by Thomas R. Horrall
[Thomas R. Horrall is a principal and the technical director of Acentech, Inc., based in Cambridge, Massachusetts. Formerly, part of Bolt Beranek and Newman, Acentech Provides consulting and design services in architectural acoustics and noise and vibration control.]
In the early 1970s at the acoustical consulting firm Bolt Beranek and Newman, my colleagues and I were interested in electro-acoustically simulating the sound fields of typical auditoriums and concert halls. In the course of that work it became clear that reasonable simulations permitting some freedom of listening location required use of multichannel loudspeaker playback in an acoustically dead environment. Consequently, there were 12 primary loudspeaker channels in what we called our "acoustics simulator," as well as a common subwoofer channel (back then, we just called it a woofer). Signals for the front two channels were conventional two-channel stereo from special tapes we made or from anechoic recordings. Signals from the remaining 10 loudspeakers were derived by processing the two main channels with suitable delays and artificial reverberation until their individual impulse responses matched those in real halls (or physical models of them).
The subwoofer was needed to assure that each of the 12 main channels, which were arranged to provide sound to the listening lo cation from specific directions, could also provide flat low-frequency response. Without a subwoofer, the deep-bass response from each main loudspeaker was quite different and not very uniform, primarily because of different floor interference frequencies, which varied according to loudspeaker elevation. Also, the room was not completely anechoic at very low frequencies, and that further affected the response flatness. Ultimately, we found it necessary to provide two subwoofers, each operating over a different frequency range.
No position for a single subwoofer could be found within the room that resulted in flat enough frequency response in the seating area, even over the relatively restricted bass operating range.
The simulations we achieved were relatively accurate, and it was certainly possible to distinguish major acoustical features of different halls modeled using the setup. Of course, we also experimented with playing conventional, commercially available two-channel stereo recordings through the simulator. Even that material sounded surprisingly believable, despite the double-signature problem: the superposition of the acoustical characteristics of the hall in which the recording was originally made and those of the hall being simulated by means of the signal processing. The common practice of recording with microphones placed much closer to the performers than typical audience members would sit in an actual performance probably helped with this problem by assuring that early hall reflections were not captured, or at least were minimized, on the recording.
A home theater with extensive acoustical treatment. Note
use of treatments on the ceiling.
On one occasion, a particularly vivid demonstration of a living room's effect on loudspeaker sound was afforded by the semi-anechoic simulator environment. Monophonic sound was, in turn, played back from two loudspeakers equalized to have nearly identical response. One loudspeaker was a large, highly directional, two-way professional type with horn-loaded drivers. The other was a popular three-way acoustic-suspension design that sprayed energy more or less equally in every direction. Two more different full-range loudspeakers can hardly be imagined, yet even the most highly trained and discriminating listeners had difficulty telling the two apart in the semi-anechoic simulator environment, where virtually all the sound arrived directly from the loudspeaker. Few experienced listeners would claim that these loudspeakers sounded even remotely similar when they were heard in a conventional living room.
Another interesting observation was how completely flat, two dimensional, and lifeless even the best conventional two-channel stereo recordings sounded when played back through only the two main stereo loudspeakers in this anechoic environment, without any signal processing or supplementary loudspeakers operating.
Again and again we switched supplementary loudspeakers off and always felt deprived by the lack of envelopment and immersion in the sound field, even when the two presentations were carefully level-matched. Anyone who has listened to music in an anechoic environment will readily appreciate what I mean. Yet, these same recordings were exciting and involving whenever the effects loud speakers were switched on, in a way never achieved when they were played back on even the best stereo equipment in a more conventional semi-reverberant living room environment, where most listeners thought them to be superb. Clearly, the normal reflected sound in the living room environment contributes positively to the sound of two-channel stereo, even if it fails to provide the degree of spaciousness and realism apparent from the same recordings in a multichannel system. People experience sound from many directions in most indoor environments. They judge sound to be unnatural if it arrives from too narrowly defined a direction.
