME
STAMPS
AD9546
COMMON
CLOCK
NOTES
1. TS TO TS CONSTITUTE TIME STAMP SEQUENCE 1 TO TIME STAMP SEQUENCE 3.
1
3
2. TC TO TC CONSTITUTE TIME CODE SEQUENCE 1 TO TIME CODE SEQUENCE 3.
1
3
Figure 55. Distributed Clock System with Analog Loopback
Rev. 0 | Page 71 of 205
AD9546
Data Sheet
details). Furthermore, selecting a loopback path does not
ANALOG LOOPBACK FEATURE
automatically complete the loopback path. To complete the
loopback path, the user must enable the M4 output driver by
programming Register 0x0106, Bit 7 = 1.
Figure 55 shows the analog loopback feature of the AD9546
color coded in red. Figure 56 shows a detailed diagram of the
internal structure of the analog loopback feature. The analog
loopback feature uses the REFB and REFBB pins as the analog
loopback input and the M4 pin as the analog loopback output.
The REFB and REFBB inputs both provide loopback paths at
the output of the single-ended receiver (the A path), the input
to the TDC (the B path) and a replica of the reference
demodulator synchronization signal (the C path). Note that the
A path does not provide access to the output of the differential
receiver.
MEASURING ROUND TRIP DELAY
The analog loopback feature enables the measurement of a
round trip delay. The main goal of a round trip delay
measurement is to quantify the interconnect delay, which tends
to be the dominant time error contributor in a distributed clock
system. A round trip delay measurement assumes the following
three conditions:
•
•
The DPLL channel operates in zero delay mode
The magnitude of the round trip delay is less than the
period of consecutive time stamp events associated with
TS1, TS2, or TS3 (see Figure 55)
The user selects the desired loopback path via Bits[2:0] of
Register 0x2D02 per Table 46. The user selects the loopback
path based on the operating configuration of the REFB or
REFBB input.
•
The frequency of the common clocks is the same as the
frequency of the common clock reference
Table 46. Analog Loopback Path Selection
Bits[2:0] (Decimal)
Loopback Path
Operating in zero delay mode is a not a strict requirement, but
greatly simplifies the round trip delay measurement process.
The second and third conditions are not strict requirements,
but if violated, increase the complexity of the round trip delay
computation to a level beyond the scope of this data sheet.
0 (Default)
Disable loopback
1
2
3
4
5
REFB receiver output to M4
REFBB receiver output to M4
REFB divider output to M4
REFBB divider output to M4
REFB demodulator synchronization
output to M4
REFBB demodulator synchronization
output to M4
Reserved
To measure the round trip delay, the user programs one of the
UTS units of the master node to accept time stamps from the
physical clock converter associated with the common clock
reference, TS1 (refer to Figure 55). The UTS converts the TS1
time stamps to TC1 time codes. The user also programs a
second UTS unit to accept time stamps from the physical clock
converter associated with the desired analog loop back path. As
an example, assume the loopback path is from the lower slave
node in Figure 55. In this case, the UTS converts the TS2 time
stamps to TC2 time codes.
6
7
The selection of a loopback path (1 decimal through 6 decimal
in Table 46) substitutes the looped back analog clock signal for
the normal M4 status selection associated with Register 0x0186
(or Register 0x0106; see the Status and Control Pins section for
Rev. 0 | Page 72 of 205
Data Sheet
AD9546
REG 0x2D02, BITS[2:0]
3
REG 0x0106, BIT 7 = 1
Mx PIN
FUNCTION
LOGIC
1-OF-6 LOOPBACK
PATH SELECTOR
M4
A
C
B
SINGLE- ENDED
DIFFERENTIAL
REFB
TIME
TDC
÷R
STAMPS
EMBEDDED
CLOCK
DEMODULATOR
SYNC
SINGLE- ENDED
TIME
STAMPS
TDC
REFBB
NOTES
÷R
EMBEDDED
CLOCK
DEMODULATOR
SYNC
1. A RANGE OF BITS USES A COLON SEPARATOR
Figure 56. Analog Loopback Detail
The TC1 and TC2 time codes occur sequentially in pairs: TC1[x]
and TC2[x], TC1[x + 1] and TC2[x + 1], and so on. Each time
code pair provides a round trip delay measurement, RTD, given by
difference between the mean specified component delays and
the actual component delays also exists. Assuming the
interconnect delay dominates, an approximation of the forward
(or reverse) path interconnect delay, Δt, is
RTD[x] = TC2[x] − TC1[x]
Δt = ½ × RTDCORR
Ideally, each RTD[x] is the same value, but in practice, the
RTD[x] values vary slightly due to various noise contributors.
Generally, the user takes an average of several RTD[x] values to
yield an improved estimate of round trip delay. The round trip
delay measurement includes all delays in the loopback path:
interconnections and active elements (like buffers and other
semiconductor components). However, Table 29 provides static
delay values for sections of the internal loopback paths, which
the user can subtract from the round trip delay measurement to
account for the delay associated with the internal paths of the
device. The resulting value, RTDCORR, constitutes the corrected
round trip delay, which consists of the total interconnect delay
(forward and reverse paths) along with delay contributions of
any other components in the path. The small error due to the
With a value for RTDCORR and knowledge of the mean
component delays in the signal path, the user can program
corrective offsets in the appropriate nodes to counteract the
path delay.
EXCESSIVE ROUND TRIP DELAY
In some applications, interconnect lengths may result in a
round trip delay that exceeds the period of the underlying clock
period. Such a situation breaks the symmetry of the TC1 and
TC2 time code pairs. Techniques exist to overcome excessive
round trip delay, but such is beyond the scope of this data sheet.
Rev. 0 | Page 73 of 205
AD9546
Data Sheet
SYSTEM CLOCK PLL
The crystal path supports crystal resonators in the 25 MHz to
80 MHz frequency range. An internal maintaining amplifier
provides the negative resistance required to induce oscillation.
SYSTEM CLOCK PLL OVERVIEW
The system clock PLL (see Figure 57) comprises an Integer N
frequency synthesizer with a fully integrated loop filter and
VCO. The VCO output is the AD9546 system clock with a
frequency range of 2250 MHz to 2415 MHz. The XOA and
XOB pins constitute the input to the system clock PLL to which
a user connects a clock source or crystal resonator.
The internal amplifier expects an AT cut, fundamental mode
crystal with a maximum motional resistance of 100 Ω for
crystals up to 52 MHz, and 50 Ω for crystals up to 80 MHz. The
following crystals, listed in alphabetical order, may meet these
criteria:
AD9546
LOCK
DETECTOR
VCO
XTAL
CALIBRATION
•
•
•
•
•
•
•
AVX/Kyocera CX3225SB
ECS, Inc. ECX-32
Epson/Toyocom TSX-3225
Fox FX3225BS
NDK NX3225SA
Siward SX-3225
2×
XOA
XOB
PFD,
SYSTEM
CLOCK
CHARGE PUMP,
LOOP FILTER
÷J
1, 2, 4, 8
2250MHz TO
2415MHz
DIRECT
÷K
4 TO 255
Figure 57. System Clock PLL Block Diagram
Suntsu SCM10B48-49.152 MHz
SYSTEM CLOCK INPUT FREQUENCY
DECLARATION
Analog Devices, Inc., does not guarantee the operation of the
AD9546 with these crystals, nor does Analog Devices endorse
one crystal supplier over another. The AD9546 reference design
uses a readily available high performance 49.152 MHz crystal
with low spurious content.
Proper operation of the AD9546 requires the user to declare the
input reference frequency to the system clock PLL by
programming the 40-bit unsigned integer in Bits[39:0] of Register
0x0206 to Register 0x0202. The programmed value constitutes the
nominal frequency, in units of mHz, applied to the XOA and
XOB pins. The AD9546 evaluation software frequency
planning wizard calculates this value for the user.
Crystal resonators are subject to the specified maximum power
dissipation requirement of the manufacturer. As such, some
crystal resonators may require the use of a resistor to absorb
some of the power delivered by the maintaining amplifier to
satisfy the maximum power dissipation requirement of the
crystal resonator. Figure 58 shows the proper way to connect
the power limiting resistor, RLIMIT, to achieve optimal performance.
SYSTEM CLOCK SOURCE
The XOA and XOB pins serve as the input connection to the
system clock PLL, giving the user access to a crystal path (see
Figure 30) or a direct path (see Figure 31). Path selection is via
Register 0x0201, Bit 3, where a Logic 0 (default) selects the
direct path and Logic 1 selects the crystal path. The optimal
reference source for the system clock input is a crystal
resonator in the 50 MHz range or an ac-coupled square wave
source (single-ended or differential) with an 800 mV p-p
amplitude.
AD9546
XOA
XOB
R
LIMIT
C
C
TUNE
TUNE
Crystal Path
Figure 58. Power Limiting Resistor Connection
Select the crystal path by programming Register 0x0201, Bit 3, to
Logic 1. Refer to Figure 30 for connecting a crystal resonator to
the XOA and XOB pins. The CTUNE capacitors shown in Figure 30
relate to CLOAD and CSTRAY as follows:
Frequency Doubler (2× Multiplier)
When a crystal resonator is the frequency source to the system
clock, the user has the option of doubling the reference
frequency via Bit 0 of Register 0x0201. Programming a Logic 1
doubles the input reference frequency to the system clock PLL,
which improves the in-band noise of the PLL by 2 dB to 3 dB.
CTUNE = 2 × (CLOAD – CSTRAY)
where:
C
C
TUNE is a tuning capacitor.
LOAD is the load capacitance, per the specification of the crystal
The frequency doubler cannot be used with the direct path
input of the system clock PLL.
manufacturer.
CSTRAY is any additional parasitic capacitance.
As an example, when CLOAD = 10 pF and CSTRAY = 2 pF to 5 pF,
the value of CTUNE is approximately 15 pF.
Rev. 0 | Page 74 of 205
Data Sheet
AD9546
Direct Path
SYSTEM CLOCK PLL OUTPUT FREQUENCY
Select the direct path by programming Register 0x0201, Bit 3, to
Logic 0. Refer to Figure 31 for connecting a TCXO or OCXO
with a 3.3 V output to the XOA and XOB pins.
Calculate the system clock frequency, fS, as follows:
K
J
K
fS = fOSC
×
The direct path supports low frequency LVPECL, LVDS, CMOS,
or sinusoidal clock sources as a reference to the system clock PLL.
For a sinusoidal source, however, it is best to use a frequency of
50 MHz or greater. The low slew rate of lower frequency sinusoids
tends to yield nonoptimal noise performance.
where:
f
OSC is the input frequency.
K is the feedback divide ratio.
J is the input divide ratio. J = ½ when using the 2× frequency
multiplier.
The direct path has a differential receiver with a self bias of
0.75 V dc. Generally, the presence of the bias voltage
necessitates the use of ac coupling between the external source
and the XOA and XOB pins. Furthermore, when using a 3.3 V
CMOS oscillator as the system clock PLL reference source, in
addition to ac coupling, it is important to use a voltage divider
to reduce the 3.3 V swing to a maximum of 1.14 V (note that
the optimal voltage swing is 800 mV p-p).
The user must choose fOSC, K, and J such that fS satisfies the
VCO range of 2250 MHz to 2415 MHz.
SYSTEM CLOCK PLL LOCK DETECTOR
The system clock PLL features a simple lock detector that com-
pares the time difference between the reference and feedback
clock edges. The user can check the status of the lock detector
via Register 0x3001, Bit 0, where Logic 1 indicates locked and
Logic 0 indicates unlocked.
Degraded phase noise performance typically occurs when the
frequency at the input of the PFD of the system clock PLL is
less than 50 MHz. For optimal phase noise performance, use
the quartz crystal resonator path with a crystal resonator
frequency ≥ 50 MHz, the frequency doubler enabled, and the
device configured to use system clock Compensation Method 3
(see the System Clock Compensation section).
SYSTEM CLOCK STABILITY TIMER
Because time processing blocks within the AD9546 depend on
the system clock generating a stable frequency, the system clock
PLL provides an indication of the status. The status of the
system clock PLL is available to the user as well as directly to
certain internal time keeping blocks.
PRESCALE DIVIDER
At initial power-up, the system clock status is unknown and
reported as being unstable. However, after the user programs
the system clock registers and the system clock PLL VCO
calibrates, the system clock PLL locks shortly thereafter.
The system clock PLL includes an input prescale divider
programmable for divide ratios of 1 (default), 2, 4, or 8. The
purpose of the divider provides flexible frequency planning for
mitigating potential spurs in the output clock signals of the
AD9546. The user selects the divide ratio via Register 0x0201,
Bits[2:1]. The corresponding divide value is 2J, where J is the
decimal value of the 2-bit number comprising Register 0x0201,
Bits[2:1].
Even though the system clock PLL may indicate a lock status,
the user can introduce an additional holdoff period, in units of
milliseconds (10−3 sec), via an unsigned 20-bit integer in
Bits[19:0] of Register 0x0209 to Register 0x0207
(Register 0x0209, Bit 3 denotes the MSB). There are two special
values: 0 and 1,048,575. A value of 0 causes the system clock to
indicate an unstable status regardless of the actual state of the
system clock (useful for debugging purposes, because the user
can force an unstable status indication even though the PLL is
stable). A value of 1,048,575 is invalid because it makes Register
0x3001, Bit 1 undefined when the system clock PLL is
unlocked.
For example, a value of 10 (binary) in Register 0x0201 Bits[2:1],
means J = 2 (decimal), making the divide ratio = 4 (2J = 22 = 4).
FEEDBACK DIVIDER
The output of the system clock PLL constitutes the system clock
frequency, fS. The system clock frequency depends on the value of
the feedback divider (K in Figure 57). The feedback divide ratio has
a range of 4 to 255, which the user programs via Register 0x0200,
Bits[7:0]. The programmed register value is the divide ratio. For
example, a programmed value of 100 (0x64 hexadecimal) yields a
divide ratio of 100.
SYSTEM CLOCK CALIBRATION
The VCO in the system clock PLL requires a calibration
sequence. At power-up, part of the initialization sequence of
the device includes system clock calibration. In general, the
user is not required to calibrate the system clock manually.
However, changing any of the system clock parameters that
result in a new VCO frequency requires a manual system clock
calibration via Register 0x2000, Bit 2, is not an autoclearing bit.
Therefore, the full programming sequence is to write Logic 1,
assert an IO update, write Logic 0, and assert an IO update.
Furthermore, whenever the system clock PLL is in the process
Rev. 0 | Page 75 of 205
AD9546
Data Sheet
of calibrating, status Bit 2 of Register 0x3001 indicates busy
(Logic 1).
example). An alternate solution uses the integrated system
clock compensation of the AD9546, which enables the use of a
crystal resonator as the system clock source. The basic concept
is the connection of a crystal resonator (25 MHz to 80 MHz) to
the XOA and XOB pins per Figure 30, which capitalizes on the
exceptional phase noise performance of the crystal.
Bit 1 of Register 0x2000 provides an alternate means to
calibrate the system clock PLL (this bit also calibrates APLL0
and APLL1).
Although the AD9546 has a variety of power-down modes,
including a specific system clock power-down, it is not
necessary to calibrate the system clock when recovering from
any of the power-down modes. However, recovering from a
power-down of the system clock necessitates calibration of the
APLLs due to the loss of the system clock during its power-down.
However, the crystal resonator lacks the desired stability. Instead,
the stability arises via the connection of a highly stable source
(for example, TCXO, OCXO, or GPS) to an unused REFx or
Mx input. The stable source then serves as the reference
frequency to the system clock compensation feature of the
device (see the System Clock Compensation section). The
system clock compensation feature allows the device to benefit
from both the stability of the highly stable source and the
exceptional phase noise performance of the crystal resonator.
Thus, the system clock compensation feature provides a
mechanism to employ a low loop bandwidth DPLL (as required
by the application) even though the system clock source is a
relatively unstable crystal resonator.
SYSTEM CLOCK STABILITY COMPENSATION
An application requiring low DPLL loop bandwidth
necessitates the use of a very stable system clock source. Using a
relatively unstable system clock source like a crystal resonator
or crystal oscillator, for example, in conjunction with the DPLL
having a very narrow loop bandwidth (less than approximately
50 Hz) may cause the DPLL to unlock or experience
intermittent unlock events. The typical solution requires the
use of a very stable system clock source (an OCXO, for
Rev. 0 | Page 76 of 205
Data Sheet
AD9546
REFERENCE CLOCK INPUT RESOURCES
The AD9546 has eight reference clock inputs as shown in
Figure 59. The eight inputs comprise two groups of four:
The primary reference inputs have configurable input receivers
(see the Reference Receivers section) and dedicated pins. The
user can configure the REFA and REFAA inputs as a single
differential receiver or as two independent, single-ended
receivers. The REFB and REFBB inputs can be configured
likewise.
•
Four primary reference inputs (REFA, REFAA, REFB, and
REFBB)
•
Four auxiliary reference inputs (auxiliary REF0, auxiliary
REF1, auxiliary REF2, and auxiliary REF3)
The auxiliary reference clock inputs have nonconfigurable,
single-ended CMOS receivers, and rather than having
dedicated pins, they are accessible only through Mx Pins (see
the Status and Control Pins section).
The two groups are functionally identical except for the
configuration of their input receivers.
Each reference clock input has a dedicated reference monitor,
reference divider (R divider), TDC, and reference demodulator.
REFA
TIME STAMPS
TIME STAMPS
TDC
÷R
÷R
REF
MON
DEMOD
TDC
REFAA
REF
MON
DEMOD
TIME STAMPS
TIME STAMPS
REFB
REFBB
AUX REF0
TDC
M0
M1
TIME STAMPS
÷R
TIME STAMPS
TIME STAMPS
TIME STAMPS
M2
REF
MON
Mx PIN
FUNCTION
LOGIC
M3
M4
DEMOD
AUX REF1
AUX REF2
AUX REF3
SDO/M5
CSB/M6
AD9546
Figure 59. Reference Inputs Block Diagram
Rev. 0 | Page 77 of 205
AD9546
Data Sheet
REFERENCE RECEIVERS
Bits[7:4] of Register 0x0300 and Register 0x0304 per Table 47.
Bits[7:4] are only effective when single-ended mode is in effect
per Bit 0 of Register 0x0300 and/or Register 0x0304.
REFERENCE RECEIVERS OVERVIEW
The four dedicated reference clock input pins (REFA, REFAA,
REFB, and REFBB) support up to four reference clock signals.
Two of the four pins designate the A reference input path, and
the other two pins designate the B reference input path. Both
the A and B reference inputs are configurable as either one
differential input or as two single-ended inputs. Each A and B
reference input path has a dedicated REFx input mode bit
(where x is A or B) for selecting single-ended or differential
modes in Bit 0 of Register 0x0300 for REFA and
Register 0x0304 for REFB. Logic 0 (default) selects single-ended
mode, and Logic 1 selects differential mode. To accommodate
input signals with slow rising and falling edges, both the
differential and single-ended input receivers employ hysteresis.
Hysteresis also ensures that a disconnected or floating input
does not cause the receiver to oscillate.
The ac-coupled mode has an internal bias termination equivalent
to a 0.6 V dc source with a 23.5 kΩ series resistance. The
dc-coupled modes have a 47 kΩ resistor connected to ground.
The internal pull-up mode has a 47 kΩ resistor connected to
1.2 V.
Single-Ended DC-Coupled Interface (1.2 V or 1.8 V CMOS)
Refer to Figure 32 for connecting a 1.2 V or 1.8 V CMOS driver to
a reference input. Be sure to select the corresponding dc-coupled
1.2 V or 1.8 V CMOS mode per Table 47.
Single-Ended DC-Coupled Internal Pull-Up Resistor
Refer to Figure 34 for connecting an open-collector or open-drain
driver to a reference input. Be sure to select the internal pull-up
mode per Table 47, which provides an internal pull-up resistor
(~47 kΩ) to 1.2 V.
The input receivers of the auxiliary references (accessible via
appropriately configured Mx control pins) are not configurable.
Instead, they have dc-coupled, single-ended, 1.8 V CMOS
receivers.
Single-Ended AC-Coupled Interface
Refer to Figure 33 to ac couple a driver to a reference input. Be
sure to select the ac-coupled interface mode per Table 47,
which provides the appropriate internal bias network.
SINGLE-ENDED MODE
When configured for single-ended operation, the input receivers
accommodate either ac- or dc-coupled input signals via
Table 47. Single-Ended Mode Bit Field Settings
Register
Bits[7:4]1
XX00
XX01
XX10
XX11
00XX
01XX
10XX
11XX
XX00
XX01
XX10
XX11
00XX
01XX
10XX
11XX
Reference
Description
0x0300
REFA
REFA
REFA
REFA
REFAA
REFAA
REFAA
REFAA
REFB
REFB
REFB
REFB
REFBB
REFBB
REFBB
REFBB
AC-coupled 1.2 V interface (default)
DC-coupled 1.2 V CMOS
DC-coupled 1.8 V CMOS
Internal pull-up resistor (disable pull-down resistor)
AC-coupled 1.2 V interface (default)
DC-coupled 1.2 V CMOS
DC-coupled 1.8 V CMOS
Internal pull-up resistor (disable pull-down resistor)
AC-coupled 1.2 V interface (default)
DC-coupled 1.2 V CMOS
0x0304
DC-coupled 1.8 V CMOS
Internal pull-up resistor (disable pull-down resistor)
AC-coupled 1.2 V interface (default)
DC-coupled 1.2 V CMOS
DC-coupled 1.8 V CMOS
Internal pull-up resistor (disable pull-down resistor)
1 X means don’t care.
Rev. 0 | Page 78 of 205
Data Sheet
AD9546
For differential mode operating frequencies in excess of
DIFFERENTIAL MODE
10.24 MHz, use the ac-coupled differential mode and connect
the reference input clock signal to the reference input pins
through a series dc blocking capacitor. For differential mode
operating frequencies < 10.24 MHz, use dc-coupled differential
mode. For direct connection of a low frequency (<10.24 MHz)
LVDS input reference source, use dc-coupled LVDS mode,
which provides the necessary level shifting.
When configured for differential operation, the input receivers
accommodate either ac-coupled or dc-coupled input signals per
Table 48. Bits[3:2] are only effective when the corresponding
differential mode is in effect.
Table 48. Differential Mode Settings
Register Bits[3:2]
Reference
REFA, REFAA
REFA, REFAA
REFA, REFAA
Unused
Description
0x0300
00
01
10
11
00
AC-coupled (default)
DC-coupled
DC-coupled LVDS
Unused
DC-Coupled LVDS (<10.24 MHz) Interface
Use Figure 35 to make a dc-coupled connection to a standard
LVDS driver having a 1.2 V common-mode output bias. Be
sure to select the dc-coupled LVDS mode per Table 48, which
provides the appropriate internal level shifting for the input
receiver. This configuration supports input frequencies up to
10.24 MHz.
0x0304
REFB, REFBB
AC-coupled interface
(default)
01
10
11
REFB, REFBB
REFB, REFBB
Unused
DC-coupled 1.2 V
CMOS
DC-coupled 1.8 V
CMOS
Differential AC Coupling (>10.24 MHz)
For differential input signals in excess of 10.24 MHz, refer to
Figure 36 to make an ac-coupled connection to an external
driver. Be sure to select the ac-coupled mode per Table 48.
Unused
The ac-coupled mode has an internal bias termination on each
differential input pin, which has a Thevenin equivalent of a
0.6 V dc source in series with a 23.5 kΩ series. The dc-coupled
mode expects the external driver to provide a 0.6 V common-
mode bias. The dc-coupled LVDS mode expects a 1.2 V
common-mode source with LVDS signal levels, as delivered by
a standard LVDS driver.
Differential DC Coupling (>10.24 MHz)
Use Figure 37 to make a dc-coupled connection to a differential
driver having a 0.6 V common-mode output bias. Be sure to
select the dc-coupled mode per Table 48.
Rev. 0 | Page 79 of 205
AD9546
Data Sheet
REFERENCE DIVIDERS (R DIVIDERS)
Each reference input has a dedicated divider, Rx (where x
differentiates between the eight reference inputs: A, AA, B, BB,
or auxiliary REF0 to auxiliary REF3). The primary purpose of
the Rx dividers is to reduce the input reference frequency
(assuming it is greater than 200 kHz) to a value between 1 Hz
and 200 kHz to satisfy the input frequency bounds of the TDC.
1000 is required. Thus, the required bit field value is 999
(0x000003E7 hexadecimal).
Table 49. Reference Divider Address Ranges
Rx Reference
Register Address Range
0x0400 to 0x0403
0x0420 to 0x0423
0x0440 to 0x0443
0x0460 to 0x0463
0x0480 to 0x0483
0x04A0 to 0x04A3
0x04C0 to 0x04C3
0x04E0 to 0x04E3
REFA
REFAA
REFB
REFBB
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
The user programs Rx via the 30-bit unsigned bit fields per
Table 49. The divide ratio of Rx depends on the value, R, the user
programs in the bit fields. The user must program the bit field
with a value one less than the desired divide ratio. Thus, the Rx
dividers provide divide ratios from 1 to 1,073,741,824.
For example, for an input reference frequency of 125 MHz and
the desired Rx output frequency of 125 kHz, a divide ratio of
Rev. 0 | Page 80 of 205
Data Sheet
AD9546
REFERENCE MONITOR
delivers digital (numeric) time stamps to the reference monitor,
which processes the time stamps to determine the status of the
input clock signal. Each reference TDC can connect to either
DPLL input via the digital cross point mux (see Figure 1). Even
when a reference does not connect to one of the DPLLs via the
cross point mux, the corresponding reference monitor
continues to observe the reference signal and generate status
signals accordingly.
REFERENCE MONITOR OVERVIEW
The AD9546 has eight independent reference monitors, one for
each reference input. Figure 60 shows the REFA input along
with its associated reference monitor. The reference monitor
diagram of Figure 60 applies identically for the REFAA, REFB,
REFBB, auxiliary REF0, auxiliary REF1, auxiliary REF2, and
auxiliary REF3 inputs by replacing all implied references to
REFA with the other reference inputs.
The reference monitor uses certain user-programmable
parameters (see the Reference Monitor Controls section) to
perform various monitoring functions. The reference monitor
provides status indicators (see the Reference Monitor Status
Indicators section) that allow the user to assess the quality of
the reference input signal.
The user has the option of powering down unused primary
references via Register 0x2001, Bits[3:0] (see the Reference
Clock Input Resources section for a definition of primary
references).
Each reference input has a dedicated input divider (for
example, RA in Figure 60) with a programmable divide ratio
(see the Reference Dividers (R Dividers) section). Because the
reference monitor follows the R divider in the signal path, the
reference monitor essentially undersamples the reference input
clock signal for divide ratios greater than one.
The AD9546 uses the status provided by the reference monitor
to control certain internal functions (automatic reference
switching, for example). The user has access to the status
signals via the register map or as a physical signal via the Mx
pins (see the Status and Control Pins section and Interrupt
Request (IRQ) section for more details).
Between the R divider and the reference monitor is a TDC (see
the Time to Digital Converter (TDC) section). The TDC
TO DIGITAL CROSS
POINT MUX
REG 0x0403 TO 0x0400, BITS[29:0]
REFERENCE
R
INPUT
REG 0x2001,
BIT 0
30
COMPONENTS
REFERENCE MONITOR
VALIDATION TIMER
USER
CHANGES
tVALID
TIME
STAMPS
RESET
÷R
REFA
TDC
A
1kHz
RESET
tVALID
20
REG 0x0412 TO 0x0410,
BITS[19:0]
CLK PERIOD
END
TIMEOUT
VALID
REG 0x2002, BIT 0
REG 0x3005, BIT 4
START
DONE
RESET
USER CHANGES R
BYPASS
REG 0x2004, BIT 0
REG 0x2003, BIT 0
INVALIDATE
SYSTEM CLOCK
UNSTABLE
USER CHANGES
INPUT MODE
TIME STAMP
PROCESSOR
AND JITTER
ESTIMATOR
1
0
FAULT
USER CHANGES tREF
,
REG 0x3005, BIT 3
PHASE STEP
DETECTED
THRESHOLD, HYSTERESIS,
OR JITTER TOLERANCE
SLOW
FAST
LOS
tREF
REG 0x3005, BIT 0
REG 0x3005, BIT 1
REG 0x3005, BIT 5
REG 0x040B TO 0x0404, BITS[59:0]
REG 0x040E TO 0x040C, BITS[23:0]
REG 0x040F, BITS[2:0]
THRESHOLD
NO SIGNAL
HYSTERESIS
DISCONTINUITY
JITTER TOLERANCE
REG 0x0414 TO 0x0413, BITS[15:0]
EXCESSIVE
JITTER
EXCESS
JITTER
REG 0x3005, BIT 2
JITTER
TOLERANCE
≠ 0
NOTES
1. REGISTER ADDRESSES SHOWN FOR REFA ONLY
2. A RANGE OF BITS USES A COLON SEPARATOR
3. CERTAIN STATUS AND CONTROL BITS CAN BE ACCESSED VIA THE MX PINS
Figure 60. Reference Monitor
Rev. 0 | Page 81 of 205
AD9546
Data Sheet
REFERENCE MONITOR BASE PERIOD
REFERENCE MONITOR CONTROLS
The reference monitor operates directly on TDC time stamps,
which occur with each rising edge from the output of the
R divider. Because time stamps result from rising edges at the
TDC input, which occur at intervals of the input reference
period multiplied by R (the reference divider ratio), the base
period (tBASE) of the reference monitor is R times the reference
period.
The reference monitor requires certain advance information to
function properly. The user provides this information by
programming the appropriate parameters as described in the
Reference Period (tREF), Reference Period Threshold, Hysteresis,
Jitter Tolerance, Validation Time (tVALID), Timeout, Bypass, and
Invalidate sections.
Reference Period (tREF
)
REFERENCE MONITOR STATUS INDICATORS
Each reference input has a dedicated unsigned 60-bit bit field in
the register address ranges shown in Table 51 for specifying
The reference monitor associated with each reference provides
six status indicators: valid, fault, LOS, slow, fast, and excess
jitter. These indicators are associated with the D4, D3, D5, D0,
D1, and D2 bits, respectively, but the register address depends
on the reference per Table 50.
t
REF. tREF constitutes the nominal period of the associated input
reference and represents time in units of attoseconds (10−18 sec)
with an upper limit of ~1.15 sec.
For example, given the REFAA input has a nominal 2.048 MHz
input signal. The corresponding period is 1/(2.048 × 106) sec,
which, when multiplied by 1018, yields 488,281,250,000
attoseconds. Program Register 0x042B to Register 0x0424 with
the 60-bit value corresponding to 488,281,250,000 attoseconds
(0x 000 0071 AFD4 98D0 hexadecimal).
Table 50. Reference Monitor Status Register Address
Reference Input
Register Address
REFA
0x3005
REFAA
0x3006
REFB
0x3007
REFBB
0x3008
0x301F
0x3020
0x3021
Table 51. Nominal Period Address Ranges
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
Reference Input
Register Address Range
0x0404 to 0x040B
0x0424 to 0x042B
0x0444 to 0x044B
0x0464 to 0x046B
0x0484 to 0x048B
0x04A4 to 0x04AB
0x04C4 to 0x04CB
0x04E4 to 0x04EB
REFA
REFAA
0x3022
REFB
REFBB
Valid
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
The valid status indicator is Logic 1 at the expiration of the
optional validation time (see the Reference Validation section).
Fault
The fault status indicator is Logic 1 when the LOS, slow, fast, or
excess jitter status is true.
Reference Period Threshold
The reference period threshold is the maximum relative
deviation from tREF the reference monitor uses to declare a
nonfaulted reference as out of tolerance (that is, slow or fast).
To specify the threshold parameter, each reference input has a
dedicated register group comprising an unsigned 24-bit integer
per the register address ranges shown in Table 52. The 24-bit
number represents fractional units of parts per billion (ppb) up
to a maximum of approximately 17 million ppb (1.7%), with
the default setting yielding 100 ppm.
Loss of Signal (LOS)
The LOS status indicator is Logic 1 if the reference monitor
determines the input signal is absent, very slow (>1.15 × tBASE
or a phase discontinuity occurred (see the Discontinuity
Detection section).
)
Slow
The slow status indicator is Logic 1 if the period of the input
signal is greater than the maximum period prescribed by the
threshold setting (see the Reference Period Threshold section).
Table 52. Offset Limit Address Ranges
Reference Input
Register Address Range
0x040C to 0x040E
0x042C to 0x042E
0x044C to 0x044E
0x046C to 0x046E
0x048C to 0x048E
0x04AC to 0x04AE
0x04CC to 0x04CE
0x04EC to 0x04EE
Fast
REFA
REFAA
REFB
REFBB
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
The fast status indicator is Logic 1 if the period of the input
signal is less than the minimum period prescribed by the
threshold setting (see the Reference Period Threshold section).
Excess Jitter
The excess jitter status indicator is Logic 1 if the reference
monitor determines the input signal jitter exceeds the limits
specified by jitter tolerance (see the Jitter Tolerance section) or
a phase discontinuity occurred (see the Discontinuity Detection
section).
Rev. 0 | Page 82 of 205
Data Sheet
AD9546
Because the threshold parameter specifies a fractional period
(δp), it relates to fractional frequency (δf) as follows:
The SF value constitutes a fractional portion of the value of
threshold and defines the amount of hysteresis.
δp = −δf/(1 + δf)
(1)
Hysteresis (ppb) = Threshold × SF
For example, consider the case where the expected maximum
relative deviation of the input reference frequency for REFAA
is 75 × 10−6 (75 ppm). That is, δf = 75 × 10−6.
From Equation 1, δp = −74.994375 × 10−6, or −74.994375 ppm,
or −74994.375 ppb. Because the threshold parameter is an
unsigned number with integer units of ppb, threshold = 74,994.
Therefore, program Register 0x042C to Register 0x042E with the
24-bit value corresponding to 74,994 (0x 01 24F2 hexadecimal).
For example, given threshold = 74,994 ppb and SF = 0.0625,
determine the hysteresis.
Hysteresis (ppb) = 74,994 ppb × 0.0625 = 4,687
The reference monitor compares reference period samples
against the tREF, threshold, and SF parameters to make fault and
unfault decisions per Figure 61.
TIME
MINIMUM DEVIATION FROM
THE BASE PERIOD FOR AN
To properly verify small values of the threshold parameter, the
reference monitor requires a very precise timing reference. The
reference monitor uses time stamps from the TDC to assess the
input signal parameters. Because the underlying precision of
the TDC relates to the system clock, small threshold values
require a suitably precise system clock or a suitably
compensated system clock (see the System Clock
THRESHOLD
UNFAULTED REFERENCE
TO BECOME FAULTED
THRESHOLD × SF
MAXIMUM DEVIATION FROM
THE BASE PERIOD FOR A
FAULTED REFERENCE TO
BECOME UNFAULTED
THRESHOLD × (1 − SF)
BASE PERIOD = tREF × R
BASE
PERIOD
Compensation section).
Hysteresis
Figure 61. Reference Monitor Hysteresis
Jitter Tolerance
The hysteresis parameter applies to a faulted reference, whereas
the threshold parameter applies to a nonfaulted reference. The
reference monitor makes fault and unfault decisions based on
how reference period samples relate to tREF, threshold, and
hysteresis. To specify the hysteresis parameter for a particular
reference, use Bits[2:0] of the register address associated with
the reference per Table 53.
Use the jitter tolerance parameter to define the maximum rms
jitter the reference monitor accepts before indicating an
excessive jitter condition. Each reference input has a dedicated
register group for entering jitter tolerance per the register
address ranges shown in Table 55. The jitter tolerance parameter
is an unsigned 16-bit integer with units of nanoseconds
(10−9 sec) up to a maximum of approximately 65.5 µs.
Table 53. Hysteresis Register Address
Reference Input
Register Address Range
Table 55. Jitter Tolerance Address Ranges
REFA
REFAA
REFB
REFBB
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
0x040F
0x042F
0x044F
0x046F
0x048F
0x04AF
0x04CF
0x04EF
Reference Input
Register Address Range
0x0413 to 0x0414
0x0433 to 0x0434
0x0453 to 0x0454
0x0473 to 0x0474
0x0493 to 0x0494
0x04B3 to 0x04B4
0x04D3 to 0x04D4
0x04F3 to 0x04F4
REFA
REFAA
REFB
REFBB
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
The programmed value of the hysteresis parameter defines a
hysteresis scale factor (SF) per Table 54.
For example, given an expected jitter limit of 75 ns rms, then
jitter tolerance = 75 (0x004B hexadecimal).
Table 54. Hysteresis Selection Table
Bits[2:0] (Decimal)
Scale Factor (SF)
When jitter tolerance = 0, the reference monitor does not include
its estimation of jitter as part of the excess noise status decision.
However, even when jitter tolerance = 0, the reference monitor
continues to perform jitter estimation to continue making out
of tolerance decisions.
0
1
2
3
4
5
6
7
0
0.03125
0.0625
0.125 (default)
0.25
0.5
0.75
0.875
Rev. 0 | Page 83 of 205
AD9546
Data Sheet
Validation Time (tVALID
)
auxiliary REF1, auxiliary REF0, REFBB, REFB, REFAA, and
REFA.
In some applications, a reference signal must maintain its
validity (that is, not faulted) for some minimum duration
(tVALID) before the system considers the reference valid. The
reference monitor has a built in timer to handle such timed
validation requirements. Each reference monitor has an
associated unsigned 20-bit integer for assigning tVALID per the
register address ranges shown in Table 56.
Programming timeout = 1 causes the validation timer to time
out immediately, giving the reference valid status (valid = 1).
Bypass
In certain circumstances, a user may find it necessary to bypass
the reference monitor control of the validation timer. To this
end, each reference input has a dedicated control bit (bypass)
via Register 0x2004, Bits[7:0], where each bit in the register
corresponds to a particular reference. Specifically, starting with
D7 and ending with D0, the corresponding references are
auxiliary REF3, auxiliary REF2, auxiliary REF1, auxiliary REF0,
REFBB, REFB, REFAA, and REFA.
Table 56. Validation Timer Address Ranges
Reference Input
Register Address Range
0x0410 to 0x0412
0x0430 to 0x0432
0x0450 to 0x0452
0x0470 to 0x0472
0x0490 to 0x0492
0x04B0 to 0x04B2
0x04D0 to 0x04D2
0x04F0 to 0x04F2
REFA
REFAA
REFB
REFBB
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
The bypass function does not actually bypass the reference
monitor itself, but only the behavior of valid status.
When a reference has bypass = 0 (default), the normal fault and
unfault conditions prescribed by the reference monitor control
the start of the validation timer, which controls the valid state.
Per Figure 60, when bypass = 0, the user has the option to force
an unfaulted reference to invalid status via the invalidate
control mechanism (see the Invalidate section). However, the
user cannot force a faulted reference to be valid via the
invalidate control mechanism when bypass = 0.
The programmed value of tVALID carries units of milliseconds,
up to a maximum of approximately 1048 sec (about 17.5 min).
For example, a validation time of 10 min requires
t
VALID = (10 min) × (60 sec/min) × (1000 ms/sec)
= 600,000 ms (0x927C0 hexadecimal)
When a reference has bypass = 1, starting the validation timer
falls under user control via the invalidate control mechanism
(see the Invalidate section). That is, programming invalidate =
0 is equivalent to the reference monitor exhibiting an unfaulted
condition. Conversely, programming invalidate = 1 is
equivalent to the reference monitor exhibiting a faulted
condition. The validation timer responds accordingly to the
imposed fault and unfault conditions.
The validation timer starts when the reference goes from a
faulted condition to an unfaulted condition. When the
validation timer expires (that is, in Figure 60, the done output
of the validation timer = 1), then the reference exhibits valid
status (valid = 1).
After the validation timer starts, if a faulted condition occurs,
the validation timer restarts. This behavior ensures that
reference validation occurs only after a reference maintains an
unfaulted condition for the full duration of the validation
timer.
Note that regardless of the state of the bypass or invalidate
control bit, the fast, slow, LOS, and excess jitter status
indicators of the reference monitor function normally and
remain accessible to the user via their respective status registers
(see Figure 60).
Note that programming a new value for tVALID causes the
validation timer to reset. Resetting the validation timer
invalidates the reference even if the reference was valid prior to
changing tVALID. However, even though the reference exhibits an
invalid status when the validation timer is reset, the reference
monitor continues to indicate the appropriate fault or
unfaulted condition based on the quality of the reference input
signal.
Invalidate
In certain circumstances, a user may find it necessary to force a
reference from a valid state to an invalid state. To this end, each
reference input has a dedicated control bit (invalidate) via
Register 0x2003, Bits[7:0], where each bit in the register
corresponds to a particular reference. Specifically, starting with
D7 and ending with D0, the corresponding references are
auxiliary REF3, auxiliary REF2, auxiliary REF1, auxiliary REF0,
REFBB, REFB, REFAA, and REFA.
The user can force a reference to invalid status without
affecting the normal operation of the reference monitor (see the
Invalidate section).
Timeout
A scenario for using invalidate = 1 when bypass = 0 is when the
reference signal must be traceable to a primary timing source,
but the user has external knowledge that the source is no longer
traceable. Under these circumstances, the reference signal may
still satisfy the parameters for being in tolerance. Therefore, the
reference monitor indicates the reference as unfaulted.
The user has the option to force the validation timer to
immediately time out via Register 0x2002, Bits[7:0], where each
bit in the register corresponds to a particular reference.
Specifically, starting with D7 and ending with D0, the
corresponding references are auxiliary REF3, auxiliary REF2,
Rev. 0 | Page 84 of 205
Data Sheet
AD9546
However, because the user knows the source is no longer
traceable, the user can force the reference to be invalid by
making invalidate = 1 (even though the reference monitor sees
the reference as being in tolerance). The user can reinstate a
valid status when conditions permit by making invalidate = 0.
However, if the reference signal goes out of tolerance (as
determined by the reference monitor) during the time that
invalidate = 1, then when the user makes invalidate = 0, the
reference remains invalid. That is, the user cannot force a
reference with an invalid status to a valid status via the fault
reference bit when bypass = 0.
REFERENCE MONITOR DECISION TIME
The reference period estimation algorithm takes into account
the value of the threshold parameter specified by the user (see
the Reference Monitor Controls section) and the actual jitter
present on the reference signal. In general, less averaging
required by the jitter estimator to bring the variance within a
suitable range means a shorter decision making time. The key
point is that both the value of the threshold parameter and the
amount of jitter present on the input signal govern the decision
time for the reference monitor to declare an out of tolerance
condition. That is, the decision time is not deterministic.
A scenario for using bypass = 1 (that is bypassing reference
monitor control of the validation timer) is when a reference
signal is of the gapped clock variety. A reference signal
employing a gapped clock is problematic for the reference monitor
because it cannot reliably discern an in or out of tolerance
condition. As such, an otherwise unfaulted reference can be
deemed faulted from the point of view of the reference
monitor. Under such conditions, the user can bypass reference
monitor control of the validation timer (and thereby valid and
invalid status) by making bypass = 1. With bypass = 1, the
invalidate control bit allows the user to mimic the fault and
unfault conditions normally generated by the reference
monitor.
Although jitter plays a role in the decision time of the reference
monitor under normal operation, jitter has little effect on the
decision time when the reference period is much greater than
the base period or when the reference signal disappears
completely. The reason is the reference monitor has advance
knowledge of the expected period of the reference signal
(namely, R × tREF). Thus, the reference monitor can rely on
internal timing (rather than TDC time stamps) to keep watch
for expected time stamps from the TDC. If the reference
monitor fails to observe the arrival of a new time stamp after a
period of 1.15 × R × tREF, the monitor declares an LOS.
Note that for an input signal that is slow (but not absent), after
the first assertion of LOS status, the LOS status toggles on each
subsequent occurrence of the expected period of the reference.
Thus, the LOS bit is not a reliable indicator of the loss of the
reference signal. Instead, use the fault status as an indicator of
the loss of the signal.
DISCONTINUITY DETECTION
The reference monitor detects the following discontinuities in
the period of the reference signal:
•
An absent reference signal or a reference period that
deviates by more than 3.125% of the expected value.
A reference period sample deviating by more than 64 times
REFERENCE VALIDATION
•
Some applications require a reference to be in an unfaulted
condition for a prescribed period before exhibiting valid status.
To accommodate these applications, the reference monitor
includes a programmable validation timer. A reference
transition from a faulted state to a nonfaulted state indicates a
start event for the validation timer, which begins counting
down. Upon expiration of the timer, valid = 1, which indicates
the reference is available for use (see Figure 60).
JEST plus ½ of jitter tolerance, where JEST is the current
estimate of the jitter variance.
•
Short term jitter exceeds long-term jitter by ~50%.
Detection of a discontinuity causes the discontinuity status
signal (see Figure 60) to be Logic 1.
REFERENCE PERIOD JITTER ESTIMATION
Because the reference monitor uses numeric time stamps to
measure the reference period, it can estimate the mean and
variance of the reference signal as it observes period samples
(that is, the difference between successive time stamps from the
TDC). The mean is an estimate of the reference period, and the
variance is an estimate of the jitter present on the reference
signal. The reference monitor uses the mean and jitter variance
estimate to make in tolerance or out of tolerance decisions by
comparing a span of 16 times the current jitter variance
estimate ( 8σ) against the tREF, threshold, and hysteresis
parameters. This statistical approach allows the reference
monitor to make tolerance decisions with a high degree of
confidence in a nearly optimal minimum observation time.
It is important to note that the validation timer stops counting
down and resets upon the occurrence of the faulted status (fault
= 1). A subsequent unfault condition causes the timer to
reinitiate a countdown from the programmed tVALID value.
Thus, the reference only attains valid status when the reference
remains unfaulted for the full duration of the validation timer.
The user has the option to force the validation timer to jump to
the end of its timing function by programming timeout = 1 (see
the Reference Monitor Controls section and Figure 60). In this
way, if a faulted reference has returned to a nonfaulted state
and is awaiting validation, the user can override the timer if
necessary, and immediately bring the reference to a valid status.
Programming timeout = 1 has no effect on a faulted reference
because a fault condition resets the validation timer.
Rev. 0 | Page 85 of 205
AD9546
Data Sheet
•
Optionally, when the reference experiences a significant
phase step as determined by the DPLL (assuming the user
enables phase step detection via Bit 7 of Register 0x2105
and Register 0x2205).
REFERENCE MONITOR RESET
The reference monitor associated with a specific reference
resets under certain conditions. A reset restarts the period and
jitter estimation process for the reference monitor. The
following list indicates the conditions that cause the reference
monitor to reset (also shown in Figure 60):
The last item requires that the user program a nonzero phase
step limit value. See the Phase Step Limit section for details
regarding the phase step limit feature.
•
•
When the system clock indicates unstable status
When the user changes the mode of the input receiver, the
R divider value, the tREF value, the ΔtREF value, the jitter
tolerance value, or the hysteresis value
Rev. 0 | Page 86 of 205
Data Sheet
AD9546
REFERENCE DEMODULATOR
reference divider as well as the demodulator input. When the
input reference signal carries embedded clock modulation
events, the demodulator provides a means of detecting them.
On detection of a modulation event on its input clock, the
demodulator synchronizes the associated reference divider and
causes the associated reference TDC to tag the time stamp
generated by the modulation event.
REFERENCE DEMODULATOR OVERVIEW
The AD9546 has eight reference demodulators: one for each
REFx input and one for each auxiliary REFx input (see the
Reference Clock Input Resources section). The REFx receivers
are configurable as single-ended or differential. The auxiliary
REFx receivers are single-ended only (nonconfigurable).
When REFA is configured for differential operation, only the
demodulator associated with REFA is available (the
demodulator associated with REFAA is unused). The same
applies for REFB.
INTERACTION BETWEEN DEMODULATOR AND
REFERENCE MONITOR
The reference monitor validates based on the input carrier
signal. However, the first modulation event after enabling the
reference demodulator causes a resynchronization of the
R divider. The resynchronization of the R divider starts a new
reference monitor validation sequence, which means the
reference invalidates. The user has access to the resynchronization
strobe via an appropriately configured Mx pin.
When enabled, a demodulator recognizes modulated carrier
cycles on its associated reference input clock (a modulated cycle
is a phase deviation imposed on the falling edge of an input
clock cycle). The demodulator is compatible with embedded
clock signals generated by Analog Devices clock chips that
support embedded output clock modulation (the AD9545, for
example). For details on embedded clock modulation, see the
Distribution Embedded Output Clock Modulation section.
The resynchronization of the R divider is a singular event that
happens when the demodulator must force the R divider to
reset on grid (see the Demodulator Synchronization section
regarding the modulation grid). Thus, after enabling the
reference demodulator, to avoid the disruption of reference
validation due to the resynchronization of the R divider, use a
reference validation timer value greater than the period
Figure 62 shows a basic block diagram of the reference
demodulator (legacy mode). The Letter X denotes a reference
demodulator associated with a specific reference input, and M
and N associate a specific register address with Demodulator X
per the register legend in Figure 62.
between modulation events (see the Validation Time (tVALID
)
Referring to Figure 62, the reference input clock signal passes
through the reference receiver and drives the clock input of the
section).
REFERENCE
RECEIVER
REFERENCE
DIVIDER
REFERENCE
INPUT
TIME
TDC
STAMPS
SYNC
TAG
DEMODULATOR X
Mx PIN
FUNCTION
LOGIC
Mx
PINS
DEMOD IN
ENABLE
DEMOD CLK
REGISTER LEGEND
REF INPUT
X
REGISTER
REGISTER M, BIT 3
REGISTER M, BIT 7
REGISTER M, BIT 2
REGISTER M, BIT 6
REGISTER M, BITS [5:4]
REGISTER M, BITS [1:0]
REGISTER N, BIT 0
M
N
AUTOMATIC POLARITY DETECT
EXPECTED POLARITY
PERSISTENCE
REFA
0x0302 0x0301
0x0303 0x0301
0x0306 0x0305
0x0307 0x0305
REFAA
REFB
REFBB
2
2
AUX REF0 0x030A 0x0301
AUX REF1 0x030B 0x0301
AUX REF2 0x030E 0x0305
AUX REF3 0x030F 0x0305
SYNC EDGE
SENSITIVITY
BANDWIDTH
TEXT = BIT(S) IN THE REGISTER MAP
TO THREE OTHER
DEMODULATORS
Figure 62. Reference Demodulator: Legacy Mode
Rev. 0 | Page 87 of 205
AD9546
Data Sheet
the first modulation event. For example, when the first
MODULATED CYCLES AND MODULATION EVENTS
modulation event has a period deviation of +Δt, then the
second modulation event must have a period of –Δt (or vice
versa). Because balanced modulation modulates the first and
second modulated cycles of a modulation event in opposite
fashion, balanced modulation events do not shift the dc balance
of the carrier signal.
The reference demodulator detects the arrival of modulated
carrier cycles. A modulated cycle spans one period of the input
reference clock (tCARRIER) and consists of a predefined phase
deviation relative to the expected falling edge of the input clock.
The definition of a modulated cycle is when the duration of the
second half of a carrier cycle is lengthened or shortened by
some period of time, Δt (see Figure 63). In general, Δt is
constant for a given application and defines the modulation
depth, where the magnitude of the modulation depth is always
less than ½.
The first modulated cycle in a modulation event determines the
polarity of the modulation event. A modulated cycle of –Δt
equates to Logic 0, whereas a modulated cycle of +Δt equates to
Logic 1.
ENABLING A DEMODULATOR
∆t
tCARRIER
Modulation Depth =
To enable a specific demodulator, use Bit 3 of Register M per
the register legend in Figure 62. Logic 0 (default) disables the
demodulator and holds it in a reset state. Logic 1 enables the
demodulator to detect modulation events.
The reference demodulator uses two modulated cycles to
identify a modulation event, although the second modulated
cycle can be unmodulated in the case of unbalanced
modulation (see Figure 63).
MODULATION POLARITY DETECTION
The demodulator performs no type of error detection on the
modulated carrier cycles.
When Bit 7 = 1 of Register M per the register legend in Figure 62,
the demodulator must automatically determine the polarity of
the first modulated cycle of a modulation event (see Figure 63).
However, the autodetect feature reduces the overall sensitivity
of the demodulator.
BALANCED AND UNBALANCED MODULATION
The demodulator recognizes two types of modulation events:
unbalanced and balanced. Figure 63 shows generalized
examples of the modulation event types and their polarity.
The reference demodulator is most sensitive when it is aware of
the modulation polarity in advance. The user can take
advantage of the maximum sensitivity of the demodulator by
programming Bit 2 in Register M per the register legend in
Figure 62 (assuming Bit 7 = 0 in Register M per the register
legend in Figure 62). The value of Bit 2 indicates the expected
polarity of Δt in the first modulated cycle of a modulation event
(see Figure 63). Logic 0 (default) indicates negative Δt, and
Logic 1 indicates positive Δt.
Unbalanced modulation occurs when the second modulated
cycle of a modulation event is unmodulated. Note that
unbalanced modulation only modulates the first modulated
cycle of a modulation event. Therefore, an unbalanced
modulation event shifts the dc offset of the carrier signal
(relative to an unmodulated carrier).
Balanced modulation occurs when the second modulated cycle
of a modulation event has the opposite polarity modulation of
MODULATION EVENT
MODULATED CYCLE
UNMODULATED
CYCLE
DC SHIFT
50%
UNBALANCED
MODULATION
(LOGIC 0)
–Δ
t
UNBALANCED
MODULATION
(LOGIC 1)
50%
DC SHIFT
+Δt
NO DC
SHIFT
BALANCED
MODULATION
(LOGIC 0)
50%
50%
+Δ
t
–Δ
t
NO DC
SHIFT
BALANCED
MODULATION
(LOGIC 1)
tCARRIER
tCARRIER
+Δt
–Δt
2 × tCARRIER
Figure 63. Modulation Events
Rev. 0 | Page 88 of 205
Data Sheet
AD9546
demodulator clock signal only occurs when the demodulator
DEMODULATOR SENSITIVITY
detects a modulation event. When D6 = 1 (default), the
demodulator learns the modulation period (which is tGRID) and
continues to generate demodulator clock pulses at the tGRID rate.
Figure 64 shows the timing of the demodulator clock signal.
Modulated input signals exhibiting a small change in pulse width
require more sensitivity than those that exhibit a large change.
To control the demodulator sensitivity to pulse width changes,
use Bits[1:0] in Register M per the register legend in Figure 62.
The demodulator clock signal is available via an appropriately
configured Mx status pin (see the Status and Control Pins section).
The demodulator defaults to its most sensitive setting, 00 (binary).
Increasing the binary value decreases the sensitivity, with 11
(binary) constituting the least sensitive setting. Decreased
sensitivity implies the need for more robust modulation events
(that is, larger Δt in Figure 63) for reliable demodulation.
Note that if the demodulator loses synchronization (for
example, when incoming modulation events go off grid), the
demodulator requires at least three consecutive tGRID intervals
(with valid modulation events) to resynchronize the persistence
mechanism. Because the demodulator requires three consecutive
DEMODULATOR PERSISTENCE
When the demodulator detects a modulation event, it generates
a single demodulator clock pulse (see Figure 62) lasting one
tGRID intervals (with persistence enabled) for resynchronization,
the occurrence of an occasional off grid modulation event does
not cause the persistence mechanism to realign the demodulator
output clock and resynchronize the R divider. Instead, the
persistence mechanism effectively filters out the occasional
occurrence of off grid modulation events.
tCARRIER period. Because demodulation events normally occur at
intervals of tGRID, the demodulator clock signal consists of single
pulses occurring at the tGRID rate. The demodulator clock pulse
synchronizes the R divider and causes the reference TDC to tag
the accompanying time stamp. Thus, tagged reference time stamps
occur at the tGRID rate (see the Demodulator Synchronization
section).
