Thrice Shy: Multichannel Music Formats Further Considered
By Daniel C. Sweeney, Ph.D.
Freelance Contributor to The Audio Critic
This is the second part of a critical examination and evaluation of mutichannel audio technology, with regard to both hardware and software, present and future, fact and hype.
In this, the second in a series dealing with the issues of multichannel music reproduction, I will examine what are perhaps the two key considerations for music lovers: (1) the adequacy of multichannel recording techniques in terms of what is known about localization by the human ear and (2) the fidelity of multichannel playback systems according to the same criterion. I will also have a few words to say concerning the use of music systems in video applications.
Antecedents: the Two-Channel Models
Within the realm of stereo recording practice, several distinct approaches have been developed for deriving differential inputs from the wavefronts representing the performance itself and then distributing the information so obtained between two channels, so as to create spatial effects upon playback. Without being in any way exhaustive, one can cite various binaural techniques (arguably not stereo at all in strictest definition), M-S stereo recording (also known as intensity stereo), the crossed figureeight method associated with Blumlein, angled cardioids, spaced omnis, and synthetic techniques whereby a multitude of microphones is assigned to a number of musical instruments and voices in any of an infinite number of arrangements. It should also be noted that the last technique or family of techniques, ubiquitous in popular recording and derisively dismissed by purists as multiple mono, can be combined with minimalist techniques so that the minimalist core engenders the overall spatial perspective, while the additional microphones provide accents, so to speak.
While cogent arguments have been persistently advanced to the effect that only the original Blumlein tech- niques constitute true stereo, such positions have retarded not at all the diversification of two-channel recording techniques, and today extreme pluralism characterizes musical recording practice for all forms of music. And, correspondingly, extreme pluralism also characterizes the playback array, with loudspeakers exhibiting all manner of directivity characteristics vying for consumer allegiance, and consumers themselves deploying such speakers in all manner of listening spaces.
With such extreme diversity existing at both ends of the signal chain within the two-channel realm, little in the way of precise matching of pickup to playback system is possible, though on a strictly rational level there is indeed something to be said for matching speakers to microphones —for example, playing spaced omnimiked recordings over omnidirectional loudspeakers, or crossed figure-eight Blumlein recordings over toed-in dipoles. Still, we enjoy satisfactory listening experiences despite this rampant heterogeneity, and we are accustomed to expect acceptable stereo without too close attention being paid to the establishment of correspondences between the recording setup and the playback system.
From Two to Many
Here the question must be asked, what does such heterogeneity within the two-channel universe portend for multichannel? Will similar centrifugal tendencies manifest themselves in the age of multichannel, and to what effect?
Damned good question, in fact the key question in the whole discussion, because it immediately gives rise to at least two other crucial questions. Just how do we use these extra channels when we can’t come to any consensus on how to use two? And what happens when we try to juggle intensity and phase relationships among sev-eral channels?
Now one might suppose that with all the confidence being expressed within the hardware industry about the benefits of multichannel, definite answers to such questions would have been framed. But sadly the opposite is the case. The general level of insight on the problems and possibilities of multichannel is lower than was the case a quarter century ago when the abortive quad revolution was launched, and of all the many individuals I’ve interviewed on the subject over the last year, only two, John Eargie and Tom Holman, seemed to have steeped themselves thoroughly in the theoretical underpinnings of multichannel recording and to be proceeding on the basis of a well-defined model of human hearing. And, interestingly, both individuals were quick to admit that much remains to be learned regarding recording for multichannel playback.
In other words you, gentle reader, will be part of a grand experiment should you add channels to your music system at this early date. Multichannel recordists of various persuasions will present you with a vast mélange of program material embodying a dizzying profusion of “takes” on the deployment of five channels, and you’ll be left with the task of sorting it all out.
But lest this scenario appear unrelentingly grim, let me hasten to point out that multichannel boosters—who include nearly everyone on the hardware side—are quick to object that subjectively successful two-channel recordings have been made with any and all of the common techniques, and that therefore no reason can be adduced that five-channel formats should not prove equally adaptable.
On the surface that argument seems plausible enough, but ignored is any due consideration of the fundamental change that occurs when the listener is surrounded by loudspeakers rather than facing a pair of them. Ignored as well is the much greater precision afforded by five discrete channels over two, which is to say that the information density of the recording becomes correspondingly greater, and directional cues are brought into play that scarcely exist in a stereo system. And finally there is the weighty issue of multichannel pans, where phase and intensity differences are juggled to produce phantoms between adjacent pairs of speakers, or even steered into the interior listening space described by the circuit of speakers, rather than being confined to the periphery as is always the case with two-channel stereo.
