VP2611CGGH1R [MITEL]
H.261 Encoder; H.261编码器型号: | VP2611CGGH1R |
厂家: | MITEL NETWORKS CORPORATION |
描述: | H.261 Encoder |
文件: | 总14页 (文件大小:153K) |
中文: | 中文翻译 | 下载: | 下载PDF数据表文档文件 |
VP2611
H.261 Encoder
Supersedes June 1996 edition, DS3487 - 4.0
DS3487 - 4.1 December 1998
FEATURES
DESCRIPTION
TheVP2611VideoCompressionSourceCoderformspart
of a chip set used in video conferencing, video telephony and
multimedia applications. It produces data which conforms to
the H261 standard for video compression with rates between
64K and 2M bits per second. With a 27 MHz clock the device
will accept data produced to full CIF resolution at 30 Hz frame
rates. The pipeline latency through the device is only 3 macro
block periods.
The VP2611 contains all the elements necessary for the
compression algorithm. It incorporates a Motion Vector Esti-
mator which performs a +/- 7 pixel search. The decision to use
interorintraframecompressionismadebythedevice,andthe
selected data blocks are read from the frame store. New or
difference data is then passed through a Discrete Cosine
Transformer and quantized. Data from the quantizer is also
inverse quantized and passed through an Inverse Discrete
Cosine Transformer. This re-constructed data is then written
to the frame store for use in the next frame period.This frame
store is managed by an internal DRAM controller, and no
external logic is needed.
The input data must be in YUV space, and must also
conform to the six sub blocks per macro block format defined
by H261. Any conversion from RGB format is performed by
the VP510 Colour Space Converter. Any reduction in spatial
resolution, down to CIF or QCIF requirements, is done by the
VP520 Three Channel Video Filter.
The quantized data is zig-zag scanned and run length
coded before being output, together with block information
and motion vectors.
■
■
■
■
■
Fully integrated H261 video encoder
Up to full CIF resolution and 30 Hz frame rates
Inputs YUV data in 8 x 8 sub block format
Outputs run length coded coefficients
On chip motion vector estimator with +/-7 pixel search
window
■
■
Addresses and control generated internally for DRAM
frame store
QFP package
ASSOCIATED PRODUCTS
■
■
■
■
■
VP510 Colour Space Converter
VP520S CIF/QCIF Converter
VP2612 Video Multiplexer
VP2614 Video Demultiplexer
VP2615 H.261 Decoder
SYSTEM
CONTROLLER
USER
INTERFACE
VIDEO
SYNC
R
VP510
COLOUR SPACE
CONVERTER
G
B
Y
RLC DATA
VP2612
VIDEO
MULTIPLEXER
H261
BIT
STREAM
64kb to 2Mb/s
VP520
3 CHANNEL
VIDEO FILTER
VP2611
INTEGRATED
VIDEO ENCODER
REQYUV
FRMIN
FLAGS
Cr/Cb
MBLK'S
NTSC
PAL
COMP VIDEO
DECODER
ADDR
DATA
CIF FRAME
STORE
16 X128K
CIF FRAME
STORE
16X128K
TX BUFFER
32K X 8
CCIR601 RESOLUTION
CIF RESOLUTION
Y
Y
720 X 288
720 X 240
Cr/Cb 360 x 288
Cr/Cb 360 x 240
PAL
Y 352 X 288
Cr/Cb 176 x 144
NTSC
Fig 1 : Typical Video Conferencing Transmission System
VP2611
PIN DESCRIPTIONS
R/W1
R/W2
Read/Write control for external DRAM 1.
YUV7:0
PCLK
This input bus accepts YUV data one pixel at a
time from the preprocessor, clocked in on the
rising edge of PCLK.
Read/Write control for external DRAM 2.
N/C if 256k DRAMs.
OE1
OE2
Output Enable control for external DRAM 1
or ADR8.
This signal is used to strobe in data at the YUV
port and must be derived by dividing SYSCLK
with an integer greater than one.
Output Enable control for external DRAM 2.
N/C if 256k DRAMs.
FRMIN
This input should be pulled high to prepare the
VP2611 to code a new frame. It must be held
high for at least one SYSCLK cycle and then
must be pulled low again before the next frame
begins. The VP2611 will respond to the rising
edge of FRMIN by asserting REQYUV
ADR7:0
Address output for the external DRAMs.
CBUS7:0
Bi-directional data bus for use by a Microproce-
ssor. Data and insructions are clocked on and
off the chip on the rising edge of CSTR.
appproximately 184 SYSCLK cycles later.
CSTR
Data strobe for the CBUS port.
REQYUV
DBUS7:0
This output is pulled high to request that YUV
data be input for a new MacroBlock. It is pulled
low again 1871 SYSCLK cycles later. It re-
mains low during Dummy MacroBlocks and
during the lay period between frames.
CEN
An enabling signal for the CBUS port.
CADR
When high, this signal defines CBUS as a data
bus, and when low as an instruction input.
This output bus serves several functions as
defined by DMODE3:0. In addition to providing
the quantized coefficients and motion vectors,
it is used to output control information.
SYSCLK
System clock, run at 27MHz maximum. The
clock must be high for between 35% and 65%
of each clock cycle. This clock is used for all
internal operations.
DMODE3:0
DCLK
Output flag port for DBUS7:0 bus. The value at
this port identifies the data type appearing on
DBUS7:0 during the same period.
RESET
Active low power on reset which must be held
low for at least 2064 cycles.
