Martin Colloms makes such statements in High Performance Loudspeakers, 5th addition, regarding "recent psycho-acoustic research" though he does not provide direct references. Sorry I can not provide more detail.
I'm not so concerned with the argument over whether it is audible or not. I'm in the camp that if it can be eliminated without introducing other deleterious effects, then it should be.
I have no argument with that, it's just that it never happens that way. Even if its just cost, everything has implications. If you can't "scale" the audiblity, how do you scale the "cost/benifit"?
I spent a little time looping the above shown waveforms in Audacity switching between them. I could not detect any significant difference. At some times I though I did but after a few switches they sounded the same.
When listening to some songs I heard more bass without but that is understandable as there is a small dip at 40Hz due to non perfect IR.
If my experiments are unrelated to square waves and "time alignment" then I would like to know.
But for now I believe a phase shift is not that important when it comes to audio perception. At lest not a smooth shift as in my tests.
If I could I would like to try the same thing where body vibrations comes in. Where you feel the kick as well as hear it. Might end up being a different sensation if the kick is shaped differently.
Any comments?
When listening to some songs I heard more bass without but that is understandable as there is a small dip at 40Hz due to non perfect IR.
If my experiments are unrelated to square waves and "time alignment" then I would like to know.
But for now I believe a phase shift is not that important when it comes to audio perception. At lest not a smooth shift as in my tests.
If I could I would like to try the same thing where body vibrations comes in. Where you feel the kick as well as hear it. Might end up being a different sensation if the kick is shaped differently.
Any comments?
I am not sure that I understand what type of phase shift you introduced. But for the purposes of this discussion, a simple experiment might be to shift the phase (edit: not quite correct; see farther below) of the bass, mid, and treble ranges by different amounts and then compare the original and shifted versions of sounds with fast attacks, such as a plucked string, a bass drum kick, a sudden piano key or chord strike, et al.
It might be interesting to try to determine the minimum phase difference between one or more of the ranges that was audible, for some common test sound(s). It might also be interesting to plot a sound's amplitude in the time domain and compare the original and shifted versions, and maybe subtract the original from each shifted version to see the delta waveform. The frequency domain amplitude plot comparisons might be quite interesting, as well.
Actually, I think that the above described doing something more like a simple time delay for each range, rather than true phase shifting, since a phase shift would imply a different delay at each frequency. The simpler time-alignment comparisons would be interesting too. But it wouldn't quite be what this thread is about.
So, ideally maybe, or eventually, for that type of experiment, you might want to be able to start with a plot or table of phase versus frequency, to control the processing, instead of just using a fixed time shift for each range.
Or maybe create a potentially-useful tool and just imagine simulating three speakers and go ahead and start with gain vs frequency and phase vs frequency plots for each of the three, plus an adjustable time delay for each one, to also simulate non-co-planar mountings (time-alignment variability).
I guess for that you would also need to make filters to separate the original sound into the bass, mid, and treble ranges, and then sum them back together after processing. But, to make it simpler, you could also just pretend it's one full-range speaker, and use only one plot or table of phase vs frequency to control the processing.
I might not have some of the details right, but you probably get the idea.
Last edited:
I have seen and heard a demo by the Swiss AES section where an almost (i.e. somewhere between transient-perfect and time-aligned) transient perfect crossover was compared to an LR4. The crossover frequency was 80 Hz. This difference was not much but it was clearly audible.
It could very well be that a crossover with the same amount of total phase-shift but a smaller rate of change of phase-shift (i.e. less group-delay and non peaking group delay) couldn't be distinguished from a transient-perfect one. A crossover that behaves like that would be an in-phase crossover with a Q of less than 0.7. The (small) disadvantage of such a crossover would be the asymmetrical slopes. This familiy of crossovers (the LR is just the one member of it with symmetrical slopes) offers good lobing behaviour.
The main problem IMO is that we are still lacking valid figures about how much group-delay is audible and at which frequency range. The Blauert and Laws paper can't be used for that purpose IMO since their tests used a different form of group-delay distortion than what is caused by a usual crossover.
