DatasheetQ Logo
Electronic component search and free download site. Transistors,MosFET ,Diode,Integrated circuits

VP2611 Просмотр технического описания (PDF) - Mitel Networks

Номер в каталоге
Компоненты Описание
производитель
VP2611
Mitel
Mitel Networks Mitel
VP2611 Datasheet PDF : 14 Pages
1 2 3 4 5 6 7 8 9 10 Next Last
VP2611
Frame Store Manager
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.
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.
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.
Discrete Cosine Transform
This circuit performs a Discrete Cosine Transform on each
8x8 sub block, whether in inter or intra mode. In intra mode,
eight bit pixel data is used, with a ninth implied sign bit ( all pixel
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.
Quantize
This section quantizes the results of the DCT by dividing
the 12 bit output from the DCT with a host supplied value. The
5 bit quantization value supplied corresponds to division of the
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.
4
Zig Zag Scan
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.
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.
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
For Intra Coded DC Coefficients :
REC = 8*LEVEL
except if LEVEL=255 when REC=1024
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.
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).

Share Link: 

datasheetq.com  [ Privacy Policy ]Request Datasheet ] [ Contact Us ]