TCK
TMS
TDI
Test clock for JTAG.
This output pulses high for a minimum of 37ns
each time new data is output on DBUS or
DMODE. It can be used as an edge sensitive
strobe signal or a level sensitive "valid" signal.
Test Mode Select for JTAG.
Input JTAG test data.
Output JTAG test data.
SW15:0
This bidirectional port is connected to the
frame store.
TDO
TRST
Reset JTAG controller (active low).
RAS
CAS
Row Address Strobe output for the external
DRAMs.
NOTE:
"Barred" active low signals do not appear with a bar in the
main body of the text.
Column Address Strobe output for the external
DRAMs.
FORWARD PATH
YUV
BLOCK
FORMAT
DCT
RLC
Motion Vectors
Q
Zig Zag
SUB
Q Step
IQ
INTER/INTRA
DECISION
PROCESSOR
Force
Intra
DATA
BUS
IDCT
LOW
PASS
FILTER
MOTION
VECTOR
ESTIMATOR
BUS
FLAGS
ADD
Force
Intra
CONTROL
LOGIC
Block Info
Force
Filter
Search
Window
Predicted
block
Force
Filter
FRAME STORE INTERFACE
DATA
HOST DATA & CONTROL
ADDRESS
CONTROL
Fig 2 : Simplified Block Diagram
2
VP2611
OPERATION OF MAJOR BLOCKS
Motion Vector Estimator
180
140
100
x = 1.125y
MC Off
The motion estimator calculates the mean absolute error
( MAE ) for each possible position of the combined luminance
block in a search window from the previous frame. The
combined luminance block consists of 16 x 16 pixels, and in
thesearchwindowthisisdisplacedbetween-7to+7vertically,
and -8 to +7 horizontally. The two lsb's of each pixel are
discarded and the MAE value is contained within 14 bits.
The minimum MAE value, representing the best match
between the previous and current block, is passed to the
motion compensation decision block, together with the posi-
tion of this best fit in the search window. The zero displace-
ment MAE value is also passed to this block, which then
decides whether the best fit is sufficiently better than the zero
displacement fit. It uses the characteristic shown in Figure 3,
where the 14 bit MAE is a Hex value. In the area to the right
of the line all points defined by the two MAE values will cause
motion compensation to be applied. In this case the best fit
MAE value is used by the inter/intra decision processor,
otherwise the zero displacement value is used.
CO
80
AB
5F
MC On
40
20
40
80 CO 100 140 180
Zero Movement Absolute Error in Hex
Fig 3 : MC Decision Slope
counter will be disabled.
The user may overide the internal Inter/Intra decision at
any time using the CBUS control port. A user generated
forced inter mode will overide an internally generated `Force
Intra'.
Inter/Intra Decision Processor
The MAE value passed by the motion compensation
decision block is compared to the simplified variance of the
current block. This simplified variance is calculated by sum-
ming the moduli of the differences between each luminance
pixel and the mean luminance value over the whole macrob-
lock. Eight bit pixels are used, and the variance value is
expressedin14bitsbydiscardingthetwolsb'sfromtheactual
16 bit result. It can then be directly compared to the 14 bit MAE
value.
If the MAE value is below a user defined threshold inter
modecodingisalwaysselected. Thedefaultthresholdis3, on
a scale from 0 to 255 using the 8 msb's from the 14 bit value.
Abovethisthresholdintermodeisonlyselectedifthevariance
of the current block is greater than or equal to the MAE value
in use.
Low Pass Filter
The macroblock selected from the previous frame in
motion compensated inter mode coding, will be filtered before
it is subtracted from the current block. This decision can be
overidden externally by the system controller. The Filter uses
a simple [ 1 2 1 ] characteristic in both vertical and horizontal
dimensionsasspecifiedinH.261onthemacroblockboundaries
[010] is used.
In order to avoid gradual picture degredation, every 61st
Macroblock input to the VP2611 is coded in intra mode
regardlessoftheabovedecision.As61isaprimenumber,this
will ensure that each macroblock will be transmitted in intra
mode at least once in every 61 transmissions. If FIX MAC-
ROBLOCK or SKIP PICTURE is invoked this `Force Intra'
SYMBOL
t RAC
PARAMETER
MINIMUM
-
MAXIMUM
Access time from RAS
105ns or under
t CAC
t RP
t CP
Access time from CAS
RAS precharge time
CAS precharge time
-
25ns or under
50ns or under
15ns or under
-
-
t RAS
t CAS
t REF
RAS pulse width
CAS pulse width
Time to refresh 256 rows
90ns or under
50ns or under
-
-
-
0.25ms or over
N.B. All times are quoted assuming 27MHz operation. For lower clock
frequencies increase the above values proportionately.
Table 1 : External DRAM timing requirements
3
VP2611
Frame Store Manager
Zig Zag Scan
The previous picture is stored in an external CIF DRAM
frame store, which is connected by a glueless interface. The
internal Frame Store Manager controls all read, write, and
refresh operations to these DRAMs. No provision is made to
allow the use of smaller DRAM's, if only QCIF operation is
required.
This is essentially an address generator which reorders
the DCT coefficients according to the standard zig-zag scan
pattern. This has the effect of concentrating the significant
coefficients at the beginning of the sub-block, improving the
efficiency of the Run Length Coder.
