Back to REGONAUDIO

In ISSUE 82 of The Absolute Sound, I wrote an article [reprinted on this site] about how digital signal processing (DSP) was opening up a new realm of home audio in which the bad effects of the listening room could be cancelled out ("Digital Room Correction: The Wave of the Future," Issue 82, p.38). That article concentrated on the theory and the possibilities rather than the actual listening experience, because my listening at that time, positive though it was, had amounted to one short session, and that was in a system other than my own. Now I have had a chance to listen to the SigTech AEC 1000 system in my own listening room over a period of weeks, and I am even more impressed than before. The revolution I was anticipating in my initial write-up is actually here.

This time around, I am going to concentrate on what the SigTech system does in sonic and musical terms, with rather minimal repetition of how it does it. For a detailed explanation of how it works, you should refer to the Issue 82 article, reprinted here.

The SigTech is a triumph of scientific and mathematical theory. Scientific theory has often promised to do greater things for audio than just supplying the basic parts and the broad outlines of how things work. Sometimes, oversimplified versions of science, a sort of technological pseudo-science masquerading as the real thing, has been used as a bludgeon in audio design, applied without sensitivity to the musical results. But this time, technology really is being used in the service of music.

What I Listened To

With the help of James Prescott from SigTech, I ended up with four filters, one for the (floor-standing) Snell B Minors, one for the Spendor SP1/2s on their dedicated stands, and two for the QuacrESL-63s, one for the Quads on their (short) stands, and one for them on the floor, but tilted back so that the listener's ears would be on-axis. The Snells and Spendors were review samples, but I had listened to them quite extensively already. The Quad 63s were my own speakers.

I should mention that the software package to establish the filters, while complex in what it does, is simple to use. Even a complete computer novice should have no trouble learning the procedure.

The filters can be used in several ways. For one, the SigTech can be inserted in the tape monitor loop via the input and analogue output. In this setup the SigTech unit converts the input to digital, does its corrections digitally, and converts the result to analogue again. In this mode of operation, "Bypass" on the control board still does A-to-D, then D-to-A, but does no room/speaker corrections of the digits. One can switch in anyone of the four filters stored by the unit.

Alternatively, the SigTech unit can do its digital corrections directly on a digital input, and produce either a digital output or an analogue output. In my case, I used direct digital in from a CD transport and digital out to an outboard D-to-A, when I wanted to deal with CD as purely digitally as possible. (There seems to be nothing to be gained from inserting an extra A-to-D, D-to-A conversion if one is dealing with digital source material to begin with.) On analogue material, you can also use the digital output, letting the SigTech do A-to-D but using your own outboard D-to-A converter.

Let me remind you what the "filters" do. A filter "program" is written for a given speaker in a given position in a particular room. The filter does thereafter a linear correction of both phase and amplitude, independently of the music signal so that the actual sound arriving from the speaker at the listening position is identical to the original music signal---identical including the effects of the room and the inaccuracies of the speaker! "Identical" here means, of course, identical within the limits of the correction system's resolution. I'll return to that point later. But first I would like to concentrate on the listening experience.

What the SigTech Did for Tonal Character

The correction process acts to correct both tonal balance and spatial (imaging/ soundstaging) presentation. And in practice, these two aspects are unified; that is, the correction process presents itself to the listener as a single, unified improvement. But from the viewpoint of describing the experience in words-at least in the usual vocabulary of audio description---it is probably best to separate the correction of tonal colorations from the improvements in spatial performance.

The three speakers I worked with are, already, without benefit of the SigTech system, among the most tonally neutral, most accurately balanced speakers that have ever existed. The Quads are well-known as among the least colored of all speakers. The Snell B Minors share the midrangetreble ultra neutral balance of their larger brothers, the Snell Bs, and in many rooms may be even smoother through the midbass than their big brothers. (The B Minors have a different and, in my view, better tweeter than the Bs, but have the same midrange drivers and the same mid-tweeter configuration geometrically.) The Spendor SP1/2s are a new version of that classic British box monitor, the Spendor SP1. This new version pushes the already high neutrality of the original SP1 even further. Thus in all three cases, the SigTech system was trying to improve on the neutrality of speakers that most audio people would regard as very neutral indeed.

