Regarding "Phase Distortion Audibility"
In your latest article Human Hearing, Part 2, Phase Distortion, at:
http://www.audioholics.com/techtips/roomacoustics/HumanHearingPhaseDistortio.php
there are some questionable presentations of technical information.
The very first figure on page one (marked via hand notation "C"), the one showing the waveform distortion of what appears to be a square wave after it has been run through a 4th order Linkwitz-Riley crossover, has several misleading and omitted details.
First, it should clearly say that the bottom trace is the SUMMED output of the crossover outputs, that is, the summation of the high pass and the low pass outputs of the 4th order L-R crossover. Yes, the notation on the diagram indicates this to a professional, but not clearly for very many layman. It should be spelled out in the text.
Second, the waveform is NOT a portion of a square wave, even though many people would look at it and think that it was, it is actually two long duration "pulses", of unknown duration, separated by a similar amount of time as the pulse duration. The clue to this, is that the pulses are rising up from the baseline of the waveform, but never go below the baseline.
(The spectrum analysis also supports this, as the spectral pattern indicates raised pulses, rather than a portion of a square wave)
If the time scale is the same as for the 2nd figure, which is not an unreasonable assumption given the way they are presented, then it would appear that the duration of these pulses is 10 mS.
We also do not know what frequency the L-R crossover is set to, although it would appear to be some where higher in frequency than the apparent period of the waveform shown.
One could also look at these as if they were a sequence of 1 and 1/2 square wave cycles, that have been offset from the baseline of 0 volts by 1/2 the peak to peak value of the square wave.
Third, from looking at this kind of result, one could erroneously draw the conclusion that a 4th order L-R crossover has no difference between the steady state waveform distortion and the transient waveform distortion, while in fact, there IS a difference for these two conditions.
The bottom waveform shown in that first figure does not appear to be entirely correct, in that it does not show the transient distortion due to the start and stop of the pulse train (or off-set squarewave).
This transient distortion can be clearly seen at the very start or final discontinuation of a more or less steady square wave signal, as having a different appearance once the initial or final waveform transient has occured, and one is within the "steady-state" portion of the waveform signal, which would be all wave transitions within the confines of the waveform excluding those first and last transitions.
For an illustration of this, see:
"A Novel approach to Linear Phase Loudspeakers Using Passive Crossover Networks", Erik Baekgaard, JAES Vol 25, No 5 (1977, May), specifically Fig. 14.
Now it may be that the pulses will not show this to the fullest, and it would take a true square wave with the transistions going both above and below the 0 volt baseline in order to fully manifest, but the point is, that typical audio signals will exhibit this kind of transient behavior, rather than the totally asymmetrical pulse behavior shown.
Perhaps this is nit-picking, but I find it relevant in the way this data is presented, along with the various comments on the published articles, in that it appears to be presented from the point of view that phase distortion does not matter for typical musical material, rather than as an exploration of wheteher or not this is so. In other words, the editor and/or author had already made up his mind, and was presenting the data solely in a way to back up his POV.
Otherwise, why go to the trouble to use such an unusual waveform, rather than a more typical square wave?
The main reason that it seems to me that the article and data are presented with a twist to them, is the inclusion of selected comments by Dr. Floyd Toole, which I presume was the work of Gene as the editor. Dr. Toole is quoted in such a way as to minimize the importance of phase shift audibility:
Quote: "It turns out that, within very generous tolerances, humans are insensitive to phase shifts."
Now if one reads the original papers, and the conclusions of the original author(s), and even reads the writings of Mr. Sanfilipo, without these added comments by Toole, one could easily come to the conclusion that there were indeed instances where phase shift was audible, and that maintaining correct phase relationships seems to be a desirable thing to do, as long as doing so does not severly degrade other parameters.
But the added comments tend to muddy those waters more than they really should.
I feel that correcting and clarifiying the diagrams on the first page might help matters some, as well as including less added comments by Dr. Toole, without the counter-balance of another POV present, including those of the original authors of the referenced papers.
It should be noted for the record that virtually none of the speakers that Dr. Toole is associated with in the form of Harmon International, maintain correct phase relationships across the audio spectrum, inso much as it is possble to do so given the typical bandwidth limitations of loudspeaker systems. This might be a potential reason for him to minimize the importance of phase distortion.
Finally, I would like to point out that the final conclusion #3 at the end of the article is really only signficantly true for a room with no room acoustics treatment. In an acoustically well treated room, phase distortion, as well as envelope distortion, is more readily observed and noticed.
Jon Risch