DEMODULATOR BANDWIDTH
The demodulators operate within two overlapping frequency
bands: Band 0 and Band 1. The user must select the appropriate
demodulator bandwidth (based on the input carrier frequency)
via Bit 0 in Register N per the register legend in Figure 62.
Logic 0 selects Band 0, and Logic 1 (default) selects Band 1.
Note that if the incoming modulation events stop occurring,
the demodulator clock pulses cease and the TDC stops
generating tags. Assuming the TDC tags route to one of the
DPLLs, the absence of tags causes the DPLL to unlock. The
persistence feature provides a mechanism for the demodulator
to continue to generate demodulator clock pulses at the tGRID
rate when modulation events are no longer present at the input,
thereby preventing the DPLL from losing lock.
Note that demodulator bandwidth control applies to a group of
four demodulators rather than an individual demodulator.
Thus, the user must ensure that all demodulators of the same
group that are actively demodulating have carrier frequencies
within the specified frequency range.
The user controls the persistence functionality via Bit 6 in
Register M per the register legend in Figure 62. When Bit 6 = 0, the
tGRID
tGRID
BASE
EDGE
BASE
EDGE
MODULATION EVENTS OCCUR
AT INTERVALS OF tGRID
tCARRIER
SYNC EDGE
SYNC EDGE
MODULATION
EVENT
MODULATION
EVENT
MODULATED
CARRIER
tGRID
tGRID
tGRID
DEMOD
CLK TIMING
TIME
FOR THIS EXAMPLE: tGRID = 10 × tCARRIER
Figure 64. Demodulator Synchronization
Rev. 0 | Page 89 of 205
AD9546
Data Sheet
tCARRIER
(8 + x) × tCARRIER
K × (8 + x) × tCARRIER
tGRID
Figure 65. Grid Timing Diagram
creates a fixed timing relationship between incoming modula-
tion events and the occurrence of tagged time stamps. Specifically,
every Kth time stamp (see Figure 65) is a tagged time stamp.
DEMODULATOR SYNCHRONIZATION
The demodulators expect modulation events to occur at regular
periodic intervals, tGRID, with tGRID comprising an integer
number of tCARRIER periods (see Figure 65) such that
The demodulator denotes the next rising edge of the carrier
following the modulation event as the base edge. The user can set
the timing of the demodulator clock pulse to occur a predefined
number of tCARRIER periods after the base edge via the 2-bit
unsigned bit field comprising Bits[5:4] in Register M per the
register legend in Figure 62. The decimal value of this bit field,
0 (default), 1, 2, or 3, corresponds to the desired synchronization
edge in Figure 64.
t
GRID = K × (8 + x) × tCARRIER
where:
K is an integer K ≥ 1.
x is an integer with x ≥ 0.
Essentially, the value of x relates to the divide ratio, R, of the
reference divider as follows:
The user must program a bit field value other than the default
value of zero (Bits[5:4] = 0 is not supported).
R = 8 + x
The net effect is that the reference divider produces an output
clock of period, tCARRIER × (8 + x). The reference TDC time
stamps the rising edge of each divider output clock cycle with K
time stamps occurring in each tGRID period.
If a modulation event occurs off grid or the modulation rate
changes to a new tGRID period, the demodulator must
resynchronize. The demodulator requires at least three tCARRIER
periods (including the first modulated cycle) to resynchronize
when Bit 6 in Register M per the register legend in Figure 62 is
Logic 0 and at least eight tCARRIER periods when Bit 6 is Logic 1.
When the demodulator identifies a modulation event within
the carrier signal, it generates a demodulator clock pulse with
timing according to Figure 64. The demodulator clock pulse
synchronizes the reference divider and causes the reference
TDC to tag the associated time stamp (see Figure 62), which
Rev. 0 | Page 90 of 205
Data Sheet
AD9546
DISTRIBUTION CLOCK OUTPUT DRIVERS
DISTRIBUTION CLOCK OUTPUT DRIVERS
OVERVIEW
Table 58. Output Driver Current
Bits[2:1] Value (Decimal)
Output Current (mA)
The AD9546 has 10 clock output pins capable of generating up
to 10 output clock signals. The 10 pins comprise two groups of
clocks differentiated by the PLL channel available as the clock
source driving the group. PLL0 can clock the Output 0 group,
and PLL1 can clock the Output 1 group.
0
1
2
3
7.5 (default)
12.5
15
Not applicable
OUTPUT MODE CONTROL
The Output 0 group comprises six of the 10 clock output pins
with the six pins consisting of three pin pairs. Each pin pair
represents an output subgroup (A, B, and C) with the pins
designated as
Each output pin pair has two channel dividers and two output
drivers. Figure 66 shows the connectivity between the dividers
and drivers. The divider to driver connectivity depends on the
selected output mode.
•
•
•
OUT0AP and OUT0AN
OUT0BP and OUT0BN
OUT0CP and OUT0CN
SINGLE-DIVIDER, DIFFERENTIAL DRIVERS
Qxy
DRIVER
DRIVER
OUTxyP
OUTxyN
SOURCE
SINK
FROM
PLL0
OR
+1.8V
The Output 1 group comprises the remaining four output pins
consisting of two pin pairs. Each pin pair represents an output
subgroup (A and B) with the pins designated as
PLL1
Qxyy
•
•
OUT1AP and OUT1AN
OUT1BP and OUT1BN
SINGLE-DIVIDER, SINGLE-ENDED DRIVERS
Qxy
DRIVER
OUTxyP
OUTxyN
Each output driver has three programmable parameters
controlling
SOURCE
SINK
FROM
PLL0
OR
+1.8V
PLL1
•
•
•
Output current direction (source or sink)
Output current magnitude
Driver configuration
Qxyy
DRIVER
DUAL-DIVIDER, SINGLE-ENDED DRIVERS
The three control parameters for a given output are bits that
reside in the same register with a register address specific to a
specific output per Table 57.
Qxy
DRIVER
OUTxyP
OUTxyN
SOURCE
SINK
FROM
PLL0
OR
+1.8V
PLL1
Table 57. Output Driver Control Register Address
Qxyy
DRIVER
Output
OUT0A
OUT0B
OUT0C
OUT1A
OUT1B
Register Address
0x10D7
0x10D8
0x10D9
0x14D7
Figure 66. Output Driver Modes
The user controls the connectivity of the dividers and drivers of
each subgroup pin pair via Bits[4:3] of the appropriate register
address (see Table 57). The value of the two bits selects the
divider to driver relationship per Table 59.
0x14D8
OUTPUT CURRENT CONTROL
Table 59. Driver Output Mode
Each pin associated with a subgroup pin pair (OUTxyP/OUTxyN,
where x is 0 or 1 and y is A, B, or C) has an output driver and a
channel divider associated with it. Both drivers of an output pin
pair act as either current sources or current sinks. The user
programs the source or sink option via Bit 0 of the appropriate
register address in Table 57. Logic 0 (default) sets both drivers
to current sink mode, and Logic 1 sets both drivers to current
source (or HCSL) mode.
Bits[4:3] Value
(Decimal)
Output Driver Mode
0
1
2
3
Single-divider, differential (default)
Single-divider, single-ended
Dual-divider, single-ended
Not applicable
The single divider modes use only the Qxy divider (with the
Qxyy divider powered down). The dual divider mode makes
use of both the Qxy and Qxyy dividers.
To control the magnitude of the current, use Bits[2:1] of the
appropriate register address in Table 57. Table 58 relates the
value of the two bits to the magnitude of the driver current.
Rev. 0 | Page 91 of 205
AD9546
Data Sheet
In single-divider differential mode, the clock signal at the
OUTxyN pin phase is inverted relative to the OUTxyP pin. In
single-divider single-ended mode, the clock signals at the
OUTxyP and OUTxyN pins are in phase relative to one
another. In dual-divider single-ended mode, the Qxy divider
drives the OUTxyP pin and the Qxyy divider drives the
OUTxyN pin. In this mode, each output of the subgroup can
have a different frequency and phase offset.
a 50 Ω pull-up resistor to 1.42 V on each output pin. This
configuration requires programming the driver settings as
follows: single-divider, differential drivers; current sink; and
driver current = 7.5 mA. This arrangement yields a 375 mV
swing across the output pins with 1.23 V common mode.
When using an additional 100 Ω differential termination (RL) at
the receiver, however, ac coupling is necessary (as shown in
Figure 38), as well as programming a drive current of 15 mA.
This arrangement yields 375 mV swing across the output pins
with 1.05 V common mode (thus the ac coupling requirement
to preserve the 1.2 V common-mode bias of the LVDS receiver).
OUTPUT DRIVER CONFIGURATIONS
Each driver receives a clock input signal from its associated
channel divider (see Figure 66). The current source (or sink) of
the driver is either on or off, depending on the logic level
output by the divider. That is, when the divider output is
Logic 0, the driver output current is 0 mA. Conversely, when
the divider output is Logic 1, the drive output current is the
programmed output current value. The Qxy divider has normal
and inverted logic outputs, which means when the Qxy divider
connects to both drivers and one current source is on, the other
is off, allowing a differential output signal.
To drive an LVDS-compatible receiver that can handle boosted
signal swing, use a Thevenin equivalent termination per Figure 43,
which exhibits the equivalent of a 50 Ω pull-up resistor to
1.60 V on each output pin. This configuration requires
programming the driver settings as follows: single-divider,
differential drivers; current sink; and driver current = 15 mA.
This arrangement yields 750 mV swing across the output pins
with 1.23 V common mode.
CML Output
Differential HCSL Output
To configure an output for CML signals, see Figure 40. When
using a 50 Ω pull-up resistor to 1.2 V, this configuration
requires programming the driver settings as follows: single-
divider, differential drivers; current sink; and driver current =
7.5 mA. This arrangement yields a 375 mV swing across the
output pins with 1.01 V common mode.
To drive a standard HCSL receiver, refer to Figure 39. This
configuration requires programming the driver settings as
follows: single-divider, differential drivers; current source;
driver current = 15 mA. In this mode, the driver expects a 50 Ω
termination to ground on each driver output pin. The 15 mA
drive current yields 750 mV swing across each 50 Ω load.
When using a 50 Ω pull-up resistor to 1.5 V or 1.8 V, program
a drive current of 15 mA. This arrangement yields a 750 mV
swing across the output pins with 1.125 V common mode for a
1.5 V supply and 1.425 V common mode for a 1.8 V supply.
LVDS Output
To drive an LVDS receiver using ac coupling, refer to Figure 38.
This configuration requires programming the driver settings as
follows: single-divider, differential drivers; current source; and
driver current = 7.5 mA. In this mode, the driver expects a 50 Ω
termination to ground on each driver output pin and dc
blocking capacitors with a 100 Ω differential termination at the
receiver. The expected termination arrangement yields 188 mV
swing across the 100 Ω load but with 188 mV common mode
across the output pins, and thus, the need for dc blocking
capacitors to preserve the common-mode bias of 1.2 V of the
receiver.
Dual, Single-Ended, In Phase Outputs
To configure the output to produce the same signal (in phase)
on each pin of an OUTxyP/OUTxyN pin pair, refer to Figure 44.
This configuration requires programming the driver settings as
follows: single-divider, single-ended drivers and current source.
Select the driver current to yield an acceptable voltage swing
based on the load resistance, RL. The output is essentially a
current source (on or off per the logic state of the Qxy divider
output) with an external pull-down resistor. Thus, the output
signal swing is between ground and V = I × RL.
To drive an LVDS receiver directly with dc coupling, use a
T network, as shown in Figure 41. This configuration requires
programming the driver settings as follows: single-divider,
differential drivers; current sink; and driver current = 7.5 mA.
This arrangement yields a 375 mV swing across the output pins
with 1.24 V common mode. When using an additional 100 Ω
differential termination (RL) at the receiver, however, ac coupling
is necessary (as shown in Figure 38). The additional termination
also requires programming a drive current of 15 mA. This
arrangement yields a 375 mV swing across the output pins with
0.67 V common mode (thus, the ac coupling requirement to
preserve the 1.2 V common-mode bias of the LVDS receiver).
Independent, Single-Ended Outputs
To configure the output to produce independent signals on the
two output pins of an OUTxyP/OUTxyN pin pair, refer to
Figure 46. This configuration requires programming the driver
settings as follows: dual-divider, single-ended drivers and
current source. Select the driver current to yield an acceptable
voltage swing based on the load resistance, RL. Note that the
output is essentially a current source (on or off per the logic
state of the Qxy and Qxyy dividers) with an external pull-down
resistor. Thus, the output signal swing is between ground and V =
I × RL.
To drive an LVDS receiver using a Thevenin equivalent
termination, refer to Figure 42, which exhibits the equivalent of
Rev. 0 | Page 92 of 205
Data Sheet
AD9546
Table 61. Autotmatic Unmute Conditions
OUTPUT DRIVER RESET
Bits[1:0]
(Decimal)
Because driver pairs (OUT0AP and OUT0AN, for example) are
configurable as differential or single-ended, when the Q dividers
associated with a driver pair stop toggling (due to a divider
reset, for example), the output state of the driver pair is unknown.
To remedy this uncertainty, the user can force a driver pair to a
known state by means of the various driver reset bits.
Automatic Unmute Condition
Immediate (default)
0
1
2
3
Upon activation of a hitless profile
Upon phase lock (hitless mode only)
Upon frequency lock (hitless mode only)
When Bits[1:0] = 0 (decimal), the drivers unmute immediately
upon release of the synchronization request. When Bits[1:0] = 1
(decimal), the drivers do not unmute until release of the
synchronization request and subsequent activation of a hitless/zero
delay profile. When Bits[1:0] = 2 (decimal), the drivers do not
unmute until release of the synchronization request and
subsequent phase lock of the appropriate PLL channel (assu-
ming a hitless/zero delay profile). When Bits[1:0] = 3 (decimal),
the drivers do not unmute until release of the synchronization
request and subsequent frequency lock of the appropriate PLL
channel (assuming a hitless/zero delay profile).
Bit 2 of Register 0x2101 allows the user to reset all the driver
pairs of Channel 0. Bit 2 of Register 0x2201 allows the user to
reset all the driver pairs of Channel 1. The user can reset individual
output driver pairs via Bit 5 per the appropriate register
addresses in Table 60.
Table 60. Output Driver Reset Register Address
Output
Register Address
OUT0AP/OUT0AN
OUT0BP/OUT0BN
OUT0CP/OUT0CN
OUT1AP/OUT1AN
OUT1BP/OUT1BN
0x2102
0x2103
0x2104
0x2202
The user can also selectively opt out of the automatic unmuting
feature on a per output basis. For PLL0, use Bits[7:2] of
Register 0x10DC, where Bits[7:2] correspond to the driver
associated with OUT0CN, OUT0CP, OUT0BN, OUT0BP,
OUT0AN, and OUT0AP, respectively. For PLL1, use Bits[5:2]
of Register 0x14DC where Bits[5:2] correspond to the driver
associated with OUT1BN, OUT1BP, OUT1AN, and OUT1AP,
respectively.
0x2203
When a driver pair is in a differential configuration, resetting
the driver pair forces the OUTxyP driver to a Logic 0 state and
the OUTxyN driver to a Logic 1 state. When a driver pair is in a
single-ended configuration, resetting the driver pair forces the
OUTxyP and OUTxyN drivers to a Logic 0 state.
OUTPUT MUTING
After an automatic mute event occurs for a particular PLL
channel, the user can clear the automatic mute state by setting
Bit 4 of Register 0x2107 for PLL0 and Register 0x2207 for PLL1.
However, this action is generally unnecessary, because the
device automatically clears the automatic mute state whenever
the user changes the state of an automatic unmute bit and issues
an IO update operation.
Manual Output Muting
To mute all the drivers associated with PLL0 or PLL1, use Bit 1
in Register 0x2101 and Register 0x2201 for PLL0 and PLL1,
respectively. To mute individual drivers, use Bits[3:2] per the
appropriate register addresses in Table 60, where Bit 3
corresponds to the OUTxyN driver and Bit 2 corresponds to
the OUTxyP driver.
Output Mute Retiming
Automatic Output Unmuting
If the driver output mutes immediately upon receiving a mute
command, it may result in the output signal shutting off before
the normal falling edge of the associated Q divider output.
Early shutoff of the output signal produces a runt pulse (that is,
the normal Logic 1 period of the Q divider is cut short).
Automatic unmuting works in conjunction with the
autosynchronization function (see the Autosynchronization
Trigger section). Part of the synchronization sequence involves
synchronously unmuting the output drivers. The AD9546
provides the user options for the timing of the unmuting of the
output drivers.
By default, the AD9546 uses a retiming block to prevent the
generation of a runt pulse. However, the user can bypass the
retiming block via Bit 5 per the appropriate register addresses
in Table 57. Bypassing the retiming block causes the output
driver to mute immediately upon receiving a mute command.
To select the automatic unmute conditions, use Bits[1:0]
of Register 0x10DC for PLL0 and Register 0x14DC for PLL1.
Table 61 shows the automatic unmute conditions invoked by
these bits. The automatic unmute control applies to PLL0 and
PLL1 independently.
Rev. 0 | Page 93 of 205
AD9546
Data Sheet
DISTRIBUTION DIVIDERS (Q DIVIDERS)
A JESD204B and burst controller enables each Q divider to
DISTRIBUTION DIVIDERS OVERVIEW
generate a clock signal supporting JESD204B or to generate
flexible burst patterns supporting gapped clock and similar
applications (see the Distribution N Shot/PRBS Output
Clocking section).
Five pairs of Q dividers correspond to the five pairs of clock
output drivers (see the Distribution Clock Output Drivers
section). Unlike typical clock dividers, which count the rising
edge (or falling edge) of its input clock, the Q dividers count
both the rising and falling edges of its input clock. Dual-edge
counting enables features not possible with a typical counter.
Furthermore, the Qxy divider of each Q divider pair is capable
of modulating (that is, time shifting) falling edges of the output
clock signal. The modulation of the clock edge location
supports embedded clock applications such as transporting a
lower frequency clock signal over a higher frequency carrier
clock (see the Distribution Embedded Output Clock
Modulation section).
The Q dividers respond to several controllers that implement
the various features of the Q dividers. Figure 67 shows a
diagram of the Q dividers and associated controllers. Although
not explicitly shown, the output of each Q divider routes to its
corresponding output driver.
At power-up, the synchronization controller holds the disabled
Q dividers (that is, there are no output clocks). Consequently,
the user must issue a synchronization request (see the
Distribution Output Clock Synchronization section) following
a power-up or reset to initialize the Q dividers.
Each Q divider has independent phase offset (delay) capability
as governed by a phase offset controller (see the Distribution
Phase Offset Control section). In addition, a sophisticated and
flexible synchronization mechanism provides synchronized
output clock timing for any or all Q divider output signals (see
the Distribution Output Clock Synchronization section).
TO Q DIVIDER PAIR 2 TO PAIR 5
4
8
8
8
Q DIVIDER PAIR 5
Q DIVIDER PAIR 1
TO
OUTPUT
DRIVERS
5
MODULATION
CONTROLLER
MODULATION
N SHOT/PRBS
PHASE OFFSET
SYNC
Qxy
10
10
10
2
SYNCHRONIZATION
CONTROLLER
SYSTEM
CLOCK
2
2
PHASE OFFSET
CONTROLLER
PLL0/PLL1
P DIVIDER
SYNC
N SHOT/PRBS
CONTROLLER
Qxy
PHASE OFFSET
N SHOT/PRBS
Figure 67. Functional Block Diagram of Q Dividers and Controllers
Rev. 0 | Page 94 of 205
Data Sheet
AD9546
In some applications, when more than one Q divider pair uses
the system clock as its clock source, it may be desirable to
synchronize at least two of those Q divider pairs using the
system clock synchronization mask bit of Register 0x10DA and
Register 0x14DA. To do so, initiate a synchronization sequence
(see the Distribution Output Clock Synchronization section),
then program the corresponding mask system clock
synchronization bit to Logic 1. To reconfigure the affected Q
dividers by changing the divide ratio, for example, the user
must first program the corresponding system clock
Q DIVIDER CLOCK SOURCE SELECTION
Each pair of Q dividers shares a common clock input (see
Figure 67) that originate from one of two sources.
•
From the output of the associated APLL channel of the
Q divider.
•
From the output of the system clock PLL VCO
(~2.3 GHz).
The first source is the normal (default) operating mode, which
provides frequency translation from a reference input to the
distribution output via the corresponding PLL channel. The
second source is the output frequency of the system clock PLL,
which is useful for clocking certain peripheral devices (a
microprocessor, for example). In this scenario, the user can
employ an optional external electronically erasable programmable
read only memory (EEPROM) to download a device
configuration that provides an output clock signal at power up
via the system clock source path (see the EEPROM Usage
section).
synchronization mask bit to Logic 0.
INTEGER DIVISION
The Q divider divide ratio (Q) depends on a 32-bit unsigned
integer in the register ranges shown in Table 63. To program
the divide ratio, write the value of Q into the corresponding
address range and then set the IO update bit.
The range of Q is 1 to 4,294,967,295.
For example, to establish a divide ratio of Q = 1,000,000, the
required 32-bit value is 1,000,000 (0x 000F 4240 hexadecimal).
Table 62 shows the register address and bits associated with
selecting the clock source for the Q dividers. Logic 0 (default)
selects the output of the corresponding APLL as the Q divider
clock source, whereas Logic 1 selects the output of the system
clock PLL as the Q divider clock source.
Table 63. Q Divider Divide Ratio Address Ranges
Q Divider
Register Address
0x1100 to 0x1103
0x1109 to 0x110C
0x1112 to 0x1115
0x111B to 0x111E
0x1124 to 0x1127
0x112D to 0x1130
0x1500 to 0x1503
0x1509 to 0x150C
0x1512 to 0x1515
0x151B to 0x151E
Q0A
Q0AA
Q0B
Q0BB
Q0C
Q0CC
Q1A
Q1AA
Q1B
Q1BB
Table 62. Q Divider Clock Source Selection
Q Divider
Register Address
Bit
D1
D2
D3
D1
D2
Q0A and Q0AA
Q0B and Q0BB
Q0C and Q0CC
Q1A and Q1AA
Q1B and Q1BB
0x10DA
0x10DA
0x10DA
0x14DA
0x14DA
When the AD9546 is configured with the system clock as the
clock source for a Q divider pair, use Bit 0 (system clock
synchronization mask) of Register 0x10DA and Register 0x14DA
to prevent those particular dividers from being resynchronized
every time a synchronization trigger occurs (see the Distribution
Output Clock Synchronization section). When this bit is set,
any Q divider associated with a specific PLL channel and with
the system clock as the clock source is immune to synchronization
events. This feature is particularly useful for outputs that serve
as the clock source to a microprocessor, for example, because
an unmasked synchronization event disrupts the clock signal of
the processor.
A value of 0 (default) sets a divide ratio of 1, which is an
unsupported divide ratio. Therefore, at power-up or after
resetting the device, the user must program the Qxy divider
ratio bit fields with values greater than 1 to enable the
corresponding Q dividers.
Although it is possible to program a different divide ratio for
each Q divider of a Q divider pair, the recommendation is to
avoid doing so because a Q divider pair with differing divide
ratios results in two different output frequencies and can cause
cross coupling issues (due to the close physical proximity of the
corresponding output drivers). The intent is for Q divider pairs
to operate with the same output frequency but with the option
to control the relative phase between the outputs (see the
Distribution Phase Offset Control section).
Rev. 0 | Page 95 of 205
AD9546
Data Sheet
HALF INTEGER DIVISION
Q DIVIDER CONSTRAINTS
The Q dividers are unique in that they support half integer
division, a benefit of the Q dividers counting both the rising
and falling edges of the input clock. To enable half integer
division for the corresponding Q divider, use Bit 5 (half integer
divide) of the appropriate register per Table 64.
Changing the divide ratio of any one of the Q dividers invokes a
synchronization process (see the Autoreconfiguration
Synchronization Trigger section). The act of changing the
divide ratio of a single Q divider causes a disturbance of all the
output clock signals of the associated PLL channel. Therefore, it
is not possible to change the divide ratio of one Q divider
without disturbing all the output clock signals of the associated
PLL channel.
Table 64. Q Divider Half Integer Divide Register Address
Q Divider
Register Address
Q0A
0x1108
When the system clock is the selected clock source for a
Q divider pair (see the Distribution Output Clock Synchronization
section), an unlock event on the system clock may cause the
Q divider pair to enter an invalid state. Recovery from this state
requires a divider reset. Therefore, after a system clock unlock
event, the user must follow these steps:
Q0AA
Q0B
Q0BB
Q0C
Q0CC
Q1A
0x1111
0x111A
0x1123
0x112C
0x1135
0x1508
Q1AA
Q1B
Q1BB
0x1511
0x151A
0x1523
1. Calibrate the system clock (see the System Clock
Calibration section).
2. Calibrate APLL0 and APLL1 (see the VCO Calibration
section).
3. Reset any Q dividers (see the Q Divider Reset section)
using the system clock as an input clock source.
4. Synchronize output distribution (see the Distribution
Output Clock Synchronization section).
When the half integer divide bit is Logic 1, it effectively adds an
extra 0.5 to the divide ratio prescribed by the corresponding
32-bit Q divider ratio. The value of the half integer divide bit
does not become effective until the user sets the IO update bit.
For example, assume the Q divider ratio is 100. If the associated
enable half integer divide bit is Logic 1, the total divide ratio is
100.5.
When using embedded output clock modulation to modulate
the Q divider output (see the Distribution Embedded Output
Clock Modulation section), a minimum Q divider divide ratio
is necessary.
Q DIVIDER RESET
Each Q divider has a dedicated Q divider reset bit via Bits[1:0]
of the registers shown in Table 65, where Bit 0 applies to the
first Q divider in the row (for example, Q0A) and Bit 1 applies
to the second Q divider in the row (for example, Q0AA).
Qxy ≥ 8
where:
x = 0 or 1.
y = A, B, or C.
The Q divider reset bits allow the user to reset the Q dividers
manually in applications that do not require output clock
synchronization. In most cases, however, there is no need to set
these bits during normal operation because the synchronization
controller automatically handles the Q divider reset function
(see the Distribution Output Clock Synchronization section).
When employing the N shot triggering mechanism to trigger a
Q divider (see the N Shot Triggering section), a minimum
Q divider divide ratio of 8 is necessary.
Qxy ≥ 8
where:
x = 0 or 1.
y = A, B, or C.
Table 65. Output Driver Reset Register Address
Q Divider
Register Address
Q0A and Q0AA
Q0B and Q0BB
Q0C and Q0CC
Q1A and Q1AA
Q1B and Q1BB
0x2102
0x2103
0x2104
0x2202
HITLESS/ZERO DELAY FEEDBACK
When configured for hitless or zero delay operation, the DPLL
of the affected PLL channel requires a feedback clock signal to
its N divider from one of the distribution outputs. This
feedback mechanism is via the hitless or zero delay feedback
and synchronization block of the N shot/PRBS controller in
Figure 68. Selection of the desired feedback path is via the
frequency translation profiles. See the Internal Zero Delay
(Hitless) Mode section in the Frequency Translation Loops
section for programming details.
0x2203
Rev. 0 | Page 96 of 205
Data Sheet
AD9546
DISTRIBUTION PHASE OFFSET CONTROL
Divide Ratio is the value of the 32-bit integer stored in the
OUTPUT PHASE OFFSET OVERVIEW
corresponding Q divider register in Table 63.
Half Integer Divide is the value of the corresponding Q divider
half integer divide bit (0 or 1).
QN is the complete divide factor (for example, 101.5) of a
specific Q divider.
The phase offset controller (see Figure 67) governs the
application of phase offsets to the individual Q dividers. The
controller implements two categories of phase offset: initial
phase offset and subsequent phase offset.
An initial phase offset applies after the device is powered up or
reset and the user issues a synchronization request that is
completed (see the Distribution Output Clock Synchronization
section). The completed synchronization request results in the
establishment of the initial phase offset. Subsequent phase
offsets apply for all subsequent synchronization requests (that
is, any synchronization requests occurring after the initial
power-up or reset synchronization request). See the
Distribution Output Clock Synchronization section regarding
synchronization requests.
Thus, for QN = 101.5, E = 203.
The value of E relates to the phase offset angle, θ, as follows:
θ = 360° × (Phase Offset/E)
where Phase Offset is the integer stored in the phase offset
registers (the phase offset register address ranges appear in
Table 66).
The phase offset angle equation implies that the programmed
phase offset value must be less than E (that is, the phase offset
value has a range of 0 to E − 1). Note that programming an
invalid phase value results in the phase offset controller taking
no action other than setting an error status in Bits[5:0] of
Register 0x310E for PLL0 and Bits[3:0] of Register 0x320E for
PLL1. For Register 0x310E, Bits[5:0] correspond to Q0CC,
Q0C, Q0BB, Q0B, Q0AA and Q0A, respectively. For Register
0x320E, Bits[3:0] correspond to Q1BB, Q1B, Q1AA, and Q1A,
respectively. The user can access the phase control error results
via an appropriately configured Mx status pin (see the Status
and Control Pins section). The Mx pin output signal constitutes
a logical OR of the six status bits associated with PLL0 or the
logical OR of the four status bits associated with PLL1.
Initial and subsequent phase offsets require the Q divider pulse
width control feature to be turned off (see the Output Pulse
Width Control section).
The initial phase offset depends on a 33-bit unsigned integer in
the register ranges shown in Table 66. The four lower addresses
for each Q divider in Table 66 carry the 32 LSBs of the 33-bit
integer, and Bit 6 of the upper address carries the MSB of the
33-bit integer (Bit 7 is unused and Bits[5:0] apply to other Q
divider features).
Table 66. Q Divider Phase Offset Control Address Ranges
Q Divider
Register Address
0x1104 to 0x1108
0x110D to 0x1111
0x1116 to 0x111A
0x111F to 0x1123
0x1128 to 0x112C
0x1131 to 0x1135
0x1504 to 0x1508
0x150D to 0x1511
0x1516 to 0x151A
0x151F to 0x1523
SUBSEQUENT PHASE OFFSETS
Q0A
Q0AA
Q0B
Q0BB
Q0C
Q0CC
Q1A
Q1AA
Q1B
Q1BB
Subsequent phase offsets involve writing a new 33-bit phase
offset value to the appropriate Q divider per Table 66 and then
setting the IO update bit. The magnitude of the applied phase
offset is the same as described in the Initial Phase Offset
section.
Unlike the initial phase offset, the phase controller implements
subsequent phase offsets in stepwise fashion as a sequence of
phase steps, where the 33-bit phase offset value denotes the
amount of phase offset present at the termination of the sequence.
The controller executes the phase steps at a rate commensurate
with the Q divider output period. This mechanism makes it
possible to soften the phase transient that results when applying
subsequent phase offsets. The reason is the stepwise
implementation of the phase offset effectively limits phase
transients to some maximum amount per output cycle of the
Q divider, which effectively constitutes a phase rate of change
limiter. The limited rate of change of the phase replaces the
relatively large instantaneous phase step of a phase adjustment
with a series of small phase steps, thereby reducing the spectral
content normally associated with phase adjustment.
INITIAL PHASE OFFSET
The initial phase offset of a Q divider is dependent on the
divide ratio and state of the half integer divide bit, which
together constitute the number of rising and falling input clock
edges that span one period of the Q divider output. For details
on the divide ratio and half integer divide, see the Distribution
Dividers (Q Dividers) section.
In terms of the values defining the divide ratio of the Q divider,
E = (2 × Divide Ratio) + Half Integer Divide
= 2 × QN
where:
E is the total number of input edges per output period of the Q
divider.
Rev. 0 | Page 97 of 205
AD9546
Data Sheet
When the Q divider is in the process of phase slewing as a part
of the phase offset sequence, the phase offset controller indicates
phase slew active status via Bits[5:0] of Register 0x310D and
Bits[3:0] of Register 0x320D. For Register 0x310D, Bits[5:0]
correspond to Q0CC, Q0C, Q0BB, Q0B, Q0AA, and Q0A,
respectively. For Register 0x320D, Bits[3:0] correspond to
Q1BB, Q1B, Q1AA, and Q1A, respectively.
Table 67. Maximum Phase Slew Step Size
Bits[2:0]
(Decimal)
Maximum Phase
Step Size (Degrees) Comment
0
360 ÷ E
One input half-cycle
Two input half-cycles
~11.25°
~22.5°
~45°
~90°
~180°
~360°
1
180 ÷ E
2
3
4
5
360 × floor(E/32)/E
360 × floor(E/16)/E
360 × floor(E/8)/E
360 × floor(E/4)/E
360 × floor(E/2)/E
360 × (E − 1)/E
The user can access the phase slew active results via an
appropriately configured Mx status pin (see the Status and
Control Pins section). The Mx pin output signal constitutes a
logical OR of the six status bits associated with PLL0 or the
logical OR of the four status bits associated with PLL1.
6
7 (Default)
Phase Slew Mode
During the execution of a subsequent phase offset, the phase
controller can slew phase in two different modes. The user selects
the mode via Bit 3 per the same address given in Table 64.
Maximum Phase Slew Step Size
During the execution of a subsequent phase offset, the user
controls the maximum step size of the phase steps for a
particular Q divider via Bits[2:0] per the same address given in
Table 64. The value stored in Bits[2:0] establishes the maximum
phase step size per Table 67 (in which E is the number of Q
divider input edges per output period (see the Initial Phase
Offset section) and floor(x) means to round x to the nearest
integer in the direction of −∞). Keep in mind that Table 67
defines a maximum phase step size during phase slewing, but
the controller may use smaller values to ensure the terminal
phase value does not exceed the programmed phase offset
value.
When Bit 3 = 0 (default), the phase controller slews phase in the
direction that reduces the output frequency during the stepwise
phase adjustment sequence. Alternatively, when Bit 3 = 1, the
phase controller slews phase in the direction requiring the
fewest number of steps. However, when the half integer divide
bit is Logic 1 and the desired phase shift is within one Q divider
input half-cycle of the midpoint of an output cycle, the phase
controller may choose the slightly longer direction.
When Bit 3 = 1, the output frequency may change in either
direction, as required, for any given stepwise phase adjustment
sequence (due to the phase controller choosing the direction of
the fewest number of steps).
When Bits[2:0] = 7 decimal, the phase controller requires only
one step to reach the desired phase offset regardless of the size of
the phase offset. Therefore, the bit setting effectively deactivates
the phase slewing feature. When Bits[2:0] = 7, the phase
controller behaves as though the phase slew mode bit is Logic 0
(that is, lag only) even if the phase slew mode bit is Logic 1 (see
the Phase Slew Mode section).
Output Pulse Width Control
By default, the Q divider output generates a clock signal with a
50% pulse width. However, the user has the option to control
the pulse width of the Q divider output. The control granularity
is one half cycle of the input clock of the Q divider. Pulse width
control is via Bit 4 per the same address given in Table 64.
When Bits[2:0] = 0 or 1 decimal, the slewing operation is
limited to half-cycles or full cycles, respectively, of the input
clock. Because a 0 or 1 setting represents the smallest possible
maximum phase step for most Q divider divide ratios, most
choices of phase offset result in phase slewing as though the
maximum phase slew step was set to 0 or 1.
When Bit 4 = 0, the phase offset controller uses the
programmed 33-bit phase offset as a phase offset (as described
in the Initial Phase Offset section). When Bit 4 = 1, the 33-bit
phase offset value defines a pulse width rather than a phase
offset.
Maximum phase slew step values greater than 1 decimal require
the user to ensure that the maximum phase step size is greater
than or equal to one input half-cycle. Otherwise, the phase slewing
function is invalid, in which case the device flags an error via the
control error bits described in the Initial Phase Offset section.
For example, when QN = 5 (that is, E = 10 (see the Initial Phase
Offset section)), one half-cycle constitutes 36°. As such,
maximum phase slew step values of 2 or 3 are invalid, and the
controller sets the appropriate error flag.
The pulse width, in percent, is given by
Pulse Width (%) = 100 × Phase Offset/E
where:
Pulse Width defines the portion of an output clock cycle from
the rising edge to the falling edge.
Phase Offset is the integer stored in the phase offset registers.
Table 66 shows the phase offset register address ranges.
E is as defined in the Initial Phase Offset section.
(2)
Rev. 0 | Page 98 of 205
Data Sheet
AD9546
For example, determine the value necessary to yield a 14% pulse
width given E = 7301. Solving Equation 2 for Phase Offset yields
the following:
The user has the option to adjust the output pulse width only
prior to execution of the first subsequent phase offset. After
invoking a subsequent phase offset, pulse width control is no
longer available. To change the output pulse width after
executing the first subsequent phase offset, the user must issue
a Q divider synchronization (see the Distribution Output Clock
Synchronization section), at which point the Q divider is ready to
accept a new initial phase offset, a new output pulse width, and
subsequent phase offsets.
Phase Offset = Pulse Width × E/100
= 14 × 7301/100
= 1022 (nearest integer)
= 0x 0 0000 03FE (hexadecimal)
Rev. 0 | Page 99 of 205
AD9546
Data Sheet
DISTRIBUTION N SHOT/PRBS OUTPUT CLOCKING
Regarding the Q dividers, Qxy denotes a specific Q divider in
the clock distribution section, whereas Qxy denotes the output
of divider Qxy (where x is 0 and y is A, AA, B, BB, C, or CC).
For example, Q0A is the output of Q divider Q0A.
N SHOT/PRBS CLOCKING OVERVIEW
The Q dividers work in conjunction with several controllers, one
of which is the N shot/pseudorandom binary sequence (PRBS)
controller (see Figure 67). The N shot/PRBS controller provides
the AD9546 with the capability of generating SYSREF and
device clock clocks per the JESD204B standard as well as gapped
and pseudorandom clock signals.
Note that the Q0AA, Q0BB, and Q0CC outputs route directly to the
N shot/PRBS controller, whereas Q0A, Q0B, and Q0C route
through the modulation controller. As such, the Q0A, Q0B, and
0C outputs are the only outputs subject to modulation (see the
Distribution Embedded Output Clock Modulation section).
However, disabled (default) modulation channels have an
internal bypass allowing the Q0A, Q0B, and Q0C outputs to pass
directly through the modulation controller.
Q
Figure 68 shows a block diagram of the N shot/PRBS clock
controller, which includes a more detailed representation of the
relationship between the Q dividers, the modulation controller
and the N shot/PRBS controller. Although Figure 68 is specific
to PLL0, it is also representative of PLL1, because there are
dedicated controllers for each PLL channel.
N SHOT/PRBS CONTROLLER
N SHOT RETIMING MUX
N SHOT/PRBS GENERATOR A
REG 0x10D6,
BIT 4
REG 0x10D6,
BIT 0
PRBS GENERATOR A
SIGA
OUT
1
0
Mx PIN
FUNCTION
LOGIC
Mx
PINS
N SHOT GENERATOR A
N SHOT
REQUEST
TRIGGER
MODE
GAP
OUT
TO
OUTPUT
DRIVERS
REG 0x2101,
BIT 0
REG 0x10D3, BIT 6
REG 0x10D2, BITS[7:0]
REG 0x10D3, BITS[5:0]
N
MODULATION
CONTROLLER
Q DIVIDER PAIR
AA
FROM
SYSTEM
CLOCK
B
Q
0A
OUT0AN
BB
OUT0BP
OUT0BN
OUT0CP
OUT0CN
C
FROM
PLL0
CC
Q
TRIGGER
MODE
GAP
0AA
REG 0x10DA,
BIT 1
N
Q DIVIDER PAIR
Q DIVIDER PAIR
Q
0B
Q
0BB
HITLESS/
ZERO DELAY
FEEDBACK
AND SYNC
FEEDBACK
Q
0C
Q
0CC
TO
DPLL0
REG 0x1203, BITS[4:2]
REG 0x1202, BITS[4:0]
N DIVIDER
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES SHOWN ARE SPECIFIC TO Q DIVIDER PAIR Q0x (WHERE x = A, AA, B, BB, C, OR CC), PLL CHANNEL 0,
N SHOT/PRBS GENERATOR A, AND TRANSLATION PROFILE 0.0
3. SEE THE MODULATION SYNCHRONIZATION SECTION FOR DETAILS ABOUT THE SNCx AND SIGx SIGNALS
Figure 68. Block Diagram of JESD204B/Gapped Clock Controller
Rev. 0 | Page 100 of 205
Data Sheet
AD9546
Each Q divider has a dedicated pair of generators: one for
generating randomized clock signals (PRBS) and another for
supporting burst or gapped clock applications (N shot), as
shown by the expanded detail of N shot/PRBS Generator A in
Figure 68. Two bits are associated with each Q divider with one
bit for the N shot function and one bit for the PRBS function
per Table 68.
output levels generated at the rate of the associated Q divider.
Figure 69 shows a representation of the output clock signal.
The spectrum of a normal clock signal (square wave) is a series
of line spectra consisting of the fundamental clock frequency
and its associated harmonics. The main purpose of a PRBS
clock (or spread spectrum clock) is to broaden the spectral line
of the fundamental while at the same time reducing the
magnitude of the harmonic spectral lines.
Table 68. Register Address and Bit Assignments for the
N Shot and PRBS Functions
Because the user can program more than one output of the
AD9546 as a spread spectrum clock, each of the 10 PRBS
generators produces a unique PRBS sequence, which prevents
the possibility of correlated crosstalk between outputs enabled
for PRBS clock generation. All 10 PRBS generators use
17th-order generator polynomials, which means the PRBS
sequence is inherently periodic over a period of 217 − 1 (131,071)
Q divider output cycles.
Register
N Shot
PRBS
Q Divider Address
Function Bit
Function Bit
Q0A
Q0AA
Q0B
Q0BB
Q0C
Q0CC
Q1A
Q1AA
Q1B
Q1BB
0x10D4
0x10D4
0x10D4
0x10D4
0x10D5
0x10D5
0x14D4
0x14D4
0x14D4
0x14D4
D0
D2
D4
D6
D0
D2
D0
D2
D4
D6
D1
D3
D5
D7
D1
D3
D1
D3
D5
D7
N SHOT (JESD204B AND GAPPED CLOCKING)
N Shot Overview
The ability of the AD9546 to generate JESD204B clock signals
and gapped clock signals derives from the N shot generators.
Each N shot generator operates at the rate of the Q divider
output. The user has control over whether the N shot generator
produces a periodic output clock pattern or a single-burst
pattern. In either case, triggering a pattern is via Bit 0 of
Register 0x2101 for PLL0 and Register 0x2201 for PLL1.
Alternatively, an external signal can serve as the N shot trigger
via an Mx pin (see the Status and Control Pins section).
Table 69 shows how the PRBS and N shot function bit values
select different output clock functions.
Table 69. N Shot/PRBS Output Clock Function Selections
PRBS Function N Shot Function
Output Clock Function
Normal clock (Q divider)
Burst or gapped clock
Randomized clock
0
0
1
1
0
1
0
1
As shown in Figure 68, the N shot parameters (trigger, mode,
gap, and N) apply to all N shot generators in a PLL channel. As
such, the N shot parameters are common to all outputs in a
channel that have burst or gapped clock functionality enabled (per
Table 69).
Not applicable
Although it is physically possible to set both function bits to
Logic 1, the user must avoid doing so because Logic 1 is an
undefined mode and may result in unspecified device behavior.
RANDOMIZED CLOCK (PRBS)
Using the N shot generators imposes a minimum divide ratio of
8 on Q dividers configured for N shot functionality.
When the randomized clock function is in effect, the output
clock signal is a pseudorandom sequence of high and low
Q DIVIDER PERIOD BASED ON ITS INPUT CLOCK AND DIVIDE RATIO
Q DIVIDER
OUTPUT
PRBS
SEQUENCE
0
1
1
0
0
0
1
0
1
0
1
1
1
0
0
1
CLOCK
OUTPUT
TIME
Figure 69. PRBS Clock Signal
Rev. 0 | Page 101 of 205
AD9546
Data Sheet
When triggered, the N shot generator produces an output
N Shot Pattern Generation
sequence (see the bottom trace of Figure 70). At the end of the
sequence, the N shot generator remains in its default state
(Logic 0).
An N shot pattern consists of a specified clock segment and gap
segment. To assign the desired number of clock segment cycles,
use the 6-bit unsigned integer in Bits[5:0] of Register 0x10D3 for
PLL0 and Register 0x14D3 for PLL1. To assign the desired
number of gap segment cycles, use the 8-bit unsigned integer in
Bits[7:0] of Register 0x10D2 for PLL0 and Register 0x14D2 for
PLL1.
In burst operation, the output appears as a sequence of Q divider
cycles per the programmed clock segment value (in Figure 70, the
clock segment is 6). The gap period is effectively an extension of
the default output condition, which is the low state (in Figure 70,
the gap segment is 10).
As shown in Figure 70, the clock and gap segments are in units
of the Q divider output period. The N shot generator treats a
clock segment value of 0 as comprising zero clock cycles and a
gap segment value of 0 as comprising 1 clock cycle. Thus, the
net effect of programming a clock segment of 0 and a gap
segment of 0 is no output.
In periodic operation, the logic level of the trigger matters (rather
than the rising edge as in burst mode). The N shot generator,
starting from its default state (Logic 0), waits for the trigger to
assume a Logic 1 state. At this point, the N shot generator begins
repeatedly generating the burst pattern (clock segment pulses,
then gap segment pulses), per Figure 72. The N shot generator
continues generating burst patterns until the trigger assumes a
Logic 0 state. The occurrence of the Logic 0 state informs the
generator to stop synchronously. That is, if the generator is not in
the gap portion of the burst, the generator stalls at the end of the
current pulse as shown in Figure 73. Furthermore, returning the
trigger signal to Logic 1 causes the pattern to resume from where
it previously stopped (that is, with the absent pulses shown in
Figure 73).
The N shot generator is capable of burst or periodic gapped
operation. To select burst or periodic operation, use Bit 6 of
Register 0x10D3 for PLL0 and Register 0x14D3 for PLL1.
When Bit 6 = 0, the N shot generators operate in burst mode
(see Figure 71). When Bit 6 = 1, the N shot generators operate
in periodic gapped mode (see Figure 72).
In burst operation, the rising edge of the trigger signal constitutes
the trigger event. Prior to the burst, the generator is in the default
state (Logic 0), which the generator holds until triggered (see
the N Shot Triggering section).
Q DIVIDER PERIOD BASED ON ITS INPUT CLOCK AND DIVIDE RATIO
Q DIVIDER
OUTPUT
CLOCK SEGMENT = 6
GAP SEGMENT = 10
CLOCK
OUTPUT
TIME
Figure 70. N Shot Plus Gap Clock Signal
RISING EDGE OF
TRIGGER INITIATES
BURST
BURST ENDS AFTER CLOCK AND GAP Q DIVIDER CYCLES
Q DIVIDER PERIOD (PER ITS INPUT CLOCKAND DIVIDE RATIO)
Q DIVIDER
OUTPUT
CLOCK SEGMENT = 6
GAP SEGMENT = 10
TRIGGER
BURST
CLOCK
OUTPUT
TIME
Figure 71. Burst Clock Signal
Rev. 0 | Page 102 of 205
Data Sheet
AD9546
START
PATTERN
GENERATION
FALLING EDGE
TERMINATES
PERIODIC BURST
TRIGGER
CLOCK
BURST
PATTERN
BURST
PATTERN
BURST
PATTERN
BURST
PATTERN
BURST
PATTERN
OUTPUT
TIME
Figure 72. Periodic Gapped Clock Signal
END OF
PREVIOUS
BURST
END OF
FINAL
BURST
Q DIVIDER PERIOD (PER ITS INPUT CLOCKAND DIVIDE RATIO)
Q DIVIDER
OUTPUT
CLOCK SEGMENT = 6
GAP SEGMENT = 10
TRIGGER
ABSENT PULSES
CLOCK
OUTPUT
TIME
PULSES
STALL
Figure 73. Stalling a Periodic Gapped Clock Signal
For the direct triggering method, the trigger signal applies
N Shot Triggering
directly to the trigger input of the N shot generators. Thus, the
trigger signal is the trigger event.
The N shot trigger signal originates from one of two sources.
•
•
Bit 0 of Register 0x2101 or Register 0x2201
An external signal applied via an Mx pin
Note that in the following paragraphs, the terms slowest and
fastest appear in reference to the output clock signal of the Q
dividers. Slowest and fastest refers to the largest and smallest Q
divider phase offset value, respectively (see the Initial Phase Offset
section).
In either case, a Logic 1 constitutes a trigger event. In general,
the user must apply Logic 1 to trigger the N shot generators and
then return the trigger source to Logic 0. Otherwise, holding
the trigger source in a Logic 1 state indefinitely may lead to
unwanted retriggering of the N shot generators on subsequent
distribution synchronization events (see the Distribution
Output Clock Synchronization section).
For the retimed triggering method, the trigger signal routes to
the N shot retime block instead of directly to the N shot
generators. The rising edge of the trigger signal initializes the
retiming block such that the retiming block waits for the rising
edge of the slowest Q divider enabled for N shot operation
(user specified). The trigger signal, qualified by the rising edge
of the slowest Q divider, constitutes a retimed trigger event.
The retiming block sends a trigger signal to the N shot
generators coincident with the retimed trigger event as shown in
Figure 81.
The N shot generators respond to the trigger signal based on
Bit 6 of Register 0x10D3 or Register 0x14D3. As described in
the N Shot Pattern Generation section, Bit 6 makes the trigger
input of the N shot generators edge or level sensitive for
generating burst or periodic gapped clock signals, respectively (per
Figure 71 and Figure 72).
The trigger mechanism for delivering a trigger signal to the
N shot generators appears in the upper left section of Figure 68.
The N shot generators support two triggering methods: direct
and retimed.
The retimed trigger event occurs with a latency of three rising
edges of the slowest Q divider. Furthermore, when using the
retimed trigger mechanism, the associated Q dividers must
have a divide ratio of at least 32. A minimum setup time of
48 Q divider input half cycles is required between the retiming
output (slowest) and the subsequent N shot enabled output
(fastest).
To select the desired triggering method, use Bit 0 of
Register 0x10D6 for PLL0 or Register 0x14D6 for PLL1. Logic 0
(default) selects direct, whereas Logic 1 selects retimed. For a
description of Bit 4 of Register 0x10D6 in Figure 68, see the
Modulation Trigger section.
Rev. 0 | Page 103 of 205
AD9546
Data Sheet
The retimed trigger is on the rising edge of the slowest N shot
enabled Q divider to accommodate multiple N shot generators
that are providing multiple output clocks. Using the rising edge
of the slowest Q divider output as a retiming mark ensures that
all N shot generators begin clocking with the fastest output as
the earliest of the group, even when the Q dividers have
different programmed phase offsets and regardless of when the
N shot request occurs.
The AD9546 automatically selects the appropriate N shot
enabled Q divider for trigger retiming, such that the output
with the largest Q divider phase offset value is always the
retiming clock.
Rev. 0 | Page 104 of 205
Data Sheet
AD9546
DISTRIBUTION EMBEDDED OUTPUT CLOCK MODULATION
within a PLL channel, whereas the period parameter, tMOD, is
unique to each modulator.
MODULATION CONTROLLER OVERVIEW
The AD9546 has the capability to embed a low frequency clock
within a high frequency carrier. Referring to Figure 68, only the
primary output of a Q divider pair routes to the modulation
controller, whereas the secondary output bypasses the modulation
controller and routes directly to the N shot/PRBS controller. Thus,
only the primary distribution clock outputs support embedded
clock modulation capability (for example, Output OUT0AP
supports modulation, whereas Output OUT0AN does not).
MODULATION MAGNITUDE
To set the modulation magnitude, use the 16-bit unsigned
integer (modulation step) in Register 0x10C0 to Register 0x10C1
for PLL0 and Register 0x14C0 to Register 0x14C1 for PLL1.
The modulation step value carries units of one-half of the
period of the input clock to the Q divider associated with the
modulator, yielding the following relationship:
Embedded clock modulation consists of changing the pulse
width of the designated Q divider output clock in synchrony
with a binary modulation signal to produce modulation events.
A modulation event always spans two Q divider clock cycles
where the first clock cycle changes duty cycle, but the second
clock cycle may or may not change duty cycle (see the Balanced
and Unbalanced Modulation section).
D = Modulation Step/(2 × Qxy)
where:
D is the duty cycle deviation (that is, the time deviation of the
modulation edge from nominal, normalized to the Q divider
output period). Figure 75 shows D as Δt/tQ.
(3)
Qxy is the divide ratio of the relevant Q divider (x is 0 or 1, and
y is A, B or C).
An expanded diagram of the modulation controller appears in
Figure 74. Although Figure 74 is specific to PLL0, it is also
representative of PLL1, because there is a dedicated modulation
controller for each PLL channel.
Because the modulation step is common to all the modulators
in a PLL channel, whereas Qxy is unique to each Q divider, D
in Equation 3 is not necessarily the same for all the modulators
in a PLL channel.
Any one or more Q divider outputs with a single-letter
subscript (for example, Q0A but not Q0AA) can operate as an
embedded clock modulator. To enable embedded modulation,
use Bit 0 of the registers shown in Table 70.
Given the divide ratio for Q divider Q0A is 1001, find the
modulation step bit field value necessary for Modulator A to
yield 5% modulation.
Modulation of 5% implies D = 0.05. Substituting the
appropriate values into Equation 3 yields
Table 70. Enable Embedded Modulation Register Address
Q Divider
Register Address
0.05 = Modulation Step/(2 × 1001)
Therefore,
Q0A
0x10CF
Q0B
0x10D0
Q0C
Q1A
Q1B
0x10D1
0x14CF
0x14D0
Modulation Step = 100 (rounded to nearest integer)
= 0x64 (hexadecimal)
Given the same modulation step value as in the preceding
example (modulation step = 100), find the modulation magnitude
for Modulator B assuming Q0B has a divide ratio of 8025.5.
Logic 1 selects the designated Q divider for embedded clock
modulation, whereas Logic 0 (default) bypasses the modulation
controller (via the mux in Figure 74).
D = Modulation Step/(2 × Qxy)
= 100/(2 × 8025.5)
Modulation control consists of two parameters: Δt and tMOD
.
Parameter Δt defines the desired magnitude of the modulation
edge variation, and tMOD defines the modulation period (see
Figure 75). The modulated signal consists of a time step of
magnitude Δt occurring at regular intervals of period, tMOD. The
magnitude parameter, Δt, is common to all the modulators
= 0.00623 (0.623%)
Rev. 0 | Page 105 of 205
AD9546
Data Sheet
MODULATION CONTROLLER
MODULATOR A
TRIGGER
MODE
28
REG 0x10C5 TO 0x10C2,
BITS[27:0]
REG 0x10CF, BIT 1
REG 0x10CF, BIT 2
16
16
REG 0x10C1 TO 0x10C0,
BITS[15:0]
REG 0x10CF,
BIT 0
0
1
SIGA
Q
0A
INA
IN
TRIGGER
OUT
MODULATION
BLOCK
MODE
SNCA
SYNC
REG 0x10CF,
BIT 3
Q
0AA
N SHOT ENBL
16
Q
Q
INB
INC
SIGB
SNCB
SIGC
0B
IN
OUT
MODULATOR B
MODULATOR C
STEP
TRIGGER
MODE
0BB
SYNC
OUT
16
Q
Q
0C
IN
STEP
TRIGGER
MODE
0CC
SNCC
SYNC
SIGAA
SIGBB
SIGCC
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES SHOWN ARE SPECIFIC TO Q DIVIDER Q0x (WHERE x = A, AA, B, BB, C, OR CC),
PLL CHANNEL 0, AND MODULATOR A
Figure 74. Block Diagram of Embedded Clock Modulation Controller
MODULATION PERIOD (tMOD
)
MODULATION MAGNITUDE
(D = Δt/tQ
)
NOMINAL
(50%)
NOMINAL
(50%)
Δt
TIME
Q DIVIDER
OUTPUT PERIOD (tQ
)
Figure 75. Modulation Magnitude and Period
Rev. 0 | Page 106 of 205
Data Sheet
AD9546
possible to produce the desired modulation period exactly. In
this example, the actual tMOD value is 1.0000210084 ms (per
Equation 4).