What I’m saying is that we may find two-channel stereo acceptable precisely because it is a loose and lossy approximation of what we actually hear in an actual performance space, and further that this approximate nature of two-channel may be what permits the multiplicity of techniques. To cite an analogy, it is a far more difficult and exacting task to make a 3-D motion picture that provides an acceptable illusion of reality than it is to accomplish the same goal with conventional single-camera flat- perspective movie footage, the 3-D format imposing upon the film maker a myriad of highly frustrating constraints as a price of the added dimensionality. At least some of the available evidence suggests that multichannel may be similarly problematic and that five-channel may prove to be less than the unalloyed enhancement that industry spokesmen would have you believe. But I mention this only as a possibility. The fact is that we simply don’t know with any certainty how five-channel will accommodate different recording techniques, because nobody has enough experience to speak definitively on the subject.
If this statement appears remarkable, consider the fact that a mere handful of recordings were specifically miked for multichannel back in the quad era. Almost everything released in quad was simply remixed from the same multitrack master used for the stereo release, with a quad pan-pot being used in an attempt to place sounds at intermediate positions between speakers. True, during the last few years, quite a bit of mixing has been done for movie formats utilizing five full-range channels, but the recording of motion-picture sound tracks diverges in both its aims and its techniques from the modes that define purist music recording. Movie sound tracks place their stress on arbitrary effects, not on the recreation of a specific performance event, and, moreover, film sound is characterized by extremely layered compositions far exceeding most multichannel pop recordings in density and complexity. Thus it is fair to say that, in regard to music, the art of recording for multichannel is still in an embryonic state of development.
I should mention here that in the researching of another article on this same general subject I went to great lengths to interview recording engineers with a clear commitment to multichannel playback and with extensive experience in the same. Unfortunately, I found that only a small minority of recordists have a clear commitment, and almost no one has much experience, John Eargle and Brad Miller being just about the only individuals who can really be considered veterans. Lots of people have experimented at one time or another, but these two individuals are the only true gurus, and, since neither is regularly employed by major labels anymore, you won’t get the benefit of their experience—if 5.1 or some other multichannel standard ever does become the dominant music format.
Bedrock: the Sampling of Acoustic Space with Two or with Multiple Channels
A sense of space is created on a recording principally by two means: by capturing localization cues that enable listeners to identify the positions of instruments and voices within a recreated acoustic space, and by capturing the pattern of reflections and reverberation that conveys the hall sound (even to the approximate dimensions of the recording venue). Bear in mind that these two perceived aspects of spatiality are not at all the same thing and should be clearly distinguished in discussions as to how an audio system recreates the space of a performance, though, unfortunately, they are often confused in subjective discussions of loudspeaker imaging.
First let's discuss sound localization. I will assume that most individuals reading this journal are aware that interaural arrival-time differences and interaural amplitude differences constitute the most significant localization cues for humans (if you’re unfamiliar with terms, consult any standard text on recording). Although most members of the audio press appear to remain ignorant of the fact, there is a large body of evidence for the existence of a third important localization cue known as the HRTF (head related transfer function), a psychoacoustic phenomenon resulting from the differential diffractive effects imposed by the structures of the head and outer ears on sounds arriving at varying angles of incidence. Less conclusive evidence may be cited for the existence of one other major cue arising from interaural phase differences.
Several other localization cues have been posited as well, but the above mentioned three are surely primary, with phase differences occupying an insecure and ambiguous position.
In stereo recording, amplitude cues are stressed, and often they are the only cues provided; however, among the classic techniques, spaced omnis and angled cardioids capture arrival time differences effectively as well. Interestingly, coincidental techniques such as crossed figure eights or M-S normally do not capture arrival- time differences. Significantly, no microphone techniques other than binaural techniques utilizing dummy heads with dummy ears can capture the HRTFs, though these can be synthesized by appropriate postequalization.
Now what about recording for multichannel playback? Do the recording techniques devised for the new media improve upon the situation and provide full employment for all localization cues?
The answer to that is a qualified no, and to understand why this is the case one must realize that for all the hoopla surrounding 5.1, relatively little research has been devoted to tapping its possibilities. (Here it must be said that various individuals and organizations such as the ARA [Acoustic Renaissance for Audio], Tom Holman, and John Eargie have proposed an open standard that would allow for selectable channel number as well as variable bit and sampling rates.)
