• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Importance of impulse response

Hi ! thanks a lot for the very valuable reply Can you confirm about passive crossovers negative influence on the dynamic behaviour of a speakers ?
i am deciding to go or not with an active crossover speaker (and powered)
There are a lot of reasons besides a pretty impulse response to go with active crossover and multiple amps. https://sound-au.com/bi-amp.htm
 
There are a lot of reasons besides a pretty impulse response to go with active crossover and multiple amps. https://sound-au.com/bi-amp.htm
Funnily enough many active crossovers have no ability to improve the impulse response over a passive one. If the crossovers can be made linear phase, FIR filters can be used or a time reversed all pass can be implemented then a real difference can be seen, that can also be done by use of a convolution filter at the source. The group delay can be undone afterwards.

As to audibility this masters thesis conducted an interesting experiment where removing the group delay distortion made the sound worse. I have also found this myself that the ability to arbitrarily alter phase also gives you the ability to destroy the sound of the speaker :)

https://www.politesi.polimi.it/bitstream/10589/186927/6/Tesi_Schgor.pdf
 
Funnily enough ... removing the group delay distortion made the sound worse. I have also found this myself that the ability to arbitrarily alter phase also gives you the ability to destroy the sound of the speaker :)

https://www.politesi.polimi.it/bitstream/10589/186927/6/Tesi_Schgor.pdf
And again, it seems You didn't read the full article. It is a master's thesis, hence is writen by a quite unexperienced engineer who is given some time to fool around. With the headphone--that's the part You are referring to from the summary/conclusion alone, why didn't he just use negative feedback on all, "phase" and amplitude together?! We don't see any--I mean any, validation of his chosen "phase correction" thing. Maybe his phase correction isn't working good, if at all. To feed a test-signal in would have sufficed. But, nope. Hmmm ... ? Not the best thesis ever, but at least he displays allpass filters as transistor circuits.

Please read and understand the articles You shell out, and point to the argument by reference. Otherwise people lose time in reading unrelated stuff.

Tja, what else to say about the audiophoolery? What about the "phase", "group delay" etc, if we stick a microphone into a grand piano with necessarily differing distance, hence "phase" to the strings, or part of the strings?

All that fuzz only to *not* listening to music.
 
Your initial post spiked a lot of controversy.
Yeah...
I just had to ask that, because for a long while I thought that it was super important. Since, not only my favorite speakers that I have owned yet were designed with that in mind as I mentioned in my first post(I didnt buy them for impulse response, I learned about it later), but aIso the best to my ears that I have heard yet were praised in a review for having good impulse response, even though I almost never see it being mentioned in any review, so I was like "I guess I like good impulse response then". Well, maybe its a coincide.
1668360998596.png
 
By what does the so called 'time-reversed' help in discriminating a monaural phase thing? "... were used in order to ..." is not an explanaition, it is a statement. If You compare the usage here to the 'time-reverse' used in reference #9 You'll see that there is no relation whatsoever.

#9 => https://www.researchgate.net/public..._head-related_transfer-function_phase_spectra
KSTR has mentioned ...
The simulation demonstrates that "ringing" is not the result of a "slow" response, but is the response to a resonance. Ringing in the response occurs only when the peaking filter frequency is ...
Mixed up terms, why? The 'ringing' is not resonance. It is the result of the band-filtering. The filter doesn't have to be 'peaking'.
So companies that say stuff like this are basically lying?


seikaku.png
Its true, they are lying. A "bass sound" cannot look like the top line in the graph. If it looked like that it was a broadband sound, not bass in particular. The delay aspect in the bottom graph is chosen arbitrary, maybe looking for the max amplitude? The sluggishness is just mathematics. A subwoofer is intended to *not* reproduce higher registers, so it is band limited. Band limitation inevitably leads to a broadening of the 'pulse'. It is mathematically so, it has to be so, we *want* it to be so.

Once all speakers, namely sub and the others work together, the puls is restored. Because the other, also bandlimited, would show the same broadening, but with opposing phase, hence cancelling out the broadening of the sub. So far the theory. Now comes the wonder: it works! It just does so, despite all the audiophoolery of everlasting doubt if the stereo could possibly do any good.

Same as with cables, people don't accept things to be just working.

