Facing the Codec Challenge Page 4

High bit rates = good, low bit rates = bad. Take a look at the difference between WMA/128 and WMA/64 on solo harpsichord (Track 8). The lower bit rate was judged distinctly inferior ("sour," said Orange, and Red mentioned "flutter"), while at 128 kbps both Red and Orange found it too close to call and confused the WMA track for the original. The same situation is even more vividly conveyed if you compare the results for Real/132 and Real/64, especially the castanets (Track 4) and solo violin (Track 7).

The RealAudio results are also instructive in another way. While going under the same name, RealAudio 8 at 132 kbps is quite a different codec than RealAudio 8 at 64 kbps. At the higher bit rate, it's essentially identical to Sony's ATRAC3. But at 64 kbps and below, RealAudio 8 turns into a Real-designed codec that's apparently intended to make compression artifacts less objectionable rather than creating fewer of them to begin with. Given our test results, maybe Real should have tried to make ATRAC3 work at 64 kbps instead.

You can't get CD quality at 64 kbps - at least not with these codecs. Not only did both Real/64 and WMA/64 sound distinctly inferior to Real/132 and WMA/128, respectively, but they both sounded decidedly worse than MP3/128. On the basis of these results, Microsoft is not justified in claiming "CD-quality at 64 kbps" for WMA, which would entail no discernible difference between WMA/64 and WMA/ 128, contrary to what we found. Nor would Real be justified in making the same claim for RealAudio 8. In fact . . .

None of these codecs are "CD-quality" at the bit rates tested. If any of them had been, one or more of the graphs would have had all its scores clustered around the zero point (within ±0.5, say), indicating the inability of all three listeners to distinguish that codec at that rate from the original, uncompressed CD audio.

Winners and Also-Rans In this round, the codec that came closest to the "CD quality" ideal was WMA/128, which was somewhat surprising given its so-so performance in our previous tests. WMA has clearly been greatly improved since its earlier version, which stumbled badly over the castanets and harpsichord tracks. This time WMA did much better than MP3/128 on those tracks.

To judge by a "the more below zero the score, the worse the sound" rule of thumb, the next best results were returned by Real/132, followed closely by MP3/128. The two 64-kbps codecs fall into another - and inferior - sonic category altogether, though WMA/64 clearly beat Real/64.

So both WMA128 and Real/132 are better, overall, than the now aging MP3 system. Incorporating these codecs in traditional high-fidelity equipment is no cause for alarm, if music is encoded at a high enough bit rate - 128 kbps is the minimum that any music lover should accept for serious listening. And if the pay-per-download services are genuinely concerned about sound quality, they ought to steer away from MP3, particularly given that both WMA and RealAudio offer the added incentive of rights management and copyright control.

The big question remaining is how the acknowledged Cadillac of codecs, namely AAC, would do among the present company. When it gains more prominence than it has as yet - only LiquidAudio.com is using it in a big way - we'll boot up our comparison software once again.


ARTICLE CONTENTS

X