Speaker time alignment, does it matter?

[QMuse]

I think this cannot be settled until you try with the exact same speaker and blindly have someone turn phase correction on and off. It would be possible with any HEDD speaker for example, since they do phase correction via a VST plugin that you could turn on and off.

I have tried exactly that and wasn't able to hear any difference. Btw, such experiments have been tried many times in various studies and general conclusion is that you need low reflection room to hear the difference and that the difference is subtle even when listening in a non-anechoic environment.

 

[thewas]

When looking at step response reflections should be filtered with FDW otherwise you are getting a mess, and that's the same reason why it is hard to hear phase correction in a non-anechoic environment.

Regarding vertical angles, usually the ratio between drivers distance vs distance to LP is large enough for that not to matter. For example, I don't see practically any change in step response when I measure at 4 meters distance at 80cm height (ears height at sitting position) and 180cm height (ears height at standing position).

As you say usually at usual recreational listening with high distances where reflected sounds dominates it doesn't really make a difference, but rather at monitoring in the near field with damped room acoustics but also there the audible difference is not large.

 

[QMuse]

As you say usually at usual recreational listening with high distances where reflected sounds dominates it doesn't really make a difference, but rather at monitoring in the near field with damped room acoustics but also there the audible difference is not large.

Exactly. Although even in a segment of professional studio monitors only a small number of speakers are phase corrected for clean step response. Most of the monitors with built-in DSP use it only to fix frequency response and not for phase correction.

 

[thewas]

The difference between the two systems, despite having essentially flat AFR, was not subtle. The phase coherent speaker was able to produce louder transients, despite both systems being SPL matched. The other stark difference was in soundstage - both systems could portray almost identical apparent width, however the phase coherent system portrayed depth much better and localisation was more distinct, probably due to point source radiation.

The question is though are those audible differences really only related to the phase coherence or also or mainly to other phenomena like different radiation pattern, baffle step and diffraction artefacts?

 

[Snarfie]

I got eyestrain from peering into the time windows from DCM I had, so I got Vandersteens.

+1
I have the Vandersteen Model 1 just bought last year. In combination with Room correction software/DSP they sound considering my specific room acoustics just amazing staging is breathtaking. I think the combination of taking care of time-alignment en phase coherent behavior fixed by hardware (based on first order cross overs, custom made speakers/baffles an what more ) an DSP did the job. The only disadvantage they need enough space from walls to let them sound amazing. 20cm off in a specific direction an they fall apart still not bad but the magic is not their.

Enclosed interview with Richard gives some inside in his development process. http://www.soundstagenetwork.com/interviews/int07.htm
Specific this part is interesting where he explain (at the time) that digital measuring equipment support his way of development over almost 30 years.

dB: What are we missing or what don’t most people know or understand about the relationship between the audibility of amplitude differences versus time/phase differences?

RV: Here’s an example: If you set up a test so that you can vary frequency response over a controlled frequency band, say half an octave. And you set the test up very carefully so that when you increase or decrease the sound level in this half octave, that there is no change at all in the time and phase domains, you find that 3dB differences are just barely audible. But if you make a 0.5dB change and shift the phase at the same time, the difference is immediately noticeable. Only when phase or time delay were changed along with amplitude did the 0.5dB level changes become obvious. The digital test equipment that has been available for a while is really useful for this kind of testing. You can select any level change you want and get it with zero time/phase changes. You can also select time delays and phase shifts with or without level changes. It was harder to do this kind of testing in the pre-digital days, but we’d done some things just well enough to know we we’re on the right road. Today’s digital test equipment just helps prove the point even more convincingly. Time/phase distortion is simply more audible to us than most of us realize or recognize. The magazines and most manufacturers don’t get this yet. [grins] After all, it has only been 20+ years since time delay and phase shift were looked at seriously by us and some other companies. [still grinning]

 

[QMuse]

+1
I have the Vandersteen Model 1 just bought last year. In combination with Room correction software/DSP they sound considering my specific room acoustics just amazing staging is breathtaking. I think the combination of taking care of time-alignment en phase coherent behavior fixed by hardware (based on first order cross overs, custom made speakers/baffles an what more ) an DSP did the job. The only disadvantage they need enough space from walls to let them sound amazing. 20cm off in a specific direction an they fall apart still not bad but the magic is not their.