During the coding of each macroblock columns of the
search window are read from these DRAMs, and finally the
"best fit" macroBlock is obtained. At the completion of coding
the fully processed new macroblock is written to the DRAM's,
after it has been decoded again. In this way the frame store
maintains a bit-accurate duplicate of the image seen by the
Decoder (excepting transmission errors).
Several configurations are possible to make the required
128Kx16 store. Two 64K x 16 DRAMs could be employed; in
this case use the default 1M DRAM mode when setting up the
chip. Otherwise, a single 256K x 16 DRAM or four 256K x 4
DRAMs could be used. In these last two cases use OE1 as
ADR8, RW1 as R/W, and do not connect RW2 and OE2. Also,
use the Setup instruction at the CPORT to put the device into
4M DRAM mode.
Run Length Coder
Each coefficient output from the zig zag scan is examined.
If it is non-zero, then the Run Length Coding circuit will pass
the coefficient magnitude to the output port along with its zero
count i.e. the number of zero magnitude coefficients preced-
ing it within the same 8x8 sub-block.
Inverse Quantize
This circuit replicates the operation of the inverse quan-
tizer in the decoder. It reconstructs the 12 bit DCT coefficients
from the 8 bit quantized inputs, using the 5 bit quantization
value. This is achieved using the following formulae.
Table 1 details the critical timing parameters which the
external DRAM must meet with SYSCLK running at 27MHz.
Note that, if used at slower speeds, the requirements on the
DRAM timing are relaxed with the exception of refresh. The
number of refresh cycles the VP2611 produces is directly
proportional to the SYSCLK frequency.
If QUANT is odd :
REC = QUANT*(2*LEVEL+1) : LEVEL > 0
REC = QUANT*(2*LEVEL-1) : LEVEL < 0
If QUANT is even :
REC = QUANT*(2*LEVEL+1)-1 : LEVEL > 0
REC = QUANT*(2*LEVEL-1)+1 : LEVEL < 0
Discrete Cosine Transform
For Intra Coded DC Coefficients :
REC = 8*LEVEL
except if LEVEL=255 when REC=1024
ThiscircuitperformsaDiscreteCosineTransformoneach
8x8 sub block, whether in inter or intra mode. In intra mode,
eightbitpixeldata isused,withaninthimpliedsignbit(allpixel
data is positive ). In inter mode the difference between the
current and best fit previous block is used. This will be a two's
complement number. Twelve bit coefficients are produced by
the DCT, and passed on to the quantizer.
If LEVEL=0 then REC=0 in all cases.
The reconstructed values (REC) are passed through a
Clipping Circuit in case of arithmetic overflow.
Thus, the Inverse Quantizer restores the DCT coefficients
to their original value but with quantisation error.
Quantize
This section quantizes the results of the DCT by dividing
the 12 bit output from the DCT with a host supplied value. The
5bitquantizationvaluesuppliedcorrespondstodivisionofthe
12 bit coefficients ( range ± 2048 ) by values from 2 to 62, but
in steps of 2. This variable quantization strategy allows the
volume of data generated by the encoder to be adjusted
dynamically, depending on the fullness of the transmission
buffer. For H.261 applications it uses the quantisation value
provided at the control port during the previous Macroblock
period (or at some earlier time). An option is provided which
allows two quantisation values to be used, one for use with
inter coded macroblocks, and the other for use with intra
coded macroblocks.
As specified in H.261, the DC coefficient of an Intra coded
Block is treated differently and the 12 bit value is always
divided by 8.
When the quantization value is small, and the DCT coef-
ficient is large, there is a danger of overflow in the eight bit
output. To avoid this a clipping circuit is included at the output
of the quantizer, which saturates at the maximum values.
Inverse DCT
This circuit replicates the operation of the Inverse Cosine
Transform in the Decoder, and outputs 9 bit signed pixel data
(intra mode) or pixel difference data (inter mode). The IDCT
fully meets the CCITT specification.
Reconstruction Adder
In Inter Mode, the IDCT data is added to the best fit block
from the previous frame store. In Intra mode, the IDCT data is
simply added to zero. After the adder, the sign bit is removed
from the result to give 8 bit pixels. Clipping circuits ensure that
any pixels with values exceeding 255 are clipped to 255, and
any with negative values are clipped to zero (such values are
possible due to quantization noise).
4
VP2611
2064 cycles
MB2
SUBBLOCK ORDER WITHIN MACROBLOCK
MB1
MB3
MB2
MB4
MB3
MB3
MB1
MB1
YUV Input
Frame Store Read
DUMMY
DUMMY
DUMMY
DUMMY
MB1
1
3
2
4
5
6
Control Decisions
Frame Store Write
DBUS Output
MB1
MB2
DUMMY
DUMMY
DUMMY
DUMMY
U
V
Y
Fig 4: MacroBlock Pipelining
OPERATION OF INTERFACES
Macroblock Delays
PIXEL ORDER WITHIN SUBBLOCK
00 01 02 03 04 05 06 07
08 09 10 11 12 13 14 15
The VP2611 has a three macroblock pipeline delay be-
tween pixel inputs and run length coded outputs. This is
illustrated in Figure 4. Whilst the second macroblock is being
input, the best fit macroblock from the previous frame is being
identified and then read from the frame store. At this time any
ControlDecisionswhicharetoeffectthefirstmacroblockmust
be supplied by the host controller. The run length coded
outputs for the first macroblock are not available until the
fourth macroblock is supplied at the input pins.
