As "so typical"? 🙄
You may have *intended* something different here, but the context clearly references phase-induced timing errors. 😉
No, that quote clearly acknowledges "group delay" errors as being potentially audible, but just as clearly rules out "phase errors" as being important, which is exactly what I said it said and completely different than what you said I said. If you don't see that then I am not willing to explain it to you.
"If you don't see that then I am not willing to explain it to you."
Now that *is* "typical" - from you. 😉 (..and unlike you, I can provide references that it is in fact "typical" behavior from you.)
I *never* stated or implied that you didn't acknowledge group delay errors at higher freq.s as being inaudible. This is what I stated:
"..Earl suggested that phase-induced timing errors at higher freq.s require a great deal of error to be at all audible (in that the amount of phase rotation at these higher freq.s results in a comparatively low time delay).."
In particular:
"..a great deal of error to be at all audible.."
This statement certainly doesn't preclude audibility.
Additionally, note the use of the word "suggested", which again - may not have anything to do with your intent. A suggestion, an implication, or a manner in which your communication could be interpreted, and from my reading - is *likely* to be interpreted. Further, I never *quoted* you, so where is the "so typical.. misquote"?
With respect to your previous post:
"..clearly rules out "phase errors" as being important..".
- No it doesn't.
1. You have limited the freq.s (..to higher freq.s)
2. You have offered group delay as a *substitute* for phase at these higher freq.s.
#1 certainly doesn't rule out the issue of phase.
#2 doesn't rule out the issue of phase either, it just changes the perspective and nomenclature.
Moreover the point I was making before from the Physics Today article is that:
Phase AT HIGHER FREQUENCIES can have an effect at lower freq.s with respect to timing and harmonics that extend below the typical 1 kHz freq. threshold for timing errors. And this is just with respect to localization, so even the limitation of higher freq.s does NOT "clearly rule(s) out "phase errors" as being important".
Last edited:
And speaking of.. Thuneau (a member here), has this to offer:
DSP Loudspeaker Phase Arbitrator. Transient Perfect loudspeaker processor
there is a software thingy that fixes phase i don't remember the name
but yeah you would need something like it
FIR XO alone won't do it because it will only perfectly sum signals that were perfect to begin with but since each individual driver has phase roll it won't work
you have to take each individual driver and equalize it flat over intended range, then use the software thingy to fix the phase of it and only after that use FIR XO to sum multiple drivers
but yeah it should be doable. you might need to build the speaker into the wall though to avoid diffraction.
but yeah you would need something like it
FIR XO alone won't do it because it will only perfectly sum signals that were perfect to begin with but since each individual driver has phase roll it won't work
you have to take each individual driver and equalize it flat over intended range, then use the software thingy to fix the phase of it and only after that use FIR XO to sum multiple drivers
but yeah it should be doable. you might need to build the speaker into the wall though to avoid diffraction.
Not at all. FIR filters can do just about anything you like and can treat amplitude and phase independently at the same time. There seems to be a great deal of confusion about FIR filters. I have seen this over the years. Many people mistakenly interchange FIR filters for linear phase filters. However, for example, if you want an LR 4 acoustic amplitude response with linear phase an FIR filter can equalize a driver amplitude to the LR4 amplitude and correct the phase to linear with a single filter kernel. In fact, it is rather trivial. You just divide the transfer function of the driver by the linear phase target and you have the transfer function for the FIR filter. This can then be used for convolution in the frequency domain.
Or you can start with the transfer function of a completed speaker and easily design an FIR filter which linearizes the system phase w/o altering the amplitude. Again, it is fairly trivial.
i know John. i guess i didn't express myself clearly.
the software thingy i referred to is in itself FIR.
ability to process phase and amplitude independently is the power of FIR.
my point was that to get square wave you will need to fix up the phase for individual drivers first and then PRESERVE it in the crossover. so you need to use FIR twice. the first time to change phase only and a second time in what we call a crossover. of course i suppose you can combine the steps.
when i said FIR XO alone can't do it err i guess its hard to explain what i meant - but not what you took it to mean.