Enclosed interview with Richard gives some inside in his development process. http://www.soundstagenetwork.com/interviews/int07.htm
Specific this part is interesting where he explain (at the time) that digital measuring equipment support his way of development over almost 30 years.

dB: What are we missing or what don’t most people know or understand about the relationship between the audibility of amplitude differences versus time/phase differences?

RV: Here’s an example: If you set up a test so that you can vary frequency response over a controlled frequency band, say half an octave. And you set the test up very carefully so that when you increase or decrease the sound level in this half octave, that there is no change at all in the time and phase domains, you find that 3dB differences are just barely audible. But if you make a 0.5dB change and shift the phase at the same time, the difference is immediately noticeable. Only when phase or time delay were changed along with amplitude did the 0.5dB level changes become obvious. The digital test equipment that has been available for a while is really useful for this kind of testing. You can select any level change you want and get it with zero time/phase changes. You can also select time delays and phase shifts with or without level changes. It was harder to do this kind of testing in the pre-digital days, but we’d done some things just well enough to know we we’re on the right road. Today’s digital test equipment just helps prove the point even more convincingly. Time/phase distortion is simply more audible to us than most of us realize or recognize. The magazines and most manufacturers don’t get this yet. [grins] After all, it has only been 20+ years since time delay and phase shift were looked at seriously by us and some other companies. [still grinning]

To my knowledge there are no blind test to confirm such claim, but there are many who confirm the opposite is the case. Btw, whenever you apply gain at a certain frequency phase will change a little but that change is far from being audible. In fact, it has been proven that even a huge change in phase is not audible.

Btw, I'm also suspicious about an expert who uses the term "time and phase domain" as that is virtualy the same thing.

 

[tuga]

No, when you time-align speaker with passive XO it is valid throughout the room.

But if you align the drivers physically won't the alignment be valid for a small area only?

 

[BenB]

+1
I have the Vandersteen Model 1 just bought last year. In combination with Room correction software/DSP they sound considering my specific room acoustics just amazing staging is breathtaking. I think the combination of taking care of time-alignment en phase coherent behavior fixed by hardware (based on first order cross overs, custom made speakers/baffles an what more ) an DSP did the job. The only disadvantage they need enough space from walls to let them sound amazing. 20cm off in a specific direction an they fall apart still not bad but the magic is not their.

Enclosed interview with Richard gives some inside in his development process. http://www.soundstagenetwork.com/interviews/int07.htm
Specific this part is interesting where he explain (at the time) that digital measuring equipment support his way of development over almost 30 years.

dB: What are we missing or what don’t most people know or understand about the relationship between the audibility of amplitude differences versus time/phase differences?

RV: Here’s an example: If you set up a test so that you can vary frequency response over a controlled frequency band, say half an octave. And you set the test up very carefully so that when you increase or decrease the sound level in this half octave, that there is no change at all in the time and phase domains, you find that 3dB differences are just barely audible. But if you make a 0.5dB change and shift the phase at the same time, the difference is immediately noticeable. Only when phase or time delay were changed along with amplitude did the 0.5dB level changes become obvious. The digital test equipment that has been available for a while is really useful for this kind of testing. You can select any level change you want and get it with zero time/phase changes. You can also select time delays and phase shifts with or without level changes. It was harder to do this kind of testing in the pre-digital days, but we’d done some things just well enough to know we we’re on the right road. Today’s digital test equipment just helps prove the point even more convincingly. Time/phase distortion is simply more audible to us than most of us realize or recognize. The magazines and most manufacturers don’t get this yet. [grins] After all, it has only been 20+ years since time delay and phase shift were looked at seriously by us and some other companies. [still grinning]

Changing the frequency magnitude response without changing the time domain (impulse response) simply isn't possible. It could theoretically be possible to change the frequency magnitude response without changing the frequency phase response, using DSP. From everything I've read from anyone not named Richard Vandersteen, it is subtle magnitude distortions that are much more audible than large phase changes.

 

[BenB]

As was I, until I did the listening test. The big question is whether our brains actually can discern between this:


And this:

As you can tell, the phase coherent setup can physically reach 25% higher sound pressure transient due to the fact that the tweeter wavefront rides on the rest and sums properly. I attribute that to the higher perceived loudness on sharp transients. On more sustained signals the loudness difference wasn't that apparent.