16 17 18 20 21 22 23
19
24 25 26 27 28 29 30 31
32 33 34 35 36 37 38 39
40 41 42 43 44 45 46 47
48 49 50 51 52 53 54 55
56 57 58 59 60 61 62 63
Fig 5 : Ordering of Pixels
then theoretically the average rate need only be 384/1871
times the SYSCLK rate. Note that PCLK must always be
obtained by dividing SYSCLK by an integer greater than one.
WhentheVP520CIF/QCIFConverterissupplyingtheVP2611
with data, it provides a peak PCLK rate equivalent to SYSCLK
divided by two, and an average rate of SYSCLK divided by
four.
The mimimum gap between REQYUV going active is
2064 SYSCLK periods. In full CIF mode "dummy" macrob-
locks are internally inserted between rows, in order to give the
chip sufficient time to load a new search window. No new YUV
data must be loaded during these dummy macroblocks, and
REQYUV will remain inactive. No dummy macroblocks are
required in QCIF mode. With a 27MHz SYSCLK all macrob-
locks will be coded in less than a 30Hz frame rate period, and
there will be a period of inactivity before FRMIN goes active
again. During this period the output bus will remain static at all
ones, and no output strobe ( DCLK ) will be produced.
YUV Input Port
The YUV port accepts pixel data from the preprocessor in
block format as illustrated in Figure 5. Within a complete
system the VP2611 is always the master device, and must be
supplied with macroblock data when it makes a demand. The
orderinwhichpixelsaresuppliedispre-determined, andmust
be strictly maintained. There are 64 pixels per sub-block and
4 luminance and 2 chrominance sub-blocks per macroblock.
Themacroblocksthemselvesaredividedintogroupsofblocks
( GOB's ), and the sequence specified in H.261 must also be
maintained.Notethat,sincethechrominanceresolutionishalf
the luminance resolution both vertically and horizontally, then
the two chrominance blocks cover the same picture area as
the four luminance blocks.
The pre-processor producing macroblock data must pro-
duce a frame start signal ( FRMIN ) when it has a complete
frame of data available. This resets the input controller within
the VP2611, which will then generate sequential GOB and
macroblock numbers for the coded outputs referenced to this
input.
FRMIN must go high for at least one system clock period,
and must go low before the next frame is available. The
VP2611 responds to FRMIN with a request for macroblock
data ( REQYUV ), which occurs approximately 184 SYSCLK
periods after FRMIN. It must then receive a complete macrob-
lock within 1871 SYSCLK periods, and at the end of this time
REQYUV will go inactive. The VP2611 must be provided with
a PCLK signal to strobe in the data. This must be derived from
SYSCLK, and must only be present when there is valid data
at the input. Data must meet the set up and hold times with
respect to PCLK as specified in Figure 6.
SCLK/2
20ns
PCLK
20ns
0ns
10ns
YUV7:0
N.B. All timings given are MINIMUM values.
Fig 6 : Timing at YUV Port
The maximum peak rate for PCLK is the SYSCLK rate
divided by two, but since there are 384 bytes per macroblock
5
VP2611
DBUS Output Port
START MB
WAIT
The DBUS port is used to pass data and control informa-
tion directly to the VP2612 Video Multiplexer. The type of data
on the output pins is identified by the DMODE 3:0 outputs,
using the codes shown in Table 2. An output strobe is also
produced ( DCLK ) which always goes high one system clock
period after the data defined by DMODE 3:0 becomes valid.
This edge is used to strobe the data into the Video Multiplexer,
and thus the data set up time is always one SYSCLK period
minus differential output delays.
(2 cycles)
IS IT
A DUMMY
BLOCK?
yes
no
CONTROL
(2 cycles)
(2 cycles)
(2 cycles)
(2 cycles)
The number of SYSCLK periods during which data re-
mains valid is dependent on the type of data, and DCLK
remains high for this same period. It goes low as the result of
the same SYSCLK rising edge which produces a change in
DMODE 3:0. The output delays with respect to SYSCLK are
illustrated in Figure 8, and Figure 9 shows a typical output
sequence during which DCLK remains high for several cycles
as the sub-block number ( code 7 ) is produced. During a Wait
State
GOB
MB
CBP
QUANT
(2 cycles)
(2 cycles)
(2 cycles)
( code 15 ) no DCLK transitions are produced. The actual
sequence of output events which occur for each macroblock,
and the duration of each event, are illustrated in Figure 7.
HORZ MV
VERT MV
The output events are defined in more detail below;
Control Decisions : This byte shows which control decisions
have been taken for the forthcoming macroblock. DBUS0
will be high if a Fixed Macroblock (FIX MB) was enforced
i.e. no new data will be transmitted this macroblock.
DBUS1 indicates whether Inter (high) or Intra (low) coding
was used for the macroblock. DBUS2 will be high if the
macroblock was filtered, and DBUS3 will be high if motion
compensation was used. DBUS5 will be high if the current
frame is being coded in FAST UPDATE mode. In this
mode the complete frame will be intra coded. DBUS6 will
be high if the current frame is a SKIP FRAME i.e. not being
coded - so no coefficients will be transmitted. DBUS4 and
DBUS7 are not used.
ARE
no
ANY BLOCKS
CODED?
yes
WAIT
(32 cycles)
SUB BLK NO (15 cycles)
RUN LENGTH
(2 cycles)
(2 cycles)
MAGNITUDE
WAIT
(1 cycle)
DMODE3:0
FUNCTION
ARE
ALL COEFFS
O/P?