Here are measurements from the speaker the o/p refers to (or something similar, but cheaper, its the top line specimen):

Yeah...
I just had to ask that, because for a long while I thought that it was super important. Since, not only my favorite speakers that I have owned yet were designed with that in mind as I mentioned in my first post(I didnt buy them for impulse response, I learned about it later), but aIso the best to my ears that I have heard yet were praised in a review for having good impulse response, even though I almost never see it being mentioned in any review, so I was like "I guess I like good impulse response then". Well, maybe its a coincide.
View attachment 243176
I cannot comment on this. It is dead proof to say, that the finest differences in time response aren't the final and most subtle clue to make these great. I love to see, that You're satisfied with the speaker's specifics.
 
It is just the amplitude spectrum over subsequent periods of time. Vulgo, the frequency response, amplitude, not phase.
That is false statement. Also amplitude spectrum over subsequent periods of time will be distorted (compared to original source) with transient signals if group delay is frequency dependent. That is very basic signal theory. That also defines that frequency dependent excess group delay causes dynamic errors with non-constant signals such as music.
The sad fact that even educated reviewers print 'impulse' graphs, and to double up the 'step' response too, doesn't mean it makes any sense at all.
False.
I personally think it is some b/s nobody ever thought about. Because the overly prudent engineers of the software used brought that in to just show off. Shi* happens ... , forgiven, but it remains shi*.
Fact is that your output quite total nonsense.
 
That is false statement. Also amplitude spectrum over subsequent periods of time will be distorted (compared to original source) with transient signals if group delay is frequency dependent. That is very basic signal theory. That also defines that frequency dependent excess group delay causes dynamic errors with non-constant signals such as music.
Your quote of my post was misleadingly shortened. I said regarding the distorted 'signal' as opposed to maintained 'information' therein: "As I stated earlier, all the information is there as far it is relevant for the hearing, let alone the enjoyment of music, or even the information contained in reproduced human talk ... . It is just the amplitude spectrum over subsequent periods of time. Vulgo, the frequency response, amplitude, not phase."

What You don't get is the idea that the information for the human being, listening to an audio reproduction, is the meaning of the program. The intellectual content--in the sense of intelligence, namely to perceive something as real. What it says to the mind after concious processing. Humans are neither auditory machines nor just vertebrates. There is no stereo without a living soul to believe in it.

So I finally asked in this thread at least: what exactly is the 'information' missing with an easily maintained, regular phase/time distorted 'signal'? Please specify what You are chasing after, I mean, what 'information'. It would help to define the missing part in real terms on the 'information' level.

There You'll find a recording venue depicted; so many microphones ( https://linkwitzlab.com/Audio_production/sf_symphony.htm ) Which of them is the time-correct one? (LoL) Hint: 30cm (12") relative distance are equivalent to 1ms of a time delay.

Still, given a certain image of an 'impulse' the human eye cannot under any circumstances determine the frequency response, the phase response or the group delay. The picture has all the data in, but interwoven in a too complex way. That is why audiophools fall for it. Because only one discriminator is left to the naked eye, namely ideal or not. Not ideal, so I cannot trust my stereo, so I happily cannot accept the enjoyment of music. Buy another one ... .

Fact is that your output quite total nonsense.
A pointed statement, if You can prove it scientifically.
 
Last edited:
By what does the so called 'time-reversed' help in discriminating a monaural phase thing? "... were used in order to ..." is not an explanaition, it is a statement. If You compare the usage here to the 'time-reverse' used in reference #9 You'll see that there is no relation whatsoever.

#9 => https://www.researchgate.net/public..._head-related_transfer-function_phase_spectra
Yes. The authors were making a statement to state why they used time reversed impulse responses in their study. So? If you think that their statement is wrong, then convince us by giving us proof. They referenced paper #9 because the paper #9 authors used similar stimuli in their study, and therefore what the authors of this paper did wasn't without precedent. Isn't that a "relation"?

ref9.png


Mixed up terms, why? The 'ringing' is not resonance. It is the result of the band-filtering. The filter doesn't have to be 'peaking'.
So what then is resonance? And what do you mean by "band-filtering"? However, I'd give you half credit as your last statement is correct, even though I am quite certain that you didn't mean it that way. A dip in the FR will also cause ringing. And yes, both are resonances.

tone_burst.gif

Fact is that your output quite total nonsense.
A pointed statement, if You can prove it scientifically.
You have made numerous factually false and unintelligible statements, and that's plenty sufficient proof.
 
