>From: "Manfredi, Albert E" <albert.e.manfredi@xxxxxxxxxx> >To: <opendtv@xxxxxxxxxxxxx> >Date: Fri, 23 Jul 2004 16:33:33 -0400 >John Shutt wrote: >> You keep referring to this, but it is my understanding that >> the reason why >> most chip manufacturers ignored the training sequence and >> went with blind >> adaptation was the fact that the training signals were too few and far >> between to use in tracking any sort of dynamic multipath. The two big problems many of us identified long ago with the ATSC training signals were (1) not long enough and (2) not frequent enough. Both problems seemed to have the same root cause: lack of sufficient understanding of terrestrial UHF propagation -- at least as it pertains to DTV -- at the time the standards were written. Not long enough compared to the multipath delays measured in the real world, and not frequent enough compared to the rate of variation of RF channels measured in the real world. >> If I recall correctly, it was the blind equalization >> techniques that NxtWave >> and Motorola were referring to when they wrote their Summer 1999 press >> releases that claimed to "crack the code" of 8-VSB. >Yes for NxtWave. I don't remember what Motorola said. Nxtwave said they "cracked the code." Motorola said they "solved the multipath problems" that Sinclair and others pointed out in 1st gen receivers. Neither statement was entirely true, but not false either. Gen 2 was a huge improvement in multipath performance over Gen 1 or the A/54 reference design, which was based on the original Zenith 'blue rack.' In my humble opinion, the performance improvements from Gen 2 to Gen 4 were only modestly incremental. Gen 5 appears to be a much more substantial leap forward. >But just because they took the blind equalization path as a >first cut at improving receivers certainly can't be >interpreted to mean they thought that was the best one >could do. True statement. Blind equalization of some form is still an important part of the solution, but it's not the only thing that can be done. The training signals are still inadequate -- neither they nor the RF channel characteristics have changed -- but "inadequate" doesn't necessarily mean "completely useless." >> I also seem to recall that Dr. McDonald also was of the >> opinion that if >> there were more training signals, then 8-VSB would have been >> much easier to >> equalize for dynamic multipath, but blind adaptation >> techniques are required with the current training signals. >Could be. Doug might have changed his mind on that, though. >He can speak up and take away the suspence. His last comment >was that using the existing sequence would go a long way, as >I recall. He also prefers a different PN sequence than the one >used now, but it takes up some extra bit rate because it's >twice or three times longer. Besides, it's incompatible with >existing receivers. Doug's scheme, or more correctly the Patel-Limberg-McDonald scheme, was a brilliant attempt to fix the ATSC training signals in the least incompatible manner. They approached the problem in much the same way DVB-T did: make the DTV signal itself carry all the information necessary for a simple receiver to acquire and track the characteristics of the unknown RF channel. Unfortunately, the desire for backward compatibility seems to have precluded any real attempt to fix the deficiencies of the ATSC 8-VSB signal structure. So rather than putting the burden on the signal generator and keeping the receivers simple (i.e., cheap), we still have the situation where clever and probably very complex DSP hardware is required in the receiver to achieve results we can call "excellent." At least someone has finally done that. -- Frank ---------------------------------------------------------------------- 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.