The Colossus of Audio
Developed by Louis Dorren, inventor of Quadracast, the national FM quadraphonic broadcast standard, Colossus is described by him as a 16-bit, linear-encoding, 4-channel digital recording system utilizing a "proprietary code." No other details have been forthcoming to date, and won't be "until the system has patent protection." This month, I received for review several Compact Discs, from different sources, all of which had been mastered by Colossus, and there was enough musical material among them for me to form some conclusions about how this new PCM system performs. (See the reviews elsewhere in this issue.) What those conclusions boil down to is that Colossus is a first-order technological breakthrough: It is the first recording system which, to judge by the evidence I have so far heard, does not sound like a recording system at all!
So, what does Colossus sound like? Since it is impossible to describe the sound of something that seems to have no sound of its own, I can only describe what it doesn't do. First, and perhaps most important, is that Colossus appears to have absolutely none of the so-called "digititis" that has plagued so many CDs to date. Its sound is at once much more liquid (less dry) and more delicately detailed than anything I have heard before, it is (by comparison) noticeably less veiled, and it seems to reproduce all instrumental timbres more felicitously. It does not sound at all "digital."
It is significant that all of these qualities were heard when playing the Colossus CDs through a "conventional," albeit top-of-the-line, Sony CD player, rather than one of the audiophile-type models like the Cambridge Audio or California Audio Labs units. This alone would seem to be proof that the dryness and grundge of many previous CD playbacks are not inherent in the CD system, and that in fact the CD system is capable of much better sound than anyone had imagined, exhibiting an extremely high degree of transparency, liquidity, and detail. Where then have those qualities been getting lost until now? I think we can deduce the answer to that, too, from my observations and from what is known about the recordings I auditioned.
Since the Colossus digital-word code is unique, its recordings must be bit-converted to Sony's PCM-1610 (CD) format before they can be mastered for CD. The conversion is done on Harmonia Mundi's BW-102 unit. Yet the sonic qualities of the system are apparently preserved, even after the conversion to CD format. Clearly, then, the difference between the sound of Colossus and that of other digital systems originates somewhere ahead of the digital domain, otherwise that difference would be lost in conversion to and playback from the CD format. Could it then be the A/D section where Colossus is doing something more "right" than other systems?
Why is it that CDs made from analog masters always seem to have much less digital grundge than ones that were digitally mastered? Probably because analog tape is inherently bandwidth-limiting; it chops highs. At the usual tape speed of 15ips, the upper limit of a professional tape recorder is (typically) around 17kHz, and even at 15kHz, the maximum recordable level is 6dB or so below the tape's maximum output level at 1kHz. It is important, too, to note that this HF attenuation effectcaused by a combination of tape saturation and self-erasuresimulates that of a dynamic limiter, not of a low-pass filter. That is, it sets a fixed limit to how strong an HF signal can be recorded, rather than just attenuating high frequencies by a certain amount. (When simple attenuation is taking place, as in a filter, you can still increase HF output just by boosting HF level. But when it's a matter of limiting, raising the HF level makes no difference in how strongly the signal is recorded; the maximum cannot be exceeded.)
Thus, when we copy from that analog tape to a digital format, the HF limiting acts to improve the effectiveness of the anti-aliasing filter that's ahead of the A/D converter. The result will be less aliasing. And since the digital grundge is what we hear to be markedly diminished on ADD discs, it is probable that aliasing is the source of much of the grundge we hear from CDs.
Making a digital master from microphones is a different animal altogether. A mike has no HF limiting action, but behaves instead like a low-pass filter, providing progressive attenuation of all frequencies above the natural resonance of its diaphragm. In other words, even if it rolls off rapidly above 20kHz, it will still be feeding some energy to the PCM recorder at frequencies as high as 30kHz. And the stronger that ultrasonic content, the stronger will be the aliasing products.
Condenser or capacitor microphones are still the de facto standard type for professional recording, and they are notorious for having a high-end peak in their response. This was a valued trait in the days of analog, because it helped to retain adequate high-end range in a system that tended to lose highs at every step along the way. It is a liability in PCM recording, though, because it increases the amount of extreme HF energy delivered to the PCM system, thus increasing the amount of aliasing that takes place in A/D conversion. The customary way in which microphone designers have attempted to sidestep the HF peaking problem has been to raise the frequency of the peak to beyond audibility. B&K took this route with their line of omni mikes, which is one of the things that made them so popular with audiophile-oriented recording companies. (The other was their freedom from colorations due to angle of incidence.) But while that "inaudible" ultrasonic HF peak makes the signal sound much better through the monitor speakers while recording, it in fact worsens the aliasing problems of A/D conversion, because the peak frequency is farther above the critical Nyquist point (the input frequency where aliasing sets in), and thus produces foldover frequencies extending farther down into the audible range.
On the basis, then, of all the evidence on hand, it would appear that the "secret" of Colossus may be nothing more than unprecedentedly effective anti-alias filtering. (Lou Dorren has so much as admitted this, but not quite.) The fact that it is also claimed to use no error concealmentso-called extrapolationwould only be of incidental significance, since concealment is required only when error correction fails to restore signal integrity, and it has been shown that, even with CD, that situation arises very rarely (see Sidebar). It does not happen nearly often enough to account for the consistently audible superiority of the Colossus system.
Ultimately, I expect the design teams at Sony, JVC, Mitsubishi, et al to find ways of improving the much-neglected (to date) A/D conversion step in their pro digital processors, to the point where they sound as good as Colossus's. But for the time being at least, Colossus is far enough ahead of the field in sound quality that it may be some time before it is equalled, let alone bettered, by any other system (footnote 1). The fact that it, alone of all the video-based PCM systems, now offers discrete 4-channel recording capability (and is claimed to have the capacity for 8 or more), at a time when surround-sound appears to be gaining acceptance in home audio systems, would seem to give it an even wider margin of superiority over its competition.
As of now, and for the foreseeable future, any sound-conscious record company not using Colossus for all original digital mastering is doing itself and its customers a disservice. Lou Dorren's system is clearly at least an order of magnitude better than anything else currently available, and that, after all, is what "state-of-the-art recording" is all about.
Footnote 1: Information about the loan, rental, or purchase of a Colossus processor, usable with any video recording medium (¾" U-Matic preferred) can write to By-the-Numbers, PO box 8359, Incline Village, NV 89450.J. Gordon Holt [21 years later, this address is no longer valid.Ed.]