panomaniac said:
Oddly enough, one of the best systems I ever heard was purely mechanical - recording and playback. An Edison Disk system. Not cylinder, disk. It was at the little "Museum of Technology" in Palo Alto, California. The player looked like a deluxe version of the typical Victrola, but used Edison's vertical needle movement on a 12" disk.
The owner of the museum saw that I was interested and asked if I cared to listen to it. "Of course!" Well, I was floored. The sound was astonishing.Not Hi-Fi, no - there was surface noise and limited bandwidth, but it was so lifelike, pure, so good.
As the disk played it drew in people from all over the museum. They all had to come see what sounded so good, find the source of that great sound. And they all commented on how good it sounded. "What IS that thing?" was a common refrain.
Don't even remember what the record was. A man singing with a small orchestra behind, IIRC. The amazing thing was how real, how good, how musical this old machine and the purely mechanical recording sounded. How have we strayed so far?
I may have heard the same phonograph some time in the early Seventies, when I was visiting my retired parents in Berkeley. I was idly strolling along Telegraph Avenue, and heard a really good opera singer inside a small arcade. Drawn off the busy street by the sound, I walked inside, turned a corner and was astonished to find a big Edison acoustic phonograph playing a thick, blue-colored (must have vertical-cut) 12" disk. I stood and listened to the whole length of the record - it really was a good approximation of a somebody stranding right there and singing, and singing damn good classical opera.
Was it hifi? No. But it did some things hifi systems don't - in important ways, it sounded real. No "electronic" colorations at all, and the mechanical colorations had somehow been ingeniously concealed for the human voice. The orchestral backing was pretty funky-sounding, but there wasn't much of it, which was just as well.
This was a truly educational experience - I was already indoctrinated into high-end audio, having subscribed to J. Gordon Holt's early Stereophile for several years at that time, owned exotica like a Thorens TD-125, Rabco SL-8E, and a Stanton 681A cartridge, and persuaded my sister to buy Fulton FMI-80 speakers (which I think she still owns - good speakers).
But that top-of-the-line Edison gramophone made me think about a lot of unquestioned assumptions I'd made about audio. It wasn't all about frequency response, impulse response, and freedom from resonance. There's a very direct and immediate perception of realness - so strong it brought me off the never-ending circus of Telegraph Avenue on a summer's day - into a secluded courtyard, and to a revelatory experience I never would have expected.
This openness to the unexpected has a been a gift, one of the deepest and most essential parts of the human spiritual endowment, and I am absolutely sure we are all born with it. I am afraid, though, that culture, and worse, education, beats it out of many of us, leaving room for only small deviations from the "known" and "true". I am also afraid that perception itself is strongly affected by prior experience, shutting off entire worlds of perception if we "already know" that certain things are impossible.
I probably offend many Western-trained rationalists with my somewhat unconscious acceptance of Taoist and Buddhist notions of "relative" and "absolute" truth. In the simplest terms, Taoists and Buddhists believe that any truth that can be described, or written down in a book, is a "relative" truth, subject to change with culture and time, and "absolute" truths are essentially ineffable, beyond all description (without gross distortion), and can only be directly experienced.
As a result, all of the world's religions are extremely distorted and limited descriptions of experiences (by the founders, saints, and prophets) that cannot be described with any accuracy at all.
This notion, of course, is abhorrent to all fundamentalists, who view Scripture as literally hand-written by God, or Newton, depending on persuasion. I guess it makes sense - if you believe Absolutes already exist, and not only that, must be defended to-the-death against barbarians and infidels, the entire notion of "absolute" and "relative" truths is the most appalling and dangerous heresy possible, because it is so seductive and intellectually appealing.
I also believe that Science as a social movement has very slowly assumed the role of religion in Western culture, right down to the white gowns, ceremonies like the Nobel Prizes, hidden power struggles, and the growth of competing sects and guilds. There are a lot of unquestioned metaphysical assumptions that are pretty obvious to Taoists and Buddhists - this doesn't undermine scientific method, which will always be useful, but the science-as-worldview does have dogmatic and emotionally defended religious overtones.
I take a more relativistic stance that basically "everything we know" could easily be proven utterly, and completely, wrong in the next several hundred years, and thus hardly makes a serious claim to any kind of "absolute" truth. A short look at the history of ideas indicates what happened in the past is happening to us right now - we are not exempt from history, but are right in the middle of it. What we have is an extremely powerful working tool (and should not be discarded), but it isn't necessarily the same thing as reality itself.
