Albert Manfredi wrote :
Obviously, if the analog is even less capable than I assume Olivier was suggesting, all you're proving is that the analog example is even further away from the Shannon limit compared with HDTV in the 6 MHz channel.
The Shannon limit certainly does come in play.
Since i am not among the happy few that have watched HDTV over DTT in France, i am not in position to do a direct meaningful comparison with PAL/SECAM at the moment, so i suppose i can indulge in a little more math : Duration of the useful part of a line in PAL is 52 µs. Luma bandwidth is 5 Mhz, chroma is 0.6 MHz. This makes for 520 luma pixels and 62.4 chroma (there are two chroma components but as PAL halves the vertical colour resolution the two factors cancel out). Using the SNR = 6.02 n + 1.761 formula for 40 dB would give 6.35 bits per pixel. Since the video information itself uses only 70% of full scale, i will reduce that to 6 bits. This would yield : 582.4 x 576 x 6 x 25 = 50 Mbps for an equivalent "digital PAL". This can be compared with an uncompressed 720x576@25 4:2:0 signal which needs 124 Mbps. The Shannon channel capacity theorem would allow for 106 Mbps in an 8Mhz channel at 40 dB (and 80 Mbps for 6Mhz). So this information should indeed be correctly transmitted to the viewers in the case of PAL. For US HDTV, we start with 746 Mbps (4:2:0) , which is compressed down to 15 or 17 Mbps. 6Mhz at 15 dB will allow for a maximum bitrate of 30.17 Mbps anyway.Of course, since PAL is uncompressed it contains much redundancy. A simple intra frame lossless compression could probably divide the datarate by 2, down to 25 Mbps. Removing temporal redundancy would possibly allow for an additional reduction. So we would probably end up below the 17 Mbps figure of HDTV, but perhaps not so far below. I would say the figure alone would still allow for some doubt as to how much additional information the HDTV signal actually provides.
It boils down to how much data you really need to reconstruct a given picture in a way that will satisfy the human eye. I don't know if there is an accurate mathematical model for that somewhere, though perhaps if there are indeed test tools that can predict the quality that will be perceived by the viewer, as mentionned on this reflector, someone may have done the modelling. Now to come back to that 31 Mhz bandwidth , i still think it is irrelevant as long as the information has gone through a 6Mhz channel on its way to the viewer . After all, you can upsample a SDTV signal to 1920x1080, but if you run a spectrum analyzer on it, you would still get the 6Mhz spectrum of the original signal all the same. But never mind the figures, i am all for HDTV anyway (in case there is some doubt). It does its job on a large screen in a digital cinema (very little compression) It can be gorgeous on some demo setups in shops (often run from a local hard disk, possibly with limited compression, too). But material from satellite retransmissions , closer to our discussion, seems much less convincing to me, though it may be due to other factors like the lack of maturity of MPEG4 encoders (one HD-1 retransmission i saw looked nice to me but it was MPEG2), or the use of upsampled SD material . And at least, i can see with my own eyes at home that a DTT SDTV channel is at least a good as its analog equivalent on a good day, though it uses one fifth the bandwidth and one tenth the power. So if things scale up correctly for HDTV (and they should since higher resolution pictures compress better) i should expect reasonably faithful retransmission of HD which would be an improvement on SECAM, indeed. ---------------------------------------------------------------------- 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.