I'm skeptical on the benefits of time aligning non-coincident driver systems. PSI and Dutch&Dutch seem to think otherwise.

I don't see how you can be confident that any changes to your perception of the sound were caused by the different phase characteristics, when you were listening to 2 difference systems that have different frequency responses, different dispersion, and different distortion. Even if the on-axis frequency response is very well matched, differences in dispersion characteristics will swamp any differences in phase.

 

[QMuse]

Changing the frequency magnitude response without changing the time domain (impulse response) simply isn't possible. It could theoretically be possible to change the frequency magnitude response without changing the frequency phase response, using DSP. From everything I've read from anyone not named Richard Vandersteen, it is subtle magnitude distortions that are much more audible than large phase changes.

Yes, it can be done with DSP but it would cause pre-ringing.

 

[BenB]

Yes, it can be done with DSP but it would cause pre-ringing.

Well certainly they can't both be minimum phase.

 

[QMuse]

Well certainly they can't both be minimum phase.

Exactly.

 

[KaiserSoze]

I participated in a blind test between my Neumann KH310 speakers and a prototype speaker from a local hi-end monitor company.

The KH310, while a highly proficient device, hasn't been designed in a way to do anything with time coherence.
...
The difference between the two systems, despite having essentially flat AFR, was not subtle. The phase coherent speaker was able to produce louder transients, despite both systems being SPL matched. The other stark difference was in soundstage - both systems could portray almost identical apparent width, however the phase coherent system portrayed depth much better and localisation was more distinct, probably due to point source radiation.

Needless to say, I sold my KH310's shortly after.

Yet we are told by many of the smartest technical people in the industry that there isn't any evidence that time/phase coherency is audible except with respect to achieving a flat response through the crossover regions (in which case it is really the phase coherency that matters and time coherency is merely a means to that end).

For me it is difficult to let go of my early experiences. In the early '70s, when I was 19 or 20, in a big old record store that also sold some loudspeakers, I heard a speaker that I'd never heard of. It was immediately apparent to me that there was something different about the speaker I was listening to. This was before college, when I was very technically naive. The music that was playing had some percussion instruments, and to me it was apparent that this speaker was reproducing those sounds in a way that was more realistic than I was used to hearing through amplified loudspeakers. The difference was so stark that my memory of the experience including my reaction remains vivid. Five or six years later I related this experience to a professor that I'd become acquainted with, who had taught electronics and who was a hobbyist, having built several speakers and tube amps. To describe to him what I had heard with this speaker that seemed special, I said that it was able to reproduce the sounds similar to a finger-snap, in a way that most other speakers couldn't. He then introduced me to the concepts of "transient response" and "time domain". Given the chronology of this experience, it is difficult for me to believe that I was just imagining having heard this sound quality, which I identified and understood in an unsophisticated but meaningful way, years before I was the least bit aware of the formalized concepts of transient response and the time domain, or was aware that Kloss had designed these speakers specifically with the intent to achieve these very qualities that I had heard.

We are continually told by some very smart people that steep crossover slops are superior for the various good reasons they give, and they tell us that there are no unwanted consequences of the steep slopes, or of anything else that is harmful to the transient response (i.e., the Helmholtz resonator).

On a few past attempts to discuss this with different people, I have referred to the sound of high quality headphones. I would say something like, "A major reason that these headphones sound so much better than loudspeakers is that there is a single driver with no crossover. The response is also very flat throughout the audible range, such that there is very little phase wandering and thus superior transient response." The typical response I've gotten is something like, "Headphones? I don't listen to headphones." People with a modicum of knowledge will of course mention the avoidance of room effects. But rarely does anyone draw attention to the potential for exceptionally good transient response in a good set of headphones, or to the potential that this is the reason that good headphones sound good (for people who hear a quality in headphones that they don't hear in speakers and who like headphones for this reason). My point is not to suggest headphones as an alternative. My point is that, if in fact this is a property that good headphones have and that isn't shared with most speakers, and if it is detectable to most people (notwithstanding that people don't typically understand why headphones sound different in a way they like), then audibility of good vs. bad transient response is real, contrary to the prevailing expert opinion.