MODULATION PERIOD
The modulation period is the interval between modulation events
(see Figure 75). To control the modulation period parameter,
use the 28-bit unsigned integer (modulation counter) in the
register address ranges shown in Table 71. The modulation
counter value carries units of the period of the output clock of
the associated Q divider.
BALANCED AND UNBALANCED MODULATION
Modulation consists of periodic modulation events occurring at
regular intervals, tMOD, with a single modulation event spanning
two output clock cycles of the Q divider associated with the
modulator. The modulator applies the specified pulse width
variation (per the modulation step and modulation counter) to
one or both clock cycles of the modulation event. Modulation
of both clock cycles constitutes balanced modulation, whereas
modulation of only one clock cycle constitutes unbalanced
modulation.
The modulation counter value must be greater than or equal to six.
Table 71. Modulation Period Address Ranges
Modulator
Register Address
0x10C2 to 0x10C5
0x10C6 to 0x10C9
0x10CA to 0x10CD
0x14C2 to 0x14C5
0x14C6 to 0x14C9
Q0A
Q0B
Q0C
Q1A
With balanced modulation, the modulator applies opposite
polarity to the modulation steps of the two consecutive cycles
of the modulation event (see Figure 76). As such, balanced
modulation maintains the average dc level of the waveform at
its nominal 50% amplitude point.
Q1B
The three lower addresses for each Q divider in Table 71 carry
the 24 LSBs of the 28-bit integer, and Bits[3:0] of the upper
address carry the four MSBs of the 28-bit integer (Bits[7:4] are
unused).
With unbalanced modulation, the modulator applies the
modulation step (Δt) to only the first pulse of the modulation
event, leaving the second pulse unaltered (see Figure 77) and
resulting in a waveform with a dc level slightly less than its
nominal 50% amplitude point.
The modulation period (tMOD) depends on the input clock
frequency (fIN), the divide ratio of the associated Q divider, and
the modulation counter value.
t
MOD = Modulation Counter × (Qxy/fIN)
(4)
To select between unbalanced and balanced modulation, use
Bit 2 per the register address in Table 70 for a given Q divider.
Logic 0 (default) selects balanced modulation, whereas Logic 1
selects unbalanced modulation.
Suppose the input clock to Q Divider Q0A comes from APLL0
with its VCO operating at 2.38 GHz. Because the VCO drives
a divide-by-2 block prior to the Q divider clock input, the
Q divider input frequency (fIN) is 1.19 GHz. Find the value of
the modulation counter value required to yield a modulation
period (tMOD) of 1 ms (0.001 sec) for a Q divide ratio of 107.5.
By default, the first pulse of a modulation event has −Δt (and
the second pulse +Δt if balance modulation is in effect).
However, the user can force alternate polarity, where the first
pulse of a modulation event has +Δt (and the optional second
pulse has −Δt). Polarity control is via Bit 1 per the register
address in Table 70 for a given Q divider. Logic 0 (default)
applies negative polarity to Δt for the first pulse of a
Substituting the appropriate values into Equation 4 yields
10−3 = Modulation Counter × (107.5/(1.19 × 109))
Therefore,
modulation event, whereas Logic 1 applies positive polarity to
Δt for the first pulse (see Figure 78), which causes a positive shift
in dc offset in the case of unbalanced modulation.
Modulation Counter = 11,070 (nearest integer)
= 0x 0000 2B3E (hexadecimal)
The desired modulation period (tMOD) is 1 ms. The modulation
counter value must be an integer. However, it is not always
Rev. 0 | Page 107 of 205
AD9546
Data Sheet
MODULATION EVENT
FIRST
MODULATION
CYCLE
SECOND
MODULATION
CYCLE
–Δt
+Δt
DC = 50%
50%
NOMINAL
(50%)
NOMINAL
(50%)
TIME
Figure 76. Balanced Modulation Waveform
MODULATION EVENT
FIRST
MODULATION
CYCLE
SECOND
MODULATION
CYCLE
SECOND PULSE
UNALTERED
–Δt
50%
DC < 50%
TIME
NOMINAL
(50%)
NOMINAL
(50%)
Figure 77. Unbalanced Modulation Waveform
MODULATION EVENT
FIRST
MODULATION
CYCLE
SECOND
MODULATION
CYCLE
–Δt
+Δt
DC = 50%
BALANCED
MODULATION
50%
+Δt
DC > 50%
UNBALANCED
MODULATION
50%
TIME
SECOND PULSE
UNALTERED
NOMINAL
(50%)
NOMINAL
(50%)
Figure 78. Alternate Modulation Polarity
Rev. 0 | Page 108 of 205
Data Sheet
AD9546
associated with tag modes in the translation profiles (see the
Translation Profiles section) per Table 72.
MODULATION SYNCHRONIZATION
The periodic modulation events produced by the modulators
are useful as timing markers for synchronization purposes.
Both the balanced and unbalanced modulation schemes use
two Q divider output cycles to define a modulation event (see
Figure 76 and Figure 77). In the case of unbalanced
modulation, the second cycle is unaltered but is still part of the
modulation event. As such, the first rising edge of the Q divider
output following the second modulated cycle marks the base
edge for synchronization purposes (see Figure 79).
Table 72. Tag Mode Register Address
Translation Profile
Register Address
0.0
0.1
0.2
0.3
0.4
0.5
1.0
1.1
1.2
1.3
1.4
1.5
0x1203
0x1223
0x1243
0x1263
0x1283
0x12A3
0x1603
0x1623
0x1643
0x1663
0x1683
0x16A3
The synchronization information available from the
modulators requires the use of a secondary signal path. This
path is the reason the modulators have SIGx and SNCx output
signals. The SIGx signals are the modulated waveforms that
ultimately propagate to the distribution output pins.
The SNCx signals are internal only and carry synchronization
information attached to the modulation events corresponding
with the associated SIGx signal (for example, SNCA associates
with SIGA). The SNCx signal is a pulse coincident with the base
edge of the corresponding SIGx signal, as shown in Figure 80.
Because the SNCx signal identifies the modulation base edge,
the user can synchronize the DPLL N divider with a specific
synchronization edge per Figure 80. To specify a synchronization
edge (0, 1, 2, or 3) for the N divider, use Bits[1:0] of
Register 0x10CE for PLL0 and Register 0x14CE for PLL1. The
decimal value of the bit pair corresponds to the desired
synchronization edge in Figure 80.
As shown in Figure 68, the SNCx signals route to the DPLL via
the hitless/zero delay feedback and synchronization block of
the N shot/PRBS controller. The SNCx signals provide
synchronization to the N divider of the corresponding DPLL
channel. The device automatically selects the appropriate SNCx
signal for the Q divider chosen for feedback (when modulation is
in effect for that particular Q divider). The SNCx selection
depends on Bits[4:2], the 3-bit tag mode bit field in the registers
The choice of synchronization edge (0, 1, 2, or 3) puts an
additional constraint on the value of the modulation counter
(see the Modulation Period section) as follows:
Modulation Counter ≥ Sync Edge + 7
MODULATION EVENT
TIME
BASE
EDGE
Figure 79. Modulation Synchronization Edges
MODULATION PERIOD
BASE
EDGE
BASE
EDGE
SYNC EDGE
SYNC EDGE
0
1
2
3
0
1
2
3
MODULATION
EVENT
MODULATION
EVENT
SIGx
SNCx
TIME
Figure 80. Modulation Synchronization
Rev. 0 | Page 109 of 205
AD9546
Data Sheet
events and then stops. Conversely, when N shot request mode
= 1, the modulator continuously generates modulation events
when the N shot request signal (see Figure 68) is Logic 1. This
behavior allows synchronous enable or disable of modulation
events without losing synchronization during modulation.
MODULATION TRIGGER
By default, when modulation is in effect for a given Q divider,
the modulator applies the modulation immediately. However, it
is also possible to delay modulation until triggered. To enable
this feature for a given Q divider, use Bit 3 in the register
address shown in Table 70. Logic 0 (default) disables the
triggering feature, whereas Logic 1 enables it.
The modulators, like the N shot generators, provide a retimed
trigger option (see the N Shot Triggering section). However, in
the case of the modulators, retiming relates to modulation
events rather than the carrier clock edges associated with the Q
divider. By default, the retiming block retimes to the carrier
clock edges. To select retiming to modulation events, program
Bit 4 to Logic 1 in Register 0x10D6 (PLL0) or Register 0x14D6
(PLL1).
When enabled, the modulators use the trigger and mode signals
originating within the N shot/PRBS controller (see Figure 68).
The mode behavior differs slightly from that of the N shot/PRBS
controller. Specifically, when N shot request mode (Bit 6 in
Register 0x10D3 for PLL0 or Register 0x14D3 for PLL1) = 0,
instead of the modulator continuously generating periodic
modulation events, it generates only five periodic modulation
MINIMUM 48 Q DIVIDER INPUT HALF CYCLES
Q
0A
Q
0AA
Q
0B
TRIGGER
SIGNAL
RETIMED
TRIGGER
N SHOT OUT0AP
N SHOT OUT0AN
N SHOT OUT0BP
3-OUTPUT CYCLE LATENCY
TIME
Figure 81. Trigger Retiming (with Q0AA as the Retiming Q Divider)
Rev. 0 | Page 110 of 205
Data Sheet
AD9546
DISTRIBUTION OUTPUT CLOCK SYNCHRONIZATION
SYNCHRONIZATION OVERVIEW
OUTPUT SIGNALS AFTER POWER-UP OR RESET
The AD9546 has a built in synchronization controller that
provides the user with a flexible and sophisticated means to
synchronize its multiple output clock signals. The synchronization
controller properly sequences internal events to ensure proper
synchronizing of the output clocks of the device.
A particularly important aspect of the synchronization controller
involves the control of the output clocks following a device
power-up or reset. On power-up or reset, the synchronization
controller waits for a synchronization trigger before activating
the output clocks. Until a synchronization trigger occurs, the
synchronization controller holds all the distribution Q dividers
in a reset state, which prevents any clock signals from
appearing at the OUTxyN and OUTxyP pins. Therefore, the
user must initiate the synchronization sequence to have any
clock signals appear at the output.
There are three ways to trigger the synchronization sequence.
•
Manually (via Bit 3 of Register 0x2000, via Bit 3 of
Register 0x2101 for PLL0 and Register 0x2201 for PLL1,
via an op code in an EEPROM download, or via an Mx
pin)
MANUAL SYNCHRONIZATION TRIGGER
•
•
Autoreconfiguration (changes to Q divider values)
Autosynchronization
A manual synchronization trigger occurs when the user sets
one of three manual synchronization bits. Programming Bit 3 =
1 of Register 0x2000 initiates the synchronization sequence for
all the AD9546 output clocks. Alternatively, the user can
synchronize only those outputs associated with one of the PLL
channels by programming Bit 3 = 1 of Register 0x2101 for PLL0
or Bit 3 = 1 of Register 0x2201 for PLL1. The same
Autosynchronization is the preferred option for output clock
synchronization.
A synchronization trigger event is channel specific, applying to
either PLL0 or PLL1. The exception is manual synchronization,
which can apply to both channels via Bit 3 of Register 0x2000.
Assertion of Bit 3 does not imply synchronization of both
channels relative to each other but causes a simultaneous triggering
of the synchronization process for both channels.
synchronization features available via these bits are also
available via any of the Mx pins configured as an input.
Setting any of the D3 bits causes the synchronization sequence
to execute up to the point just before unmuting all drivers. At
that point, the synchronization controller stalls, waiting for the
user to clear the appropriate manual synchronization bit. This
delay allows the user to update the Q divider ratios before the
controller unmutes the output drivers, which does not occur
until the user clears the relevant manual synchronization bit.
A synchronization trigger initiates the synchronization
sequence. Following a trigger event, the synchronization
controller checks whether Bit 2 of Register 0x10DB or Register
0x14DB is set. If so, the synchronization controller waits for a
hitless profile to become valid before muting all outputs (because
reference synchronization is only viable for zero delay/hitless
translation modes).
AUTORECONFIGURATION SYNCHRONIZATION
TRIGGER
Next, the synchronization controller confirms that the output
drivers are muted and then puts the P dividers and Q dividers
in a reset state (for the controller to load new Q divider
configuration information, for example, a new divide ratio).
Then the synchronization controller again checks whether Bit 2
of Register 0x10DB or Register 0x14DB is set so the controller
can align the release of the output clocks with a rising edge of
the active reference. However, if Bit 2 of Register 0x10DB or
Register 0x14DB bit is clear, the synchronization controller
checks for release of manual synchronization (that is, a clearing
of Bit 3 of Register 0x2000, or Bit 3 of Register 0x2101 (for
PLL0) or Bit 3 of Register 0x2201 (for PLL1). The
An autoreconfiguration synchronization trigger occurs when
the user changes any of the Q divider ratios and subsequently
sets the IO update bit. However, the autoreconfiguration
synchronization trigger applies only after the synchronization
controller has already completed a synchronization sequence as
the result of a power-up (or reset). After a power-up (or reset)
synchronization sequence completes, any subsequent updates
to the Q divider ratios constitutes an autoreconfiguration
synchronization trigger.
Autoreconfiguration is a convenience feature, because it
synchronizes the outputs automatically when the user changes
one or more Q divider values and subsequently sets the IO
update bit. This feature alleviates the need to initiate a manual
synchronization trigger to change Q divider values. Be aware,
however, that changing a single Q divider value affects all the
clock signals of the corresponding PLL channel. Specifically,
changing the value of any Q0x (where x = A, AA, B, BB, C, or
CC) results in disturbing all the Output 0 clock outputs because
it initiates a synchronization sequence for those outputs.
Likewise, changing the value of any Q1x divider (where x = A,
synchronization controller also checks that the APLL indicates
calibrated and locked status before proceeding.
Next, the synchronization controller unmutes the output
drivers and releases the appropriate Q dividers from their reset
state. Finally, the synchronization controller returns to waiting
for a subsequent synchronization trigger.
The output clock synchronization sequence initiates only when
the system clock is locked and stable.
Rev. 0 | Page 111 of 205
AD9546
Data Sheet
AA, B, or BB) results in disturbing all the Output 1 clock
outputs. These output clock disturbances result from the
synchronization process. However, the disturbances are glitch
free, because the synchronization sequence mutes the affected
output drivers during synchronization.
PLL. When autosynchronization mode = 2, the
autosynchronization event constitutes an indication of phase
lock by the DPLL.
When autosynchronization mode = 3, the autosynchronization
event constitutes an indication of frequency lock.
The aforementioned output clock disturbances are avoidable,
however, when the clock source to the Q dividers is the system
clock (see Table 62) and the system clock synchronization mask
bits are Logic 1 (see the Q Divider Clock Source Selection
section) for a subset of the device outputs.
After the autosynchronization trigger occurs for a PLL channel,
the autosynchronization trigger generation sequence terminates.
Subsequent autosynchronization sequences on the same PLL
channel are not possible unless the user programs Bit 0 = 1 of
Register 0x2107 or Register 0x2207, or by updating Bits[1:0] of
Register 0x10DB or Register 0x14DB.
AUTOSYNCHRONIZATION TRIGGER
Although the AD9546 provides the user with the ability to
synchronize the distribution outputs via a manual synchroniza-
tion trigger, it is not a recommended action. Instead, the
AD9546 provides an autosynchronization trigger feature that
applies only after meeting certain conditions, which overcomes
potential usability issues associated with manual trigger scenarios.
The immediate autosynchronization mode (see Table 73) is an
effective replacement for manual synchronization triggering.
REFERENCE SYNCHRONIZATION
Output clock synchronization establishes a coincident starting
time among output clocks (usually on a per PLL channel basis).
Output clock synchronization, however, is generally
asynchronous with respect to input reference clock signal
edges. The reference synchronization feature allows the user to
make a rising edge of the input reference clock an additional
gating item for synchronization of the output clocks. Reference
synchronization establishes a small fixed phase relationship
(~30 ns) between the input and output clocks (assuming the
output clock frequency is an integer multiple of the input
reference frequency). Establishing an initial fixed phase
relationship between the input and output clocks minimizes the
DPLL acquisition time by making it deterministic rather than
completely random.
The user selects the desired autosynchronization trigger
conditions via Bits[1:0] of Register 0x10DB for PLL0 and
Register 0x14DB for PLL1. Table 73 shows the selectable
autosynchronization conditions.
Table 73. Autosynchronization Mode Selections
Bits[1:0] (Decimal)
Autosynchronization Mode
0
1
2
3
Disabled (default)
Immediate
On DPLL phase lock
On DPLL frequency lock
The reference synchronization feature requires a hitless profile
be active when reference synchronization is in effect. The
reference synchronization feature does not function with a
phase buildout profile.
Autosynchronization enables the AD9546 to execute a
distribution synchronization sequence without any user
intervention. The nonzero autosynchronization selections
specify conditions for generating an autosynchronization
trigger.
Like most of the other synchronization features, reference
synchronization applies on a per PLL channel basis. To enable
the reference synchronization feature, use Bit 2 (reference
synchronization) of Register 0x10DB for PLL0 and Register
0x14DB for PLL1. Setting the reference synchronization bit,
which is Logic 0 by default, enables reference synchronization
of those outputs associated with the corresponding PLL
channel.
When autosynchronization mode = 0 (default), only manual or
autoreconfiguration triggers are possible. When autosynchroniza-
tion mode ≠ 0, the autosynchronization trigger feature is in
effect.
When using reference synchronization in conjunction with
autosynchronization mode (that is, autosynchronization mode
= 1, 2, or 3), be sure to program the reference synchronization
bit to Logic 1 and assert the I/O update bit prior to
When autosynchronization is active (autosynchronization
mode ≠ 0), nothing happens until the system clock PLL
indicates a locked status. The autosynchronization mechanism
then enters an autosynchronization trigger detection loop
awaiting the specified autosynchronization event. When
autosynchronization mode = 1, the autosynchronization event
constitutes the indication of a locked status by the system clock
programming autosynchronization mode = 1, 2, or 3.
Rev. 0 | Page 112 of 205
Data Sheet
AD9546
FREQUENCY TRANSLATION LOOPS
Translation Profile 1.0 through Translation Profile 1.5 for
DPLL1). Each translation profile consists of the following
programmable parameters:
FREQUENCY TRANSLATION LOOPS OVERVIEW
The frequency translation capability of the AD9546 comprises
two independent dual-PLL channels (Channel 0 and Channel 1).
Each PLL channel comprises a DPLL followed by an APLL.
Both channels have programmable reference dividers at the
input and programmable channel dividers at the output. Figure 82
shows a block diagram of a single channel. REFx and REFy can
each be any of the reference inputs, but not the same reference
input.
•
•
•
•
•
•
•
•
Profile enable
Profile priority
Translation mode selection
Input reference source selection
N divider value (integer, fraction, and modulus)
DPLL loop filter bandwidth
Reference and feedback TDC tagging options
DPLL fast acquisition options
The DPLL channels (DPLL0 and DPLL1) are capable of low
loop bandwidths (µHz) well suited for jitter cleanup
applications involving the 1 pulse per second (pps) input
signals that originate from global positioning system (GPS) and
global navigation satellite system (GNSS) receivers. The DPLL
is of the fractional-N variety but is capable of Integer N
operation when the fractional part is set to zero.
Table 74 shows the translation profile address ranges.
Table 74. Translation Profile Register Address Ranges
Translation Profile
Start Address
0x1200
0x1220
0x1240
0x1260
0x1280
0x12A0
0x1600
0x1620
0x1640
0x1660
0x1680
0x16A0
End Address
0x1217
0x1237
0x1257
0x1277
0x1297
0x12B7
0x1617
0x1637
0x1657
0x1677
0x1697
0x16B7
0.0
0.1
0.2
0.3
0.4
0.5
1.0
1.1
1.2
1.3
1.4
1.5
The DPLL provides an output frequency of up to approximately
400 MHz. The DPLL also plays a role in the automatic
reference switching capability of the AD9546 and its ability to
provide hitless and phase buildout functionality.
The APLL is of the Integer N variety with an integrated VCO
and provides the high frequency upconversion capability of the
cascaded DPLL/APLL pair. Unlike the two DPLLs, which are
identical, the two APLLs differ slightly. Specifically, they have
nonoverlapping VCO ranges with APLL0 spanning 2.424 GHz
to 3.232 GHz and APLL1 spanning 3.232 GHz to 4.040 GHz.
TRANSLATION PROFILES
When a given profile is in effect, the DPLL uses the parameters
as specified in that profile. That is, the DPLL has no memory of
the given profile when switching to another profile.
DPLL0 and DPLL1 each have six dedicated sections in the
register map allocated for translation profiles (Translation
Profile 0.0 through Translation Profile 0.5 for DPLL0 and
XOA XOB
LFx
OTHER DPLL
FEEDBACK TDC
DIGITAL
CROSS
POINT
MUX
SYSTEM
CLOCK
AD9546
fS
AUXILIARY
REFERENCE TDCS
DPLL
APLL
OUTPUT
DISTRIBUTION
INVERSE USER TIME
STAMPERS (IUTS)
TIME STAMP PROCESSOR,
DIGITAL LOOP FILTER,
TUNING WORD PROCESSOR
AND NCO
PFD
AND
VCO
÷2
Q
LOOP
FILTER
AUXILIARY NCOS
OUTxyN/
OUTxyP
PHASE
BUILD
OUT
FEEDBACK
TDC
MUX
N + FRAC/MOD
REFx
REFy
R
R
M DIVIDER
0
1
TDC
N DIVIDER
INTERNAL ZERO DELAY
REFERENCE TDCS
EXTERNAL ZERO DELAY (MAY INCLUDE ADDITIONAL FREQUENCY TRANSLATION)
NOTE
1. REFx AND REFy CAN EACH BE ANY OF THE REFERENCE INPUTS, BUT NOT THE SAME REFERENCE INPUT.
Figure 82. Frequency Translation Loop Block Diagram
Rev. 0 | Page 113 of 205
AD9546
Data Sheet
The translation profiles control only the frequency translation
of the DPLL. The overall frequency translation of the channel
depends on the following:
Source selection is via Bits[4:0] in the register at the start
address of the translation profile (see Table 74) plus an offset of
1 (decimal). The decimal value associated with Bits[4:0] assigns
the source, as shown in Table 75.
•
•
•
•
N divider value (depending on the translation mode)
M divider value (via the APLL0 and APLL1 controls)
R divider value (via the REFx and auxiliary REFx controls)
Q divider value (via the distribution controls)
Table 75. Reference Source Selection
Bits[4:0]
Source
0
REFA
1
2
3
4
5
6
7
REFAA
REFB
REFBB
Each For detailed information on the use and programming of
the N, M, R, and Q divider values, see the DPLL Feedback
Divider (N Divider), APLL Feedback Divider (M Divider),
Reference Dividers (R Dividers), and Distribution Dividers (Q
Dividers) sections.
Feedback from DPLL0 (for PLL1 only)
Feedback from DPLL1 (for PLL0 only)
Auxiliary REF0
PROFILE ENABLE
Auxiliary REF1
For a DPLL channel to have access to a translation profile, the user
must enable the profile. To enable a translation profile, program
Bit 0 to Logic 1 (default is Logic 0) in the register at the start
address of the translation profile (per Table 74). Only enabled
translation profiles are selectable by the DPLL. When the DPLL
selects a translation profile, the selected translation profile
becomes an active translation profile (see the Reference
Switching section for what constitutes an active translation
profile). Certain functional blocks within the AD9546 make use
of the active and inactive state of a translation profile.
8
9
10
11
12
13
14
15 to 31
Auxiliary NCO 0
Auxiliary NCO 1
Not applicable
Auxiliary REF2
Auxiliary REF3
Inverse User Time Stamper 0
Inverse User Time Stamper 1
Not applicable
ACTIVE REFERENCE INDICATION
When a DPLL selects a translation profile, the translation
profile becomes active (see the Profile Enable section), which
causes the user assigned reference input (see the Input
Reference Source Selection section) associated with that profile
to become the time stamp source to the DPLL.
PROFILE PRIORITY
Profile priority works in conjunction with the reference switching
capability of the AD9546 (see the Reference Switching section).
The user programs a priority value via Bits[5:1] in the register at
the start address of the translation profile per Table 74. When a
DPLL must switch from one input reference to another, the
DPLL can do so according to the 5-bit priority value
Note that when a reference becomes the time stamp source to a
DPLL, the reference becomes an active reference. The user has
access to the active or inactive state of a reference via an
appropriately configured Mx status pin.
programmed by the user. A priority of 0 constitutes the highest
priority, and a priority of 31 constitutes the lowest priority.
In summary, indication of active reference status is the result of
a DPLL activating a translation profile (which contains a user
assigned reference source). As such, do not confuse the active
reference status with the valid reference status, which is a status
condition associated with the reference source itself (not a
translation profile).
INPUT REFERENCE SOURCE SELECTION
The AD9546 provides the user with several options for the
reference source that supplies the input clock signal to the
DPLL. These options include any one of the four REFx inputs,
any of the four auxiliary REFx inputs, either of the two auxiliary
NCOs, either of the two inverse user time stampers, and a
special configuration that uses the feedback signal from the
other DPLL channel. Each translation profile can specify a
different reference source.
Rev. 0 | Page 114 of 205
Data Sheet
AD9546
Even during a hitless acquisition, the AD9546 builds out the
TRANSLATION MODES
initial phase offset between the reference and feedback signals
(as though the device is performing a phase buildout
acquisition) to help mitigate the disturbance at the output.
However, unlike the case of a phase buildout acquisition, the
phase slew rate limiter is operational during a hitless
acquisition (see the Phase Slew Rate Limit section for details on
the phase slew rate limiter) and applies slew rate limiting to the
buildout phase transition as needed.
Overview
The translation modes govern the way the DPLL responds
when it acquires a reference signal. In general, at the start of a
reference acquisition, the feedback of the DPLL and reference
signals do not have the same phase relationship. As such, when
the DPLL begins an acquisition, it likely exhibits a significant
output disturbance as the loop attempts to compensate for the
phase mismatch. To deal with the output disturbance that
typically results from a reference acquisition, the AD9546 offers
two translation mode types: phase buildout and hitless.
For further information on how a PLL responds to switching
from one reference to another under hitless and phase buildout
operating modes, see the AN-1420 Application Note.
A phase buildout acquisition virtually eliminates the output
disturbance that results from an initial phase mismatch. During
a phase buildout acquisition, the DPLL effectively measures the
initial phase offset between its feedback and reference inputs
and inserts the measured phase offset into the feedback path of
the loop. The measured phase offset is the phase buildout offset.
Insertion of the phase buildout offset into the loop effectively
eliminates the initial phase difference between the feedback and
reference signals, thereby minimizing the disturbance at the
output of the DPLL. The insertion of the phase buildout offset
implies a phase difference between the output of the DPLL and
reference clock signals. This phase difference is the expected
behavior for a phase buildout reference acquisition.
Translation Mode Selection
PLL0 and PLL1 can operate in any one of three modes:
•
•
•
Phase buildout mode
Internal zero delay mode (hitless)
External zero delay mode (hitless)
The zero delay modes constitute hitless rather than phase
buildout operation. For details on each mode, see the Phase
Buildout Mode section, Internal Zero Delay (Hitless) Mode
section, and External Zero Delay (Hitless) Mode section.
The user selects the operating mode via the translation profile
registers. That is, each profile gives the user the option of selecting
any one of the three operating modes for that specific profile. To
select the operating mode, use Bits[1:0] in the register at the
start address of the translation profile in Table 74 plus an offset
of 3 (decimal). Bits[1:0] select the operating mode per Table 76.
Note that during a phase buildout acquisition, the phase slew
rate limit function does not operate on the buildout phase
transition (see the Phase Slew Rate Limit section of the Digital
PLL (DPLL) section for details on the phase slew rate limiter).
Unlike a phase buildout acquisition, a hitless acquisition
ultimately results in a phase match of the output of the DPLL
and reference clock signals. As such, a hitless translation mode
is a prerequisite for zero delay operation. Given a phase
mismatch between the reference of the DPLL and feedback
signals at the start of an acquisition, the DPLL temporarily
changes its output frequency to steer the phase difference
toward zero. The AD9546 optimizes this process by assessing
the initial phase relationship between the reference and
feedback signals and steering toward the reference clock edge
exhibiting the smallest relative phase offset.
Table 76. Frequency Translation Modes
Bits[1:0] (Decimal)
Translation Mode
0
1
2
3
Phase buildout
Internal zero delay
Not applicable
External zero delay
Because a hitless acquisition undergoes frequency changes to
bring about phase alignment, there can be a significant phase
disturbance at the output while the loop locks to the phase of
the reference (even if the initial feedback and reference frequencies
are identical). The user can minimize this disturbance by placing a
limit on the DPLL output frequency excursion (caused by a
phase rate of change (δθ/δt) at its input) via the phase slew limit
feature (see the Phase Slew Rate Limit section). Furthermore,
the user can place a limit on the overall output frequency
excursion (due to both an input frequency offset and δθ/δt) by
using the frequency clamp feature (see the Tuning Word Offset
Clamp section).
Rev. 0 | Page 115 of 205
AD9546
Data Sheet
However, the NCO, VCO, TDCs, and the APLL PFD inputs
each have specific frequency limits. The following formulas
relate the output frequency of the NCO (fNCO), fVCO, fTDC, and
PHASE BUILDOUT MODE
Figure 83 shows the phase buildout configuration. In phase
buildout mode, when the AD9546 switches from one input
reference to another, there is virtually no phase disturbance at
the output. Because the AD9546 uses time stamps generated by
the reference TDCs, the difference between the time stamps
from the old and new reference equates to a phase difference.
Thus, the AD9546 can assess the phase offset and automatically
compensate at switchover (see the Initial Phase Skew Refinement
Steps section for more detail).
f
PFD to fREFx, fOUTx, and divider values. Note that each formula has
two solutions: one with respect to the input frequency (fREFx
)
and the other with respect to the output frequency (fOUTx). The
user must ensure that fNCO, fVCO, fTDC, and fPFD are within their
specified frequency bounds.
f
TDC = fREFx/R
= (2 × Q/(M × (N + FRAC/MOD))) × fOUTx
(5)
The frequency translation factor for phase buildout mode is
where:
REFx is the reference frequency for a specific reference input, for
example, REFA, REFAA, REFB, or REFBB.
OUTx is the output frequency for a specific distribution output,
for example, OUT0C or OUT1A.
f
FRAC
MOD
N +
× M
fOUTx
=
f
fREF x
2 × R × Q
f
NCO = fPFD = (2 × Q/M) × fOUTx
= ((N + FRAC/MOD)/R) × fREFx
VCO = 2 × Q × fOUTx
where:
N is the N divider integer part.
FRAC is the N divider fractional part.
MOD is the N divider fractional modulus.
M is the divide ratio of the APLL feedback divider.
R is the divide ratio of the input reference divider.
Q is the divide ratio of the distribution divider.
(6)
(7)
f
= ((N + FRAC/MOD) × M/R) × fREFx
When the reference source is auxiliary NCO 0, auxiliary NCO 1, or
the feedback from the other DPLL, then in Equation 5 through
Equation 7, use R = 1.
XOA XOB
LFx
OTHER DPLL
FEEDBACK TDC
DIGITAL
CROSS
POINT
MUX
SYSTEM
CLOCK
AD9546
fS
AUXILIARY
REFERENCE TDCS
DPLL
APLL
OUTPUT
DISTRIBUTION
INVERSE USER TIME
STAMPERS (IUTS)
TIME STAMP PROCESSOR,
DIGITAL LOOP FILTER,
TUNING WORD PROCESSOR,
AND NCO
PFD
AND
VCO
÷2
Q
LOOP
FILTER
AUXILIARY NCOS
R
OUTxyN/
OUTxyP
PHASE
BUILD
OUT
FEEDBACK
TDC
MUX
N + FRAC/MOD
REFx
M DIVIDER
0
1
TDC
N DIVIDER
REFERENCE TDCS
Figure 83. Phase Buildout PLL Configuration
Rev. 0 | Page 116 of 205
Data Sheet
AD9546
frequency (fREFx) and the other with respect to the output
frequency (fOUTx). The user must ensure that fNCO, fVCO, fTDC, and
PFD are within their specified frequency limits.
INTERNAL ZERO DELAY (HITLESS) MODE
Figure 84 shows the internal zero delay configuration. Internal
zero delay mode is a hitless mode (rather than phase buildout
operating mode), in which the AD9546 gradually changes the
output phase as DPLLx realigns to the phase of the new
reference. The user has some control over the magnitude of the
output disturbance via the phase slew limit of the AD9546 and
tuning word clamp features (see the Phase Slew Rate Limit
section and Tuning Word Offset Clamp section, respectively).
f
f
f
f
TDC = fREFx/R
= fOUTx/N
(8)
(9)
NCO = fPFD = (2 × Q/M) × fOUTx
= ((2 × Q × N)/(R × M)) × fREFx
VCO = 2 × Q × fOUTx
In internal zero delay mode, the feedback path of the PLL origi-
nates at the output of a user selected Q divider in the output
distribution section. The user selects the feedback Q divider
selection via the translation profile registers. To select the
feedback Q divider, use Bits[4:0] (unsigned integer) in the
register at the start address of the translation profile in Table 74
plus an offset of 2 (decimal).
= (2 × Q × N/R) × fREFx
(10)
When the reference source is auxiliary NCO 0, auxiliary NCO 1,
IUTS 0, IUTS 1, or the feedback from the other DPLL, then in
Equation 8 through Equation 10, use R = 1.
Table 77. Zero Delay Feedback Path Selection
Bits[4:0]
Internal Zero
Delay
External Zero
Delay
(Decimal)
DPLL
The decimal value of Bits[4:0] selects the Q divider per the
Internal Zero Delay column in Table 77. With DPLL N divider
set for integer only (that is, no fractional component—a
requirement for zero delay operation), the internal zero delay
configuration ensures no phase offset between the reference
TDC and the output of the user selected Q divider.
0
0
1
1
2
2
3
3
4
4
5
5
6
6
7
7
0
1
0
1
0
1
0
1
0
1
0
1
0
1
0
1
OUT0AP
OUT1AP
OUT0AN
OUT1AN
OUT0BP
OUT1BP
OUT0BN
OUT1BN
OUT0CP
Not applicable
OUT0CN
REFA
REFA
REFAA
REFAA
REFB
REFB
REFBB
REFBB
The frequency translation factor for internal zero delay mode is
fOUTx
fREFx
N
R
Auxiliary REF0
Auxiliary REF0
Auxiliary REF1
=
To provide true hitless operation, internal zero delay mode
requires a constraint on the relationship between fREFx and fOUTx
Not applicable Auxiliary REF1
Not applicable Auxiliary REF2
Not applicable Auxiliary REF2
Not applicable Auxiliary REF3
Not applicable Auxiliary REF3
.
Namely, fOUTx/fREFx must be an integer greater than or equal to 1.
The NCO, VCO, TDCs, and the APLL PFD inputs each have
specific frequency limits. The following formulas relate fNCO
VCO, fTDC, and fPFD to fREFx, fOUTx, and divider values. Note that
each formula has two solutions: one with respect to the input
,
f
8 to 31
Not applicable Not applicable Not applicable
XOA XOB
LFx
OTHER DPLL
FEEDBACK TDC
DIGITAL
CROSS
POINT
MUX
SYSTEM
CLOCK
AD9546
fS
AUXILIARY
REFERENCE TDCS
DPLL
APLL
OUTPUT
DISTRIBUTION
INVERSE USER TIME
STAMPERS (IUTS)
TIME STAMP PROCESSOR,
DIGITAL LOOP FILTER,
TUNING WORD PROCESSOR,
AND NCO
PFD
AND
VCO
÷2
Q
LOOP
FILTER
AUXILIARY NCOS
R
OUTxyN/
OUTxyP
FEEDBACK
TDC
MUX
N + FRAC/MOD
REFx
M-DIVIDER
INTERNAL ZERO DELAY
0
1
TDC
N DIVIDER
REFERENCE TDCS
Figure 84. Internal Zero Delay PLL Configuration
Rev. 0 | Page 117 of 205
AD9546
Data Sheet
The DPLL provides the user with two status bits to indicate
when the DPLL enters or exits hitless operation. When the
DPLL enters hitless operation, Bit 4 latches to Logic 1 in
Register 0x3011 for DPLL0 or Register 0x3016 for DPLL1.
Likewise, when the DPLL exits hitless operation, Bit 3 latches to
Logic 1 in Register 0x3011 for DPLL0 or Register 0x3016 for
DPLL1. Because these are latched status bits, to obtain visibility
of subsequent DPLL state transitions into and out of hitless
operation, the user must clear the corresponding bits in the
IRQ map (Bit 4 and Bit 3, respectively, in Register 0x200C for
DPLL0 and Register 0x2011 for DPLL1).
where:
Ry is the divide ratio of the R divider associated with REFy.
REFy refers to any reference input (other than the REFx input).
Rx is the divide ratio of the R divider associated with REFx.
REFx refers to any reference input (other than the REFy input).
Z is the external frequency translation factor such that
fREFy
Z =
fOUTx
To provide true hitless operation, external zero delay mode
requires a constraint on the relationship between fREFx and fREFy
.
Namely, fREFy/fREFx must be an integer greater than or equal to 1.
EXTERNAL ZERO DELAY (HITLESS) MODE
The NCO, VCO, TDCs, and the APLL PFD inputs each have
Figure 85 shows the external zero delay configuration appears
in. External zero delay mode, like internal zero delay mode, is a
hitless operating mode (see the Internal Zero Delay (Hitless)
Mode section). As such, the same status bits associated with
entering and exiting hitless operation still apply.
specific frequency limits. The following formulas relate fNCO
,
f
VCO, fTDC, and fPFD to Z, fREFx, fOUTx, and divider values. Note that
each formula has two solutions: one with respect to the input
frequency (fREFx) and the other with respect to the output
frequency (fOUTx). The user must ensure that fNCO, fVCO, fTDC, and
In external zero delay mode, the feedback path of the PLL is via an
external connection from an appropriate OUTxy output to a REFx
or an auxiliary REFx input. The user chooses the desired input
path via the same 5-bit integer used to select the feedback Q
divider in the internal zero delay configuration, but with the
decimal value applying to the External Zero Delay column in
Table 77. Typically, this feedback path is merely a direct connec-
tion, but the external path may also include an additional
frequency translation component. To discriminate between the
normal and feedback reference inputs in Figure 85, they appear
as REFx and REFy, respectively. Although Figure 85 shows REFx
and REFy, auxiliary REFx, and auxiliary REFy are equally valid.
f
PFD are within their specified frequency limits.
f
f
f
TDC = fREFx/Rx
= (Z/Ry) × fOUTx
(11)
(12)
(13)
NCO = fPFD = ((2 × Q × Ry)/(Rx × M × Z)) × fREFx
= (2 × Q/M) × fOUTx
VCO = ((2 × Q × Ry)/(Rx × Z)) × fREFx
= 2 × Q × fOUTx
When the reference source is auxiliary NCO 0, auxiliary
NCO 1, or the feedback from the other DPLL, then in Equation
11 through Equation 13, use Rx = 1.
The frequency translation factor from REFx (auxiliary REFx) to
OUTxyP/OUTxyN for external zero delay mode is
Although the feedback divider (N divider) does not appear in
Figure 85, the N divider requires special treatment for external
zero delay operation. See the DPLL Feedback Divider (N
Divider) section for details.
fOUTx
fREFx Rx × Z
Ry
=
XOA XOB
LFx
OTHER DPLL
FEEDBACK TDC
DIGITAL
CROSS
POINT
MUX
SYSTEM
CLOCK
AD9546
fS
AUXILIARY
REFERENCE TDCS
DPLL
APLL
OUTPUT
DISTRIBUTION
INVERSE USER TIME
STAMPERS (IUTS)
TIME STAMP PROCESSOR,
DIGITAL LOOP FILTER,
TUNING WORD PROCESSOR,
AND NCO
PFD
AND
VCO
÷2
Q
LOOP
FILTER
AUXILIARY NCOS
OUTxyN/
OUTxyP
REFx
REFy
R
R
M DIVIDER
REFERENCE TDCS
Z
NOTE
1. REFx AND REFy CAN EACH BE ANY OF THE REFERENCE INPUTS, BUT NOT THE SAME REFERENCE INPUT.
Figure 85. External Zero Delay PLL Configuration
Rev. 0 | Page 118 of 205
Data Sheet
AD9546
SOURCE PROFILES
SWITCHOVER
CONTROL
SOURCE PROFILES OVERVIEW
AD9546
The source profiles give the user a method to define how an
input source, when selected by a DPLL channel via the active
translation profile, interacts with the DPLL. The user has access
to the 14 source profiles (one profile for each source) via the
register map with each source profile having a start and end
address per Table 78. Each source profile gives the user control
of the following parameters:
REFERENCE 1
REFERENCE 2
REFx
Figure 86. External Reference Switching
Phase transient threshold detection works by monitoring the
output of the DPLL phase detector for phase transients but in a
manner that is somewhat jitter tolerant. Otherwise, the phase
transient threshold detector is prone to false positives.
•
•
•
•
•
DPLL phase lock detector
DPLL frequency lock detector
Phase step limit
Skew adjustment
Initial phase skew refinement steps
To control the phase transient detector, use Bits[31:0]
(unsigned) of the appropriate source profile at the start address
shown in Table 78 plus an offset of 10 to 13 (decimal).
Bits[31:0] constitute the phase step threshold value. A phase
step threshold value of zero (default) disables the phase
transient threshold detector, whereas a nonzero value serves to
activate the phase transient threshold detector as well as define
the desired threshold (ΦTHRESH) in units of picoseconds per the
following equation:
Table 78. Source Name vs. Source Profile Address Range
Source Name
Start Address
0x0800
0x0820
0x0840
0x0860
0x0880
0x08A0
0x08C0
0x08E0
0x0900
0x0920
0x0940
0x0960
0x0980
0x09A0
End Address
0x0811
0x0831
0x0851
0x0871
0x0891
0x08B1
0x08D1
0x08F1
0x0911
0x0931
0x0951
0x0971
0x0991
0x09B1
REFA
REFAA
REFB
REFBB
Auxiliary NCO 0
Auxiliary NCO 1
DPLL0
DPLL1
IUTS 0
Φ
THRESH = Phase Step Threshold × 10−12
(14)
The phase transient threshold detector is not active unless the
DPLL indicates frequency locked status.
Determine the value of the phase step threshold value necessary
for a 12 ns limit (that is, ΦTHRESH = 12 × 10−9). Solving
Equation 14 for phase step threshold yields
Phase Step Threshold = ΦTHRESH/10−12
= (12 × 10−9)/10−12
IUTS 1
Auxiliary REF0
Auxiliary REF1
Auxiliary REF2
Auxiliary REF3
= 12,000
= 0x 0000 2EE0 (hexadecimal)
A source profile links to a DPLL translation profile (see the
Translation Profiles section) via Bits[4:0] of the translation
profile register start address shown in Table 74 plus an offset of
1 (decimal). The selected source (per Bits[4:0] in the associated
translation profile) identifies the source name in Table 78.
To reduce the likelihood of threshold violations induced by
jitter, the user must choose ΦTHRESH to be at least two times the
expected rms jitter (σJITTER) associated with the input reference
signal.
Φ
THRESH ≥ 2 × σJITTER
(15)
DPLL PHASE/FREQUENCY LOCK DETECTOR
As such, in the previous example for a phase step threshold of
12,000, an input signal with 12 ns rms jitter is likely to produce
false positives, because it violates the inequality in Equation 15.
To reduce the likelihood of a false positive, the inequality in
Equation 15 implies that a phase step threshold of 24,000 is a
better choice. However, even with a value of 24,000, there is still
a slight probability of a jitter sample exceeding 2 × σJITTER. As
such, multiplying σJITTER by four to six is an even better choice.
See the DPLL Lock Detectors section for details concerning the
phase and frequency detectors of the DPLL.
PHASE STEP LIMIT
Although the AD9546 can switch between multiple reference
inputs, some applications use only one input and handle
reference switching externally (see Figure 86). This
arrangement forfeits the ability of the AD9546 to mitigate the
output disturbance associated with a reference switchover,
because the reference switchover is not under the control of the
AD9546. However, the AD9546 offers a phase transient
threshold detection feature to help identify when an external
reference switchover occurs and acts accordingly.
When a phase transient occurs that exceeds the prescribed
value, one or both of the following two events occur, depending
on the state of Bit 7 of Register 0x2105 (for PLL0) and
Register 0x2205 (for PLL1):
•
•
The DPLL initiates a new acquisition sequence
The reference monitor is reset
Rev. 0 | Page 119 of 205
AD9546
Data Sheet
When Bit 7 is Logic 0 (default), detection of a phase step causes
only the first event to occur. By initializing a new DPLL
acquisition sequence, the DPLL can take advantage of the fast
acquisition feature (see the DPLL Fast Acquisition Options
section), assuming it is active, which is helpful for very low loop
bandwidth applications. In addition, a new acquisition manages
the impact of the phase step by either building out the phase or
slewing to the new phase in a hitless manner.
Monitor section). Manual reference invalidation imposes the
least impact on the steady state operation of the device. The
only steady state impact is that the validation timer of the
associated reference must be set to a duration that is longer
(with suitable margin) than the duration between the assertion
of the forced fault condition and the occurrence of the external
event.
SKEW ADJUSTMENT
When control Bit 7 is Logic 1, detection of a phase step causes
both events to occur. Because exceeding the phase step
threshold in this case implies an external switch to a new
reference, resetting the reference monitor forces it to establish
new reference statistics (see the Reference Monitor section).
Skew adjustment allows the user to associate a fixed phase
offset with a reference input, which is useful in applications
with redundant GNSS or GPS reference sources, for example.
That is, a user can have two or more GNSS/GPS sources that
have identical frequency but exhibit a fixed time offset due to a
mismatch between antenna cable lengths.
The phase transient threshold detector provides the user with a
status indicator for the occurrence of a phase step detection via
Bit 0 of Register 0x3011 and Register 0x3016 (for PLL0 and
PLL1, respectively). Because these status indicators are latched
IRQ bits, the user must clear them via Bit 0 of Register 0x200C
and Register 0x2011 (for PLL0 and PLL1, respectively) to
obtain visibility of subsequent threshold violations detected by
the phase step detector (see the Interrupt Request (IRQ)
section).
To activate the skew adjustment feature, use Bits[23:0] (signed)
of the appropriate source profile at the start address shown in
Table 78 plus an offset of 14 to 16 (decimal). Bits[23:0]
constitute the phase skew value. A phase skew value of zero
(default) disables the phase skew adjustment feature, whereas a
nonzero value serves to activate the phase skew adjustment
feature and define the desired phase skew in units of ps. The
time skew associated with phase skew value is given by the
following equation:
Mitigating Phase Step Limit False Positives
Time Skew = Phase Skew × 10−12
(17)
When enabled, the phase transient threshold detector operates
continuously while the associated reference is the active
reference for the DPLL (DPLL0 or DPLL1), assuming the DPLL
is frequency locked. As such, any phase disturbance at the input
to the phase detector, including user induced phase adjustment
per the DPLL Phase Offset Control section and the Skew
Adjustment section, is subject to violating the threshold of the
phase transient threshold detector. To mitigate false triggering
of the phase transient threshold detector (when enabled) due to
intentional phase adjustments, the user can employ the phase
slew rate limiter (see the Phase Slew Rate Limit section).
For example, determine the value of phase skew bit necessary
for a time skew of –35 ns. Solving Equation 17 for phase skew
yields the following:
Phase Skew = Time Skew/10−12
= (–35 × 10−9)/10−12
= –35,000
= 0x FF 7748 (hexadecimal)
INITIAL PHASE SKEW REFINEMENT STEPS
The following formula relates the maximum phase slew rate
(MPSR) assigned to the phase slew rate limiter necessary to
prevent inadvertent triggering of the phase transient threshold
detector due to a user induced phase disturbance:
Whenever the AD9546 performs a reference switchover, it
performs an initial assessment of the time offset between the
old and new reference. The device does not use the phase of the
new reference to determine the phase offset but compares the
time stamps of the feedback TDC with the time stamps from
the TDC of the new reference (see Figure 87). This comparison
allows the device to insert an opposing time offset into the
servo loop of the DPLL, a phase buildout operation (see the
Frequency Translation Loops section).
MPSR ≤ (P × f)/7
where:
P is the phase transient threshold detector limit (in ps).
(16)
f is the frequency (in Hz) at the input of the DPLL phase detector.
The inequality in Equation 16 ignores other contributors to
phase error such as jitter, frequency offset, and propagation
delay variation. Because MPSR in Equation 16 constitutes an
upper limit, it is recommended to use a lower value of MPSR to
provide margin (a 25% reduction is reasonable).
For a phase buildout translation loop, initial phase offset
insertion is an essential component of a reference switch over
requiring phase buildout. However, the AD9546 also supports
hitless translation profiles, for which phase buildout is not
necessary. The AD9546 has the capability of initial phase offset
insertion to build out the initial phase offset for a hitless
switchover as well. By applying phase buildout at the start of a
hitless switchover, the AD9546 can significantly reduce the
time necessary to complete a hitless switchover acquisition.
If the user has advance knowledge of the timing of an external
event such as the switching of the reference input clock source
via an external mux, rather than using the phase transient step
detector, a better solution is to invalidate the associated
reference manually (see the Invalidate section of the Reference
Rev. 0 | Page 120 of 205
Data Sheet
AD9546
Because the feedback path of the DPLL includes the loop filter,
the feedback signal has the benefit of reduced jitter and an
inherent resistance to change (a consequence of the typically
narrow bandwidth of the loop filter). As such, when the
AD9546 switches to a new reference, the phase and frequency
of the old reference tends to persist in the feedback path of the
DPLL. The feedback path persistence provides some time for
the DPLL to compare the new reference signal to the old one.
single sampled edge, which includes the contribution of any jitter
present on the reference signal.
A nonzero phase skew refinement steps value, K, sets the
number of phase samples the AD9546 analyzes as part of the
phase skew refinement process. That is, instead of taking the
first phase sample (jitter included) as the phase buildout value,
the phase skew refinement feature processes the first K phase
samples to assess the reference jitter and to determine the phase
buildout value. As such, the phase skew refinement feature
extends the time required to determine a phase buildout value
following a reference switchover, but the extra time yields a
more accurate phase buildout value.
As shown in Figure 87, the reference likely exhibits jitter. The
presence of jitter implies uncertainty in the measured time offset
between the feedback and reference signals. This uncertainty, in
turn, leads to a potential error in the determination of the
correct phase buildout value in the DPLL.
The phase skew refinement process operates under the
assumption that the feedback and reference clocks are of the same
frequency, or at least very close. If they are not the same
frequency, the frequency offset appears as a linear phase slew,
which quickly becomes the dominant phase contributor and
masks any jitter that may be present on the reference signal.
Therefore, a reference switchover between references of
dissimilar frequency results in degraded performance of the
phase skew refinement feature.
To help mitigate jitter induced errors in the assessment of the
phase buildout value, the AD9546 provides a phase skew
refinement feature. To activate the phase skew refinement
feature, use Bits[7:0] (unsigned) of the appropriate source
profile at the start address shown in Table 78 plus an offset of
17 (decimal). Bits[7:0] constitute the phase skew refinement
steps value.
A phase skew refinement steps value of zero (default) disables
the phase skew refinement feature. With the phase skew
refinement feature disabled, the phase buildout value is an
unfiltered snapshot of the reference and feedback time offset at a
SUPPRESSED
JITTER
THE OLD REFERENCE SIGNAL TEMPORARILY PERSISTS IN THE FEEDBACK SIGNAL
FEEDBACK
SIGNAL
REFERENCE JITTER
REFERENCE
SIGNAL
NOMINAL TIME OFFSET BETWEEN OLD REFERENCE SIGNAL FEEDBACK AND NEW REFERENCE SIGNAL
UNCERTAINTY
REFERENCE
SWITCHOVER
Figure 87. DPLL Reference and Feedback Signals
Rev. 0 | Page 121 of 205
AD9546
Data Sheet
DIGITAL PLL (DPLL)
DPLL OVERVIEW
DPLL LOOP CONTROLLER
The DPLL is a completely digital implementation of a PLL.
Figure 88 shows the fundamental building blocks of an APLL
and a DPLL. An APLL typically relies on a VCO as the
frequency element for generating an output signal, where the
output frequency depends on an applied dc voltage. A DPLL,
conversely, uses an NCO, which relies on a digital frequency
tuning word (FTW) to produce the output frequency. A VCO
inherently produces a timing signal because the VCO is, by
definition, an oscillator, whereas the AD9546 NCO requires an
external timing source, the system clock. The fundamental
difference between an APLL and a DPLL is that the VCO in an
APLL can tune to any frequency within its operating
bandwidth, whereas the NCO in a DPLL can only tune to
discrete frequencies (by virtue of the FTW).
The DPLL has several operating modes: freerun, holdover, and
active. To ensure seamless transition between modes, the DPLL
has a loop controller. The loop controller sets the appropriate
DPLL operating mode based on the prevailing requirements of
automatic reference switching or manual control settings.
Switchover
Switchover occurs when the loop controller switches directly
from one input reference to another (see the Reference
Switching section). To handle a reference switchover, the
AD9546 briefly enters holdover mode, loads the DPLL
parameters from the source profile associated with the new
reference, and then immediately recovers.
Holdover or Freerun
ANALOG
Typically, the holdover or freerun state is in effect when all
input references are invalid. However, the user can force the
holdover mode even when one or more references are valid by
programming Bit 1 to Logic 1 of Register 0x2105 (for DPLL0)
or Register 0x2205 (for DPLL1). In holdover mode, the output
frequency remains fixed (to the extent of the stability of the
system clock). The accuracy of the AD9546 in holdover mode
depends on the device programming and availability of the
tuning word history.
PHASE
FIXED LOOP
ERROR BANDWIDTH VOLTAGE
PHASE
DETECTOR
LOOP
FILTER
VCO
DIVIDER
APLL
To force freerun mode, see the Freerun Tuning Word section.
NUMERIC
FREQUENCY
PROGRAMMABLE
LOOP BANDWIDTH
TUNING WORD
When the DPLL is in holdover or freerun mode, there is no
active translation profile available (see the Frequency Translation
Loops section). However, the DPLL is still attached to a
translation profile by virtue of Bits[2:0] of Register 0x102A (for
DPLL0) or Register 0x142A (for DPLL1). Bits[2:0] define the
translation profile with a value of x, where x is 0 to 5, that
associates with the DPLL when in holdover or freerun mode.
NUMERIC
COEFFICIENTS
NUMERIC
PHASE
ERROR
SYSTEM
CLOCK
NCO
DIGITAL
PHASE
DETECTOR
DIGITAL
LOOP
FILTER
DIVIDER
Because the DPLL attaches to the profile specified by Bits[2:0],
the recommendation is for the user to program Bits[2:0] with
one of the translation profiles the user has configured. For
example, if the user configures Translation Profile 0,
Translation Profile 2, and Translation Profile 5 for use by
DPLL0, program Bits[2:0] with a decimal value of 0, 2, or 5.
DPLL
Figure 88. APLL vs. DPLL
The DPLLs in the AD9546 have a phase detector (that operates
on numeric time stamps at its reference and feedback inputs
rather than physical clock signals) and a digital loop filter with
programmable bandwidth. The digital loop filter output yields
a digital FTW (instead of a dc voltage, as in the case of an
analog PLL) that produces a corresponding NCO output
frequency.
