Polk SDA-1 Stereo Dimensional Array speaker (Equip. Report, Jan 1983)

Home | Audio Magazine | Stereo Review magazine | Good Sound | Troubleshooting

Polk "Reinvents" the Loudspeaker

JUST TO LOOK AT IT, the Polk SDA-1 Stereo Dimensional Array doesn't seem much different from a great many other floor-standing dynamic loudspeakers. To hear it, though, is quite another matter. Our first, all too brief audition simply bowled us over: The width, depth, and precision of the stereo image were astounding, as though a very good image-enhancement unit had been plugged into the system along with the speakers. We now know that this is, in effect, just what was going on, although the implementation is primarily acoustical, rather than electronic. And we also know, after extended listening, that we're no less astonished than we were that first day at the system's sometimes mind-boggling powers of sonic persuasion.

The problem the SDA-1 seeks to over come is known as acoustical crosstalk, which is also the prime target of several purely electronic devices, most notably the Carver Sonic Hologram Generator and the Sound Concepts IR-2100 Image-Restoration system. (The SDA-1's approach is more like that of the Sound Concepts.) Acoustical crosstalk occurs when a signal that should ideally be heard by only one ear is heard by the other as well. Unfortunately, nonideal behavior is inevitable in ordinary stereo systems. Sound from the right speaker, which really should be heard only by the right ear, sneaks around the head to the left ear, where it competes with the desired signal from the left speaker. And the same effect happens from left to right.

----------------

Polk SDA-1 Stereo Dimensional Array floor-standing loudspeaker system, in walnut veneer cabinet.

Dimensions. 15 1/4 by 43 3/4 inches (front), 11 3/4 inches deep.

Price: $1,700 per stereo pair.

Warranty: "limited”, five years parts and labor.

Manufacturer: Polk Audio, 1915 Annapolis Road, Baltimore, Md. 21230, USA.

-----------------------

Cancellation of acoustical crosstalk is achieved by means of a second array of drivers, called the dimensional array, mounted to the outer side of each speaker cabinet's front baffle, slightly more than the width of a human head away from the regular stereo array. In the example shown here, a signal emerges from the left speaker's stereo array, arriving first at the left ear and slightly later, because of the additional distance traveled, at the right ear (where it constitutes acoustical crosstalk).

Simultaneously, a signal comprising a standard right channel component and an out of-phase left-channel component emerges from the right dimensional array. The placement of the right dimensional array causes the path length from it to the right ear to be almost the same as that from the left stereo array to the right ear. Consequently, the in-phase component from the left stereo array and the out-of-phase component from the right dimensional array arrive at the right ear at the same time and cancel each other, thereby eliminating the left-channel acoustical crosstalk.

---------------

Acoustical crosstalk is purely an artifact of stereophonic reproduction: It cannot occur in nature, where sound comes from only one direction for a given source--not two, as in stereo. It is a problem because of its potential for confusing the brain's sound-localization system, which depends primarily on differences in arrival time at the two ears to determine the direction from which a sound is coming. For example, a sound coming from your left should arrive at your left ear before it gets to your right ear. The magnitude of this interaural delay depends on the distance between your ears and on the angle from which the sound is coming. The interaural delay will be greatest when the sound source is exactly 90 degrees to your left or right and zero when the source is directly in front of you.

Anextreme example will serve to illustrate how acoustical crosstalk can pro vide false localization cues. Consider a recording in which the sound source is far off to the left and the microphones are K spaced apart by a distance greater than the diameter of your head. When played back over a conventionally arranged stereo system, the acoustical crosstalk signal from the left speaker will arrive at the right ear before the direct signal from the right speaker. The precedence effect will cause the brain to suppress the later-arriving right-channel signal. The brain will then dutifully localize the sound as coming from the direction of the left speaker.

By this logic, it follows that in a conventional stereo system no sound can ever be localized outside the limits defined by the positions of the speakers, regardless of where the sound actually originated at the recording site. This is especially damaging to ambience information, which naturally arrives from all directions and serves to give a live performance its sense of depth and spaciousness. In conventional stereo, this information is all squeezed up front between the speakers.

To generate a more nearly correct image, each Polk SDA-1 speaker has two arrays of drivers, plus a woofer system that operates from about 100 Hz down. The woofer system consists of two 6 1/2-inch active drivers and a large passive radiator to which they are acoustically coupled. Above each of the woofer cones (which are side by side on the baffle) is a column of two drivers: a 6 1/2-inch mid-woofer and a dome tweeter slightly more than 1-inch in diameter. The inner (or "stereo") column is driven by a normal left- or right-channel signal; the outer (or "dimensional") column, how ever, is driven by either an L-R signal (in the case of the left speaker) or an R-L signal (in the case of the right speaker). These signals are derived by means of a passive matrixing network and a cable linking the two speakers.

The trick is that the two arrays are separated by a distance just slightly greater than the diameter of a human head. Knowing that, consider the following simple (if rather artificial) case, in which the output from the amplifier to the speakers consists solely of left-channel information (see diagram). That signal will emerge from the inner "stereo" array of the left speaker. In addition, an L-R and an R-L signal will be derived by the matrix-in this special case, an L and a-L signal, respectively, since there is no right-channel signal. The-L signal emerges from the right speaker's outer "dimensional" array at the same time the L signal emerges from the left speaker's stereo array. The only element of the signal from the left stereo array that is desired is the one to the left ear: The other path, to the right ear, is acoustical crosstalk.