Even though these speakers are at the top level of neutrality as speakers go, it is also true that this doesn't say much about their neutrality in absolute terms. Most speakers are quite wildly colored. It is hard indeed to make a truly neutral speaker, so hard that even the best efforts are bound to fall short, at least so far in audio history. And most speakers fall short by a lot.

For reasons that aren't entirely clear to me, the audiophile community has seemingly decided that High End speakers are neutral enough. But the truth is, if a truly neutral comparison standard is available, or has been developed in one's mind by careful listening, then the vast majority of High End speakers reveal themselves as colored indeed. (Of course, non-High End speakers are often even worse.)

The Quads are indeed well-balanced and low in coloration as speakers go, but they still have some observable imbalances and colorations, notably a certain midbass leanness (which increases as they are raised further from the floor on various kinds of stands) and a discontinuity in the lower-to-mid treble that tends to make the speaker sound mid-rangy and slightly nasal. I am pointing these out not to criticize the Quads, which remain among the finest speakers around, but rather to make two other points.

First, that a speaker almost legendary for good balance and low coloration nonetheless has some observable colorations and imbalances. And second, these problems that I had observed over long-term listening experience were immediately identified by the SigTech measurement and analysis process. That is, one could see on the graphs produced by the system that these problems occurred. [ Before and after graphs for the Quads are in the introductory article] And, moreover, the SigTech filter corrected the problems admirably.¹

I have some feeling that maybe not everyone wants to know about this. But the truth is, when the Quads were made more neutral a la SigTech, they really did sound more neutral, and in a way that really counts in musical terms. Even though the Quads are far lower in coloration than almost any other speaker, nonetheless, when the chance to erase the colorations arrived, how almost addictively attractive that chance seemed!

I was less intimately familiar with the Snell B Minors; but I was far enough into the review process with them to have identified a certain sonic character, a kind of dimness in the upper midrange followed by a brightness at higher frequencies if the tweeters were not turned down a good bit below the "flat" position. Once again, the SigTech nailed the problem in its graphical analysis, and repaired it in listening terms. Similarly, I had noticed a "hole" in the mid bass, presumably from the floor cancellation in that region. The SigTech system again took measure of the problem, and fixed it.

I am mentioning these cases not to pat myself on the back but rather to point out that the SigTech system does seem to correlate well with listening experience, at least with mine. It is not at all obvious what kind of in-room measurements-of frequency response in its various versions are actually most closely correlated with what one hears as tonal character. What the previous paragraphs are intended to suggest is that SigTech is on the right track in this business.

The most spectacular example of all this was the Spendor SP1/2. Ever since I first encountered this speaker at Winter CES 1993, and even more since I received my review samples (long before the SigTech arrived in particular), I had been struck by their amazing neutrality. They were curtailed at the most extreme frequencies-no deep, deep bass, a slightly subdued top in listening terms. But over the whole of the rest, where the music is, they seemed almost flawless in tonal terms, or at least far, far superior to almost all others.

Of course, there were small colorations observable, but small was the operative word, in a world where most speakers make rather large errors. In particular, the almost ubiquitous floor-cancellation midbass hole was minimal. And except for curtailment at the frequency extremes, the SP1/2s seemed startlingly smooth and neutral. I had not yet written my review, but I was fascinated with their tonal correctness to the point of inflicting live vs. recorded tests on anyone willing to listen.

Enter SigTech, and, sure enough, the correction filter had only subtle tonal effects, except for bringing up the frequency extremes, in particular, bringing in more mid to upper treble. I don't want to put words in anyone's mouth, but Prescott-who has of course heard many times over the large changes the SigTech induces in most speakers-seemed very struck by how little the Spendors needed correcting or, more precisely, how little change was noticeable when the correction was switched in. This was as far as colorations went. As I shall discuss later, all the speakers including the Spendors underwent gratifying improvements in spatial performance when SigTech-corrected.