Yet, if it is audible, why are the many learned experts so convinced that it isn't? Is it possible that bias interferes with their ability to conduct a proper test for audibility of differences in transient response fidelity? To conduct a proper test of this hypothesis, it would be necessary to build two speakers that are identical in every respect except that in one, phase undergoes rotation from low to high speaker in a manner like the graph above for the KH310, while in the other, there is very little phase rotation, ideally none at all. This would be a very difficult thing to do, and as such, there may never have been a proper test of this question. Perhaps with a single-driver speaker, listened to close up, digitally equalized to achieve a near-flat response, one without any phase rotation intentionally introduced, the other with phase rotation introduced to mimic speakers like the KH310. Is it possible to do something like this with DSP?

 

[QMuse]

There's really no need for long discussion as the answer why we can't hear time aligned speaker is really simple, and once again it is all about reflections.

Take a look at the phase response of Kef LS50 which has been corrected perfectly for phase response. As you can see phase is smooth and flat and excess phase is 0 throughout the entire frequency range.

Looks really good, isn't it? Indeed it does, except that reflections were removed with FDW of 6 cycles. Now let's have a look how it looks when reflections are added when FDW is removed:

Not pretty anymore, right? No wonder you can't hear the result of that perfect phase correction/time alignement anymore with all this mess coming from added reflections arriving each of their own at different times.

 

[BenB]

On a few past attempts to discuss this with different people, I have referred to the sound of high quality headphones. I would say something like, "A major reason that these headphones sound so much better than loudspeakers is that there is a single driver with no crossover. The response is also very flat throughout the audible range, such that there is very little phase wandering and thus superior transient response."

If you like the sound of headphones, and if you want to test whether you like it because it lacks the phase distortion that crossovers cause, it's very easy to test.

Create an all-pass filter that adds phase distortion modeled after crossovers, and insert that into the signal path when you listen to headphones. If the phase distortion ruins the sound of headphones for you, then that's the culprit. I did exactly this test over a decade ago using foobar, which I used in two different ways. First, it allows you to run the convolver (you probably need a plug-in for that) and toggle it on and off. Secondly, it allows for ABX testing (another plug-in). In order to use that to test the perceptibility of phase distortion, I started with numerous high quality audio samples, and created phase-distorted versions in matlab. I then used foobar to test myself on whether or not I could hear the difference.

Eventually, I was able to find combinations of crossover slope and frequency that caused audible differences. I think the threshold for me was 4th order at 1 kHz. Higher crossover orders, and lower frequencies made it easier to hear. I don't really remember what happened when I pushed the frequencies super low. I think the perception may have maximized in the 200-400 Hz range. Below 100 Hz there isn't too much transient energy. Even when I found this audible range, the difference was subtle and it wasn't at all clear to me that I had any preference for the original clips over the phase distorted versions.

I did try the same audio clips on my speakers, and I was able to pass ABX tests with high order, low frequency (4th order at 250 Hz?) phase distortion, despite room reflections. Again, without preference.

 

[steve59]

These dsp's I'm using have bass alignment. Their literature says for bass to arrive at the same time the woofers would need to be 27 feet closer/farther apart, anyhow it is switchable and the few times I turned it off I can't say either way. Maybe a larger room or listening in the far field would show more obvious differences.

 

[KaiserSoze]

There's really no need for long discussion as the answer why we can't hear time aligned speaker is really simple, and once again it is all about reflections.

Take a look at the phase response of Kef LS50 which has been corrected perfectly for phase response. As you can see phase is smooth and flat and excess phase is 0 throughout the entire frequency range.

View attachment 69872

Looks really good, isn't it? Indeed it does, except that reflections were removed with FDW of 6 cycles. Now let's have a look how it looks when reflections are added when FDW is removed:

View attachment 69873

Not pretty anymore, right? No wonder you can't hear the result of that perfect phase correction/time alignement anymore with all this mess coming from added reflections arriving each of their own at different times.

uh oh.

I'm looking for you to now explain why you are asserting that the presence of room reflections (which are obviously not coherent with respect to the sound coming directly from the speaker), make it impossible for a listener to hear the difference between a speaker with ideal transient response and a speaker where phase rotates through 360 degrees multiple times from low frequency to high frequency. It is obvious that this is the assertion you have made, but you didn't make any effort at all to back up this assertion. Instead, you included these graphs as though you expected the graphs would support this assertion, but all they really do is demonstrate that which is patently obvious and not the least bit subject to dispute, i.e., that reflections are not coherent with the sound directly from the speaker.