GOB Number
MB Number
Control Decisions
Quant Value
Horizontal MV
Vertical MV
Coded Blk Pattern
Sub-Block No
Zero Run Count
RLC Coefficient
Not used
0000
0001
0010
0011
0100
0101
0110
0111
1000
1001
1010
1011
1100
1101
1110
1111
no
yes
(wait variable time to make total
time since start of sub-block up
to 335 cycles)
WAIT
ARE
ALL BLOCKS
O/P?
no
yes
Not used
Not used
Not used
Not used
WAIT
(variable cycles)
END MB
Wait State
Fig 7 : DBUS Port Flow Chart
Table 2 : DBUS Functions
6
VP2611
CBUS3:0
INSTRUCTION
SCLK
DCLK
Input VAR Threshold
Reserved
Input Inter Quantiser
Input Intra Quantiser
Input Setup Data
Input Control Functions
Reserved
0000
0001
0010
0011
0100
0101
0110
0111
1000
1001
1010
1011
1100
1101
1110
1111
20ns max
25ns max
20ns max
DATA FROM
VP2611
DATA VALID
DMODE
3:0
DATA VALID
Reserved
Output GOB Number
Output MB Number
Reserved
Output Control Decisions
Output Setup Data
Reserved
Fig 8: Timing diagram
GOB Number : At the start of each new macroblock, the
current GOB Number is output on DBUS3:0. (DBUS3 is
MSB).
Reserved
Overide internal clock doubler
MB Number : After the GOB Number, the macroblock
Number is output on DBUS5:0 (DBUS5 is MSB).
Table 3 : CBUS Instruction Codes
Coded Block Pattern : This byte contains a 6 bit linear code
that indicates which of the sub-blocks actually contain
coded data. DBUS6 will be high if sub-block 1 contains
coded data, through to DBUS 1 being high if sub-block 6
contains coded data. DBUS7 and DBUS0 are not used.
Note that if the macro block is not motion compensated
and the coded block pattern is all zero's, the fixed macro
block bit will be set in the control decisions byte.
Sub-block Number : An identifier for the run length coded
coefficients which are about to be made available. DBUS
2:0 contain the coded sub-block number from 1 to 6. All
zero sub-blocks will not be produced at the outputs, and
their corresponding numbers will not appear.
Zero Run Count : The number of zero valued coefficents
preceding the next non zero coefficient is available on
DBUS5:0 (DBUS5 is MSB). Normally, DBUS7:6 are low,
except to signify the end of a Sub-block, when they will
both be high. Zero Run Count is always followed by a
coefficient, even at the end of a sub-block.
Quant Value :The quantisation value used in processing the
current macroblock is output on DBUS4:0 (DBUS4 is
MSB). This represents an actual quantisation level be-
tween 2 and 62, in steps of 2 and as defined in H.261.
RLC Coefficient : This byte contains the 8 bit coefficient value.
It will always be a non-zero value, except when the
previous Zero Run Count signalled the end of sub-Block.
A zero value is then possible since, as stated above, the
run count is always followed by a coefficient byte, which
may be zero if the last coefficient is zero.
Horizontal MV : If motion compensation is used, the horizontal
componentofthemotionvectorwillbeoutputon DBUS4:0
(DBUS4 is MSB). This 5 bit value represents a two's
complement number in the range +/-15
(although only vectors in the range -8 to +7 are currently
possible with the VP2611).
Wait State : This indicates that no valid data is being output
from the DBUS port during this cycle. No DCLK is pro-
duced for this state.
Vertical MV : If motion compensation is used, the vertical
componentofthemotionvectorwillbeoutputon DBUS4:0
(DBUS4 is MSB). This 5 bit value represents a two's
complement number in the range ±15 ( although only
vectors in the range ±7 are currently possible with the
VP2611).
Pins which are "not used" for certain functions will be
forced low.
This diagram shows a typical Sub-block being output from the VP2611.
DCLK
DMODE
DBUS
15
X
7
2
8
0
9
15
X
8
9
15
8
9
0
15
X
4
1
-2
X
252
Both msb's are high showing end of block.
Fig 9: DBUS Timing
7
VP2611
tion is given in Figure 10, and when writing data or instructions
to the VP2615 the set up and hold times which are referenced
to the rising edge of CSTR must be maintained.
CBUS Control Port
The CBUS control port is used to input control and setup
information and also to output status information. In order to
save on pin count, a microprocessor driving this port is
required to execute two I/O instructions in order to transfer a
single byte of information to or from the device. The first
transfer is always a write operation, with a low level on the
single address line which is used by the interface. Data on the
bus then defines the instructions listed in Table 3. The second
transfer can be a read or write operation as necessary, but the
address line must then be high with the set up time given in
Figure 10.
When a write instruction has been defined CADR should
be pulled high, valid data presented to CBUS7:0 and then
strobed in using CSTR. Other system I/O transfers can occur
between defining a write operation and supplying the data to
be written, assuming CEN is not active during those other
transfers. If CSTR does not go active because of I/O transfers
to other devices, then CEN can remain active low between the
instruction and data.
When a read instruction has been specified the requested
data will then be output on CBUS7:0 after the access time
specified from CEN going low, assuming that CADR was
already high. Otherwise the data will become valid after the
access time specified from CADR going high after CEN was
In addition to the single addresss line (CADR), data
transfers use a control strobe (CSTR) which is only effective
when a chip enable is present (CEN). Detailed timing informa-
WRITING DATA FROM THE CBUS:
This diagram shows a typical instruction and associated data field being written to the device.