And again, it seems You didn't read the full article. It is a master's thesis, hence is writen by a quite unexperienced engineer who is given some time to fool around.
Yes it is which is why I described it as such in the post above. I tend to think of it as being written by someone with an interest and current study in the field unless after reading it I form the same opinion as you.
Please read and understand the articles You shell out, and point to the argument by reference. Otherwise people lose time in reading unrelated stuff.
Instead of making upfront assumptions I like to read it and make my own mind up how much validity to give to any part of a document. I have no intention of spoon feeding anyone and I am not making an argument, just pointing out some information I have found and saved that I think has relevance to the topic.

There is very little audio research I have come across that can not have holes poked in it, whether that was down to lack of time, resources or knowledge is not always clear.
Never the less I can usually find something useful to take out of it.
 
Still, given a certain image of an 'impulse' the human eye cannot under any circumstances determine the frequency response, the phase response or the group delay. The picture has all the data in, but interwoven in a too complex way.
A raw impulse is difficult to read because it is dominated by high frequencies, so a filtered or processed version like a step response is much easier to use in determining features of a speaker. Certainly some humans are capable of extracting useful information from them.

You can see from looking at a step response if the frequency response of the speaker is flat or if it slopes down, you can tell this from the initial step rise, you can tell if it has extended bass response from how long it takes for the step to fall below zero, you can tell if the speaker has group delay from the shape of the step, you can tell the polarity of the drivers and get some idea of the crossover topology.

All of this takes experience or comparison to known quantities. If all measurements are available then it is much easier to see nuance by looking at the individual graphs and no reading of tea leaves is necessary.
 
Last edited:
Thanks and all honor to the clarification. I read all three thoroughly. I took it as a training in reasoning as a starter for the weekend. Regularly it would be helpful to be a bit more specific with references.
 
Yes. The authors were making a statement to state why they used time reversed impulse responses in their study. So? If you think that their statement is wrong, then convince us by giving us proof. They referenced paper #9 because the paper #9 authors used similar stimuli in their study, and therefore what the authors of this paper did wasn't without precedent. Isn't that a "relation"?

View attachment 243213
O/k, that goes nowhere. U should read reference #9. I'm sure You won't need to come back to me.

So what then is resonance? ... A dip in the FR will also cause ringing. And yes, both are resonances.
Likewise. Bandpass filter has same differential equation as resonance, but not filter in general e/g lowpass.

I err'ed with following U'r interpretations. Sorry for coming back to U.

As I said, define what is missing with not so perfect 'pulse' on the 'information' as opposed to 'signal' level. Compare to how microphones are used and mixed in the production process. Do U really think this discussion goes up-hill? I feel that the o/p is fed up for a while.
 
Last edited:
Other studies have shown when you have the exact same speaker with the exact same crossover slopes etc, the linear phase version is mostly preferred over the common allpass phase version. The more crossover points you have at lower frequencies (<1kHz) the stronger the preference. Subwoofer XO at ~80Hz is especially bad for "speed" and "compactness" of bass transients.
Is 80hz crossover too high or too low?
 
Other studies have shown when you have the exact same speaker with the exact same crossover slopes etc, the linear phase version is mostly preferred over the common allpass phase version. The more crossover points you have at lower frequencies (<1kHz) the stronger the preference. Subwoofer XO at ~80Hz is especially bad for "speed" and "compactness" of bass transients.

Remember, "transient" does NOT mean a sharp and short pulse here, rather it means a short but tonal burst of just a few cycles of a sine wave with a smooth envelope, often called "blips". Those are even standardized, by now (CEA-2010 Burst, 6.5 cycles, Hann window envelope). Phase distortion now manifests itself in different arrival times of those bursts at different frequencies (the lower ones coming later, typically) even though they started at the same time (or had their "center of gravity" aligned) in the source signal. And that's what we can perceive in a direct comparison, some more than others...
What study shows preference for linear phase speakers as opposed to allpass? I would like to read this, I didn't think people were doing listening tests of such a thing.

Normally I'd assume that creating a speaker with two different filters having the same directivity/magnitude but different phase characteristics would be impossible but I guess it's possible now.
 