About the same time as our simulator experiments, commercial quadraphonic recordings became available. Although originally made from four-channel material, these recordings either suffered from unacceptable amounts of distortion and noise or, more typically, were actually only two channels matrixed to provide four on playback. While capable of entertaining effects, especially when de coded with steering-logic processors, the matrixed systems were unable to provide the instantaneous separation among channels required to keep musical instruments imaged believably in front and at the same time provide a musically satisfying and realistic surround of reverberant sound. No two-transmission-channel matrix system I've ever heard has remotely approached the spatial realism afforded by the simulator, and this judgment extends to present day analog surround sound, which also uses a two-channel matrixed transmission format. In 1986, however, Yamaha introduced a digitally based simulator in a box-a device that mimicked the signal processing incorporated in the BB&N simulator, albeit in a somewhat limited way. This system, the Yamaha DSP-1, was soon followed by similar digital processors from a number of manufacturers. Most of these are available today, several in second- and third-generation incarnations, and some crude versions are even built into popularly priced A/V receivers and amplifiers.
Although limited to four or six channels (the two main stereo channels plus two or four supplementary effects channels), the best of these surround processors can provide a convincingly realistic simulation of real architectural spaces. To do so, however, they must be auditioned in a relatively dead acoustical environment or the double-signature problem presents itself, and in a particularly obvious way if the space being modeled is similar in scale to that of the listening room. If reflections from the listening room surfaces and the space to be simulated-say, a nightclub-occur with roughly similar delays and amplitudes, there is serious confusion, making the illusion at best unconvincing and, at worst, unpleasantly reverberant. Even if the simulated space is much larger (a concert hall, for example), the reflections from listening room walls mask important spatial cues, compromising the illusion. Another problem is that while they can do a good job of synthesizing a fairly large number of discrete reflections from different directions, most processors lack the computing power required to synthesize high-quality later reverberation. As a result, most rely on reverberation encoded within the original two-channel recording and optimized for two-channel playback, which is an unfortunate compromise.
Very recently, discrete multichannel digital recording and trans mission systems, such as Dolby Digital (AC-3), have become avail able to the consumer, and these promise to deliver very high-quality music and video-soundtrack material to the home at reasonable cost. Instead of relying on expensive synthesis circuitry, such a sys tem can record reverberation information directly in the software and retrieve it with a relatively simple and cost-effective discrete channel surround decoder. The results can probably be superior acoustically to that provided by earlier consumer simulator-type processors. In addition, the new technology enables very precise lo cation of sound sources and effects around the listener, which is of particular interest for video soundtracks.
So far, consumer releases of such discrete multichannel recordings have all been movie soundtracks on laserdisc, but the technology will be extended to other consumer formats-including DVD, DSS, and HDTV, all of which have committed to multichannel dig ital audio in some form. Although these are all combined audio/video formats, audio-only multichannel delivery will surely not be far behind. Initially, it may well use one or more of these audio/video formats. In any case, the optimum acoustical conditions in the listening room for both video-soundtrack and audio-only material are quite similar, and it is therefore useful to consider them together.
Speaker placement and Listening positions
The current standard complement for home theater loudspeakers is five primary speakers (plus, usually, a subwoofer). This appears unlikely to change in the foreseeable future, since all consumer video delivery formats comply with it, whether via two-channel matrix or 5.1-channel discrete means (the ".1" represents a dedicated, limited-bandwidth, low-frequency effects channel). A five-loudspeaker home theater system is thus unlikely to be come obsolete, although those with particular interest in audio-only applications may want to supplement these five with two or more additional loudspeakers.
Of the five primary loudspeakers, the center-channel speaker de livers virtually all of the dialog in soundtrack presentations. Col oration of speech is particularly easy to identify, and using a single loudspeaker for dialog prevents the comb-filter frequency response typical of two-channel presentations of center-panned speech to listeners seated at different distances from left and right loudspeakers. A center loudspeaker also provides good directional realism:
Regardless of the listener's position within the room, the visual and aural sources are coincident. It is interesting to note that the earliest public stereophonic presentations of music, in the 1930s, used a center channel as well as left and right loudspeakers. Perhaps stereo sound regressed initially in an effort to achieve commercial viability. Although three-channel professional tape recorders were avail able and used for recording much of the original material, it was difficult enough to put two channels of sound on a vinyl record, let alone three. In any case, after another 60 years or so it has finally become practical to deliver three and more discrete channels of sound into the home.
Fig. 1--Sound transmission characteristics of a cinema projection screen (A)
and of an acoustically transparent video projection screen (B). Note that in
both cases response is smoother if the speaker is several inches behind the
screen. (After Fukuhara et al., JAES)
The highest-quality home theater will employ loudspeakers (or the center speaker, at least) behind the screen, although this is impractical unless the screen is separate from the projector. This is the same arrangement used in movie theaters, where the center speaker is concealed behind the perforated projection screen so that dialog emanates from the screen's center.