Hi Guys
With all the impulses being shown it brings up a question or an observation about what impulse responses actually show you. It is most often said “its time” but what does it say?.
What I mean is that I guess you’d say impulse responses are one way to look at the speaker but what they tell you is not always obvious or intuitive.
For example, in this thread Edward West posted the magnitude and phase of a perfect speaker and a second curve with the effect of a conventional perfect crossover in the middle.
In both cases the impulse response is posted.
http://www.audio-trek.net/diyAudio/MinPhaseIntro/MinimumPhase.htm
So if one takes a “perfect” speaker attached to a perfect band pass filter set, what does that filter do to the impulse response? Look further in Edward’s direction.
For say a simple 2nd order filter at each end, one gets a traditional good looking impulse response. One can see that the impulse on its return stays below zero for some time, satisfying the equal area criteria Earl mentions.
If one switches to sharper high pass filter slope at the same F, one see’s the time of the recovery period below zero shortens / deepens, sharper filters also cause a slow oscillation above and below zero at a period related to the lf corner.
Raising the high pass corner F causes a deeper return past zero and shorter period.
In creasing the sharpness of the low pass corner one see’s strong ringing in the recovery period, which might be interpreted as a resonance.
Lowering the low pass frequency, increases the period of the impulse.
While all of these produce radically different impulse responses, the magnitude and phase shows directly what the speaker was doing. Time, as to how the speaker reproduces a broad band waveshape is shown in the acoustic phase response, where phase lags, this portion is late, where it leads, it is ahead (possible with a sine wave signal that has existed long enough in time.).
The impulse (via Heyser’s view) is “Voltage” from the real part (in the resistive plane) of the Energy vs Time curve. As one hears pressure (in both real and imaginary planes) and that actuates the microphone, it is the energy vs time which seems like it might be more useful. Here, due to the log presentation and energy from both planes, reflections and stored energy show up very well although weighted towards the high frequency end .
Lastly, I was hoping it would be up at our web site by now but I have found a reality test which might be useful. It was used at a recent Synaudcon seminar with some puzzlement and quandary. In Pro-sound, many of the dual FFT and sequence based measurement systems people use do not actually measure acoustic phase, while they have a curve for it.
I got in hot water on PSW some years ago for suggesting this but I figure it’s worth mentioning here, if the phase isn’t right, I wonder if the impulse response is “right” too.
Anyway here is the test should you want to find out if a given measurement system is a real Poncho or a Sears Poncho (with respect to Frank Z.) so far as measuring actual acoustic phase.
Take a good loudspeaker controller or passive crossover with a time delay, set it up with a 4 th order high pass filter at 100Hz and a 2nd order low pass at 5KHz.
This emulates a perfect band limited driver.
The phase for this filter will look similar to the one shown in Edward’s earlier post, it comes down at the low frequency end and goes down again at the hf end.
The slope and phase span related to the order of filter used.
What Edward’s post showed is the theoretical response of a perfect bandpass, with no time delay in the system. We want a perfect response too so we use a speaker controller / crossover.
To be a real speaker under test however, one must also add say 3-4ms of time delay to that filter (in addition to the latency already supplied digitally).
Now, test that closed system as “the speaker” in fact a “perfect” speaker with a known magnitude and phase.
Most measurement systems show the phase at the hf end going up instead of down, (some show the phase at the lf end with too little slope).
At least with TDS, that hf error (phase going up at the hf end) suggests that the time reference is slightly off, the lf slope error I have no idea what the is related to.
Anyway, since we are dealing with phase, it is good to know that the system actually measures it properly.
Earl, John
So lets say you had a driver with a peak at its hf end, say centered at 3KHz and say 10dB in magnitude for ease of discussion.
Lets say you use an fir or iir filter to completely fix the magnitude and phase.
The “yes but” part is that with that mound being a result of acoustic gain after the motor, that gain is still present after being “fixed” and may not be inconsequential.
That same “after the motor” gain magnifies what ever distortion the motor produced at sub harmonics of that, like magnifying the 3rd harmonic about 3.3 times when driven at 1KHz.
I don’t think that distortion magnification effect goes away with this kind of correction so even with DSP, one would especially want it to be an extra low distortion driver .
