MPEG-2 vs. MPEG-4?

No matter who ends up "winning," the Blu-Ray and HD DVD format war has probably entered its most dangerous period. For right when new formats are launched, you'll find advocates of one system or another putting forth unsubstantiated claims and various forms of quasi- and pseudo-science to back their side. The public is particularly vulnerable to such mis- or disinformation because it spreads rapidly via the Internet, and because influential first adopters (such as readers of Sound & Vision are intensely, even competitively (but often uncritically) interested in any information they can get.

The early days of DVD provide a fine example of how unproven assertions can affect the survival of a technology. Back then, a few "golden-eared" listeners - some of whom also write for a couple of A/V magazines (not this one) - claimed that DTS multichannel encoding sounded "better" than Dolby Digital encoding. These conclusions did not come from any scientifically controlled listening tests, but merely from relatively casual listening to commercial DVDs. Their opinions were uncritically taken up by many in the public and were crucial to the survival of DTS into the high-def era.

Giving those audio pundits the benefit of the doubt that they had no commercial ax to grind, there's a good reason why they had to draw their conclusions from casual listening: a controlled and fair test pitting Dolby Digital against DTS would be damn difficult to pull off.

I know, because we tried to do one.

To have a fair listening test, besides having to set up a "double-blind" presentation (so that neither the supervising tester nor the listener knows whether it's DTS or Dolby Digital that's being played), we needed an instantaneously switched, level-matched, multichannel, multi-format decoding system. But we couldn't find any commercial hardware that met all our requirements.

Where we actually had the toughest time, though, was in getting suitable program material. It's hard enough to find a DVD with multichannel Dolby Digital and DTS soundtracks where you know for sure that the same master tapes were used for each. But we also couldn't verify that the mastering engineer(s) didn't equalize, compress or otherwise "sweeten" one encoding differently than the other. We considered mastering our own Dolby Digital and DTS recordings so we could control the entire process, but getting a wide range of original movie soundtrack material proved impossible, even with assistance from Dolby and DTS.

Performing a similar controlled and fair test using the new high-definition discs will be equally daunting, which is why you won't likely see many. Of particular interest this time around is the decision by Sony Pictures Home Entertainment to use DVD-type MPEG-2 video encoding in early Blu-ray discs - especially since they raised the issue at a recent press event. While Sony plans to use MPEG-2 for Blu-ray - albeit at bit rates several times higher than used on DVDs - HD DVD's backers will be relying primarily on the newer AVC codec (a.k.a. H.264 or MPEG-4 Part 10) or the Microsoft VC-1 codec. While any of the three codecs can be used with either disc system, Sony claimed that their high-bit rate MPEG-2 encoding produces more artifact-free video than the others, at least for now.

Sony indeed showed a remarkably clean - but not absolutely artifact-free - MPEG-2 encoding of some movie excerpts, including one from A Knight's Tale. But the company did not demonstrate Sony's MPEG-2 encoding against AVC or VC-1. The company merely showed a table of results from a Japanese test in which viewers were shown MPEG-2 and AVC at various bit rates. At a very high bit rate of 24 megabits per second (Mbps), Sony said, 100% of the viewers found the MPEG-2 high-def video quality "acceptable," against 70% who found AVC acceptable. At a more realistic bit rate of 20 Mbps, MPEG-2's edge shrank to just 5% (60% vs. 55% who found AVC acceptable). That's a small difference that might easily be accounted for by a host of factors known to influence visual-codec tests (such as the choice of program material). The scary part is that even 60% "acceptability" is below what I was hoping either of the new systems would provide, assuming these are scientifically valid results.

I think Sony was being unnecessarily defensive about using "old-fashioned" MPEG-2 - what matters is the visual result, not how you get there - and that their press event can just be seen merely as part of the jousting that occurs in any format war. Microsoft, for its part, can trot out any number of viewing tests showing the superiority of their VC-1 to even AVC, much less MPEG-2. I don't trust any of them, and neither should you. The jousting here, unlike most such medieval encounters, is intended to be fatal.

Back to Tech Talk Main Back to Homepage What's New on S&V