I said regarding the distorted 'signal' as opposed to maintained 'information' therein: "As I stated earlier, all the information is there as far it is relevant for the hearing, let alone the enjoyment of music, or even the information contained in reproduced human talk ... . It is just the amplitude spectrum over subsequent periods of time. Vulgo, the frequency response, amplitude, not phase."
Is this written drunk or just overly complicated English?
What You don't get is the idea that the information for the human being, listening to an audio reproduction, is the meaning of the program. The intellectual content--in the sense of intelligence, namely to perceive something as real. What it says to the mind after concious processing. Humans are neither auditory machines nor just vertebrates. There is no stereo without a living soul to believe in it.
Is this written drunk or just overly complicated English?
So I finally asked in this thread at least: what exactly is the 'information' missing with an easily maintained, regular phase/time distorted 'signal'? Please specify what You are chasing after, I mean, what 'information'. It would help to define the missing part in real terms on the 'information' level.
Is this written drunk or just overly complicated English?
This forum might be targeted also for others that native English so please try to avoid idioms, complex sentence structures and rare abbreviations if you hope that others understand what you really try to tell or ask.
Still, given a certain image of an 'impulse' the human eye cannot under any circumstances determine the frequency response, the phase response or the group delay.
The picture has all the data in, but interwoven in a too complex way. That is why audiophools fall for it. Because only one discriminator is left to the naked eye, namely ideal or not. Not ideal, so I cannot trust my stereo, so I happily cannot accept the enjoyment of music. Buy another one ... .
Sound is not amplitude responses or phase responses flying in the air. Those responses are not reality and accurately calculable with continuously changing signals. Just results of mathematical analyses within certain (limited) time period and dependent on window function and other FFT parameters. Unambiguous result can be expected with long static signal (steady state).
Sound is much closer to pressure as a function of time. Badly imperfect time domain is difficult...impossible to convert to frequency domain without FFT tool, but that is not the point/purpose. Information is in time domain, and it's not our problem if you don't understand for example ETC which is visual and easy for human being because "illogical" jumps due to phase are eliminated. So don't blame and call others audiophools. They could understand time domain and have found correlation between perceived sound and p(t) curves, though GD(f) is easier measure of timing quality especially at LF...MF.

Reliability of studies about audibility/perceptibility of frequency dependent excess or normal group delay is part of this discussion. Frequency dependent GD is perceivable and also audible (though far from easy) and obvious imperfection in sound reproduction which makes all this quite simple at least for all who are able perceive differences and problems due to poor timing. Own priorities and preferences are taken into account and opponents will be ignored due to repeated fake news and selective generalizations based on faith in poor science.
 
Is 80hz crossover too high or too low?
Well, actual frequency does not matter much, it's more about the additional crossover point increasing the number of ways by one.

Assume you had a main speaker which is good down to 40Hz, say a 3-way speaker.

When you properly integrate a subwoofer you now have a 4-way system, splitting the range below the midrange driver in two sections, one above the subwoofer XO and one below it.

This split generates an increased group delay at and below the subwoofer XO frequency and that the offending property, assuming a classical split function like Linkwitz-Riley 4th order acoustic or the like is being used.

And 80Hz (THX standard frequency) is pretty much at or even above the low-frequency core event of kick drums or the attack of plucked upright bass etc, now coming in later than the rest.

Hope it is clearer now what I meant to say.
 
Is this written drunk or just overly complicated English?

Is this written drunk or just overly complicated English?

Is this written drunk or just overly complicated English?
This forum might be targeted also for others that native English so please try to avoid idioms, complex sentence structures and rare abbreviations if you hope that others understand what you really try to tell or ask.


Sound is not amplitude responses or phase responses flying in the air. Those responses are not reality and accurately calculable with continuously changing signals. Just results of mathematical analyses within certain (limited) time period and dependent on window function and other FFT parameters. Unambiguous result can be expected with long static signal (steady state).
Sound is much closer to pressure as a function of time. Badly imperfect time domain is difficult...impossible to convert to frequency domain without FFT tool, but that is not the point/purpose. Information is in time domain, and it's not our problem if you don't understand for example ETC which is visual and easy for human being because "illogical" jumps due to phase are eliminated. So don't blame and call others audiophools. They could understand time domain and have found correlation between perceived sound and p(t) curves, though GD(f) is easier measure of timing quality especially at LF...MF.

Reliability of studies about audibility/perceptibility of frequency dependent excess or normal group delay is part of this discussion. Frequency dependent GD is perceivable and also audible (though far from easy) and obvious imperfection in sound reproduction which makes all this quite simple at least for all who are able perceive differences and problems due to poor timing. Own priorities and preferences are taken into account and opponents will be ignored due to repeated fake news and selective generalizations based on faith in poor science.
Kimmo, do you have a link to a study which shows the audibility and preferences related to frequency variable group delay?
 