Lastly, so far as looking at a systems time response in a proportion way, what do you guys think about the Wavelet view? It appears to be something like a CSD but done with a view of acoustic size / wavelength, not time per say.. Seems logical.
I found a nice write-up for those interested.
http://users.rowan.edu/~polikar/WAVELETS/WTpart1.html
Best,
Tom Danley
With all the impulses being shown it brings up a question or an observation about what impulse responses actually show you. It is most often said “its time” but what does it say?.
What I mean is that I guess you’d say impulse responses are one way to look at the speaker but what they tell you is not always obvious or intuitive.
For example, in this thread Edward West posted the magnitude and phase of a perfect speaker and a second curve with the effect of a conventional perfect crossover in the middle.
In both cases the impulse response is posted.
http://www.audio-trek.net/diyAudio/MinPhaseIntro/MinimumPhase.htm
So if one takes a “perfect” speaker attached to a perfect band pass filter set, what does that filter do to the impulse response? Look further in Edward’s direction.
For say a simple 2nd order filter at each end, one gets a traditional good looking impulse response. One can see that the impulse on its return stays below zero for some time, satisfying the equal area criteria Earl mentions.
If one switches to sharper high pass filter slope at the same F, one see’s the time of the recovery period below zero shortens / deepens, sharper filters also cause a slow oscillation above and below zero at a period related to the lf corner.
Raising the high pass corner F causes a deeper return past zero and shorter period.
In creasing the sharpness of the low pass corner one see’s strong ringing in the recovery period, which might be interpreted as a resonance.
Lowering the low pass frequency, increases the period of the impulse.
While all of these produce radically different impulse responses, the magnitude and phase shows directly what the speaker was doing. Time, as to how the speaker reproduces a broad band waveshape is shown in the acoustic phase response, where phase lags, this portion is late, where it leads, it is ahead (possible with a sine wave signal that has existed long enough in time.).
The impulse (via Heyser’s view) is “Voltage” from the real part (in the resistive plane) of the Energy vs Time curve. As one hears pressure (in both real and imaginary planes) and that actuates the microphone, it is the energy vs time which seems like it might be more useful. Here, due to the log presentation and energy from both planes, reflections and stored energy show up very well although weighted towards the high frequency end .
Lastly, I was hoping it would be up at our web site by now but I have found a reality test which might be useful. It was used at a recent Synaudcon seminar with some puzzlement and quandary. In Pro-sound, many of the dual FFT and sequence based measurement systems people use do not actually measure acoustic phase, while they have a curve for it.
I got in hot water on PSW some years ago for suggesting this but I figure it’s worth mentioning here, if the phase isn’t right, I wonder if the impulse response is “right” too.
Anyway here is the test should you want to find out if a given measurement system is a real Poncho or a Sears Poncho (with respect to Frank Z.) so far as measuring actual acoustic phase.
Take a good loudspeaker controller or passive crossover with a time delay, set it up with a 4 th order high pass filter at 100Hz and a 2nd order low pass at 5KHz.
This emulates a perfect band limited driver.
The phase for this filter will look similar to the one shown in Edward’s earlier post, it comes down at the low frequency end and goes down again at the hf end.
The slope and phase span related to the order of filter used.
What Edward’s post showed is the theoretical response of a perfect bandpass, with no time delay in the system. We want a perfect response too so we use a speaker controller / crossover.
To be a real speaker under test however, one must also add say 3-4ms of time delay to that filter (in addition to the latency already supplied digitally).
Now, test that closed system as “the speaker” in fact a “perfect” speaker with a known magnitude and phase.
Most measurement systems show the phase at the hf end going up instead of down, (some show the phase at the lf end with too little slope).
At least with TDS, that hf error (phase going up at the hf end) suggests that the time reference is slightly off, the lf slope error I have no idea what the is related to.
Anyway, since we are dealing with phase, it is good to know that the system actually measures it properly.
Earl, John
So lets say you had a driver with a peak at its hf end, say centered at 3KHz and say 10dB in magnitude for ease of discussion.
Lets say you use an fir or iir filter to completely fix the magnitude and phase.
The “yes but” part is that with that mound being a result of acoustic gain after the motor, that gain is still present after being “fixed” and may not be inconsequential.
That same “after the motor” gain magnifies what ever distortion the motor produced at sub harmonics of that, like magnifying the 3rd harmonic about 3.3 times when driven at 1KHz.
I don’t think that distortion magnification effect goes away with this kind of correction so even with DSP, one would especially want it to be an extra low distortion driver .