when i say crossover to me that doesn't include equalization. to me equalization is a separate thing. but to you equalization is probably a function of a crossover too. it is really just a matter of semantics.
so when i say you have to equalize it first then pass it to crossover to you it doesn't make sense but we're talking about the same thing really.
my point is speakers introduce phase error twice - first each individual driver has its own phase error and then crossover adds additional error to that.
both of these will need to be dealt with if u want a square wave to come out. in my nomenclature so to speak FIR XO refers to the solution to just the second part of this problem - removing the error introduced by the crossover by removing an analog crossover.
but then u still need to remove the error introduced by individual speaker drivers. this takes FIR technology too but i personally don't consider this a crossover function. that's all.
by the way i think we need to be very careful when we say "this is not audible" or "that is not audible"
we are not really smart enough to make such conclusions. just like those scientists who declared that margarine is good for you - sometimes the problem may appear simple but in fact be much more complex and you can be certain of the answer and be wrong at the same time.
point is i no longer believe claims like "phase is not audible" or "bass cannot be localized" etc. this all belongs in the same category as "margarine is good for you, cholesterol is bad". simple minded humans trying to draw even simpler conclusions about very complex subjects.
i think this is where we can apply precautionary principle. given that we aren't smart enough to know anything let's not go out of our way to screw up things more than we need to. let's not mix down bass to one sub next to your right ear and let's not create incredible amounts of phase distortion simply because somebody thought that he has a brain and declared that it's not audible.
we are not really smart enough to make such conclusions. just like those scientists who declared that margarine is good for you - sometimes the problem may appear simple but in fact be much more complex and you can be certain of the answer and be wrong at the same time.
point is i no longer believe claims like "phase is not audible" or "bass cannot be localized" etc. this all belongs in the same category as "margarine is good for you, cholesterol is bad". simple minded humans trying to draw even simpler conclusions about very complex subjects.
i think this is where we can apply precautionary principle. given that we aren't smart enough to know anything let's not go out of our way to screw up things more than we need to. let's not mix down bass to one sub next to your right ear and let's not create incredible amounts of phase distortion simply because somebody thought that he has a brain and declared that it's not audible.
Certainly phase errors are compounded by crossover and drivers but the system has some given phase response (at some reference point) that can be corrected with a single inverse FIR filter. The whole is made up of the sum of the parts. It is only the sum that needs to be corrected, not the individual parts. You need not worry about how the sum is obtained.
I don't think most people mean the terms in the scientific sense of proven to 99% confidence. When I say that I usually mean that "for all practical purposes" phase is not a significant criteria in audibility. There is probably some situtation where it is audible, I have no doubt, but to obsess over it while leaving significant things uncontrolled is poor design. A big pet pieve of mine is the audiofool tendency to raise insignificant issues to the realm of the all important. Interconnect discussion anyone?
To me, phase and group delay are just two different ways of plotting the same data.
As to audibility, anyone who uses music ripped to a PC can download the demo version of phase arbitrator and give it a try. If you decide it's worth the hassle, a license is $90.
DSP Loudspeaker Phase Arbitrator. Transient Perfect loudspeaker processor
Also, recent Pioneer receivers, at least the more expensive ones, include a switchable "full band phase control" which is supposed to do something similar although I've never seen any independent measurements of how well it works.
what you say is true in theory.
however such solution probably won't work too well off-axis.
Nietzsche explained how it works. Everything we do is moral, right and good. Everything people on the other side do is i immoral, wrong and evil.
Everybody has the need to feel good about themselves. Since on average every person is no better than other people the only way to feel good about yourself is to apply the strategy described by Nietzsche.
Furthermore you get maximum *leverage* when you split off as few people to your own side and as many as possible to the other. You have to make a case that you are a unique snowflake in other words.
For example - 90% of people like clean, tight, strong, punchy bass. Perhaps 0.01% of people will claim that they like the sound of silver wire. You want to be with the silver wire people if you *really* want to feel good about yourself. Therefore you will proclaim that bass is irrelevant while only silver wire is listenable.