Please note also that when you say, "No wonder you can't hear ...", this statement takes your assertion as a forgone conclusion, thereby suggesting through innuendo that your assertion is an established fact beyond dispute. This is argumentatively impolite.

 

[KaiserSoze]

If you like the sound of headphones, and if you want to test whether you like it because it lacks the phase distortion that crossovers cause, it's very easy to test.

Create an all-pass filter that adds phase distortion modeled after crossovers, and insert that into the signal path when you listen to headphones. If the phase distortion ruins the sound of headphones for you, then that's the culprit. I did exactly this test over a decade ago using foobar, which I used in two different ways. First, it allows you to run the convolver (you probably need a plug-in for that) and toggle it on and off. Secondly, it allows for ABX testing (another plug-in). In order to use that to test the perceptibility of phase distortion, I started with numerous high quality audio samples, and created phase-distorted versions in matlab. I then used foobar to test myself on whether or not I could hear the difference.

Eventually, I was able to find combinations of crossover slope and frequency that caused audible differences. I think the threshold for me was 4th order at 1 kHz. Higher crossover orders, and lower frequencies made it easier to hear. I don't really remember what happened when I pushed the frequencies super low. I think the perception may have maximized in the 200-400 Hz range. Below 100 Hz there isn't too much transient energy. Even when I found this audible range, the difference was subtle and it wasn't at all clear to me that I had any preference for the original clips over the phase distorted versions.

I did try the same audio clips on my speakers, and I was able to pass ABX tests with high order, low frequency (4th order at 250 Hz?) phase distortion, despite room reflections. Again, without preference.

I certainly do admire your determination in finding out for yourself whether you can detect whether a bunch of phase rotation has been applied to the sound. This is an outstandingly good suggestion, however at present I do not possess the skill, the tools, or the ambition to do something like this. But please know that you have planted the seed in my brain, so maybe I'll take a slow but steady approach to being able to do an experiment of this very sort. I certainly would like to, because it would settle something that has been a minor frustration for forty+ years.

 

[BenB]

I certainly do admire your determination in finding out for yourself whether you can detect whether a bunch of phase rotation has been applied to the sound. This is an outstandingly good suggestion, however at present I do not possess the skill, the tools, or the ambition to do something like this. But please know that you have planted the seed in my brain, so maybe I'll take a slow but steady approach to being able to do an experiment of this very sort. I certainly would like to, because it would settle something that has been a minor frustration for forty+ years.

When I did that test, I tried to get forum members to repeat it (might have been the madisound forum, or PE... I can't remember at this point), but I didn't get much traction. This forum is a lot more technical, so it may go over better here. I can generate all-pass filters to mimic the phase distortion of various crossovers, and post those. I could also filter some short audio files and post in a lossless format, but that opens up a whole discussion about what to include. I don't have that much free time to commit to this. It would be great to get some help from others if they wanted to provide instructions for downloading foobar (or something similar) and installing the appropriate plug-ins. Also, anyone with matlab or octave (which is free) can perform the filtering... there are probably many others programs capable of it as well. If there is substantial interest, I could certainly support an effort to bring others into this kind of testing, though my time is somewhat limited due to numerous responsibilities and other projects.

What support would you need to make it possible to perform this investigation yourself?

I do think that knowing the impact of phase distortion is important for audio consumers and producers, so that engineering efforts can be targeted at the right areas. Engineering is the optimization of compromises, and we have to know which ones we can get away with, and which ones we can't.

 

[GelbeMusik]

... it's very easy to test.

Not so easy. Did You check that the only parameter You changed was the phase / group delay / time alignment whatever You name what You are after? There were so and so many more or less focussed investigations and anecdotes which tell the contrary. So, better to double check.

... important for audio consumers and producers, so that engineering efforts can be targeted at the right areas.

You think of greening the sahara? Or get rid of nuclear waste? Avoid nano plastic in Your daily diet? Flee to mars, terraforming a planet? Maybe we better gonna design for marsian atmosphere. Audio is so important for wealthy consumers.