Lastly, so far as looking at a systems time response in a proportion way, what do you guys think about the Wavelet view? It appears to be something like a CSD but done with a view of acoustic size / wavelength, not time per say.. Seems logical.
I found a nice write-up for those interested.
http://users.rowan.edu/~polikar/WAVELETS/WTpart1.html
Best,
Tom Danley
Tom Danley said:
Tom, there is no argument that the gain remains and that any nonlinearities or other stimuli that enter the system between the equalization and the radiating surface will be amplified by the native gain. But the gain is the linear part of the system which can be corrected for in the input. We have no control of the nonlinearities unless we employ feedback.
Equalization will only correct the linear aspects of the system and will only be correct for the one point in space where they are derived for. Not unlike equalizing a speaker for flat in room response at the listening position. It's good for that position only. So if I eq driver 1 with a big resonance peak, to have the identical response as driver 2 with a smooth response at some point in space that doesn’t mean that 15 degrees off axis the two driver will still measure the same using the same eq. This is another reason why (and here I agree with soongsc) that even after eqing two drivers may sound very different in a real environment. Even if the drivers are perfectly linear. Eqing them to have identical direct (let's just say on axis) response doesn’t imply the same off axis response.
About wavelets I can really comment. Interesting that the page you reference was written in ’95 and not updated or continued yet. But isn’t this something that is more like histogram? I.E it tells me what frequencies are present and what their magnitude is as a function of time?
Lynn Olson said:
That's funny as hell, Lynn. Yeah, it may have been the exact same Edison phonograph and even the same record.
Your description is pretty much what I remember. The voice was stunning.A lot of us know that Mr. Edison conducted a number of "Fool the Audience" demos of his record player. The listeners were said not to be able to tell the difference between live and the Edison recordings. We laugh at that from our perspective of a century hence. And we "know" how bad those old phonographs sounded. Well, hearing that Edison machine really changed my mind.
Thru my old buddy Jean Hiraga I was aware that a number of Japanese "Hi-Fi Nuts" were lovers and collectors of old acoustic recordings and playback gear. Never gave it much thought until that day in Palo Alto. Over the years I've talked to a few Edison fans who told me that his record players were the very high end of the market. Top quality build, very expensive. Not a huge success.
Here is an article by one of those Edison fans, Mary Bellis.
Edison Disc on About.com
It gives me quite a chuckle to talk about this old technology right here in the middle of an extremely detailed discussion of impulse response and other arcane technological. You guys must think we're nutz. Kinda are, yeah. But I assure you, within the scope of voice and probably small ensemble, the old Edison Disc player would blow away most of what you've heard, at least on an emotional level. There is a very direct connection to the music that's lacking in most modern stuff.
OK, back to impulse measurements.
Martin J. King has put a paper on his site which investigates the open baffle speaker. It is found here:
http://www.quarter-wave.com/General/OB_Design.pdf
http://www.quarter-wave.com/General/OB_Design.pdf
Well, while we conduct this parallel discussion (which I thoroughly enjoy, by the way) I'd like to bring up the related matter of "out-of-the-room" sound. By this I mean the occasional experience of hearing a vivid impression of a singer being in the house, while you yourself are in different part of the house. The illusion frequently collapses just as you enter the living room, but before you see anything in the room, so I'm not too sure it's a matter of visual perception overriding the sound.
Now when this happens (I've heard it myself many times), there's a lot going on. First, you're taken by surprise - "my hifi doesn't sound like THAT, who's the musical guest". On the physical plane, the sound is taking many different paths to get to us, but there's still a first-arrival sound. However, it's only a matter of chance from which direction that first sound was originally emitted by the loudspeaker, and unlikely it was in the fairly narrow favored frontal plane and on the horizontal axis.
Not only that, as we walk down the hall towards the sound, the relationship of the first-arrival waves continually changes, yet sonically, we hear almost no timbral change at all, much like the constancy of color-temperature perception in different environments and different light sources. This, I think, is the toughest one to explain using the standard models of hearing.
We know with color vision, the constancy of color-temperature is actually extraordinarily complex in a physiological sense, and difficult for movie makers and photographers to compensate except on a scene-by-scene esthetic basis. Photographers and movie-makers have been fighting with this for more than a half-century, and auto-color-balance is still unsatisfactory except for very undemanding amateur use. Professionals always balance color on a carefully controlled subjective basis, using precision-calibrated display systems and viewing environments. And all we're discussing here is color balance!
