Miscellaneous

Editor's Note: Published in 1998, this was the final review written by Stereophile's founder, the late J. Gordon Holt in the 37 years he was associated with the magazine. In it he expounds on his passion for experiencing recorded music in surround sound. Our continuing focus on two-channel products and recordings was one of the reasons Gordon eventually resigned, in August 1999.—John Atkinson

The Electronics Industry Association reports that one in four US households has a surround-sound system. What they don't report is that no one listens to music on it. Even when the family stereo and the home theater share the same loudspeakers, the surrounds go off when a CD goes on. (Many receivers turn the surrounds off automatically when you switch them to non-video input.) Americans seem about as uninterested in surround-sound for music as they are about Guatemalan politics, as a quick tour of any CES or Stereophile HI-FI Show will confirm. With very few exceptions, exhibits with a video screen will have multichannel surround-sound; ones that don't will have two-channel stereo.

Why should this be? After all, real sounds happen in three-dimensional space, and music sounds much more realistic when that space is reproduced in 3-D instead of through a sonic picture window. But perception is what drives the marketplace, not reality, and the public perception in 1998 seems to be that surround-sound is for movies and stereo is for music.

Pro-audio compnay Rocktron begs to disagree. They're the company behind Circle Surround, and they're convinced there are vast hordes of music lovers out there who itch to switch from two channels to five, just waiting for the right multichannel system to come along before they take the plunge. I wish I could share Rocktron's optimism, and not just because I'm a shameless surround-sound fanatic myself. But the historical evidence just doesn't justify optimism. Multichannel music has already failed once in the marketplace—28 years ago.

Horrifical history
In 1960, stereo was a booming business but in 1970, it was showing signs of market saturation. So the industry conspired to say, "Gee, now that the multitudes have stereo, wouldn't it be peachy-keen if we could sell them two more amplifier channels and two more speakers?" And it came to pass, and it was called quadraphonic sound because educated people used to know Latin in those days. But the multitudes were not impressed, being perfectly happy with their old-fashioned stereo. Of course, the multitudes had never recognized quality in anything until the arbiters of public taste declared it to be good, but alarmingly, the audio pundits and record critics didn't embrace quadraphonic sound either. For very good reasons.

First of all, it had too many faces. There was discrete quadraphonic on two-track and four-track open-reel tapes. (This was by far the best delivery system for quad, but the tapes were just too expensive.) There was a discrete system that consigned the surround signal to a 30kHz RF carrier, and was completely incompatible with all the others. There were simple passive boxes that extracted the hall reverb from ordinary music recordings and routed it to rear loudspeakers, and there were sophisticated encode/decode systems featuring inscrutable names like SQ, QS, EV-4, and UVX. And each system, of course, was touted as the only correct way to do surround-sound. The multitudes, confused about which system to buy, bought none.

The fact of the matter was, most of these systems were more alike than different. All the encode/decode systems used a sum-and-difference matrixing system to shoehorn front and rear center channels into the main stereo pair. Front center, of course, was a left-signal/right-signal mix, because L+R is what produces a phantom center in stereo. The rear channel was recorded out-of-phase in the front left and right channels (footnote 1), so that the process that recovers one signal will cancel out the other. Adding the main left and right extracts the front center signal and cancels the out-of-phase rear signal, while subtracting front left and right from each other extracts the rear signal and cancels the front signal. Unfortunately, the channel separation between adjacent channels is a lousy 3dB, so a technique called logic steering was ultimately used to monitor the stereo signals, "decide" which is the dominant channel at a given instant, and subtract that channel's signal from the ones adjacent to it. For instance, when left front is dominant, its signal is canceled from left back and center front. When the surround channel is dominant, its signal is subtracted from the left and right fronts (footnote 2).

Anyone who's into home theater may notice that I've just described Dolby Pro-Logic. In fact, Dolby's famous theater surround system was based on the same Peter Scheiber patents as the old quad systems. The main difference is that Pro-Logic is still with us because it works, while the early quad systems aren't because they didn't. Logic steering introduced too late, which meant unacceptable front/back crosstalk for all the early adopters of quadraphonics. And the active decoders that did use logic steering suffered from sluggish, over-responsive, or erratic steering, while the additional amplifying circuitry fouled the main front signals. Audiophiles were not impressed.

Then there were quad's artistic affronts to music. To ensure that people noticed how different quad was, increasing numbers of recordings threw more and more sounds into the surround channels until the listener was completely encircled by instruments, which music lovers found insulting. With the multitudes and the audiophiles and the music lovers turned off by quad's excesses and shortcomings, who was gonna buy it? It expired with barely a whimper in 1975.

But that was then, this is now . . .
. . . and surround-sound is no longer the alien concept it was 28 years ago. THX, AC-3, and DTS promote it every day in hundreds of movie theaters across the land, Dolby Pro-Logic in countless home theaters has created a perception that surround-sound is worth investing money and space in, and surround-sound has caught on as an indispensible feature of every car stereo.

But is anyone going to use it for CDs at home? If they do, it had better not be via Pro-Logic, because Pro-Logic wants an encoded recording with a consistent surround signal that the steering can grab hold of. Unencoded recordings, which are the vast majority of CDs, contain extractable surround in the form of hall reverberation, but consistent it is not. Reverberation is random energy, which is why roughly half of it at any given moment is left/right anti-phase. But that randomness causes Pro-Logic to steer here, there, and elsewhere with dizzying unpredictability. This is why many surround decoders offer a Music Surround mode that extracts the center and surround signals but doesn't apply steering to them. It's also the reason consumers who've tried using Pro-Logic with music recordings have declared surround-sound an abomination.