In fact, no really new recording techniques have yet been developed for 5.1, though John Eargie has suggested some of the directions in which recordists might proceed. The two individuals who have been most active in 5.1 music recording, Brad Miller of Mobile Fidelity International and Tom Jung of dmp, both use techniques dating from the quad era, and it is safe to say that recording for multichannel, such as it is, represents no significant advance over the quad experiments. People are talking about a revolution, but the basic recording technology is stuck in the ‘70s.
And just how did our distant ancestors back in the ‘70s record for four-channel playback?
Several formulae are mentioned in the literature, including three-channel matrix techniques which I will disregard here as having no direct application, but basically two approaches obtained.
The first approach, owing much to purist stereo techniques, was to use a fairly conventional minimalist miking setup such as angled cardioids up front and then to place another pair of mikes—often widely spaced omnis —in the rear of the hall. A variant was the use of cardioids in the rear, facing away from the back wall. Most of the purists thought that rear channels should be chiefly used for reproducing ambience, and some even suggested recording left rear and right rear out of phase to inhibit the formation of a rear phantom image.
The second approach was an extension of the panpotted multitrack methods which had become the norm in popular recording by 1970. As in stereo multitrack recording, each instrument and voice was assigned a separate track, and performers might even be recorded apart from one another in booths or gobos. The performance, which could be entirely synthetic, was essentially assembled during mixdown, and performers were distributed about a synthesized acoustic space via the mixing console and the pan-pot. Extending the approach to quad meant simply mixing down to four rather than two channels and involved similarly arbitrary decisions as to what went where.
The first approach, building on classic stereo recording techniques, was explored by many eminent audio professionals, including Eargle, F. Alton Everest, Peter Scheiber, and E. Roerback Madsen, but it chiefly found expression in their published ruminations rather than in actual recordings. Later though, at the close of the decade of the ‘70s, one particularly dogmatic form of the purist approach represented by the Ambisonics camp did result in actual recordings, albeit only a few dozen altogether. Unhappily, the cult tendencies and tendentious proselytizing of its advocates served to marginalize the Ambisonics format, and today it seems unlikely to exert much influence on five-channel music recordings. Nevertheless, the purist approach lives on in Dolby matrix recordings put out on the Delos label and in the Mobile Fidelity International recordings engineered by Brad Miller, and it is fairly certain to see some future applications at least in the realm of audiophile recordings.
The second approach, i.e., the synthetic approach, resulted in recordings which, by and large, significantly detracted from the credibility of multichannel. Rear channels were given equal weight to those in front, and placement of instruments in back of the listening space was frequently sought. Circumferential dynamic pans were also attempted. This type of treatment was the norm in the hundreds of quad remixes of popular releases.
During quad's heyday, such as it was, these aggressive surround mixes invariably came to naught due to the very poor rear-channel separation of the dominant SQ format, and today quad defenders are quick to marshal the low-separation argument to explain all of the failings of quad as a popular medium. Certainly low separation effectively thwarted any attempt to place sources at points beyond the frontal soundstage, and yet I believe that limitations of the low-separation matrices were not the only factor in the mass rejection of quad in popular music. The real problem lay with the listener.
If current scientific theories of human localization are even remotely correct, the interaural amplitude, arrival time, and phase differences only permit pans in front or in back of the listener. Images cannot be panned between a front left and rear left speaker or a front right and rear right speaker because, in the case of amplitude differences, the opposite ear is completely shadowed for all positions, while arrival time and phase differences vary by too small a degree to produce stable phantoms. Furthermore, in the case of pans behind the listener between rear left and rear right speakers, the precision of phantom localization appears to be far less than in the front quadrant. Simply put, four- and five-channel formats, used like pan-potted stereo, demand a kind of listener who doesn’t exist, a listener whose localization abilities are uniform throughout the horizontal plane.
If humans indeed lack the ability to form stereo phantoms in all quadrants, then the only option open to the recordist who wishes to emphasize the rear channels and provide the multichannel convert with his money’s worth is to localize sounds at the rear speaker locations— hard left back and hard right back. And that will give us the same hole-in-the-middle effects and ping-pong lateralization that was popular in the earliest days of stereo. None of which augurs well for a multichannel renaissance in the field of popular music.
Not unexpectedly, most marketing people in the hardware industry grow very cross when presented with such information because for them pronounced backchannel effects are crucial in the promotion of the format. Indeed I’ve been told by such types that I’m disloyal to the industry for even presuming to mention scientific studies that produce these negative findings, and that even if the studies are accurate, DSP will solve the problem. But it seems to me that all the industry loyalty in the world isn’t going to change the nature of human perception, and if pop-music producers attempt to make recordings that flout the laws of psychoacoustics, as they did in the ‘70s, a successful reintroduction of multichannel becomes unlikely. And as for DSP solving the problem, the problem lies in the way human beings construct a soundfield from a finite number of channels. What they’re asking for is a software fix for a wetware (neurological) limitation. And that human limitation is worth examining.