Recovery from Holdover
When in holdover mode, and an enabled translation profile
becomes available, the device exits holdover operation. The
loop controller restores the DPLL to closed-loop operation, which
begins to lock onto the selected reference, and sequences the
recovery of all loop parameters based on the profile settings for
the active reference.
If Bit 1 = 1 in Register 0x2105 (for DPLL0) or Register 0x2205
(for DPLL1), the device does not automatically exit holdover
when a valid translation profile is available. However,
automatic recovery of a valid translation profile can occur after
clearing Bit 1.
Rev. 0 | Page 122 of 205
Data Sheet
AD9546
When the desired N divider divide ratio requires a fractional
DPLL FEEDBACK DIVIDER (N DIVIDER)
component, the FRAC and MOD control elements are
necessary. To program the value of FRAC, use Bits[23:0]
(unsigned) of the appropriate translation profile at the start
address shown in Table 74 plus an offset of 16 to 18 (decimal).
To program the value of MOD, use Bits[23:0] (unsigned) of the
appropriate translation profile at the start address shown in
Table 74 plus an offset of 19 to 21 (decimal).
The DPLL feedback N divider consists of user-programmable
integer part and fractional part. The fractional part, however, is
only available in buildout mode (see the Frequency Translation
Loops section).
Rather than offering direct programming control of the
N divider, the AD9546 makes use of translation profiles (see the
Translation Profiles section). Within each translation profile,
the following four control elements are associated with the N
divider:
For the special case of FRAC = 0, the user must also program
MOD = 0 to ensure proper device functionality. Furthermore,
the value of FRAC must be less than the value of MOD.
•
•
•
•
Buildout N divider integer part (N buildout)
Hitless N divider integer part (N hitless)
Buildout fractional multiplier (FRAC)
Buildout fractional modulus (MOD)
For phase buildout applications that require the N divider to be
integer only, program MOD = 0.
For example, suppose the desired fractional part of the N divider
divide ratio is 0.387429. Then, the corresponding fraction is
FRAC/MOD = 387,429/1,000,000. Generally, the user reduces
the fraction, FRAC/MOD, to its lowest terms, but in this
example, the fraction is irreducible. Therefore, FRAC = 387,429
(0x 0005 E965 hexadecimal) and MOD = 1,000,000 (0x 000F
4240 hexadecimal).
N buildout and N hitless are 32-bit, unsigned integers (only
Bits[29:0] are meaningful because Bits[31:30] must be zero).
FRAC and MOD are 24-bit, unsigned integers.
There are 12 occurrences of the four control elements
coinciding with the 12 translation profiles. Table 74 in the
Frequency Translation Loops section shows the address ranges
associated with the 12 translation profiles. Note that there are
constraints on programming the N divider, as described in the
Frequency Translation Loops section.
N Divider in Internal Zero Delay Mode
When operating in zero delay (hitless) mode, the AD9546
behaves like an Integer N PLL. Thus, the N divider divide ratio
must be an integer (that is, no fractional part). As such, the
AD9546 ignores the FRAC and MOD control elements when
operating in zero delay mode.
N Divider for Phase Buildout Mode
In phase buildout mode, the N divider divide ratio consists of
an integer and fractional part, as follows:
Note that zero delay operation makes use of a different integer
part N divider control element than buildout operation. Thus,
zero delay operation requires programming of N hitless rather
than N buildout. To program the value of N hitless, use
Bits[31:0] (unsigned) of the appropriate translation profile at
the start address shown in Table 74 plus an offset of 8 to 11
(decimal). The programmed value of N hitless constitutes one
less than the desired divide ratio. That is, an N hitless value of 0
constitutes an integer divide ratio of 1. Therefore, N hitless
provides integer divide ratios from 1 to 230 (1,073,741,824).
N divider divide ratio = (N buildout + 1) + FRAC/MOD
To program the value of N buildout, use Bits[31:0] (unsigned)
of the appropriate translation profile at the start address shown
in Table 74 plus an offset of 12 to 15 (decimal). The programmed
value of N buildout constitutes one less than the desired integer
part of the divide ratio. That is, an N buildout value of 0
constitutes an integer divide ratio of 1. Therefore, N buildout
provides integer divide ratios from 1 to 230 (1,073,741,824).
For example, for an N divider divide ratio 1,000,000,
N buildout = 999,999 (0x 000F 423F hexadecimal).
Note that operating in zero delay (hitless) mode still requires
the user to program N buildout to ensure proper operation of
the DPLL loop filter. However, the value of N buildout is not
the same as N hitless. For internal zero delay operation,
N buildout relates to N hitless per the following (rounded to the
nearest integer):
Note that changing the value of N buildout or MOD in a
currently active translation profile causes the associated DPLL
to reacquire its reference input signal. However, the DPLL can
accommodate on-the-fly changes to the FRAC value without
causing a reacquisition.
N Buildout = N Hitless × 2 × Q/M
where:
Q is the divide ratio of the Q divider.
M is the divide ratio of the M divider (see Figure 82).
Rev. 0 | Page 123 of 205
AD9546
Data Sheet
DPLL Loop Filter Base Coefficients
N Divider in External Zero Delay Mode
The response characteristic of a digital loop filter is a function of
digital coefficients. In the case of the AD9546, there are base
coefficients that establish a normalized frequency response
characteristic. The user scales the normalized response to
achieve an absolute response (see the DPLL Loop Filter
Bandwidth section).
There is no internal feedback divider associated with the DPLL
in external zero delay operation because the feedback path is
external. Therefore, there is no requirement to program the
N hitless control element. However, the user must still program
N buildout to ensure proper operation of the DPLL loop filter.
Program N buildout as follows (refer to Figure 85):
The AD9546 allows two independent sets of base coefficients,
Loop Filter 0 (LF0) and Loop Filter 1 (LF1). Each set consists of
four coefficients: alpha, beta, gamma, and delta, where Alpha 0
associates with LF0, Alpha 1 with LF1, and so on. Each of the
four coefficients has two components: a significand and an
exponent. The user has access to the coefficients via the
addresses and bit ranges shown in Table 79.
N Buildout = round((2 × Q × Ry × fOUTx)/(M × fREFy)) + 1
where:
round() is a function to round the value contained within the ()
to the nearest integer.
Q is the divide ratio of the Q divider.
Ry is the divide ratio of the R divider associated with the REFy
input in Figure 85.
Each DPLL has access to the LF0 and LF1 base coefficient sets,
per Figure 90. Thus, DPLL0 and DPLL1 can use either of the
base coefficient sets. The translation profiles (see the
Translation Profiles section) establish which base coefficient set
a DPLL uses. To select LF0 or LF1, use Bit 7 of the appropriate
translation profile at the start addresses shown in Table 74 plus
an offset of 3 (decimal). Logic 0 (default) selects LF0, and
Logic 1 selects LF1.
f
OUTx is the output frequency at the OUTxyP and/or OUTxyN
output.
M is the divide ratio of the M divider.
f
REFy is the frequency at the REFy input.
The calculated value of N buildout cannot be less than 1 or
greater than 1,073,741,824.
DPLL LOOP FILTER
The loop filter for DPLL0 and DPLL1 consists of a digital
infinite impulse response (IIR) filter architecture. The digital
filter is analogous to the third-order analog loop filter shown in
Figure 89, where input is on the left (from the phase detector)
and output is on the right (to the VCO).
Table 79. Loop Filter Base Coefficients
Coefficient
Address Range
0x0C00 to 0x0C01
0x0C02
0x0C03 to 0x0C04
0x0C05
Bit Range
0 to 15
0 to 7
0 to 15
0 to 7
0 to 15
0 to 7
0 to 15
0 to 7
0 to 15
0 to 7
0 to 15
0 to 7
0 to 15
0 to 7
0 to 15
0 to 7
Alpha 0 Significand
Alpha 0 Exponent
Beta 0 Significand
Beta 0 Exponent
R3
R2
C2
C1
C3
Gamma 0 Significand
Gamma 0 Exponent
Delta 0 Significand
Delta 0 Exponent
Alpha 1 Significand
Alpha 1 Exponent
Beta 1 Significand
Beta 1 Exponent
0x0C06 to 0x0C07
0x0C08
0x0C09 to 0x0C0A
0x0C0B
0x0C0C to 0x0C0D
0x0C0E
0x0C0F to 0x0C10
0x0C11
Figure 89. Third-Order Analog Loop Filter
Whereas the response of an analog loop filter is a function of
physical components subject to thermal drift and aging, the
response of a digital loop filter is a function of numeric coefficients
and is not subject to thermal drift and aging. Figure 90 is a block
diagram showing the digital loop filters and their control
elements.
Gamma 1 Significand
Gamma 1 Exponent
Delta 1 Significand
Delta 1 Exponent
0x0C12 to 0x0C13
0x0C14
0x0C15 to 0x0C16
0x0C17
For completeness, Figure 90 shows the fast acquisition control
block of the AD9546 as part of the loop filter control because
the fast acquisition feature dynamically adjusts the loop
bandwidth as part of the fast acquisition process (see the DPLL
Fast Acquisition Options section).
The AD9546 default settings provide values for LF0 and LF1,
resulting in a default response characteristic. The LF0 default
values yield an open-loop response characteristic with 70°
phase margin (nominal). The LF1 default values yield an open-
loop response characteristic with an 88.5° phase margin
(nominal). Furthermore, the LF1 default coefficients yield a
nearly flat closed-loop response with <0.1 dB peaking.
Generally, the response characteristics of a digital filter depend
solely on the numeric coefficients. However, the DPLL loop
filters are special in that they require additional input
parameters: a bandwidth scale factor and the numeric value of
the DPLL feedback divider (see the DPLL Feedback Divider (N
Divider) section for details). The DPLL automatically resets the
internal state of the loop filter when the loop is not active (for
example, during entry into holdover or freerun mode).
Rev. 0 | Page 124 of 205
Data Sheet
AD9546
DIGITAL
LOOP
FILTER
FROM
DIGITAL PFD
TO
NCO
PHASE LOCK
DETECT
DPLL0
LOOP
FILTER
COEFFICIENTS
BW
ADJ
FAST
ACQ 0
REG 0x120F TO 0x120C, BITS[31:0]
BW
REG 0x1207 TO 0x1204, BITS[31:0]
REG 0x1203, BIT 7
0
1
1
LOOP FILTER 0
COEFFICIENTS
LOOP FILTER 1
COEFFICIENTS
REG 0x0C0B TO 0x0C00, BITS[95:0]
LF0
REG 0x0C17 TO 0x0C0C, BITS[95:0] LF1
0
REG 0x1603, BIT 7
REG 0x1607 TO 0x1604, BITS[31:0]
BW
FAST
REG 0x160F TO 0x160C, BITS[31:0]
ACQ 1
BW
ADJ
LOOP
FILTER
COEFFICIENTS
PHASE LOCK
DETECT
DPLL1
DIGITAL
LOOP
FILTER
FROM
DIGITAL PFD
TO
NCO
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
2. EXCEPT FOR LF0 AND LF1, REGISTER ADDRESSES ARE SPECIFIC TO TRANSLATION PROFILE 0.0 AND 1.1
Figure 90. Digital Loop Filter Block Diagram
For example, given a desired loop bandwidth of 50 Hz, the
DPLL Loop Filter Bandwidth
required scale factor is
Scale Factor = 50 Hz/10−6 Hz
= 50,000,000
Because the base coefficients establish a normalized frequency
response, the loop filter requires a scale factor to set the
absolute frequency response. The scale factor, which sets the
bandwidth (BW), resides in the translation profiles (see the
Translation Profiles section), as shown in Figure 90. The scale
factor is Bits[31:0] (unsigned) residing in the appropriate
translation profile at the start address shown in Table 74 plus an
offset of 4 to 7 (decimal).
= 0x 02FA F080 (hexadecimal)
Note that the 32-bit scale factor implies a maximum DPLL loop
bandwidth of nearly 4.3 kHz (232 × 10−6 Hz). However, the
presence of the NCO gain tuning word filter in the loop (see the
NCO Gain Tuning Word Filter Bandwidth section) puts a
constraint on the maximum DPLL loop filter bandwidth. Table 80
relates the maximum DPLL loop filter bandwidth as a function of
the value of the selected NCO gain filter bandwidth. The active
translation profile of the DPLL determines whether the LF0 or
LF1 column applies for a given DPLL channel.
The frequency units associated with the scale factor depend on the
normalization of the LF0 and LF1 base coefficients (see the DPLL
Loop Filter Base Coefficients section). In the case of the default
base coefficients, the normalized frequency units are µHz
(10−6 Hz) for both LF0 and LF1.
To calculate the scale factor necessary to set the 0 dB open-loop
bandwidth, divide the desired loop bandwidth by the
normalized frequency associated with the selected set of base
coefficients, LF0 or LF1. In the case of the default LF0 and LF1
values, the normalized frequency is 10−6 Hz.
Rev. 0 | Page 125 of 205
AD9546
Data Sheet
Table 80. Maximum DPLL Loop Filter Bandwidth
input and feedback signals to numeric time stamps (see the
Time to Digital Converter (TDC) section for details).
NCO Gain Filter
Bandwidth
Selection
LF0 Maximum
Loop Bandwidth
(Hz)
LF1 Maximum
Loop Bandwidth
(Hz)
Although Figure 91 shows the N divider connected directly to
the NCO output, this diagram is a simplification of the actual
feedback path (see Figure 82 in the Frequency Translation
Loops section). However, regarding the operation and control
of the DPLL, this simplification is valid in the context of the
following paragraphs.
0
1850
305
1
2
3
4
5
6
7
8
925
152.5
76.3
38.1
19.1
9.53
462.5
231.3
115.6
57.81
28.91
14.45
7.227
3.613
1.807
0.9033
0.4517
0.2258
0.1219
0.0565
Frequency tuning of the DPLL is by virtue of an NCO, which
employs a sigma-delta modulator (SDM) architecture. The
SDM has an internal integer divider that divides down the
system clock frequency with the output of the divider
constituting the output of the NCO. The SDM effectively
modulates the modulus of this divider to produce an output
frequency that is a fractionally scaled down version of the
system clock frequency based on an input 48-bit FTW. Because
the NCO is SDM-based, it employs noise shaping that
redistributes its modulation noise away from the NCO output
frequency (the APLL, which follows the DPLL, suppresses the
out of band modulation noise of the SDM).
4.77
2.38
1.191
0.596
0.298
0.149
0.745
0.037
0.019
0.009
9
10
11
12
13
14
15
DPLL NCO
The output frequency of the NCO (fNCO) depends primarily on
the numeric value of the 48-bit FTW and the frequency of the
system clock (fS) per the following equation:
The DPLL NCO requires an external clock source. In the case
of the AD9546, the external clock source drives the XOA and
XOB input pins, from which an integrated PLL synthesizer
generates a clock signal that is approximately 2.4 GHz.
f
NCO = fS × FTW/248
For a given fS and a desired fNCO, compute FTW as
The DPLL normally operates in closed-loop fashion (that is, as
a PLL), which is the normal operating mode, wherein the loop
filter is the FTW source for the NCO. However, under certain
conditions, the DPLL operates in an open-loop configuration.
Figure 91 differentiates between the two configurations by
means of a switch. A loop controller determines whether the
DPLL is in closed-loop operation (switch closed) or open-loop
operation (switch open), while an FTW processor determines
the FTW source for the NCO.
FTW = round(248 × fNCO/fS)
(18)
where round(x) is a function to round x to the nearest integer.
The NCO automatically converts the 48-bit FTW into two
components: an integer part (INT) and a fractional part
(FRAC). INT and FRAC relate to FTW as
INT = floor(248/FTW)
FRAC = 2−40 × round(240 × ((248/FTW) − INT))
(19)
(20)
XOA XOB
where:
7 ≤ INT ≤ 13.
0.05 ≤ FRAC ≤ 0.95.
floor(x) is a function that leaves x unchanged if x is an integer.
Otherwise, x becomes the nearest integer in the negative
direction.
LOOP
SYSTEM
CLOCK
AD9546
CONTROLLER
FREERUN
LOCK
DETECTORS
TUNING WORD
REG 0x1005 TO 0x1000,
BITS[45:0]
NUMERIC
DIGITAL CROSS
POINT MUX
COEFFICIENTS
46
The constraints on INT and FRAC necessarily impose
limitations on the choice of FTW. For example, let fS =
2.30 GHz and fNCO = 245.76 MHz, which yields (per
Equation 18).
DIGITAL
PHASE
DIGITAL
LOOP
FTW
PROCESSOR
TDC
NCO
48-BIT
FTW
DETECTOR
FILTER
TDC
N DIVIDER
FTW = 30,076,213,163,657
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES ARE SPECIFIC TO DPLL0
Then, per Equation 19 and Equation 20,
INT = 9
Figure 91. DPLL Block Diagram
Figure 91 also shows the lock detectors (see the DPLL Lock
Detectors section). For details on the feedback divider, see the
DPLL Feedback Divider (N Divider) section. For clarity, Figure 91
also shows the digital cross point mux and TDCs that feed the
digital phase detector. The TDCs convert the rising edges of the
FRAC = 0.35872395833303016843274235725403
In this case, INT and FRAC satisfy their defined constraints.
Although the preceding example validates FTW for the given fS
and fNCO, the example does not necessarily validate FTW for a
Rev. 0 | Page 126 of 205
Data Sheet
AD9546
TUNING
WORD
given application. That is, the preceding example assumes fS
and fNCO are completely static values. However, fS is only as
stable as the oscillator or resonator at the XOA and XOB pins.
Furthermore, with the DPLL locked to an input reference
signal, fNCO tracks variations in the reference frequency.
Therefore, the user must assess variations on FTW for a given
application. That is, the user must consider upper and lower
FTW values, which lead to upper and lower INT and FRAC
values as well.
TRANSITION
TIME
FTW2
INPUT
OUTPUT
FTW1
For example, assume in the preceding example that input
frequency variations cause the FTW to vary by 0.5%, leading to
two FTW values that differ from 30,076,213,163,657 by 0.5%:
TIME
Figure 92. NCO Gain Tuning Word Filter Response
Lower FTW = 29,925,832,097,839
Upper FTW = 30,226,594,229,475
To control the filter bandwidth, use Bits[3:0] (unsigned integer)
in Register 0x1009 for DPLL0 and Register 0x1409 for DPLL1.
Table 81 shows how the value of Bits[3:0] relates to the
bandwidth and resulting transition time.
The upper and lower FTW values lead to the following INT and
FRAC values:
To prevent degradation of the phase margin associated with the
DPLL loop filter (see the DPLL Loop Filter section), the user
must be careful to choose an NCO gain tuning word filter
bandwidth from Table 81 that is at least 100 times greater than
the loop bandwidth of the DPLL. This value includes any
expansion of the DPLL loop filter bandwidth by the fast
acquisition block, if enabled (see the DPLL Fast Acquisition
Options section).
INTUPPER = 9
FRACUPPER = 0.3121631426201929571107029914856
INTLOWER = 9
FRACLOWER = 0.40575272194291756022721529006958
In this case, the upper and lower INT values and upper and
lower FRAC values satisfy the constraints on INT and FRAC.
The upper and lower INT values must be the same. Otherwise,
having different values implies that the upper and lower FTW
values cross an SDM integer boundary, which can lead to poor
spurious performance. There are two ways to remedy this
problem. The first, which is less workable, is to limit the
variation on the system clock frequency and the reference input
frequency. The second is to choose a new FTW value (and, by
implication, a new fNCO value). In either case, the goal is to
constrain the variation of the FTW such that the FTW yields
identical (and valid) upper and lower INT values, as well as
valid upper and lower FRAC values.
Table 81. NCO Gain Tuning Word Filter Bandwidth Selections
Bits[3:0]
3 dB Bandwidth (Hz) Transition Time (ms)
0
1
2
3
4
5
6
7
248,000
124,000
62,000
31,000
15,500
7800
3900
1900
970
0.003
0.006
0.013
0.026
0.051
0.102
0.204
0.419
0.820
1.62
8
9
The NCO applies adjustments to INT and FRAC as necessary
when the system clock compensation feature is active (see the
System Clock Compensation section).
490
10
11
12
13
14
15
240
120
61
30
15
7.6
3.32
6.63
13.0
26.5
53.1
105
NCO GAIN TUNING WORD FILTER BANDWIDTH
Although not explicitly shown in Figure 91, the NCO contains
a digital low-pass filter, the NCO gain tuning word filter. This
filter has a single-pole response similar to a simple resistor and
capacitor low-pass filter (see Figure 92), but with a variable gain
component that compensates for the nonlinear gain of the
NCO. The NCO gain tuning word filter reduces frequency
transients that can occur when the DPLL switches between
closed-loop and open-loop operating modes (active to
holdover, for example).
The NCO gain tuning word filter has implications when using
the NCO as a traditional, open-loop digital frequency synthesizer.
For example, when the DPLL is programmed for freerun mode
(see the Freerun Tuning Word section), the DPLL operates like
a traditional NCO. That is, the user can program different
freerun tuning words to synthesize different frequencies. In a
traditional NCO, programming a new tuning word results in an
instantaneous switch from the initial frequency to the new
frequency (like the input trace shown in Figure 92). However,
Rev. 0 | Page 127 of 205
AD9546
Data Sheet
in the case of the DPLL, when programming different tuning
words, the NCO transitions from one frequency to the next
smoothly based on the programmed bandwidth of the NCO
gain tuning word filter, as shown in Figure 92 (see Table 81 for
the transition time to 99% of a step input).
water level (vertical) vs. time (horizontal) and the resulting
lock/unlock states.
PREVIOUS
STATE
LOCKED
UNLOCKED
2047
1024
0
LOCK LEVEL
DPLL LOCK DETECTORS
RATE
DRAIN RATE
DPLL Phase Lock Detector
UNLOCK LEVEL
–1025
–2048
Each DPLL channel (DPLL0 and DPLL1) contains a completely
digital phase lock detector. The user controls the threshold
sensitivity and hysteresis of the phase lock detector via the
source profiles (see the Source Profiles section).
Figure 93. Lock Detector Diagram
The user has access to the 12-bit (signed) instantaneous water
level value of the phase lock detector via Register 0x3109 to
Register 0x310A (DPLL0) and Register 0x3209 to Register 0x320A
(DPLL1). As shown in Figure 93, the pertinent water level
values appear along the left side of the tub.
The phase lock detector indicates the phase lock status via Bit 1
of Register 0x3100 for DPLL0 and Register 0x3200 for DPLL1
(Logic 0 is unlocked and Logic 1 is locked). However, because
Bit 1 is dynamic in nature, the recommendation is to use the
interrupt request (IRQ) mechanism for phase lock indication
instead. The IRQ mechanism observes the state of Bit 1 and
latches the state transitions. Specifically, Bit 0 of Register
0x3010 for DPLL0 and Register 0x3015 for DPLL1 latches a
status change from phase unlocked to phase locked as a Logic 1.
Likewise, Bit 1 of the same registers latches a status change
from phase locked to phase unlocked as a Logic 1. Because Bit 0
and Bit 1 are latched bits, however, they may represent a
condition that is no longer true. Therefore, the user must clear
the phase locked and phase unlocked status via Bit 0 and Bit 1,
respectively, of Register 0x200B for DPLL0 and Register 0x2010
for DPLL1. Otherwise, the user may lose indication of
subsequent state transitions by the phase lock detector (see the
Interrupt Request (IRQ) section).
During any given PFD phase error sample, the lock detector
either adds water with the fill bucket or removes water with the
drain bucket (one or the other, but not both). The decision of
whether to add or remove water depends on the phase lock
threshold level specified by the user via Bits[23:0] (unsigned
integer) of the appropriate source profile at the start address
shown in Table 78 plus an offset of 0 to 2 (decimal). The value
of Bits[23:0] is the desired threshold in picoseconds. Thus, the
phase lock threshold extends from 0 ps to 16.7 µs and
represents the phase error at the output of the PFD. Though the
programming range supports 0 ps as a lower limit, in practice,
the minimum value must be greater than 50 ps.
The phase lock detector compares the absolute value of each
phase error sample at the output of the PFD to the programmed
phase threshold value. If the absolute value of the phase error
sample is less than or equal to the programmed phase threshold
value, the detector control logic adds one fill bucket into the
tub. Otherwise, the detector control logic removes one drain
bucket from the tub. The magnitude of the phase error sample
(polarity is ignored), relative to the phase threshold value,
determines whether to fill or drain the bucket.
The phase lock detector behaves in a manner analogous to
water in a tub (see Figure 93). The total capacity of the tub
is 4096 units, with −2048 denoting empty, 0 denoting the 50%
point, and +2047 denoting full. The tub also has a safeguard to
prevent overflow. Furthermore, the tub has a low water mark
at −1025 and a high water mark at +1024. To change the water
level, the phase lock detector adds water with a fill bucket or
removes water with a drain bucket.
Regarding the fill and drain process, an exception to normal
operation occurs when the phase slew limiter is active. When
the phase slew limiter is actively in the limiting process, the lock
detector inhibits fill events, allowing only drain events to occur.
The user specifies the size of the fill and drain buckets via the
source profiles. To specify the phase lock fill rate, use Bits[7:0]
(unsigned integer) of the appropriate source profile at the start
address shown in Table 78 plus an offset of 3 (decimal). To
specify the phase lock drain rate use Bits[7:0] of the appropriate
source profile at the start address shown in Table 78 plus an offset
of 4 (decimal).
When more filling is taking place than draining, the water level
in the tub eventually rises above the high water mark (1024),
which causes the lock detector to indicate lock. When more
draining is taking place than filling, the water level in the tub
eventually falls below the low water mark (−1024), which
causes the lock detector to indicate unlock. The ability to
specify the threshold level, fill rate, and drain rate enables the
user to tailor the operation of the lock detector to the statistics
of the timing jitter associated with the input reference signal.
Note that, for debug purposes, the user can make the fill or
drain rate zero to force the lock detector to indicate a lock or
unlock state, respectively.
The water level in the tub is what the lock detector uses to
determine the lock and unlock conditions. When the water
level is below the low water mark (−1025), the lock detector
indicates an unlock condition. Conversely, when the water level
is above the high water mark (1024), the lock detector indicates a
lock condition. When the water level is between the marks, the lock
detector holds its previous condition. Figure 93 shows this
concept with an overlay of an example of the instantaneous
Rev. 0 | Page 128 of 205
Data Sheet
AD9546
Whenever the AD9546 enters freerun or holdover mode, the
DPLL phase lock detector indicates an unlocked state.
Table 78 plus an offset of 5 to 7 (decimal). The value of
Bits[23:0] is the desired frequency lock threshold in ps. Thus,
the frequency lock threshold extends from 0 ps to 16.7 µs. The
frequency lock threshold represents the absolute value of the
period error between the reference and feedback signals at the
input to the DPLL as follows:
For more information on how to choose the appropriate phase
lock threshold, fill rate, and drain rate values for a given
application, refer to the AN-1061 Application Note, Behavior of
the AD9548 Phase and Frequency Lock Detectors in the Presence
of Random Jitter.
Frequency Lock Threshold = |Period Error|/10−12
For example, consider a nominal frequency at the reference
input to the DPLL of 80 kHz. Under a stable lock condition, the
frequency at the reference and feedback inputs to the DPLL are
equal. To configure the frequency lock detector to make fill or
drain decisions when the feedback and reference frequency at
the input to the DPLL differ by 100 Hz, establish the frequency
lock threshold for a 100 Hz deviation by choosing fREF = 80 kHz
and fFB = 80.1 kHz (or 79.9 kHz).
DPLL Frequency Lock Detector
The operation of the frequency lock detector is identical to that
of the phase lock detector, with the following two exceptions:
•
The fill or drain decision is based on the period deviation
between the reference of the DPLL and the feedback
signals instead of the phase error at the output of the PFD.
The frequency lock detector is unaffected by the state of
the phase slew limiter.
•
Frequency Lock Threshold = |Period Error|/10−12
= |1/fREF − 1/fFB|/10−12
Like the phase lock detector, the user has access to the 12-bit
(signed) instantaneous water level value of the frequency lock
detector via Register 0x310B to Register 0x310C (DPLL0) and
Register 0x320B to Register 0x320C (DPLL1). As shown in
Figure 93, the pertinent water level values appear along the left
side of the tub.
= |1/80,000 − 1/80,100|/10−12
= 15,605 (nearest integer)
= 0x 00 3CF5 (hexadecimal)
For more information on how to choose the appropriate
frequency lock threshold, fill rate, and drain rate values for a
given application, refer to AN-1061 Application Note.
The frequency lock detector indicates frequency lock status via
Bit 2 of Register 0x3100 for DPLL0 and Register 0x3200 for
DPLL1 (Logic 0 is unlocked, and Logic 1 is locked). However,
because Bit 2 is dynamic in nature, the recommendation is to
use the IRQ mechanism for frequency lock indication instead.
The IRQ mechanism observes the state of Bit 2 and latches the
state transitions. Specifically, Bit 2 of Register 0x3010 for
DPLL0 and Register 0x3015 for DPLL1 latches a status change
from frequency unlocked to frequency locked as a Logic 1.
Likewise, Bit 3 of the same registers latches a status change
from frequency locked to frequency unlocked as a Logic 1.
Because Bit 2 and Bit 3 are latched bits, however, they may
represent a condition that is no longer true. Therefore, the user
must clear the frequency locked and frequency unlocked status
via Bit 2 and Bit 3, respectively, of Register 0x200B for DPLL0
and Register 0x2010 for DPLL1. Otherwise, the user may lose
indication of subsequent state transitions by the frequency lock
detector (see the Interrupt Request (IRQ) section).
FREERUN TUNING WORD
The closed switch in Figure 91 indicates that the DPLL is
operating in closed-loop mode, where the loop filter delivers
FTWs in real time to the NCO. In open-loop operation, the
switch is in the open position and the FTW processor provides
a static FTW to the NCO. The loop controller opens or closes the
switch as needed. For example, when the DPLL is in freerun
mode (that is, Bit 0 = 1 of Register 0x2105 or Register 0x2205),
the loop controller opens the switch and the FTW processor
routes the freerun tuning word to the NCO.
In this case, the freerun tuning word establishes the NCO output
frequency, fNCO. The user sets the value of the freerun tuning
word via Bits[45:0] (unsigned integer) in Register 0x1000 to
Register 0x1005 (for DPLL0) and Register 0x1400 to
Register 0x1405 (for DPLL1).
The difference between the period of the signal arriving at the
reference input to the DPLL and the period of the signal
arriving at the feedback input to the DPLL constitutes the
period error between the two signals. The period error relates
to fREF and the feedback frequency (fFB) as
f
NCO ≈ fS × FTW0/248
where:
NCO is the NCO output frequency.
f
fS is the system clock frequency.
FTW0 represents the value of the 46-bit freerun tuning word.
Period Error = 1/fFB − 1/fREF
The preceding fNCO formula is an approximation (≈) because
the exact NCO output frequency differs slightly (see the DPLL
NCO section).
For any given period error sample, the frequency lock detector
either adds water with the fill bucket or removes water with the
drain bucket (one or the other, but not both). The decision of
whether to add or remove water depends on the frequency lock
threshold specified by the user via Bits[23:0] (unsigned integer)
of the appropriate source profile at the start address shown in
Rev. 0 | Page 129 of 205
AD9546
Data Sheet
For example, given a system clock frequency of 2.3 GHz and a
desired NCO output frequency of 245.76 MHz, solve the fNCO
equation for FTW0 to yield the following value of the freerun
tuning word:
(IRQ) section) through Bit 2 (fast acquisition started) and Bit 3
(fast acquisition completed) in Register 0x3012 (for DPLL0)
and Register 0x3017 (for DPLL1). Because Bit 2 and Bit 3 are
latched, the user must use Bit 2 and Bit 3, respectively, of
Register 0x200D (for DPLL0) and Register 0x2012 (for DPLL1) to
clear the latched state. Otherwise, the user cannot observe
subsequent state transitions of Bit 4.
FTW0 = 248 × (fNCO/fS)
= 248 × (245.76 × 106/(2.3 × 109))
(21)
= 30,076,213,163,657 (nearest integer)
= 0x 1B5A AA00 7689 (hexadecimal)
Fast Acquisition Bandwidth Control
The fast acquisition feature is active when the user sets the fast
acquisition excess bandwidth parameter to a nonzero value.
The fast acquisition excess bandwidth parameter is Bits[3:0]
(unsigned integer) in the appropriate translation profile at the
start address shown in Table 74 plus an offset of 22 (decimal).
Although the tuning resolution of the NCO is 48 bits, the
freerun tuning word is only 46 bits because, under all normal
operating conditions, the two most significant bits of the freerun
tuning word calculation (see Equation 21) are Logic 0. Therefore,
even though the NCO tuning word is 48 bits, the freerun
tuning word uses only the 46 LSBs of the NCO tuning word.
When the fast acquisition excess bandwidth parameter is
nonzero, the fast acquisition controller uses the fast acquisition
excess bandwidth parameter value to define the maximum
starting loop bandwidth (BW0) as follows:
The value of freerun tuning word must satisfy the constraints
imposed by the NCO SDM (see the DPLL NCO section).
BW0 = DPLL Loop Bandwidth × 2 fast acquisition excess bandwidth parameter
DPLL FAST ACQUISITION OPTIONS
For example, given a DPLL loop bandwidth of 0.0001 Hz
(100 µHz) and a fast acquisition excess bandwidth parameter
value of 9, find the value of BW0 as follows:
Fast Acquisition Overview
Certain applications necessitate low loop bandwidths less than
1 Hz (for example, an application where the input reference to
the DPLL originates from a GPS/GNSS receiver with a 1 pulse
per second output). A 1 Hz reference requires a loop bandwidth
much less than 1 Hz, which can lead to frequency acquisition and
phase lock times in the range of minutes to hours.
BW0 = DPLL Loop Bandwidth × 2 fast acquisition excess bandwidth parameter
= 0.0001 Hz × 29
= 0.0512 Hz
In the preceding example, when the DPLL begins the signal
acquisition process, the starting loop bandwidth for the fast
acquisition sequence is 512 times wider than the final value.
After the DPLL acquires phase lock, the fast acquisition
controller tightens the loop bandwidth by a factor of two and
waits for phase lock. This bandwidth tightening process repeats
until the loop bandwidth reaches the specified DPLL loop
bandwidth. By using wider loop bandwidths during the signal
acquisition process, the fast acquisition feature achieves frequency
and phase lock much quicker than normal DPLL operation
without the fast acquisition feature.
To overcome the long acquisition time imposed by a sub 1 Hz
loop bandwidth, the AD9546 provides a built in fast acquisition
feature that greatly reduces the lock time of the DPLL. The fast
acquisition feature automatically changes the DPLL loop filter
bandwidth in a controlled manner. When the fast acquisition
feature is active (see the Fast Acquisition Bandwidth Control
section) the fast acquisition controller uses the following
features to control the fast acquisition process:
•
•
•
•
DPLL loop bandwidth
Fast acquisition excess bandwidth
Fast acquisition lock settle time
Fast acquisition timeout
The phase lock characteristic of the loop is subject to the DPLL
lock detector settings (see the DPLL Lock Detectors section).
The controller begins the fast acquisition process by applying a
user defined excess bandwidth factor to the specified DPLL loop
filter bandwidth (see the DPLL Loop Filter Bandwidth section in
the DPLL Loop Filter section). The controller then successively
reduces the excess bandwidth until it reaches the specified DPLL
loop filter bandwidth. At each step in the bandwidth reduction
process, however, the controller waits for the DPLL phase
detector to indicate lock status before moving on to the next
step (note the fast ACQ x functional blocks in Figure 90).
Fast Acquisition Settling Time Control
Because each step in the fast acquisition process relies on phase
lock indication to proceed to the next step, the phase lock
indication must not chatter. That is, a noisy reference can cause
the phase lock indicator to switch intermittently between lock
and unlock until the loop settles to a stable lock condition. If
the lock indicator chatters, the chatter can cause the fast
acquisition controller to advance prematurely.
To help minimize the effect of a chattering phase lock indication,
the fast acquisition controller allows the user to specify a lock
indication settling time via Bits[2:0] (unsigned integer) in the
appropriate translation profile at the start address shown in
Table 74 plus an offset of 23 (decimal).
Fast Acquisition Status Indicators
While the fast acquisition controller is in the process of
performing a fast acquisition, status Bit 4 is Logic 1 in
Register 0x3102 (for DPLL0) and Register 0x3202 (for DPLL1).
Bit 4 relates to the IRQ function (see the Interrupt Request
Rev. 0 | Page 130 of 205
Data Sheet
AD9546
The value of Bits[2:0] tells the fast acquisition controller how
long to observe a nonchanging phase lock indication before
moving on to the next step. Table 82 show the available settling
time durations.
Fast Acquisition Trigger
The Fast Acquisition Timeout Control section describes the
timing of the fast acquisition process but not what triggers the
process. The user controls what triggers the fast acquisition
process via Bits[3:0] of Register 0x2106 (for DPLL0) and
Register 0x2206 (for DPLL1).
Table 82. Fast Acquisition Lock Detect Settling Time
Bits[2:0]
Settle Time
0
1
2
3
4
5
6
7
1 ms
10 ms
50 ms
100 ms
500 ms
1 sec
10 sec
50 sec
•
•
•
•
Bit 0: fast acquisition from freerun
Bit 1: fast acquisition from holdover
Bit 2: fast acquisition during first acquisition
Bit 3: fast acquisition when no distribution outputs are
toggling
The fast acquisition controller treats the specific restrictions
implied by Bits[3:0] in logical OR fashion. That is, the
controller obeys all restrictions that apply per the set bits. Note
that when all four of these bits are Logic 0, no restrictions
apply. That is, the fast acquisition controller is completely
unrestricted, and the DPLL always uses fast acquisition (if
enabled). However, making any one of these bits Logic 1 puts
the fast acquisition controller into a restricted operating mode
where it must obey the restrictions explicitly as follows.
During operation, if the DPLL indicates a phase unlock within
the settling period, the fast acquisition controller does not
advance to the next step. Instead, the controller resets the
settling timer, waits for the next indication of phase lock, and
invokes the settling period again. This process continues until
the phase lock remains for the full duration of the prescribed
settling time.
Fast acquisition triggers under the following four sets of
conditions:
Fast Acquisition Timeout Control
To prevent the fast acquisition controller from stalling because
the DPLL fails to reach a phase locked condition, the controller
has a timeout mechanism. That is, at each step in the fast
acquisition process, the fast acquisition controller waits for the
DPLL to phase lock (and settle), but only up to a user defined
maximum amount of time.
The user specifies the maximum time for the fast acquisition
controller to wait for phase lock via Bits[6:4] (unsigned integer)
in the appropriate translation profile at the start address shown
in Table 74 plus an offset of 23 (decimal).
The available timeout values appear in Table 83. If the fast
acquisition controller does not observe phase lock and settling
within the prescribed timeout interval, the fast acquisition
controller automatically advances to the next step in the fast
acquisition process. When the fast acquisition controller
reaches the final bandwidth step and the DPLL indicates phase
lock within the specified settle time, the fast acquisition
controller sets status Bit 5 (fast acquisition complete) of
Register 0x3102 (for DPLL0) or Register 0x3202 (for DPLL1).
•
•
•
When Bit 0 = 1 and the DPLL enters closed-loop operation
from freerun mode, fast acquisition triggers.
When Bit 1 = 1 and the DPLL enters closed-loop operation
from holdover mode, fast acquisition triggers.
When Bit 2 = 1 and this is the first execution of the fast
acquisition process, fast acquisition triggers. That is, the
fast acquisition controller has not completed a fast
acquisition sequence previously (more specifically, when
the fast acquisition complete status bit is Logic 0). When
the fast acquisition complete status bit is Logic 1, the user
can clear it by writing Logic 1 to Bit 3 (autoclearing) of
Register 0x2107 (for DPLL0) or Register 0x2207 (for
DPLL1). Setting Bit 3 = 1 provides a mechanism to have
subsequent first acquisition events.
•
When Bit 3 = 1 and all clock distribution outputs of the
corresponding PLL channel are not toggling, fast
acquisition triggers.
As an example of fast acquisition trigger control, assume Bit 1
and Bit 2 are both Logic 1 and that DPLL enters closed-loop
operation from freerun mode, which does not satisfy the first
condition. However, because there is a second restriction, and
the device treats the restrictions in logical OR fashion, the
status of the second constraint applies. As such, if the entry of
the DPLL into closed-loop operation from freerun mode
happens to be the first acquisition of the DPLL, a fast
acquisition is in effect (otherwise, a normal acquisition results).
Next, assume the DPLL enters closed-loop operation from
holdover, which satisfies the first condition. As such, the DPLL
employs fast acquisition (the first acquisition restriction is
immaterial in this case).
Table 83. Fast Acquisition Timeout
Bits[6:4]
Settle Time
0
1
2
3
4
5
6
7
10 ms
50 ms
100 ms
500 ms
1 sec
10 sec
50 sec
100 sec
Rev. 0 | Page 131 of 205
AD9546
Data Sheet
Because the DPLL feedback loop results in an average offset of
zero between the two inputs of the digital phase detector, tOFST
translates to the DPLL output with the polarity shown in Figure 94.
The signed nature of Bits[39:0] allows the user to advance or
delay the DPLL output signal relative to its input. A positive
value for tOFST delays the output signal relative to the input signal.
That is, the output edge occurs later. Conversely, a negative value
advances the output signal relative to the input signal. Therefore,
the output edge occurs earlier.
DPLL PHASE OFFSET CONTROL
In general, the feedback loop of the DPLL (like the APLL) tends
to force the average phase offset between the two inputs of the
digital phase detector to zero (see Figure 94). As a result, assuming
Integer N operation (that is, the N divider is not fractional), the
DPLL input and output signals are edge aligned (or exhibit a
constant time offset due to path latency).
The AD9546 DPLLs provide the ability to impose a programma-
ble time offset (tOFST) between the input and feedback signals by
means of a summing node at the feedback input of the digital
phase detector. To insert a time offset, use Bits[39:0] (signed) of
Register 0x1015 to Register 0x1019 (for DPLL0) and Register
0x1415 to Register 0x1419 (for DPLL1). The value of Bits[39:0]
has units of ps.
The relationship between tOFST (sec) and the programmed value
for the DPLL phase offset (Bits[39:0]) is
DPLL Phase Offset = tOFST/10−12
For example, find the value of the DPLL phase offset necessary
to advance the output signal by 75 ns (that is, tOFST = −75 ns).
Two other sources for applying a phase offset also feed the
summing node (not explicitly shown in Figure 94). One is via
the source profiles (see the Skew Adjustment section of the
Source Profiles section), and the other is via the delay compens-
ation feature (see the Delay Compensation section).
DPLL Phase Offset = (−75 × 10−9)/10−12
= −75,000
= 0x FF FFFE DB08 (hexadecimal)
XOA XOB
LOOP
CONTROLLER
SYSTEM
CLOCK
AD9546
PHASE OFFSET
FREERUN TUNING WORD
LOCK
DETECTORS
REG 0x1019 TO 0x1015, BITS[39:0]
REG 0x1005 TO 0x1000, BITS[45:0]
NUMERIC
COEFFICIENTS
DIGITAL CROSS
POINT MUX
40
46
DIGITAL
PHASE
DETECTOR
DIGITAL
LOOP
FILTER
FTW
PROCESSOR
TDC
NCO
0
48-BIT
FTW
tOFST
N DIVIDER
TDC
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES ARE SPECIFIC TO DPLL0
Figure 94. DPLL Phase Offset Feature
Rev. 0 | Page 132 of 205
Data Sheet
AD9546
In some cases, it is more useful to specify fCLAMP as a fractional
offset of the NCO output frequency (in percent or parts per
million, for example).
TUNING WORD OFFSET CLAMP
The DPLL contains a frequency clamp that places bounds on
the frequency range of the DPLL output. The frequency clamp
feature (see Figure 95) is beneficial for applications in which
downstream devices cannot tolerate frequencies beyond
prescribed limits.
For example, given a system clock frequency of 2.4 GHz, an
NCO output frequency of 250 MHz and a desired frequency
offset clamp limit of 25 ppm (25 × 10−6), find the frequency
clamp value. First, determine fCLAMP as follows:
The clamp feature uses Bits[23:0] (unsigned) in Register 0x1006
to Register 0x1008 (for DPLL0) and Register 0x1406 to
Register 0x1408 (for DPLL1), which constitutes the frequency
clamp value. The frequency clamp feature is always active.
However, the default value of the frequency clamp is the
maximum clamp value, which establishes a default frequency
clamp limit of approximately 586 kHz (for a system clock
frequency of 2.4 GHz). Given a 320 MHz NCO output
frequency, a 586 kHz frequency clamp equates to an offset of
approximately 1800 ppm (or 0.18%).
f
CLAMP = 25 × 10−6 × 250 MHz
= 6250 Hz
Next, substitute fCLAMP into Equation 22.
Frequency Clamp Value = 236 × fCLAMP/fS
= 236 × 6250/(2.4 × 109)
= 178,957 (nearest integer)
= 0x 02 BB0D (hexadecimal)
When the frequency clamping circuitry is actively clamping the
frequency, status Bit 1 is Logic 1 in Register 0x3102 (for DPLL0)
and Register 0x3202 (for DPLL1). Bit 1 relates to the IRQ
function (see the Interrupt Request (IRQ) section) through
Bit 6 and Bit 7 (frequency clamp active and inactive, respectively)
in Register 0x3010 (for DPLL0) and Register 0x3015 (for
DPLL2). Because Bit 6 and Bit 7 are latched, the user must use
Bit 6 and Bit 7, respectively, of Register 0x200B (for DPLL0)
and Register 0x2010 (for DPLL1) to clear the latched state.
Otherwise, the user cannot observe subsequent state transitions
of Bit 1.
The frequency offset, fCLAMP, defines an upper and lower
frequency bound as a deviation from the frequency given by the
freerun tuning word of the DPLL. The relationship between
f
CLAMP, the system clock frequency (fS), and the frequency clamp
value is as follows:
f
CLAMP = Frequency clamp value × (fS/236)
For example, given a system clock frequency of 2.4 GHz and a
desired frequency offset clamp limit of 10 kHz (that is,
fS = 2.4 × 109 and fCLAMP = 104), find the frequency clamp value.
Solving the fCLAMP equation for the frequency clamp value,
Frequency clamp value = 236 × fCLAMP/fS
= 236 × 104/(2.4 × 109)
(22)
= 286,331 (nearest integer)
= 0x 04 5E7B (hexadecimal)
XOA XOB
SYSTEM
LOOP
CONTROLLER
AD9546
CLOCK
LOCK
DETECTORS
FREERUN TUNING WORD
REG 0x1005 TO 0x1000, BITS[45:0]
46
NUMERIC
DIGITAL CROSS
POINT MUX
COEFFICIENTS
DIGITAL
PHASE
DETECTOR
DIGITAL
LOOP
FILTER
FTW
PROCESSOR
TDC
NCO
48-BIT
FTW
24
REG 0x1008 TO 0x1006, BITS[23:0]
TDC
N DIVIDER
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES ARE SPECIFIC TO DPLL0
Figure 95. Tuning Word Offset Clamp Feature
Rev. 0 | Page 133 of 205
AD9546
Data Sheet
Register 0x1411 to Register 0x1414 (for DPLL1), which constitutes
the phase slew limit value. The DPLL phase slew limit value has
units of ps/sec or 10−12, but accuracy is typically limited to
approximately 50 ps/sec.
PHASE SLEW RATE LIMIT
The digital phase detector uses the reference and feedback
TDCs to determine the associated phase error and pass the
error along to the loop filter. In addition to the loop phase
error, however, other types of phase information can appear in
the loop, for example, phase offsets originating from any of the
following:
The phase slew rate limit feature is active for all DPLL phase
slew limit values except zero, which disables the phase slew rate
limiter (disabling the phase slew rate limiter is not recommended).
However, there is little reason to disable the rate limiter. Instead,
programming the rate limiter to the maximum value imposes a
slew rate limit corresponding to a frequency shift in excess of
4000 ppm. Most applications do not experience frequency
transients exceeding 4000 ppm. The DPLL phase slew limit default
value establishes a phase slew rate limit of 100,663,296 ps/sec
(which equates to a maximum frequency shift of approximately
100 ppm).
Changes to the DPLL phase offset (see the DPLL Phase
Offset Control section)
Changes to the phase skew offset (see the Source Profiles
section)
From the delay compensation block (see the Delay
Compensation section)
As the result of a phase buildout operation (see the Phase
Buildout Mode section)
The relationship between the desired phase slew rate limit, Δt/t,
and the value of the DPLL phase slew limit value is as follows:
For certain phase sources, the DPLL provides a phase slew rate
limiting function (see Figure 96) that places an upper bound on
the rate of change of phase passed along to the loop filter. The
phase slew rate limiter mitigates the adverse effects of injecting
a large phase step into the loop by converting the phase step to
a phase ramp with the maximum slope set by the user.
DPLL Phase Slew Limit Value = (Δt/t) × 1012
For example, given a desired upper bound to the rate of change
of phase (Δt/t) of 0.25 ppm (2.5 × 10−7), find the required DPLL
phase slew limit value as follows:
DPLL Phase Slew Limit Value = (Δt/t) × 1012
= (2.5 × 10−7) × 1012
The phase slew rate limiter operates only on phase transients
introduced by the following phase sources:
DPLL phase offset (see the DPLL Phase Offset Control
section)
Phase skew offset (see the Source Profiles section)
Initial phase correction due to activating a hitless mode
translation profile (see the Translation Modes section)
= 250,000
= 0x 0003 D090 (hexadecimal)
When the phase slew limiter is actively limiting the rate of
change of phase, status Bit 2 (DPLL phase slew limiting) is
Logic 1 of Register 0x3102 (for DPLL0) or Register 0x3202 (for
DPLL1). Bit 2 relates to the IRQ function (see the Interrupt
Request (IRQ) section) through Bit 4 (DPLL phase slew limiter
active) and D5 (DPLL phase slew limiter inactive) in Register
0x3010 (for DPLL0) and Register 0x3015 (for DPLL1). Because
Bit 4 and Bit 5 are latched, the user must use Bit 4 and Bit 5,
respectively, of Register 0x200B (for DPLL0) and Register
0x2010 (for DPLL1) to clear the latched state. Otherwise, the user
cannot observe subsequent state transitions of Bit 2.
XOA XOB
That is, the DPLL only applies phase slew limiting under two
conditions. The first is a phase offset that exists at the start of a
reference acquisition. The second is phase offsets introduced by
the user via the DPLL phase offset feature and the phase skew
offset feature.
To set the phase slew limit rate, use Bits[31:0] (unsigned) in
Register 0x1011 to Register 0x1014 (for DPLL0) and
LOOP
CONTROLLER
SYSTEM
CLOCK
AD9546
LOCK
DETECTORS
FREERUN TUNING WORD
REG 0x1005 TO 0x1000, BITS[45:0]
NUMERIC
COEFFICIENTS
DIGITAL CROSS
POINT MUX
46
DIGITAL
PHASE
DETECTOR
DIGITAL
LOOP
FILTER
FTW
PROCESSOR
TDC
NCO
48-BIT
FTW
32
REG 0x1014 TO 0x1011, BITS[31:0]
TDC
N DIVIDER
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES ARE SPECIFIC TO DPLL0
Figure 96. Phase Slew Rate Limit Feature
Rev. 0 | Page 134 of 205
Data Sheet
AD9546
Averaging Processor Delay
TUNING WORD HISTORY
By default, as soon as a translation profile becomes active (see the
Reference Switching section for what constitutes an active
translation profile), the tuning word processor resets the averaging
processor (and DPLL history available status bit) and the averaging
processor immediately starts processing tuning words from the
loop filter.
The DPLLs have a tuning word processor that handles the
application of tuning words to the NCO. The tuning word
processor embodies several of the functional blocks shown in
Figure 91, including the loop controller, FTW processor, and
the switch. The NCO can receive tuning words from the
following three possible sources:
However, the user has access to two independent mechanisms to
impose a delay between when a translation profile becomes
active and when the averaging processor begins the tuning
word averaging process: an event dependent delay and a timed
delay.
•
•
•
Freerun tuning word
Digital loop filter
Tuning word averaging processor
This section focuses on the tuning word averaging processor,
which provides the following three digital outputs residing in
the register map:
By default, both mechanisms are inactive, implying no delay.
Event dependent delays take priority over time delays: first, any
of the three possible event dependent delay selections
programmed by the user, then the timed delay programmed by
the user. Until these delays expire, the tuning word processor
ignores incoming tuning words.
•
•
•
DPLL tuning word history
DPLL history available status
DPLL history updated status
The DPLL tuning word history is a 46-bit unsigned value that
resides in Register 0x3103 to Register 0x3108 (for DPLL0) and
Register 0x3203 to Register 0x3208 (for DPLL1).
The status of the DPLL is the basis for the event dependent
delay mechanism. To invoke the event dependent delay, write a
Logic 1 to any combination of the following delay history
control bits:
The DPLL history available status is available via Bit 0 of
Register 0x3102 (for DPLL0) and Register 0x3202 (for DPLL1).
•
•
•
DPLL delay history until phase lock
DPLL delay history until frequency lock
DPLL delay history until not slew limiting
The DPLL history updated status is available via Bit 2 of
Register 0x3011 (for DPLL0) and Register 0x3016 (for DPLL1).
The user also has access to the DPLL history available status
and DPLL history updated status as a physical logic level via an
appropriately configured Mx pin.
These bits reside in Bits[5:3] of Register 0x100E (for DPLL0)
and Register 0x140E (for DPLL1). Logic 1 invokes the
described delay. The DPLL delay history until phase lock bit
(Bit 3) causes the averaging process to delay until the DPLL
phase locks. The DPLL delay history until frequency lock bit
(Bit 4) causes the averaging process to delay until the DPLL
frequency locks. The DPLL delay until not slew limiting bit (Bit
5) causes the averaging process to delay until the phase slew
limiter ceases slew limiting, assuming slew limiting occurs (see
the Phase Slew Rate Limit section).
The main purpose of the averaging processor is to compute an
average of tuning word samples when a DPLL translation profile
initially becomes active (but after expiration of any delays specified
by the delay element of the averaging processor as detailed the
Averaging Processor Delay section). After the averaging processor
collects enough samples to allow a valid tuning word average
computation, the processor sets the DPLL history available
status bit to Logic 1, indicating that the averaged tuning word
history is available. If the DPLL needs to switch to holdover
operation, the DPLL can use the averaged tuning word history
of the averaging processor. Otherwise, the DPLL uses the last
available tuning word from the loop filter or the value in the DPLL
freerun tuning word, depending on the configuration of the
averaging processor.
When more than one of the delay history control bits are Logic 1,
the implementation of the delay behaves as an AND function of
the selected conditions. That is, all the selected status
conditions must be satisfied before the averaging process
begins. The status conditions are real-time status indicators as
they follow the actual state of the DPLL. However, the moment
all selected status conditions are true, the averaging processor
waits for the prescribed holdoff period to expire (assuming the
DPLL history holdoff time is not zero) and starts the averaging
process (even if any of the status conditions become false after
the averaging processor starts averaging).
The averaging processor comprises the following three
functional elements:
•
•
•
Delay
Windowed average
Continue or reset
These functional elements respond to user input via the register
map, as explained in the Averaging Processor Delay section, the
Averaging Processor Windowed Average section, and the
Averaging Processor Continue or Reset section.
Rev. 0 | Page 135 of 205
AD9546
Data Sheet
The user specifies the desired delay, or holdoff period, Bits[7:0]
(unsigned) in Register 0x1010 (for DPLL0) and Register 0x1410
(for DPLL1). A value of zero implies no delay. Bits[7:0]
constitute the DPLL history holdoff time. The delay relates to
the DPLL history holdoff time value as follows:
Changing the value of the DPLL history accumulation timer
exhibits different behavior depending on the current state of the
history circuit. The history processor only uses DPLL history
accumulation timer during holdoff and active operation, in
which case the history processor uses the new value (though
there may be a delay associated with the old value before the
update takes effect). The history circuit does not change states
upon a change of the value of the DPLL history accumulation
timer.
