Upward Mobility: From 2 Channels to Surround
The first problem was that the technology just wasn't ready: the QS and SQ matrix systems worked by combining the four source channels into the two available from LP, whence they could never be independently recovered, while JVC's ambitious CD4 discrete system multiplexed two channels into each groove wall, using processes similar to those used in stereo FM radio. As a result, the lacquer cutting head was supposed to cut, and the replay stylus to trace, signal frequencies of up to 45kHz. If the LP system could achieve this at all, then the discs certainly couldn't be expected to do so for long.
As if this were not sufficient bad news, the early quad systems were also, to put it politely, psychoacoustically naïve. Pair-wise mixing between channels was capable of generating a virtual image between the front speakers, as we already knew from two-channel stereo. But it worked poorly for sound sources behind the listener, and didn't work at all at the sides. By the time a properly conceived soundfield re-creation solution arrived in the form of Ambisonics, it was too late: the first attempt at introducing surround sound was already in terminal decline. Along with all the technical problems, the existence of competing, incompatible systems had ensured that consumers sat on their hands.
The audio industry really ought to have learned the lessons and got multichannel right the second time around, but we all know that the cutthroat world of commerce often doesn't work that way. And so we find ourselves today with two multichannel music carriers—DVD-Audio and SACD—that are withering on the vine. The developers of HD DVD and Blu-ray have ambitions to replace DVD-A and SACD, but there is little evidence that these new competing media will be greeted with any more enthusiasm as carriers of audio per se.
This explains, at least in part, why there is currently a renaissance of interest in so-called upmix systems that generate, usually, a five-channel feed from a two-channel stereo source. The thinking is that, as multichannel audio carriers have failed to capture a wide audience but multichannel home-theater systems are now in many people's homes, effort should be made to enhance the experience of listening to "legacy" two-channel source material over such systems. In reality, this effort should have been made years ago—but better late than never. With copious digital signal processing (DSP) power increasingly available, some of it is being pointed in this direction.
All of which will surely sound sacrilegious to many Stereophile readers, either because they (mistakenly) believe two-channel stereo to be some sort of absolute and anything more elaborate as necessarily ersatz, or because they have experienced earlier generations of upmix processing and been unimpressed with the results. But by first recounting some of the long history of this endeavor and then describing some of the more recent developments, I hope to modify that view, and even ennoble the attempt to deliver a more spatially convincing and immersive experience from two-channel source material. It's no replacement for good multichannel, but it can be an improvement over conventional stereo that is worth having.
There are two distinct threads to this story that deserve separate consideration. The first relates to improving the accuracy and stability of frontal images by adding one or more speakers between the main stereo pair. Of particular relevance here is the simplest case of generating a single extra, center channel, since this ties in with the three frontal speakers of a standard 5.1-channel setup.
The second theme is the generation of rear-channel surround feeds that provide a more immersive listening experience. Let's be clear from the outset that what will be described here are what have been termed natural spatialization algorithms: that is, processes that do not add anything to the original signal in the way of artificial reverberation. Instead, what they attempt to do is extract reverberant information from within the stereo signal and present it in such a way as to enhance the listener's sense of envelopment. A key requirement in any such system that hopes to win audiophile approval is that the process of extracting and presenting the ambient information should do nothing to compromise the accuracy of the frontal soundstage, either timbrally or spatially. That's a tall order, but the signs are that it may at last be achievable.
Using a center speaker to enhance the reproduction of two-channel stereo signals is an idea that dates back as far as the early stereo experiments conducted at Bell Laboratories in the early 1930s. With film sound foremost in mind, the Bell team developed a three-channel, three-loudspeaker stereo system, the use of a center channel being favored because it better locked the sonic image to the screen image for viewers offset to one or other side of the auditorium, and better tracked the movement of sound sources across the screen. In the course of development, though, Bell compared its full three-channel stereo system with various alternatives, including two-channel recordings reproduced using a derived center channel (footnote 1). Paul Klipsch revived this idea in the late 1950s, performing a series of experiments (footnotes 2–5) that led him to conclude that using three loudspeakers to reproduce two-channel stereo signals offers superior results.
The method used by Bell and Klipsch to generate the signal for the third, centrally placed loudspeaker was as simple as could be. The left and right stereo channels were summed and subject to appropriate, perhaps variable, attenuation. So, using L to represent the left-channel signal and R the right-channel signal, the center speaker was fed k(L+R) , where k is less than 1 and represents the applied attenuation.
Although Klipsch was enthusiastic about the benefits of his 2, 2, 3 approach (two recording channels, two transmission channels, three loudspeakers)—claiming that "All the stereo recording systems are amenable to using a derived central channel and the expense is small for a large gain in stereo geometry"4—the idea of using a third loudspeaker did not catch on and the concept faded from view.
If this was due, at least in part, to a third loudspeaker and amplifier being financially and domestically unacceptable in the early 1960s, then the arrival of home-theater systems in the early 1990s—equipped with a center-channel speaker as standard—represented an opportunity to resurrect the concept. This was duly exploited when Michael Gerzon presented a paper describing an elaborated 2, 2, 3 system at the 91st Convention of the Audio Engineering Society in October 1991 (footnote 6).
Although the idea was inspired by home theater and the need to provide better locking of the audio and visual images over a wider listening area, that wasn't the only benefit of Gerzon's system. He also claimed improved image focus and reduced listening fatigue for a listener in the stereo hot seat. What Gerzon was suggesting with the latter claim was that a central loudspeaker can reduce the amount of unconscious work the brain has to do to construct a credible soundstage from just two channels.
Drawing on the psychoacoustic knowledge he had incorporated into Ambisonics, what Gerzon described in his paper were improved methods for replaying M channels over N loudspeakers, where N is greater than M. In the simplest case, where M=2 and N=3—ie, the generation of a center channel from a two-channel stereo signal—his scheme became known as Trifield processing when the rights to it were bought by Trifield Productions Ltd. in the UK. Trifield, in common with all of Gerzon's "optimum reproduction matrices," is frequency-dependent, which gives it much improved performance compared with Klipsch's simple bridged center channel.
Footnote 1: J.C. Steinberg and W.B. Snow, "Physical Factors in Auditory Perspective," Electrical Engineering, Vol.53, January 1934.
Footnote 2: P.W. Klipsch, "Stereophonic Sound with Two Tracks, Three Channels by Means of a Phantom Circuit (2PH3)," JAES, Vol.6, April 1958 (available from www.aes.org).
Footnote 3: P.W. Klipsch, "Wide-Stage Stereo," IRE Transactions on Audio, Vol.AU-7, July–August 1959.
Footnote 4: P.W. Klipsch, "Signal Mutuality in Stereo Systems," IRE Transactions on Audio, Vol.AU-8, September–October 1960.
Footnote 5: P.W. Klipsch, "Stereo Geometry Tests," IRE Transactions on Audio, Vol.AU-10, November–December 1962.
Footnote 6: M.A. Gerzon, "Optimum Reproduction Matrices for Multispeaker Stereo," JAES, Vol.40 Nos.7/8, July/August 1992 (available from www.aes.org).