What normally happens when one attempts to pan in every quadrant is that the soundfield becomes highly unstable, with the formation of inside-the-head phantoms similar to those experienced in headphone listening and sounds seeming to pop up in unpredictable places like shells falling on a battle field. The effect is interesting enough on first listen, but whether it will continue to beguile even the mass of unsophisticated consumers is questionable. It has been said that no one has ever lost money by underestimating the taste of the American public, and here we have an excellent test case.
To Market, to Market...
Obviously I feel very strongly that the aggressive use of rear channels in pop recordings will result in crassly gimmicked presentations, with little appeal to serious listeners. But of course serious and sophisticated listeners form a tiny minority, while the Beavis and Butt-head sensibility is overwhelmingly dominant. Knowing this, fivechannel boosters suggest that gimmicked, ping-pong presentations will not only prove satisfactory but will eventually be insisted upon by the great American listening public.
Perhaps they're right, and such multichannel monstrosities will become the norm in music reproduction. But in the sunny pronunciamentos of the 5.1, camp I see several key questions being begged.
One of those questions has to do with the way listeners will set up systems in their home. Multichannel playback systems are not forgiving of haphazard placement, and the familiar consumer response to stereo—put one speaker behind a sofa and the other on a bookshelf— is not going to provide much of anything when extended to five full-range channels. Of course today, if industry sales figures are to be trusted, most of the new audio systems going out the door are already multichannel combinations, albeit designed for Dolby matrix playback, but no one knows exactly how these systems are set up in actual use or how many such systems are really operating at anything close to their potential. How many people, for instance, actually connect surround speakers to their Dolby Pro Logic AV receivers? And of those that do, how many have the speakers where they can produce a surround effect? Clearly such questions demand answers because if 5.1 is presented on systems consisting of three front speakers around the fireplace and a couple of eightinch- high rear speakers somewhere on the floor in the back of the room, the results will do little to justify the overheated hype being generated by the industry.
Other unanswered questions pertain to marketing considerations—considerations which may ultimately be more important.
In introducing any multichannel playback format, one inevitably has to deal with the matter of backward compatibility, always a major concern to record producers in the past. Ensuring full stereo and mono compatibility in a five-channel mix isn’t easy, and yet radio stations, which are not equipped to transmit 5.1 and may never be, are not going to stand for incompatible mixes.
There’s also the issue of industry support on the software level. The fact is that no prominent recording artist or music producer has come out in support of 5.1 for music, and no major label has said a word about issuing music recordings in that format. The support is coming strictly from the hardware side and from a few individuals working for small audiophile labels.
Hardware supporters of 5.1 counter by arguing that five-channel music recordings can simply ride on the coat-tails of video, and they point out, correctly, that 5.1 is already available on laser disc. But that’s a weak argument because laser disc itself is a format with negligible penetration after more than fifteen years on the market. The popular-music recording industry is not going to content itself with addressing the universe of laser disc supporters. The numbers just aren’t there. For 5.1 music recording to piggyback on video, either HDTV (which will support discrete multichannel) is going to have to come on a lot faster than anyone expects, or the as yet unreleased [now just barely released—Ed.] DVD (digital versatile disc) is going to have to succeed magnificently.
Most 5.1 boosters in fact pin their hopes on DVD, but its success cannot be considered a foregone conclusion —witness the failure of Philips’s current 5¼-inch digital video disc, which bears many points in common with DVD, industry protests notwithstanding. And, really, why should either small-disc video format succeed? The initial target market of early adopters already has a video disc, the venerable 12-inch laser disc, and it already has 5.1 on the 12-inch format. Why then should such buyers opt for a new format and one that is burdened with the uncomfortable issue of video compression to boot? And why should the specialty retailers who have loyally supported laser disc for two decades come out in favor of something designed to obsolesce the older format instantly and put them in ruinous competition with mass merchandisers, who are the primary target for DVD?
Representatives of the Japanese majors blithely brush such objections aside by saying that there’s plenty of room for multiple formats in the specialty market, but the history of the industry has in every instance proven otherwise. What this means in terms of marketing strategy is that the DVD forces are going to have go straight to the mass market, push the hell out of the thing, and somehow convince all the video store owners to get into dual inventory or perhaps just scrap tape altogether. And maybe that will happen. Maybe the Japanese and the software manufacturers will initiate the aggressive—no, giveaway —pricing necessary to gain the consumer critical mass quickly, though they’ve certainly never done so with any previous format. Anyone want to pay me to do a video-software retailer survey on the subject?
In all the years I’ve been writing about consumer electronics, I’ve never encountered a tougher call than 5.1. I wish I could predict its fate with certainty because I think the response of the music lover ultimately has to hinge on market considerations involving this new format. If DVD fails, then where will 5.1 be? If Delos, dmp, and the like are the only ones who are going to be issuing 5.1 music recordings, are you prepared to revamp your music system? And what if some of the audiophile labels go for the Dolby AC-3 version and others for DTS? That’s already happening now, with John Eargle favoring Dolby, and Brad Miller and Tom Jung solidly behind DTS.
THE PLAYBACK CHAIN
Might we now stand back for a moment and attempt to summarize the situation at the software end? Recording techniques have not solidified, nor is past recording practice for multichannel generally in accord with what is currently understood about human localization capabilities. And even where recording techniques are informed by present-day science, the resources available to the five-channel recordist appear to be insufficient to map a three-dimensional acoustic space with total verisimilitude.
Correspondence between pickup and playback systems has scarcely been investigated systematically at all, either during the quad heyday or at the present. It goes without saying that detailed recording/playback standards are not in place—except in the case of the nonstarter Ambisonics format. (Ambisonics is being heavily promoted by the ARA group as the recording standard for the proposed audio-only high-density disc; my bet is it won’t fly, but that’s another article.)
Finally, a well-organized marketing campaign is not in evidence. Instead, 5.1 enthusiasts have simply assumed the inevitability of multichannel dominance and have not gone to the trouble of enlisting music-industry support, nor of educating the public beyond placing puff pieces in a few audio buff books.
Compared to the cogent, disciplined, and energetic promotional activities of the Compact Disc Group, the efforts of the 5.1 camp have been amazingly weak. Such slackness speaks highly of the latter group’s confidence but does not reflect historical realities. During the last fifteen years nearly twenty new consumer software formats have appeared, but, of those, only a couple have succeeded, suggesting that concerted marketing efforts are well-nigh indispensable.
In sum, the software situation of multichannel music playback is disorganized in the extreme, at least as disorganized as the quad revolution was twenty-five years ago. True, in quad the format rivalry was infinitely worse, though SQ was heavily dominant, but, on the other hand, software support was certainly there along with a powerful publicity campaign. If quad failed with the full resources of both the software and hardware manufacturers behind it, what are the chances of 5.1 with admittedly better technical specifications but with little marketing impetus? I don’t know.
My guess is that a real conversion from stereo to multichannel is not in the offing as regards music programming. In the longer term—say, five years from now—it may occur, but absent a coordinated marketing campaign I can’t see how it can happen now.
Of course, I could be wrong. Of the two runaway format successes of the last fifteen years, Dolby Pro Logic did in fact succeed without the disciplined promotional efforts that marked the progress of the other winner, the compact disc. But Dolby encoding was already present on most stereo movie software anyway and was fully backwards compatible, and thus Pro Logic could be promoted successfully from the hardware side alone, since the software was already taken care of. A similar situation does not obtain with 5.1.
We shall see. My advice is to ponder long and hard before rethinking your music system for five discrete channels. It may be a long time coming.
Even if 5.1 remains a minority video-only format— a distinct possibility--the enthusiast listener is still confronted with the dominance of matrix surround throughout the video realm and with the decision of whether to attempt to use a single system for stereo music listening as well as for home-theater sound reinforcement.
Much ink has been spilt in blathering on the suitability or lack thereof of audiophile speakers for movie sound and, conversely, home-theater rigs for music listening, and indeed the contention has grown rather ugly of late, with old-line audiophiles dismissing the home-theater crowd as tin-eared Philistines, while the latter mark the former as irrelevant fogies. And while one might wish to stand above the fray, doing so becomes difficult in a journal whose allegiance is clearly to the cause of promoting serious music. Still one can attempt to avoid the acrimony and to concentrate on the design issues, which should be the focus of the discussion.
Can One System Do Both?
To begin to answer the questions concerning the efficacy of dual-use systems, let us pose a hypothetical that surely applies to many of our readers.
Let us suppose that you as a music lover already own a high-priced, well-reviewed pair of stereo loudspeakers. It could be a pair of Avalon Ascents, or Quad ESL-63’s, or Infinity Epsilons, to name just a few currently popular contenders embodying various design models. Let us further assume that you have a large sum of money in amplification. How might you retrofit this system for video applications, or need you retrofit it?
Or might a better strategy be simply to build an entirely separate system for video?
Or just sell off that audiophile relic and buy an ambitious dual-use system, such as the JBL Synthesis or the Snell THX Reference?
Or is a retrofit even necessary when an audiophile system is pressed into service as a video PA?
Certainly, you can play stereo video material in straight two-channel form instead of decoding the center and surround channels, and that sort of makeshift would relieve you of the task of making modifications in your existing system. But, at the very least, such a strategy would rob you of surround effects and would compromise lateral pans across the front for all but the listener in the sweet spot. In my view, an optimal dual-use system perforce must include multiple speakers.
That being the case, could such a multispeaker system be built using an existing stereo music system as its core? Undoubtedly something could be done along those lines, but I believe that the results would be less than ideal.
To understand why this might be so, one must examine three areas: speaker matching, program requirements, and room interactions.
In terms of speaker matching, the user faces two distinct sets of problems-- integrating a center speaker with the stereo pair and matching the surround speakers to all three front speakers.
Most people who've seriously investigated multichannel playback agree that the frontal array takes precedence, and that matching of front speakers is of paramount importance in achieving credible soundstaging. Indeed, listening experiments indicate that when the center speaker is not timbrally matched to left and right, pans—both static and dynamic—are seriously degraded. On this basis, center-channel augmentation would appear to require a single speaker of the same model designation as those used for stereo, or else something that is very closely matched in timbre.
Obviously, with mirror-imaged music speaker systems you're screwed. There's simply no easy way to add a center channel with the right characteristics. But even when both speakers are the same and the manufacturer is willing to sell you a single speaker, you may have problems.
The problems arise both from the differing natures of multichannel movie sound and stereo music programming, and from the loudspeaker directivity characteristics best suited to either application.
What follows here is controversial but cannot be avoided on that ground, as it is central to the whole discussion.
Stereo music software, particularly the purist recordings of acoustical music which we cherish, contains three basic types of spatial information: the direct sound at two locations represented by the stereo microphone pair, the early reflections from the front of the hall (generally captured by the same pair of microphones), and the diffuse reverberant soundfield from the back of the hall, which often is picked up by auxiliary ambience mikes. In consort these three types of spatial information suggest both the dimensions of the acoustical space and the placement of performers within it.
In two-channel stereo all of this information is folded together into two channels, where it is imperfectly decoded by the ear-brain. On the other hand, in film recording ambience is allotted to the surround channel, while early reflections are generally not an issue, due to the synthetic nature of the recordings. In five-channel music recording the channel allocation of early reflections is anyone’s guess.
In actual two-channel playback in the home, the natural ambience in the music recording is usually supplemented by room reflections stimulated by the off-axis output of the loudspeakers themselves. The alternative, the LEDE (live-end-dead-end) concept whereby early refections from the front of the listening room are suppressed and supplemental simulated ambience is provided by rear-wall diffusion, has much to recommend it from a theoretical perspective but has never gained much popularity outside of the recording-studio milieu because of the expensive and ungainly room treatment accessories necessary to achieve it. (One might add that the speakers deemed appropriate for LEDE applications are such are scarcely to be found in the consumer marketplace— combining, as they must, phase linearity with narrow directivity characteristics; the old Win SM-1O fits the bill nicely, but nothing else of which I am aware.)
Here I must point out that, strictly speaking, the front-wall, side-wall, floor, and ceiling reflections engendered by a wide dispersion stereo pair are not really analogous to the diffuse soundfieid one experiences in a large listening space because of the uneven temporal and spatial distribution of acoustic energy. But in subjective terms the two are sufficiently alike to create favorable reactions among the majority of sophisticated listeners. Within a normal listening environment, wide-dispersion stereo speakers are generally deemed more natural-sounding than their beamier counterparts, though they cannot equal the imaging precision of narrow-directivity designs. There’s also evidence that the acoustical crosstalk produced by wide-dispersion speakers is conducive to the creation of a more stable phantom center.
In the two-channel stereo mode, where such matters have been extensively studied, the off-axis output of the loudspeaker exerts several significant effects in addition to affecting apparent source placement and spatial perspective. Off-axis output influences listener perceptions of overall tonal balance, determines the width of the stereo window, and, in the lower frequencies, affects the coupling of the speaker with the room and the incidence of room modes and wall resonances. Furthermore, for reasons we needn’t examine here, optimization of directivity characteristics for one purpose, such as achieving a fairly uniform perceived frequency response in varying domestic listening environments, is almost always had at the cost of compromised performance in other areas, hence the endless unresolvable debates on the relative merits of dipoles, bipoles, monopoles, and omnis. In toto, off-axis response virtually defines the subjective character of a loudspeaker and for that reason is subject to the most careful attention from competent loudspeaker designers.
Now, in multichannel systems the issue of directivity becomes vastly more complicated (as if it weren’t complicated enough) due to the fact that ambience is chiefly assigned to side or rear speakers, a fact which, according to some, eliminates the need for or desirability of wide-dispersion speakers up front. Indeed, according to the folks at THX, who have probably researched the subject more extensively than anyone else, wide-dispersion designs for the front speakers are to be positively avoided in a multichannel playback system.
So what happens to all those carefully crafted omnis, dipoles, and bipoles which have found such favor with audiophiles in the past? According to Tony Grimani of Lucasfilms, such designs are apt to wane in acceptance as more and more customers warm to the idea of dual-use systems. In other words, you can consign your Quad ESL- 63’s to the dustbin of history. Instead you’re enjoined to embrace the THX concept for front speakers, which is essentially that of a short line source—albeit only an approximate one.
And what happens when you use such narrowdirectivity speakers for stereo music listening? That’s where opinions differ. Those in THX fold, which includes most of the mainstream audio press, assert that THX speakers surpass conventional designs in all applications, but there are numerous dissenters, including many members of the subjective press. One can certainly argue that narrow-directivity designs, by avoiding floor and ceiling bounce, are inherently capable of better fidelity, but that isn’t going to make people love them. The fact is that THX speakers sound identifiably different than the quasi-point-source radiators that comprise the bulk of the well-regarded audiophile music speakers, and that, I think, accounts for the rejection of THX designs in the audiophile press.
I find it interesting in this regard that JBL, in designing their supposedly state-of-the-art Synthesis systems, endowed the front speakers with dual driver arrays—one set for video and one for music—and that the directivity characteristics of the two arrays differed markedly. And the fact that Floyd Toole has been involved in the design of the Synthesis series gives us reason to believe that the notion of different directivity characteristics for different program material is not altogether unfounded.
So what does all this tell us about the feasibility of dual-use systems?
I have already indicated the difficulties of modifying a traditional music system for video use. As for scrapping the music system and going for a designated dual-use system, such as any of the THX-certifled systems, I’d say audition the dual-use system carefully on music before you do so. I’ve never heard a video system that I considered an ultimate music system, but that’s a personal opinion. Most of my peers would disagree. I would note, however, that video-oriented speakers, aside from having directional characteristics that are arguably ill-suited for two-.channel reproduction, never seem to approach the build quality of the best audiophile speakers. Really premium drivers are not used, cabinet quality is soso, and crossover components are not of the highest quality. Of course, there are those who will argue that topquality drivers and inert cabinets are needless extravagances, and that no one can hear the difference between a carefully designed system using average components and one where cost is no object. But then, if you believed that, you'd have never bought audiophile speakers in the first place.
Terra Incognita: the Anterior Portion of the Listening Space
Now what about the rear speakers? For some reason the industry has fixated on the rear (Freudian overtones here?), and both speaker manufacturers and the audio press have become embroiled in two related disputes—whether the rear speakers should be of the same type as the fronts, and whether the rears should produce an output characterized by a preponderance of direct or diffuse sound.
Lucasfilms THX has taken a very clear and carefully reasoned position on the matter, advocating diffuse di-polar rear speakers having directional characteristics that are obviously different from those of the front triad. On the other hand, multichannel advocates with roots in the quad era, such as Gary Reber (editor of Widescreen Review) and Brad Miller of Mobile Fidelity International, insist that like speakers be used in all positions. Certain prominent high-end speaker designers, most notably Jim Thiel, also support the notion of like speakers all around.
While the dispute may appear to represent a classic clash of reason against mysticism, pitting the modern, scientifically informed THX gang against an unlikely alliance of fuzzy-minded high-enders and quad-era relics, in fact the roots of the controversy go all the way back to the earliest days of quad, and the issues are nowhere near as simple as the above construction might imply.
Most early researchers of multichannel playback assumed that, as a matter of course, like speakers would be used all around. Quad was seen as an extension of twochannel stereo, and like speakers appeared desirable and even necessary. And this attitude was not confined to scientific ignoramuses by any means. The mathematically minded prophets of Ambisonics always assumed that like speakers would be used.
However, within the laboratories of those two stalwarts of the Boston sound, AR and Advent, experiments were conducted in which omnidirectional rear speakers were used along with conventional monopole front speakers. Whether Tom Holman, the architect of THX, was directly aware of those experiments I can’t say, but living where he did at the time he must have absorbed the notion that rear-channel output should be diffuse, a notion which found further expression in the a/d/s delay-derived rear channel, which often was assigned to small bookshelf speakers supplementing a main pair of towers in the front.
So who's right?
If matrix surround were all we had, diffusive rear speakers would clearly hold the advantage, despite the fact that the reverberant characteristics of a large hail can only be approximated in the home. But with 5.1 we have the capability of isolating sounds in left and right locations toward the rear of the listening space. Wouldn’t more conventional speakers be more appropriate in that instance?
Lucasfilms says no, but it seems to me here that their usual rigorous logic is less in evidence here. If THX rear speakers are specifically designed to thwart localization to the speaker position, which they certainly are, how in the world can they reproduce sources that are specifically intended to be localized at the speaker positions? Mr. Grimani suggested to me that, after initial experimentation, mixers will end up using the rear channels pretty much as they always have, eliminating the conflict, but I tend to think that the very presence of split rears in the format is always going to encourage their use. They simply add an additional element of razzle-dazzle.
But to return to our discussion of music systems, should one then attempt to supplement one’s expensive pair of stereo speakers by adding another pair of the same make and model in the back of the listening room?
I am certain that most of the manufacturers of expensive stereo music speakers would heartily encourage you to do just that, but I am not certain I can recommend that approach as either cost-effective or just plain effective. Experimental evidence suggests that there is indeed a problem of localizing even diffuse back-channel information at the rear speaker locations, which of course conduces to an unconvincing simulation of ambience. If this is the case, why then would you want to pay an additional five to fifteen thousand dollars for another pair of premium speakers which ultimately won’t give you what you want? Especially when you can simulate rear-channel ambience by much less expensive means, by investing in a set of diffusive room-treatment devices from such manufacturers as RPG, ASC, or Systems Development Group. ( Such devices will work with any speakers and eliminate matching problems.)
Multichannel music recording is in an early state of development and may well remain in that state for years to come. From my view, the scanty software and insufficient format specifications should indicate extreme caution in making sizable investments in hardware at this time—if music listening is your primary use for your audio system.
If you wish to enjoy video surround-sound tracks, you are better off obtaining a dedicated system than attempting to retrofit your music system. Should that system be THX? I am reluctant to suggest that no uncertified system will do, but I will say that THX-certified systems enjoy the special advantage of being through-engineered in every particular and thus being relatively easy to set up and calibrate. No non-THX system that I am aware of reveals the same attention to every aspect of the signal chain. On the other hand, I am not sure that the THX approach to rear channels is ideal for the new 5.1 format. Unfortunately, no one else is attempting to design for the new format with anything approaching the thoroughness of Lucasfilms, so you can’t assume that the THX approach has been automatically obsolesced by the appearance of the new format.
Finally, I would mention again that to obtain the full potential of multichannel playback we need more than five full-range channels. Six would be better, eight would be much better. Extra channels could be obtained by combining matrix technology with 5.1, and John Eargle has in fact suggested that approach. Recently, as I indicated, many of the backers of an audio-only version of DVD have come out in support of an open standard that would permit more than five channels at the discretion of the recording engineer. Nevertheless, the feeling is widespread that the dominating presence of video will dictate the general use of a five-channel configuration for music as well as movies.
On a practical level, any attempt to introduce multichannel in the perfectionist market must inevitably come up against critical cost constraints. High-quality stereo systems are already inordinately expensive, and very few individuals are likely to treble their expenditures to obtain the extra channels. Hardware manufacturers realize this and many propose to address the problem by lowering speaker quality on the grounds that listeners will be so awed by the extra channels as to ignore nuances of sound reproduction. Indeed, the same attitude is already manifest in the some formats themselves, where extra channels have been obtained at the cost of severe data compression.
I will end by noting that in my many conversations with manufacturers, both high-end and mass-market, the prevailing note in the discussions has been how 5.1 is going to make a killing in the marketplace, while, sadly, few individuals were inclined to mention the specifics of how they were going to exploit the potential of the format to create more convincing musical reproduction. In the midst of these discussions I was frequently reminded of my duty to remain unfailingly enthusiastic about the benefits of the new format—for the good of the industry.
Unfortunately, I remain at some level a consumer and I can’t help suspecting that in the frantic effort to cash in, musical ends may be slighted. But in the words of one manufacturer who will go nameless, “You have no right to put your own selfish listening pleasure above the good of the industry.” And I guess he’s right.
<< Previous Page | Article Index | Next Page >>