The problem of "realism" in audio goes much deeper than the simpler and better-understood mechanisms of color balance in vision. When it's right, we feel "yeah, that's good", and feel a mood elevation if we like the music. When it's close-but-no-cigar, the feelings are more complex, but musical satisfaction (an emotion) is decreased or entirely removed. When the error is gross, we're actively repelled, and want to leave. This is primarily an emotional percept with an after-the-fact attempted interpretation and rationalization of the sonic quality.
This isn't as simple as turning the gamma or H&D curve knobs for RGB, YUV, or CMYK colorspaces - we aren't even sure which knobs to turn, or worse, maybe they don't exist yet!
Now when this happens (I've heard it myself many times), there's a lot going on. First, you're taken by surprise - "my hifi doesn't sound like THAT, who's the musical guest". On the physical plane, the sound is taking many different paths to get to us, but there's still a first-arrival sound. However, it's only a matter of chance from which direction that first sound was originally emitted by the loudspeaker, and unlikely it was in the fairly narrow favored frontal plane and on the horizontal axis.
Not only that, as we walk down the hall towards the sound, the relationship of the first-arrival waves continually changes, yet sonically, we hear almost no timbral change at all, much like the constancy of color-temperature perception in different environments and different light sources. This, I think, is the toughest one to explain using the standard models of hearing.
We know with color vision, the constancy of color-temperature is actually extraordinarily complex in a physiological sense, and difficult for movie makers and photographers to compensate except on a scene-by-scene esthetic basis. Photographers and movie-makers have been fighting with this for more than a half-century, and auto-color-balance is still unsatisfactory except for very undemanding amateur use. Professionals always balance color on a carefully controlled subjective basis, using precision-calibrated display systems and viewing environments. And all we're discussing here is color balance!
The problem of "realism" in audio goes much deeper than the simpler and better-understood mechanisms of color balance in vision. When it's right, we feel "yeah, that's good", and feel a mood elevation if we like the music. When it's close-but-no-cigar, the feelings are more complex, but musical satisfaction (an emotion) is decreased or entirely removed. When the error is gross, we're actively repelled, and want to leave. This is primarily an emotional percept with an after-the-fact attempted interpretation and rationalization of the sonic quality.
This isn't as simple as turning the gamma or H&D curve knobs for RGB, YUV, or CMYK colorspaces - we aren't even sure which knobs to turn, or worse, maybe they don't exist yet!
Our understanding of vision and mentioned chromatic adaptation is much better than our hearing. But in both it is quite easy to fool if you don't know the original. With digital camera you can turn picture by single 3x3 convolution into the mostly like an original picture within medium dynamic range. I've scanned and processed a thousands of negatives and resulting pictures are damn good even if they are far from original colours. That's a beauty of analogue medium. People like coloured sound too.
Lynn Olson said:
Lynn,
I have experienced this phenomenon. From my perspective the three most important characteristics that a speaker system has to possess to exhibit this are:
1) Uniform power response with decent dispersion across the full spectrum. Since the sound that leaves a room is essentially a sample of the total power put into the room, uniform power response is necessary to get believable spectral balance in the out-of-room-sample.
2) Low energy storage in the speaker system. Since the out-of-room-sample is an integration of all the sound being input to the room, un-natural sounds such as dissonant panel resonances will skew the spectral decays. During in-room listening, this can be masked by the first-arrival. But listening from another room, all the emitted sound is on "equal footing" so the un-natural sources need to limited.
3) Really good midrange presence. This is a characteristic which I have real difficulty describing, but I know it when I hear it
It's a liveliness which I have heard in a number of drivers - all of which happen to be paper.While I am sure there are other contributing characteristics, I believe these to be the major requirements to achieve the 'who is the guest performer in my house?" experience.
Edward
Nice thinking, Edward. Seems to make sense to me.
Any one else care to comment? It's an interesting phenomenon, the "out of room - it's real" trick.
As much as I love Quad ESL panels, and as truly wonderful their sound, I don't remember them ever doing that. Lynn, do you?
As for color science... wow, don't get me started. It's what I do for a living, so I don't have the energy left to think about it off hours.
Any one else care to comment? It's an interesting phenomenon, the "out of room - it's real" trick.
As much as I love Quad ESL panels, and as truly wonderful their sound, I don't remember them ever doing that. Lynn, do you?