Delay = DPLL History Holdoff Time × (tACCUM/8)
where tACCUM is the time specified by the DPLL history
accumulation timer value (see the Averaging Processor
Windowed Average section for details on the DPLL history
accumulation timer).
Updating the value of the DPLL history accumulation timer
while the history circuit is actively collecting history causes
future intervals to conform to the new value. If the elapsed
period of the current interval exceeds that of the new interval,
the next interval begins immediately. Otherwise, the current
interval conforms to the new value.
The delay (if enabled) is also in effect following a pause/restart
trigger event (see the Averaging Processor Continue or Reset
section).
Averaging Processor Windowed Average
DPLL History Accumulation Timer
However, if the history circuit is in the holdoff state (see the
Averaging Processor Delay section), behavior is slightly
different. Because the history circuit averages over 1/8th
subintervals (see the Tuning Word Averaging Process section),
the progress of the averaging circuit matters. Specifically, the
number subintervals already completed remains unchanged
regardless of the actual time elapsed, and future subintervals
use the new period while the current subintervals continue to
use the old value.
In general, the averaging processor collects tuning word
samples generated by the loop filter and computes an average
of the tuning word samples. The user specifies an averaging
window for the rolling average time span via Bits[27:0]
(unsigned) of Register 0x100A to Register 0x100D (for DPLL0)
and Register 0x140A to Register 0x140D (for DPLL1).
Bits[27:0] constitute the DPLL history accumulation timer value.
The units associated with Bits[27:0] are milliseconds allowing
averaging time spans from 0.001 sec to 268,435.455 sec
(~74.5 hours). The default value is 10 (decimal), yielding a
default averaging window time span of 10 ms.
Changing the DPLL history holdoff time value (see the Averaging
Processor Delay section) has no effect until the beginning of the
next holdoff period (in the entirety of the period, not a
subinterval). If the behavior associated with the holdoff period,
as a result of updating either the history period or the holdoff
value, must reflect the new value as soon as possible, the user can
restore the history circuit to its initialization state via the DPLL
clear history bit (see the DPLL Clear History Bit section).
The time span of the averaging window relates to the value of
the DPLL history accumulation timer as follows:
Time Span = DPLL History Accumulation Timer × 10−3
The recommended minimum averaging window time span is at
least 10 times the minimum expected period of the signal at the
input to the phase/frequency detector of the DPLL.
Tuning Word Averaging Process
During operation, the averaging processor performs averaging
by partitioning the time span defined by the value of the DPLL
history accumulation timer into eight subintervals. During each
subinterval, the averaging processor computes the average of
the tuning word samples for that subinterval. A full tuning
word average, as defined by the value of the DPLL history
accumulation timer, constitutes the average of eight
consecutive subinterval averages.
As an example, determine the value of the DPLL history
accumulation timer necessary to provide a 15-minute
windowed average. First, solve the time span equation for the
DPLL history accumulation timer as follows:
DPLL History Accumulation Timer = Time Span × 1000
= (15 × 60) × 1000
= 900,000
At the termination of each successive subinterval (given the
averaging processor has set the DPLL history available status
bit), the averaging processor triggers the following events:
= 0x 00D BBA0 (hexadecimal)
DPLL history accumulation timer = 0 is a special case. This
value bypasses the averaging processor by routing tuning word
samples from the digital loop filter to both the DPLL NCO and
to the DPLL tuning word history. Thus, programming DPLL
history accumulation timer = 0 makes it possible to monitor
tuning words directly from the loop filter as they are applied to
the DPLL NCO. The DPLL history available status bit (Bit 0 of
Register 0x3102 for DPLL0 and Register 0x3202 for DPLL1) is
Logic 0 when DPLL history accumulation timer = 0.
•
•
•
Update the rolling average computation
Save the result to the DPLL tuning word history
Update the DPLL history update status bit
As such, at the end of each subinterval, the averaging processor
stores the result of the updated rolling average computation in
the DPLL tuning word history. At the same time, the averaging
processor sets the DPLL history updated status bit to Logic 1.
The DPLL history updated status bit provides the user with a
Rev. 0 | Page 136 of 205
Data Sheet
AD9546
mechanism for knowing when the tuning word history
processor refreshes the DPLL tuning word history. However,
the DPLL history updated bit is a latched bit. Therefore, this bit
indicates a refresh of the DPLL tuning word history, but not
necessarily the most recent history.
subinterval averaging window, which remains the case for all
subsequent subintervals.
The quick start feature is useful when the time span dictated by
the DPLL history accumulation timer is very large (hours, for
example). The quick start feature allows the averaging
processor to make an initial tuning word average available after
¼ of the time indicated by the DPLL history accumulation timer.
DPLL Clear History Bit
The user can clear the averaging history at any time via Bit 1 of
Register 0x2107 (for DPLL0) and Register 0x2207 (for DPLL1).
Logic 1 forces the averaging processor to clear the averaging
circuitry, thereby erasing any previous tuning word history
information, and forces the DPLL history available bit to
Logic 0 (until a newly computed initial average is available).
The DPLL single sample history and DPLL quick start history
bits are not mutually exclusive. Both can be set to provide both
options simultaneously with the state of the averaging
processor, dictating which option is applicable for the
prevailing conditions. Following a switch from active closed-
loop operation to holdover operation, the progress made by the
averaging processor in computing the initial tuning word
average directs the action of the tuning word processor
according to the state of the DPLL single sample history bit and
the DPLL quick start history bit.
Programming the DPLL clear history bit to Logic 1 does not
clear the contents of the DPLL tuning word history. That is, the
last update of the DPLL tuning word history remains intact
until the processor calculates a new average and sets the DPLL
history updated bit, thereby notifying the user a new average is
available.
DPLL Persistent History Bit
The DPLL persistent history bit works in conjunction with the
control bits described in the Averaging Processor Continue or
Reset section. The DPLL persistent history bit is Bit 0 of
Register 0x100E (for DPLL0) and Register 0x140E (for DPLL1).
DPLL Single Sample History Bit
When the DPLL must make a switch from active closed-loop
operation to holdover operation, it attempts to use the tuning
word history from the averaging processor. However, if the
averaging processor has not yet set the DPLL history available
status bit (indicating no history is available), the DPLL uses the
DPLL freerun tuning word as the tuning word for the NCO by
default. Alternatively, the user can have the DPLL use the most
recent tuning word from the loop filter instead. To use the most
recent tuning word from the loop filter, program Bit 1 to Logic
1 of Register 0x100E (for DPLL0) and Register 0x140E (for
DPLL1).
When Bit 0 = 0, the tuning word processor resets the averaging
processor based on the control bits described in the Averaging
Processor Continue or Reset section. When Bit 0 = 1, this
setting prevents the tuning word processor from clearing the
averaging processor based on the control bits described in the
Averaging Processor Continue or Reset section. Instead, the
averaging processor holds onto the previous computation. That
is, the computation pauses or persists.
Averaging Processor Continue or Reset
DPLL Quick Start History Bit
By default, the averaging processor continues to compute a
rolling average regardless of the state of the DPLL. The following
three bits provide optional control over resetting or pausing the
averaging processor based on the DPLL state:
Normally, when the DPLL is in active closed-loop operation,
the averaging processor must process tuning words for at least
the amount of time prescribed by the DPLL history accumulation
timer. The prescribed time is necessary to compute a full tuning
word history average and to set the DPLL history available bit
(thereby indicating the averaging processor has processed
enough tuning words to compute an initial average). However,
the user has the option to allow the averaging processor to
indicate the history is available earlier by writing a Logic 1 to
Bit 2 of Register 0x100E (for DPLL0) and Register 0x140E (for
DPLL1). Under this condition, the averaging processor sets the
DPLL history available status bit after the first two subintervals
expire and updates the DPLL tuning word history. Updates of
the DPLL tuning word history continue every subinterval
thereafter.
•
•
•
DPLL pause history while phase unlocked
DPLL pause history while frequency unlocked
DPLL pause history while phase slew limiting
These bits pause or reset the averaging process depending on
the DPLL persistent history bit. When DPLL persistent history = 0,
these three bits cause the averaging processor to reset. When
DPLL persistent history = 1, these three bits cause the averaging
processor to pause and then resume when the conditions
dictate.
The DPLL pause history while phase unlocked bit is Bit 0 of
Register 0x100F (for DPLL0) and Register 0x140F (for DPLL1).
Programming Bit 0 = 1 pauses or resets the averaging processor
while the DPLL is not phase locked.
During the first eight subintervals of the quick start history
sequence, the second and third subintervals exhibit a two-
subinterval averaging window. The fourth, fifth, sixth, and
seventh subintervals exhibit a four-subinterval averaging
window. Finally, the eighth subinterval exhibits an eight-
Rev. 0 | Page 137 of 205
AD9546
Data Sheet
The DPLL pause history while frequency unlocked bit is Bit 1
of Register 0x100F (for DPLL0) and Register 0x140F (for
DPLL1). Programming Bit 1 = 1 pauses or resets the averaging
processor while the DPLL is not frequency locked.
Delay compensation generates the temperature dependent time
offset value, Δt, based on a fifth degree polynomial, as shown in
Figure 97. The independent variable of the polynomial is
temperature (T in Figure 97). The five polynomial coefficients
(C1 to C5), which the user enters via the register map, scale with
corresponding powers of T. Note that there is no need for a
constant term (C0) in the delay compensation block, because the
DPLL phase offset feature (see the DPLL Phase Offset Control
section) provides that function. The Δt computation implied by
Figure 97 is
The DPLL pause history while phase slew limiting bit is Bit 2 of
Register 0x100F (for DPLL0) and Register 0x140F (for DPLL1).
Programming Bit 2 = 1 pauses or resets the averaging processor
while the phase slew limiter is slew limiting (see the Phase Slew
Rate Limit section).
The user can set any combination of these bits. Unlike the
averaging processor delay bits, however, when more than one
of these bits are Logic 1, the implementation of the delay
behaves as an OR function of the selected conditions. That is,
satisfying any of the selected status conditions causes the
averaging processor to reset or pause.
5
∆t = C Tk
∑
k
k =1
The temperature (T) input into the open-loop direct compensation
block originates from temperature sensor circuitry (see the
Temperature Sensor section). The user has the option of choosing
the temperature source: either the internal temperature sensor
(default) or a user supplied temperature via the register map.
When using the internal temperature sensor, the device
inherently uses the appropriate scaling units. However, when
using the register map option, the user must provide the proper
scaling (see the External Temperature Source section for
details).
DELAY COMPENSATION
Overview
Delay compensation provides an automatic phase offset to the
DPLL to counteract temperature dependent delay variation.
Figure 97 is a block diagram of the DPLL and delay compensation
block. Delay compensation is only in effect while the DPLL is
active. Furthermore, the response time of the delay
compensation at the DPLL output is subject to the loop
bandwidth of the DPLL.
XOA XOB
LOOP
CONTROLLER
SYSTEM
CLOCK
AD9546
LOCK
DETECTORS
FREERUN TUNING WORD
REG 0x1005 TO 0x1000, BITS[45:0]
NUMERIC
COEFFICIENTS
46
DIGITAL
PHASE
DETECTOR
DIGITAL
LOOP
FILTER
TDC
FTW
PROCESSOR
NCO
48-BIT
FTW
DPLL0 PHASE
OFFSET
40
REG 0x1019 TO 0x1015, BITS[39:0]
N DIVIDER
TDC
Δ
t
16
T
5
4
3
2
T
T
T
T
C
C
C
C
C
1
5
4
3
2
FILTER
3
DELAY COMPENSATION
REG 0x1029, BITS[2:0]
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES ARE SPECIFIC TO DPLL0
Figure 97. Delay Compensation
Rev. 0 | Page 138 of 205
Data Sheet
AD9546
Because −29 is greater than the quantization limit of −127, the
C1 exponent is
Delay Compensation Coefficients (Ck) Programming
Coefficient Ck (where k = 1 to 5) applies scale factors to the
appropriate powers of T. The Ck coefficients carry units of
sec/°C. Each coefficient comprises a significand component and
an exponent component. The user programs the significand
and exponent components independently in the register map.
The coefficient components reside in the register map per
Table 84.
C1 Exponent = C1_ExpVal
= −29
= 0xE3 (hexadecimal)
The C1 significand is
C1 Significand = C1 × 215 − C1_ExpVal
= 10−9 × 215 − (−29)
Table 84. Delay Compensation Coefficient Address Ranges
Register Address
= 17,592 (nearest integer)
= 0x44B8 (hexadecimal)
Delay Compensation Filter
DPLL Coefficient
Significand
Exponent
0x101C
0x101F
0x1022
0x1025
0x1028
0x141C
0x141F
0x1422
0x1425
0x1428
0
0
0
0
0
1
1
1
1
1
C1
C2
C3
C4
C5
C1
C2
C3
C4
C5
0x101A to 0x101B
0x101D to 0x101E
0x1020 to 0x1021
0x1023 to 0x1024
0x1026 to 0x1027
0x141A to 0x141B
0x141D to 0x141E
0x1420 to 0x1421
0x1423 to 0x1424
0x1426 to 0x1427
The Δt values produced by the delay compensation block depend
on temperature measurements (internal or via the register map)
as an input parameter. Any noise associated with those
measurements is a potential noise source on Δt, which can lead
to a degradation of the phase noise performance of the DPLL.
To mitigate potential noise injection, the delay compensation
block uses a filter that applies a smoothing function to the raw
Δt values.
The significand and exponent values are signed (twos
complement) numbers and program in exactly the same manner
as the significand and exponent of the Tk values for system
clock compensation (see the Compensation Method 1 section
for details on converting decimal coefficient values to signed,
twos complement, significand and exponent values).
The phase slew limiter (see the Phase Slew Rate Limit section)
does not process time variations injected into the DPLL by the
delay compensation filter. That is, these variations affect the
DPLL output without intervention by the phase slew limiter.
The user controls the filter bandwidth via Bits[2:0] (unsigned)
of Register 0x1029 (for DPLL0) and Register 0x1429 (for
DPLL1). The filter comprises a single-pole response yielding a 3
dB bandwidth per Table 85 (which also shows the transition
time to 99% of a step input, associated with a step change at the
input to the filter).
For example, suppose a user makes measurements that indicate
the output clock signals of the AD9546 exhibit a delay variation
of −1 ns/°C. To compensate for this variation, apply a 1 ns/°C
correction. In polynomial form,
f(T) = C5T5 + C4T4 + C3T3 + C2T2 + C1T
Table 85. Delay Compensation Filter Bandwidth
where f(T) is the correction value in units of seconds as a
function of temperature (T).
Bits[2:0]
3 dB Bandwidth (Hz) Transition Time (ms)
000
001
010
011
100
101
110
111
240
120
60
30
15
7.6
3.8
1.9
3.27
6.58
13.0
26.5
53.1
99.5
199
Because the delay variation in this example is strictly linear with
respect to temperature, all the coefficients are zero except for
C1. Thus, in this example, the compensation polynomial
simplifies to the following:
f(T) = C1T
To apply 1 ns/°C compensation, C1 = 10−9 sec/°C. Referring to
the procedure in the Compensation Method 1 section, calculate
the following for C1:
398
log | C1 |
log2
C1_ExpVal =
+1
-9
|
log |10
log 2
=
+1
= −29
Rev. 0 | Page 139 of 205
AD9546
Data Sheet
The no tagging mode is the normal operating mode of the DPLL.
That is, the digital PFD processes every reference and feedback
time stamp whether the time stamp is a tagged time stamp or not.
TIME STAMP TAGGING OPTIONS
The input to the DPLL consists of time stamps arriving from
two TDCs: a reference input TDC and a feedback TDC.
Furthermore, the user can enable tagging of time stamps
originating from those TDCs (see the Tagged Time Stamps
section). The user can also enable the DPLL to operate based on
tagged time stamps (rather than every time stamp) originating
from the reference and/or feedback TDCs.
The reference tagging only mode tells the digital PFD to use
tagged reference time stamps for phase alignment purposes.
The user must ensure the reference is set up to provide tagged
time stamps. When the reference is one of the REFx or auxiliary
REFx inputs, the reference demodulator is the source of tagged
time stamps (see the Reference Demodulator section). When
the reference is an auxiliary NCO, the auxiliary NCO is the
source of tagged time stamps (see the Auxiliary NCOs section).
The use of tagged time stamps by the DPLL only applies for zero
delay (hitless) mode. The DPLL ignores tagged time stamps in
phase buildout mode. See the Frequency Translation Loops section
for an explanation of zero delay and phase buildout modes.
The feedback tagging only mode tells the digital PFD to use
tagged feedback time stamps for phase alignment purposes.
The user must ensure the feedback path is set up to provide
tagged time stamps via output clock modulation (see the
Distribution Embedded Output Clock Modulation section).
The user establishes the tagged time stamp functionality of the
DPLL via Bits[4:2] of the appropriate translation profile at the
start address shown in Table 74 plus an offset of 3 (decimal).
There are five tagged operating modes per Table 86.
The combined reference and feedback tagging modes tell the
digital PFD to use tagged time stamps from both the reference
and feedback TDCs. The user must ensure that the reference
and feedback paths are set up to provide tagged time stamps.
Furthermore, the user must ensure that the tag rate of the
reference path matches the tag rate of the feedback path.
Table 86. Tagged Time Stamp Operating Modes
Bits[4:2]
Tag Mode
0
1
2
3
No tagging (default)
Reference tagging only
Feedback tagging only
Combined reference and feedback tagging with
equal carrier rates
There are two variants of the combined reference and feedback
tagging modes: equal and unequal carrier rates. The combined
reference and feedback tagging modes require equal tag rates
on both paths. There is no such requirement on the carrier
rates for both paths. However, to function properly, the DPLL
requires advance knowledge of the relationship between the
reference and feedback carrier rates, which is the reason for the
two variants. The user must select the appropriate variant based
on whether the reference and feedback TDC carrier rates differ.
4
Combined reference and feedback tagging with
unequal carrier rates
5 to 7
Not applicable
With respect to tagged time stamps, the carrier rate is the rate at
which the TDC generates time stamps (that is, the underlying rate
of the input clock to the TDC). The tagged rate is the rate at which
the TDC generates tagged time stamps. The tagged rate is
always an integer submultiple of the carrier rate.
Rev. 0 | Page 140 of 205
Data Sheet
AD9546
CASCADED DPLL CONFIGURATION
The cascaded DPLL configuration applies to applications that
lock both DPLLs to the same reference source with both DPLLs
remaining phase locked, even if all reference sources are lost. To
understand cascaded configurations, first consider device
operation without cascaded capability, as shown in Figure 98.
As a prerequisite, the reader must be familiar with the
Translation Profiles section and the Reference Switching
section. Figure 98 and Figure 99 show simplistic diagrams of
the DPLL showing a phase buildout configuration, but the zero
delay (hitless) configuration is valid as well.
source, assuming automatic reference switching is in effect (see
the Reference Switching section). Presumably, the user has
defined translation profiles in such a way that DPLL0 finds
Translation Profile 0.b and DPLL1 finds Translation Profile 1.y
(where b and y are variables that can be any integer value from
0 to 5, but not the same value as the a and z variables) with both
profiles assigning the same valid reference source (Reference
Source 2, in this example). As such, both DPLLs again lock to a
common reference as desired.
The far right hand side of Figure 98 shows the result of losing
Reference Source 2, leaving no valid reference sources. In this
case, both DPLLs have no recourse but to switch to holdover or
freerun mode. However, because the DPLLs may not have the
same FTW in effect when they switch to holdover or freerun
mode, their outputs are no longer phase locked, which breaks
the original requirement that both DPLLs remain phase locked
when all reference sources are lost. The cascaded configuration
provides a solution to this problem.
The far left side of Figure 98 shows two valid reference sources
with DPLL0 and DPLL1 using Translation Profile 0.a and
Translation Profile 1.z, respectively, assigned to Reference
Source 1, where a and z are variables that can be any integer
value from 0 to 5. Therefore, both DPLLs lock to a common
reference as desired.
The middle portion of Figure 98 shows the result of losing
Reference Source 1. The loss of Reference Source 1 results in
both DPLLs independently searching for a new reference
TRANSLATION PROFILE 0.a AND
TRANSLATION PROFILE 1.z ASSIGN
REFERENCE SOURCE 1
TRANSLATION PROFILE 0.b AND
TRANSLATION PROFILE 1.y ASSIGN
REFERENCE SOURCE 2
BOTH DPLLS SWITCH TO
HOLDOVER/FREERUN
TRANSLATION
DIGITAL
DIGITAL
CROSS
POINT
MUX
TRANSLATION
PROFILE 0.a
DIGITAL
CROSS
POINT
MUX
PROFILE0.b
CROSS
POINT
MUX
DPLL0
DPLL0
DPLL0
TO
APLL0
TO
APLL0
TO
APLL0
FTW
NCO
NCO
LF
PFD LF NCO
N
ALL
REFERENCE
SOURCES
LOST
PFD
REFERENCE
SOURCE 1
LOST
REFERENCE
SOURCES
REFERENCE
SOURCES
N
TDC
1
TRANSLATION
PROFILE 1.z
TRANSLATION
PROFILE 1.y
TDC
2
TDC
2
DPLL1
DPLL1
DPLL1
TO
APLL1
TO
APLL1
TO
APLL1
FTW
PFD LF NCO
N
PFD
NCO
NCO
LF
N
Figure 98. Noncascaded DPLL Operation
Rev. 0 | Page 141 of 205
AD9546
Data Sheet
Figure 99 shows cascaded DPLL configuration and its operation.
The main difference between Figure 98 and Figure 99 is that
DPLL1 has Translation Profile 1.x configured to use the
feedback path of DPLL0 as a reference input. Setting up
Translation Profile 1.x in this way establishes the potential for
cascaded DPLL operation. The fact that Translation Profile 1.x
is set up to use the feedback path of DPLL0 as a reference input
identifies DPLL1 as the secondary DPLL and DPLL0 as the
primary DPLL in the cascaded DPLL configuration (see the
right side of Figure 99).
When DPLL0 loses its reference, it has no recourse but to
switch to holdover or freerun mode because no reference
sources are available. However, when Translation Profile 0.b
becomes inactive, because Translation Profile B is the inactive
profile index for DPLL0, the feedback path of the DPLL0 is
available as a valid reference for DPLL1. Although DPLL0
switches to holdover or freerun, DPLL1 does not because it has
an available reference source: the feedback path of DPLL0.
DPLL1 may experience a momentary switch to holdover or
freerun mode before switching to cascaded DPLL operation.
The result is cascaded DPLL operation, in which the reference
input of DPLL1 (the secondary DPLL) connects to the feedback
path of DPLL0 (the primary DPLL). In this way, the two DPLL
outputs are phase locked, which preserves the original
requirement of phase locked DPLL outputs even when all
reference sources are lost.
The existence of Translation Profile 1.x is one of two conditions
necessary to enable cascaded DPLL operation. The other
condition is that the primary DPLL must have an assigned
inactive profile. The user makes this assignment via Bits[2:0] of
Register 0x102A (for DPLL0) or Register 0x142A (for DPLL1). By
default, Bits[2:0] = 000 (binary), which assigns Translation
Profile 0.0 (for DPLL0) or Translation Profile 1.0 (for DPLL1)
as the inactive profile assignment, but the user can specify any
one of the six translation profiles as the inactive profile via
Bits[2:0]. Programming Bits[2:0] with a value of x, where x is 0
to 5 (decimal) selects the corresponding Translation Profile 0.x
or 1.x, where x can be any integer value from 0 to 5). Figure 99
indicates Translation Profile 0.b as the inactive profile to
demonstrate a nondefault example.
The role of the primary and secondary DPLLs is reversible
because the role is dependent on how the user programs the
translation profiles and assigns the inactive profile index.
For the cascaded DPLL configuration to become active, the
following conditions are necessary:
•
A translation profile associated with the secondary DPLL
that assigns the feedback path of the primary DPLL as a
reference input must exist.
Consider the left side of Figure 99, which is the same as in
Figure 98 except for the fact that DPLL0 has Translation Profile B
assigned as its inactive profile. Like Figure 98, both DPLLs lock
to Reference Source 1. The middle portion of Figure 99 applies
when Reference Source 1 is lost. Like Figure 98, both DPLLs
lock to Reference Source 2. When Reference Source 2 is lost (as
shown in the right portion of Figure 99), however, and
assuming the user has configured a translation profile
(Translation Profile 1.x in this example) to use the feedback
path of the primary DPLL as a reference source, the following
events occur.
•
The primary DPLL must have an inactive translation
profile assignment (by default, the inactive profile is
Translation Profile 0 unless otherwise specified).
The primary channel of the APLL must be calibrated and
indicate locked status.
•
•
The primary DPLL must be inactive, that is, either in
freerun or holdover operation.
PROFILE B IS NOW
DPLL0
TRANSLATION PROFILE 0.a AND
CONFIGURED
TRANSLATION PROFILE 0.b AND
TRANSLATION PROFILE 1.y ASSIGN
REFERENCE SOURCE 2
INACTIVE, BUT DPLL0
BECOMES A VALID
TRANSLATION PROFILE 1.z ASSIGN
WITH
INACTIVE
INDEX
ASSIGNED TO
PROFILE B
REFERENCE SOURCE BY
VIRTUE OF THE INACTIVE
INDEX ASSIGNMENT
REFERENCE SOURCE 1
TRANSLATION
PROFILE 0.b
TRANSLATION
PROFILE 0.a
DIGITAL
CROSS
POINT
MUX
DIGITAL
CROSS
POINT
MUX
DIGITAL
CROSS
POINT
MUX
DPLL0
(PRIMARY)
DPLL0
DPLL0
TO
APLL0
TO
APLL0
TO
APLL0
PFD
LF NCO
PFD LF NCO
N
FTW
NCO
ALL
REFERENCE
SOURCES
REFERENCE
SOURCE 1
LOST
REFERENCE
SOURCES
LOST
REFERENCE
SOURCES
N
N
TDC
1
TRANSLATION
PROFILE 1.y
TRANSLATION
PROFILE 1.x
TRANSLATION
PROFILE 1.z
DPLL1
(SECONDARY)
TDC
2
TDC
2
DPLL1
DPLL1
TO
APLL1
TO
APLL1
TO
APLL1
PFD
LF NCO
PFD LF NCO
N
PFD
LF NCO
N
N
DPLL1 USES DPLL0 AS A
REFERENCE SOURCE
(DEFINED IN PROFILE 1.x)
Figure 99. Cascaded DPLL Configuration
Rev. 0 | Page 142 of 205
Data Sheet
AD9546
•
The user cannot specify fractional feedback division for the
primary DPLL (that is, the device ignores the fractional
portion of the feedback divider when operating in the
cascaded DPLL configuration). Furthermore, the user
must properly program the buildout N divider value in the
targeted inactive profile (even if the targeted inactive
profile is a hitless (internal) profile). See the DPLL
Feedback Divider (N Divider) section for details relating to
the buildout N divider.
CAVEATS OF CASCADED DPLL OPERATION
The recommendation is to avoid using reference synchronization
when employing cascaded DPLL operation (see the Reference
Synchronization section).
When the profile targeted by the assigned inactive profile index
becomes inactive, certain aspects of the target profile
configuration still matter (even though the target profile is
inactive), as follows:
•
The translation type (phase buildout or hitless—see the
Translation Modes section). For phase buildout mode, the
APLL associated with the primary DPLL must be in a
calibrated and locked state. For hitless (internal) mode, the
APLL associated with the primary DPLL must be in a
calibrated and locked state, and the selected distribution
path must be active (that is, a synchronization event has
occurred (see the Distribution Output Clock
Synchronization section) on the channel associated with
the secondary DPLL).
The device uses the information related to the feedback
configuration of the primary DPLL (even though the
primary DPLL is effectively operating open loop, that is,
holdover or freerun).
•
•
The user cannot specify the external zero delay configuration
for cascaded DPLL operation.
The device uses the information related to time stamp
tagging according to the primary DPLL. Specifically, the
tagging mode defined for the feedback path of the primary
DPLL determines whether time stamp tags are available on
the reference input path of the secondary DPLL. Thus, the
targeted inactive profile for the primary DPLL must use a
value of 2, 3, or 4 as the tag mode. In addition, the translation
profile for the secondary DPLL must use a value of 1, 3, or 4
as the tag mode (see Table 86 regarding the tag modes).
•
Rev. 0 | Page 143 of 205
AD9546
Data Sheet
ANALOG PLL (APLL)
for a given output frequency and to accommodate proper
operation over a broad temperature range, the AD9546
incorporates an automated VCO calibration feature.
APLL OVERVIEW
Figure 100 shows a block diagram of the APLL, which is valid
for both APLL0 and APLL1. The main components of the
APLL comprise the following:
VCO Calibration
To initiate VCO calibration of APLL0 or APLL1 independently,
write a Logic 1 to Bit 1 of Register 0x2100 (for APLL0) and
Register 0x2200 (for APLL1). Bit 0 is not autoclearing.
Therefore, the user must immediately restore Bit 0 to Logic 0,
because maintaining a static Logic 1 state may cause unexpected
device behavior.
•
•
•
•
An integrated VCO
A combined PFD and charge pump
A loop filter
A feedback divider (M divider)
The purpose of the APLL is
•
•
•
To provide a low noise output clock signal
To upconvert the DPLL output frequency to ~3 GHz
To suppress the spurious artifacts of the NCO
To calibrate both APLL channels together, instead of each channel
independently, write a Logic 1 to Bit 1 of Register 0x2000. Bit 1
calibrates the system clock PLL as well as both APLLs.
One option for checking the status of the APLL VCO calibration
process is via Bits[5:4] of Register 0x3100 (for APPL0) and
Register 0x3200 (for APLL1). Bit 4, when Logic 1, indicates the
APPL is busy calibrating. Bit 5, when Logic 1, indicates the
APPL is done calibrating. Bit 4 is also available as a physical
output signal via an appropriately configured Mx pin.
VOLTAGE CONTROLLED OSCILLATOR (VCO)
The VCO provides a spurious free, low noise RF clock signal in
the 3 GHz range that feeds the output distribution section, with
APLL0 as the source for the OUT0x outputs and APLL1 as the
source for the OUT1x outputs. The VCOs for APLL0 and
APLL1 cover nonoverlapping frequency ranges, as shown in
Table 87. The output frequency of the APPL is one half of the
VCO frequency due to the P divider (see Figure 100).
The other option for checking the status of the APLL VCO
calibration process is via Bits[1:0] of Register 0x3014 (for APLL0)
and Register 0x3019 (for APLL1). Because Bit 0 and Bit 1
represent the latched state transitions of Bit 4 and Bit 5,
respectively, they may represent a condition that is no longer
true. Therefore, the user must clear Bit 0 and Bit 1 via Bit 0 and
Bit 1, respectively, of Register 0x200F for APLL0 and Register
0x2014 for APLL1. Otherwise, the user may lose indication of
subsequent state changes of the VCO calibration process.
Table 87. APLL VCO Frequency and Gain
APLL
VCO Frequency (MHz)
VCO Gain (MHz/V)
0
1
2424 to 3232
3232 to 4040
60
100
The VCO has a frequency range that is approximately
800 MHz, which consists of many narrow, overlapping
frequency bands. To ensure the VCO uses the appropriate band
3.9nF
220nF
LFx
LDOx
7
2
REG 0x1080, BITS[6:0]
REG 0x1080, BIT 7
REG 0x1083, BIT 0
REG 0x1083, BITS[2:1]
REG 0x1083, BIT 3
AD9546
2
CHARGE
PUMP
CONTROL
REG 0x1082, BITS[1:0]
LOOP
FILTER
CONTROL
3
3
REG 0x1082, BITS[4:2]
REG 0x1082, BITS[7:5]
P DIVIDER
VCO
TO
PHASE
DETECTOR AND
CHARGE PUMP
LOOP FILTER
OUTPUT
÷2
DISTRIBUTION
REG 0x3100, BIT 3
REG 0x3014, BIT 2
REG 0x3014, BIT 3
REG 0x3100, BIT 4
REG 0x3100, BIT 5
REG 0x3014, BIT 0
8
LOCK
DETECTOR
REG 0x1081,
BITS[7:0]
VCO
CAL
REG 0x2100, BIT 1
REG 0x3014, BIT 1
NOTES
1. A RANGE OF ADDRESSES OR BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES ARE SPECIFIC TO APLL0
Figure 100. APLL Block Diagram
Rev. 0 | Page 144 of 205
Data Sheet
AD9546
For example, given ICP = 743 µA, determine the charge pump scale
value. Solving Equation 23 for the charge pump scale yields
APLL FEEDBACK DIVIDER (M DIVIDER)
The user programs the APLL feedback M divider via Bits[7:0] in
Register 0x1081 (for APLL0) and Register 0x1481 (for APLL1).
The divide ratio of the M divider is the decimal value of
Bits[7:0], yielding divide ratios from 1 to 255. A value of 0 is the
same as 1.
Charge Pump Scale = ICP/8 µA
= 743 µA/8 µA
= 93 (nearest integer)
= 0x5D (hexadecimal)
For example, given a desired divide ratio of 27, program
Bits[7:0] = 27 (decimal) or 0x1B (hexadecimal).
In automatic mode, the charge pump current scale value is
ineffective. Instead, the APLL automatically adjusts the charge
pump current (ICP) based on the value of the M divider
according to Table 88. This automatic adjustment yields a
relatively constant loop bandwidth for M divider values from
1 to 63. The loop bandwidth is ~250 kHz for APLL0 and
~300 kHz for APLL1. Note that automatic charge pump control is
only valid over a subset of M divider values. Given this
constraint, it is generally best practice to avoid automatic mode.
PHASE FREQUENCY DETECTOR (PFD)
The PFD detects the instantaneous phase error between the
feedback signal from the M divider and the output from the
DPLL. The phase error essentially drives the servo loop of the
APLL in a manner that ultimately nulls out the phase difference
between the two signals. The PFD bandwidth is wide enough to
handle signals originating from the NCO of the DPLL (up to
~350 MHz nominal).
Table 88. APLL Charge Pump Current in Automatic Mode
The APLLs incorporate lock detection circuitry that indicates
when they achieve a frequency locked condition. Locked status
indication is available via Bit 3 of Register 0x3100 (for APLL0)
and Register 0x3200 (for APLL1), where Logic 1 indicates
locked status. Bit 3 is also available as a physical output signal
via an appropriately configured Mx pin.
M Divider Value
ICP (µA)
M × 16
1016
1 to 63
64 to 255
Regardless of the operating mode (manual or automatic), the
charge pump has a provision for applying a constant dc offset
current to the output of the charge pump. Injection of an offset
current overcomes some of the spectral artifacts associated with
charge pump nonlinearity when the APLL is in a locked state.
Generally, noise performance improves significantly when this
feature is active and properly adjusted.
The IRQ section of the register map provides latched status bits
reflecting state transitions of the lock detector status via
Bits[3:2] of Register 0x3014 (for APLL0) and Register 0x3019 (for
APLL1). Bit 2 latches to Logic 1 on an unlocked to locked state
transition, whereas Bit 3 latches to Logic 1 on a locked to
unlocked state transition. Because Bit 2 and Bit 3 are latched
bits, the user must clear Bit 2 and Bit 3 via Bit 2 and Bit 3,
respectively, of Register 0x200F (for APLL0) and Register
0x2014 (APLL1). Otherwise, the user may lose indication of
subsequent state changes of the lock detector.
To enable or disable the dc offset current feature, use Bit 0 of
Register 0x1083 (for APLL0) and Register 0x1483 (for APLL1).
Logic 1 (default) enables this feature, whereas Logic 0 disables it.
To control the polarity (positive or negative) of the offset current,
use Bit 3 of Register 0x1083 (for APLL0) and Register 0x1483
(for APLL1). Logic 0 (default) is positive, whereas Logic 1 is
negative. The magnitude of the offset current depends on the
value of Bits[2:1] (unsigned) of Register 0x1083 (for APLL0)
and Register 0x1483 (for APLL1). The value of this bit field sets
the dc offset current as a fraction of ICP per Table 89, but with a
granularity of 8 µA. Thus, the offset current is in integer steps of
8 µA, with fractions of an 8 µA step rounded in the direction of
zero (that is, −53 µA rounds to −48 µA).
CHARGE PUMP
The charge pump consists of a pair of constant current sources
that deliver charge to or remove charge from the loop filter
based on the output of the phase detector. The transfer of
charge increases or decreases the voltage applied to the VCO,
which steers the VCO frequency to match the input frequency
and ultimately bring about a phase locked condition.
The charge pump has two operating modes, manual and
automatic, selectable via Bit 7 of Register 0x1080 (for APLL0)
and Register 0x1480 (for APLL1). Logic 1 (default) selects
manual mode, whereas Logic 0 selects automatic mode.
Table 89. APLL Charge Pump DC Offset Current
Bits[2:1] (Decimal)
DC Offset Current (% of ICP)
0
1
2
3
50
25 (default)
12.5
6.25
In manual mode, the user programs the charge pump current via
Bits[6:0] (unsigned) of Register 0x1080 and Register 0x1480.
The actual charge pump current (ICP) relates to Bits[6:0] (charge
pump scale) as follows:
In general, the default values of the dc offset current and charge
pump scale yield optimal overall performance. For this reason, the
recommendation is to use only the manual charge pump mode
(refer to the Charge Pump section regarding manual charge
pump mode).
I
CP = Charge Pump Scale × 8 µA
CP has a range of 0 µA to 1016 µA. The default value of
Bits[6:0] is 0x14 (20 decimal), which yields ICP = 160 µA.
(23)
I
Rev. 0 | Page 145 of 205
AD9546
Data Sheet
To program R1, C2 and R3 use Register 0x1082 (for APLL0)
and Register 0x1482 (for APLL1). Bits[7:5] (unsigned) control
R1, Bits[4:2] (unsigned) control C2, and Bits[1:0] (unsigned)
control R3. Table 90 shows how the decimal value of the bit
fields that select R1, C2 and R3 relate to the physical loop filter
component value.
APLL LOOP FILTER
The phase errors detected by the PFD drive the charge pump.
The charge pump transfers charge to the loop filter in proportion
to the amount of phase error. The loop filter stores the charge,
which creates a dc voltage that steers the VCO frequency in a
manner that nulls out the phase error.
Table 90. APLL Loop Filter Component Values
The loop filter is a resistor and capacitor filter with a low-pass
response characteristic that serves three purposes.
Bit Field Value
(Decimal)
R1 (Ω)
0
250
500
750
1000
1250
1500
C2 (pF)
R3 (Ω)
•
•
Storage of charge originating from the charge pump
Attenuation of the high frequency components of the
phase error signal
0
1
2
3
4
5
6
7
8 (default) 200 (default)
24
40
56
72
88
104
250
333
500
n/a
n/a
n/a
n/a
•
Stabilization of the feedback loop
Figure 101 shows the loop filter schematic. Three of the
components (R1, R3, and C2) are programmable, allowing the
user to tailor the response of the loop filter. The two external
capacitors (3.9 nF and 220 nF) are connected to the LFx and
LDOx pins. These two components are necessary for proper
operation of the loop filter and the internal LDO, respectively.
1750 (default) 120
In general, the default filter settings yield optimal performance.
Although altering the response of the APLL loop filter is not
necessary, the user can contact Analog Devices for assistance
with determining alternative component values for the APLL
loop filter to tailor the response to specific requirements.
3.9nF
220nF
LDOx
LFx
AD9546
R1
C2
30pF
FROM
CHARGE
PUMP
R3
TO
VCO
Figure 101. APLL Loop Filter
Rev. 0 | Page 146 of 205
Data Sheet
AD9546
REFERENCE SWITCHING
REFERENCE SWITCHING OVERVIEW
FORCED HOLDOVER MODE
The AD9546 has a sophisticated reference switching mechanism
administered by an internal controller. In general, the DPLL
operates in closed-loop mode, wherein the DPLL locks to the
frequency and phase of a reference input signal. However, if a
reference becomes invalid, the AD9546 can automatically
switch to a different reference input, depending on how the
user programs translation profiles (see the Translation Profiles
section). In the absence of any valid references, the AD9546
switches to one of the two open-loop operating modes:
holdover or freerun. The DPLL selects freerun, holdover, or
manual/automatic source selection per the DPLL mode
selection flowchart of Figure 102. The three procedures in
Figure 102 (holdover, translation profile validation, and
priority-based translation profile selection) have separate flow
charts per Figure 103, Figure 104, and Figure 105, respectively.
Although the automatic reference switching capability of the
AD9546 can cause the DPLL to switch to holdover mode, the
user can force the DPLL to holdover mode by writing Logic 1 to
Bit 1 of Register 0x2105 (DPLL0) or Register 0x2205 (DPLL1).
In holdover mode, like freerun mode, the DPLL operates open
loop and produces a static output frequency. However, the
DPLL uses one of three static tuning words dictated by the
tuning word history processor (see the Tuning Word History
section). Figure 103 shows a summary of the holdover tuning
word selection.
Note that freerun mode overrides holdover mode (see Figure 102).
Furthermore, freerun mode and holdover mode both override
the manual/automatic source selection modes (see the
Manual/Automatic Translation Profile Selection section).
MANUAL/AUTOMATIC TRANSLATION PROFILE
SELECTION
The DPLL has three status indicators to notify the user when
the DPLL begins the switch to a new translation profile
(inferring a new reference input) or when the DPLL enters
freerun or holdover mode. These indicators reside in Bits[2:0]
of Register 0x3101 and Register 0x3201, as follows:
Assuming Bits[1:0] = 1 decimal in Register 0x2105 (DPLL0) or
Register 0x2205 (DPLL1), the DPLL selects a translation profile
based on Bits[3:2] of Register 0x2105 (DPLL0) or
Register 0x2205 (DPLL1). Bits[3:2] support four translation
profile selection modes per Table 91.
•
•
•
Bit 2: switching translation profiles
Bit 1: holdover mode
Bit 0: freerun mode
Table 91. DPLL Translation Profile Select Mode Settings
Bits[3:2]
(Decimal)
Description
These indicators also inform corresponding IRQ status bits via
Bits[7:5] of Register 0x3011 and Register 0x3016, as follows:
0
Automatic, priority-based translation profile
selection
Manual translation profile selection with fallback
to priority-based translation profile selection
Manual translation profile selection with fallback
to holdover
Manual only translation profile selection
•
•
•
Bit 7: switched translation profiles
Bit 6: entered freerun mode
Bit 5: entered holdover mode
1
2
3
Bit 7 latches to Logic 1 when the DPLL is in the process of
switching to an active translation profile. Bit 6 latches to Logic
1 when the DPLL enters freerun mode. Bit 5 latches to Logic 1
when the DPLL enters holdover mode. Because Bits[7:5] are
latched bits, the user must clear them via Bits[7:5] of Register
0x200C and Register 0x2011 to obtain visibility of subsequent
state transitions to reference switching, freerun, or holdover
(see the Interrupt Request (IRQ) section).
Manual Only Translation Profile Selection
The manual only translation profile selection mode in Table 91
(Bits[3:2] = 3 decimal) is not a recommended operating mode
because this setting includes several operational caveats.
Contact Analog Devices for application support on using the
manual only translation profile selection mode.
Manual Translation Profile Selection with Fallback to
Holdover
FORCED FREERUN MODE
Although the automatic reference switching capability of the
AD9546 can cause the DPLL to switch to freerun mode, the
user can force the DPLL to freerun mode by writing Logic 1 to
Bit 0 of Register 0x2105 (DPLL0) or Register 0x2205 (DPLL1).
In freerun mode, the DPLL operates open loop and produces a
static output frequency as dictated by the 46-bit freerun tuning
word in Register 0x1000 to Register 0x1005 (DPLL0) and
Register 0x1400 to Register 0x1405 (DPLL1). Refer to the
Freerun Tuning Word section for details.
In the manual translation profile selection with fallback to
holdover mode (Bits[3:2] = 2 decimal), the user specifies a
particular translation profile for manual operation using
Bits[6:4] of Register 0x2105 (DPLL0) or 0x2205 (DPLL1). The
decimal value, y, of Bits[6:4] identifies the desired translation
profile, x.y (y is 0 to 5 and x is 0 or 1 for DPLL0 or DPLL1,
respectively). If the specified translation profile is not valid, the
DPLL switches to holdover operation (open loop). However,
depending on the status of the tuning word history processor
(see the Tuning Word History section), the DPLL may select
freerun operation instead of holdover.
Rev. 0 | Page 147 of 205
AD9546
Data Sheet
Manual Translation Profile Selection with Fallback to
Priority-Based Translation Profile Selection
Revertive and Nonrevertive Reference Switching
Revertive reference switching means that if the reference
associated with a translation profile (Translation Profile 0, for
example) becomes invalid and the DPLL switches to a new
translation profile (Translation Profile 1, for example), and the
reference associated with Translation Profile 0 becomes valid
again, the DPLL switches back (reverts) to Translation Profile 0,
even though the reference associated with Translation Profile 1
is valid.
In the manual translation profile selection with fallback to
priority-based translation profile selection mode (Bits[3:2] = 1
decimal), the user specifies a particular translation profile for
manual operation in the same manner as described in the
Manual Translation Profile Selection with Fallback to Holdover
section. However, if the selected profile is not valid, rather than
switching to holdover mode, the DPLL attempts to use the
priority-based selection process (see the Automatic, Priority-
Based Translation Profile Selection section for a description of
priority-based selection). If the DPLL is unable to find a
translation profile per the priority selection process, the DPLL
switches to holdover or freerun operation per Figure 103.
Nonrevertive reference switching means that if the reference
associated with a translation profile (Translation Profile 0.0, for
example) becomes invalid and the DPLL switches to a new
translation profile (Translation Profile 0.1, for example), the
DPLL continues to use Translation Profile 0.1 even if the
reference associated with Translation Profile 0.0 becomes valid
again. See the Automatic, Priority-Based Translation Profile
Selection section on how to assign a priority value to a
translation profile.
Automatic, Priority-Based Translation Profile Selection
The priority associated with a particular translation profile
depends on the programmed value of Bits[5:1] of the following
register addresses: 0x1200, 0x1220, 0x1240, 0x1260, 0x1280,
and 0x12A0 for DPLL0 and 0x1600, 0x1620, 0x1640, 0x1660,
0x1680, and 0x16A0 for DPLL1. A lower value of Bits[5:1]
means a higher priority of the associated input reference
source. That is, 31 is the lowest (least favored) selection
priority, and 0 is the highest (most favored) selection priority.
The automatic, priority-based translation profile selection
process provides for revertive and nonrevertive reference
switching using the difference between the priority of the
current translation profile and the priority of the previous
switched from translation profile to make the revertive or
nonrevertive decision. Specifically, if the priority value of the
current translation profile is at least 8 more than the priority
value of the switched from translation profile, the current
translation profile has significantly lower priority than the
switched from translation profile. Therefore, the DPLL reverts
to the switched from translation profile as soon as its reference
source becomes valid. Alternatively, if the priority value of the
current translation profile is 0 to 7 more than the priority value
of the switched from translation profile, the current translation
profile has only marginally lower priority than the switched
from translation profile. Therefore, the DPLL does not revert,
but remains with the current translation profile until its
reference source becomes invalid (or the user selects a new
translation profile manually).
In the automatic, priority-based translation profile selection
mode (Bits[3:2] = 0 decimal), the DPLL relies solely on the
automatic priority selection process shown in Figure 105 for
choosing a translation profile. However, if the DPLL is unable
to find a translation profile per the priority selection process,
the DPLL switches to holdover or freerun operation per
Figure 103.
As shown in Figure 105, the DPLL selects the most favored
translation profile when the current profile becomes invalid.
Translation Profile Validation
As shown in Figure 102, the manual and automatic translation
profile selection modes require validation of the selected
translation profile. Figure 104 shows a flowchart of the
translation profile validation process.
In summary, the priority separation between the current
translation profile and the previously switched from translation
profile determines whether reference switching is revertive or
nonrevertive. Specifically, revertive switching requires the
priority value of the switched from translation profile to be at
least 8 lower than the priority value of the current translation
profile. Otherwise, nonrevertive reference switching applies.
That is, a priority separation of 8 marks the boundary between
revertive and nonrevertive reference switching.
Part of the translation profile validation process requires
verification of the validity of the reference source associated
with the translation profile. Because the reference source
associated with a translation profile can be any of the following
five different source types (per Table 75), determining what
constitutes a valid reference depends on the source type:
•
•
REFA, REFAA, REFB, or REFBB reference input
Auxiliary REF0, auxiliary REF1, auxiliary REF2, or
Auxiliary REF3
•
•
•
IUTS0 or IUTS1
Auxiliary NCO 0 or NCO 1
DPLL0/DPLL1 feedback path
Figure 104 shows the source dependent validation decisions in
the reference source validation portion of the flowchart.
Rev. 0 | Page 148 of 205
Data Sheet
AD9546
(DPLL0) and Register 0x300A (DPLL1) (see the Interrupt
Active Profile
Request (IRQ) section).
When the DPLL selects a specific translation profile, whether
manually or automatically, the selected profile becomes active.
When a translation profile is active, the corresponding bit is
Logic 1 in Register 0x3009 (DPLL0) or Register 0x300A
(DPLL1). Bits[5:0] associate with Translation Profile x.5
through Translation Profile x.0, respectively, where x = 0 or 1.
An active translation profile is not the same as a valid translation
profile. That is, a profile can be valid without being active.
When all translation profiles for a particular DPLL are inactive, the
user can find the most recently active profile via Bits[6:4] of
Register 0x3101 (DPLL0) and Register 0x3201 (DPLL1). The
value of Bits[6:4] corresponds to the most recently active
profile. If a translation profile is currently active (as indicated
by Bits[5:0] in Register 0x3009 and Register 0x300A), the value
of Bits[6:4] in Register 0x3101 (DPLL0) and Register 0x3201
(DPLL1) identify the currently active profile.
The IRQ status section of the register map indicates Logic 0 to
Logic 1 transitions of Bits[5:0] of Register 0x3013 (DPLL0) and
Register 0x3018 (DPLL1). Because the IRQ bits are latched bits,
the user must clear them via Bits[5:0] of Register 0x200E (DPLL0)
and Register 0x2013 (DPLL1) to obtain visibility of subsequent
Logic 0 to Logic 1 transitions of Bits[5:0] in Register 0x3009
Rev. 0 | Page 149 of 205
AD9546
Data Sheet
START
FORCED
FREERUN
EXIT
DPLL USES
YES
YES
YES
REG 0x2105,
BIT 0
HOLDOVER
= 1?
REG 0x1005 TO 0x1000,
BITS[45:0]
PROCEDURE
ENTER
NO
FORCED
HOLDOVER
REG 0x2105,
BIT 1
= 1?
NO
AUTOMATIC,
NO
PRIORITY-BASED
TRANSLATION
PROFILE SELECTION
PRIORITY
SELECTION
PROCEDURE
REG 0x2105,
BITS[3:2]
PASS?
= 0?
NO
YES
SELECT TRANSLATION
PROFILE 0.x PER
DPLL USES SELECTED
TRANSLATION PROFILE
REG 0x2105, BITS[6:4]
MANUAL ONLY
TRANSLATION
PROFILE
1
SELECTION
YES
REG 0x2105,
BITS[3:2]
= 3?
NO
ENTER
TRANSLATION
PROFILE
VALIDATION
PROCEDURE
EXIT
DPLL USES
TRANSLATION
PROFILE 0.x
YES
PROFILE
VALID?
MANUAL
TRANSLATION
PROFILE
NO
SELECTION
WITH FALLBACK
TO HOLDOVER
YES
REG 0x2105,
BITS[3:2]
= 2?
REG 0x2105, BITS[3:2] = 1
NO
MANUAL TRANSLATION PROFILE
SELECTION WITH FALLBACK TO
PRIORITY- BASED TRANSLATION
PROFILE SELECTION
1
FOR DEBUG ONLY. THE DPLL USES TRANSLATION PROFILE 0.x WITHOUT VALIDATING THE SOURCE. THEREFORE, THE DPLL
IS NOT GUARANTEED TO FUNCTION NORMALLY. CONTACT ANALOG DEVICES FOR DETAILS.
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES ARE SPECIFIC TO DPLL0
Figure 102. DPLL Mode Selection Flowchart
Rev. 0 | Page 150 of 205
Data Sheet
AD9546
ENTER
HOLDOVER
PROCEDURE
HISTORY
AVAILABLE
YES
REG 0x3102,
BIT 0
= 1?
NO
SINGLE
SAMPLE
HISTORY
YES
REG 0x100E,
BIT 1
= 1?
NO
DPLL USES
DPLL USES
DPLL USES MOST
RECENT TUNING WORD
REG 0x1005:0x1000,
BITS[45:0]
REG 0x3108:0x3103,
BITS[45:0]
FREERUN
HISTORY
EXIT
NOTES
1. A RANGE OF ADDRESSES OR BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES ARE SPECIFIC TO DPLL0
Figure 103. Holdover Flowchart
Rev. 0 | Page 151 of 205
AD9546
Data Sheet
ENTER
TRANSLATION PROFILE
VALIDATION PROCEDURE
SYSTEM CLOCK
STABLE?
EXAMPLE
YES
NO
REG 0x3001,
BIT 1
= 1?
TRANSLATION
PROFILE 0.2
ENABLED?
NO
NO
NO
REG 0x1240,
BIT 0
= 1?
= 1?
= 1?
YES
APLL0
CALIBRATED?
REG 0x3100,
BIT 5
YES
APLL0
LOCKED?
REG 0x3100,
BIT 3
REFERENCE SOURCE VALIDATION
YES
DPLL1
FEEDBACK?
AUX NCO
0 OR 1?
REFA, REFAA,
REG 0x1241,
BITS[4:0]
REG 0x1241,
BITS[4:0]
REFB, REFBB,
Aux REF0, REF1,
REF2, REF3
NO
NO
NO
REG 0x1241,
BITS[4:0]
=
=
= 5?
0 TO 3,6,7,
11 TO 14?
8 OR 9?
OR IUTS0/IUTS1?
YES
YES
YES
EXAMPLE
DPLL1
FREERUN?
NO
NO
REG 0x3020,
BIT 4
REG 0x3201,
BIT 0
AUX REF1
VALID?
= 1?
= 1?
DPLL1
HOLDOVER?
YES
YES
NO
REG 0x3201,
BIT 1
= 1?
EXAMPLE
TRANSLATION
PROFILE 0.2
BUILDOUT
YES
YES
REG 0x1243,
BIT 0
= 0?
= 0?
APLL1
CALIBRATED?
EXAMPLE
NO
REG 0x3200,
BIT 5
NO
TRANSLATION PROFILE
0.2 HITLESS
EXAMPLE
= 1?
TRANSLATION PROFILE 0.2
EXTERNAL ZERO DELAY
NO
REG 0x1243,
BIT 1
YES
APLL1
LOCKED?
EXAMPLE
TRANSLATION PROFILE 0.2
INTERNAL ZERO DELAY
NO
YES
0.2
YES
REG 0x3200,
BIT 3
= 1?
EXTERNAL
DISTRIBUTION
FEEDBACK CHANNEL NOT
POWERED DOWN?
NO
NO
ZERO DELAY
REFERENCE SOURCE
VALID?
YES
YES
DISTRIBUTION
FEEDBACK PATH
SYNCED (OR SYNC REQUEST
NO
AWAITING TRIGGER)?