--------------

Now, the right-channel dimensional array is displaced just enough to the right that its output (-L) will arrive at the right ear at the same time as the acoustical crosstalk signal (L) from the left stereo array, so that the two cancel. That eliminates the false localization cue that otherwise would have existed because of the crosstalk. But there is now only one signal, which arrives only at the left ear; localization requires signals at both ears.

The desired localization cue is sup plied by the L signal from the left dimensional array. Although it travels to both ears, it arrives at the left ear after the L signal from the nearer stereo array and is therefore ignored by the left ear (because of the precedence effect). And because of the dimensional array's displacement to the left, it takes twice as long to get to the right ear as did the now cancelled crosstalk signal from the left stereo array. Consequently, the perceived interaural delay is twice as long as it would be if the sound source were in the direction of the left loudspeaker. This causes the brain to localize the sound to the left of the left speaker. (According to Polk, systems that cancel the crosstalk without supplying this additional cue can severely distort the central image by stretching it too much toward the sides.) How far to the left depends on the listener's distance from the loudspeakers. If he is exactly as far away from an imaginary line connecting the two speakers as the speakers are apart from each other, he will hear the sound exactly 90 degrees to his left. Moving farther away will cause it to move gradually around an arc toward the left speaker; moving closer will create an apparent interaural delay larger than is possible in nature. (For that reason, Polk recommends that you not sit too close to the speakers.) Real situations are more complicated and bring factors such as loudness differences into play as secondary localization cues, but the basic idea is clear. And it is, in its essence, elegantly simple, despite the necessary complexity of our exegesis.

Testing the speaker proved to be at least as difficult a task as explaining it.

Diversified Science Laboratories reports that measurements of the SDA-1 account for more than 7% of the total data it has accumulated on loudspeakers since we began our current testing program in June 1981. Frequency response measurements were especially difficult, since the loud speakers cannot be operated individually and their outputs vary so much according to the characteristics of the input signal. The curves shown here were made using a mono drive signal with both speakers pushed together against the rear wall. There is no special logic to this nonstandard arrangement, other than that the results seem pass ably consistent with what we hear and with the general trend of the many other curves DSL generated. (There do, for example, seem to be mild prominences at around 500 Hz and 10 kHz.) We admit that we really have no idea what to make of an off-axis curve made in this way. And we note that these are neither the best nor the worst that we obtained. For the record, the smoothest curve was taken off axis with the right- channel speaker several feet from the rear wall and the left-channel speaker in another room, muffled with blankets and with the whole system driven by a right-channel input.

At any rate, of all the loudspeaker response curves DSL has ever run for us, the ones obtained with the Polks are by far the least consistent or predictive. But given the SDA-1's unusual operating principle, we're not too surprised. Take the curves shown here with several large grains of salt and judge the speakers by ear.

Impedance measurements also posed special problems, because the impedance varies according to how "stereo" the drive signal is. A mono signal, which exercises only the "stereo" arrays, gives a fairly smooth curve that never drops below 4.8 ohms. Left- or right-channel drive brings the dimensional arrays fully into operation, along with the stereo arrays, dropping the minimum impedance to 2.4 ohms. In nor mal use, the impedance would vary continuously between these two extremes. Some amplifiers may have difficulty with this load, although we expect most will get along fine, provided you don't run another pair of speakers in parallel.

DSL measured sensitivity, power handling, and distortion in the most conservative possible way-using a mono drive signal with one speaker in the room against the rear wall and the other in another room muffled with blankets. Sensitivity turns out to be high anyway, which means that it's probably somewhat higher still with more typical signals. In the 300-Hz tone-burst test, the Polk accepted the full output of DSL's amplifier-62 1/2 volts peak, equivalent to 27 dBW, or 488 watts, into 8 ohms.

Total harmonic distortion (THD) is generally quite low on the SDA-1 , even under these worst-case conditions. At a moderately loud sound pressure level (SPL) of 85 dB, THD averages less than 1/4% over DSL's entire test range (30 Hz to 10 kHz) and less than 1/2% from 100 Hz up, and it doesn't rise appreciably until a very loud 95 dB SPL is reached, where it still averages less than 1% from 100 Hz up.

In the listening room, we wound up placing the SDA-1s against the rear wall, where they not only sound somewhat better balanced, but also seem to image better than when placed away from the wall (a first in our experience). The overall sound is agreeably smooth, with an occasional tendency to what listeners have variously referred to as a slight brightness or hardness. And the system tends to make some recordings with large amounts of artificial reverberation mixed into them sound annoyingly echoey. These have been the only complaints, however.

The SDA-1 's strong suit (to put it mildly) is its imaging, which ranges from very good to flabbergasting, depending on the material. It seems to be at its best with simply milted jazz and classical recordings or with heavily produced rock, which it can make devastatingly dramatic. With good classical discs, the soundstage seems to open up, presenting a greater sense of depth and enveloping the listener more fully in the recorded ambience. The effect seems more subtle on most heavily multimiked material, but remains ingratiating, nonetheless.

But it's on fancy rock recordings that the system can really strut its stuff-solo instruments thrown out far to the left or right, beyond the confines of the speakers, the sensation of almost falling into the sound, singers made larger than life. (Try the Beach Boys' "In My Room," for example, or Pink Floyd's The Wall, if you'd like a shiver or two.) It's not natural, of course, but the sort of record we're talking about here is not exactly what you'd call organically grown, and it really is great good fun. We find ourselves listening to unfamiliar recordings on other speakers and saying to each other, "I wonder what this would sound like on the Polks?" And we're going to miss being able to find out when the time comes to send them back to Baltimore.

These are by no means inexpensive speakers: Most of you probably can't afford to run right out and buy a pair. We do suggest, however, that you at least pretend that you can long enough to get an audition. It's worth the trouble just for the experience.