Measuring With MLSSA

Wouldn't you just know it. As soon as I decide on a formal regime of measurements to accompany Stereophile's loudspeaker reviews—see Vol.12 No.10, October 1989, p.166—along comes some hot new technology that changes everything. Robert Harley reported in last month's "Industry Update" column how impressed he and I were with the new MLSSA measurement system from DRA Laboratories.

We were sufficiently impressed that we persuaded Stereophile's publisher Larry Archibald to splash out and buy one.

MLSSA is an expanded acronym for MLS or "Maximum Length Sequence." The system's progenitors, Douglas D. Rife of DRA Labs and John Vanderkooy of the University of Waterloo, explained at some length in the June 1989 issue of the Journal of the Audio Engineering Society (footnote 1) what a Maximum Length Sequence is and how it can be used as the basis for an audio measurement system. A sonic MLS sounds like white noise and, indeed, it appears to have the spectrum of white noise. But as it is a pseudo-random signal, actually having a formal structure, a dedicated analyzer can compare the original MLS signal with the version of it output by the DUT (Device Under Test). By performing a cross-correlation operation, it can derive the DUT's impulse response. Performing a Fast Fourier Transform (FFT) on this impulse response will then give the DUT's transfer function (amplitude and phase responses).

If the DUT is a loudspeaker, then as long as the computed impulse response is free from room reflections, the system will automatically produce the speaker's anechoic frequency response without recourse to expensive anechoic chambers. In addition, by successively windowing the calculated impulse response one time sample along and performing the FFT, the system shows how the frequency response equivalent to that impulse response decays with time, revealing the presence of cabinet and drive-unit resonances. The Heyser Energy-Time Curve can also be easily calculated. If the calculated impulse response is long enough to include all the room reflections, then performing the FFT will give the transfer function of the speaker and the room, allowing the reverberation time, intelligibility index, and direct/reflected ratio to be calculated. By connecting a high-value resistor in the hot connection to a loudspeaker, the speaker's complex impedance can be calculated by looking at the impulse response, hence transfer function, of the resultant voltage divider.

MLSSA is undoubtedly a powerful tool, but it actually appears to offer nothing more than what is already offered by traditional FFT or TDS techniques. Why then is Stereophile so excited about it?

• Unlike narrow-pulse–based FFT testing—the system I have used until now, for example—the MLS stimulus signal has a very narrow dynamic range; ie, it can be played at reasonable levels without the danger of driving the DUT into nonlinear behavior. It intrinsically, therefore, has a sufficiently high signal/noise ratio that it will operate in reasonably noisy environments (like a typical listening room) without the user having to average a large number of samples.

• As well as being accurate, it's fast. I measured all the loudspeakers reviewed in this issue in just one normal working day. With my old system, using a Heath-Zenith 8-bit 'scope and a QuickBasic FFT program I developed myself, this would have taken twice as long (footnote 2) and I hadn't even started to write the software (assuming that I could) to perform all the fancy post-processing functions MLSSA includes as a matter of course.

• It's cheap. Just $2750 buys the complete package: a long card to fit in one of a PC's expansion slots, and all the software. The card contains the MLS stimulus generator (which can also be programmed to produce pulse and squarewave signals), a 12-bit A/D converter sampling at up to 150kHz, a 9-bit DAC, and a 4- or 8-pole anti-aliasing filter, programmable as to bandpass and type, all completely under software control. The system can also be used as a conventional digital-storage oscilloscope, with then its comprehensive FFT capabilities used to measure distortion. About the only practical limitation is that the host PC must carry a math coprocessor and have a hard disk, no big deal these days. We have installed Stereophile's MLSSA system in a Compaq "luggable" 80286 computer so that we can easily assess our reviewers' listening rooms as well as easily and quickly measure review loudspeakers in Dick Olsher's, Robert Harley's, Larry Archibald's, and my rooms.

I envisage MLSSA being a useful tool in Stereophile's long-term quest to provide some correlations between what is measured and what is heard. I heartily recommend it.—John Atkinson

Footnote 1: "Transfer-Function Measurement with Maximum-Length Sequences," Douglas D. Rife and John Vanderkooy, JAES, Vol.37 No.6, June 1989, p.419.—John Atkinson

Footnote 2: For interest's and paranoia's sakes, I used MLSSA to analyze the impulse responses captured by the Heath 'scope for the loudspeaker reviews published in Stereophile since May 1989. The anechoic responses were effectively identical to those published, which came as a big reassurance considering John (Waveform) Ötvös's crack in the December 1989 issue (p.243) about "journalists attempting to bolster the technical credibility of their subjectivity, using...lap computers with the latest trendy FFT programs." Trendy they may be, Mr. Ötvös, but if they're accurate, then why get so hot under the collar?—John Atkinson