How we hear
Let me take a moment here to rattle off a few pertinent facts and observations about how we hear. We are extremely good at localizing sounds in the area covered by our field of view. As sound sources go up, down and behind we are less good at localization. We use the high-frequency content of the reflected sounds to help localize the sources of sound and the low-frequency content of reflections to "localize" the boundaries of the room. Phase-locked volleys of early reflections up to approximately 50 ms. after the direct sound are "fused" together into a single perceptual construct. From a psychological point of view, then, a "sound" consists of a bunch of phase-locked early reflections that are integrated by the auditory system, rather than consisting of the single direct emission from the loudspeaker. Filtered early room reflections are
more audible as being separate from the integrated sound than those with the same timbre as the direct sound. Filtered reflections from above, below and behind seem to be more disturbing to the localization mechanism than those from the sides. Laterally reflected energy increases the sense of envelopment from music playback - live or recorded. Highly damped and anechoic spaces are unnatural and unmusical places to listen to music in - live or recorded. To go a little further with this, let me briefly consider loudspeakers. All loudspeakers radiate sound in every direction. If we soffit mount them, we constrain the output to a hemisphere, but the total acoustic power output is essentially the same. We tend to concentrate our attention on the axial response of the speaker as a matter of test convenience and simplification. And in fact, the direct sound of the speaker is very important. If there are response problems on axis, it will not be an excellent sounding speaker, period. What is frequently overlooked, also as a matter of engineering convenience, is the off-axis response, the extreme off-axis response and the overall power response of the loudspeaker. Anyone who is used to measuring loudspeakers knows that the performance specification I quoted above (140° horizontal dispersion across the spectrum) is pretty wild. Many have said, "Can't be done."
The correct answer is, "Couldn't be done until now." Those funny looking speakers on the August Mix cover have a horizontal coverage angle of over 180°. All the way up to 16 kHz. the amplitude response is smooth, even at 90° off-axis. This behavior, which we call Panoramic Power Response, is made possible by the use of a device I call an Acoustic Lens, a pair of which you can see sitting on top of the cylindrical woofer section. Panoramic power response is, to be slightly crass, the stuff of other loudspeakers' wet dreams. With this performance capability, the rules of the studio design game change.
Conventional loudspeakers have off-axis response curves that are increasingly rolled off as we move off-axis around the loudspeaker. Worse, because of the different directivity patterns of the individual drivers, most loudspeakers have increasingly lumpy response curves as we move around to the side of the speaker. All of this lumpy low-pass sound is emitted into the room. And yes, we hear it. You better believe we hear it. Just because we've gotten used to it doesn't mean we don't hear it!
Conventional wisdom says that directional loudspeakers and rooms that damp or diffuse early reflections are good. Usually, all that is being done with such a treatment is to add even more low pass filtering to the "lumpy low-pass" reflected sound and to the room tone in general. The loss in high-frequency information particularly hurts the localization of phantom images and phantom reverberance cues, as well as darkening the overall perceived timbre. However, with a loudspeaker that does not have the limitations in its dispersion that conventional speakers suffer from, accurate lateral reflections are maintained, yielding better images, particularly of ambience, and a brighter, more open, more spacious and natural range of timbres.
COMMENTS
Jan 21, 2005 10:32 AM
To the Editor of Mix Magazine:
While Manny LaCarrubba's ultra-wide-dispersion loudspeakers may or may not "sound good" ("The Wide-Dispersion Listening Space," November 1999), his conclusions about control room design are a misapplication of basic acoustical principles.
Mr. LaCarrubba would be correct that lateral reflections from a room's side walls are important for localization cues, if the loudspeaker were a guitar or a singer or a violin. But it's not. The information necessary to form a sound image (a.k.a. the localization cues or "room sound") is already contained within the audio signal, owing to the complex set of reflections that have combined at the microphone in the recording space, whether it's happening right now on the other side of the glass or during a session forty years ago. If the reflection information already present in the sound you're trying to reproduce is convoluted by early reflections in the control room, you can no longer say with certainty whether what you hear is actually in the signal or just a consequence of the peculiar characteristics of the listening space.
Yes, anechoic spaces are "unnatural and unmusical," but that's due to our perceptual inability to reconcile what we hear in the playback of recorded sound with what we hear when we talk or turn our heads or move around within the space. Anechoic spaces are also very unforgiving in allowing the listener to move outside a narrowly defined "sweet spot" without giving up all sense of spatial image. That's why acoustical diffusion is useful in a control room, and why "robbing the reflected energy" of its directional and temporal information is precisely the point.
Being able to tell exactly where the loudspeakers are as you listen to music is not a goal in optimizing a listening environment. Quite the opposite - a well-defined image (particularly in a multi-channel system) should not seem to emanate from discrete monitor locations. The importance of lateral reflections to a listener's sense of envelopment and spaciousness has been well known in the acoustical community for decades, and they are essential in any performance space. In a properly designed control room or listening space, however, early reflections and comb filtering are not "good data." Confusing these two room types does not help to further the science of control room acoustics.
Richard Schrag
Russ Berger Design Group
Dallas, Texas
Jan 21, 2005 10:33 AM
As Manny LaCarrubba's partner, I have a few comments in regard to Richard Schrag's letter.