As for color science... wow, don't get me started. It's what I do for a living, so I don't have the energy left to think about it off hours.
FWIW I have noticed this effect many times over: severly off axis or somewhere else in the room, this is where my system sounds most real, not just with voices but with almost any material. Only exception maybe orchestral or other material with lots of reverb already on the record.
My interpretation is twofold:
- we need a fair amount of reflections in the 10-50 ms range for optimal enjoyment, and they even enhance intelligibility. This comes up over and over in Grieseinger, Toole, Linkwitz, Geddes etc. As Earl Geddes points out, I believe in the Summa white paper, is that in an average small room you're better off having a directional speaker because otherwise your first reflection comes too soon. So as far as I understand he recommends listening to directional speakers off axis, and pointed at the opposite side wall at 45 degrees, to maximise the distance for the first reflection to arrive in your ears, to minimize early reflectionns from the wall closest to the speaker, and yet to maximise medium arrival time room reflections (!).
So, in a typical domestic environment, you must be in a different room or the hallway to get your 10-50 ms spread of reflected sound that is apparently optimal. Dipoles btw IF you can afford to place them well afar the front wall have it easier since say a distance of 1.8 m or 6 ft to the front wall already gets the reflection of the rear radiation a delay of 10 ms as compared to the direct sound.
- close miking makes for an unnatural presentation of the source - no one typically listens at 2 inches from any source. But on-axis close to a speaker, this is what you hear: the sound spectrum that was recorded at 2 inches.
My interpretation is twofold:
- we need a fair amount of reflections in the 10-50 ms range for optimal enjoyment, and they even enhance intelligibility. This comes up over and over in Grieseinger, Toole, Linkwitz, Geddes etc. As Earl Geddes points out, I believe in the Summa white paper, is that in an average small room you're better off having a directional speaker because otherwise your first reflection comes too soon. So as far as I understand he recommends listening to directional speakers off axis, and pointed at the opposite side wall at 45 degrees, to maximise the distance for the first reflection to arrive in your ears, to minimize early reflectionns from the wall closest to the speaker, and yet to maximise medium arrival time room reflections (!).
So, in a typical domestic environment, you must be in a different room or the hallway to get your 10-50 ms spread of reflected sound that is apparently optimal. Dipoles btw IF you can afford to place them well afar the front wall have it easier since say a distance of 1.8 m or 6 ft to the front wall already gets the reflection of the rear radiation a delay of 10 ms as compared to the direct sound.
- close miking makes for an unnatural presentation of the source - no one typically listens at 2 inches from any source. But on-axis close to a speaker, this is what you hear: the sound spectrum that was recorded at 2 inches.
Tom you wrote
"The “yes but” part is that with that mound being a result of acoustic gain after the motor, that gain is still present after being “fixed” and may not be inconsequential."
Absolutely true, but it is inconsequential if nonlinear distortion in a loudspeaker is not an audible effect.
About wavelets. This is my perspective (from a sonar point of view).
We take data. We want to detect certain things which are in that data. The optimal detector depends very much on what it is we are looking for. Except to look at very broad mostly random spectra, the FFT is seldom, if ever, the optimal detector. Wavelets are ideal when there are randomly occurring deterministic signal like sine waves that come and go. The wavelets will lock right into them.
To measure distortion in an amplifier I use frame locked sine waves and a detector that is basically a Fourier series, on data that has been sychronously averaged at the frame rate. With this technique I can measure almost 20 dB below the noise floor of the measurement system - a sound card in this instance. A standard FFT use on a 1 kHz sine wave cannot get down to much lower than a few dB above the noise floor.
In a mathematical sense the FFT has say 1024 degrees of freedom, but its not smart enough to allocate these degrees of freedom in anything but a fixed manner. Thats why its called a non-parametric estimator. Now a wavelet can target its degrees of freedom to specific waveforms yield much higher performance. It is called a parametric estimator. There are lots of parametric estimators and I wrote a few papers on this about 20 years ago.
So the bottom line is, the nature of the data you are trying to detect will dictate the best detector. Wavelets win sometimes, Proney series, or Fourier series win at other times. There is no one ideal.
A bit lengthy, but it is an interesting subject.
I write all of my own measurement algorithms because none of the existing software are flexible enough. The FFT is the workhorse because of its speed and simplicity, but its seldom the only techniqu I use. It has terrible LF resolution for example.