Thats also how you get people who will say things like " i hate Pink Floyd " or " i hate Martin Logan ". They're going for the maximum uniqueness award.
These are also the same people who will go to a conspiracy theory forum to tell everybody that Federal Reserve is in fact run by a bunch of saints.
Last edited:
Well I tested my preferred monitors tonight with a phase response offset impulse response which approximately linearized them. I ABXed it successfully so I guess I heard a real difference. But the thing is imo the difference was negligible and not easy to hear.
I found a recording which I could expose the slight deficiency so that I could more easily ABX between the a non-linearized and a linearized example. The main thing I honed in on was a lack of attack on a sound which had a slow sweeping attack near the crossover region. The difference sounding like one example swelling with no attack and the other attacking and then swelling. Another thing that sticks out to me is that as I listened to the test track it gradually became harder for me to differentiate between the two - this also happened the last time I tried this experiment but much faster and to the point where I couldn't differentiate at all.
Approximate FR and PR
Approximate Offset
Approximate Convolved Speaker output
Most of those FR and PR irregularities I am thinking are due to the mic I am using (BLUE Baby Bottle) which is not a measurement mic and was just used to give me an idea of what the crossover filters are doing to the PR. So basically the best I can hope for is an offset at the crossover frequency (1.8kHz) and everything else is approximate.
I found a recording which I could expose the slight deficiency so that I could more easily ABX between the a non-linearized and a linearized example. The main thing I honed in on was a lack of attack on a sound which had a slow sweeping attack near the crossover region. The difference sounding like one example swelling with no attack and the other attacking and then swelling. Another thing that sticks out to me is that as I listened to the test track it gradually became harder for me to differentiate between the two - this also happened the last time I tried this experiment but much faster and to the point where I couldn't differentiate at all.
Approximate FR and PR
An externally hosted image should be here but it was not working when we last tested it.
Approximate Offset
An externally hosted image should be here but it was not working when we last tested it.
Approximate Convolved Speaker output
An externally hosted image should be here but it was not working when we last tested it.
Most of those FR and PR irregularities I am thinking are due to the mic I am using (BLUE Baby Bottle) which is not a measurement mic and was just used to give me an idea of what the crossover filters are doing to the PR. So basically the best I can hope for is an offset at the crossover frequency (1.8kHz) and everything else is approximate.
Last edited:
😀 ..and the link I provided originally.
And yes, only at it's best from one point in space.. (which can be said for just about any system.) But on the other hand many tweeter systems extend down low enough, and are usually spaced closely to the next pass-band's driver so that it should NOT substantially crap-out the performance off-axis. Of course if the drivers are concentric at mid and higher freq.s, better still (..at lest in this limited respect).
I think Dennis has it correct.
Play with it, see if it does something for you or not - at least from the perspective of a hobby.
For those more concerned with cranking-out research, set-up a suitable double blind test with it with proper "controls".
Either way, it can potentially help resolve the issue for each interested party.
It is true in theory and in practice. What is also true in theory and practice is that any phase linearizion scheme is only correct at a single observation point. Moving off axis means the system phase changes so the correction would also have to change. And, in fact, building a loudspeaker with a TP crossover only yield TP response at the desing point as well. A loudspeaker's phase response varies with listeing position 3 dimensionally.
Last edited:
i know. but don't you think a system that equalized phase in 1 shot as you suggest would have greater error off-axis than a system that linearized phase of drivers individually before combining them and equalizing phase again ?
i mean lets have a little logical experiment. lets say each driver splits impulse response into two impulses ( bad driver ). and then crossover combines them to produce 4 impulses ( 2 + 2 ). you equalize all that so that there is just 1 impulse instead of 4. by taking off 4 impulses at the combined output you add 4 impulses to each individual speaker so now each speaker has 6 impulses but they combine to just 1. now you move off axis and you have 12 impulses ...
versus my approach. each driver is equalized to a single impulse. the comibned output is equaized again to single impulse. you move off axis and you have 2 impulses.
isn't 2 closer to 1 than 12 ?
am i thinking about this wrong ?
Last edited:
- Status
- Not open for further replies.