10ns
10ns
CEN
20ns
10ns
20ns
CADR
CSTR
20ns
20ns
10ns
10ns
20ns
20ns
20ns
20ns
10ns
20ns
CBUS
I/P
INSTRUCTION
DATA IN
READING INFORMATION ON CBUS :
This diagram shows a typical instruction and associated data field being read from the device.
10ns
10ns
CEN
Th
10ns
20ns
CADR
CSTR
20ns
20ns
10ns
20ns
10ns
20ns
20ns
20ns
*
20ns
*
20ns
50ns
10ns
*
20ns
CBUS
INSTRUCTION
DATA OUT
If Th is less than 5 ns then CBUS may be driven by the VP2615until CEN going high eventually turns off
the drivers. It will not prevent correct data being read when CEN again goes active
N.B. All timings shown are minimum values except those marked * which are maximums.
Fig 10 : Use of the Control Port
8
VP2611
Decision circuitry will be overidden according to CBUS1;
if CBUS1 is HIGH then all subsequent macroblocks will be
intra coded, if it is LOW they will be inter coded. When
CBUS2 is HIGH the on-board Filter Decision circuitry is
overiddenaccordingtoCBUS3;ifCBUS3isHIGHthenthe
filter will be forced on, if it is LOW the filter will be forced off.
If CBUS4 is HIGH then FIX MB will be implemented, and
no new data from the current macroblock will be coded. A
two macroblock delay exists between defining the Force
Inter/Intra, Force Filter or FIX MB decisions through the
control bus and data being affected at the outputs. These
decisions will stand for all subsequent macroblocks until
they are again changed. If CBUS5 is HIGH a FAST
UPDATE will be performed on the next frame and all
blocks will be coded in intra mode. If CBUS6 is HIGH then
the next frame will not be transmitted ( SKIP FRAME ).
Notethatthesetwoglobalframebitsdonottakeeffectuntil
the start of the next frame, and stay in effect for all frames
until they are removed. If CBUS7 is HIGH, then the on-
board Force Update Controller will be overidden, and the
user will have to enforce their own Force Update policy
using the Force Intra command. RESET will cause the
options to default to those defined by the LOW state. Note
that SKIP FRAME has priority over any other bits and that
FIXMB has priority over all bits bar SKIP FRAME. See
note below.
low. Note that in the data read phase CADR must always go
high before CSTR goes high, with the set up time specified.
When CEN goes high, or CADR goes low, the CBUS will go
high impedance after the delay specified.
Note that the access times under the conditions given
above are only true when the gap between CSTR going high
in the instruction phase, and CEN going low in the data phase,
is greater than the minimum specified in figure 10.
Only the four LSBs, CBUS3:0, are used when writing
instructions to the VP2611. The remaining bits, CBUS7:4,
should be pulled low while the instruction is strobed into the
VP2611.
The instructions listed in Table 3 are described below in
greater detail;
Input VAR Threshold: VAR is the difference between the best
fit MAE value and the variance of the current macroblock.
The VAR Threshold is the best fit MAE value below which
Inter Frame Prediction is always used, no matter what the
variance of the current block. Above this threshold inter
mode coding is only used if the best fit value is less than
the current block variance. The default value is 3, within a
range of 0 - 255 using the eight most significant bits of the
14 bit value. In normal operation values below 15 should
be used.
Output GOB Number: This instruction will output the GOB
Number on CBUS3:0, for the data currently being output
on DBUS. CBUS7:4 are not used (always low).
Input Inter Quantiser:Coefficents of inter coded macroblocks
will be quantized using the value on CBUS4:0 following
this instruction. Internally this represents a 6 bit number
with the lsb always zero, giving a value between 0 and 62
in steps of two. Where only one quantization value is to be
used for both inter and intra cases, this instruction should
be used. On reset the value will default to the maximum
allowed. See note below.
Output MB Number: This instruction will output the macrob-
lock number on CBUS5:0, for the data currently being
output on DBUS. If CBUS6 is low it indicates that the
macroblock number has just changed, or is about to
change. New Quantization Value or Control Function
words should not be written at this time since it is uncertain
which macroblock they will effect. CBUS7 is not used
(always low).
Input Intra Quantiser: This instruction is similar to the above,
except that it defines the quantization level for intra mode
codingwhenitistobedifferenttothatofintermodecoding.
See note below.
Output Control Decisions: This instruction will output the
details of several control decisions on the CBUS. CBUS0
shows whether the MacroBlock currently being output on
DBUS was inter or intra coded (0=Intra). CBUS1 shows
whether motion compensation was used (1=MC used).
CBUS3 shows whether the macroblock was passed
through the loop filter or not (1=Filtered). CBUS4 will be
high if the FIX MB instruction was enforced. CBUS5 will be
high if FAST UPDATE is currently being undertaken.
CBUS6 will be high if SKIP FRAME is in force. CBUS2 and
CBUS7 are not used.
Input Setup Data: This instruction allows several user defined
optionstobespecified,usingindividualbitsinthefollowing
data word. If CBUS0 is LOW the device will work in full CIF
mode , if HIGH it uses the QCIF mode. If CBUS3 is HIGH
both inter and intra quantization values will be used,
otherwise a common value will be used. If CBUS5 is high
then the motion compensation circuits will be disabled. If
CBUS6 is high, then the device will be configured to use
256Kx16or256Kx4DRAM's,otherwiseitwillassumethe
use of two 64K x 16 DRAM's. The default conditions after
RESET are those selected by the Low level. CBUS1,
CBUS2, CBUS4 and CBUS7 are not used but must be low
during the definition phase. This instruction may be used
any time after RESET has gone high, but the video input
bus must not be active. If a subsequent mode change
between CIF and QCIF is made then a further RESET is
needed.