YES
VALID
INVALID
EXIT
NOTES
1. A RANGE OF ADDRESSES OR BITS USES A COLON SEPARATOR
2. REGISTER ADDRESSES ARE SPECIFIC TO DPLL0, TRANSLATION PROFILE 0.2 (EXAMPLE), AND AUXILIARY REF1 (EXAMPLE)
Figure 104. Flowchart for Translation Profile Validation
Rev. 0 | Page 152 of 205
Data Sheet
AD9546
ENTER
PRIORITY SELECTION
PROCEDURE
NO
IS THERE AN
ACTIVE PROFILE?
YES
IDENTIFY THE
PROFILE AS
PROFILE x.a
NO
IS PROFILE x.a
VALID ?
YES
LIST ALL VALID PROFILES
EXCLUDING PROFILE x.a
IDENTIFY ALL
VALID PROFILES
NO
IS THERE A
VALID PROFILE?
YES
IS THERE
A VALID PROFILE
IN THE LIST?
NO
YES
NOTE THE VALID PROFILE
WITH THE LOWEST
PRIORITY VALUE
NOTE THE PROFILE IN
THE LIST WITH THE
LOWEST PRIORITY VALUE
IDENTIFY THE
PROFILE AS
PROFILE x.b
IDENTIFY THE
PROFILE AS
PROFILE x.a
IS THE PROFILE x.b
PRIORITY VALUE AT LEAST
8 LESS THAN THE PROFILE
x.a PRIORITY VALUE?
NO
SELECT
PROFILE x.a
YES
SELECT
PROFILE x.b
FAIL
PASS
EXIT
Figure 105. Priority-Based Translation Profile Selection Flowchart
Rev. 0 | Page 153 of 205
AD9546
Data Sheet
TIME TO DIGITAL CONVERTER (TDC)
TDCs are at the core of the DPLL technology in the AD9546.
The purpose of a TDC is to continuously observe a clock signal,
note the occurrence of an edge transition, and generate a digital
time stamp identifying when the edge occurred.
INPUT CLOCK
TIME STAMPS
TDC
ONE PER INPUT
CLOCK EDGE EVENT
TIME BASE CLOCK
Because a TDC effectively digitizes rising edge clock events by
means of time stamps, it enables digital (that is, numeric)
processing of clock signals. In a conventional PLL, for example,
a PFD circuit measures the relative phase of the reference and
feedback signals and provides an analog representation of the
phase difference. Instead of an analog representation of the
phase difference, the AD9546 uses a pair of TDCs (one in the
reference path and one in the feedback path) to generate
reference and feedback time stamps, constituting a numeric
representation of the phase difference. With phase information
in numeric form, the AD9546 implements the entire PLL
function digitally by employing a numeric PFD, digital loop
filter and an NCO.
Figure 106. TDC Diagram
The AD9546 contains 10 TDCs.
•
•
•
Four at the output of the reference dividers associated with
the four reference inputs
Four at the output of the reference dividers associated with
the auxiliary reference inputs
Two associated with the two DPLLs at the output of the
feedback divider
Each TDC produces digital time stamps identifying each rising
edge of its input clock. That is, the TDCs generate time stamps
at a rate commensurate with their respective input clock period.
The user must, however, ensure the TDC input clock rate is
within the specified range of 1 Hz to approximately 200 kHz.
Figure 106 shows a simplified diagram of a TDC. The TDC has
a time base clock (which gives the TDC its sense of time), an
input clock (the signal in question), and an output consisting of
numeric time stamps.
The TDC time stamps derive from an internal time base that
originates with the device system clock. As such, the TDCs
benefit from the ability of the AD9546 to provide system clock
compensation, assuming the user activates system clock
compensation via Register 0x0280 to Register 0x0282 (see the
System Clock Compensation section).
Rev. 0 | Page 154 of 205
Data Sheet
AD9546
TIME STAMPS
Auxiliary NCOs and Inverse User Time Stamper (IUTS)
sections), but without using a TDC.
TIME STAMPS OVERVIEW
The AD9546 has 14 time stamp sources.
In general, the time stamp sources require advance knowledge
of the period of the applied input signal. For example, the
REFA TDC requires the user to enter the nominal period of the
REFA input signal. The user programs the nominal input
period via the register map.
•
•
•
•
•
•
•
•
•
•
•
•
•
•
REFA TDC
REFAA TDC
REFB TDC
REFBB TDC
DPLL0 feedback TDC
DPLL1 feedback TDC
Auxiliary REF0 TDC
Auxiliary REF1 TDC
Auxiliary REF2 TDC
Auxiliary REF3 TDC
Auxiliary NCO 0
Auxiliary NCO 1
IUTS 0
Time stamps produced by the time stamp sources are not
readily accessible by the user. Instead, the user accesses time
stamps from the various time stamp sources by means of a user
time stamp processor (see the User Time Stamp Processor
(UTSP) section and User Time Stamper (UTS) section).
DIGITAL CROSSPOINT MUX
The digital cross point mux is an interconnect matrix allowing
an array of time stamp destinations access to the various time
stamp sources within the AD9546. However, not every time
stamp source can connect to every time stamp destination.
Table 92 summarizes the interconnections of the various time
stamp sources to potential destinations, where a yes indicates a
valid connection and a blank cell means invalid connection.
IUTS 1
Most of the time stamp sources in the AD9546 are TDCs. The
Auxiliary NCOs and IUTSs also generate time stamps (see the
Table 92. Time Stamp Source to Destination Matrix
Time Stamp Destination
DPLL0
Reference
Input
DPLL1
Reference
Input
SYSCLK
Compensation PLL
Reference Input
Time Skew
Measurement
Input
User Time
Stamp
Processor Input (UTSx) Input
User Time
Stamper
Time Stamp Source
REFA TDC
REFAA TDC
REFB TDC
REFBB TDC
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
DPLL0 Feedback TDC
DPLL1 Feedback TDC Yes
Auxiliary REF0 TDC
Auxiliary REF1 TDC
Auxiliary REF2 TDC
Auxiliary REF3 TDC
Auxiliary NCO 0
Auxiliary NCO 1
IUTS 0
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
IUTS 1
Rev. 0 | Page 155 of 205
AD9546
Data Sheet
This pattern continues to 65,535. A decimal value of 0 (default)
disables time stamp tagging (no tags).
TAGGED TIME STAMPS
A tagged time stamp is a time stamp with a unique marker, or
tag, attached to it. Tagged time stamps allow the AD9546 to use
a lower frequency clock embedded within a higher frequency
clock. For example, by tagging every 10th edge of a clock signal,
there is a carrier signal (that is, the original clock signal) and a
divide by 10 version of the carrier embedded within the signal
resulting from the tagged time stamps.
The user can also shift the time position of the tagged flags
Bits[15:0] in Register 0x280D to Register 0x280E (for Auxiliary
NCO 0) and Register 0x284D to Register 0x284E (for Auxiliary
NCO 1). Bits[15:0] are an autoclearing bit field. The value, K,
represented by Bits[15:0], is a signed integer (twos complement)
indicating a number of fundamental periods of the auxiliary
NCO. Thus, for K = −3, the time stamp processor tags time
stamps three time stamps earlier than the Nth time stamp
locations indicated by the tag ratio described previously. K is
effectively a phase shift of the tagged time stamp period, where
K < N.
The various time stamp sources within the AD9546 can
generate tagged time stamps (the auxiliary NCOs, for example).
Likewise, certain time stamp destinations within the AD9546
are capable of processing only tagged time stamps while
ignoring untagged time stamps (the user time stamp
processors, for example, described in the User Time Stamp
Processor (UTSP) section). However, the two DPLLs are the
most important of the time stamp destinations capable of using
tagged time stamps (see the Time Stamp Tagging Options
section).
Note that when the auxiliary NCO is the reference to the DPLL
for an active translation profile (see the Translation Profiles
section), shifting the time position of a tag is not valid and must
be avoided. When using the tag shift feature, be sure that any
translation profiles using the auxiliary NCO as a DPLL
reference are not enabled, or if they are enabled, that the
associated DPLL is in freerun or holdover mode.
Tagged Auxiliary NCO Time Stamps
The auxiliary NCOs generate time stamps coincident with the
expiration of each period defined by the programmed auxiliary
NCO frequency (that is, at the rollover point of the phase
accumulator). For example, an NCO frequency of 100 Hz
results in the generation of time stamps at a 100 Hz rate with
each time stamp 10 ms greater than the previous one.
Tagged Reference Time Stamps
Tags associated with time stamps originating from the reference
TDCs are the result of demodulating an embedded clock (see
the Reference Demodulator section). Synchronization events
associated with demodulation generate tagged time stamps.
The user can program the DPLL to use the tagged time stamps
from a reference TDC.
The time stamp generator gives the user the option of tagging
auxiliary NCO time stamps via Bits[15:0] in Register 0x280B to
Register 0x280C (for Auxiliary NCO 0) and Register 0x284B to
Register 0x284C (for Auxiliary NCO 1).
Tagged Feedback Time Stamps
Tagged time stamps originating from the DPLL feedback path
are the result of synchronization events associated with the
modulation controller (see the Modulation section). The user
can program the DPLL to use the tagged time stamps from a
feedback TDC.
The value, N, represented by Bits[15:0] (tag ratio) results in every
Nth time stamp being tagged. For example, Bits[15:0] = 1 decimal
means every other time stamp is tagged. Similarly, a decimal
value of 2 means every third time stamp is tagged. Likewise, a
decimal value of 3 means every fourth time stamp is tagged.
Rev. 0 | Page 156 of 205
Data Sheet
AD9546
USER TIME STAMP PROCESSOR (UTSP)
UTSP 1). However, the user must assert an IO update operation
(see the IO Update section) prior to reading Bits[79:0]. The
UTSPs perform time stamp conversions as the time stamps
arrive from their prescribed time stamp sources. As such, the
UTSPs are event driven. Because of this event driven process,
the user cannot be certain that the data residing in Bits[79:0] is
indeed relevant. There are two methods for mitigating this
problem.
UTSP OVERVIEW
The user can view time stamps from the various time stamp
sources (see the Time Stamps section) via the UTSPs. The
AD9546 has two UTSPs with their respective output results
available via the register map, as shown in Figure 107. The
UTSPs convert time stamps originating from a given time
stamp source (see Table 92) to a time scale meaningful to the
user. The time scale conversion is necessary because the time
stamp sources use an arbitrary time scale internal to the device,
which is meaningless to the user.
The first method is to read Bits[79:0] periodically. However, to
avoid missing time stamps, the user must read Bits[79:0] at a
rate greater than the expected rate of the time stamp source
assigned to the input of the UTSP. Because consecutive time
stamps arriving from a time source cannot have the same value
(by definition), it guarantees that a time stamp reading
different from the last constitutes a new time stamp.
The user selects the desired time scale source for a UTSP via
Bits[7:6] of Register 0x2A12 (for UTSP 0) and Register 0x2A13 (for
UTSP 1). The value of Bits[7:6] selects the desired time scale per
Table 93. Note that either one of the UTSPs can use either one of
the auxiliary NCO time scales.
The second method is to use the IRQ feature set of the AD9546
(see the Interrupt Request (IRQ) section). Whenever a UTSP
completes a time stamp conversion, the corresponding status Bit
2 (for UTSP 0) or Bit 3 (for UTSP 1) of Register 0x300F
(assuming the status bit is unmasked) is set. Upon the status bit
becoming Logic 1, assert an IO update operation, and read
Bits[79:0] (see Figure 107).
Table 93. UTSP Time Scale Source
Bits[7:6] (Decimal)
Description
0
1
2
3
Auxiliary NCO 0 time scale (default)
Common time scale
Auxiliary NCO 1 time scale
Common time scale
To attach a particular time stamp source to a UTSP, use
Bits[4:0] (unsigned) of Register 0x2A12 (for UTSP 0) and
Register 0x2A13 (for UTSP 1). The value of Bits[4:0] select the
desired time stamp source per Table 94.
The event driven nature of the UTSP means that the user may
miss one or more user time stamps if multiple time stamps
occurred before the user reads Bits[79:0]. Bits[7:0] in
Register 0x3A1E (for UTSP 0) and Register 0x3A2A (for UTSP 1)
provide a way for the user to check for any missed input time
stamps. When Bits[7:0] indicate a value of zero, then the
current user time stamp is the latest one in a contiguous series.
When Bits[7:0] indicate a nonzero value, then the user missed
the indicated number of time stamps. Therefore, the current
time stamp constitutes the first one in a new series of time
stamps. Missed time stamps are lost and are not recoverable.
Table 94. UTSP Input Time Stamp Source
Bits[4:0] (Decimal)
Description
0
REFA TDC
1
REFAA TDC
2
REFB TDC
3
REFBB TDC
4
5
6
7
8
9
10
11
DPLL0 feedback TDC
DPLL1 feedback TDC
Auxiliary REF0 TDC (default for UTSP 0)
Auxiliary REF1 TDC (default for UTSP 1)
Auxiliary NCO 0 time scale
Auxiliary NCO 1 time scale
Auxiliary REF0 TDC
Auxiliary REF2 TDC
Auxiliary REF3 TDC
IUTS 0
In operation, the user must assert an IO update operation,
which refreshes Bits[79:0] and Bits[7:0] with the most recent
values. After asserting an IO update operation, the
recommendation is to read Bits[79:0] immediately followed by
reading Bits[7:0]. Doing so enables immediate assessment of
the value of Bits[7:0] as an integral part of capturing the time
stamp value. Bits[7:0] clear automatically when read.
12
13
Because the user specifies the expected input period for a TDC
via the register map, the TDC knows approximately how much
time must elapse between time stamp events. If an event
happens far outside the expected event period, the TDC flags
the time stamp associated with that event as an error via Bit 0 of
Register 0x3A1F (for UTSP 0) and Register 0x3A2B (for UTSP 1).
In general, the user must discard a time stamp when Bit 0 is
Logic 1.
14
IUTS 1
15 to 31
Not applicable
READING USER TIME STAMPS
The converted input time stamps are available at the output of
the UTSP via Bits[79:0] in Register 0x3A14 to Register 0x3A1D
(for UTSP 0) and Register 0x3A20 to Register 0x3A29 (for
Rev. 0 | Page 157 of 205
AD9546
Data Sheet
REG 0x2A12, BITS[7:6]
80
TIME STAMP SOURCES
DIGITAL CROSS
POINT MUX
LOCAL TIME SCALE
80
80
80
TIME
STAMPS
AUX NCO 0 TIME SCALE
REFERENCE
REFERENCE TIME
SCALE FOR UTSP 0
TDCS
AUX NCO 1 TIME SCALE
USER TIME
STAMP SOURCE 0
80
8
REG 0x3A1D TO 0x3A14, BITS[79:0]
TIME
STAMPS
AUXILIARY
REG 0x3A1E, BITS[7:0]
REG 0x3A1F, BIT 0
REG 0x300F, BIT 2
USER TIME STAMP
PROCESSOR 0
(UTSP 0)
REFERENCE TDCS
REG 0x2A12, BIT 5
TIME
STAMPS
DPLL FEEDBACK
TDCS
USER TIME
STAMP SOURCE 1
TIME
STAMPS
80
8
INVERSE
REG 0x3A29 TO 0x3A20, BITS[79:0]
REG 0x3A2A, BITS[7:0]
REG 0x3A2B, BIT 0
USER TIME
STAMPERS (IUTS)
USER TIME STAMP
PROCESSOR 1
(UTSP 1)
REG 0x2A13, BIT 5
TIME
STAMPS
AUXILIARY NCOS
REG 0x300F, BIT 3
AUX NCO 0
80
80
80
TIME SCALE
REFERENCE
TIME SCALE
FOR UTSP 1
LOCAL TIME SCALE
REG 0x2A12, BITS[4:0]
REG 0x2A13, BITS[4:0]
80
AUX NCO 1
TIME SCALE
AUX NCO 0 TIME SCALE
AUX NCO 1 TIME SCALE
REG 0x2A13, BITS[7:6]
CLOCK SOURCE
(DERIVED FROM SYSTEM CLOCK)
NOTE
1. A RANGE OF BITS USES A COLON SEPARATOR
Figure 107. User Time Stamp Processors
For example, suppose the selected time scale is auxiliary
NCO 0, which the user had programmed with a frequency of
10052.630025 Hz (corresponding to a period of
99.476455167760936272992897696939 μs). Thus, the time
associated with the example time stamp (177.0547628421) is
the time stamp value multiplied by the period of selected time
scale (99.476455167760936272992897696939 μs), which is
0.017612780178 sec (rounded to the nearest ps). That is, the
time stamp in this example indicates a time of approximately
17.6 ms.
INTERPRETING USER TIME STAMPS
The outputs of the UTSPs comprise 80-bit words with the least
significant bit weighted as 2−40. The units associated with a time
stamp converted by a UTSP depends on the selected time scale.
Specifically, the 80-bit user time stamp consists of a 40-bit
integer part (the 40 leftmost bits) and a 40-bit fractional part
(the 40 rightmost bits), as shown in Figure 108. Each count of
the integer part represents one period of the selected time scale.
The 40 fractional bits represent the fractional part of one period
of the selected time scale with a resolution of 2−40
.
Time stamps have a meaningful resolution of approximately
1 ps. Thus, digits in a time stamp calculation representing less
than 1 ps must be ignored.
80-BIT USER TIME STAMP
79 ...
40 39
...
0
INTEGER PART FRACTIONAL PART
–40
2
DECIMAL
POINT
In a digitized clocking application, the AD9546 produces a local
time scale. Because the local time scale advances at a rate
commensurate with external time keeping (rather than based
on periods of a programmed frequency as with the auxiliary
NCOs), the local time scale is better suited for relating time
stamps to external time.
Figure 108. User Time Stamp
For example, assume the 80-bit word is the hexadecimal
number, 0x 0000 0000 B10E 04F0 07C1, in which case the
decimal equivalent is 194,673,770,497,985.
Because the scale of the fractional part is 2−40, the time stamp
value (rounded to 10 decimal places) is
Time Stamp Value = 194,673,770,497,985 × 2−40
= 177.05476284207634307676926255226
= 177.0547628421
This time stamp value represents periods of the selected time
scale.
Rev. 0 | Page 158 of 205
Data Sheet
AD9546
TAGGED TIME STAMPS
UTSP WITH SYSTEM CLOCK COMPENSATION
The UTSPs can operate on tagged time stamps from the various
time stamp sources shown in Figure 107. When the selected time
stamp source is generating tagged time stamps, the user can
program the UTSP to only pay attention to the tagged time stamps,
while ignoring untagged time stamps. To enable tagged time stamp
processing, program Bit 5 of Register 0x2A12 (for UTSP 0) and
Register 0x2A13 (for UTSP 1) to Logic 1.
The AD9546 allows the user to apply system clock
compensation (see the System Clock Compensation section) to
the time stamp sources as well as to the auxiliary NCOs. This
poses a potential problem when the user selects an auxiliary
NCO as a time scale for a UTSP. Specifically, when the user
elects to apply system clock compensation to a time stamp
source and that time stamp source is the input to a UTSP, then
the user must apply the same system clock compensation to the
auxiliary NCO selected as the time scale for that UTSP. Otherwise,
the time stamps that are output by the UTSP are in error.
Rev. 0 | Page 159 of 205
AD9546
Data Sheet
TIMING SKEW MEASUREMENTS USING TWO TDCs
TIME STAMP SOURCES
DIGITAL CROSS
POINT MUX
TIME
STAMPS
REFERENCE
TDCS
SKEW
REFERENCE
TIME
62
62
SOURCE
STAMPS
REG 0x3A33 TO 0x3A2C, BITS[61:0]
AUXILIARY
REFERENCE TDCS
TAG
OFFSET
REG 0x2A15, BIT 5
REG 0x3A33, BIT 7
TIME
STAMPS
SKEW
REG 0x300F, BIT 4
MEASUREMENT
PROCESSOR
DPLL FEEDBACK
TDCS
SKEW
TARGET
SOURCE
REG 0x3A3B TO 0x3A34, BITS[61:0]
REG 0x3A3B, BIT 7
TIME
STAMPS
DRIFT
INVERSE
USER TIME
REG 0x2A16, BIT 5
TAG
STAMPERS (IUTS)
REG 0x300C, BIT 5
WINDOW
4
TIME
STAMPS
AUXILIARY NCOS
REG 0x2A15, BITS[4:0]
REG 0x2A16, BITS[4:0]
REG 0x2A14, BITS[3:0]
CLOCK SOURCE
(DERIVED FROM SYSTEM CLOCK)
NOTE
1. A RANGE OF BITS USES A COLON SEPARATOR
Figure 109. Skew Measurement Block Diagram
The UTSP (see the User Time Stamp Processor (UTSP)
section) allows the user to read time stamps originating from
various time stamp sources—the time stamps indicating the
occurrence of a rising edge at the signal source. Sometimes,
however, it is desirable to know the relative phase offset
between the rising edges of two different signal sources rather
than the corresponding time associated with each edge. To this
end, the AD9546 provides the ability to measure and report the
time difference (timing skew) between two time stamps
(usually, but not exclusively, from different sources). The time
difference between time stamps relates directly to phase offset.
Table 95. Skew Measurement Window Settings
Bits[3:0]
N for Skew Offset
N for Skew Drift
0
2
2
1
4
4
2
8
8
3
4
5
6
7
8
9
10
11
12
13
14
16
32
64
16
16
16
16
16
16
16
16
16
16
16
16
128
256
512
1024
2048
4096
8192
16384
32768
Time skew (phase offset) only has meaning when two clock
signals operate at the same frequency. Otherwise, the phase
offset is not constant but drifts over time. Figure 109 shows a
block diagram of the skew measurement processor.
Time skew measurements involve a reference time stamp
source and a target time stamp source. The user assigns the
reference time stamp source via Bits[4:0] of Register 0x2A15 and
the target time stamp source via Bits[4:0] of Register 0x2A16.
The decimal value of Bits[4:0] selects the time stamp source per
Table 94, except that the default selection is REFA for both the
reference time stamp source and target time stamp source.
The selected skew window size bit field value sets the number
of averaging samples for both skew offset and skew drift
measurements. However, a given selection may yield a different
number of averaging samples for offset measurements vs. drift
measurements, as shown in Table 95.
Unlike the results from the UTSP, which constitute individual
measurements, the offset and skew measurements from the
skew measurement processor constitute an averaged group of
sequential measurements. The user selects the number, N, of
samples averaged per measurement via Bits[3:0] of
Register 0x2A14 according to Table 95.
The skew offset and drift measurements constitute a low-pass
filtered sequence of N samples. That is, the skew measurement
processor smooths out (that is, averages) the measurement
variation over N samples. After an N sample average is
established, subsequent measurements include the next sample
and the preceding N − 1 samples.
Rev. 0 | Page 160 of 205
Data Sheet
AD9546
As such, the processor requires a minimum of N samples to
compute the desired average, after which the processor
continues to update the average one sample at a time. The skew
measurement processor provides the user with an indication
that it has captured the first block of N samples by setting Bit 7
of Register 0x3A33 (for an offset measurement) or
For example, consider a 62-bit skew offset reading of 0x 3FFF
FFF8 A673 24B5 (hexadecimal), which is −31,567,174,475
decimal. To find the skew offset time, scale the decimal value by
2−16 ps, which results in −481,676.8566131591796875 ps. The
negative sign implies the target source leads the reference
source by approximately 482 ns.
Register 0x3A3B (for a drift measurement).
The result of a skew drift measurement appears in Bits[61:0]
(signed, twos complement) of Register 0x3A34 to
Register 0x3A3B. The decimal value of Bits[61:0] has units of
2−16 picoseconds per unit interval (ps/UI). That is, the value is
relative to the measured period of the reference source. Skew
drift is a measure of the rate at which the skew offset varies
cycle by cycle.
Bit 7 is Logic 0 for the first N − 1 measurements in a sequence
and Logic 1 for the Nth measurement. As such, when the user
reads the result of a skew offset or skew drift measurement, the
value is a fully averaged result if Bit 7 (associated with the
measurement type: offset or drift) is Logic 1. When Bit 7 is
Logic 0, offset or drift readings captured exhibit incomplete
averaging but are still useable. However, the variance of the
results is greater than for a fully averaged measurement.
For example, consider two signals with a nominal frequency of
100 Hz. A skew drift reading of 0x 3FFF FFF8 A673 24B5
(hexadecimal) is −31,567,174,475 decimal. To find the skew
drift, scale the decimal value by 2−16, which results in
−481,676.8566131591796875 ps/UI. Because the reference
period is 10 ms (1/100 Hz), the result indicates the target source
period is decreasing 482 ns every 10 ms, implying a frequency
offset of −48.2 ppm (−482 ns/10 ms = −48.2 × 10−6).
Unlike the results from the UTSPs, the skew measurement
processor results are in absolute units of picoseconds. To access
the time skew measurement processor results, use the skew offset
and skew drift bit fields (Register 0x3A2C to Register 0x3A33
and Register 0x3A34 to Register 0x3A3B, respectively).
The result of a skew offset measurement appears in Bits[61:0]
(signed, twos complement) of Register 0x3A2C to
Two signals with the same frequency have no drift and must
yield a drift measurement of zero. However, a frequency
difference that is too great can exceed the limits of the skew
processor, which is 1/16th UI. Upon reaching this limit, the
skew measurement processor sets the status Bit 5 of Register
0x300C to Logic 1.
Register 0x3A33 and notification that a measurement is ready
appears as an IRQ in Bit 4 of Register 0x300F. The decimal
value of Bits[61:0] constitute time in units of 2−16 ps (a range of
approximately 35 sec). Skew offset measurements assume the
reference and target sources are of nearly the same frequency.
That is, skew is meaningless for two sources with differing
frequency, because the skew constantly drifts in such a case.
The skew measurement processor measures skew relative to the
assigned skew reference source (the processor computes a
coarse average of the reference period). The skew measurement
processor determines the polarity of the skew offset
measurement by considering a time window spanning ½ unit
interval (UI), where 1 UI is the measured period of the
reference source. A negative value means time stamps from the
target source lead time stamps from the reference source by as
much as ½ UI, whereas a positive value means time stamps
from the target source lag time stamps from the reference
source by as much as ½ UI. Note that jitter can affect the
polarity of the skew measurement result.
TAGGED SKEW MEASUREMENT TIME STAMPS
The skew measurement processor can operate on tagged time
stamps from the various time stamp sources shown in
Figure 109. When the selected skew reference and/or skew
measurement time stamp source is generating tagged time
stamps, the user can program the skew measurement processor
to only pay attention to the tagged time stamps, while ignoring
untagged time stamps. To enable tagged skew reference time
stamp processing, program Bit 5 of Register 0x2A15 to Logic 1.
To enable tagged skew target time stamp processing program
Bit 5 of Register 0x2A16 to Logic 1.
Rev. 0 | Page 161 of 205
AD9546
Data Sheet
AUXILIARY NCOS
Auxiliary NCO Center Frequency
AUXILIARY NCO OVERVIEW
Center frequency assignment is via Bits[55:0] (unsigned) of
Register 0x2800 to Register 0x2806 (for auxiliary NCO 0) and
Register 0x2840 to Register 0x2846 (for auxiliary NCO 1).
Bits[55:0] carry units of 2−40 Hz. Thus, the center frequency
value of Bits[55:0] relates to the center frequency, fCENTER, as
The AD9546 has two auxiliary NCOs as shown in Figure 111. The
NCOs generate time stamps and/or physical output signals based
on a very precise programmable frequency specified by the user.
The auxiliary NCOs also provide a time scale to the UTSPs (see the
User Time Stamp Processor (UTSP) section). The NCOs have
frequency tuning resolution of 2−40 Hz, or approximately 1 pHz.
f
CENTER = Center Frequency Value × 2−40
AUXILIARY NCO FREQUENCY
For example, given a desired center frequency of 10 kHz, find the
required 56-bit register value.
The user programs the output frequency of the auxiliary NCOs
(fAUX) by assigning a center frequency and an offset frequency.
By default, the offset frequency value is zero, which means the
default output frequency is the center frequency. Figure 110
details the relationship between the center frequency (56 bits) and
offset frequency (32 bits) register values.
Center Frequency Value = fCENTER × 240
= 104 × 240
= 10,995,116,277,760,000
= 0x 27 1000 0000 0000 (hexadecimal)
The center frequency upper limit is 65,536 Hz (256 × 2−40 Hz).
The auxiliary NCO is effectively a direct digital synthesizer (DDS)
with a phase accumulator. The DDS accumulates a given phase
value, θ, at a prescribed rate, fACCUM
.
fACCUM = fS/96
where fS is the system clock frequency (see the System Clock PLL
section). The auxiliary NCO calculates θ such that the
accumulator rolls over with a period of precisely 1/fAUX
.
OUTPUT = CENTER + OFFSET
56-BIT CENTER FREQUENCY
... 0
CENTER 55
...
40
24
39
23
32-BIT OFFSET FREQUENCY
–40
2
Hz
OFFSET 31
...
...
0
–24
2
Hz
INTEGER Hz
FRACTIONAL Hz
DECIMAL
POINT
Figure 110. Auxiliary NCO Frequency Tuning
Rev. 0 | Page 162 of 205
Data Sheet
AD9546
When the user programs Bits[39:0] and asserts the IO update
bit, the AD9546 automatically imposes a specified upper limit
on how fast the phase can change (phase slew). See the
Auxiliary NCO Phase Slew Limit section for details.
Auxiliary NCO Offset Frequency
Offset frequency assignment is via Bits[31:0] (unsigned) of
Register 0x2807 to Register 0x280A (for auxiliary NCO 0) and
Register 0x2847 to Register 0x284A (for auxiliary NCO 1).
Bits[31:0] carry units of 2−24 Hz. The offset frequency value of
Bits[31:0] relates to the offset frequency, fOFFSET, as
Absolute Phase Offset
Absolute phase offset adjustment is in effect when Bit 0 = 0 of
Register 0x280F (for auxiliary NCO 0) and Register 0x284F (for
auxiliary NCO 1). For absolute phase offset adjustments, the phase
offset bits, Bits[39:0], constitute a signed absolute time offset with
1 ps resolution. The signed absolute time offset allows an offset
range of approximately 0.55 sec ( 10−12 sec × 239). However,
the user must limit programmed offset value to 1 UI (that is,
plus or minus one period of the auxiliary NCO).
f
OFFSET = Offset Frequency Value × 2−24
For example, given a desired offset frequency of 100 Hz, find the
required 32-bit register value.
Offset Frequency Value = fOFFSET × 224
= 102 × 224
= 1,677,721,600
Note that programming a phase offset value in excess of the
1 UI limit causes the device to set the status in Bit 5 or Bit 7 in
Register 0x3002 corresponding to auxiliary NCO 0 or auxiliary
NCO 1, respectively.
= 0x 6400 0000 (hexadecimal)
The offset frequency upper limit is 256 Hz (232 × 2−24 Hz).
The main purpose for the offset frequency is to give an external
processor or controller that is limited to 32-bit processing
capability the ability of rapidly changing the output frequency.
For example, it is cumbersome for a 32-bit machine to compute
the 56-bit center frequency and then deliver the computation to
the register map.
For example, given a programmed auxiliary NCO 0 frequency
of 10 kHz (per the center frequency and offset frequency
settings) and a desired phase offset of −15.5°, determine the
necessary value of Bits[39:0] in Register 0x2814 to Register
0x2818, the absolute time offset.
The 32-bit offset frequency option alleviates this problem
because the user can initially program an appropriate 56-bit
center frequency and then use 32-bit offset values to change the
output frequency. The 32-bit offset value allows a 32-bit
machine to operate in its native 32-bit environment and
facilitates rapid changing of the output frequency. The offset
values offer less precise tuning than the center frequency
(2−24 Hz vs. 2−40 Hz resolution).
First, ensure the phase offset is less than 360° to satisfy the
1 UI maximum phase offset requirement. The desired phase
offset of −15.5° is within the 1 UI limit.
Next, determine the time, t, associated with a −15.5° phase
offset.
t = (−15.5°/360°)/(10 kHz)
= −4.3055555556 µs (to 10 decimal places)
Finally, convert the value of t to units of ps.
Bits[39:0] = t/(10−12 sec)
Depending on the tuning range for a given application, the user
can use a combination of the center and offset values to minimize
the number of 8-bit registers required to cover the range, yet
still yield 2−40 Hz frequency resolution. For example, the user can
select a center and offset value that allows the offset to remain
fixed over the tuning range yet constrains tuning changes of the
center value to the lower 8, 16, 24, or 32 bits of the center
frequency value. Using only the LSBs of the center frequency
allows 32-bit (or less) register operations while still maintaining
2−40 Hz frequency tuning resolution.
= (−4.3055555556 × 10−6 s)/(10−12 sec)
= −4,305,556 (nearest integer)
= 0x FF FFBE 4D6C (40-bit hexadecimal)
Relative Phase Offset
Relative phase offset adjustment is in effect when Bit 0 = 1 of
Register 0x280F (for auxiliary NCO 0) and Register 0x284F (for
auxiliary NCO 1). For relative phase offset adjustments, the phase
offset bits, Bits[39:0], constitute a signed relative offset specified
as a fraction of the period of the auxiliary NCO. Thus, Bits[39:0]
cover a range from −½ UI to 1 LSB less than +½ UI.
AUXILIARY NCO PHASE OFFSET
The user can adjust the phase of the auxiliary NCO via
Bits[39:0] (signed, twos complement) in Register 0x2814 to
Register 0x2818 (for auxiliary NCO 0) and Register 0x2854 to
Register 0x2858 (for auxiliary NCO 1). The specific functionality
of the phase adjustment depends on the value of Bit 0 of
Register 0x280F (for auxiliary NCO 0) and Register 0x284F (for
auxiliary NCO 1), as explained in the Absolute Phase Offset
section and the Relative Phase Offset section.
For example, given a programmed auxiliary NCO 0 frequency
of 10 kHz (per the center frequency and offset frequency settings)
and a desired phase offset of −15.5°, determine the necessary
value of Bits[39:0] in Register 0x2814 to Register 0x2818, the
relative time offset.
First, the phase offset must be within 180° to satisfy the ½ UI
maximum offset requirement. The desired phase offset of
−15.5° is within the ½ UI limit.
Rev. 0 | Page 163 of 205
AD9546
Data Sheet
Next, convert −15.5° to fractional UI.
ELAPSED TIME ADJUSTMENT
Fractional UI = −15.5°/360°
As shown in Figure 111, the user can read the auxiliary NCO
time scale via Bits[79:0] of Register 0x3A00 to Register 0x3A09
(for auxiliary NCO 0) and Register 0x3A0A to Register 0x3A13
(for auxiliary NCO 1). The format of the time scale is a 40-bit
integer part (most significant bits) and a 40-bit fractional part
(least significant bits). The integer part keeps track of elapsed
periods of the auxiliary NCO. The fractional part effectively
represents the phase of the accumulator within the current
cycle period.
= −0.0430555556 (to 10 decimal places)
Because Bits[39:0], the phase offset bits, constitute a signed, twos
complement 40-bit number spanning 1 UI, calculate the value
of Bits[39:0] as follows:
Bits[39:0] = −0.0430555556 × 240
= −47,340,083,974 (nearest integer)
= 0x F4 FA4F A4FA (40-bit hexadecimal)
AUXILIARY NCO PHASE SLEW LIMIT
The contents of the time scale registers indicate the elapsed
time up to the moment of the most recent IO update.
When the user requests the application of a phase offset on one
of the auxiliary NCOs, the AD9546 does not necessarily apply
the full phase offset all at once. Instead, Bits[31:0] (unsigned) of
Register 0x2810 to Register 0x2813 (for auxiliary NCO 0) and
Register 0x2850 to Register 0x2853 (for auxiliary NCO 1)
control whether the phase offset is immediate or slew limited.
The user has the option of adjusting the 40-bit integer portion
of the time scale register based on Bits[39:0] of Register 0x2819
to Register 0x281D (for auxiliary NCO 0) and Register 0x2859
to Register 0x285D (for auxiliary NCO 1). The specific
functionality of the adjustment depends on the value of Bit 1 of
Register 0x280F (for auxiliary NCO 0) and Register 0x284F (for
auxiliary NCO 1), as explained in the Direct Elapsed Time
Adjustment section and the Incremental Elapsed Time
Adjustment section.
When the phase slew limit bits, Bits[31:0], have a decimal value
of zero, phase offset is immediate. That is, the auxiliary NCO
applies the specified phase offset value all at once. However,
when Bits[31:0] are nonzero, phase offset is slew limited. That
is, the auxiliary NCO gradually introduces phase changes by
regulating the rate at which the phase changes occur to prevent
exceeding the limit imposed by the value of Bits[31:0]. The
auxiliary NCO keeps track of the total phase change and
terminates the phase slew limiting process when the total phase
change satisfies the value established by the auxiliary NCO
phase offset parameters (see the Auxiliary NCO Phase Offset
section).
Bits[31:0] carry units of 2−36 UI/UI, where UI is the unit interval
of the associated auxiliary NCO (1 UI is the reciprocal of the
programmed frequency of the auxiliary NCO). That is, the user
specifies the maximum rate of change of phase as fractions of a
cycle per cycle. Note the user must program Bits[31:0] prior to
invoking a phase offset.
An application benefiting from elapsed time adjustment is time
of day indication, where the user adjusts the integer part of the
time scale register to match UTC time. Such an application
assumes programming the auxiliary NCO to 1 Hz, which
equates to a period of 1 sec.
Direct Elapsed Time Adjustment
Direct elapsed time adjustment is in effect when Bit 1 = 0 of
Register 0x280F (for auxiliary NCO 0) and Register 0x284F (for
auxiliary NCO 1). For direct elapsed time adjustments, Bits[39:0]
constitute an unsigned integer that directly replaces the
corresponding 40-bit integer portion of the auxiliary NCO time
scale register.
Incremental Elapsed Time Adjustment
Incremental elapsed time adjustment is in effect when Bit 1 = 1
of Register 0x280F (for auxiliary NCO 0) and Register 0x284F
(for auxiliary NCO 1). For incremental elapsed time adjustments,
the elapsed time adjustment bits (Bits[39:0] in Register 0x281D to
Register 0x2819 for auxiliary NCO 0 and Register 0x285D to
Register 0x2859 for auxiliary NCO 1), constitute a signed, twos
complement integer that increments or decrements the 40-bit
integer portion of the auxiliary NCO time scale register relative
to its current value.
For example, assume the programmed frequency of the
auxiliary NCO is 60 Hz (therefore, 1 UI = 1/60 sec) and the
desired limit for the rate of change of phase is 1 μs/sec
(10−6 sec/sec). Note that sec/sec and UI/UI are fundamentally
equivalent. Therefore, 10−6 sec/sec is 10−6 UI/UI. Because
Bits[31:0] are in units of 2−36 UI/UI, the required value for
Bits[31:0] is 10−6/2−36 = 68,719 (rounded to the nearest integer)
or 0x 0001 0C6F hexadecimal. Assuming the user programs
Bits[31:0] to 68,719 (decimal), when the user invokes the phase
offset specified by the auxiliary NCO phase offset parameters
(see the Auxiliary NCO Phase Offset section), the AD9546
automatically ensures that the rate of change of phase does not
exceed the specified value of 1 μs per second.
AUXILIARY NCO TIME STAMPS
The auxiliary NCO generates internal time stamps marking the
rollover of its phase accumulator. Because the accumulator
updates with a quantized period of 1/fACCUM, the accumulator
rarely rolls over to exactly zero (that is, most of the time, the
accumulator rolls over with a remainder). Therefore, the
auxiliary NCO performs the appropriate interpolation to
When Bits[31:0] are nonzero and while the auxiliary NCO is in
the process of limiting the phase slew rate, the device sets a
status bit in Register 0x3002: Bit 4 for auxiliary NCO 0 or Bit 6
for auxiliary NCO 1.
Rev. 0 | Page 164 of 205
Data Sheet
AD9546
ensure that the internally generated time stamps precisely mark
the zero-crossing point of the accumulator.
auxiliary NCO 1). To program E, use Bits[7:4] of the same
registers.
Pulse Width = (1/fACCUM) × (1 + S × 2(E + 5)
For example, given fS = 2.4 GHz, S = 6, and E = 2, then
Pulse Width = (1/fACCUM) × (1 + S × 2(E + 5)
= (96/fS) × (1 + 6 × 2(2 + 5)
The internal time stamps generated by the auxiliary NCO relate
to the internal time scale of the AD9546.
)
Because the internal time stamps truly mark the rollover period
of the auxiliary NCO accumulator as well as relate to the
internal time scale, the time stamps can serve as an input
reference source to either of the DPLLs. This feature allows the
AD9546 to have a completely self contained reference source to
the DPLLs (that is, no external reference source required).
)
)
= (96/(2.4 × 109)) × (1 + 6 × 2(2 + 5)
= (96/(2.4 × 109)) × (1 + 6 × 27)
= 30.76 ꢀs
)
Because the auxiliary NCO time stamps are an available
reference source, the user has access to them through the
UTSPs (see the User Time Stamp Processor (UTSP) section).
For example, suppose the user programs auxiliary NCO 0 for
fAUX = 5 kHz. To read the time stamps from auxiliary NCO 0,
the user can program UTSP 0 to use auxiliary NCO 0 as both a
time stamp source and as a time scale for UTSP 0. Then,
reading the first time stamp yields a reference value related to
the time scale of auxiliary NCO 0. However, the next time
stamp has a value of 1 unit greater than the first, because the
time scale has units of 200 μs (per fAUX). Unfortunately, this
method of time stamp conversion by UTSP 0 is not conducive
to external use.
However, given a desired pulse duration (tP), one must
determine the appropriate E and S values for a given system
clock PLL VCO frequency (fS). To do so, let
K = (tP × fS)/3072 – 1/32
Next, determine E as
E = ceil(log(K/15)/log(2))
where:
0 ≤ E ≤ 15.
E = ceil(x) means E = x if x is an integer. Otherwise, E is the
nearest integer to x in the positive direction.
Finally, determine S as
However, by relating the time stamps to an appropriate time
scale, the time stamp conversion can become meaningful. To
S = round(K/2E)
where:
0 ≤ S ≤ 1.
do so, program auxiliary NCO 1 for fAUX = 1 Hz. Because fAUX
1 Hz the time scale for auxiliary NCO 1 carries units of 1 sec.
Next, program UTSP 0 to use the time scale from auxiliary
NCO 1 instead of the time scale from auxiliary NCO 0 (but
keep auxiliary NCO 0 as the time stamp source to UTSP 0).
Then, reading the first time stamp yields a reference value
=
round(x) means round x to the nearest integer.
When the computation of E or S yields a result outside the
specified range, use the appropriate limit. For example, if the
computation of E yields a result of –2, then use E = 0.
(related to the time scale of auxiliary NCO 1) and the next time
stamp corresponds to a value precisely 200 μs later than the
first with subpicosecond precision.
The K, E, and S values for tP = 100 μs and fS = 2.4 GHz are K =
78.09375, E = 3, and S = 10.
E = 3 and S = 10 are within their respective range limits.
Because E and S must be integers, the resulting pulse duration
is quantized. Given the K, E, and S results for a desired pulse
period of tP = 100 μs, the actual pulse period is
AUXILIARY NCO PULSE OUTPUT
In addition to generating time stamps, the auxiliary NCOs also
produce a physical output signal (see Figure 111). The output
signal is effectively a single pulse commensurate with the
rollover of the internal phase accumulator of the auxiliary
NCO. The auxiliary NCO quantizes the pulse timing such that
the peak-to-peak pulse jitter is approximately 1.4 ns.
tP = (96/fS) × (1 + S × 2(E + 5)
)
= 102.44 μs
Because the output frequency (fAUX) of the auxiliary NCO and
the output pulse width (tP) are independently programmable
parameters, the user must ensure
The user can route the output pulse signal to any one of the Mx
pins by defining the desired Mx pin as an output and selecting
the desired auxiliary NCO as the output source (see the Status
and Control Pins section).
tP ≤ 1/fAUX
Although the user has the option of generating pulses
commensurate with tagged time stamps (see the Tagged
Auxiliary NCO Time Stamps section), the tP ≤ 1/fAUX limit on tP
still applies (that is, based on fAUX and not the tag frequency).
The user programs the duration of the pulse via a significand
(S) value and exponent (E) value. To program S, use Bits[3:0] of
Register 0x281E (for auxiliary NCO 0) and Register 0x285E (for
Rev. 0 | Page 165 of 205
AD9546
Data Sheet
CENTER
56
32
REG 0x2806 TO 0x2800,
BITS[55:0]
AUXILIARY
NCO 0
56
40
FREQ
OFFSET
REG 0x280A TO 0x2807,
BITS[31:0]
TIME
REG 0x2818 TO 0x2814,
BITS[39:0]
PHASE
STAMPS
TO DIGITAL
CROSS POINT
MUX
REG 0x280F, BIT 0
PHASE TYPE
32
40
REG 0x2813 TO 0x2810,
BITS[31:0]
PHASE SLEW
AUX NCO 0
TIME SCALE
80
REG 0x281D TO 0x2819,
BITS[39:0]
TIME ADJUST
TIME ADJUST TYPE
Mx PIN
SWITCH
MATRIX
REG 0x280F, BIT 1
Mx
PINS
4
4
PULSE
REG 0x281E, BITS[7:4]
REG 0x281E, BITS[3:0]
M0
PULSE EXPONENT
PULSE SIGNIFICAND
REG 0x3002,
BIT 5
16
16
STATUS
REG 0x280C TO 0x280B,
BITS[15:0]
TAG RATIO
TAG SHIFT
REG 0x3002,
BIT 4
CSB/M6
REG 0x280E TO 0x280D,
BITS[15:0]
DERIVED FROM
SYSTEM CLOCK
CLOCK
SOURCE
CENTER
56
REG 0x2846 TO 0x2840,
BITS[55:0]
AUXILIARY
NCO 1
TIME
56
40
STAMPS
FREQ
REG 0x3A09 TO 0x3A00,
BITS[79:0]
OFFSET
32
REG 0x284A TO 0x2847,
BITS[31:0]
AUX NCO 1
TIME SCALE
TO UTSP 0
AND UTSP 1
80
REG 0x2858 TO 0x2854,
BITS[39:0]
PHASE
REG 0x3A13 TO 0x3A0A,
BITS[79:0]
REG 0x284F, BIT 0
PHASE TYPE
32
40
REG 0x2853 TO 0x2850,
BITS[31:0]
PHASE SLEW
PULSE
REG 0x285D TO 0x2859,
BITS[39:0]
TIME ADJUST
TIME ADJUST TYPE
REG 0x3002, BIT 7
REG 0x3002, BIT 6
REG 0x284F, BIT 1
REG 0x285E, BITS[7:4]
REG 0x285E, BITS[3:0]
STATUS
4
4
PULSE EXPONENT
PULSE SIGNIFICAND
16
16
REG 0x284C TO 0x284B,
BITS[15:0]
TAG RATIO
TAG SHIFT
REG 0x284E TO 0x284D,
BITS[15:0]
NOTE
1. A RANGE OF BITS USES A COLON SEPARATOR
Figure 111. Auxiliary NCO Block Diagram
Rev. 0 | Page 166 of 205
Data Sheet
AD9546
TEMPERATURE SENSOR
TEMPERATURE
SENSOR
REG 0x2906 TO 0x2905, BITS[15:0]
16
REG 0x2902, BIT 0
16
0
1
16
IRQ MAP
TO SYSTEM CLOCK
COMPENSATION
REG 0x300C, BIT 4
THRESHOLD
DETECTOR
REG 0x3002, BIT 0
REG 0x2902, BIT 1
16
0
1
TO DPLL0 DELAY
COMPENSATION
16
REG 0x2904 TO 0x2903, BITS[15:0]
REG 0x2902, BIT 2
16
0
1
REG 0x3004 TO 0x3003, BITS[15:0]
REG 0x2901 TO 0x2900, BITS[15:0]
TO DPLL1 DELAY
COMPENSATION
16
NOTES
1. A RANGE OF BITS USES A COLON SEPARATOR
Figure 112. Temperature Sensor Block Diagram
TEMPERATURE SENSOR OVERVIEW
INTERNAL TEMPERATURE SENSOR
Figure 112 shows a block diagram of the temperature sensor
and supporting circuitry. Three AD9546 functional blocks have
the capability of using temperature values.
The AD9546 has an internal temperature sensor with a digital
output (see Figure 112). The sensor reflects the temperature of
the silicon die, which the user can read via Bits[15:0] of Register
0x3003 to Register 0x3004. The temperature sensor indicates
temperature as a 16-bit signed (twos complement) number
scaled by 2−7, as shown in Figure 113. This format provides a
range of 256ꢀC with 0.007ꢁꢀC resolution.
System clock compensation
DPLL0 delay compensation
DPLL1 delay compensation
Each of these functional blocks can receive temperature values
from two sources.
16-BIT TEMPERATURE
...
...
15
7
6
0
The internal temperature sensor (default)
Bits[15:0] of Register 0x2900 to Register 0x2901
FRACTIONAL
PART
INTEGER PART
DECIMAL
POINT
–7
2
TEMPERATURE SOURCE SELECTION
Figure 113. Temperature Format
The user selects the desired temperature source for each of the
temperature compensation functional blocks via Bits[2:0] of
Register 0x2902. Bit 0 controls the routing of the temperature
source to the system clock compensation block (see the
Compensation Method 1 section). Bit 1 and Bit 2 control the
routing of the temperature source to the delay compensation
block of DPLL0 and DPLL1, respectively, (see the Delay
Compensation section). For Bits[2:0], Logic 0 (default) selects
the internal temperature sensor, whereas Logic 1 selects the
external temperature source (see the External Temperature
Source section).
The sensor samples at approximately 25 MHz and performs a
212-sample average, providing updates at a rate of
approximately 6.1 kHz. The intent of the temperature sensor is
for tracking temperature variation rather than providing a
measure of absolute temperature.
Given internal temperature = 0x11ꢁD (4493 decimal), find the
corresponding temperature, T.
T = 4493 × 2−7ꢀC
= 35.1015625ꢀC
Given internal temperature = 0xFꢁ33 (−1997 decimal),
T = −1997 × 2−7ꢀC
= −15.6015625ꢀC
Rev. 0 | Page 167 of 205
AD9546
Data Sheet
Temperature Alarm Feature
EXTERNAL TEMPERATURE SOURCE
The temperature sensor routes to a threshold detector. The
threshold detector detects when temperature readings are
outside the range specified by a high temperature threshold and
a low temperature threshold. The high temperature threshold
resides in Bits[15:0] of Register 0x2905 to Register 0x2906. The
low temperature threshold resides in Bits[15:0] of Register
0x2903 to Register 0x2904. The high and low temperature
thresholds use the temperature format shown in Figure 113.
The external temperature source constitutes a temperature
value programmed into Bits[15:0] of Register 0x2900 to
Register 0x2901. Although the internal temperature sensor
properly scales temperature values for use by the internal
temperature compensation functional blocks automatically, the
user must scale the external temperature values properly.
The external temperature bit field represents temperature in
units of °C × 2−7 (see Figure 113). This format yields a
temperature range of 256°C with 0.0078°C resolution.
If the output value of the temperature sensor falls outside the
specified limits, status Bit 0 of Register 0x3002 assumes a Logic
1 state (otherwise, it is Logic 0). Status Bit 0 allows the user to
check the status of the threshold detector in real time.
For example, to program an external temperature value of
95°C,
External Temperature = 95/2−7
Status Bit 0 of Register 0x3002 is the source for Status Bit 4 of
Register 0x300C. Status Bit 4 allows the user to set up automatic
notification of a temperature threshold detection event via the
IRQ mechanism (see the Interrupt Request (IRQ) section).
= 12,160
= 0x2F80 (hexadecimal)
Rev. 0 | Page 168 of 205
Data Sheet
AD9546
SYSTEM CLOCK COMPENSATION
In the context of system clock compensation, consider a given
system clock frequency error expressed in terms of FFE.
Applying an FFE correction factor to the NCOs and TDCs
compensates for the FFE error of the system clock source. FFE
as a correction factor to the NCOs and TDCs compensates for
the FFE of the system clock source. The AD9546 has the option
of applying system clock compensation via two methods: open-
loop method and closed-loop method.
SYSTEM CLOCK COMPENSATION OVERVIEW
The NCOs and the TDCs of the AD9546 derive their
timekeeping from the system clock (see the System Clock PLL
section). Therefore, the frequency accuracy of any of the NCOs
relates directly to the accuracy of the system clock. Likewise, an
inferred frequency based on the difference between successive
TDC time stamps is subject to the accuracy of the system clock.
Therefore, the stability of the system clock is crucial to the
accuracy of the NCOs and TDCs within the AD9546.
Open-Loop Method
Figure 114 shows the open-loop method, with the system clock
source driving the system clock PLL, which in turn serves as the
clock source for a representative NCO within the AD9546.
Although Figure 114 shows an NCO, this section also applies to
a TDC.
The stability of the system clock, in turn, relates directly to the
stability of the system clock source. The system clock source is
the frequency source driving the XOA and XOB pins. As such,
an ideally stable system clock source is desirable. In practice,
however, the system clock source suffers from frequency
instability caused by aging, variations in temperature, and
similar physical factors. Any frequency instability introduced
by the system clock source translates to a frequency instability
in the NCOs and TDCs.
SYSTEM
CLOCK
SOURCE
fSRC
Because the NCOs and TDCs are fundamentally numeric (digital)
in nature, it is possible to tune the NCOs and TDCs
numerically to counteract the system clock instability. That is,
with a known frequency error associated with the system clock
source, the user can apply a corresponding correction
(numerically) to the NCOs and TDCs, which is the underlying
concept of system clock compensation.
XOA XOB
AD9546
COMPENSATION
SYSTEM
CALCULATOR
CLOCK PLL
fS
1 + FFE
COMP
CORRECTED
FTW
fNCO
FTW
NCO
The system clock compensation operates on fractional frequency
error (FFE) rather than absolute frequency error. For a given
nominal frequency, f0, the FFE of a deviated frequency, f, is
Figure 114. Open-Loop Method
FFE = (f – f0)/f0
or
The system clock source provides the primary frequency, fSRC,
with a nominal value of f0. The system clock PLL, which is the
clock source to the NCO, multiplies fSRC by a constant, KPLL
(that is, fS = KPLL × fSRC). Assuming the FFE for system clock
compensation (FFECOMP) = 0 in Figure 114, the NCO produces an
output frequency, fNCO, proportional to the applied numeric
FTW (that is fNCO = fS × FTW × KNCO, where KNCO is the
FFE = f/f0 − 1
In the case of TDCs, the difference between successive time
stamps is the period of the underlying frequency. The
difference between successive time stamps gives rise to the
concept of fractional period error (FPE).
proportionality constant). Therefore, express fNCO in terms of
f
SRC as
NCO = fSRC × KPLL × FTW × KNCO
The ideal (error free) fNCO is then
NCO_IDEAL = f0 × KPLL × FTW × KNCO
FPE = (p − p0)/p0
or
f
FPE = p/p0 − 1
f
where:
p0 is the nominal period.
p is the deviated period.
where f0 is the nominal frequency of the system clock source. If
the system clock source experiences a fractional error, FFE,
Because frequency relates to period as f = 1/p, FFE and FPE
relate as follows:
f
NCO = f0 × (1 + FFE) × KPLL × FTW × KNCO
Apply a fractional correction, FFECOMP, to compensate for FFE
as follows:
FFE = −FPE/(FPE + 1)
or
f
NCO = f0 × (1 + FFE) × (1 + FFECOMP) × KPLL × FTW × KNCO
If FFECOMP = –FFE,
NCO = f0 × (1 + FFE) × (1 – FFE) × KPLL × FTW × KNCO (24)
FPE = −FFE/(FFE + 1)
f
Rev. 0 | Page 169 of 205
AD9546
Data Sheet
The term, (1 + FFE) × (1 − FFE), simplifies to 1 − FFE2. Given
that FFE2 = FFECOMP × FFE, if both FFECOMP and FFE are less
than 1 ppm (10−6), a reasonable assumption, then FFE2 is less
than 10−12 (one part in a trillion). For FFE values < 1 ppm, the
value of FFE2 is exceedingly small. Under such conditions, FFE2
is practically zero, and Equation 24 simplifies to the following:
frequency of the stable frequency source is constant, by
definition, which is a necessary condition of the closed-loop
method. As such, fNCO is constant because the feedback loop of
the DPLL ensures fNCO = N × fREF. Therefore, if fSRC changes (due
to temperature, for example), the DPLL feedback loop causes
FTW to change in a manner that maintains a constant fNCO
.
