Sounds a great read. I have fond memories of Quads. I did have a Quad 405.1 power amp for some time in the 70's and 80's.

I look forward to receiving July's Issue.

No, phase shift vs frequency is not in itself indicative of phase distortion. If the amount of phase shift is linearly dependent on frequency (ie if the phase shift at 2f is twice that at f) then this is equivalent to a pure time delay, and no alteration of the waveform occurs. Whereas if the phase shift is not linearly dependent on frequency, waveform change can occur on complex signals. This is phase distortion. There are various means of depicting it graphically, of which group delay is probably the most widely used but arguably not the best. For more on this, if you're interested, read Marshall Leach's AES Journal paper "The Differential Time-Delay Distortion and Differential Phase-Shift Distortion as Measures of Phase Linearity" (obtainable as a download, for a fee, from www.aes.org). All the analogue low-pass and high-pass filters normally encountered in audio introduce phase distortion, the audibility of which has been controversial for decades. FIR (finite impulse response) digital filters, by contrast, are generally linear-phase (ie non-phase-distorting) although this isn't necessarily the case.

Sorry for the explanation being so condensed: this would take a lot of space, and many diagrams, to explain satisfactorily from the ground up.

Pleased you liked the article.

Quote:
No, phase shift vs frequency is not in itself indicative of phase distortion. If the amount of phase shift is linearly dependent on frequency (ie if the phase shift at 2f is twice that at f) then this is equivalent to a pure time delay, and no alteration of the waveform occurs. Whereas if the phase shift is not linearly dependent on frequency, waveform change can occur on complex signals. This is phase distortion. There are various means of depicting it graphically, of which group delay is probably the most widely used but arguably not the best. For more on this, if you're interested, read Marshall Leach's AES Journal paper "The Differential Time-Delay Distortion and Differential Phase-Shift Distortion as Measures of Phase Linearity" (obtainable as a download, for a fee, from www.aes.org). All the analogue low-pass and high-pass filters normally encountered in audio introduce phase distortion, the audibility of which has been controversial for decades. FIR (finite impulse response) digital filters, by contrast, are generally linear-phase (ie non-phase-distorting) although this isn't necessarily the case.

Sorry for the explanation being so condensed: this would take a lot of space, and many diagrams, to explain satisfactorily from the ground up.

Pleased you liked the article.

Thanks for the response. I get your explanation about phase distortion being a nonlinearity in the phase vs. frequency relationship. But I still don't understand both why frequency vs. phase change that is linear is not a problem and why a speakers electromechanical rolloff causes phase distortion. Yes, I am to cheap and probably to uneducated to read and understand the paper you refered to.

Quote:
No, phase shift vs frequency is not in itself indicative of phase distortion. If the amount of phase shift is linearly dependent on frequency (ie if the phase shift at 2f is twice that at f) then this is equivalent to a pure time delay, and no alteration of the waveform occurs. Whereas if the phase shift is not linearly dependent on frequency, waveform change can occur on complex signals. This is phase distortion. There are various means of depicting it graphically, of which group delay is probably the most widely used but arguably not the best. For more on this, if you're interested, read Marshall Leach's AES Journal paper "The Differential Time-Delay Distortion and Differential Phase-Shift Distortion as Measures of Phase Linearity" (obtainable as a download, for a fee, from www.aes.org). All the analogue low-pass and high-pass filters normally encountered in audio introduce phase distortion, the audibility of which has been controversial for decades. FIR (finite impulse response) digital filters, by contrast, are generally linear-phase (ie non-phase-distorting) although this isn't necessarily the case.

Sorry for the explanation being so condensed: this would take a lot of space, and many diagrams, to explain satisfactorily from the ground up.

Pleased you liked the article.

Hi Keith,

I just stumbled across this article posted on the website yesterday when searching for some other info. Your discussion and first investigation into this was a very good read and quite well noted considering a short novel could be written on the details of the problem.

I spend more of my time and work in the Home Theater realm of audio, but my experience parallels much of what you have observed and cited from the work done at B&W.

With some of the product work I did in the professional audio market with Tom Danley, he was able to make some very significant improvements in the time-based performance of a speaker, and I ended up doing a lot of further investigation and experimenting with adjusting the subwoofer-main speaker interaction in a typical home theater setup. In most every case I have observed a measureable reduction in the total "smearing" in time of a system (without hugely compromising other performance factors), the observation has been that of improved sound, especially on precussive or plucked instruments.

There is of course that important point of not compromising other basic performance characteristics in the process. Linear output at varying levels and total power response are of course the most commonly compromised.

I wanted to also note that there are now a variety of currently available tools from various markets that will allow much further experimentation from what you describe in your article. Those options range from FIR filtering of an impulse response using a PC, to professional audio products with FIR capability which can be programmed with almost any MatLAB function you can dream up. The other option to consider in the subwoofer section ties into your explanation of boundary interaction.

While some have chosen to force subwoofers to have flat half space response to 8-14Hz in search of the phase/group delay profile you describe, in any real, enclosed room this is not what you get. While what I'm describing should be nothing new, note that if the room is even remotely enclosed, you will observe a 6-12dB/octave gain starting around the half-wavelength equivalent to the long diagonal of the room (upper to lower opposite corner). At these low frequencies and long wavelengths, the phase/time behavior is indeed the combined response of the subwoofer and room. Of course a conventional radiator still has a ~90 deg phase shift in its flat region, but this certainly gives you a much more managable task to tackle with DSP. I have had sucess in observing flat in-room response to the 5-12Hz range with modest EQ when using subwoofers with a sealed roll off that is not high passed electronically. Many sealed subs on the market do have high pass filters in the chain, so you can't presume that any sealed sub will perform as I describe, but this would allow for a next step in the experimentation you have already written about.

The other part of the equation that gets a little tricky is actually being sure of the phase measurements you are taking. Most measurement systems assume a linear phase system, where this is often true over various frequency ranges, this falls appart as soon as you have changing directivity of a speaker and a few other cases. The only system I am aware of that takes a direct measurement of acoustic phase are those employing a TDS measurement, namely the Gold-Line TEF systems or the newer EASE-RA system. The MLSSA system will be plenty accurate to make marked improvements, but it is possible to make wide-band measurements look better than reality.

If you plan to do further testing or experimenting, do feel free to contact me or continue discussion here.

Best Regards,

Mark Seaton
Seaton Sound, Inc.