Here is an additional ilustration of frequency response corrections, in this case of the floor-standing DALI Grand speakers, which are anechoically quite flat. Note the vast improvement in smoothness of the bass and midbass in particular, through correction of room effects.

What can I say? I am only one person. But, for me at least, the SigTech system really works for the removal of colorations. It identifies the major problems and moves the speakers toward ideal neutrality. It doesn't do this without limitation. Some innate character of the speaker remains. For example, the metal dome sound of the metal dome tweeter of the Snell B Minors remained even after correction. So did a certain "box" character of the bass of the Spendors at high levels. If the correction process was perfect, the speakers would all end up sounding tonally identical, after their respective corrections. And that doesn't quite happen. But major moves towards neutrality are achieved. And the resulting nearly perfect neutrality is musically vital. Once experienced, it is unforgettable and hard to relinquish.

What the SigTech Did for Space and Imaging

However important tonal correctness is theoretically-and for me musically, too-there are audiophiles for whom it is secondary to the spatial aspects of musical reality. But for those people, too, the SigTech system has remarkable things to offer. There is a band of frequencies, roughly 100 to 1000 Hz, over which phase differences are a vital aspect of stereo imaging (cf., my article in Issue 64, "Directional Hearing: How to Listen to Stereo," pp. 40-52, reprinted on this website). But in this frequency range it is hard to avoid having room reflections confuse the recorded phase relationships. The SigTech can fix this up in principle. And it seems to do so in practice, too.

On well recorded material, the SigTech system can, in listening terms, effectively erase the listening room and the speakers as sound sources, and simply provide the original acoustic environment of the recording. Much of this effect was previously obtainable from either nearfield listening or by using large planar radiators which effectively put the listener in the near(er) field, subjectively, at fairly large listening distances. But the SigTech correction raises this to new levels.

The SigTech really does erase reflections. Look carefully at the in-room impulse response shown here. (These are the floor standing DALI Grands.) The strong early reflection is simply gone after correction except for the very high frequency part.

Tonal balance interacts with space perception. So to talk about the space effects of the SigTech, it makes sense to talk about the speakers that underwent the least tonal correction, namely the Spendor SP1/2s. The SP1/2s are box speakers, but they are far from boxy in the usual coloration sense. Moreover, they are nearly inaudible as sources when producing centered or close to centered images. However, for images to the left or to the right, a slight sense of discontinuity arises in the sound-field.

It is well known, and more or less clear intuitively, that the ear-brain uses reflections as part of the sonic mechanism of locating objects in rooms, as well as for sensing room volume. So it isn't surprising that cancelling room reflections makes for less sense of where each speaker is. If you have tried listening to your speakers outdoors, you know how this works. Except the ground (floor) reflection, outside is reflection-free. The SigTech with the Spendors gives you much of that effect indoors. The result is the sense of seamless sound-field and original acoustic space that is such an audiophile goal.

Another interesting aspect of the spatial situation is that the SigTech generates increased perceived localization of sounds in the midrange down through the mid bass. Sound sources that involve hard transients and hence high frequencies, say a snare drum or a wood-block, tend to be well localized by any well set up system. All that is necessary is good pair-matching of the speakers and minimal early high frequency reflections from the side walls. (The latter is easily arranged by suitable wall damping.) But, as I already pointed out, the phase in formation that acts to localize lower frequency sources is easily obscured by room effects.

With the SigTech this problem is much ameliorated. To take a specific instance, the lower organ notes on the Dorian recording of Bach's Goldberg Variations (played by Jean Guillou) acquired a far more precise location with the the SigTech than without. In reality, the bottom notes of the organ are too low to be locatable via their fundamental frequencies, but are localizable via their harmonics and their initial transients.