"The “yes but” part is that with that mound being a result of acoustic gain after the motor, that gain is still present after being “fixed” and may not be inconsequential."
Absolutely true, but it is inconsequential if nonlinear distortion in a loudspeaker is not an audible effect.
About wavelets. This is my perspective (from a sonar point of view).
We take data. We want to detect certain things which are in that data. The optimal detector depends very much on what it is we are looking for. Except to look at very broad mostly random spectra, the FFT is seldom, if ever, the optimal detector. Wavelets are ideal when there are randomly occurring deterministic signal like sine waves that come and go. The wavelets will lock right into them.
To measure distortion in an amplifier I use frame locked sine waves and a detector that is basically a Fourier series, on data that has been sychronously averaged at the frame rate. With this technique I can measure almost 20 dB below the noise floor of the measurement system - a sound card in this instance. A standard FFT use on a 1 kHz sine wave cannot get down to much lower than a few dB above the noise floor.
In a mathematical sense the FFT has say 1024 degrees of freedom, but its not smart enough to allocate these degrees of freedom in anything but a fixed manner. Thats why its called a non-parametric estimator. Now a wavelet can target its degrees of freedom to specific waveforms yield much higher performance. It is called a parametric estimator. There are lots of parametric estimators and I wrote a few papers on this about 20 years ago.
So the bottom line is, the nature of the data you are trying to detect will dictate the best detector. Wavelets win sometimes, Proney series, or Fourier series win at other times. There is no one ideal.
A bit lengthy, but it is an interesting subject.
I write all of my own measurement algorithms because none of the existing software are flexible enough. The FFT is the workhorse because of its speed and simplicity, but its seldom the only techniqu I use. It has terrible LF resolution for example.
Hi
As for me I am considering this experience as something related to timing ( phase ? group delay ?) rather than to timbre
You can have it even with TV sets occasionally – though I don't get it from my current setup
Most astonishing is the effect in its extreme - when you are listening to what is played in the next floor.
Here we have a very limited bandwidth and though you can tell if that bass is " walking right " or not.
Greetings
Michael
panomaniac said:
As for me I am considering this experience as something related to timing ( phase ? group delay ?) rather than to timbre
You can have it even with TV sets occasionally – though I don't get it from my current setup
Most astonishing is the effect in its extreme - when you are listening to what is played in the next floor.
Here we have a very limited bandwidth and though you can tell if that bass is " walking right " or not.
Greetings
Michael
I've had the same experience.EdwardWest said:
Lynn,
I have experienced this phenomenon. From my perspective the three most important characteristics that a speaker system has to possess to exhibit this are:
1) Uniform power response with decent dispersion across the full spectrum. Since the sound that leaves a room is essentially a sample of the total power put into the room, uniform power response is necessary to get believable spectral balance in the out-of-room-sample.
2) Low energy storage in the speaker system. Since the out-of-room-sample is an integration of all the sound being input to the room, un-natural sounds such as dissonant panel resonances will skew the spectral decays. During in-room listening, this can be masked by the first-arrival. But listening from another room, all the emitted sound is on "equal footing" so the un-natural sources need to limited.
3) Really good midrange presence. This is a characteristic which I have real difficulty describing, but I know it when I hear it
While I am sure there are other contributing characteristics, I believe these to be the major requirements to achieve the 'who is the guest performer in my house?" experience.
Edward
We had the speaker shown in my avatar in a small 3MX3M room once, and many people wondered why it sounded good even when listening from outside the room and where ever you stand in the room. The interesting thing is that a metal cone was used and controlling cone vibration was an interesting process. There was a point where we could see energy distribution changes in the CSD and impulse, but really had a hard time deciding which was better. The stored energy just wouldn't go away. From a driver design point of view, if the surround had good loss capability and the mechanical characteristics were designed to allow maximum transfer of stored energy into the surround instead of reflecting back, then there will be much less problem.
"out of room" experience
I've experienced this with my Infinity RS II's in years past, and even more so with my linesource homebrew BG dipoles.
I called it "jump factor", when you, or someone sitting nearby, suddenly has a startle moment, feeling a "presence" in an adjacent area or room, especially noticeable on solo piano, acoustic guitar, or well recorded female vocals.