Because the FFE of the stable source is zero (the source is error
free, presumably), then any FFE associated with fNCO must be
attributable to the FFE of the system clock source. That is, the
FFE of the system clock source translates deterministically to a
deviation of FTW (from its nominal value, FTWNOM).
f
NCO = f0 × KPLL × FTW × KNCO
This expression for fNCO is the same as fNCO_IDEAL. Therefore,
applying the fractional correction term, FFECOMP, nullifies the
FFE associated with the system clock source.
The preceding example demonstrates the viability of
compensating for the system clock FFE. However, FFE
compensation is of little value without a means of predicting
FFE so that the appropriate correction can be applied. In this
regard, the open-loop method relies on two assumptions.
In the closed-loop method, FTW variations lead to a correction
factor, 1 + FFECOMP. The FTW analyzer shown in Figure 115 is
an integral component of the AD9546. Applying the correction
factor to other designated NCOs and TDCs in the AD9546 is
the basis of the closed-loop method.
SYSTEM
•
Temperature variation dominates the FFE of the system
clock source
CLOCK
SOURCE
•
The FFE vs. temperature (T) behavior of the system clock
source takes the form of an x-order polynomial
fSRC
XOA XOB
In the case of the AD9546, x = 5. Therefore, the latter
assumption is expressible as
AD9546
SYSTEM
CLOCK PLL
fS
FFESRC = c5T5 + c4T4 + c3T3 + c2T2 + c1T1 + c0
(25)
DPLL
where the coefficients, cx, define the shape of the FFESRC curve.
c0 constitutes a static offset, whereas the remaining coefficients
relate to powers of T. Note also, that any particular power of T
can be artificially eliminated by assigning its corresponding
coefficient to zero. For example, reduce the polynomial to
second order by making the c5, c4, and c3 coefficients in
Equation 25 equal to zero. The open-loop method assumes
advance knowledge of the cx values, which the user programs
into the appropriate AD9546 register map locations.
STABLE
fREF
DIGITAL
LOOP
FILTER
DIGITAL
PHASE
DETECTOR
fNCO
FREQUENCY
SOURCE
NCO
÷N
FTW
ANALYZER
1 + FFE
COMP
OTHER NCO OR TDC
1 + FFE
COMP
The overall concept of the open-loop method is to program the
AD9546 with the appropriate coefficients, cx, such that the
polynomial closely models the temperature dependent behavior
of the system clock source. The AD9546 has a built in
compensation calculator to implement the FFE polynomial (see
the Compensation Method 1 section for more details). The
AD9546 provides the means to apply the correction factor
generated by the compensation calculator to the NCOs and
TDCs designated by the user.
CORRECTED
FTW
FTW
Figure 115. Closed-Loop Method
The dynamic and automatic compensation capability of the
closed-loop method typically offers improved performance
over the open-loop method but requires the availability of a
stable frequency source. In general, the stable frequency source
must be at least 10 times more stable than the system clock
source for the closed-loop method to be viable.
Closed-Loop Method
The closed-loop method, unlike the open-loop method,
requires no advance knowledge of the FFE behavior of the
system clock source. Instead, the closed-loop method relies on a
DPLL and the availability of a very stable external frequency
source (GPS, for example) as a reference input to the DPLL (see
Figure 115).
The AD9546 employs two forms of the closed-loop method.
The first form uses one of the DPLL channels (DPLL0 or
DPLL1) as the DPLL shown in Figure 115. The second form
uses an independent, special purpose DPLL (the auxiliary
DPLL) as the DPLL shown in Figure 115. The latter form allows
the user to implement closed-loop system clock compensation
without sacrificing one of the DPLL channels.
In operation, assuming the DPLL is locked and stable, the FTW
applied to the NCO is relatively constant, especially under the
assumption that the frequency of both the stable frequency
source and the system clock source are constant. However, the
Rev. 0 | Page 170 of 205
Data Sheet
AD9546
source frequency. Table 96 shows a hypothetical example of a
25 MHz source in the two leftmost columns.
COMPENSATION METHOD 1
Compensation Method 1 employs the open-loop method of
system clock compensation (see the Open-Loop Method
section). As shown in Figure 116, Compensation Method 1
takes the polynomial coefficients (Cx) and a temperature value
(T) to compute a compensation factor (CF1). Compensation
Method 1 differs slightly from the open-loop method by the
inclusion of a digital filter (see the Digital Filter section).
Because the AD9546 implements compensation in terms of
fractional error, it is necessary to convert the frequencies in
Table 96 to FFE values (refer to the System Clock Compensation
Overview section). Table 96 shows the resulting FFE values.
However, Compensation Method 1 generates CF1 as a time
error correction rather than a frequency error correction.
Therefore, the user must convert the FFE data to FPE data
(refer to the System Clock Compensation Overview section for
the conversion formula). Table 96 shows the resulting FPE
values. From the calculated FPE values, find the coefficients
(Cx) that best fit the FPE data, which generally involves
mathematical regression techniques using specialized
mathematical software tools.
COEFFICIENT
OUTPUT FILTER
CUTOFF
C
C
C
C
C
C
0
1
2
3
4 5
POLYNOMIAL
CALCULATOR
DIGITAL
FILTER
T
CF
1
Figure 116. Compensation Method Number 1
Applying such regression techniques to the hypothetical FPE
values in Table 96 yields the following coefficients:
The coefficient values, C0 to C5, are programmed into the register
map by the user (see the Programming the Coefficients (Cx) for
Compensation Method 1 section). The temperature value comes
from the internal temperature sensor or a temperature register
(see the Temperature Sensor section for details). The CF1
compensation factor applies to the designated NCOs and TDCs
to correct for frequency variations of the system clock source.
CF1 embodies the factor, 1 + FFECOMP, shown in Figure 114.
C0 = 1.0046936 × 10−3
C1 = −2.8927765 × 10−6
C2 = −1.9110192 × 10−7
C3 = 2.2493907 × 10−9
C4 = 2.7943570 × 10−11
C5 = −2.4817310 × 10−13
How to Derive Coefficients from Frequency Data
To compensate for the frequency variation of the system clock
source, knowledge of the frequency vs. temperature profile is a
necessity. Generally, the temperature profile of the source is a
table of temperature values with corresponding values of the
Table 96. Example 25 MHz Source Data
Temperature (°C)
Frequency (MHz)
24.9757265855
24.9745666538
24.9751062576
24.9758628851
24.9780535472
24.9789426612
24.9809631193
24.9810049826
24.9800392641
24.9777091418
24.9742106356
FFE (×10−6)
−970.93658
−1017.33385
−995.74970
−965.48459
−877.85811
−842.29355
−761.47523
−759.80070
−798.42944
−891.63433
−1031.57458
FPE (×10−6)
971.88021
1018.36987
996.74220
966.41766
878.62943
843.00361
762.05552
760.37843
799.06744
892.43005
1032.63982
−20
−10
0
10
20
30
40
50
60
70
80
Rev. 0 | Page 171 of 205
AD9546
Data Sheet
Figure 117 shows the FPE vs. temperature curve resulting from
these coefficients. The trace is the predicted FPE vs.
temperature curve based on the coefficients, and the solid
points are the temperature vs. FPE data from Table 96.
1400
Table 97. System Clock Compensation Coefficients
Register Address
Exponent
Coefficient
Significand
C1
C2
C3
C4
C5
0x028E to 0x028F
0x0291 to 0x0292
0x0294 to 0x0295
0x0297 to 0x0298
0x029A to 0x029B
0x0290
0x0293
0x0296
0x0299
0x029C
1300
1200
1100
1000
900
C1 through C5 apply scale factors to corresponding powers
of T. The C1 through C5 coefficients have the following format:
Cx = Sx × 2E
x
where:
800
x is the coefficient index.
Sx is the significand component.
Ex is the exponent component.
700
600
–50 –40 –30 –20 –10
0
10 20 30 40 50 60 70 80 90 100
In practice, Cx is a base 10 number, as are Sx and Ex. However,
the AD9546 requires Sx and Ex to be signed integers. The
following procedure explains how to convert base 10 Cx values
to the necessary signed integers corresponding to the
significand register value (Cx_S) and the exponent register value
(Cx_E).
TEMPERATURE (°C)
Figure 117. FPE vs. Temperature
Given a set of coefficients like C0 to C5, the user must convert
the numeric values to suitable register contents (see the
Programming the Coefficients (Cx) for Compensation Method 1
section for details).
There are three cases for Cx.
Programming the Coefficients (Cx) for Compensation
Method 1
•
•
•
The trivial case (Cx = 0)
The case where Cx is less than the quantization limit
The case where Cx is greater than the quantization limit
Compensation Method 1 makes use of six coefficients, Cx,
where the index, x, ranges from 0 to 5. Coefficient C0 is
represented by a 40-bit (signed) coefficient value (CV) residing
in Register 0x0289 to Register 0x028D.
For the first two cases (that is, Cx = 0 or Cx is less than the
quantization limit),
C
x_S = 0 (0x0000 hexadecimal)
x_E = 0 (0x00 hexadecimal)
The relationship between the value of CV and the value of C0 is
CV = C0/2−45
C
To test if Cx is less than the quantization limit, check whether
the following inequality is true:
The value of CV has no units and is proportional to the system
clock period.
For example, convert C0 = 1.0046936 × 10−3 to its
corresponding register value (C0_RegVal).
log |Cx |
log2
+1 < −127
(26)
C
0_RegVal = C0/2−45
= 1.0046936 × 10−3/2−45
where:
log(x) is the logarithm of x (for example, ln(x), log2(x), log10(x)).
|x| is the absolute value of x.
⌊x⌋ is the notation for signifying the nearest integer to x in the
direction of −∞.
= 35,349,513,305 (nearest integer)
= 0x 08 3AFE C459 (hexadecimal)
Coefficient C1 to Coefficient C5 are represented by employing
two components: a 16-bit (signed) significand and an 8-bit
(signed) exponent. The significand and exponent components
reside in the register map per Table 97.
If Cx is nonzero and the inequality in Equation 26 is false, then
log | Cx |
E =
+1
log 2
C
x_E = E
x_S = round(Cx × 215 − E
C
)
Rev. 0 | Page 172 of 205
Data Sheet
AD9546
Using C1 as an example, convert C1 = −2.8927765 × 10−6 to its
corresponding register values. First, check that C1 is greater
than the quantization limit.
This value is greater than the quantization limit of −127.
Therefore,
C
3_E = E = −28 = 0xE4 (hexadecimal)
3_S = round(C3 × 215 – E
log | C1 |
C
)
E =
+1
= round(2.2493907 × 10−9 × 215 − (−28)
)
log 2
−6
= 19,786
log | −2.8927765× 10
|
=
+ 1
log2
= 0x4D4A (hexadecimal)
Using C4 as an example, convert C4 = 2.7943570 × 10−11 to its
corresponding register values. First, check that C4 is greater
than the quantization limit.
= −18
This value is greater than the quantization limit of −127.
Therefore,
log | C4 |
C
1_E = E = −18 = 0xEE (hexadecimal)
1_S = round(C1 × 215 − E
E =
+1
log 2
C
)
−11
= round(−2.8927765 × 10−6 × 215 − (−18)
)
log | 2.7943570× 10
|
=
+ 1
log2
= −24,849
= −35
= 0x9EEF (hexadecimal)
Using C2 as an example, convert C2 = –1.9110192 × 10−7 to its
corresponding register values. First, check that C2 is greater
than the quantization limit.
This value is greater than the quantization limit of −127.
Therefore,
C
4_E = E = −35 = 0xDD (hexadecimal)
4_S = round(C4 × 215 − E
C
)
log | C2 |
E =
+1
= round(2.7943570 × 10−11 × 215 − (−35)
)
log 2
−7
= 31,462
log | −1.9110192× 10
|
=
+ 1
log2
= 0x7AE6 (hexadecimal)
Using C5 as an example, convert C5 = −2.4817310 × 10−13 to its
corresponding register values. First, check that C5 is greater
than the quantization limit.
= −22
This value is greater than the quantization limit of −127.
Therefore,
log | C5 |
log 2
C
2_E = E = −22 = 0xEA (hexadecimal)
2_S = round(C2 × 215 − E
E =
+1
C
)
−13
= round(−1.9110192 × 10−7 × 215 − (−22)
)
log | −2.4817310 ×10
|
=
+1
log 2
= −26,265
= −41
= 0x9967 (hexadecimal)
Using C3 as an example, convert C3 = 2.2493907 × 10−9 to its
corresponding register values. First, check that C3 is greater
than the quantization limit.
This value is greater than the quantization limit of −127.
Therefore,
C
5_E = E = −41 = 0xD7 (hexadecimal)
5_S = round(C5 × 215 − E
C
)
log | C3 |
E =
+1
= round(−2.4817310 × 10−13 × 215 − (−41)
)
log2
−9
= −17,883
log | 2.2493907×10
|
=
+1
log2
= 0xBA25 (hexadecimal)
= −28
Rev. 0 | Page 173 of 205
AD9546
Data Sheet
Use the following formulas to verify the register value
computations, Cxʹ, closely match the desired coefficient value:
The AD9546 automatically bypasses the digital filter under the
following conditions:
C0ʹ = C0_RegVal × 2−45
(27)
•
•
The system clock is unstable.
The difference between the input and output of the digital
filter exceeds ~125 ppm.
−15
Cx
= Cx S ×2C
x _ E
′
(28)
|
x≠
0
_
Note that there may be a slight difference between the values
produced by Equation 27 and Equation 28 and the original
coefficient values (Cx) due to quantization effects.
None of these conditions occurs during normal operation.
Thus, the filter is nearly always present.
COMPENSATION METHOD 2
An example of coefficient verification follows:
C0ʹ = 35,349,513,305 × 2−45
Compensation Method 2 employs the closed-loop method of
system clock compensation (see the Closed-Loop Method section).
As shown in Figure 118, Compensation Method 2 makes use of
one of the DPLLx channels (where x is 0 or 1). The user must
specify which DPLL to use via Bit 0 of Register 0x0287. Logic 0
(default) selects DPLL0, and Logic 1 selects DPLL1 as the source of
CF2. CF2 embodies the factor, 1 + FFECOMP, in Figure 115.
= 1.004694 × 10−3 (compare: C0 = 1.0046936 × 10−3)
C1ʹ = −24,849 × 2−18 − 15
= −2.892804 × 10−6 (compare: C1 = −2.8927765 × 10−6)
C2ʹ = −26,265 × 2−22 − 15
= −1.911030 × 10−14 (compare: C2 = −1.9110192 × 10−7)
C3ʹ = 19,786 × 2−28 − 15
= 2.249408 × 10−9 (compare: C3 = 2.2493907 × 10−9)
C4ʹ = 31,462 × 2−35 − 15
Compensation Method 2 generates the correction factor, CF2,
which the user can apply to other NCOs and TDCs within the
AD9546. However, a compensation factor, COMPx, can be applied
to the NCO of DPLLx in Figure 118. Because it is possible for the
COMPx factor to include CF2 (see the Integrated Compensation
Subsystem section for details), the user must ensure that the NCO
of the DPLL channel selected for Compensation Method 2 is not
itself a recipient of CF2 via COMPx.
= 2.794387 × 10−11 (compare: C4 = 2.7943570 × 10−11
)
C5ʹ = −17,883 × 2−41 − 15
= −2.481765 × 10−13 (compare: C5 = −2.4817310 × 10−13
)
SYSTEM
CLOCK
SOURCE
Digital Filter
The CF1 values produced by the polynomial calculator block in
Figure 116 depend on temperature measurements (internal or
external) as an input parameter. Any noise associated with
those measurements is a potential noise source on CF1, which
can lead to a degradation of phase noise performance. To
mitigate potential noise injection by the compensation
mechanism, Compensation Method 1 employs a filter that
applies a smoothing function to the raw CF1 values.
fSRC
XOA
XOB
AD9546
SYSTEM
CLOCK PLL
fS
STABLE
FREQUENCY
SOURCE
DPLLx
COMP
x
REFERENCE
DIVIDER
AND TDC
fREF
DIGITAL
FTW
fNCO
PFD AND
LOOP
NCO
FILTER
The user controls the filter bandwidth via Bits[2:0] of
Register 0x0288 according to Table 98. The filter comprises two
collocated poles that yield a 3 dB bandwidth per Table 98.
FROM
FEEDBACK
TDC
FTW
ANALYZER
CF
2
Table 98. Digital Filter Bandwidth
Bits[2:0] (Binary)
3 dB Bandwidth (Hz)
000
001
010
011
100
101
110
111
156
78
39
20
10
5
Figure 118. Compensation Method Number 2
2
1
Rev. 0 | Page 174 of 205
Data Sheet
AD9546
Because the system clock variations are the dominant
COMPENSATION METHOD 3
contributor to CF3, CF3 reflects the corrections necessary to
compensate for the temperature induced stability errors on the
system clock.
Compensation Method 3 is unavailable in digitized clocking
applications (see the Common Clock DPLL section).
Compensation Method 3 employs the closed-loop method of
system clock compensation (see the Closed-Loop Method
section). As shown in Figure 119, Compensation Method 3
produces correction factor CF3, where CF3 embodies the factor,
1 + FFECOMP, in Figure 115. Compensation Method 3 and
Compensation Method 2 are similar, but Compensation
Method 3 uses the auxiliary DPLL rather than one of the DPLL
channels. Because Compensation Method 3 relies on the
availability of a stable external clock source applied to one of
the REFx inputs or one of the auxiliary REFx inputs (via an Mx
pin), the user must assign the TDC associated with the external
clock source to the auxiliary DPLL via Bits[4:0] of
Accuracy error is the difference between the defined nominal
frequency and the actual frequency. The defined nominal
frequency derives from programmed reference period stored in
the register associated with the selected reference. The actual
frequency is fREF, the true frequency of the stable reference (see
Figure 119). The accuracy error is the difference between the
programmed reference period and the actual nominal period of
the stable reference (1/fREF).
Because CF3 is a composite of both the stability and accuracy
error sources, the programmed reference period directly affects
the accuracy error component of CF3. Thus, accurate entry of
the stable reference clock period is crucial, because the accuracy
error propagates to any COMPx involving CF3 (see the
Integrated Compensation Subsystem section).
Register 0x0284. Table 99 shows the assignment codes for the
various TDCs.
Table 99. Auxiliary DPLL TDC Assignment Codes
Auxiliary DPLL Loop Bandwidth
Bits[4:0] (Decimal)
TDC Description
REFA TDC (default)
REFAA TDC
The user controls the loop bandwidth of the auxiliary DPLL
servo loop (BWCOMP) via Bits[15:0] (unsigned) in
0
1
Register 0x0285 to Register 0x0286. Note that BWCOMP is 0.1 Hz
scaled by the 16-bit decimal value of Bits[15:0].
2
3
REFB TDC
REFB TDC
4 to 5
6
7
8 to 10
11
12
Unused
For example, given a desired value of BWCOMP = 247.63 Hz,
determine the appropriate value of Bits[15:0] as follows:
Auxiliary REF0 TDC
Auxiliary REF1 TDC
Unused
Auxiliary REF2 TDC
Auxiliary REF3 TDC
Unused
Bits[15:0] = BWCOMP/0.1
= 247.63/0.1 Hz
= 2476 (nearest integer)
= 0x 09AC (hexadecimal)
13 to 15
The auxiliary DPLL can itself receive compensation from other
sources via COMPx in Figure 119 (see the Integrated
Compensation Subsystem section for details). Unlike
Compensation Method 2, the CF3 compensation factor does
not include the contribution of COMPx. That is, CF3 comprises
only the residual error to represent the frequency deviation
associated with the system clock source only.
The programmed bandwidth affects how long it takes the
auxiliary DPLL to lock to the stable reference: a lower
bandwidth means it takes the auxiliary DPLL longer to lock.
The programmed bandwidth also affects the ability of the
auxiliary DPLL to track variations in the frequency (fS) of the
system clock. Specifically, a relatively stable, slowly varying
system clock frequency allows the use of a low loop bandwidth,
whereas a system clock exhibiting more rapid frequency
variations requires a wider loop bandwidth. For example, using
an OCXO as the system clock reference allows the use of a
0.1 Hz loop bandwidth, whereas a standard crystal oscillator
necessitates a loop bandwidth in excess of 50 Hz.
Assuming the auxiliary DPLL is locked and has fully settled, the
period of the stable reference must exactly match the
programmed reference period stored in the register associated
with the selected reference (see Figure 119). Any deviation is an
indication of error associated with the system clock period and
leads to a corresponding CF3 value. There are two error
sources: stability error and accuracy error.
Auxiliary DPLL Reference Monitor Status
The auxiliary DPLL has a dedicated reference monitor to
provide indication that the reference period is within limits.
The reference monitor relies on the defined reference period
(user input via the appropriate registers).
Stability error involves deviation from the mean value over
time (the mean corresponding to a specific operating condition
like 25°C, for example). As temperature varies over time, the
frequency of the oscillator serving as the source for the system
clock varies in response to the temperature changes. Given the
prerequisite that the stable reference is significantly more stable
than the system clock source, the dominant variations
The auxiliary DPLL reference monitor status is available in real
time via Bit 2 (auxiliary DPLL reference status) of
Register 0x3002. The auxiliary DPLL reference fault and unfault
states are Logic 1 and Logic 0, respectively.
appearing in the feedback path are those of the system clock.
Rev. 0 | Page 175 of 205
AD9546
Data Sheet
The auxiliary DPLL reference monitor status is also available
via the IRQ mechanism (see the Interrupt Request (IRQ)
section) by means of Bit 2 and Bit 3 of Register 0x300C. Bit 2
and Bit 3 provide latched information regarding state
transitions of the auxiliary DPLL reference status bit. Bit 2
latches to Logic 1 when the auxiliary DPLL reference status bit
transitions from Logic 0 to Logic 1, whereas Bit 3 latches to
Logic 1 when the auxiliary DPLL reference status bit transitions
from Logic 1 to Logic 0.
Because Bit 2 and Bit 3 represent the latched state transitions of
the auxiliary DPLL reference status bit, they may represent a
condition that is no longer true. Therefore, the user must clear
Bit 2 and Bit 3 via Bit 2 and Bit 3, respectively, of Register 0x2007.
Otherwise, the user may lose indication of subsequent state
changes of the auxiliary DPLL reference status bit.
SYSTEM
CLOCK
SOURCE
fSRC
XOA
XOB
AD9546
SYSTEM
CLOCK PLL
fS
AUXILIARY DPLL
SYSTEM CLOCK
COMPENSATION
SUBSYSTEM
DIGITAL
CROSS
POINT
MUX
SYSCLK REF FREQ
TIME SCALE
GENERATOR
REG 0x0206 TO 0x0202, BITS[39:0]
INTERFACE
TIME
SCALE
COMP
X
TIME
PERIOD
ERROR
REFERENCE
DIVIDER
AND TDC
DIGITAL
LOOP
fREF
STAMPS
TIME STAMP
CF
3
PROCESSOR
FILTER
REFx
OR Mx
STABLE
FREQUENCY
SOURCE
LOCK
DETECTOR
REFERENCE
MONITOR
REG 0x0286 TO 0x0285, BITS[15:0]
REG 0x040B TO 0x0404, BITS[59:0]
REG 0x042B TO 0x0424, BITS[59:0]
REG 0x044B TO 0x0444, BITS[59:0]
REG 0x046B TO 0x0464, BITS[59:0]
REG 0x048B TO 0x0484, BITS[59:0]
REG 0x04AB TO 0x04A4, BITS[59:0]
REG 0x04CB TO 0x04C4, BITS[59:0]
REG 0x04EB TO 0x04E4, BITS[59:0]
REFERENCE
PROGRAMMED
REFERENCE
PERIOD
SOURCE
SELECT
5
REG 0x0284, BITS[4:0]
NOTE
1. A RANGE OF BITS USES A COLON SEPARATOR
Figure 119. Compensation Method 3
Rev. 0 | Page 176 of 205
Data Sheet
AD9546
such, the auxiliary DPLL holds the CF3 value that existed just
Auxiliary DPLL Lock Detector
prior to entering the holdover state. Therefore, it is not possible
to ascertain further stability information until the loop resumes
normal operation. When the condition that forced holdover no
longer exists, however, the auxiliary DPLL returns to closed-
loop operation and resumes updating CF3 (that is, when the
lock detector indicates a lock condition).
The auxiliary DPLL has a built in lock detector for indicating
lock status. The lock and unlock thresholds for the lock
detector are fixed and not user programmable.
The auxiliary DPLL lock detector status is available in real time
via status Bit 1 (auxiliary DPLL lock status) of Register 0x3002.
The auxiliary DPLL lock and unlock states are Logic 1 and
Logic 0, respectively.
INTEGRATED COMPENSATION SUBSYSTEM
The three compensation methods comprise the integrated
compensation subsystem of the AD9546. The compensation
subsystem allows the user to employ any combination of the
three compensation methods to any of the NCOs and TDCs
within the AD9546 (barring a few exceptions per the
Compensation Assignment Guidelines section). Figure 120
shows a simplified block diagram of the compensation
subsystem.
The auxiliary DPLL lock detector status is also available via the
IRQ mechanism (see the Interrupt Request (IRQ) section) by
means of Bit 0 and Bit 1 of Register 0x300C. Bit 0 and Bit 1
provide latched information regarding state transitions of the
auxiliary DPLL lock status bit. Bit 0 latches to Logic 1 when the
auxiliary DPLL lock status bit transitions from Logic 0 to Logic
1, whereas Bit 1 latches to Logic 1 when the auxiliary DPLL lock
status bit transitions from Logic 1 to Logic 0.
In Figure 120, note the implied feedback path between
DPLL0/DPLL1 and Compensation Method 2 and between the
auxiliary DPLL and Compensation Method 3. The reason for
the implied feedback is that the indicated compensation
destination (the auxiliary DPLL, for example) can receive
compensation from any of the compensation sources via the
combiner and distributor. Thus, the compensation destination
can potentially apply self compensation, which leads to an
undesired positive feedback loop. The user must avoid such
loops (see the Compensation Assignment Guidelines section).
Because Bit 0 and Bit 1 represent the latched state transitions of
the auxiliary DPLL lock status bit, they may represent a
condition that is no longer true. Therefore, the user must clear
Bit 0 and Bit 1 via Bit 0 and Bit 1, respectively, of Register
0x2007. Otherwise, the user may lose indication of subsequent
state changes of the auxiliary DPLL lock status bit.
Auxiliary DPLL Holdover
Under normal conditions, the auxiliary DPLL continuously
updates CF3 computations with each rising edge event to the
TDC associated with the stable reference. Under certain
conditions, however, the auxiliary DPLL can enter a holdover
state. For example,
For digitized clocking applications (see the Digitized Clocking
section), the common clock DPLL replaces the auxiliary DPLL.
Thus, in a digitized clocking application, CF3 originates from
the common clock DPLL. The implication is that, when
applying system clock compensation in a digitized clocking
application, the burden of stability falls on the common clock
reference to the common clock DPLL.
The user programs auxiliary DPLL bandwidth = 0
The reference monitor indicates a fault
While in the holdover state, the auxiliary DPLL is effectively in
an open-loop condition (that is, the NCO output is static). As
COMPENSATION
METHODS
COMPENSATION
DESTINATIONS
DISTRIBUTOR
COMBINER
CF
CF
CF
1
2
3
COMPENSATION
METHOD 1
REFx TDCS AND
AUXILIARY TDCS
COMP
0
1
2
3
4
5
6
7
COMP
COMP
COMP
COMP
COMP
COMP
COMP
AUXILIARY NCO 0/NCO 1
DPLL0/DPLL1
SLEW
RATE
LIMITER
COMPENSATION
METHOD 2
COMPENSATION
METHOD 3
AUXILIARY DPLL
Figure 120. Integrated Compensation Subsystem
Rev. 0 | Page 177 of 205
AD9546
Data Sheet
Each compensation method produces its own compensation
factors (CF1, CF2, and CF3). The combiner translates the three
CFx factors to eight possible combined compensation factors,
COMPx, according to Table 100. The index, x, of COMPx is the
3-bit value programmed in the associated compensation
destination bit field per Figure 121.
Table 101. Compensation Rate Limiting
Bits[2:0] (Binary)
Rate Limit (ppm/sec)
000
001
010
011
100
101
110
111
None
0.715
1.430
2.860
5.720
11.44
22.88
45.76
Table 100. Composite Compensation Factor (COMPx)
Compensation Factor
COMPx Index (x) Index
Combiner
Output
0
1
2
3
4
5
6
7
Not applicable
1
2
1, 2
3
1, 3
2, 3
1, 2, 3
0
CF1
CF2
CF1 × CF2
CF3
CF1 × CF3
CF2 × CF3
CF1 × CF2 × CF3
Compensation Assignment Guidelines
When using a UTSP (see the User Time Stamp Processor
(UTSP) section) with an auxiliary NCO selected as the time
scale, both the auxiliary NCO and the UTSP time stamp source
must use the same COMPx factor. That is, the user must
program identical values in the bit field of the REFx or auxiliary
REFx TDC compensation destination and in the bit field of the
auxiliary NCO 0 or auxiliary NCO 1 compensation destination
(whichever is the selected time scale for the UTSP). See Figure 121
for the compensation destination bit fields.
The distributor gives the user the ability to connect any COMPx
output from the combiner to any compensation destination.
The AD9546 implements the distributor through six 3-bit bit
fields, where each bit field corresponds to a specific
compensation destination per Figure 121. The user assigns a
COMPx factor to a specific destination by writing a 3-bit value
to the corresponding compensation destination bit field. See
the System Clock Compensation Programming Registers
section for a register programming example.
To avoid a positive feedback loop in the integrated
compensation subsystem, do not assign COMPx sources
involving CF2 in Table 100 to the following compensation
destinations:
•
DPLL0 NCO compensation destination when Bit 0 of
Register 0x0287 is Logic 0
Any of the COMPx selections (except for COMP0) constitute a
real-time sequence of compensation values originating from the
various compensation sources. This real-time sequence
continuously applies to the associated compensation destination.
Because the COMPX outputs respond to the various stimuli
associated with the compensation sources, COMPx tends to
vary over time, causing the associated compensation
destination to vary accordingly. However, if the frequency
variation vs. time slope is too large, the slope can appear as an
unwanted noise source to compensation destinations employing
closed-loop compensation (such as DPLL0, DPLL1, and the
auxiliary DPLL).
•
DPLL1 NCO compensation destination when Bit 0 of
Register 0x0287 is Logic 1
Likewise, do not assign COMPx sources involving CF3 in Table 100
to the auxiliary DPLL (the reason Bit 6 of Register 0x280 is
shaded out in Figure 121).
SYSTEM CLOCK COMPENSATION PROGRAMMING
REGISTERS
System clock compensation comprises three compensation
methods and six compensation destinations. Three 8-bit
registers constitute the function of the distributor in Figure 120.
The three registers function as a connection matrix with six
3-bit bit fields distributed over the three registers (see Figure 121).
To mitigate the potential noise injection caused by COMPx
exhibiting an excessive frequency variation vs. time slope, the
combiner employs programmable rate limiting via a slew rate
limiter. The slew rate limiter effectively prevents the rate of
change of frequency originating from compensation sources
from exceeding a preset limit.
Each 3-bit bit field associates with one of the following six
compensation destinations:
•
•
•
•
•
•
Auxiliary DPLL
REFx TDCs and auxiliary REFx TDCs
Auxiliary NCO 0
Auxiliary NCO 1
DPLL0
The user controls rate limiting via Bits[2:0] of Register 0x0283
according to Table 101. The rate limit values are in units of
parts per million per second (ppm/sec or 10−6/sec) and
represent the maximum rate of change of COMPx that occurs
given the value of Bits[2:0]. Because rate limiting applies
globally to COMP1 through COMP7, the user must base the rate
limit selection on the stability requirements of the most
frequency sensitive compensation destination.
DPLL1
Rev. 0 | Page 178 of 205
Data Sheet
AD9546
The value of the 3-bit bit field for a given destination establishes
the COMPx factor applied to that destination (see the
Integrated Compensation Subsystem section for details).
For example, to provide system clock compensation to
Auxiliary NCO 0 using a combination of Compensation
Method 1 and Compensation Method 3, follow these steps:
The compensation methods relate to columns in the bit field
matrix of Figure 121, whereas the compensation destinations
relate to rows. Organizing the compensation bit field matrix in
this way enables any compensation destination to receive any
one of the eight possible COMPx signals (see Figure 120).
Certain compensation destinations have constraints on which
compensation methods apply (see the Compensation
Assignment Guidelines section).
1. Determine the appropriate compensation destination bit
field per Figure 121. In this case, Bits[2:0] of
Register 0x0281 apply because the compensation
destination is auxiliary NCO 0.
2. Determine the 3-bit code to write to Bits[2:0] of
Register 0x0281 per the leftmost column of Table 100
corresponding to the appropriate COMPx. In this case, the
code is 5 (101 binary) because it uses CF1 and CF3 per the
middle column of Table 100. In conclusion, by programming
Bits[2:0] = 5 in Register 0x0281, auxiliary NCO 0 receives
both Compensation Method 1 and Compensation Method 3.
DATA
BITS
D7
D6
D5
D4
D3
D2
D1
D0
AUXILIARY DPLL NCO
COMPENSATION
DESTINATION
REFx AND AUXILIARY REFx TDC
COMPENSATION DESTINATION
NOT USED
NOT USED
0x0280
AUXILIARY NCO 1
AUXILIARY NCO 0
NOT USED
NOT USED
NOT USED
NOT USED
0x0281
0x0282
COMPENSATION DESTINATION
COMPENSATION DESTINATION
DPLL1 NCO
COMPENSATION DESTINATION
DPLL0 NCO
COMPENSATION DESTINATION
CF
CF
CF
CF
CF
CF
1
3
2
1
3
2
Figure 121. System Clock Compensation Registers
Rev. 0 | Page 179 of 205
AD9546
Data Sheet
STATUS AND CONTROL PINS
When switching the operation of an Mx pin from an output to
an input, the recommendation is as follows. First, program the
Mx pin input function to no operation and assert the IO update
bit. This configuration avoids input/output contention on the
Mx pin or other undesired behavior because, prior to the
assertion of the IO update bit, the device continues to drive the
Mx pin. Following the assertion of the IO update bit, the device
releases the Mx pin but ignores the logic level on the pin due to
the programmed no operation function. The recommendation is
to avoid using a high impedance source on an Mx pin configured
as an input because doing so may cause excessive internal
current consumption. Second, drive the Mx pin with Logic 0 or
Logic 1 via the desired external source and program the
associated Mx pin register from no operation to the desired
function.
STATUS AND CONTROL PINS OVERVIEW
The AD9546 features seven independently configurable digital
CMOS status or control pins (M0 to CSB/M6, hereafter
referred to as Mx). Configuring an Mx pin as a status pin causes
that pin to be an output. Conversely, configuring an Mx pin as a
control pin causes that pin to be an input. During power-up or
reset, the Mx pins temporarily become inputs and only allow the
device to autoconfigure. Figure 122 is a block diagram of the Mx
pin functionality.
Mx PIN FUNCTION LOGIC
DEVICE
RESET
POWER-UP
AUTOCONFIGURATION
I/O
CONTROL
MULTIFUNCTION PINS AT RESET OR POWER-UP
CONTROL
CONTROL
FUNCTION
SELECT
At power-up or in response to a reset operation, the Mx pins
enter a special operating mode. For a brief interval following a
power-up or reset operation, the Mx pins function only as
inputs (the internal drivers enter a high impedance state during
a power-up/reset operation). During this brief interval, the
device latches the logic levels at the Mx inputs and uses this
information to autoconfigure the device accordingly. The Mx
pins remain high-Z until either an EEPROM operation occurs,
in which case M1 or M2 become an I2C master, or the user (or
EEPROM) programs them to be outputs.
DESTINATIONS
Mx PIN
CONTROL
REGISTERS
I/O UPDATE
STATUS
SOURCE
SELECT
STATUS
SOURCES
Figure 122. Mx Pin Logic
The Mx pin control logic uses special register write detection
logic to prevent the Mx pins from behaving unpredictably when
the Mx pin function changes, especially when changing mode
from input to output or vice versa.
If the user does not connect external pull-up or pull-down
resistors to the Mx pins, the M3 and M4 pins have internal pull-
down resistors to ensure a predictable start-up configuration.
In the absence of external resistors, the internal pull-down
resistors ensure that the device starts up with the serial port in
SPI mode and without automatically loading data from an
external EEPROM (see Table 102). Although the M0, M1, M2,
SDO/M5, and CSB/M6 pins are high impedance at startup,
connect external 100 kΩ pull-down or pull-up resistors to these
pins to ensure a deterministic start-up condition.
When an Mx pin functions as an output, it continues operating
according to the prior function, even after the user programs
the corresponding registers. However, assertion of an IO update
causes the corresponding pins to switch to the new function
according to the newly programmed register contents.
Changing from one output function to another output function
on an Mx pin does not require special timing to avoid
input/output contention on the pin.
Table 102 shows the Mx pin start-up conditions. M0, M1, and
M2 do not appear in Table 102 because these pins have no
explicit function during a power-up or reset operation.
When an Mx pin functions as an input, programming a specific
Mx pin function register causes all the Mx pin control functions
to latch their values. Assertion of an IO update switches to the
newly programmed pin function, at which time normal behavior
resumes. When switching from one input function to another
input function on the same pin, the logic state driven at the
input to the pin can change freely during the interval between
writing the new function to the corresponding register and
asserting an IO update.
Table 102. Mx Pin Function at Startup or Reset
Start-Up/Reset
Function
Mx Pin
Logic 1
Logic 0
M3
EEPROM load
function
Load from
EEPROM
Do not load from
EEPROM
(default)
M4
Serial port
function
I²C mode
SPI mode
(default)
When switching the operation of an Mx pin from an input to
an output function, the recommendation is that the external
drive source become high impedance during the interval
between writing the new function and asserting the IO update
bit.
SDO/M5 I2C address
offset
CSB/M6 I2C address
offset
See Table 103 See Table 103
See Table 103 See Table 103
Rev. 0 | Page 180 of 205
Data Sheet
AD9546
When the start-up conditions select the serial port to be I2C
mode (that is, M4 is Logic 1 at startup), the SDO/M5 and
CSB/M6 pins determine the I2C port device address offset, per
Table 103. The logic levels in Table 103 only apply during a
power-up or reset operation.
A 2-bit modifier augments the status or control functionality of
an Mx pin. Each Mx pin has a dedicated modifier bit pair per
Table 106. The modifier bits affect the associated Mx pin
differently depending on the Mx pin operating mode: status or
control. The specific meaning of the 2-bit code appears in
Table 107 for status functions and in Table 108 for control
functions.
Table 103. I²C Device Address Offset
CSB/M6
SDO/M5
M4
Address Offset
Not applicable
1001000 (0x48)
1001001 (0x49)
1001010 (0x4A)
1001011 (0x4B)
Table 106. Modifier Bits of Mx Pin Functionality
X1
X1
0
1
1
1
Register Address
[D7:D6] [D5:D4] [D3:D2] [D1:D0]
0
0
1
1
0
1
0
1
0x0100
0x0101
M3
Unused
M2
M6
M1
M5
M0
M4
1
STATUS FUNCTIONALITY
1 X means don’t care.
Configuring an Mx pin for status functionality gives the user
access to specific internal device status and IRQ functions in
the form of a hardware pin that produces a logic signal.
DEFINING AN Mx PIN FOR STATUS OR CONTROL
To define an Mx pin as a status or control pin, use Bit 7 in the
register address associated with the Mx pin per Table 104.
Logic 0 (default) defines the associated Mx pin as a control pin,
whereas Logic 1 defines the associated Mx pin as a status pin.
The drive strength of an Mx pin when configured for status
functionality is programmable via Bits[6:0] of Register 0x0109,
where Bits[6:0] correspond to M6 to M0, respectively). Logic 0
(default) selects normal drive strength (~6 mA) and Logic 1
selects weak drive strength (~3 mA).
Table 104. Mx Pin Mode Register Addresses
Mx Pin
Register Address
When configured as a status pin, the modifier bits affect the status
functionality per Table 107. The PMOS open-drain mode requires
an external pull-down resistor. The NMOS open-drain mode
requires an external pull-up resistor. The open-drain modes enable
the implementation of wire-OR’ed functionality of multiple Mx
status pins (including Mx status pins across multiple AD9546
devices or other compatible devices—for example, to implement
an IRQ bus).
M0
M1
M2
M3
M4
SDO/M5
CSB/M6
0x0102
0x0103
0x0104
0x0105
0x0106
0x0107
0x0108
Table 107. Mx Status Pin 2-Bit Modifier Codes from Table 106
The status or control function for an Mx pin depends on an
8-bit code stored in the Mx pin function register addresses
associated with a specific Mx pin per Table 105.
Code Mode
Description
00
01
10
11
CMOS,
Default. Output is Logic 0 for a deasserted
active high internal function and Logic 1 for an
asserted internal function.
CMOS,
Because the Mx pin mode and code reside in different registers,
the following programming procedure is necessary. First, write
the desired mode via Bit 7 of the appropriate mode register (see
Table 104) with Bits[6:0] = 0. Then, write the desired code to
the corresponding code register (see Table 105). Finally, assert
an IO update. When programming more than one Mx pin, it is
acceptable to wait until all the desired Mx pin mode and code
registers are programmed before asserting an IO update.
Output is Logic 1 for deasserted internal
function and Logic 0 for an asserted
internal function.
active low
PMOS,
Output is high impedance for a deasserted
open drain internal function and active high for an
asserted internal function.
NMOS,
Output is high impedance for a deasserted
open drain internal function and active low for an
Table 105. Mx Pin Status or Control Code Register Addresses
asserted internal function.
Mx Pin
Register Address
M0
0x0182
M1
0x0183
M2
0x0184
M3
0x0185
M4
0x0186
SDO/M5
CSB/M6
0x0187
0x0188
Rev. 0 | Page 181 of 205
AD9546
Data Sheet
Consider a case where the M0, M2, M3, and CSB/M6 pins are
CONTROL FUNCTIONALITY
configured for control functionality and are assigned to the IO
update control function. That is, Bits[7:0] = 0x01 in
Register 0x0182, Register 0x0184, Register 0x0185, and
Register 0x0188.
Configuring an Mx pin for control functionality gives the user
control of specific internal device functions via an external
hardware logic signal. When configured as a control pin, the
modifier bits affect the control functionality per Table 108.
In addition, M0 is assigned for AND operation (00), M2 for NOT
OR (11) operation, M3 for NOT AND (01) operation, and M6 for
OR (10) operation (that is, the 2-bit codes according to Table 108
in the appropriate bit positions of Register 0x0100 and
Register 0x0101 per Table 106).
Table 108. Mx Control Pin 2-Bit Modifier Codes
Code Boolean Description
00
AND
Logical AND the associated Mx control pin
with the other Mx control pins assigned to
the same control function.
01
NOT AND Invert the logical state of the associated Mx
control pin and AND it with the other Mx
control pins assigned to the same control
function.
With these settings, the IO update function behaves according to
the following Boolean expression:
((NOT M2) OR M6) AND M0 AND (NOT M3)
10
11
OR
Logical OR the associated Mx control pin with
the other Mx control pins assigned to the
same control function.
Invert the logical state of the associated Mx
control pin and OR it with the other Mx control
pins assigned to the same control function.
In this case, an IO update occurs when M0 is Logic 1 and M3 is
Logic 0, and either M2 is Logic 0 or M6 is Logic 1.
When an Mx control pin acts on a control function individually
(rather than as part of a group, per the previous example), the
Boolean functionality of the codes in Table 108 reduces to two
possibilities. Namely, Code 00 and Code 10 specify a Boolean
true (the Mx pin logic state applies to the corresponding
control function directly), whereas Code 01 and Code 11 specify a
Boolean false (the Mx pin logic state applies to the corresponding
control function with a logical inversion).
NOT OR
The Boolean functionality of aggregated Mx control pins follows a
hierarchy whereby logical OR operations occur before logical AND
operations. The OR and NOT OR operations are collectively
grouped into a single result. A logical AND is then performed
using that result and the remaining AND and NOT AND
operations.
Rev. 0 | Page 182 of 205
Data Sheet
AD9546
INTERRUPT REQUEST (IRQ)
populates the IRQ mask with a Logic 1 to unmask the desired
IRQ events. Writing a Logic 1 to an IRQ mask bit may result in
immediate indication of an IRQ if the corresponding IRQ event
is already asserted (that is, the device previously registered the
corresponding IRQ event).
IRQ OVERVIEW
The AD9546 monitors certain internal device events, potentially
allowing them to trigger an IRQ event. Three groups of
registers (see Figure 123) control the IRQ functionality within the
AD9546.
IRQ CLEAR BITS
•
•
•
IRQ status registers (Register 0x300B through
Register 0x301E)
IRQ mask registers (Register 0x010C through
Register 0x011F)
The IRQ clear bits comprise a bit for bit correspondence with
the IRQ status bits. Writing a Logic 1 to an IRQ clear bit forces
the corresponding IRQ event to Logic 0, thereby clearing that
specific IRQ event. The IRQ clear registers are autoclearing.
Therefore, after writing a Logic 1 to any IRQ clear bit in
Register 0x2006 to Register 0x2019, the device automatically
restores the IRQ clear bit to Logic 0. The IRQ remains asserted
until the user clears all of the IRQ clear bits associated with
PLL0 IRQ, PLL1 IRQ, common IRQ, or any IRQ signal shown
in Figure 123).
IRQ clear registers (Register 0x2006 through Register 0x2019)
The IRQ logic can indicate an IRQ event status result for any
specific device events via the logical OR of the status of all the
IRQ status bits. In addition, the IRQ logic offers IRQ event
status results for select groups of specific IRQ events, namely,
the PLL0 IRQs, PLL1 IRQs, and common IRQs (see Figure 123).
The PLL0 IRQ group includes all device events associated with
DPLL0 and APLL0. The PLL1 IRQ group includes all device
events associated with DPLL1 and APLL1. The common IRQ
group includes events not associated with PLL0 or PLL1 (like
the system clock, EEPROM, reference monitors, and others).
Although it is not recommended, in certain applications, it
may be desirable to clear an entire IRQ group all at one time.
Register 0x2005 provides four bits for clearing IRQ groups. Bit
0 clears all IRQ clear bits. Bit 1 clears the IRQ clear bits
associated with the common IRQ bits. Bit 2 clears the IRQ clear
bits associated with the PLL0 IRQ bits. Bit 3 clears the IRQ
clear bits associated with the PLL1 IRQ bits.
IRQ STATUS BITS
The IRQ status bits maintain a record of specific IRQ events.
The occurrence of a specific device event results in the setting
and latching of the corresponding bit in the IRQ status registers
(assuming the user enables those specific IRQ events (see the
IRQ Mask Bits section)).
Alternatively, the user can program any of the multifunction
pins as an input for clearing an IRQ group, which allows
clearing an IRQ group with an external logic signal rather than
by writing to Register 0x2005 (see Figure 123).
The recommendation for clearing IRQ events is to first service
the specific IRQ event (as needed) and then clear the specific
IRQ for the given IRQ event. Clearing IRQ groups via
Register 0x2005 or via an Mx pin requires great care. Clearing
an IRQ group all at one time may result in the unintentional
clearing of one or more asserted IRQ events. Clearing asserted
IRQ events eliminates the record of the associated device
events, subsequently erasing any history of those events having
occurred.
IRQ MASK BITS
The IRQ mask bits comprise a bit for bit correspondence with
the IRQ status. Writing a Logic 1 to a mask bit allows
(unmasks) the corresponding specific device event to propagate
to the IRQ status bits. A Logic 0 (default) prevents (masks) the
corresponding specific device event to propagate to the IRQ
status bits.
The presence of the IRQ mask allows the user to select certain
device events for generating an IRQ event, while ignoring
(masking) all other specific device events from contributing to
an IRQ (PLL0 IRQ, PLL1 IRQ, common IRQ, or any IRQ signal in
Figure 123).
The default state of the IRQ mask bits is Logic 0. Therefore, the
device is not capable of generating an IRQ until the user
Rev. 0 | Page 183 of 205
AD9546
Data Sheet
IRQ MASK BITS
REG 0x010C TO 0x011F
0: DISABLE (DEFAULT), 1: ENABLE
IRQ MASK
IRQ
STATUS
BITS
MASKED
IRQ EVENTS
DEVICE
EVENT
SOURCES
LATCHED
IRQ
EVENTS
IRQ
EVENTS
REG 0x300B
TO 0x301E
MASKED IRQ EVENTS
GROUPED IRQ CLEAR
GROUPED IRQ EVENTS
IRQ CLEAR BITS
REG 0x2006 TO 0x2019
REG 0x2005, BIT 0
REG 0x2005, BIT 2
REG 0x2005, BIT 3
REG 0x2005, BIT 1
CLEAR CLEAR CLEAR
CLEAR
ALL
IRQS
COMMON PLL0
PLL1
GROUP GROUP
IRQ IRQ
PLL0
IRQS
PLL1
IRQS
COMMON
IRQS
GROUP
IRQ
ANY IRQ
Mx PIN FUNCTION LOGIC
Mx PINS
Figure 123. IRQ System Diagram
Rev. 0 | Page 184 of 205
Data Sheet
AD9546
WATCHDOG TIMER
The watchdog timer is a general-purpose programmable timer
capable of triggering a specific IRQ event (see Figure 124). The
timer relies on the system clock, however. Therefore, the
system clock must be present and locked for the watchdog
timer to be functional. The bit fields associated with the watchdog
timer reside in the Mx pin status and control function section of
the register map.
Because Bit 2 of Register 0x300B is a latched IRQ bit, the user
must clear it via Bit 2 of Register 0x2006 to obtain visibility of
subsequent watchdog timeout events. The user also has access to
the watchdog timeout status by assigning the watchdog timer to
an Mx status pin. In this case, the timeout event of the
watchdog timer is a pulse spanning 96 system clock periods
(approximately 40 ns).
The user sets the period of the watchdog timer via Bits[15:0] in
Register 0x010A and Register 0x010B. The 16-bit unsigned
timeout value carries units of milliseconds and provides a range
of 1 ms to 65.535 sec.
There are two ways to reset the watchdog timer, thereby
preventing it from indicating a timeout event. The first method
is by writing a Logic 1 to Bit 7 (an autoclearing bit) in Register
0x2005. Alternatively, the user can program any of the Mx pins
as a control pin to reset the watchdog timer, which allows the
user to reset the timer using a hardware pin rather than using
the register map.
A value of zero (default) disables the watchdog timer. The relative
accuracy of the timer is approximately 0.1% with an uncertainty
of 0.5 ms.
There are two typical cases for employing the watchdog timer,
both of which assume the watchdog timer output appears at the
output of an appropriately configured Mx status pin. The first case
is for an external device (an FPGA or microcontroller, for
example) to monitor the watchdog timer output, using the
output as a signal to carry out periodic housekeeping functions.
The second case is to have the watchdog timer output
connected to the external device such that the assertion of the
watchdog output resets the external device. In this way, under
normal operation, the external device repeatedly resets the
watchdog timer by either writing Logic 1 to the clear watchdog
bit or by asserting an Mx control pin configured for clearing the
watchdog. Therefore, as long as the external device keeps
resetting the watchdog timer before the timer times out, the
watchdog timer does not generate an output signal. As such,
the watchdog timer does not reset the external device.
However, if the external device fails to reset the watchdog timer
before the timeout period expires, the watchdog timer
eventually times out, resetting the external device via the
appropriately configured Mx status pin.
Updating the watchdog timer value automatically clears the
timer, ensuring a correct timeout period (per the new value)
starting from the moment of the update.
The 16-bit watchdog timer value relates to the timeout period
as follows:
Watchdog Timer Value = Timeout Period × 103
To determine the watchdog timer value necessary for a timeout
period of 10 sec,
Watchdog Timer Value = Timeout Period × 103
= 10 × 103
= 10,000
= 0x2710 (hexadecimal)
If enabled, the timer runs continuously and generates a timeout
IRQ event when the timeout period expires. The user has access
to the watchdog timer status as an IRQ bit (see the Interrupt
Request (IRQ) section) via Bit 2 of Register 0x300B.
Mx PINS
Mx PIN FUNCTION LOGIC
CLEAR
WATCHDOG VIA
Mx PIN
WATCHDOG
OUTPUT VIA
Mx PIN
REG 0x2005, BIT 7
40ns
CLEAR
INTERNAL
1kHz CLOCK
TO IRQ EVENT
SOURCE LOGIC
TIMER
OUT
CLK
PRESET
16
REG 0x010B TO 0x010A, BITS[15:0]
NOTE
1. A RANGE OF BITS USES A COLON SEPARATOR
Figure 124. Watchdog Timer
Rev. 0 | Page 185 of 205
AD9546
Data Sheet
EEPROM USAGE
EEPROM Upload
EEPROM OVERVIEW
To store the AD9546 register contents in the EEPROM, the
user must write a Logic 1 to Bit 0 of Register 0x2E02. Bit 0 does
not require an IO update. Writing a Logic 1 to this bit
immediately triggers an upload sequence.
The AD9546 supports an external, I2C-compatible, EEPROM
with dedicated access. With some restrictions, the AD9546 also
supports multidevice access to a single external EEPROM on a
shared I2C common bus. The AD9546 has an on-chip I2C
master to interface to the EEPROM through the Mx pins.
The AD9546 has the equivalent of a write protect feature in that
the user must write a Logic 1 to Bit 0 of Register 0x2E00 prior to
requesting an upload to the EEPROM. Attempting to upload to
the EEPROM without first setting Bit 0 results in a fault
Because the default register settings of the AD9546 do not
define a specific frequency translation, the user must factory
program the EEPROM content before it can be downloaded to
the register map (either automatically or manually). If desired,
the user can store custom device configurations by manually
forcing an upload to the EEPROM via the register map.
indication (that is, the AD9546 asserts Bit 1 of Register 0x300B).
A prerequisite to an EEPROM upload is the existence of an upload
sequence stored in the 15-byte EEPROM sequence section of
the register map (Register 0x2E10 to Register 0x2E1E). That is,
the user must store a series of upload instructions (see the
EEPROM Instruction Set section) in the EEPROM sequence
section of the register map prior to executing an EEPROM
upload.
EEPROM CONTROLLER GENERAL OPERATION
EEPROM Controller
The EEPROM controller governs all aspects of communication
with the EEPROM. Because the I2C interface uses a 100 kHz
(normal mode) or 400 kHz (fast mode) communication link,
the controller runs synchronous to an on-chip generated clock
source suitable for use as the I2C serial clock. The on-chip oscillator
enables asynchronously, immediately on a request for
activation of the controller. Upon startup, the oscillator notifies
the controller of its availability, and the controller activates.
After the requested controller operation is complete, the
controller disables the clock generator and returns to an idle
state.
The EEPROM controller performs an upload sequence by reading
the instructions stored in the EEPROM sequence section of the
register map byte by byte, and executing them in order. That is,
the data stored in the EEPROM sequence section of the register
map are instructions to the EEPROM controller on what to
store in the EEPROM (including operational commands and
AD9546 register data).
