DTS & Data Compression

In the October 1994 Stereophile (Vol.17 No.10, p.39), I discussed my experiences with the DTS audio data-reduction code/decode switch box, which, briefly, is a two-channel box that makes use of the algorithm DTS has proposed for their version of discrete multichannel sound for laserdiscs.

Unlike the proposed Dolby AC-3 system which has been adopted by Pioneer, which will be located in the data space now reserved for one of the existing laserdisc analog tracks, the DTS proposal is to put the multichannel signal on the existing digital-track space. The Pioneer/Dolby arrangement retains the original digital tracks for digital playback, Pro Logic compatibility. DTS retains the latter, but only on the analog tracks.

In compensation, DTS makes use of the full 1.4 megabits/second data capacity of the existing digital tracks for their six channels of information. AC-3 uses only 384 kilobits/second for their system, requiring the use of a much more aggressive data-reduction algorithm when more than one channel carries information (footnote 1). DTS's data reduction—which they call Zeta—is relatively benign (they use a data rate of 240 kilobits/second/channel compared with CD's approximately 706kb/s/ch).

DTS's test box enables the listener to switch instantly between the original, unencoded program material and the perceptually encoded version. Recently, DTS sent us a sample of the same two-channel box I'd reported on in that previous article. Unfortunately, it arrived at a time when we weren't able to make full use of it in organized listening tests, but we were able to both run some measurements and make an archive copy of its sonic performance for later study. JA did the measurements—see later—and also used it to encode musical selections from recent Stereophile recordings for possible use on our Test CD 3.

Just before we had to return the unit (only three of them exist, and they're understandably in demand), I also filled two CD-Rs with program material excerpted from a number of commercial CDs. On one of these CD-Rs, I recorded the same material both with and without DTS processing; the second was made up primarily of DTS-encoded music. I chose a broad spectrum of material, though the nature of audio data-reduction is such that one can never be certain that some as-yet-unheard selection won't trip up the system. Nevertheless, these CD-Rs were then available to me and other people here at Stereophile for later auditioning, long after we had returned the DTS box. The only limitation to this test is that DTS claims full 20-bit-equivalent resolution for their system; in theory, then, use of commercial 16-bit material will not get the most out of the system.

Despite this limitation, I found the performance of the DTS system in its 240kb/s mode (the rate proposed for laserdiscs and for music-only CDs) phenomenally good. I listened to the above CD-Rs on loudspeakers as high-end as the Infinity Epsilons and the Snell Music and Cinema Reference system, and on headphones as good as the Stax SR-Omegas, so I'm reasonably certain that system resolution was not a limiting factor. At least one listener for whom I played the recordings suggested that there was a slight loss of ambience at the top end in the DTS mode, and on further listening I thought that perhaps this might be the case. But I found it to be a very elusive thing, indeed; and in the context of a full surround-sound system, where ambience is more pronounced in any event and more controllable (surround setup and level being fully under the listener's control), it's probably irrelevant.

What I didn't hear were any digital nasties added to the existing program material—no edge, glare, harshness, or loss of detail. This was consistent with my prior experience with the box in real-time A/B testing, without the intervening CD-R recordings. The recordings, incidentally, were made digital direct—the only D/A conversion was on final playback. Everyone for whom I played the CD-Rs agreed that six channels of information of this quality would be superb in a surround-sound application.

I also brought the CD-R disc which contained primarily DTS-processed material with me to the 1995 Winter CES, from which I've just returned as I write this. I played several selections in a number of rooms, and found that it was every bit as rewarding to listen to as any other standard CD I brought. The exhibitors for whom I played it seemed uniformly impressed as well.

We haven't yet been offered the opportunity to run the same sort of tests with Dolby's AC-3 system, but would certainly like to do so. Because of AC-3's bit-pool sharing system, however, it wouldn't be as simple to properly exercise AC-3 and evaluate the results as it was with DTS's system (footnote 2). In the latter, the bits available to each channel are completely independent.

Footnote 1: AC-3 was demonstrated at the recently completed Winter CES in Las Vegas, and at least two of the demonstrations I heard were very encouraging. DTS was scheduled to demonstrate, but last-minute room arrangements caused them to cancel. More on this in our complete CES report next month.—TJN

Footnote 2: The available bits are allocated between the six (actually 5.1) channels as required by the activity in each. If you only use two channels, all of the bits will be available for them, which will theoretically give better quality than if all channels are active at the same time.—TJN