I've noticed with my current setup (large BG RD 75's), with many recordings, I can walk from my listening position (~12 feet away) towards the speakers, and the image remains stable in the room, much like it does at a live venue.... to the point where, standing between the speakers sounds as if you're actually amongst the musicians. As I can walk past the setup, between the speakers, the sound retreats as it would walking "through" a stage performance, rather than sounding like "speakers in another room". Kinda unnerving at first. Naive listeners have also noticed this effect unprompted, which leads me to believe I'm not demented.
I believe this has more to do with dipole radiation and good transient response than anything else.
On another plane, here are some Visaton "NoBox" inspired OB's just finished for my daughters' boyfriend...
John L.
I've experienced this with my Infinity RS II's in years past, and even more so with my linesource homebrew BG dipoles.
I called it "jump factor", when you, or someone sitting nearby, suddenly has a startle moment, feeling a "presence" in an adjacent area or room, especially noticeable on solo piano, acoustic guitar, or well recorded female vocals.
I've noticed with my current setup (large BG RD 75's), with many recordings, I can walk from my listening position (~12 feet away) towards the speakers, and the image remains stable in the room, much like it does at a live venue.... to the point where, standing between the speakers sounds as if you're actually amongst the musicians. As I can walk past the setup, between the speakers, the sound retreats as it would walking "through" a stage performance, rather than sounding like "speakers in another room". Kinda unnerving at first. Naive listeners have also noticed this effect unprompted, which leads me to believe I'm not demented.
I believe this has more to do with dipole radiation and good transient response than anything else.
On another plane, here are some Visaton "NoBox" inspired OB's just finished for my daughters' boyfriend...
John L.
Re: Very interesting thread
In the many drivers I've measured, each and every one was minimum phase, even horribly abused and modified old Minimus 7 woofers, even old Magneplanar panels. The only noted deviation from min phase was when a driver was damaged, and rubbed.
I just remeasured my Tang Band W4657 gongs and processed them on the weekend; poorly terminated cone (huge dip above 1 khz) and terrible 6.6kHz peak, but still perfectly min phase.
We all certainly work from our own set of expectations, and sensitivities. I've been very successful in blind detecting metal drivers in commercial systems, over many systems blind (grills on, in first listen). Metal woofers I can usually eventually attune to, short term, but not tweeters. Not to be confused with excess sibilance, which in fact many speakers under represent; metal tweeters are just hard sounding to my ear.
I can appreciate AJ's perspective, as I listen to a great deal of jazz and symphonic. What I find, for me, is that any inherent advantage metal drivers show with close miced jazz, and they do show those advantages in swing and excitement, is overwhelmed by the disadvantages with symphonic, but also with stoner rock and hardcore/heavy metal, instilling in these last forms far too much "white noise" and obscuring the subtle interplay of the feedback on the guitar.
The root cause doesn’t point to just the distortion amplification of the hard material; distorted rock should be largely immune to much of this. This music is very revealing. The impact I hear is a temporal homogenizing of the distorted guitar, losing the subtle timing cues in the feedback (which can be quite beautiful: Spaceman 3, Kyuss etc.).
dlr said:
In the many drivers I've measured, each and every one was minimum phase, even horribly abused and modified old Minimus 7 woofers, even old Magneplanar panels. The only noted deviation from min phase was when a driver was damaged, and rubbed.
I just remeasured my Tang Band W4657 gongs and processed them on the weekend; poorly terminated cone (huge dip above 1 khz) and terrible 6.6kHz peak, but still perfectly min phase.
We all certainly work from our own set of expectations, and sensitivities. I've been very successful in blind detecting metal drivers in commercial systems, over many systems blind (grills on, in first listen). Metal woofers I can usually eventually attune to, short term, but not tweeters. Not to be confused with excess sibilance, which in fact many speakers under represent; metal tweeters are just hard sounding to my ear.
I can appreciate AJ's perspective, as I listen to a great deal of jazz and symphonic. What I find, for me, is that any inherent advantage metal drivers show with close miced jazz, and they do show those advantages in swing and excitement, is overwhelmed by the disadvantages with symphonic, but also with stoner rock and hardcore/heavy metal, instilling in these last forms far too much "white noise" and obscuring the subtle interplay of the feedback on the guitar.
The root cause doesn’t point to just the distortion amplification of the hard material; distorted rock should be largely immune to much of this. This music is very revealing. The impact I hear is a temporal homogenizing of the distorted guitar, losing the subtle timing cues in the feedback (which can be quite beautiful: Spaceman 3, Kyuss etc.).