Loudspeakers have a GREAT deal in common with other musical instruments - from an acoustical standpoint, they are essentially the same device, and subject to the same physical rules (except for one quirk I'll get to in a minute). Much of the information contained in Mr. Schrag's "complete set of reflections that have combined at the microphone" has actually been lost at the microphone, which cannot detect that "complex set of reflections" except as a comb-filtered two-dimensional map of pressure over time.
If we were considering the behavior of a mirror, we would probably not say that "reflection information already present in the image we're trying to reproduce is convoluted by reflections of the mirror." The optical mirror may fool the range finder on a camera, just as an acoustic mirror will fool a microphone. However, neither our eyes nor our ears are bothered much at all, unless the mirror is tinted, discontinuous or both. Our auditory system is extraordinarily well equipped to make good use of acoustical reflections, and does so with ease, integrating them with the direct sound artifacts in a way that microphones cannot approach.
The unmusicality of anechoic spaces has NOTHING to do with the specifics of playback of recorded music, as Mr. Schrag implies, and the issue has little to do with control rooms, which are not anechoic and never have been (except in a few special cases - Hidley, Newell, et al). End-user environments also have never been anechoic. Under existing conditions, then, it is inevitable that there will be early reflections involved with loudspeaker playback. We cannot pretend that low-pass filtering of those early reflections is the same as eliminating them a la anechoic treatments. Our traditional loudspeakers, as a function of their inherent directional behavior, cause such lateral low-pass filtering of early reflections. The process of "deadening" walls with absorbent materials increases the severity of such filtering.
Mr. Schrag suggests that accurate localization of a loudspeaker is undesirable. Not so. If we are going to get stereo to work reliably, it is essential that we should perceive a unique signal sent to a single loudspeaker as coming from that loudspeaker. The particular unique quirk of loudspeakers that I alluded to above is that they may be operated in phase-locked arrays. When that is done and a common signal is sent in parallel to emit, phase-locked, from two points in space simultaneously, this defeats our auditory localization system and causes us to perceive phantom images. This is, of course, the basis for stereophony. What is actually happening from a perceptual standpoint is quite interesting. The loudspeakers are no longer perceived as the source of the sound, but as early reflections of a "phantom" source whose direct artifact was not perceived, and whose presence and position is inferred by the auditory system based on those perceived early reflections. They are, psychologically speaking, the first early reflections of the sound. If they are supported by subsequent early reflections from the room that are spectrally and temporally accurate, the phantom image becomes that much stronger, palpable and precise. This is why we call such reflections "good data." And, the more accurate those reflections are, the MORE data about the recording and LESS data about the playback room will be transmitted via those reflections.
How much stronger, more palpable and precise? Manny LaCarrubba was being modest when he said the speakers "sound good." One of our country's leading mastering engineers characterized the imaging of the speakers, in a large reverberant control room, as "stunning." A senior, highly experienced and extremely successful loudspeaker designer felt the phantom image was so real he had to physically verify, with apologies, that the center channel was turned off (it was!). A professional, highly experienced test listener, doing blind testing, asserted that for the first time in her experience she thought a live player might have been substituted for the loudspeakers! Numerous audio professionals and musicians have said they believed that live performances were occurring as they first entered the playback room. Numerous audio professionals have characterized the speakers as "the best I've ever heard."
Mr. Schrag suggests that the early reflections of the playback room mask the reverberant artifacts of the recorded signal(s). Happily, the time bases of the two are both different and generally complementary in both acoustic stereophonic recording and multitrack production, so that the volley of early reflections in the playback room supporting any particular playback artifact is pretty much complete by the time that a single "reflection" of that artifact occurs in the recording of the performance space (ca. 30 ms.). The net result of this effect is a positive one, where a multiplicity of early reflections in the playback room carry direct sounds, early reflection and reverberant artifacts of the recorded sound to the listener, in a wonderfully comprehensive and rich way that our auditory system is well-suited to assimilate.
The real problem, in our opinion, is the playback room's reverberance, that wash of increasingly uncorrelated and unintegrated reflections that occur after the Precedence Effect (with its integration of early reflections into a coherent perceptual construct) has decayed. This begins to occur at approximately 40 ms. after the direct sound arrives at our ears. It is this reverberant sound from the playback room that we believe muddies up the playback of recordings, masking details and obscuring images, and that we should try to avoid in order to obtain "transparent" playback.
The design solution? Support accurate specular early reflections for 50 ms. or so, and then employ broadband suppression of all reflections and reverberance after that. It turns out this leads to an easy and dirt-cheap control room design topology (I call it a "Moulton Room") that works quite well with almost all types of speakers. It works extraordinarily well with wide-dispersion speakers.
The problem with the incoherent diffusion created by quadratic residue diffusers and similar devices in this application is that such diffusers convert direct energy into early reverberant energy. This creates a reverberant wash in the playback room precisely during the time period when we should be integrating early reflections as part of our localization.
In closing, I don't think Manny and I are confusing control rooms and performance spaces at all, and I DO think we are furthering the science (and art!) of control room design.
Sincerely,
David Moulton
Moulton Laboratories
Sausalito Audio Works