The EEPROM controller sets Bit 0 of Register 0x3000 to Logic 1
while the upload sequence is in progress as an indication to the
user that the controller is busy.
EEPROM Download
An EEPROM download transfers contents from the EEPROM
to the AD9546 programming registers and invokes specific
actions per the instructions stored in the EEPROM (see Table 109).
Automatic downloading is the most common method for
initiating an EEPROM download sequence, which initiates at
power-up of the AD9546, provided Pin M3 is Logic 1 at
power-up (see the Multifunction Pins at Reset or Power-Up
section). Alternatively, instead of cycling power to the AD9546 to
initiate an EEPROM download, the user can force the RESETB
pin to Logic 0, force Pin M3 to Logic 1, then return the
RESETB pin to Logic 1, and remove the drive source from
Pin M3.
Because the EEPROM sequence section of the register map is
only 15 bytes, the sequence typically cannot hold enough
instructions to upload a complete set of AD9546 data to the
EEPROM. Therefore, most upload sequences necessitate that the
user upload a series of subsequences. For example, to
accomplish a required upload sequence consisting of 20 bytes
of instructions, perform the following procedure:
1. Write the first 14 instructions to the EEPROM sequence
registers with the 15th instruction being a pause instruction
(see Table 109).
2. Initiate an EEPROM upload by writing Logic 1 to Bit 0 of
Register 0x2E02. When the EEPROM controller reaches
the pause instruction, it suspends the upload process and
waits for another assertion of Bit 0 of Register 0x2E02.
3. While the controller pauses, write the remaining six bytes
of the upload sequence into the EEPROM sequence
registers, followed by an end of data instruction (see
Table 109).
The user can also request an EEPROM download on demand
(that is, without resetting or cycling power to the AD9546) by
writing a Logic 1 to Bit 0 of Register 0x2E03.
Bit 0 does not require an IO update. Writing a Logic 1 to this
bit immediately triggers a download sequence.
The EEPROM controller sets Bit 1 of Register 0x3000 to Logic 1
while the download sequence is in progress as an indication to
the user that the controller is busy.
4. Initiate an EEPROM upload by writing Logic 1 to Bit 0 of
Register 0x2E02. When the EEPROM controller reaches the
end of data instruction, the upload process terminates.
Rev. 0 | Page 186 of 205
Data Sheet
AD9546
The preceding procedure is an example of an upload sequence
consisting of two subsequences. Most upload sequences require
more than two subsequences. However, the procedure is the
same. Specifically, partition a long sequence into several
subsequences by using the pause instruction at the end of each
subsequence and the end of data instruction at the end of the
final subsequence.
unaffected because there was no transfer of data to the AD9546
registers.
EEPROM Header
The EEPROM controller adds a header to stored data that
carries information related to the AD9546, such as
•
•
•
•
Vendor ID
Chip type
Product ID
Chip revision
EEPROM Checksum
When the EEPROM controller encounters an end of data
instruction (see Table 109) during an upload sequence, it
computes a 32-bit cyclic redundancy check (CRC) checksum
and appends it to the stored data in the EEPROM. Similarly,
when the EEPROM controller executes a download sequence, it
computes a checksum on-the-fly. At the end of a download
sequence, the EEPROM controller compares the newly
computed checksum to the one stored in the EEPROM. If the
two checksums do not match, the EEPROM controller asserts
Bit 3 of Register 0x3000.
At the beginning of an EEPROM download sequence, the
EEPROM controller compares the stored header values to the
values in the corresponding registers of the AD9546. If the
controller detects a mismatch, it asserts Bit 2 of Register 0x3000
and terminates the download.
EEPROM INSTRUCTION SET
The EEPROM controller relies on a combination of
instructions and data. An instruction consists of a single byte
(eight bits). Some instructions require subsequent bytes of
payload data. That is, some instructions are self contained
operations, whereas others are directions on how to process
subsequent payload data. Table 109 shows a summary of the
EEPROM controller instructions.
To minimize the possibility of downloading a corrupted
EEPROM data set, the user can execute a checksum test by
asserting the Bit 2 of Register 0x2E00, which causes the
EEPROM controller to execute a download sequence, but without
transferring data to the AD9546 registers. The controller still
computes an on-the-fly checksum, performs the checksum
comparison, and asserts Bit 3 of Register 0x3000 if the
checksums do not match. Therefore, after the device completes
the download sequence and deasserts Bit 1 of Register 0x3000, the
user can check that Bit 3 of Register 0x3000 = 0 to verify that the
test passed. However, even if the test fails, device operation is
When the controller downloads the EEPROM contents to the
AD9546 registers, it does so in a linear fashion, stepping
through the instructions stored in the EEPROM. However, when
the controller uploads to the EEPROM, the sequence is a nonlinear
combination of various parts of the register map and computed
data that the controller calculates on the fly.
Table 109. EEPROM Controller Instruction Set Summary
Instruction Code (Hexadecimal)
Response
Comments
0x00 to 0x7F
0x80
0x81 to 0x8F
0x90
0x91
0x92
0x93
0x94
0x95 to 0x97
0x98
0x99
0x9A
0x9B to 0x9F
0xA0
0xA1
0xA2
0xA3 to 0xAF
0xB0
0xB1 to 0xBF
0xC0 to 0xFD
0xFE
Register transfer
IO update
Not applicable
Calibrate all PLLs
Calibrate the system clock PLL
Calibrate APLL0 and APLL1
Calibrate APLL0
Calibrate APLL1
Not applicable
Requires a 2-byte register address suffix
Assert an IO update during download
Undefined
Calibrate the system clock PLL, APLL0, and APLL1 during download
Calibrate only the system clock PLL during download
Calibrate only APLL0 and APLL1 during download
Calibrate only APLL0 during download
Calibrate only APLL1 during download
Reserved/unused
Force DPLL0 and DPLL1 to freerun during download
Force only DPLL0 to freerun during download
Force only DPLL1 to freerun during download
Reserved/unused
Synchronize all distribution outputs during download
Synchronize only Channel 0 distribution outputs during download
Synchronize only Channel 1 distribution outputs during download
Reserved/unused
Force freerun
Force DPLL0 freerun
Force DPLL1 freerun
Not applicable
Synchronize outputs
Synchronize Channel 0
Synchronize Channel 1
Not applicable
Clear condition
Set condition
Not applicable
Apply Condition 0 and reset the condition map
Apply Condition 1 to Condition 15, respectively
Undefined
Pause
Pause the EEPROM upload sequence
0xFF
End of data
Marks the end of the instruction sequence
Rev. 0 | Page 187 of 205
AD9546
Data Sheet
Register Transfer Instructions (0x00 to 0x7F)
Condition Value
Instructions with a hexadecimal value from 0x00 through 0x7F
prescribe a register transfer operation. Register transfer
The condition value has a one to one correspondence to the
conditional instruction. Specifically, the condition value is the
value of the conditional instruction minus 0xB0. Therefore,
condition values have a range of 0 to 15. The EEPROM controller
uses condition values in conjunction with the condition map,
whereas the user uses a condition value to populate the
EEPROM load condition bit field of the register map with a
condition ID.
instructions require a 2-byte suffix, which constitutes the
starting address of the AD9546 register targeted for transfer
(where the first byte to follow the data instruction is the most
significant byte of the register address). When the EEPROM
controller encounters a data instruction, it knows to interpret
the next two bytes as the register map target address.
The value of the register transfer instruction encodes the
payload length (number of bytes). That is, the EEPROM
controller knows how many register bytes to transfer to and
from the indicated register by adding 1 to the instruction value.
For example, Data Instruction 0x1A has a decimal value of 26.
Therefore, the controller knows to transfer 27 bytes (one more
than the value of the instruction) to the target register for a load
operation or from the target register for a save operation.
Condition ID
The condition ID is the value stored in Bits[3:0] of
Register 0x2E01. The EEPROM controller uses the condition
ID in conjunction with the condition map to determine which
instructions to execute or ignore during a download sequence.
Condition Map
The condition map is a table maintained by the EEPROM
controller consisting of a list of condition values. When the
EEPROM controller encounters a conditional instruction
during a download sequence, it determines the corresponding
condition value of the instruction (0 to 15). If the condition
value is nonzero, the EEPROM controller places the condition
value in the condition map. Conversely, if the condition value
is zero, the controller clears the condition map and applies
Condition 0 (the absence of conditions). Condition 0 causes all
subsequent instructions to execute unconditionally (until the
controller encounters a new conditional instruction that causes
conditional processing).
IO Update Instruction (0x80)
When the EEPROM controller encounters an IO update
instruction during an upload sequence, it stores the instruction
in EEPROM. When encountered during a download sequence,
however, the EEPROM controller initiates an IO update event,
which is equivalent to the user asserting Bit 0 of Register 0x000F
(IO update).
Device Action Instructions (0x90 to 0xA2)
When the EEPROM controller encounters a device action
instruction during an upload sequence, it stores the instruction
in EEPROM. When encountered during a download sequence,
however, the EEPROM controller executes the specified action
per Table 109.
Conditional Processing
While executing a download sequence, the EEPROM controller
executes or skips instructions depending on the condition ID
and the contents of the condition map (except for the condi-
tional and end of data instructions, which always execute
unconditionally).
Conditional Instructions (0xB0 to 0xBF)
The conditional instructions allow conditional execution of
EEPROM instructions during a download sequence. During an
upload sequence, however, they are stored as is and have no
effect on the upload process.
If the condition map is empty or the condition ID is zero, all
instructions execute unconditionally during download.
However, if the condition ID is nonzero and the condition map
contains a condition value matching the condition ID, the
EEPROM controller executes the subsequent instructions.
Alternatively, if the condition ID is nonzero but the condition
map does not contain a condition value matching the condition
ID, the EEPROM controller skips instructions until it encounters
a conditional instruction with a condition value of zero or a
condition value matching the condition ID.
Conditional processing makes use of four elements:
•
•
•
•
The conditional instruction
The condition value
The condition ID
The condition map
Conditional Instruction
When encountering a conditional instruction during an upload
sequence, the EEPROM controller stores the instruction in the
EEPROM. When the EEPROM controller detects a conditional
instruction during a download sequence, the instruction affects
the condition map and the outcome of the conditional processing.
The condition map allows multiple conditions to exist at any
given moment. This multiconditional processing mechanism
enables the user to have one download instruction sequence
with many possible outcomes, depending on the value of the
condition ID and the order in which the controller encounters
conditional instructions. Table 110 shows an example of the use
of conditional processing.
Rev. 0 | Page 188 of 205
Data Sheet
AD9546
AD9546
DEVICE 1
AD9546
DEVICE 2
Table 110. Example Conditional Processing Sequence
Instruction
Operation
SCL SDA M1 M2
33 35
SCL SDA M1 M2
33 35
2
4
2
4
0x00 to 0x7F
A sequence of register transfer instructions
that execute unconditionally
Apply Condition 1
A sequence of register transfer instructions
that execute only if the condition ID is 1
SCL
SDA
EEPROM
0xB1
0x00 to 0x7F
SCL
SDA
CPU
SCL SDA
SCL SDA
Figure 125. Level 1 Multidevice Configuration
AD9546 AD9546
0xB2
0xB3
0x00 to 0x7F
Apply Condition 2
Apply Condition 3
A sequence of register transfer instructions that
execute only if the condition ID is 1, 2, or 3
DEVICE 1 DEVICE 2
SCL SDA M1 M2 SCL SDA M1 M2
33 35
35
2
4
33
2
4
0x91
0xB0
0x80
0xFF
Calibrate the system clock PLL
Clear condition map
IO update
SDA
SCL
SCL SDA
EEPROM
SCL SDA
CPU
Terminate sequence
Figure 126. Level 2 Multidevice Configuration
Pause Instruction (0xFE)
Multidevice Bus Arbitration
The EEPROM controller only recognizes the pause instruction
during an upload sequence. Upon encountering a pause
instruction, the EEPROM controller enters an idle state, but
preserves the current value of the EEPROM address pointer.
The EEPROM controller implements bus arbitration by
continuously monitoring the SDA and SCL bus signals for start
and stop conditions. The controller can determine whether the
bus is idle or busy. If the bus is busy, the EEPROM controller
delays its pending I2C transfer until a stop condition indicates
that the bus is available.
One use of the pause instruction is for saving multiple, yet
distinct, values of the same AD9546 register, which is useful for
sequencing power-up conditions.
Bus arbitration is essential in cases where two I2C master
devices simultaneously attempt an I2C transfer. For example, if
one I2C master detects that SDA is Logic 0 when it is intended to
be Logic 1, the master assumes that another I2C master is active
and immediately terminates its own attempt to transfer data.
Similarly, if one I2C master detects that SCL is Logic 0 prior to
entering a start state, the master assumes that another I2C
master is active and stalls its own attempt to drive the bus.
In either case, the prevailing I2C master completes its current
transaction before releasing the bus. Because the postponed I2C
master continuously monitors the bus for a stop condition, it
attempts to seize the bus and carry out the postponed transaction
on detection of such a stop condition.
The pause instruction is also useful for executing an upload
sequence requiring more space than is available in the EEPROM
sequence registers (see the EEPROM Upload section).
End of Data Instruction (0xFF)
When the EEPROM controller encounters an end of data
instruction during an upload sequence, it stores the instruction
in EEPROM along with the computed checksum, clears the
EEPROM address pointer, and then enters an idle state. When
encountered during a download sequence, however, the
EEPROM controller clears the EEPROM address pointer,
verifies the checksum, and then enters an idle state.
During EEPROM downloads, condition instructions always
execute unconditionally.
The EEPROM controller includes an arbitration timer to optimize
the bus arbitration process. Specifically, when the EEPROM
controller postpones an I2C transfer as a result of detecting bus
contention, the arbitration timer starts. If the EEPROM controller
fails to detect a stop condition within 255 SCL cycles, it attempts
to force another transaction. If the bus is still busy, the EEPROM
controller restarts the arbitration timer, and the process continues
until the EEPROM controller eventually completes the pending
transaction.
MULTIDEVICE SUPPORT
Multidevice support enables multiple AD9546 devices to share
the contents of a single EEPROM. There are two levels of
multidevice support. Level 1 supports a configuration where
multiple AD9546 devices share a single EEPROM through a
dedicated I2C bus. Level 2 supports a configuration where
multiple AD9546 devices share a single EEPROM connected to
a common I2C bus that includes other I2C master devices.
Figure 125 and Figure 126 show the Level 1 and Level 2
configurations, respectively. Only the I2C functions of the
multifunction pins are shown.
Rev. 0 | Page 189 of 205
AD9546
Data Sheet
Table 111. Template for a Multidevice EEPROM Sequence
Multidevice Configuration Example
Instructions
Comment
Consider two AD9546 devices (Device 1 and Device 2) that
share a single EEPROM and assume both devices have a
common PLL0 configuration but differing PLL1 configurations.
0x00 to 0x7F
A sequence of register transfer instructions
associated with the PLL0 configuration
common to both devices
A template for an EEPROM sequence that accomplishes this
configuration is shown in Table 111.
0xB1
0x00 to 0x7F
Apply Condition 1
A sequence of register transfer instructions
associated with the PLL1 configuration
specific to Device 1
Clear the condition map
Apply Condition 2
A sequence of register transfer instructions
associated with the PLL1 configuration
specific to Device 2
The sequence relies on conditional processing to differentiate
between Device 1 and Device 2. Therefore, the user must
program the condition ID of both devices prior to executing an
EEPROM download. Specifically, the user must program
Bits[3:0] of Register 0x2E01 in Device 1 with a condition ID of
1 and in Device 2 with a condition ID of 2.
0xB0
0xB2
0x00 to 0x7F
0xB0
0x80
0xFF
Clear the condition map
IO update
End of sequence
Rev. 0 | Page 190 of 205
Data Sheet
AD9546
APPLICATIONS INFORMATION
A key aspect of any synchronized clocking system is the timing
error caused by the propagation delay of clock signals sent
between the master and slave nodes. To properly align the
common time scale across all nodes, the system must measure
and compensate for the propagation delay between nodes. The
recommended method for quantifying the propagation delay is
measuring round trip delay in real time, which allows
continuous assessment and correction of delay that may arise
from temperature variations in the system. The AD9546 readily
accommodates assessment of round trip delay via its analog
loopback feature (see the Analog Clock Loopback section) and
time skew measurement processor (see the application example
in Figure 129 and the Clock Propagation Delay Measurement
section).
DIGITIZED CLOCKING APPLICATION
The diagram in Figure 127 shows a simplified example of a
digitized clocking system. The system contains three digitized
clocking nodes, with one of the nodes functioning as the master
node and the other two nodes functioning as slave nodes (Slave
Node 0 and Slave Node 1). The master node receives a common
clock reference signal that serves as the common clock for the
entire system.
A reference clock can be applied at an input to one of the slave
nodes. The goal is to distribute the reference clock to each of
the nodes using data transport over a common digital
communication link rather than via analog clock signals. That
is, to transport the reference clock in frequency and phase from
the selected node to the other nodes over a data link rather than
routing the reference clock throughout the system in the form
of an analog signal. Digitized clocking technology enables this
ability to transport clock signals over a digital bus (see the
Digitized Clocking section for details).
An example of a digitized clocking system appears in Figure 128,
which includes round trip delay measurement capability. The
example system comprises a timing card and a pair of line cards
interconnected via a backplane. The timing card and line cards
rely on digitized clocking to transport clock signals as time
codes via the AD9546 SPI port. Although not explicitly shown
in Figure 128, the AD9546 analog loopback feature for round
trip delay measurement requires the use of REFB or REFBB as
the loopback input and the M4 pin as the loopback return.
The master node generates the common clock as a reference
clock to the slave nodes. The common clock is the time base
shared by all nodes, a fundamental digitized clocking
requirement. The master node also sends synchronization
information to the slave nodes. Synchronization comprises two
components: a physical synchronization event signal
(embedded within the common clock signal in Figure 127) and
a synchronizing code sent over the digital communication bus
(corresponding to the time associated with the synchronization
event signal). This two-pronged synchronization approach
results in all slave nodes sharing a common epoch with the
master node, with all nodes having a common (synchronized)
time scale.
The timing card employs an IEEE 1588 servo for transporting
timing over ethernet (the traffic labels in the diagram). The
servo controls a PLL with the output of the PLL providing the
common clock for the system. The PLL produces three
identical clock signals with coincident output clock edges that
serve as the common clock input to each of the cards. This
arrangement, in combination with the AD9546 analog
loopback capability, allows the timing card to measure the
propagation delay between the timing card and each line card.
The measured propagation delay can then be compensated via
offsets applied to the digitized clocking paths.
For example, suppose the digitized clocking system has a
distributable reference clock applied to Slave Node 1. Slave
Node 1 digitizes the reference clock by creating time codes
(digital word representations of the reference clock rising
edges) derived from the common time scale. The time codes are
a proxy for the physical reference clock signal. Using digitized
clocking, Slave Node 1 sends the time codes to the master node
over the digital bus. Likewise, the master node distributes the
time codes to Slave Node 0 (and back to the Slave Node 1, if
required) over the same digital bus. Then, using digitized
clocking, the receiving nodes convert the time codes back into a
physical clock signal with the same frequency and phase as the
reference clock and aligned to within 100 ps.
The line cards handle Ethernet traffic via a physical layer (PHY)
with each line card serving multiple PHYs. Figure 128 shows
only one PHY with a connection to a UTS/IUTS pair on the
AD9546 for simplification. However, each PHY has access to a
dedicated UTS/IUTS pair on the AD9546. In this way, any one of
the PHYs can provide a received clock to the AD9546 for
digitization via a UTS. The digitized received clock is
transported over the digital bus to the timing card, where the
timing card uses the digitized receive clock as a SyncE clock for
the IEEE 1588 servo. The timing card, in turn, generates a
digitized clock for transport back to the line cards over the
digital bus. The line cards convert the digitized clock back to an
analog clock for use by the PHYs for clocking outgoing traffic.
For digitized clocking applications, the recommendation is to
operate the AD9546 in SPI mode, rather than I2C mode,
because SPI supports a much higher serial transfer rate than I2C
(50 MHz for SPI vs. 400 kHz for I2C).
Rev. 0 | Page 191 of 205
AD9546
Data Sheet
SLAVE NODE 0
DIGITIZED CLOCKING NODE
DISTRIBUTABLE
REFERENCE CLOCK
2
DIGITAL COMMUNICATION BUS
SPI/I C
RECOVERED
CLOCKS
CONTROLLER
COMMON CLOCK
(PLUS EMBEDDED SYNCHRONIZATION)
MASTER NODE
DIGITIZED CLOCKING NODE
ANALOG LOOPBACK
ANALOG LOOPBACK
2
SPI/I C
RECOVERED
CLOCK
CONTROLLER
COMMON
CLOCK
REFERENCE
COMMON CLOCK
(PLUS EMBEDDED SYNCHRONIZATION)
2
SPI/I C
RECOVERED
CLOCKS
CONTROLLER
DISTRIBUTABLE
REFERENCE CLOCK
DIGITIZED CLOCKING NODE
SLAVE NODE 1
Figure 127. Digitized Clocking Application
LINE CARD
BACK PLANE
AD9546 ANALOG LOOP BACK
ALP
AD9546 USER TIME STAMPER
UTS
IUTS
SPI
AD9546
DATA IN
CLOCK 1
AD9546 INVERSE USER TIME STAMPER
AD9546 SERIAL PORT INTERFACE
AD9546 TIME TO DIGITAL CONVERTER
COMMON CLOCK (DIRECT AND LOOPBACK PATHS)
UTS
PHY
TRAFFIC
TDC
SPI
IUTS
RECOVERED
CLOCK 1
TIME CODES: LINE CARD
TIME CODES: TIMING CARD
SyncE CLOCK
TIMING CARD
LINE CARD
DATA IN
CLOCK 2
UTS
RECOVERED
CLOCK 2
PHY
TRAFFIC
ALP
1588 TIME STAMPS
DIGITAL BUS
IUTS
TIMING CARD
LINE CARD
AD9546
DATA IN
CLOCK 3
SPI
TDC
TDC
TDC
UTS
PHY
PHY
TRAFFIC
TRAFFIC
SPI
IUTS
RECOVERED
CLOCK 3
AD9546
DATA IN
CLOCK 4
UTS
RECOVERED
CLOCK 4
ALP
SyncE
IUTS
1588
SERVO
PLL
IDENTICAL, TIME-ALIGNED
COMMON CLOCKS
Figure 128. Digitized Clocking System Example
Rev. 0 | Page 192 of 205
Data Sheet
AD9546
To measure propagation delay, the user connects the in and out
signals to any two of the eight independent reference clock
inputs of the AD9546. The reference clock inputs comprise the
REFx pins or one of the Mx control pins configured as an input
to one of the four auxiliary references (auxiliary REFx). Then,
using the skew measurement processor (see the Timing Skew
Measurements Using Two TDCs section), the user can read the
measurement results via the serial port, which indicates the
value of tP with approximately 1 ps resolution.
CLOCK PROPAGATION DELAY MEASUREMENT
Consider an application comprising a clock source (in) driving
additional components that introduce propagation delay (tP) to
the final output (out) signal (see Figure 129). The additional
components can be, for example, a combination of physical
interconnects (cables and PCB traces), clock buffers or other
elements that introduce propagation delay.
In applications requiring quantification of the propagation
delay between two clock signals, the TDCs in the AD9546
provide a means to measure the delay and make the result
available in digital form via serial port registers. Figure 129
shows a simplified propagation delay measurement setup using
the AD9546.
The measured value of tP can then be used by the controller to
adjust the timing of the clock source to compensate for the
delay through the clock components. By periodically measuring
tP and compensating accordingly, the user can correct for time
dependent changes in tP that may arise from temperature
variations.
0
0
tP
IN
OUT
CLOCK
SOURCE
DELAY
COMPONENTS
REFx/Mx
REFx/Mx
SKEW
MEASUREMENT
PROCESSOR
TDC
TDC
SERIAL
PORT
AD9546
2
SPI/I C
DELAY ADJUST
CONTROLLER
Figure 129. Propagation Delay Measurement Application
Rev. 0 | Page 193 of 205
AD9546
Data Sheet
making it possible to capture time stamps from all eight input
time stamp sources (REFx and auxiliary REFx) concurrently.
TIME STAMP MEASUREMENT
The AD9546 is particularly well suited for applications
requiring the measurement of clock frequencies and clock
timing offsets. Measurement of frequency and time offset is
made possible by digitizing (time stamping) input clock edges
using TDCs. Each of the four primary reference inputs (REFA,
REFAA, REFB, REFBB) and four auxiliary reference inputs
(auxiliary REF0, auxiliary REF1, auxiliary REF2, and auxiliary
REF3) of the AD9546 has a dedicated TDC (see Figure 130).
The UTS system stores captured time stamps in the UTS FIFO,
which the user accesses via the serial port (see the UTS
Readback FIFO section). The FIFO has limited capacity, which
means if the time stamp arrival rate to the FIFO exceeds the
FIFO extraction rate (on average), the FIFO eventually
overflows. Because a FIFO overflow causes the FIFO to stop
accepting time stamps from the UTSs, the user must extract the
FIFO contents at an average rate that exceeds the average time
stamp arrival rate to the FIFO (otherwise, time stamps are lost).
Thus, the serial port read rate puts an upper limit on the time
stamp arrival rate to the FIFO.
The AD9546 embodies a user time stamp capturing system that
gives the user access to time stamps generated by the TDCs (see
the User Time Stamper (UTS) section). The UTS system
embodies nine time stamp capturing units (UTS 0 to UTS 8),
DIGITAL CROSS
POINT MUX
AD9546
REFA
TDC
UTS 0
UTS 1
UTS 2
UTS 3
UTS 4
UTS 5
UTS 6
UTS 7
UTS 8
REFAA
TDC
REFB
TDC
REFBB
TDC
2
SPI/I C
AUX REF0
UP TO 8
CLOCK INPUTS
UTS
FIFO
SERIAL
PORT
M0
M1
CONTROLLER
TDC
AUX REF1
TDC
M2
M3
Mx PIN
MATRIX
AUX REF2
TDC
M4
SDO/M5
CSB/M6
AUX REF3
TDC
TIME
STAMPS
TIME
STAMPS
Figure 130. Time Stamp Measurement Application
Rev. 0 | Page 194 of 205
Data Sheet
AD9546
in the System Clock Compensation section). Specifically, the
IEEE 1588 SERVO
reference input serves as a frequency stable source to the system
clock compensation block (for example, an OXCO or TCXO).
The frequency stable source allows the AD9546 to use a crystal
resonator as the system clock source (XOA and XOB pins) and
compensate for the inherent crystal frequency instability,
thereby maintaining the overall timing precision of the system.
The AD9546 provides the jitter cleaning, phase shifting, and
synchronization necessary for an IEEE 1588 servo and software
stack in a Telecom-BC (T-BC) with hybrid support application
(see Figure 131). Figure 131 represents a superset of possible
connection scenarios for implementing a variety of IEEE 1588
applications rather than a specific IEEE 1588 application. The
diagram in Figure 131 supports IEEE 1588 master and slave
clock operation. The dashed box in Figure 131 constitutes
external circuitry that handles precision time protocol (PTP)
packets, software to implement IEEE 1588, and time of day
support. The integrated features of the AD9546 simplify the
task of synchronizing clocks via PTP. Furthermore, the Mx pins
of the AD9546 provide the option to implement assisted partial
timing support (APTS) via a GPS receiver providing a precision
1 pps source.
PLL0 uses the 125 MHz SyncE clock as a reference input to
generate two filtered, 125 MHz SyncE clocks. PLL1 uses one of
the auxiliary NCOs of the AD9546 as a reference input to
generate two arbitrary output clocks with stability
commensurate with the REF IN source. The system clock
compensation system ensures frequency stability of the NCOs
associated with the DPLLs and auxiliary NCOs.
The performance and feature set of the AD9546 make it ideal
for enabling the latest IEEE 1588 features, as well as the latest
ITU-T packet synchronization related standards for 4G and 5G
wireless networks.
Figure 131 implies the use of the AD9546 system clock
compensation feature (see the Compensation Method 3 section
REFA/REFAA
AD9546
÷Q0A
OUT0Ax
÷RA
REFA
DEMOD
DEMOD
TDC
TDC
÷Q0AA
REFERENCE
INPUT
DPLL0
APLL0
REFAA
÷RAA
2.424GHz TO 3.232GHz
÷Q0B
OUT0Bx
OUT0Cx
REFB
÷Q0BB
DIGITAL CROSS
POINT MUX
REFB/REFBB
REFERENCE REFERENCE
REFBB
3.232GHz TO 4.040GHz
÷Q0C
SWITCHING CONTROL MONITORS (8)
DPLL1
APLL1
÷Q0CC
AUXILIARY REFERENCE (4)
INTERNAL
ZERO DELAY
SYSTEM CLOCK
DEMOD
÷R
TDC
COMPENSATION
÷Q1A
INVERSE
USER TIME
STAMPS (2)
OUT1Ax
OUT1Bx
CLOCK OUT
CLOCK OUT
÷Q1AA
AUXILIARY
NCO (2)
TEMPERATURE
SENSOR
SYSTEM
CLOCK
÷Q1B
SERIAL PORT
USER TIME
STAMPS (9)
2
÷Q1BB
(SPI OR I C)
AND EEPROM
CONTROLLER
STATUS AND
CONTROL
PINS
AUXILIARY CLOCK
OUTPUTS
(1PPS TO 500MHz)
SYSTEM
CLOCK PLL
COMMON CLOCK
SYNCHRONIZER
MODULATION, PHASE
OFFSET, AND JESD204B
2
Mx PINS
XOA XOB
OPTIONAL
ASSISTED
PARTIAL
TIMING
SUPPORT
(APTS)
SPI/I C PORT
125MHz FILTERED/ADJUSTED
125MHz FILTERED/ADJUSTED
1PPS
GPS
125MHz (SyncE/CDR)
CLOCK INTERFACES
SERIAL PORT
PACKET
CLASSIFIER
AND
TIME
RGMII/SGMII
(WITH SyncE SUPPORT)
TIME OF DAY IN/OUT
(1PPS + UART WITH NEMA FORMAT)
OF
DAY
TIMESTAMP
MAC
CPU
MEMORY
MANAGEMENT
INTERFACE
Figure 131. Using an AD9546 in an IEEE 1588 Version 2 Application
Rev. 0 | Page 195 of 205
AD9546
Data Sheet
INITIALIZATION SEQUENCE
The suggested initialization sequence for powering on and programming the AD9546 is shown in Figure 132 to Figure 134.
START
YES
NO
SPI
EEPROM
DOWNLOAD
SET M3 HIGH
2
I C
2
SPI OR I C
APPLY VDD
(ALL DOMAINS)
SET M4 HIGH
SET M4 LOW
2
END
SELECT AD9546 I C
ADDR USING CSB/M6, SDO/M5
APPLY VDD (ALL
DOMAINS)
NO
THIS LOOP SERVES AS AN INDICATOR
THAT THE POWER-ON RESET CIRCUITRY
HAS TRIGGERED AN INTERNAL RESET.
REGISTER 0x000C = 0x56
YES
REFERENCE INPUT(S)
CAN BE APPLIED
ANY TIME HEREAFTER
CHIP LEVEL
RESET LOOP
ISSUE A
RST_COUNT =
RST_COUNT + 1
RST_COUNT = 0
CHIP LEVEL RESET
(PIN OR SOFT RESET)
SOFTWARE
WRITE REGISTER
CONTENT FROM
SETUP FILE
GENERATED
AD9546
REGISTER DUMP
WRITE:
REGISTER 0x000F = 0x01
SUBPROCESS
SYSTEM CLOCK
INITIALIZATION
x = 0
NO
RST_COUNT > 0
RAISE FLAG FOR
DEBUGGING
YES
WRITE
REGISTER 0x000F = 0x01
READ:
REGISTER 0x3000 TO REGISTER 0x3023
REGISTER 0x3100 TO REGISTER 0x310E
REGISTER 0x3200 TO REGISTER 0x320E
APLLx
NO
ENABLED
YES
SUBPROCESS:
ANALOG PLLx
INITIALIZATION
x = x +1
NO
x > 1
YES
END
Figure 132. Programming Sequence Flowchart
Rev. 0 | Page 196 of 205
Data Sheet
AD9546
START
SYSTEM CLOCK
RECALIBRATION LOOP
CAL_COUNT = 0
VCO
CALIBRATION
OPERATION
WRITE:
REGISTER 0x2000, BIT 2 = 1
NO
WRITE:
REGISTER 0x000F = 0x01
END
(TO RST_COUNT CHECK)
YES
CAL_COUNT > 1
WRITE:
REGISTER 0x2000, BIT 2 = 0
WRITE:
REGISTER 0x000F = 0x01
CAL_COUNT =
CAL_COUNT + 1
SYSTEM CLOCK
LOCKED AND
STABLE POLLING LOOP
START TIMEOUT CLOCK:
TIME = 0
TIMEOUT CLOCK:
TIME > SYSCLK_TO
NO
YES
REGISTER 0x3001,
BITS[1:0] = 0x3
1
YES
END
SYSCLK_TO IS A TIMEOUT VALUE CALCULATED BY THE CONTROLLING SOFTWARE.
1
SYSCLK_TO = 50ms + SYSTEM CLOCK VALIDATION TIME (REGISTER 0x0207 TO REGISTER 0x0209 [UNITS OF ms])
Figure 133. System Clock Initialization Subprocess
Rev. 0 | Page 197 of 205
AD9546
Data Sheet
APLL RECALIBRATION
LOOP
START
APLL REGISTER REGISTER
REGISTER
REGISTER
0
1
0x2100
0x2200
0x3100
0x3200
0x10DB
0x14DB
0x2101
0x2201
CAL_COUNT = 0
NO
VCO
CALIBRATION
OPERATION
END
(TO RST_COUNT
CHECK)
YES
CAL_COUNT > 1
WRITE:
CAL REGISTER, BIT 1 = 1
WRITE:
REGISTER 0x0F, BIT 0 = 1
WRITE:
REGISTER 0x0F, BIT 0 = 1
CAL_COUNT =
CAL_COUNT + 1
WRITE:
CAL REGISTER, BIT 1 = 0
APLL LOCK DETECT
POLLING LOOP
START TIMEOUT CLOCK:
TIME = 0
NO
NO
YES
TIMEOUT CLOCK:
TIME > 50ms
LOCK REGISTER, BIT 3 = 1
YES
WRITE:
CAL REGISTER, BIT 1 = 0
WRITE:
REGISTER 0x0F, BIT 0 = 1
AUTOSYNC
REGISTERS, BITS[1:0] = 0
NO
YES
MANUAL DISTRIBUTION
SYNCHRONIZATION
OPERATION
WRITE:
SYNC REGISTER, BIT 3 = 1
WRITE:
REGISTER 0x0F, BIT 0 = 1
WRITE:
SYNC REGISTER, BIT 3 = 0
WRITE:
REGISTER 0x0F, BIT 0 = 1
END
Figure 134. APLL Initialization Subprocess
Rev. 0 | Page 198 of 205
Data Sheet
AD9546
SERIAL CONTROL PORT
directions. The 4-wire mode uses SDIO for transferring data to
the AD9546, and SDO for transferring data from the AD9546.
The output drive strength of the SDO and SDIO pins is
selectable via Bit 7 of Register 0x0109.
SERIAL CONTROL PORT OVERVIEW
The AD9546 serial control port is a flexible, synchronous serial
communications port that provides a convenient interface to
many industry-standard microcontrollers and microprocessors.
The AD9546 serial control port is compatible with most
synchronous transfer formats, including I²C, Motorola SPI, and
Intel SSR protocols. The serial control port allows read/write
access to the AD9546 register map.
The chip select (CSB) signal is an active low control that gates
read and write operations. Assertion (active low) of CSB
initiates a write or read operation to the AD9546 SPI port. The
user can transfer any number of data bytes in a continuous stream.
The register address is automatically incremented or decremented
based on the setting of the address ascension bit (Register 0x0000).
The user must deassert CSB following the last byte transferred,
thereby ending the stream mode. CSB is internally connected to
a 10 kΩ pull-up resistor. When CSB is high, SDIO and SDO
enter a high impedance state.
The AD9546 uses the Analog Devices unified SPI protocol (see
the Analog Devices Serial Control Interface Standard). The
unified SPI protocol guarantees that all new Analog Devices
products using the unified protocol have consistent serial port
characteristics. The SPI port configuration is programmable via
Register 0x0000.
Implementation Specific Details
Unified SPI differs from the SPI port found on older Analog
Devices products, such as the AD9557 and AD9558, in the
following ways:
The Analog Devices Serial Control Interface Standard provides
a detailed description of the unified SPI protocol and covers
items such as timing, command format, and addressing. The
unified SPI protocol defines the following device specific items:
•
•
Unified SPI does not have byte counts. A transfer is
terminated when the CSB signal goes high. The W1 and
W0 bits in the traditional SPI become the A12 and A13
bits of the register address (see Table 113). This format is
much like streaming mode in the traditional SPI.
The address ascension bit (Register 0x0000) controls
whether register addresses are automatically incremented
or decremented regardless of the LSB/MSB first setting. In
a traditional SPI, LSB first mode dictated auto-incrementing
and MSB first mode dictated auto-decrementing of the
register address.
•
•
•
•
Analog Devices unified SPI protocol revision: 1.0
Chip type: 0x5
Product ID: 0x0121
Physical layer: 3-wire and 4-wire supported and 1.5 V,
1.8 V, and 2.5 V operation supported
Optional single-byte instruction mode: not supported
Data link: not used
•
•
•
Control: not used
Communication Cycle—Instruction Plus Data
•
The first 16 register addresses of devices that adhere to the
unified serial port have a consistent structure.
The unified SPI protocol consists of a two-part communication
cycle. The first part is a 16-bit instruction word coincident with
the first 16 SCLK rising edges. The second part is the payload,
the bits of which are coincident with SCLK rising edges. The
instruction word provides the AD9546 serial control port with
information regarding the payload. The instruction word
SPI/I²C PORT SELECTION
Although the AD9546 supports both the SPI and I2C serial port
protocols, only one is active following power-up (as determined
by the M4 pin during the start-up sequence). The only way to
change the serial port protocol is to reset (or power cycle) the
device. See Table 103 for the I2C address assignments.
W
includes the R/ bit that indicates the direction of the payload
transfer (that is, a read or write operation). The instruction
word also indicates the starting register address of the first
payload byte.
SPI SERIAL PORT OPERATION
Pin Descriptions
IO Update
The serial clock (SCLK) signal serves as the serial shift clock.
This pin is an input. SCLK synchronizes serial control port read
and write operations. The rising edge SCLK registers write data
bits, and the falling edge registers read data bits. The SCLK line
supports a maximum clock rate of 50 MHz.
In general, there are two sets of serial port registers: buffer
registers and active registers. When the SPI port receives data, it
deserializes the data, which collects in the buffer registers. In
most cases, the output of the buffer registers route to the input
of the active registers rather than directly to the intended circuit
blocks within the device. As such, when the user writes SPI
data, the intended functional blocks within the device are
unaffected by the newly written data.
The user can configure the SPI port hardware configuration
and data format via the Register 0x0000. The hardware
configurations are 3-wire bidirectional (default) or 4-wire
unidirectional mode and the data formats are MSB first
(default) or LSB first. The 3-wire mode uses the serial data
input/output (SDIO) signal for transferring data in both
Rev. 0 | Page 199 of 205
AD9546
Data Sheet
To make the buffer register data effective, the user must
perform an IO update operation, which transfers the contents
of all buffer registers to the active registers. The output of the
active registers causes the desired effect upon the functional
blocks within the device.
A readback operation takes data from either the serial control
port buffer registers or the active registers, as determined by
Register 0x0001, Bit 5.
SPI Instruction Word (16 Bits)
W
The MSB of the 16-bit instruction word is R/ , which
There are two methods available for issuing an IO update
operation:
indicates whether the ensuing operation is read or write. The
next 15 bits are the register address (A14 to A0), which indicates
the starting register address of the read/write operation (see
Table 113). The SPI controller ignores A14, treating it as Logic 0,
because the AD9546 has no register addresses requiring more
than a 14-bit address word.
1. Writing Logic 1 to Bit 0 of Register 0x000F
2. Configuring an Mx pin as a physical IO update pin
In the case of the first method, Bit 0 is an autoclearing bit. In
the case of the second method, applying Logic 1 to the
configured Mx pin asserts the IO update operation.
SPI MSB First and LSB First Transfers
The AD9546 instruction word and payload can be transferred
MSB first or LSB first. The default for the AD9546 is MSB first.
To invoke LSB first mode, write a Logic 1 to Register 0x0000,
Bit 6. Immediately after invoking LSB first mode, subsequent
serial control port operations are LSB first.
The user can program as many register bits as desired before
executing the IO update operation. Then, upon assertion of the
IO update operation, the all the buffer register contents transfer
to their active register counterparts.
Write
Address Ascension
When the instruction word indicates a write operation, the
payload is written into the serial control port buffer of the
AD9546. Data bits are registered on the rising edge of SCLK.
Generally, it does not matter what data is written to blank
registers. However, it is customary to use 0s. The user must
verify that all reserved registers within a specific range have a
default value of 0x00. However, Analog Devices makes every
effort to avoid having reserved registers with nonzero default
values.
If the address ascension bit (Register 0x0000, Bit 5) is Logic 0,
serial control port register addressing decrements from the
specified starting address toward Address 0x0000. If the address
ascension bit (Register 0x0000, Bit 5) is Logic 1, serial control
port register addressing increments from the starting address
toward Address 0x3A3B. Reserved addresses are not skipped
during multibyte input/output operations. Therefore, write the
default value to a reserved register and Logic 0s to unmapped
registers. It is more efficient to issue a new write command
than to write the default value to more than two consecutive
reserved (or unmapped) registers.
Read
If the instruction word indicates a read operation, the next
N × 8 SCLK cycles clock out the data starting from the address
specified in the instruction word, where N is the number of
data bytes to read. Read data appears on the appropriate data
pin (SDIO or SDO) on the falling edge of SCLK. The user must
latch the read data on the rising edge of SCLK. The internal SPI
control logic does not skip over blank registers during a
readback operation.
Table 112. Streaming Mode (No Addresses Skipped)
Address Ascension
Stop Sequence
0x0000 to 0x3A3B
0x3A3B to 0x0000
Increment
Decrement
Table 113. Serial Control Port, 16-Bit Instruction Word
MSB
LSB
I0
I15
I14
I13
I12
I11
I10
I9
I8
I7
I6
I5
I4
I3
I2
I1
R/W
A14
A13
A12
A11
A10
A9
A8
A7
A6
A5
A4
A3
A2
A1
A0
CSB
DON'T CARE
DON'T CARE
DON'T CARE
SCLK
DON'T CARE
SDIO
R/W A14 A13 A12 A11 A10 A9 A8 A7 A6 A5 A4 A3 A2 A1 A0 D7 D6 D5 D4 D3 D2 D1 D0 D7 D6 D5 D4 D3 D2 D1 D0
16-BIT INSTRUCTION HEADER REGISTER (N) DATA REGISTER (N – 1) DATA
Figure 135. Serial Control Port Write—MSB First, Address Decrement, Two Bytes of Data
Rev. 0 | Page 200 of 205
Data Sheet
AD9546
CSB
SCLK
DON'T CARE
DON'T CARE
SDIO
SDO
A9 A8 A7 A6 A5 A4 A3 A2 A1 A0
D7 D6 D5 D4 D3 D2 D1 D0 D7 D6 D5 D4 D3 D2 D1 D0 D7 D6 D5 D4 D3 D2 D1 D0 D7 D6 D5 D4 D3 D2 D1 D0
REGISTER (N) DATA REGISTER (N – 1) DATA REGISTER (N – 2) DATA REGISTER (N – 3) DATA
DON'T CARE
16-BIT INSTRUCTION HEADER
DON'T
CARE
Figure 136. Serial Control Port Read—MSB First, Address Decrement, Four Bytes of Data
tDS
tHIGH
tS
tC
tCLK
tDH
tLOW
CSB
DON'T CARE
DON'T CARE
DON'T CARE
SCLK
SDIO
R/W
A14
A13
A12
A11
A10
A9
A8
A7
A6
A5
D4
D3
D2
D1
D0
DON'T CARE
Figure 137. Timing Diagram for Serial Control Port Write—MSB First
CSB
SCLK
tDV
SDIO
SDO
DATA BIT N
DATA BIT N – 1
Figure 138. Timing Diagram for Serial Control Port Register Read—MSB First
CSB
SCLK
SDIO
DON'T CARE
DON'T CARE
DON'T CARE
DON'T CARE
A0 A1 A2 A3 A4 A5 A6 A7 A8 A9 A10 A11 A12 A13 A14 R/W D0 D1 D2 D3 D4 D5 D6 D7 D0 D1 D2 D3 D4 D5 D6 D7
16-BIT INSTRUCTION HEADER REGISTER (N) DATA REGISTER (N + 1) DATA
Figure 139. Serial Control Port Write—LSB First, Address Increment, Two Bytes of Data
tS
tC
CSB
tCLK
tHIGH
tLOW
tDS
SCLK
SDIO
tDH
BIT N
BIT N + 1
Figure 140. Serial Control Port Timing—Write
Table 114. Serial Control Port Timing
Parameter
Description
tDS
tDH
tCLK
tS
Setup time between data and the rising edge of SCLK
Hold time between data and the rising edge of SCLK
Period of the clock
Setup time between the CSB falling edge and the SCLK rising edge (start of the communication cycle)
Setup time between the SCLK rising edge and CSB rising edge (end of the communication cycle)
Minimum period that SCLK is in a logic high state
tC
tHIGH
tLOW
tDV
Minimum period that SCLK is in a logic low state
SCLK to valid SDIO (see Figure 138)
Rev. 0 | Page 201 of 205
AD9546
Data Sheet
The master always generates the stop condition to terminate
a data transfer. The SDA line must always transfer eight bits
(one byte). Each byte must be followed by an acknowledge bit.
Bytes are sent MSB first.
I²C SERIAL PORT OPERATION
The I2C interface is popular because it requires only two pins
and easily supports multiple devices on the same bus. Its main
disadvantage is its maximum programming speed of 400 kbps.
The AD9546 IꢀC port supports the 400 kHz fast mode as well as
the 100 kHz standard mode.
The acknowledge bit (A) is the ninth bit attached to any 8-bit
data byte. An acknowledge bit is always generated by the
receiver to inform the transmitter that the byte has been
received. Acknowledgement consists of pulling the SDA line
low during the ninth clock pulse after each 8-bit data byte.
To support 1.5 V, 1.8 V, and 2.5 V IꢀC operation, the AD9546
does not strictly adhere to every requirement in the original IꢀC
specification. For instance, it does not support specifications such
as slew rate limiting and glitch filtering. Therefore, the AD9546
is IꢀC compatible, but not necessarily fully IꢀC compliant.
A
The nonacknowledge bit ( ) is the ninth bit attached to any
8-bit data byte. A nonacknowledge bit is always generated by
the receiver to inform the transmitter that the byte has not been
received. Nonacknowledgement consists of leaving the SDA
line high during the ninth clock pulse after each 8-bit data byte.
After issuing a nonacknowledge bit, the AD9546 IꢀC state
machine goes into an idle state.
The AD9546 IꢀC port consists of a serial data line (SDA) and
a serial clock line (SCL). In an IꢀC bus system, the AD9546
connects to the serial bus (data bus SDA and clock bus SCL)
as a slave device. That is, the AD9546 does not generate an IꢀC
clock. The AD9546 uses direct 16-bit memory addressing
rather than 8-bit memory addressing, which is more common.
Data Transfer Process
The master initiates a data transfer by asserting a start
condition, which indicates that a data stream follows. All IꢀC
slave devices connected to the serial bus respond to the start
condition.
The AD9546 allows up to four unique slave devices to occupy
the I2C bus via a 7-bit slave address transmitted as part of an
I2C packet. Only the device with a matching slave address
responds to subsequent I2C commands. Table 103 lists the
supported device slave addresses.
The master then sends an 8-bit address byte over the SDA line,
I2C Bus Characteristics
W
consisting of a 7-bit slave address (MSB first) plus an R/ bit.
This bit determines the direction of the data transfer, that is,
whether data is written to or read from the slave device (Logic 0
indicates write, and Logic 1 indicates read).
Table 115 shows a summary of the various I2C abbreviations.
Table 115. I2C Bus Abbreviation Definitions
Abbreviation
Definition
The peripheral whose address corresponds to the transmitted
address responds by sending an acknowledge bit. All other
devices on the bus remain idle while the selected device waits
S
Sr
P
A
A
W
R
Start
Repeated start
Stop
Acknowledge
Nonacknowledge
Write
W
for data to be read from or written to it. If the R/ bit is Logic 0,
the master (transmitter) writes to the slave device (receiver).
W
If the R/ bit is Logic 1, the master (receiver) reads from the
slave device (transmitter). The read and write formats for these
commands appear in the Data Transfer Format section.
Read
Figure 141 shows an example of valid data transfer. One clock
pulse is required for each data bit transferred. The data on the
SDA line must be stable during the high period of the clock.
The high or low state of the data line can change only when the
clock signal on the SCL line is low.
Data is then sent over the serial bus in the format of nine clock
pulses, one data byte (eight bits) from either master (write mode)
or slave (read mode), followed by an acknowledge bit from the
receiving device. The protocol allows a data transfer to consist
of any number of bytes (that is, the payload size is unrestricted).
In write mode, the first two data bytes immediately after the
slave address byte are the internal memory (control registers)
address bytes (the higher address byte first). This addressing
scheme gives a memory address of up to 216 − 1 = 65,535. The
data bytes after these two memory address bytes are register
data written to or read from the control registers. In read mode,
the data bytes following the slave address byte consist of
register data written to or read from the control registers.
SDA
SCL
DATA LINE
STABLE;
DATA VALID
CHANGE
OF DATA
ALLOWED
Figure 141. Valid Bit Transfer
Figure 142 shows start and stop functionality. The start
condition is a high to low transition on the SDA line while SCL
is high. The master always generates the start condition to
initialize a data transfer. The stop condition is a low to high
transition on the SDA line while SCL is high.
When all the data bytes are read or written, stop conditions are
established. In write mode, the master device (transmitter)
asserts a stop condition to end the data transfer during the
clock pulse following the acknowledge bit for the last data byte
from the slave device (receiver).
Rev. 0 | Page 202 of 205
Data Sheet
AD9546
In read mode, the master device (receiver) receives the last data
byte from the slave device (transmitter) but does not pull SDA
low during the ninth clock pulse (a nonacknowledge bit). By
receiving the nonacknowledge bit, the slave device knows that
the data transfer is finished and enters idle mode.
Data Transfer Format
The write byte format is used to write a register address to the
RAM starting from the specified RAM address (see Table 116).
The send byte format is used to set up the register address for
subsequent reads (see Table 117). The receive byte format is
used to read the data byte(s) from RAM starting from the
current address (see Table 118). The read byte format is the
combined format of the send byte and the receive byte (see
Table 119).
The master device then takes the data line low during the low
period before the 10th clock pulse, and high during the 10th
clock pulse to assert a stop condition.
A start condition can be used instead of a stop condition.
Furthermore, a start or stop condition can occur at any time,
and partially transferred bytes are discarded.
SDA
SCL
S
P
START CONDITION
STOP CONDITION
Figure 142. Start and Stop Conditions
MSB
SDA
SCL
ACK FROM
SLAVE RECEIVER
ACK FROM
SLAVE RECEIVER
3 TO 7
8
9
3 TO 7
8
9
10
P
1
2
1
2
S
Figure 143. Acknowledge Bit
MSB
SDA
SCL
ACK FROM
SLAVE RECEIVER
ACK FROM
SLAVE RECEIVER
3 TO 7
8
9
3 TO 7
8
9
10
P
1
2
1
2
S
Figure 144. Data Transfer Process (Master Write Mode, 2-Byte Transfer)
SDA
SCL
ACK FROM
NONACK FROM
MASTER RECEIVER
MASTER RECEIVER
3 TO 7
8
9
3 TO 7
8
9
10
1
2
1
2
S
P
Figure 145. Data Transfer Process (Master Read Mode, 2-Byte Transfer), First Acknowledge from Slave
Rev. 0 | Page 203 of 205
AD9546
Data Sheet
Table 116. Write Byte Format
S
Slave
W
A
RAM address high byte
A
RAM address low byte
A
RAM
A
RAM
A
RAM
A
P
address
Data 0
Data 1
Data 2
Table 117. Send Byte Format
Slave address
S
W
A
RAM address high byte
RAM Data 0
A
RAM address low byte
A
P
P
Table 118. Receive Byte Format
Slave address
S
R
A
A
RAM Data 1
A
RAM Data 2
A
Table 119. Read Byte Format
S
Slave
address
W
A
RAM address
high byte
A
RAM address
low byte
A
Sr Slave
R
A
RAM
Data 0
A
RAM
Data 1
A
RAM
Data 2
A
P
address
I²C Serial Port Timing
SDA
tLOW
tR
tSU; DAT
tBUF
tHD; STA
tR
tF
tSP
tF
SCL
tSU; STO
tSU; STA
tHD; STA
tHIGH
tHD; DAT
S
Sr
P
S
Figure 146. I²C Serial Port Timing
Table 120. IꢀC Timing Definitions
Parameter
Description
tBUF
Bus free time between stop and start conditions
Repeated hold time start condition
Repeated start condition setup time
Stop condition setup time
Data hold time
Data setup time
tHD; STA
tSU; STA
tSU; STO
tHD; DAT
tSU; DAT
tLOW
SCL clock low period
tHIGH
SCL clock high period
tR
tF
Minimum/maximum receive SCL and SDA rise time
Minimum/maximum receive SCL and SDA fall time
tSP
Pulse width of voltage spikes that must be suppressed by the input filter
Rev. 0 | Page 204 of 205
Data Sheet
AD9546
OUTLINE DIMENSIONS
DETAIL A
(JEDEC 95)
7.10
0.30
0.25
0.18
7.00 SQ
6.90
PIN 1
INDICATOR
AREA
PIN 1
IONS
INDICATOR AR EA OP T
(SEE DETAIL A)
37
36
48
1
0.50
BSC
5.70
EXPOSED
PAD
5.60 SQ
5.50
24
13
0.50
0.40
0.30
0.20 MIN
BOTTOM VIEW
5.50 REF
TOP VIEW
END VIEW
0.80
0.75
0.70
FOR PROPER CONNECTION OF
THE EXPOSED PAD, REFER TO
THE PIN CONFIGURATION AND
FUNCTION DESCRIPTIONS
0.05 MAX
0.02 NOM
COPLANARITY
SECTION OF THIS DATA SHEET.
0.08
SEATING
PLANE
0.203 REF
COMPLIANT TO JEDEC STANDARDS MO-220-WKKD-4.
Figure 147. 48-Lead Lead Frame Chip Scale Package [LFCSP]
7 mm × 7 mm Body and 0.75 mm Package Height
(CP-48-13)
Dimensions shown in millimeters
ORDERING GUIDE
Model1
Temperature Range
Package Description
Package Option
AD9546BCPZ
AD9546BCPZ-REEL7
AD9546/PCBZ
−40°C to +85°C
−40°C to +85°C
−40°C to +85°C
48-Lead Lead Frame Chip Scale Package [LFCSP)
48-Lead Lead Frame Chip Scale Package [LFCSP]
Evaluation Board
CP-48-13
CP-48-13
1 Z = RoHS Compliant Part.
I2C refers to a communications protocol originally developed by Philips Semiconductors (now NXP Semiconductors).
©2020 Analog Devices, Inc. All rights reserved. Trademarks and
registered trademarks are the property of their respective owners.
D23266-10/20(0)
Rev. 0 | Page 205 of 205
相关型号:
©2020 ICPDF网 联系我们和版权申明