Re: [Discuss-gnuradio] ATSC real-time application?

[Charles Swiger]

On Sat, 2006-10-28 at 01:19 +1000, Kyle Zhou wrote:
> I am quite interested in the ATSC project due to some previous HDTV
> R&D experience.
> I am not quite sure about the current status of this project.
> According to my understanding, ATSC has a very high throughput
> requirement with a symbol rate=10.72M / sec
> Considering this high symbol rate and the computationally intense
> algorithm in RS decoder, Equalizer, Viterbi, etc., I am just wondering
> how powerful the computer has to be to process all these in realtime? 
> What kind of computers have been used in the gr-atsc project to
> process realtime HDTV?


None that I know of. With a 4-cpu 2Ghz system with the processes
spread out so the total utilization is 80%, there's still a maybe 
(rough guess) 7:1 process-time to real-time ratio. At least I seem
to remember taking at least 14 hours to process a 2 hour movie. The
bottleneck (cpu running 99%) was still the gnuradio0.9 part from
bit-timing-loop to field-sync-demux. Once atsc is completely ported to
the 2.x framework more can be done to optimize things.

>Recalling that a PC do not have the luxury of parallel computation. For
>instance, a 4.0G CPU has to process one symbol in 400 clock durations,
>which seems to be not enough.

That's the question - is the most intensive atsc function too
much for a commonly available cpu to process in realtime?