There is probably an easier way to think about this. But the result is still the same and it is doubtful that a decomposition will help.
Simple case is that the transducer in a cabinet has a frequency response. A short time later the sound bounces off a wall and may add back to the ongoing signal. This is a "delay-and-add filter" and produces comb filtering. But this occurs at some time after the signals onset.
Where you are heading (I think) is to try and perform some filtering for the ongoing or more steady state portion (which includes the delay-and-add filtering that produces the comb filtering).
Here is the headache. Trying to "undo" the comb filtering will only be appropriate for the steady state signal (or the long term average - feel free to pick your own integration window). Problem is that the filter is not appropriate for the "onset of the signal (again pick your own integration window).
What would be required is a filter that changes its response over time. Is this where you are headed? And yes I know that I have oversimplified the example with a single reflection etc....
Simple case is that the transducer in a cabinet has a frequency response. A short time later the sound bounces off a wall and may add back to the ongoing signal. This is a "delay-and-add filter" and produces comb filtering. But this occurs at some time after the signals onset.
Where you are heading (I think) is to try and perform some filtering for the ongoing or more steady state portion (which includes the delay-and-add filtering that produces the comb filtering).
Here is the headache. Trying to "undo" the comb filtering will only be appropriate for the steady state signal (or the long term average - feel free to pick your own integration window). Problem is that the filter is not appropriate for the "onset of the signal (again pick your own integration window).
What would be required is a filter that changes its response over time. Is this where you are headed? And yes I know that I have oversimplified the example with a single reflection etc....
gedlee said:
Well, the way I am looking at it is the Green's function represents the source to listener transfer function. In free space this is just a time delay as far as phase is concerned so in free space the phase at the listener is MP + delay if the source is MP. In a room we have the wall boundary conditions entering the picture. Above the modal region it is possible to have minimum phase depending of the relative strengths of the direct and reflected sound. Cepstral processing can tell you what is what. At lower frequency I'm just not sure either way. I have cases where I am sure that it is, but others where it isn't clear.
The reason I ask is this: The frequency response at a listening position is the sum of the response from the excitation of each mode. Each mode is a single resonance and as such will have a minimum phase characteristics and a unique impulse response. However, the net frequency response is the sum of all these resonances modes and the net impulse is the sum of all the impulses from the individual resonance modes. Now, if the response at the listening position is equalized flat (no guarantee it will be flat anywhere else
) then if it is minimum phase (plus a delay), regardless of the source type, or what modes contribute to the frequency response at that listening point, the transient response must be the same. On the other hand, if it is not MP, then while the amplitude is flat regardless of source type, the phase could be different since with different sources different modes could contribute differently to the unequalized response. This would lead to different transient response with the result being the different sources types would sound different even is the response was equalized flat at the observation point.
john k... said:
I am not sure that this follows. Just because the free field Green's function is minimum phase does not mean that the room's Green's function would be. They are most certainly NOT the same functions. In fact, in Morse, he shows how the free field Green's function can be separted off of the rooms Green's function leaving only the effect of the room. This might be worth looking at. Morse also discusses the room transient response showing how it violates characteristics of a linear time invarient system in that the room decays at frequencies which need not be present in the input. These features of acoustics tend to make electrical domain considerations of things like MP tenative at best.
I have no doubt that what you are saying is possible, but you are mixing subjective impression into the problem and this will make it almost untenable since you would have to claim a knowledge of how we perceive LF sounds and I don't think that such a clear knowledge exists. Its one thing to use Green's functions to show the spectral and spatial characteristics of a sound field, but its an entirely different matter to use that to talk about "how they sound".
gedlee said:
I think you misinterpreted what I meant. I was just pointing out the solution for Green's function is the source to listener transfer function and that in free space it is just a time delay with respect to phase. I am familiar with the in room GF;
G = g + X
where g is the free space Green’s function and X is the solution to the homogeneous form of the equation for which the wall boundary conditions are applied.
Well, I'm not mixing the subjective into the problem. I was looking for an objective explanation as to why difference sources may sound different when the source and listing position are the same if eq'ed to the same amplitude response. If the eq'ed response were MP plus at most a delay then the transient behavior would have to be the same. If it is not then different source can have different transient decays. This could objectively explain some of the subjective claims.
At this time I think I have pretty much convinced myself that while a MP response can occur, in general the response at the listening position will not be MP and therefore different sources, even after being eq'ed to flat amplitude response, or the same source type positioned differently, can have different transient response which could account for the subjective comments made by listeners. This still doesn't define better or worse, just different. Saying a dipole sounds different because it excites fewer modes is really off target. To excite fewer modes it must be very carefully positioned. But even more relevant is the observation that any source eq’ed flat for a fixed listening position will produce different transient response just by changing the position of the source (with new eq) so different modes are excited differently, assuming the response is not MP.
