Bert -I don't have an intuitive grasp of C/N ratios, and what is needed for "acceptable" reception. I'm a software guy, not a TV engineer, and tend to think of things in bits, good bits and bad bits, but already bits, with some expected (but varying) probability of being correct at any moment.
My issue with predicting reception from C/N ratios is that the amount of signal and noise appear to be averages taken over some period of time. It seems these values would actually fluctuate quite wildly from moment to moment with multi-path problems. Meanwhile, error correction is limited to correcting only over short intervals of time unless you are willing to put up with long delays and huge buffers, all not acceptable to M/H TV viewers.
So it seems you could have an acceptable C/N ratio averaged over time but the bursty error rate in multi-path conditions would still cause dropouts as the error rate in any short period exceeds the ability to error correct within that same limited period.
How well have C/N ratios actually predicted 8vsb observed performance under multi-path, assuming failures are observed as drop-outs?
> The way MPH was being used, the 8T-VSB wide channel was 14.9 > Mb/s (15.2 dB of C/N) and the narrow channel should be something less > than 2.2 Mb/s if 1/2 rate were used, at an undisclosed C/N margin.And yet, adjusting to pretend all of the the narrow channel was using 1/2 rate, only 1.2 Mb/s was actually observed, just slightly over half your prediction. Though Mark A. has indeed said the numbers were flaky.
In any event, all this is why I am skeptical of numbers that look good on paper but haven't yet been shown in real world experience. Sometimes that is the best we can do but it is easy to overlook things with a simple mathematical model. Most anything involving error correction seems to fall into this bucket since the long term average error rate is not a good enough indicator of actual real time performance.
- Tom Albert Manfredi wrote:
Tom Barry wrote:But I'm still not sure at all of the other side. That is, how much FEC overhead is needed to get similar M/H reception quality on DVB-T. Is it truly 3 or more times better at mobile or does DVB-T also have to use massive extra error correction to getquality M/H reception. If so, how much?That is, how many bits / second can reliable emerge from a moving handheld receiver using DVB-T in a 6 mhz channel, after subtracting FEC overhead?The results from tests are a good thing, of course, and Mark will have some available soon. But this doesn't preclude considering what the underlying principles are. And among these is that DVB-T does HM by addressing two components: the FEC (and GI) of the robust channel AND the constellation. HM uses QPSK, which is analogous to 2-VSB. However, in 8-VSB, if M/H were to use a different constellation, the entire multiplex would have to use that same constellation. Since only the 8-VSB one is used, that kind of limits what the M/H channel has to be built on. So you can't *directly* compare the FEC mode used in the robust channel, between the two systems. But you CAN compare the bit rates with the C/N margin requirements, which ultimately is all that matters. For DVB-T, the document that shows what one can expect is ETSI 300 EN 744. It doesn't give all results for 6 MHz channels, but you can draw the same comparisons using the 8 MHz results. What that document shows, for 8 MHz channels, is that in the mode that favors the HM channel most, you can reduce the wide channel's capacity from 24.13 to 12.06 Mb/s, to retain the same robustness you had in non-HM mode. Doing so, you will get an HM channel that gives you as much as 6.03 Mb/s of capacity, with a C/N margin of 6.5 dB of in a gaussian channel (this is with the same 1/32 GI as the wide channel). So, I suppose you could conclude that this would be the same overall effect as 8-VSB M/H with a 1/2 rate FEC, right? The 12 Mb/s taken away from the wide channel becomes 6 Mb/s in the HM channel. So the next question is, what C/N margin is required in the MPH or A-VSB, when the 1/2 rate option is used? IIRC, A-VSB gives you 4 dB of C/N for a 1/4 rate channel and a single receive antenna, and something like 1.9 dB of C/N if a diversity receive antenna is used with 1/4 rate. But I do not remember, or maybe the numbers were not given, for the 1/2 rate option. The other consideration is that when putting COFDM into HM, you are going to get a fairly big robust channel. Whereas with MPH or A-VSB, it's more like DVB-H. The robust channel can take up less of the wide channel (and of course provide less M/H capacity as a consequence). This may be desirable, though, for stations that want to retain the ability to transmit an HD and SD multiplex along with M/H. In short, using a simple 6/8 ratio: COFDM HM in a 6 MHz channel could provide you with 9 Mb/s of wide channel (16.5 dB of C/N margin required) and 4.5 Mb/s in the H/M channel (6.5 dB of C/N). Echo tolerance would not be impressive, though, because we are assuming 1/32 GI. The way MPH was being used, the 8T-VSB wide channel was 14.9 Mb/s (15.2 dB of C/N) and the narrow channel should be something less than 2.2 Mb/s if 1/2 rate were used, at an undisclosed C/N margin. Or something less than 1.1 Mb/s (~4 dB C/N) if 1/4 rate is used. The "something less" is to account for the extra overhead needed for the training and sync sequences, but that shouldn't be a big number. Bert _________________________________________________________________ Get Windows Live and get whatever you need, wherever you are. Start here.http://www.windowslive.com/default.html?ocid=TXT_TAGLM_WL_Home_082008 ----------------------------------------------------------------------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.
-- Tom Barry trbarry@xxxxxxxxxxx ---------------------------------------------------------------------- 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.