AES Journal Forum

I was very interested to see that this paper promotes the idea that non-linearity measurements based on multi-tone test signals might be more appropriate for audio. The British Broadcasting Corporation came to this conclusion in the 70's and actually produced a multi-tone test that gave good agreement between double-blind subjective assessments and objective measurements of audio non-linear distortion. The method had the disadvantage that the digital circuitry caused the test set to be more expensive than a conventional one. The patents have long since expired and the test method is now part of IEC standard 60748-4-3 for dynamic testing of ADCs. As it is now easy to implement using software on a PC and an audio interface, perhaps this comment will encourage others to investigate what this technique might now offer when they are making audio measurements.

A variation of your method was presented in a paper by the late Deane Jensen and Gary Sokolich, "Spectral Contamination Measurement", at the AES 85th Convention, 1988 (preprint #2725). In their (and my) opinion, much of what audiophiles describe as "the veil" consists of low-level, non-harmonic distortion products. I believe a major source of these is due to widespread use of op-amps under heavy negative feedback. As frequency increases (at ultrasonic frequencies, for example), the op-amps essentially run out of open-loop gain and start approaching open-loop conditions. This allows their inherent non-linearity to intermodulate distortion products from previous stages and the result is non-harmonic distortion products. It is one very good reason for passive bandlimiting at the input of every active amplifier stage. Otherwise, the effect is cumulative - and demonstrably ugly sounding. The test Jensen and Sokolich proposed is very similar to the AP (Cabot) multi-tone test.

Bill Whitlock
President & Chief Engineer
Jensen Transformers, Inc.
(and recent AES Fellow)