A soon as time permits I will make a comparison of a low-Q in-phase crossover and a "transient-perfect" one with my two-way Manger system. But this won't be a statistically and scientifically valid test of course.
Regards
Charles
It could very well be that a crossover with the same amount of total phase-shift but a smaller rate of change of phase-shift (i.e. less group-delay and non peaking group delay) couldn't be distinguished from a transient-perfect one. A crossover that behaves like that would be an in-phase crossover with a Q of less than 0.7. The (small) disadvantage of such a crossover would be the asymmetrical slopes. This familiy of crossovers (the LR is just the one member of it with symmetrical slopes) offers good lobing behaviour.
The main problem IMO is that we are still lacking valid figures about how much group-delay is audible and at which frequency range. The Blauert and Laws paper can't be used for that purpose IMO since their tests used a different form of group-delay distortion than what is caused by a usual crossover.
A soon as time permits I will make a comparison of a low-Q in-phase crossover and a "transient-perfect" one with my two-way Manger system. But this won't be a statistically and scientifically valid test of course.
Regards
Charles
Hello,
About audibility of group delay distortion below 200Hz I can give the results of comparison we have done between the quality of the bass reproduced by a Onken W bass reflex (loading 2 Altecs 416 by side) and by a huge bass horn like this one:
http://www.diyaudio.com/forums/atta...427-jean-michel-lecleach-horns-floor-horn.jpg
The superior quality of the bass horn was very neat and nobody found the bassreflex superior to the bass horn.
A part of the audible difference was probably due to a lesser distortion of the bass-horn but a majority of listeners felt the sound of instruments through the bass horn more natural (acoustic bass was amazing!). After discussion a consensus appear to conclude that the main difference was due to a lesser group delay distortion obtained with the bass horn.
(we even could listen to the bass horn full range and it was really precise, coherent ...)
Best regards from Paris, France.
Jean-Michel Le Cléac'h
About audibility of group delay distortion below 200Hz I can give the results of comparison we have done between the quality of the bass reproduced by a Onken W bass reflex (loading 2 Altecs 416 by side) and by a huge bass horn like this one:
http://www.diyaudio.com/forums/atta...427-jean-michel-lecleach-horns-floor-horn.jpg
The superior quality of the bass horn was very neat and nobody found the bassreflex superior to the bass horn.
A part of the audible difference was probably due to a lesser distortion of the bass-horn but a majority of listeners felt the sound of instruments through the bass horn more natural (acoustic bass was amazing!). After discussion a consensus appear to conclude that the main difference was due to a lesser group delay distortion obtained with the bass horn.
(we even could listen to the bass horn full range and it was really precise, coherent ...)
Best regards from Paris, France.
Jean-Michel Le Cléac'h
Salut Jean-Michel
This is a very nice integration of a bass horn !
I guess you used some digital crossover with delay - didn't you ? Otherwise the group delay distortion would have been excessive - and any audible improvement would have definitely been caused by something else than "superiour" group delay properties.
Regards
Charles
This is a very nice integration of a bass horn !
I guess you used some digital crossover with delay - didn't you ? Otherwise the group delay distortion would have been excessive - and any audible improvement would have definitely been caused by something else than "superiour" group delay properties.
Regards
Charles
I though that was pretty clear by the phase plot I posted.
The samples I showed was a kick drum and a snare (I think).
Hello,
You are right the time alignment of the different loudspeakers was based on the use of a digital delay. But my message was not related to phase distortion or group delay distortion caused by the crossover.
My message has for goal to give an example of the (probable) audibility of the group delay of the acoustic load itself at low frequency.
The group delay due to the bass reflex in the frequency interval from 20Hz to 200Hz is less constant than the one caused by the large basshorn. I think the excellent pulse reponse of the bass horn and its negligeible group delay distortion from 20Hz to 200Hz is the main factor why the basshorn was preferred to the bass reflex.
