[opendtv] DVB-T2 performance predictions

From: "Manfredi, Albert E" <albert.e.manfredi@xxxxxxxxxx>
To: <opendtv@xxxxxxxxxxxxx>
Date: Fri, 5 Sep 2008 18:50:14 -0400
I've been trying to find performance mesurements, or at least
predictions, for DVB-T2. I found this viewgraph presentation from the
RAI Research Center:

http://www.fub.it/public/Morello020408B.pdf

The part on expected performance is in slide 35.

To be considered in this slide is that it ignores the effect of GI and
the existence of active pilots, so the numbers can perhaps even better
be used to estimate the performance of the updated FEC scheme on a
potential ATSC2 upgrade. The effect is quite good.

For DVB-T2, they went from a viterbi code to low density parity check
(LDPC). You'll note in that slide that 256-QAM is somewhat less robust
at bit rates where a decision of using 256-QAM or using 64-QAM with less
FEC, can be made. These are the regions of overlap.

But conversely, when it comes to 64-QAM vs 16-QAM, the overlap is hard
to detect. This means that 64-QAM with a stronger FEC scheme works just
as well as 16-QAM with a weaker FEC, in terms of robustness vs bit rate.

So applied to ATSC, if the FEC scheme were changed from the current
viterbi to LDPC, on the same 8-VSB constellation, we should see results
exactly the same as those in slide 35.

Pretty good stuff. At the same 3.3 b/s/Hz we use now, the C/N margin
would go down from 15 dB or so we get now, to about 11.3 dB of C/N, and
the Shannon limit is 10.47 dB of C/N. Less than 1 dB of difference.
(Their Shannon curve appears to be drawn slightly off.)

In fact, just making the best combined use of the existing trellis and
RS code would probably be good for 13 dB of C/N margin, without making
any changes to the transmission standard.

Alternatively, for more usable bit rate, the existing 2/3 trellis FEC
could be reduced to a 4/5 LDPC code, which would raise spectral
efficiency to 4.5 b/s/Hz, providing 24.2 Mb/s in the same 5.38 MHz
channel, with the same 15 dB of C/N we have now.

My impression is that reducing the C/N margin in ATSC is not something
broadcasters would want, so I don't think there's any big payoff in
attempting 16T-VSB.

In fact, looking at that slide, I don't see too much advantage, in terms
of spectral efficiency, going to 256-QAM. The advantage seems to be only
to get the long guard intervals for more practical SFNs, at a price of
higher needed C/N margins.

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.
[opendtv] DVB-T2 performance predictions

Other related posts:
