Is It Real? Or Is It...
Well, though it didn't sound much like any recorded piano on any system I have heard, it was nevertheless a real piano, something a listener could determine from the shortest exposure. Similarly, when walking the corridors, you knew immediately that the sound of the guitarists in room 206—when you could hear it above the din from one of the nearby video surround-sound dems—was real.
What is it, therefore, that distinguishes real sound from reproduced so unambiguously?
Loudness has something to do with it. The 14' x 37' room in which Jim played was on the large side for a domestic system. As Enid Lumley noted, however, the piano had no difficulty in over-driving the room acoustic, something beyond the capabilities of the Klyne SK5/Krell KSA-100/KEF R107 system I was using for my own demonstrations in the same room. Yet measurements of sound-pressure level indicated that this excellent system, capable of sustaining 113dB peaks or so, would have no difficulty in keeping up with the real thing. A major difference, of course, and one which has always invalidated live-vs-recorded tests, is that, with the exception of omnidirectional or quasi-omnidirectional models (such as the Ohm Walsh 5 reviewed by Dick Olsher this month), other than at low frequencies a loudspeaker puts out its energy into a relatively restricted solid angle to its front. A real musical instrument, however, puts out energy in all directions; although the system above is probably putting out two acoustic watts into the room at full output (assuming the speakers to be around 0.5% efficient at converting electrical power into acoustic) which is around the same acoustic power being put out by the piano, the way in which that power is fed into the room is far more complex for the live instrument. In addition, live piano sound features a high ratio of peak-to-mean levels, whereas even a purist recording will compress those peaks somewhat, not to mention the inability of even superb conventional loudspeakers to avoid peak compression at realistic playback levels.
Another first-order way in which reproduced sound differs from real—one considered to be of primary importance by reviewers for the mainstream hi-fi magazines—lies in the way that instrumental timbre is preserved. In inexpensive systems, this will be a significant fault—on my car radio the other day I was astonished to find that what I had taken to be a somewhat thin-sounding soprano saxophone was in fact a clarinet—but on a true high-end system, such timbrel distortion should be minimal, perhaps diluting the difference between, say, French and German bassoons but otherwise allowing generic instruments to be differentiated.
In a thought-provoking article in this issue, contrabassoonist Lewis Lipnick argues even with this assumption, but I think that with a typical high-end system, second-order problems are more important.
Second-order effects? What is JA talking about? Does he mean harmonic distortion?
Not really. It has been known for decades that if harmonic distortion consists predominantly of low orders—second, third, etc.—and is not accompanied by musically enharmonic intermodulation products, listeners can tolerate surprisingly large amounts without feeling that the music suffers. Rather, by second-order I mean resonant problems, inaccurate imaging, and a lack of transient—ie, phase—integrity. My experience has been that deficiencies in these areas can be fundamentally destructive of the real illusion once tonal inaccuracy and harmonic distortion have been reduced below the threshold of tolerance (this varying for different people).
These errors which I think important have, in general, been regarded as benign. Inaccurate stereo imaging? As I indicated in "As We See It" in Vol.9 No.8, nearly all commercial classical recordings lack an accurate stereo image, though only a handful of us nitpickers appear to be disturbed by such spatial distortion. But as the human ear/brain is excellent at determining where sounds come from in real life, preserving the directional information intact on a recording must surely have some importance.
(JGH has often argued that the fundamental problem with stereo imaging, and one echoed in this issue by Dick Olsher, Bill Sommerwerck, and John Sunier, is that two-channel reproduction folds the reverberation enveloping the real-life listener into the frontal image. I wouldn't argue with the fact that this occurs, but I would point out that this is inescapable with stereo. Of the systems that tried to do more, only Ambisonics was based on a correct theoretical model of what information needed to be recorded/replayed, but even that failed to be generally accepted. Why? An analogy can be drawn with photography: effective three-dimensional photography has been around almost since the time of the daguerrotype, yet virtually all photographs taken are flat; the (in theory) more-limited medium actually provides sufficient information. Such, I feel, is the case with stereo vs surround-sound. Larry Archibald, however, argues with this, feeling it only to be true if the aim of photography was similar to sound reproduction, ie, to recreate the original visual field of the observer. The aim of photography is sharply different from that, however, even making use of deliberate distortion to create the desired artistic effect.)
Resonant problems? Recent work by Floyd Toole, of Canada's National Research Council, presented in a workshop on loudspeaker measurements at the recent London AES Convention (reported on in this issue), supports earlier research (footnote 1) that loudspeaker resonances cause audible problems only when of low Q. If of high Q, they fail to be sufficiently excited by music to reach full amplitude when they ring.
The evidence as presented by Dr. Toole seemed incontrovertible, but I must admit that I find it hard to believe, as it runs contrary to my own experience, particularly of tonearms. I shall retreat into pointing out that, except for performances in unsuitable halls—and via typical PA systems—music in reality reaches our ears without the modifying effects of such resonances.
Phase distortion? I remember Quad's Peter Walker telling me that phase errors did not matter—while demonstrating to me that his ESL-63 loudspeaker was sufficiently phase linear to produce a squarewave with its shape intact!
Footnote 1: By, among others, Peter Fryer, now at B&W Loudspeakers