Unfortunately, the perforations in a standard movie-theater projection screen cause severe high-frequency attenuation as well as visual artifacts when used with a video source, including significant light loss and moiré patterns. However, there are now screens de signed specifically for home theater applications that use small and random perforations and achieve satisfactory performance in all respects. Measurements of the acoustical insertion loss of such screens indicates that loudspeakers should be located several inches behind them in order to avoid significant ripples in high-frequency response (see Fig. 1). If the loudspeaker is recessed into a cavity be hind the screen, acoustically absorbing treatment should be applied to minimize undesirable cavity resonances.
In a more conventional installation, built around a rear-projection or direct-view TV, the center loudspeaker should be immediately above or below the screen. Neither location is ideal from a localization standpoint, as both result in significant vertical offset, which can be heard by an attentive listener. Placing the center loud speaker above the screen may be a little more favorable for localization, but placing it below the screen may yield smoother frequency response if a directional loudspeaker is used. The interference notch, or "floor dip," in the frequency response occurs at a higher frequency if the loudspeaker is closer to the floor, and the loud speaker may have enough vertical directivity in the treble that the amplitude of the dip is minimized. In its THX criteria for front home theater loudspeakers, Lucasfilm requires a restricted vertical radiation pattern at high frequencies and states a preference for lo cation even with or, if necessary, below the screen. For the floor dip to be minimized at lower frequencies, the loudspeaker should be placed on the floor or very slightly above. The loudspeaker listening axis should also be tipped up (or down if the speaker is above the screen) toward the listening area-a particularly important provision for loudspeakers exhibiting substantial vertical directivity.
Ideally, the front left and right loudspeakers should have the same directivity and frequency response as the center loudspeaker, to minimize timbral shifts as effects are panned across the front.
They should be located at about the same elevation for the same reason, and because significant elevation changes during a pan are also detectable and undesirable even if the timbre does not change.
However, if the loudspeakers are to be used for audio-only presentations, the floor location is not very desirable. Without the strong visual cues from an on-screen image, the listener will readily localize the sound below ear level, a somewhat unnatural aural perspective for most program material.
The lateral separation between loudspeakers depends on screen size and whether or not the loudspeakers will be used for both A/V and audio-only presentations. In a system with a large projection screen and intended strictly for home theater, the front left and right loudspeakers should be located just inside or just outside the edges of the screen to prevent discrepancies between aural and visual localization of certain sources. For smaller screens or where the same loudspeakers will also be used for high-quality audio-only re production, wider separation may be desirable. For best results with audio-only material, the loudspeakers should typically be in the range of 70° to 80° apart, as viewed from the listener's position.
For A/V material, on the other hand, Lucasfilm recommends that the spacing between the left and right loudspeakers be no greater than 11 times the screen width, even for small screens, which will almost invariably yield a narrower included angle. Some compromise may therefore be required. Alternatively, separate loudspeakers might be provided for audio alone and audio/video in the most elaborate installations.
In setting up a home theater system, it is best to keep the video screen well away from side walls so that the front left and right loudspeakers can be placed to avoid wall reflections. If the screen must go near a side wall, the wall should be effectively treated with sound-absorbing material of at least 1 inch in thickness, preferably 2 inches. The most effective place for such treatment is at the point along the wall where a direct reflection to the listener would other wise occur.
For best acoustic imaging, it is desirable for the sound from the front left and right loudspeakers to reach the listener at the same time, within a small fraction of a millisecond, as that from the center loudspeaker. This implies that their distance to the listener should be the same (within an inch or less). Some surround processors provide adjustable delay circuitry to align these arrival times without requiring alteration of the loudspeakers' front-to-back placement. If such a feature is not available, the plane of the front left and right loudspeakers should be slightly in front of the center loudspeaker. This requirement is likely to be more critical for audio-only presentations.
For video soundtracks, the best location for surround-channel loudspeakers is directly to the sides of the listening area. These loudspeakers are often of dipolar design, which results in a null, or cancellation, of sound emitted directly toward the listening area.