Output Setup Data: This instruction allows the user to verify
the internal setup previously selected. The bits have the
same significance as in the Input Setup Data Instruction.
Note
For definitive operation the output MB number should be
read first, and these bytes only changed if CBUS b is high.
Input Control Functions: This instruction specifies several
control options using individual bits in the following data
word. If CBUS0 is HIGH then the on board Inter/Intra
9
VP2611
Initialising the VP2611
On power-up, RESET should be low and must remain low
for at least 2064 cycles of SYSCLK. After RESET is pulled
high, FRMIN may be activated to start the first frame. Before
activating FRMIN for the first time, it is advisable to use the
CBUS to implement a FAST UPDATE for the first frame (i.e.
all blocks Intra coded).
Instructions are clocked into the 8 bit instruction register
(no parity bit) and the following are available.
Instruction Register Name
( MSB first )
11111111
00000000
01000000
XX001011
BYPASS
EXTEST (No inversion)
INTEST
JTAG Test Interface
SAMPLE/PRELOAD
The VP2611 includes a test interface consisting of a
boundary scan loop of test registers placed between the pads
and the core of the chip. The control of this loop is fully JTAG/
IEEE 1149-1 1990 compatible. Please refer to this document
for a full description of the standard.
The interface has five dedicated pins: TMS, TDI, TDO,
TCK and TRST. The TRST pin is an independent reset for the
interface controller and should be pulsed low, soon after
power up; if the JTAG interface is not to be used it can be tied
low permanently. The TDI pin is the input for shifting in serial
instruction and test data; TDO the output for test data. The
TCK pin is the independent clock for the test interface and
registers, and TMS the mode select signal.
TimingdetailsfortheJTAGcontrolsignalsareshowninfig
11.The maximum TCK frequency is 5 MHz.
The test registers, their positions in the boundary loop and
thecorrespondingi/opadaredetailedinTable4. Notethatthe
three state control signals also have test registers associated
with them which are labelled as TRI in Table 4. DHz is an
output enable for all signals to the DRAM. The order given in
Table 4 determines the serial data stream needed for JTAG
testing.
TDI and TMS are clocked in on the rising edge of TCK, and
all output transitions on TDO happen on its falling edge.
Pad
Type Reg No Pad
Type Reg No Pad
Type Reg No
TCK
RESET
CADR
DBUS5
DBUS6
DBUS7
SW15
OP
62
IN
IN
IN
93
92
91
90
89
88
87
86
85
84
83
82
81
80
79
78
77
76
75
74
73
72
71
70
69
68
67
66
65
64
63
31
30
29
28
27
26
25
24
23
22
21
20
19
18
17
16
15
14
13
12
11
10
9
OP
OP
OP
TRI
IN
SW1
SW0
OP
IN
61
60
59
58
57
56
55
54
53
52
51
50
49
48
47
46
45
44
43
42
41
40
39
38
37
36
35
34
33
32
CSTR
CEN
IN
Signal
IN
OP
IN
CBUS0
OP
TRI
Tsu
Thd
DHZ
RAS
TRI
OP
IN
SW14
SW13
SW12
OP
IN
TCK
Signal
CBUS1
CBUS2
CBUS3
OP
IN
CAS
RW1
OP
OP
OP
IN
OP
IN
RW2
OE1
OE2
OP
OP
OP
OP
IN
OP
IN
SW11
SW10
OP
ADR0
OP
Tprop
CBUS4
CBUS5
CBUS6
CBUS7
DCLK
OP
IN
IN
OP
ADR1
ADR2
OP
OP
OP
IN
IN
ADR3
ADR4
OP
OP
Tsu
15
Thd
5
Tprop
SW9
SW8
OP
TMS toTCK timing
OP
IN
IN
OP
ADR5
ADR6
OP
OP
15
TDI to TCK timing
5
ADR7
PCLK
OP
IN
OP
IN
IN
SW7
OP
IN
Chip i/p to TCK timing
15
5
OP
IN
IN
YUV7
YUV6
DMODE0 OP
DMODE1 OP
SW6
SW5
OP
IN
20
TCK to TDO timing
8
YUV5
YUV4
YUV3
IN
IN
IN
DMODE2
DMODE3
OP
OP
OP
IN
7
Fig 11 : JTAG Interface timing
6
DBUS0
DBUS1
OP
OP
SW4
SW3
SW2
OP
IN
5
YUV2
YUV1
YUV0
IN
4
IN
IN
DBUS2
DBUS3
DBUS4
OP
OP
OP
3
SYSCLK IN
2
IN
OP
IN
1
FRMIN
OP
REQYUV
OP
0
Table 4 Pin and JTAG test registers
10
VP2611
ABSOLUTE MAXIMUM RATINGS [See Notes]
Supply voltage VDD
Input voltage V
-0.5V to 7.0V
-0.5V to VDD+ 0.5V
-0.5V to VDD + 0.5V
Output voltageINVOUT
Clamp diode current per pin I (see note 2)
Static discharge voltage (HBM)
18mA
500V
K
Storage temperature T
-55°C to 150°C
S
Ambient temperature with power applied T
AMB
0°C to 70°C
125°C
Junction temperature
Package power dissipation
3000mW
NOTES ON MAXIMUM RATINGS
1. Exceeding these ratings may cause permanent damage.
Functional operation under these conditions is not implied.