John
I just thought of an interesting study that I did as part of my thesis that may be pertinent here.
I tried to simulate the direct/near field from the modal solutions of a room. By taking listener points ever closer to the source I wanted to plot out the direct to reveberant transition. Didn't work.
Thats when I found the analysis by Morse which explained everything. The series solution to the Greens function converges very slowly near the source - it being a singularity. He showed how you could seperate off the free field Green's function from the series solution as a singular function (which has perfect convergence) and leaves behind a new series which also converges much more rapidly. With this approach you do get the direct to reverberant transition. The fact that the near field (or direct field) is contained in the series solution was an error in Todd Welti's subwoofer paper, which I wrote to JAES correcting. The series solution does contain the near field, but it takes a lot of modes to develop it. In short, ALL modes contribute to the near field - these are called evanescent modes if you follow the literature.
I am sure that you could find the near field from this approach even for dipoles and cardiods by using Morse's approach, but you will never get it from a pure series. The terms that remain after seperating off the near field term are more than likely NOT MP because of the extra omega squared in the denominator of the series. This make the "resonances" not act like normal resonances.
I just thought of an interesting study that I did as part of my thesis that may be pertinent here.
I tried to simulate the direct/near field from the modal solutions of a room. By taking listener points ever closer to the source I wanted to plot out the direct to reveberant transition. Didn't work.
Thats when I found the analysis by Morse which explained everything. The series solution to the Greens function converges very slowly near the source - it being a singularity. He showed how you could seperate off the free field Green's function from the series solution as a singular function (which has perfect convergence) and leaves behind a new series which also converges much more rapidly. With this approach you do get the direct to reverberant transition. The fact that the near field (or direct field) is contained in the series solution was an error in Todd Welti's subwoofer paper, which I wrote to JAES correcting. The series solution does contain the near field, but it takes a lot of modes to develop it. In short, ALL modes contribute to the near field - these are called evanescent modes if you follow the literature.
I am sure that you could find the near field from this approach even for dipoles and cardiods by using Morse's approach, but you will never get it from a pure series. The terms that remain after seperating off the near field term are more than likely NOT MP because of the extra omega squared in the denominator of the series. This make the "resonances" not act like normal resonances.
john k... said:
Except that if g is the free field Green's function it IS NOT a solution to the room boundary conditions and thus X cannot be either. Its has to be a soution to a different problem since g + X is a solution to the original one. No big deal here I suppose.
This COULD define a difference IF a true difference was shown to be the case, but there is only circumstantial evidence for that. I would have thought your last sentence obvious from my data since no two listener responses are the same it follows that no two source positions to the same listening point would be either. EQ'd flat, to me, only has relavence in terms of the power response for exactly this reason. But if your were to EQ to flat exactly one single point for both a monpole and say a dipole, I don't think that you can conclude that the time responses will be different at that point (although it will be quite different at other points). It could be, but I don't see in the math why it has to be true.
Etienne88 said:
On another thread he states his position. No data, many claims.
See this post:
http://www.diyaudio.com/forums/showthread.php?postid=1518203#pos
Etienne88 said:
Thanks for the link, it was interesting reading!
As you said: "no data, many claims"... So maybe he is not a "real" scientist...!
Regards,
Etienne
Jean-Michel acknowledged to me on another thread that all of his "data" comes from his own ears and if this is contradicted by scientific data then so be it. That isn't very scientific at all.
I don't totally disagree with him, he is honest, but I'm unwilling to throw away valid data simply because my ears may not agree with it at the moment.
gedlee said:
I understand your point of view! The situation is nevertheless ambiguous...
Jean-Michel is an academic person would spend his life doing research, from that point of view he is highly scientific. Then he doesn't agree with results that are also scientific but based on the statistic response of a panel of people. I guess you know statistic better than me: there is always some spreading, most of the people will be quite close to the average (you know the 2 or 3 sigmas zone if you consider a Gaussian repartition) but there will still be persons outside. Not many but it will be! He might be one of them...
Reading a book from Mario Rossi (Audio) made me realise that. He says (in French in the original text) that the study of psychoacoustic is based on experimental methods by which the statistical response of a large number of participants to specific stimuli allows to establish a relationship between the stimuli and the auditive feeling (sensation). This is done mostly without trying to understand the ear mechanism involved.
After reading that I realise that it might be a grey zone surrounding psychoacoustic even if the study of it is perform according to scientific methods!
Regards,
Etienne
Etienne88 said:
This is a misunderstanding of the scientific method.
Harman has done numerous studies of experts listeners and non-experts. Both have exactly the same mean responses, but differ in the spread. There is not a difference in preference between an expert and a novice, the expert is just better able to get to the answer with fewer trials.