The intent is to diffuse the sound from the speakers by reflecting it from the front and rear walls of the room, creating a more non-localizable sound field. This arrangement may also provide effective surround for most music presentations, since reverberant sound in most performing spaces is highly diffused spatially. On the other hand, it may not permit as precise localization as desired for some special music mixes--for instance, where the intent is to place an instrument or soloist to the rear. The surround loudspeakers should also be elevated relative to the listeners' ears, near the ceiling unless it is very high.
Placement requirements for surround loudspeakers used with audio-only synthesis-type processors may be somewhat different.
Elevations are normally as described above, but the horizontal angles to the listening position may vary. Six-channel ambience-synthesis processors normally place the four surround, or "effects," channels either in front of and in back of the listener, in a roughly rectangular array (this is what Yamaha recommends for its units), or to the sides and behind the listener (essentially, adding two rear speakers to the conventional home theater array). The exact angles are unimportant. Surround loudspeakers should be located so they do not reflect strongly from side walls to the listening position. If walls are too close to the loudspeakers, they may require acoustical treatment for best results.
Guidelines for subwoofer placement are difficult to provide. Al though a corner location will normally result in the greatest acoustical average power into the room, the primary goal should be to place subwoofers where they will provide the flattest and most extended low-frequency response at the normal listening positions.
This may be difficult to determine in advance. The shape of the room and its consequent modal distribution have a profound effect on the bass response, as do the loudspeaker location, listener position, and the location and extent of low-frequency sound absorption in the room.
One commonly cited rule of thumb is to place loudspeakers away from the corners of a room so as to avoid exciting room modes, but a corner location may very well provide the flattest response from a subwoofer if the listening position is favorable relative to the room's modal pattern. Perhaps the most useful advice I can give about subwoofer location is that you should be flexible about it and be prepared to experiment after the room setup is completed. Ironically, subwoofers are often made very large to in crease their average output capabilities and then limited because of their size to just a few practical room locations, none of which pro vides very even response. Keep in mind that a favorably located medium-output woofer will often outperform a much higher-output woofer unfavorably located, at least in a small room such as most home theaters or listening rooms. If the room is rectangular and constructed with uniformly distributed low-frequency absorption, it is reasonably easy to predict the response at very low frequencies from the room's dimensions. If the room is irregular in some way, as is more typical, it is still possible to calculate response with room boundary element techniques, but few designers have access to such tools. Two (or more) subwoofers optimally located may give superior results to a single woofer.
On a related note, it may be best to provide for stereo drive to the subwoofers if more than one is used. That is, the best results may be obtained if the subwoofers are driven with incoherent signals, al though standard motion picture practice is to drive them monophonically. This is a subtle point, and some will no doubt take issue with it. It may be most relevant to music recorded in large rooms, and there is indeed a substantial amount of such information on many Compact Discs, if not on many film soundtracks.
The listener is an important component in the system. Proper listening location within the room is arguably as important as proper loudspeaker location. Just as poor loudspeaker location results in poor imaging and low-frequency performance, so does poor listener location. In fact, consideration of reciprocity would lead us to this conclusion. The listener should not be immediately adjacent to any wall, including the wall behind him, if possible.
Whenever practical, there should be at least several feet of space behind him, and the listening area should be well away from side walls. Although the perceived low-frequency response depends on the combination of loudspeaker and listener locations, it is probably easiest to optimize response by adjusting the location of the subwoofers for flattest response. There are other more important constraints on the listener's position, such as good stereo imaging and viewing relationship to the screen.
Fig. 2. Audibility-threshold versus delay of reflected sound relative to
the level of the direct sound Curves in (A) are for various sounds from different
angles in an anechoic environment. Curves in (B) are for speech in rooms
with different reverberation characteristics. (After Olive and Toole, JAES)
Room Acoustics
Although a full program of room treatment will enhance the overall acoustical experience, I realize that many will find that impractical and will feel that a less thoroughgoing implementation is good enough. Ideally, the listening room would not contribute any reflected sound to recorded material; the recorded environment would entirely establish the acoustical ambience of the presentation. However, with only five channels, it is helpful to have the room somewhat reflective to compensate for the limited spatial distribution of sound directly from loudspeakers. Nevertheless, per haps the most important overall acoustical recommendation is that the room should be considerably more absorptive than most domestic living spaces.
Reverberation time (RT) in a small, acoustically dead room is not too meaningful but does provide some useful information. The Society of Motion Picture and Television Engineers (SMPTE) recommends an RT of about 200 milliseconds for a small theater of about 3,000 cubic feet, which is about the size of a medium to large home theater. (This RT happens to correspond to an average absorption coefficient of about 0.5 for all boundary surfaces in a room that size and of typical proportions.) Even shorter reverberation times are recommended for smaller rooms.