2. Maximumdissipationfor1secondshouldnotbeexceeded,
only one output to be tested at any one time.
3. Exposure to absolute maximum ratings for extended
periods may affect device reliablity.
4. Current is defined as negative into the device.
STATIC ELECTRICAL CHARACTERISTICS
Operating Conditions (unless otherwise stated)
Tamb = 0 C to +70°C VDD = 5.0v ± 5%
Characteristic
Symbol
Conditions
Units
Value
Min.
Max.
Typ.
Output high voltage
Output low voltage
Input high voltage
Input low voltage
Input leakage current
Input capacitance
Output leakage current
Output S/C current
I
I
V
OH = 4mA
OL = -4mA
DD -1V for SYSCLK and PCLK
VOH
VOL
VIH
VIL
IIN
CIN
IOZ
ISC
V
V
V
2.4
-
2.0
-
-
0.4
-
0.8
+10
V
GND < V < VDD
µA
pF
µA
mA
-10
IN
10
GND < VOUT < VDD
VDD = Max
-50
10
+50
300
ORDERING INFORMATION
VP2611 CG GH1R (Commercial - Plastic QFP power package)
11
VP2611
Function
Function
Pin
Pin
Function
Pin
YUV3
NC
DCLK
NC
87
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
SW3
1
88
NC
2
YUV4
YUV5
VDD
GND
YUV6
YUV7
NC
CBUS7
VDD
89
SW4
3
90
SW5
4
CBUS6
GND
91
GND
5
92
VDD
6
VDD
93
SW6
7
CBUS5
GND
94
SW7
8
95
NC
9
PCLK
NC
CBUS4
CBUS3
CBUS2
CBUS1
NC
96
SW8
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
97
SW9
NC
98
SW10
SW11
NC
ADR7
ADR6
ADR5
VDD
GND
NC
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
GND
GND
VDD
SW12
NC
CBUS0
TRST
CEN
VDD
ADR4
ADR3
ADR2
ADR1
GND
ADR0
VDD
GND
NC
SW13
SW14
NC
NC
NC
CSTR
NC
SW15
DBUS7
DBUS6
NC
CADR
RESET
VDD
DBUS5
GND
GND
OE2
OE1
VDD
RW2
RW1
CAS
RAS
VDD
GND
NC
TCK
VDD
TMS
DBUS4
DBUS3
NC
TDI
NC
TDO
DBUS2
NC
(CLK54)
REQYUV
FRMIN
VDD
NC
DBUS1
DBUS0
DMODE3
NC
NC
SW0
SW1
SW2
NC
SYSCLK
GND
GND
NC
VDD
YUV0
YUV1
YUV2
DMODE2
DMODE1
DMODE0
NC
Pin out table for GH128 PQFP package
12
http://www.mitelsemi.com
World Headquarters - Canada
Tel: +1 (613) 592 2122
Fax: +1 (613) 592 6909
North America
Tel: +1 (770) 486 0194
Fax: +1 (770) 631 8213
Asia/Pacific
Tel: +65 333 6193
Fax: +65 333 6192
Europe, Middle East,
and Africa (EMEA)
Tel: +44 (0) 1793 518528
Fax: +44 (0) 1793 518581
Information relating to products and services furnished herein by Mitel Corporation or its subsidiaries (collectively “Mitel”) is believed to be reliable. However, Mitel assumes no
liability for errors that may appear in this publication, or for liability otherwise arising from the application or use of any such information, product or service or for any infringement of
patents or other intellectual property rights owned by third parties which may result from such application or use. Neither the supply of such information or purchase of product or
service conveys any license, either express or implied, under patents or other intellectual property rights owned by Mitel or licensed from third parties by Mitel, whatsoever.
Purchasers of products are also hereby notified that the use of product in certain ways or in combination with Mitel, or non-Mitel furnished goods or services may infringe patents or
other intellectual property rights owned by Mitel.
This publication is issued to provide information only and (unless agreed by Mitel in writing) may not be used, applied or reproduced for any purpose nor form part of any order or
contract nor to be regarded as a representation relating to the products or services concerned. The products, their specifications, services and other information appearing in this
publication are subject to change by Mitel without notice. No warranty or guarantee express or implied is made regarding the capability, performance or suitability of any product or
service. Information concerning possible methods of use is provided as a guide only and does not constitute any guarantee that such methods of use will be satisfactory in a specific
piece of equipment. It is the user’s responsibility to fully determine the performance and suitability of any equipment using such information and to ensure that any publication or
data used is up to date and has not been superseded. Manufacturing does not necessarily include testing of all functions or parameters. These products are not suitable for use in
any medical products whose failure to perform may result in significant injury or death to the user. All products and materials are sold and services provided subject to Mitel’s
conditions of sale which are available on request.
M Mitel (design) and ST-BUS are registered trademarks of MITEL Corporation
Mitel Semiconductor is an ISO 9001 Registered Company
Copyright 1999 MITEL Corporation
All Rights Reserved
Printed in CANADA
TECHNICAL DOCUMENTATION - NOT FOR RESALE
相关型号:
©2020 ICPDF网 联系我们和版权申明