The Circle Surround difference
Initially intended for music only, Rocktron's Circle Surround was slated to be the ultimate matrixed surround system. The original design included a studio encoder, panner control, and steered decoder, plus ambience-extraction circuitry for use with unencoded recordings. It was decided to follow the Dolby Pro-Logic model because it's simpler and more stable than some other matrix systems, but to do so without buying into Pro-Logic's shortcomings: its inability to steer in more than one direction at a time, its tendency to get confused by unencoded program sources, its loss of stereo separation in the presence of a strong center-front signal, and its dimensionless, monophonic surround. Then Rocktron elected to do all this the hardest way possible: in the analog domain.

Steering must respond fast enough that it will sound as if it starts in the channel it belongs in (instead of starting here and then hopping to there), and slowly enough that the steering doesn't cause audible pumping. But the conflicting requirements make any combination of time constants a compromise. The advantage of digital matrix-surround processing is that it allows the signal to be delayed in a holding buffer while the steering logic "decides" whether a subtle surround cue is simply a random phase anomaly that should be ignored, or whether it's the leading edge of a sound that really is supposed to be steered somewhere. This can yield much more accurate steering of surround effects than is possible with analog processors that make their steering decisions in real-time, and must therefore be a little more reluctant to respond to borderline cues.

Rocktron had a better idea. They divide the audio band into three control bands—low, middle, and treble—and allow each to steer the signal where it apparently belongs. This way, treble sounds can be steered very rapidly, midrange sounds less rapidly, and so on, and each can be sent in a different direction. CS's steering can change direction as the spectral content of the sound changes from moment to moment, so a vocal sibilant might come from the center, a spoken vowel from the left, and the rumble of a bus from the right, giving an impression of discrete surround.

As with Pro-Logic, any information that appears out-of-phase in both front channels is directed to the surround speakers, but Circle Surround's logic adds an enhancement. The out-of-phase signals are checked for left/right channel balance and, if the signal is equally strong in both channels, it's steered equally to both rear speakers so as to image at center rear. If the surround signal is stronger in, say, the front left channel, it's steered toward the left rear speaker, and vice versa. This is how Circle Surround delivers true stereo in the surround channels as well as up front, and how it is able to simultaneously direct three spectrally different images in any directions around the listener. (This is what the 5s in the CS's 5.2.5 designation refers to: the ability to reproduce signals from five channels.)

Pro-Logic's propensity for losing stereo spread in the presence of a strong center signal was a result of Hollywood's conviction that keeping dialog onscreen was more important than good stereo. Sounds that are hard-panned to left or right cause the center channel to shut down, but the rest of the time any surround-channel signal attenuates the front left and right channels to a degree dependent on center-channel loudness. This means the left and right stereo channels operate somewhere between full-level right or left and some degree of attenuation, relative to the center, so they can never deliver full stereo separation. Instead of allowing the center channel to mindlessly attenuate the left and right channels, CS controls the center channel depending on its signal. A strong center signal is reproduced at full level, but reduced center-channel activity reduces center-channel volume by half. Hard-panned left or right signals will cause the center channel to shut down completely.

Installation
My sample Circle Surround CS 5.2.5 decoder came with a demo CD containing a number of CS-encoded cuts plus several unencoded selections from conventional music recordings (including some dubs from LP). So before installing the 5.2.5, I listened to the demo disc through my reference Lexicon DC-1 decoder, using its Logic 7 mode. (Like Circle Surround, Logic 7 is claimed to be capable of extracting true stereo surround from matrixed sources.) The demo disc was a little scary, providing a not-very-reassuring glimpse of what our surround-sound future holds. Some of it was raucous sound-effects material, depressingly reminiscent of the stuff that heralded the advent of stereo, and most of the musical material, while directionally and spatially interesting, was instantly forgettable.

The Circle Surround 5.2.5 is not very versatile. Rocktron seems to see it as a free-standing device—the main control center of the system—but I can't see that happening because it has only one stereo input, and the last person who listened exclusively to one audio source passed away in 1994. More serious is that it can't be integrated into a full-blown surround-sound system either, because it has no 6-channel pass-through inputs, only outputs. It monopolizes all the amplifiers, preventing the use of any other surround decoder(s). In most cases, the 5.2.5 will have to be fed from the tape output of a stereo controller that will be used for source switching. And there's no high-pass provision to allow for using skimpy front and surround speakers. All channels are fed full-bass signals, which pretty much rules out the use of a satellite/subwoofer speaker system, although the subwoofer output does include an 80Hz low-pass filter.

---

Footnote 1: Strictly speaking, the surround signal lagged the left stereo channel by 90° of phase and led the right channel by 90°, so that it was anti-phase between the stereo channels but only a symmetrical 90° out-of-phase with either stereo channel.

Footnote 2: Even with steering, there's usually still some leakage of front left and right signals into the rear, so a small time delay is put in the surround channel, to harness the ears' tendency to localize sounds in the direction they are first heard from. CS claims their steering is so good they don't need it, but they do use this time delay in the Video mode.