IQBADX
SECTION 28c
IQBADXOPS
050804
www.snellwilcox.com
Version 1 Issue 5
28c.20
Extractor Operation
The SMPTE 272M standard specification allows for
up to four groups of digital audio to be embedded
in a component D1 digital video stream.
Each group consists of two stereo pairs (four
channels), giving a total of sixteen audio channels
for all four groups.
A single IQBADX module is capable of extracting
four stereo pairs, or eight channels, so two
IQBADX cards are required to extract all sixteen
possible channels.
The embedded AES audio data must be sampled
at 48 KHz, clock synchronous to the video stream.
Asynchronous operation is not supported.
The IQBADX module contains four independent
extractors, each one assignable to any of the
possible embedded audio pairs. If any of the
extractors are set to the same address they all will
output that extracted pair. Additionally, any output
can provide an AES test tone (1 KHz, -20 dB) or
‘digital silence’.
A total of exactly 1920 audio samples occur within
one frame of 625-line video. For 525-line video the
relationship is 8008 audio samples over five video
frames.
The audio data is usually distributed evenly
throughout each video frame, situated in the non-
active picture regions between the end of one line
and the start of the next. The majority of lines
contain three audio samples, some four, and
certain reserved lines may contain no samples.
Other sample counts are also possible.
These differences in sample count between lines
mean that some FIFO buffering is required to
supply output audio samples during lines
containing no samples, and to absorb the excess
from lines with several.
An audio extractor should be able to process
ancillary packets where ever and whenever they
appear. However, there are usually practical and
operational limitations such as the finite length of
the extractor FIFO buffer, and the length of the
audio delay introduced by the buffering process.
IQBADX FIFO Buffer Length :
A FIFO buffer requires that before any samples are read out a minimum number must be written in. This
number determines the latency (delay) and implies limitations on the embedded sample distribution. So-called
‘smart buffers’ make assumptions about the distribution of audio samples within the video frame and so can
utilise smaller FIFOs. The 24-bit sample distribution given in SMPTE-272M for example requires a ‘smart
buffer’ length of 57 samples with a read-write offset of 17 samples. A simple buffer however needs to be 80
samples long with 40 samples offset. In both cases the buffers can handle a maximum change in usage of 40
samples but the ‘smart’ buffer knows when this happens and so can use less memory and have a lower
latency.
For any known sample distribution the required buffer capacity can be calculated. The simple buffer needs to
be twice this in length
SMPTE-272M, 24-bit example
40 samples
Panasonic D5 VTR, 625 lines
6 samples
Sony Digital Betacam, DVW-500P
5 samples
S&W IQBAI / BAIR inserter modules
14 samples
Tektronix TSG-422
7 samples
S&W MDD-3000 digital decoder
7 samples
