Celestion SL600si loudspeaker & DLP600 digital equalizer SL600si and DLP600 1992

John Atkinson wrote again about the SL600si in August 1992 (Vol.15 No.8) and about the DLP600 DSP equalizer:

You always meet interesting people in the stairwells of hotels. Whether this is because the more interesting a person is, the more easily they become frustrated at having to wait for an elevator, or because they feel better at having expended a little physical energy, who knows? But so it proved at the 1992 Summer CES when, working my way down the Chicago Hilton's floors one by one in my usual pedestrian manner, I bumped first into David Clark, then John Eargle (footnote 1), and found myself defending the typical audiophile's rejection of equalizers.

And reject them we rightly do, for three reasons: first, as you have no idea how your loudspeakers depart from neutrality, an equalizer can only apply an arbitrary change to the overall sound. The chance of you hitting on exactly the right combination of boost and cut all the way across the audio band to produce a flat response is zero. All an equalizer can do is to act as an enormously complicated tone control.

Second, the much-vaunted ability of multi-band equalizers to compensate for room acoustic problems is a crock. The sound you hear at the listening seat is a mix of the direct sound from the loudspeakers, the early reflections of that sound from the floor, the side and rear walls, and the ceiling, and the room reverberant field. Stick a (usually) omnidirectional microphone up at the listening position, measure the 1/3-octave spectrum of the sound, and apply the inverse of that spectrum via a 1/3-octave equalizer, as is usually done, and you'll have equalized the reverberant field to be flat. On the other hand, you'll have modified (usually) the direct sound of the speakers with a large amount of boost and cut at various frequencies, which (usually, in a normal-size, normally furnished room) is the sound that dominates the perceived tonal balance. The result, even if you haven't applied so much boost at some frequencies that your amplifier runs in perpetual clip, is sound quality that makes you want to take up another hobby (footnote 2). Or at least spend more time with your Significant Other. And children.

Third, Atkinson's Third Law of Audio states in unequivocal terms that every component you insert into the analog signal chain makes the overall sound worse, even if it effects an improvement in one small area of performance. Given that even engineers of talent cannot design a simple line-level preamplifier that is totally without character (footnote 3), what do you think are the chances of someone designing an equalizer, with its literally dozens of cheap op-amp-based circuits in parallel and generally a Mickey Mouse power supply, that doesn't savage the sound? Particularly if they're trying to reach a price point within reach of non-audiophiles? (footnote 4) Yeah, right!

So, given that conventional equalizers are quite correctly regarded in Audioland as being poor-sounding pieces of cheap, amusical junk, what I am doing reviewing what, for want of a better word, is an equalizer? Yes, I know that Celestion calls their brainchild a "Loudspeaker Processor," but they're just following the semantic lead of B&W, who termed the bass equalizer for their 800 series of speakers a "filter." The difference lies in the word "digital," the "D" in front of the "LP" in DLP600, meaning that the desired tonal changes are achieved by subjecting a digital data stream representing an audio signal to a series of mathematical manipulations.

Although it's possible to design a digital equalizer that works in the same manner as a conventional analog model, applying arbitrary boosts and cuts to predefined bands of frequencies, there's a more fundamental mode of working: The correction can be applied in the time domain, by modifying the system's impulse response. In effect, the designer works out the impulse response necessary to make that of the system conform to a target, and designs a digital filter with exactly that impulse response. System impulse plus equalizer anti-impulse equals sonic perfection; a time-domain "spin" is applied to the digital data that optimally shapes the ultimate analog soundwaves as they arrive at the listener's ears (see sidebar 3).

Paradoxically, that is where this idea's biggest drawback lies. The correction is specific to a previously defined spatial point. Anywhere else in the listening room, the correcting impulse response no longer adds to that of the loudspeaker in the appropriate manner. It might even make the sound considerably worse. The designer of the digital equalizer therefore has possibly to compromise the target response somewhat, both to allow the listener to sit without his or her head in a clamp and to avoid the sound quality being inadvertently made worse in all but the sweet spot.

A digital equalizer is considerably more versatile, however, than an analog one. First, besides correcting the loudspeaker's anechoic amplitude response aberrations on the chosen axis, the overall system phase can be adjusted independently (as there is no reason for a digital filter to have a phase response connected to its amplitude response). Second, by introducing "anti-impulses" at the same time that the speaker features predictable reflections of the sound from such baffle projections as mounting-bolt heads, the grille frame, and the edges of the baffle, the effect of these on the speaker's sound can be eliminated.

Celestion's DLP600 is a small, rectangular black box that accepts a digital data input from a CD transport or DAT machine and outputs a manipulated digital data stream that can be fed to a D/A processor. The correction applied is specific to one loudspeaker, Celestion's SL600Si. There are two data inputs, one Toslink optical, the other coaxial, the former having priority. (The input automatically switches to the optical unless a digital signal is present at the coaxial input.) There are also two data outputs, again one Toslink optical and the other coaxial. Front-panel LEDs indicate Power On, Data Lock, and Equalization In/Out, the last controlled by a front-panel pushbutton. The equalization is only applied to data sampled at 44.1kHz. While 32kHz and 48kHz data are accepted, the equalization is disabled.

Footnote 1: The name of David Clark, of course, will be familiar to readers of this magazine as the organizer of many blind ABX tests for Stereo Review and the Audio Engineering Society, the results of which have been interpreted as showing that there are no audible differences between amplifiers or between cables. Mr. Clark is a Contributing Editor to Audio magazine, as is Mr. Eargle, who is also a consultant to JBL and engineers some fine-sounding classical recordings for Delos (see J. Gordon Holt's "Engineer's Choice" feature in Stereophile, Vol.14 No.8, August 1991).

Footnote 2: Though 1/3-octave equalization is ubiquitous in recording-studio control rooms, I understand that if more than a couple of dB boost or cut needs to be applied, the engineers physically readjust the room's acoustic treatment.

Footnote 3: Although Martin Colloms and Robert Harley tell me that the Audio Research LS2 comes pretty close.

Footnote 4: I often wonder, when told by the left-brain-dominant members of the Boston Audio Society that all amplifiers can be made to sound alike by judicious uses of 1/3-octave equalization, which equalizer they are actually talking about. Or are such people content to live with the results of their thought experiments without ever feeling the need to put them to the test?