Best regards from Paris, France
Jean-Michel Le Cléac'h
You are right the time alignment of the different loudspeakers was based on the use of a digital delay. But my message was not related to phase distortion or group delay distortion caused by the crossover.
My message has for goal to give an example of the (probable) audibility of the group delay of the acoustic load itself at low frequency.
The group delay due to the bass reflex in the frequency interval from 20Hz to 200Hz is less constant than the one caused by the large basshorn. I think the excellent pulse reponse of the bass horn and its negligeible group delay distortion from 20Hz to 200Hz is the main factor why the basshorn was preferred to the bass reflex.
Best regards from Paris, France
Jean-Michel Le Cléac'h
The problem with most of these "tests" is that more than one variable changes. What is necessary is to have the ability to correct the phase response without changeing anythng else, polar response, frequency response, etc.
If it comes to the testing of the audibility of phase-distortion alone I can definitely agree with you.
OTOH crossovers like LR etc offer better polar response than a transient-perfect constant-voltage crossover (specially when non-symmetrical driver arrangements are used).
So the comparison of an in-phase crossover with a transient-perfect constant-voltage crossover is also a valid one for instance. In this case one would test if it is worth trading good polar response against better transient response.
BTW: This thread is actually not about audibility of phase-distortion but how to achieve low phase distortion.
Regards
Charles
OTOH crossovers like LR etc offer better polar response than a transient-perfect constant-voltage crossover (specially when non-symmetrical driver arrangements are used).
So the comparison of an in-phase crossover with a transient-perfect constant-voltage crossover is also a valid one for instance. In this case one would test if it is worth trading good polar response against better transient response.
BTW: This thread is actually not about audibility of phase-distortion but how to achieve low phase distortion.
Regards
Charles
I think it is always important to know how much the "bottom line" is going to be affected by any particular variable. For audio, what is important is what gets to your brain and is interpreted as music there. If there are things going on in the recording and reproduction chain that don't really matter to the brain and do not effect the perception of sound, then they are not important and we should not worry about them.
Therefore, any discussion of how to achieve low phase distortion should really come along with an explanation or mentioning of whether a listener can actually hear phase distortion and to what degree that affects the listening experience compared to other issues pertinent to accurate sound reproduction.There is no point in solving problems that have no impact on the listening experience, eh? 🙂
By the way, thanks to those who have posted comments about audibility of phase distortion. I have seen many of these, but would love to see a more scientific double-blind type of study. Any out there (e.g. published, etc.)?
-Charlie
hello,
There is, imo, a way to make +/- evident the audibility of phase relations between
the harmonics of a simple synthetised sound with Audio Toolbox OSX.
Mac Audio Toolbox Download
This soft allows to tune initial phase of each harmonic in up to twelve sinus
thus hear the tone changes with a good headphone.
my hearing tests:
phase et harmoniques - Le blog de jimbee
best regards from Paris
There is, imo, a way to make +/- evident the audibility of phase relations between
the harmonics of a simple synthetised sound with Audio Toolbox OSX.
Mac Audio Toolbox Download
This soft allows to tune initial phase of each harmonic in up to twelve sinus
thus hear the tone changes with a good headphone.
my hearing tests:
phase et harmoniques - Le blog de jimbee
best regards from Paris
This has been called a Fourier synthesizer and is exactly the sort of thing used for blind testing by me in 1977. At that time the test group was music students. 80% could hear a 75 degree change in the second harmonic of the 9 harmonic wave used. All could hear a 180 degree change in the second harmonic. I know this thread is not about what can be heard but thought many would find this interesting.
hello,
For my personal system i use a substractive / delay filter based on a DXC2496
modified to achieve low phase distortion. (Special thanks to J. Kreskovsky.)
DCX2496 soustractif / delay - Le blog de jimbee
For my personal system i use a substractive / delay filter based on a DXC2496
modified to achieve low phase distortion. (Special thanks to J. Kreskovsky.)
DCX2496 soustractif / delay - Le blog de jimbee
- Status
- Not open for further replies.