Your points imply that an "expert" may prefer something that the novice doesn't, but all data points to the contrary.
No one is so "good" at listening that they don't make errors. It has been shown time and time again that data that is not taken blind is strongly influenced by biases that disapear when the evaluation is done blind.
Humans are simply not very good evaluators when their brains are free to influence the results. The brain is more powerful than the ear.
gedlee said:
this statement could not be more true. evaluating the evaluator is something that most people never take the time to do (and least frequently when the evaluators are ourselves). here is just a "short list" of cognitive biases in humans:
http://www.healthbolt.net/2007/02/14/26-reasons-what-you-think-is-right-is-wrong/
and some more (not all unique):
http://en.wikipedia.org/wiki/List_of_cognitive_biases
Ok, thank you Earl for opening my eyes! You mean that Jean-Michel might be biased when he says that he can hear phase distortion?
He says that he can hear it in a known environment (his listening room) with know music. Se page 22 of this document (http://freerider.dyndns.org/anlage/LeCleach2.zip). I don’t know if he tried with a kind of blind test… And even with a blind test, he might have been biased by the fact that he knew what he would be tested for!
“The brain is more powerful than the ear” Yes a do believe that!
And thank you Publius for the link, it was very instructive. I think I suffer from confirmation bias!
The power of the brain might also be the reason why it is more valuable for big companies to work on marketing than working on research…
Regrads,
Etienne
He says that he can hear it in a known environment (his listening room) with know music. Se page 22 of this document (http://freerider.dyndns.org/anlage/LeCleach2.zip). I don’t know if he tried with a kind of blind test… And even with a blind test, he might have been biased by the fact that he knew what he would be tested for!
“The brain is more powerful than the ear” Yes a do believe that!
And thank you Publius for the link, it was very instructive. I think I suffer from confirmation bias!
The power of the brain might also be the reason why it is more valuable for big companies to work on marketing than working on research…
Regrads,
Etienne
Etienne88 said:
We are all biased and none of us is capable of doing scientific work when we know what it is we are listeing too and for.
Your last point is one that I have stated over and over again. Only if we get away from "subjective" reviews and learn to trust the measurements (which means, of course, that they have to be good measurements of the right things) can we ever hope to increase our level of understanding and make valid purchasing decisions that are not "directed" by the marketing department.
I also can hear some kind of phase distortions but only with special signals :
- group delay in bass frequencies with sawtooth signal
- phase distortion of some typical filters with triple tone signal
But, frankly, those effects are really minor compared to many other distortions.
To do my own tests, I programmed two free testing tools and would be interested to know about other (different ?) results :
one is an IIR allpass filter
the other is a FIR filter with adjustable delays
CLS said:Beside that Big OB plan which got zero insterest, I'm thinking of building a bass module with cardioid characteristic into that system instead of dipole.
My main thinking is to get better directivity control, reduce reflections and room modes, and higher output.
Here are some thoughts:
A is well-known NaO which is a semi open back U frame.
B is a combination of monopole and dipole. There'd be cancellation on the back.
C is stolen from here: http://www.duran-audio.com/pdfs/downloads/brochures/AXYS_Target_brochure.pdf
(In the middle of page 3, you may see a cardioid configuration.)
At the first glance, C seems to be a waste of drivers because 2 of them work in a way almost identical to 1 in dipole configuration.
The AXYS system emphansizes DSP. Does this do anything good to that "waste" mentioned above?
Any thoughts and suggestions?
For subs the best way is to have the speaker mounted facing forward but have a duct from the back of the speaker that comes back round to the front. If the duct is the right length then the sound from the back will be back in phase with the front giving almost twice the sound level from your speaker.
You can work out the duct length from the speed of sound and its wavelength.
Jlo,
Even if I don't know how you performed your distortion test, I think you were biased by the fact that you knew that you were testing it... I might be wrong, but it is a bit like testing a new set of cables you more or less want them to be different than your previous set. You might thus "hear" what you want...!
Thank you for the link to your website! Your softwares might be helpful. I saw you had something about room acoustic there. You might found some valuable information on Earl's website and in his comments in the following thread LS and room as a system. Reflections are discussed.
Regards,
Etienne
Even if I don't know how you performed your distortion test, I think you were biased by the fact that you knew that you were testing it... I might be wrong, but it is a bit like testing a new set of cables you more or less want them to be different than your previous set. You might thus "hear" what you want...!
Thank you for the link to your website! Your softwares might be helpful. I saw you had something about room acoustic there. You might found some valuable information on Earl's website and in his comments in the following thread LS and room as a system. Reflections are discussed.
Regards,
Etienne
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.