[opendtv] Re: News: Glitches Come to Light in Wilmington
- From: "Manfredi, Albert E" <albert.e.manfredi@xxxxxxxxxx>
- To: <opendtv@xxxxxxxxxxxxx>
- Date: Wed, 10 Sep 2008 12:45:38 -0400
Tom Barry wrote:
> I think the FEC is part of the standard, so changing it probably
> will change the standard.
True, but using the existing viterbi convolutional code and Reed Solomon
block code in a more efficient manner would not require a change in the
transmission standard. That would allow a receiver to achieve more
robust operation without impacting anything else. That's music to the
engineer's ear.
As Al Limberg and A/54 (I think) mention, you can double the number of
correctable bytes per MPEG-2 TS frame if you use the trellis code to
locate the error bursts in the frame, and providing that info to the RS
decoder. Instead of doing two sequential but independent FEC operations.
Out of 188 bytes of MPEG-2 frame, you can increase the correctable
number of bytes from 10 to 20, which isn't half bad.
And I'm not convinced that's the best you can do, either. I think there
might be some more gains if the feedback equalizer plays a part in this
too.
> In any event the best FEC in the world won't deliver more than
> 100% of the channel capacity.
The best FEC is what leads you toward the Shannon limit. The Shannon
limit is 10.47 dB of C/N. And I'll say again, there is nothing written
in concrete about the current ~15.2 dB of C/N quoted often for ATSC. It
is the limit at which humans start noticing glitches, and it assumes
sequential error correction by the trellis and RS decoders. It is NOT
any sort of physical limit.
So, a better FEC can get you either more robustness *OR* it can eat up
less of your theoretical 30 Mb/s channel capacity, over that 8T-VSB
channel. Or a combination. And there certainly are FEC schemes out there
today that can get you within 1 or so dB of the Shannon limit.
For instance, EITHER you get 19.39 Mb/s with something closer to 10.47
dB of C/N, or you get something closer to 27 Mb/s at 15 dB C/N (for
example), all while still in a 5.38 MHz channel, or any other
combination.
> Certainly good FEC is essential but it alone is not a silver bullet.
> You have to start with a reasonably low error rate. And if you have
> bursts of high error rates that span longer than your error correction
> buffer size then you will still have dropouts during those periods.
Sure, that's why it's important to keep track of where you are wrt the
Shannon limit. The fact is, though, that 10.47 dB is a good number. Some
of the COFDM HM modes don't give you better than that in the robust
channel, and many have said that HM works very well. So better FEC is a
good thing to work on, IMO. Imagine having something as robust as the
narrow channel in HM mode, but still with 3.3 b/s/Hz of spectral
efficiency.
Bert
----------------------------------------------------------------------
You can UNSUBSCRIBE from the OpenDTV list in two ways:
- Using the UNSUBSCRIBE command in your user configuration settings at
FreeLists.org
- By sending a message to: opendtv-request@xxxxxxxxxxxxx with the word
unsubscribe in the subject line.
Other related posts:
