On a recent visit with Don and Carolyn Davis, Don showed me a letter that was written to him by Paul Klipsch in 1959. I was still in diapers at the time.

He and Don had been discussing various aspects of distortion earlier that day. Mr. Klipsch’s comments should be of interest to anyone engaged in audio system design, since he points out things that can and can’t be corrected in loudspeakers and other system components.

I was mistaken this afternoon in defining distortion broadly instead of nonlinear distortion. This latter gives rise to new frequencies not originally present and which cannot be restored.

Frequency distortion (which I prefer to call discrimination) and phase distortion (which I prefer to call time delay effect) are both restorable; for example one may cut a record on the RIM curve and restore it on playback with the proper complimentary curve, but amplitude distortion, the real culprit, is not restorable.

In an audio system one may minimize the high frequency distortion components by introducing frequency discrimination, but the original state of things can never he restored. That is why I personally reserve the term “distortion” to mean “non-linear” distortion, which has no restorative treatment, in contrast with frequency or phase effects which may be restored - at least theoretically and practically in simple cases.

Modulation distortion is a form of nonlinear distortion and arises only when nonlinear conditions exist. I cannot conceive of a way by which modulation distortion could occur except by some form of non-linearity which is the basic form of distortion I like to refer to as distortion.

For my purposes, then, Distortion (capital D), or non-linear distortion, could be defined in the following ways:

1. It introduces frequencies not originally present, and (not or; and)
2. It is non-restorable in that a “complimentary” net work could not be employed to cancel the effect.

Figure1 (click on image for full size)

An example of a nonlinear distortion would be distortion due to clipping. Figure 1 shows a 400 Hz tone as viewed on an oscilloscope and on a spectrum analyzer. The bottom graph shows the same signal after having been clipped by a mixer. Note the frequencies present in the output that are not present in the input. This is not a “restorable” form of distortion.

Figure 2 shows an example of what Mr. Klipsch referred to as “frequency distortion.” A spectrally flat signal is sent to a loudspeaker, but the signal measured at the test microphone position has bumps in its response. The minimum phase bumps can be compensated for in both magnitude and phase with some appropriate filters (Figures 3 and 4). Pb.

Fig.2: (click on image for full size) Magnitude and phase response of small bookshelf loudspeaker. Note the “humps” in the 800 Hz to 8 kHz region. The phase response associated with each bump suggests that it can be compensated for with a band-reject filter. If this is true, both the magnitude and phase curves should be “smoothable” with the application of the proper filters.

Fig. 3: (click on image for full size) The parametric equalizer section of TOA DACsvs processor parametric filters are adjustable in frequency, Q and magnitude. By tuning the filters, a “conjugate” of the loudspeaker’s response is created in both magnitude and phase. The magnitude is what is observed when adjusting the filter. The phase response comes along “for the ride.” All-pass filters allow the phase to be adjusted independently of the magnitude.

Figure 4: (click on image for full size) The combination of two “filter sets,” one being the loudspeaker and the other being the parametric equalizer. Note that the combined result of the responses is much smoother in both magnitude and phase than the original loudspeaker response. Thus, the humps in the loudspeaker response represent a ‘restorable” form of distortion. This type of equalization is the first step in equalizing a sound reinforcement system.