The SigTech system also gave a greatly improved sense of the volume of the original recording venue. I would suppose that this is related to the improvements in phase behavior and related localization cues. But, in any case, material recorded in large spaces sounded large, not so much in literal image size, but rather in the sense of volume of space that HP has discussed in the past. To my mind, this plays an important role in dissipating the feeling that reproduced music is somehow a miniaturized toy replica of the real thing.

In short, the SigTech system conveys considerable, sometimes striking benefits in spatial performance. I should emphasize in this context that, as you would suppose, I have already tried hard to optimize this aspect of audio performance by speaker placement and room treatment. We are talking here about improvements that go beyond what can be practically accomplished by conventional methods.

Further illustration of reflection cancellation by the SigTech

We have already discussed the importance for the representation of space as well as for smooth frequency response of cancelling early, discreet reflections. Here is an illustration of this process in action. The top curve shows the uncorrected impulse response of a Dynaudio floor-standing model at the listening position. The bottom curve shows the impulse response corrected by the SigTech. The red arrows indicate arrivals of early reflections. The corresponding times indicated by the blue arrows in the corrected impulse response should be compared to see how the reflections have been essentially erased, except for very high frequency components. These reflections may look relatively small in the graphical illustrations but they have large effects on frequency response in particular. As you can see in the uncorrected and corrected frequency response curves below the impulse response graphs.

Notice that not only are the reflections almost eliminated but the early arrival from the speaker the first 2 msecs is also smoother in the corrected impulse. This smoothing is associated to the correction of defects of the speaker itself, rather than room effects.

The combined result is greatly improved overall balance and greater smoothness, as the following frequency response graphs show.

Limitations of the SigTech System

From the beginnning, High End audio has always been opposed to fooling with the signal. The mass market might pursue tone controls, graphic equalizers, and ambience generators; but the purist seeks just that, purity, with only rare exceptions.

The SigTech belongs in an entirely different category from analogue EQ devices, however. First of all, it really does do the job it is supposed to do in a way that no analogue EQ device does. There are certain intrinsic limitations to correction of room effects; as I discussed in the first part of this article, higher frequency reflections could only be cancelled if one were willing to be a truly "clamped-head" listener. But within these limitations, the SigTech does its job.

There are certain choices to be made in the relationship between the measured performance of the speaker and the "filter" or correction program that results. In the future, some slightly different choices may turn out to be better; until alternative approaches are available, there is no way to make one's own judgment about that. But in absolute terms, the SigTech system works admirably.

The SigTech carries out its measurements with an (essentially) omnidirectional microphone. There might be a case to be made for using one or more directional microphones pointing in different directions and subsequently treating arrivals from different directions differently in the filter programming. It might also arguably be better for the higher frequency parts of the filter to average the measurements over a number of positions near the actual listening location, rather than using a single one-point measurement. One might also imagine incorporating some stereo-enhancing effects a la Carver and Polk as an option for the user.

These matters do not have clear-cut theoretical ideals simply because there is no clear-cut theoretical paradigm for the stereo recording-playback process as a whole. The absence of such a paradigm has been often pointed out over the years, notably by Jon Dahlquist in a guest article in Issue 3 of TAS. But the issue remains unresolved.

Recording companies have almost always made their recordings by-guess-andby-golly, checking on the results as they went along by listening to "monitor" systems of highly erratic performance, in many cases. This is pretty far from being a pattern for which ideal playback would be clearly defined. As DSP along the lines of SigTech comes to be used to standardize speaker performance, it may lead to some more precise understanding of how recordings are or should be related to playback for a more or less literal replication of reality. But there is a long way to go in terms of the recording industry actually giving this question the careful thought it needs and deserves.

The question remains right now of whether the SigTech's programmed filters damage the signal in any significant way. For digital material, my answer so far is no, in effect. A tradition is growing up in High End audio of "purist" treatment of CD material, as if the readout of a CD were a more or less exact and easily damaged literal facsimile of the music, as an analogue signal is. This doesn't really make sense to me. A digitized signal is far more robust. If you do not damage the bits and if you re-clock the signal to be jitter-free before conversion to analogue, all is well.

The SigTech system makes such big improvements to digital material that whatever degradations it makes, if any, are insignificant. (The system uses a 20-bit system to minimize losses in the calculations, and to my ears it does the job.) Such is my impression so far. I would be surprised if in any system the large and obvious improvements the SigTech makes would fail to outstrip by far whatever little artifacts might appear from the data processing.

What the SigTech System Means to High End

If people approach the SigTech system (and other similar systems to come) with an open mind, if we listen with ears and musical memories, not with prejudices, then I believe these systems will have a profound effect.

They will effect not only our day-to-day lives as listeners, but also our understanding of the whole process of reproducing recorded music. Eventually, they will effect our ideas of how music should be recorded, too.

First of all, the SigTech makes it obvious that in almost all audio systems the speaker-listening room setup is making large, musically significant alterations in the music being played. By large, I mean both large enough to be instantly, conclusively audible, and also far too large to be significantly ameliorated by changes of amplifiers, preamps, CD players, cables, or anything else. (except speakers and room acoustics)

High End audio tends to be obsessed with the kinds of differences that changes of electronics and cables make. I would strongly affirm that the room/speaker combination is, in many musically vital ways, always the weakest link in the playback chain.

Of course, many people, myself included, already believed this long before the SigTech system arrived. But the opportunity to change from uncorrected to corrected playback by punching a button makes the point excruciatingly obvious, and also obviously important in musical terms.

I should mention that the SigTech type of system will offer at long last a way to study what is really significant in the room-speaker interaction. The SigTech can be programmed not only to produce flat response, but also other types of response. In principle, such systems can be programmed to produce any given (linear) behavior, that is, any frequency response and phase response one might care to hear.

In effect, this becomes a do-it-yourself version of the Archimedes/Eureka (of KEF, Bang and Olufsen and the Danish Technical University) project for evaluating the audible effects of the listening room. Even my relatively brief experiments with the three speakers mentioned already gave preliminary indications of some principles: The floor cancellation of most speakers is very audible and annoying musically. Broadband imbalances are far more audible than narrow-band, with extremely narrow-band dips being almost totally insignificant as generators of coloration. All these are widely held beliefs, but with the SigTech you can try them out for yourself, seeing, in effect, before your very eyes what is causing the colorations you hear.

I am not trying to suggest here that SigTech-corrected sound is perfect. Rather, the point is this: Once you have heard a system in which the room-speaker component has been made nearly flat and nearly reflection free, then it becomes obvious how colored almost all uncorrected room/speaker setups are.

Writings elsewhere, and a number of conversations with audiophiles, have led me to believe that a certain skepticism is developing in High End circles about DSP room correction of the SigTech type, sometimes even before these systems have been heard at all by the self-declared skeptics. Old prejudices against signal processing, or perhaps vested interests of various sorts, apparently not only die hard but even resist any new challenge. Furthermore, the supposedly technical objections being raised are mostly just pure nonsense.² Shame on those who won't approach these new systems just on their sonic merits. So let me speak right out.

The SigTech-corrected Quads and Spendor SP1/2s generated the most tonally neutral sound I have ever heard from a system in my listening room.³ And within the limits of the digital standard, all three of the speakers produced (at normal listening distance) a sound-field of exceptional spatial realism, equalled in my experience only by near-field listening to truly coherent sources (e.g., Quads or Sound Lab Pristines). Again, within the limits of the CD digital standard, the SigTech-corrected systems offered also exceptional resolution of detail.

In short, this SigTech system works and addresses issues that are by audiophile consensus of the greatest importance. Moreover, to my ears, anyone who finds its effects subtle must have peculiar listening orientations indeed.

The differences the SigTech makes are in fact anything but subtle, compared, say, to changing from one good cable to another. The differences are positively enormous compared to putting electronics on feet or gluing little things on the sides of amplifiers.

We each have our perspectives on what is important in audio. But in some general sense, it seems natural to suppose that a big difference in sound is likely to make a big difference to musical experience. What we are talking about with the SigTech system was truly a big change in all the cases I tried, and I was correcting speakers that were already among the best, in a carefully laid out, dedicated listening room.

To my musical perception, the SigTech did indeed make a big musical difference. For example, listening to violin recordings on the SigTech-corrected Spendors was, for this violinist, positively hypnotic, so strong was the feeling of hearing the true tone of each recording.

Don't let anyone tell you this type of correction isn't important. There may be more refined versions in the future; in fact, during my review process, SigTech came out with a revision of the bass processing part of their software. And, no doubt, they and others will revise and improve the whole process as time goes by. But there is no question even as things stand now of this whole idea being anything but vitally significant.

Some years ago, Keith Howard interviewed Peter Walker for TAS [Issue 23] at the time the "new" Quad ESL -63s first went into production. Howard asked Walker how long he thought the "new" Quads would last at the forefront of speaker design. (The Quads, with their narrowed radiation pattern and ultra low distortion are ideal candidates for DSP correction, and in that sense may be current for a long, long time.) Walker's words do have a ring of prophecy, if reluctantly so: "That's prophecy, isn't it, which is awfully difficult I think it [the new Quad] should hold for a long time. I wouldn't know how to make a better loudspeaker at the moment unless you could get the room out of the system. But I don't think that's possible unless you can hold the head of the listener in one rigid position or have some means of the reproducer knowing exactly where the listener's ears are. If that can be done, then you have a new era in high fidelity."

DSP systems have certain intrinsic limitations. They cannot really deal completely with the narrow-band "infinite nulls" that can arise in the low frequencies. And they cannot correct for reflections in the highs except for a fixed listening position. SigTech has devised thoughtful and satisfying solutions to these potential problems. By band-averaging the low frequency corrections, the system does not overdrive amplifier nor speakers in a futile attempt to do narrow-band correction of "infinite nulls." Since narrow-band dips are essentially inaudible, this works perfectly. To avoid "clamped-head" high frequency problems, the system shortens the time "window" over which correction occurs as the frequency rises. This works, too: No discontinuities are caused by natural head movements, and the listening area in which improvements are noticeable from the SigTech system is as wide as the area over which anything resembling decent stereo would occur, in any case.

In my experience so far, the kind of correction that is possible for a centered but not rigidly constrained listener does encompass most of what could be carried out in the theoretically ideal situation of a "clamped-head" listener. For speakers with relatively narrow radiation patterns in the higher frequencies combined with absorbent room treatment, the goal of taking the room out of the system is being approached. And, as Walker prophesied, it really is "a new era in high fidelity."

¹ I did not have the Arcici stands available to analyze that setup. But I did verify my own conclusion that the Quads, at least in my room, are better balanced on the floor than on the short stands, and would suppose from past listening that the Arcici stands would make the midbass leanness even more extreme than do the short stands.

² An example is the carrying on about infinite nulls, where a reflection has equal amplitude but opposite polarity from the direct sound. These are, as such, uncorrectable. But, as Roy Allison's pioneering work has pointed out, they can be minimized in sonic impact by careful speaker placement-and speaker design. Moreover, they are by nature narrow in bandwidth; indeed, complete cancellation from a particular reflection occurs at discrete frequencies, i.e., no bandwidth at all! It is widely believed-and my experience confirms this-that filling in broadband dips effectively improves perceived performance enormously. And very narrow-band dips are virtually unobservable. Note, however, that filling in a broad, fairly deep dip in lower frequency response does require a great deal of output from the speakers.

³ I omitted the Snell B Minors here because the SigTech did not eliminate their top octave metal dome coloration. Otherwise, they too moved into stunning neutrality.

Back to REGONAUDIO