What study shows preference for linear phase speakers as opposed to allpass? I would like to read this, I didn't think people were doing listening tests of such a thing.

Normally I'd assume that creating a speaker with two different filters having the same directivity/magnitude but different phase characteristics would be impossible but I guess it's possible now.
I seem to recall that during the development (and later) of the now legendary Klein+Hummel O500C 3-way DSP monitor blind tests were conducted to asses the properties of the three switchable modes of phase linearisation (none = minphase filters, linphase tweeter-to-mid only, full = linphase tweeter-to-mid and linphase mid-to-woofer), but can't dig up a reference atm.

Phase linearisation of any given speaker is well outlined, conceptually, these days:
You just need to obtain, with a DRC software package, the acoustic allpass phase (excess phase) of the speaker, preferably converted to clean analytical form (with the help of curve-fitters, when required).
The time inverse of the impuse response of this allpass then is directly the phase-compensation convolution kernel for a convolver (FIR "filter"), placed on the overall input signal.
Practical examples of this approach are Grimm LS1, for example.

We now can A/B this when switching between the phase-correction kernel and dummy kernel with a "1" at the at final peak sample position in the phase-compensator.
On top of that, we can introduce regular allpass kernels (not time-inversed) to the phase-correction speaker and by this we can asses the effect of arbitrary excess phase / group delay in full isolation, comparable to headphones but with the special properties of real speaker setups.

With a good non linear-phase speaker as the foundation of such experiments, say a Neumann or Genelec, I'm really quite convinced almost anybody will a) be able to detect the (admittedly usually subtle) differences to linear-phase and b) develop a preference for the linear-phase version. It takes some training and obviously some passion for the topic in general, with a lot of evenings quickly adding up...
 
I seem to recall that during the development (and later) of the now legendary Klein+Hummel O500C 3-way DSP monitor blind tests were conducted to asses the properties of the three switchable modes of phase linearisation (none = minphase filters, linphase tweeter-to-mid only, full = linphase tweeter-to-mid and linphase mid-to-woofer), but can't dig up a reference atm.

Phase linearisation of any given speaker is well outlined, conceptually, these days:
You just need to obtain, with a DRC software package, the acoustic allpass phase (excess phase) of the speaker, preferably converted to clean analytical form (with the help of curve-fitters, when required).
The time inverse of the impuse response of this allpass then is directly the phase-compensation convolution kernel for a convolver (FIR "filter"), placed on the overall input signal.
Practical examples of this approach are Grimm LS1, for example.

We now can A/B this when switching between the phase-correction kernel and dummy kernel with a "1" at the at final peak sample position in the phase-compensator.
On top of that, we can introduce regular allpass kernels (not time-inversed) to the phase-correction speaker and by this we can asses the effect of arbitrary excess phase / group delay in full isolation, comparable to headphones but with the special properties of real speaker setups.

With a good non linear-phase speaker as the foundation of such experiments, say a Neumann or Genelec, I'm really quite convinced almost anybody will a) be able to detect the (admittedly usually subtle) differences to linear-phase and b) develop a preference for the linear-phase version. It takes some training and obviously some passion for the topic in general, with a lot of evenings quickly adding up...

Well that's just the thing. I'd like to see data on the detectability. My experience is that most acoustic phenomena that are 'subtle but detectable' are often actually imaginary. I'm not suggesting that this is the case with the different phase linearization modes, but I'd like to see some kind of blind testing. KH is not the only manufacturer doing this - Grimm is one you mentioned but I think that Neumann is doing it in their newer models, an perhaps someone else if I remember right - maybe Adam or HEDD.

Having said that, I see a speaker like the KH as a piece of metrological equipment. Just because you can't see the difference between two parts that vary by a micron doesn't mean there is some value in a micrometer that can. Might as well make the speaker as accurate as possible.
 
Is this written drunk or just overly complicated English?
Is this written drunk or just overly complicated English?
Is this written drunk or just overly complicated English?
This forum might be targeted also for others that native English ...

Sound is not ... ... obvious imperfection in sound reproduction ... quite simple at least for all who are able perceive differences and problems due to poor timing. ... opponents will be ignored due to repeated fake news and selective generalizations based on faith in poor science.
Thank You. Obviously I provoked You. Sorry.
 
Back
Top Bottom