"Andy Hall" wrote in message
...

I am familiar with the basics of COFDM as well as the rival U.S.
system and have read the papers by Stott and others.

OK, sorry, but that wasn't clear from what you said.

While COFDM should give good multipath performance, the code rate and
the guard intervals mean a compromise between the data rate and
signal robustness.

It should and does give good multipath performance, up to (a little over)
the guard interval. As we're not using SFNs for TV the guard interval
fraction has been set to the lowest DVB option (1/32) to minimise the loss
of capacity.

I understand that there are various issues around the UK
system because of the limitations imposed by the current transmission
arrangements, but I haven't looked at that in depth.

ISTM that you're thinking more here of the choice of modulation
constellation and code rate. In 1998 DTT launched with 64-QAM modulation &
2/3 code rate, giving about 24Mb/s net data rate. Following the demise of
ITV Digital the BBC and Crown Castle decided to opt for 16-QAM to increase
robustness, but to reduce the code rate to 3/4 to claw back some of the lost
capacity. This combination gives about a 4 dB advantage in a gaussian
channel. So we now have a hybrid situation with the BBC & CCI running
16-QAM 3/4 (~18Mb/s) but with the D3&4 mux (ITV, C4, etc.) and SDN (C5, S4C,
etc.) still running 64-QAM 3/4.

In practice I have seen cases where the bit error rate is noticably
worse when cabling is poor or the antenna is not aligned correctly
and there is visual evidence of multipath reception on analogue.

The higher BER there might just be due to the signal being weaker, of
course, or to long-delayed echoes.

I suspect, therefore, that as Stott admits in at least one of his
papers, while one can model some aspects of COFDM mathematically,
some issues can only be determined experimentally. It would appear
that the practicality may not be quite what the theory predicts.

I think most aspects of it are now pretty well understood and measurements
agree well with theory. The performance of different receivers in the
presence of long-delayed echoes does vary though, and depends (IIRC)
somewhat on the Viterbi implementation. Unless you can be more specific
about which of Jonathan Stott's papers you're referring to I can't really
add any more.

DAB suffers in practice from limitations arising from the
broadcasters trying to squeeze more into the spectrum than will give
the best results.

That's a completely separate issue, not COFDM related. There isn't, AFAIK,
the same flexibility as there is in DVB-T to change the transmission
parameters to trade off capacity and signal robustness.

--
Andy