It is also important to consider which surfaces will yield the greatest benefit from treatment and how the treatment should be applied. Figure 2A shows thresholds of detectability of reflections in an anechoic environment for various sources, relative to the level of the direct sound. Although the detectability of reflections clearly varies greatly depending on the type of signal-impulsive sounds, continuous music, and so forth-speech is among the types of signals with the lowest thresholds. Speech reflections with delays typical of those produced in living rooms are readily detected even if the level of the reflection is on the order of 20 dB or more below that of the direct sound. Reflection attenuation great enough to eliminate this effect implies an absorption coefficient of about 0.99.
Although that may seem extreme, similar speech reflections in a standard IEC listening room with fairly high reverberation are detectable at about-12 dB (Fig. 2B), implying that an absorption co efficient of about 0.94 would be needed to attenuate a reflection to inaudibility if the spreading losses for the direct and reflected paths are similar. Yamaha suggests "the deader, the better" in Sound Field Creation, a booklet on one of its multichannel processors, and Lucasfilm also recommends a relatively dead home theater environment for best sound.
Side-wall reflections of the output from front loudspeakers can be attenuated by using directional loudspeakers, keeping the speakers well away from walls, or applying sound-absorptive material to the walls. (Best results will be obtained by employing all three approaches.) The best way to apply absorption is to treat the lower side walls up to the height of the front loudspeakers. A 1-inch thickness of fiberglass panels wrapped in a suitable open-weave fabric will be sufficient in most cases. Aim for a total reflected path loss of at least 20 dB relative to the mid- and high-frequency direct-path loss from the front loudspeakers. Thin plywood covered with lightweight finishing treatments can provide good low-frequency absorption on all walls. Stud spacing behind the plywood should be randomized to avoid too much absorption over a single frequency range.
Front and rear walls may be similarly treated, except that surfaces above an elevation beginning just below the line of sight from the listener to the reflection of the surround loudspeakers should be acoustically diffuse to assure desirable reflections. Some acousticians suggest using diffusors on part of the rear wall.
Ideally, the ceiling should be entirely treated except for 2 or 3 feet around the perimeter, where some reflection is useful in diffusing sound from the surround loudspeakers. The ceiling surface area is typically large enough that it will be difficult to adequately reduce room reverberation without treating it. Making the ceiling treatment several inches thick will extend the frequency range over which it is effective. A suspended acoustical-tile ceiling that has a high NRC rating (greater than 0.85) is particularly economical and effective in providing low-frequency absorption.
On the floor, thick carpeting can provide useful attenuation of the highest frequencies and is good practice. However, its effective frequency range will be sufficiently restricted that other surfaces must provide the majority of absorption. Use of loudspeakers with restricted vertical dispersion will minimize the energy reaching the floor, making its treatment less important.
We are entering an era in which more than two main loudspeakers will be the norm in high-quality audio systems, which means that our thinking about rooms and speaker layouts will have to change accordingly. Although few people will be able to follow all the practices suggested here, I think that any of them will provide some sonic benefit in a multichannel system. And the more you can do, the better.
References
1. SMPTE Engineering Guideline EG 18-1994.
2. Olive, Sean E. and Floyd E. Toole, "The Detection of Re flections in Typical Rooms," Journal of the Audio Engineering Society, July/August 1989 (Vol. 37, No. 7/8).
3. Nakayama, Takeshi, Tanetoshi Miura, Osamu Kosaka, Michio Okamoto, and Takeo Shiga, "Subjective Assessment of Multichannel Reproduction," JAES, October 1971 (Vol. 19, No. 9).
4. Fukuhara, Suemei, Satoshi Kageyama, Yaeko Tai, and Koichi Yoshida, "An Acoustically Transparent Screen," JAES, December 1994 (Vol. 42, No. 12).
5. Horrall, Thomas R., "Auditorium Acoustics Simulator: Form and Uses," AES Preprint No. 761, October 1970.
6. Lucasfilm, Home THX Audio System Installation and Operation Manual.
(adapted from Audio magazine, Aug. 1996)
Also see: Crossing Over--The Misunderstood World of Cutoffs, Slopes and Response Shaping (Feb. 1997)
= = = =