the key word here is AUDIBLE distortion. and that would obviously depend on the listener.
for example the Dutch & Dutch's distortion can go up to 10% at 100Hz at 96 dB, nevertheless it's regarded as one of the best speakers in the world.
-
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!
A chat with Dr. Floyd Toole
- Thread starter hardisj
- Start date
the key word here is AUDIBLE distortion. and that would obviously depend on the listener.
for example the Dutch & Dutch's distortion can go up to 10% at 100Hz at 96 dB, nevertheless it's regarded as one of the best speakers in the world.
It is trivial to get a speaker to distort to audible levels. I do it all the time. 
Even superbly engineered active speakers distort the moment you push them beyond their amplification limit. Ditto for drivers bottoming out with deep bass. These are not areas that are researched. Getting listeners and subjecting them to very loud SPL violates many ethical rules and runs foul of employee liability laws. Testing for speaker preference is always done at recommended levels (in the 80s SPL) so doesn't represent meaningful distortion levels that I am describing.
In many cases, distortion in sub-base reaches and exceeds 100% as it gets even louder than the tone signal itself! This, we can see in measurements and almost assuredly predicts audibility.
What Dr. Toole is talking about is that when it comes to preference, distortion doesn't come into equation in the face of tonality differences. This is very true specially at the levels tested.
Even superbly engineered active speakers distort the moment you push them beyond their amplification limit. Ditto for drivers bottoming out with deep bass. These are not areas that are researched. Getting listeners and subjecting them to very loud SPL violates many ethical rules and runs foul of employee liability laws. Testing for speaker preference is always done at recommended levels (in the 80s SPL) so doesn't represent meaningful distortion levels that I am describing.In many cases, distortion in sub-base reaches and exceeds 100% as it gets even louder than the tone signal itself! This, we can see in measurements and almost assuredly predicts audibility.
What Dr. Toole is talking about is that when it comes to preference, distortion doesn't come into equation in the face of tonality differences. This is very true specially at the levels tested.
Adding on, I am able to pass double blind test of very small impairments such as lossy codecs at high rates. It is inconceivable that mechanical distortion is not audible in a speaker at very elevated levels compared to those. The issue is that we don't have a switch we can throw to change distortion and nothing else in identical speaker. So accurate testing has not been done. Research has always focused on two different products with different tonality and distortion, and showing higher distortion is not correlated with preference whereas frequency response is. This answers a different question than "does distortion matter in a speaker."
audiofooled
Addicted to Fun and Learning
- Joined
- Apr 1, 2021
- Messages
- 533
- Likes
- 594
This is just wisdom and it sometimes takes a lifetime of acquiring knowledge to be able to remain humble before a notion that there's still a lot you can learn. But of course he knows, he just said that to emphasize the absence of resonances as being the first criteria, rather than directivity, and to further illustrate all other variables he goes on to elaborate. Variables are many and no one can know for sure if just one idea is always a good one. Anyway, thank you @hardisj for sharing this interview, it was so nice to hear from Dr Toole again. What a wonderful man.
Maybe Danny Richie can explain what benefits his crossover upgrades provides
Not to pick on him, Miniscus offers a bunch of options for crossover components in their kits. Would be great to see what actual differences there are.
If tolerances are tighter (2.2uF capacitor is more precisely 2.2uF), then that’s at least something.
If longevity or temperature handling is better, then that’s something as well.
If distortion is lower, that would be amazing to see.
You remind me that I have one of his upgraded crossovers to still test!
Oh boy! I can't wait.
To me, that is the mark of true intelligence and something I strive to do in life. I would also say is the mark of a true scientist.
is acoustically generated harmonic distortion (by introducing it into the signal) different than mechanically generated harmonic distortion?
I have also heard that it’s possible to use DSP to phase cancel distortion, and there are rumors that it has been used in soundbars to assist their small drivers.
I recorded the audio and transcribed it using the speech recognition library Vosk. It doesn't tell which passages were from Erin and which were from Dr. Toole. I also haven't made any corrections (for it is far from perfect as you'd expect from some brainless AI tool).
Adding various amounts of harmonic distortion and/or IMD for blind testing shouldn't be that hard, can't we just simply add a few tones as we see fit and see if somebody hears it?
@amirm made a couple of very good points, with which I agree.
Dr. Toole was actually saying two distinct things:
One, measurements of non-linear distortion do not do a good job of predicting audibly significant distortion.
Two, with "mainstream" speakers generally, distortion is not an issue.
To be completely honest I am not very comfortable with either of these statements. Measurements of distortion are motivated primarily by the desire to determine how loudly the speaker will play before distortion reaches the level that it matters. With any speaker, there is a volume level where distortion will be so bad that it obviously matters. As such, his second statement could reasonably be interpreted as a statement to the effect that all mainstream speakers can play as loudly as they need to play. Perhaps this is true, but when interpreted in this reasonable way, it is apparent that the statement is sort of silly. Distortion is definitely an issue if a speaker is cranked up higher than the SPL it can handle, and measurements of non-linear distortion do a pretty decent job of finding where this threshold occurs for a given speaker.
It is almost certainly true that with the great majority of speakers, the anomalies in the frequency response are much more audible than the non-linear distortion. It is for this reason that Dr. Toole and many others have made comments similar to the comments he made. This is understandable, and it is appropriate to say something along the lines of what he said, but in saying it, I would try to be careful not to suggest that speakers don't distort badly if you turn them up too loud, or to suggest that measurements aren't useful for finding the threshold where this occurs.
Dr. Toole was actually saying two distinct things:
One, measurements of non-linear distortion do not do a good job of predicting audibly significant distortion.
Two, with "mainstream" speakers generally, distortion is not an issue.
To be completely honest I am not very comfortable with either of these statements. Measurements of distortion are motivated primarily by the desire to determine how loudly the speaker will play before distortion reaches the level that it matters. With any speaker, there is a volume level where distortion will be so bad that it obviously matters. As such, his second statement could reasonably be interpreted as a statement to the effect that all mainstream speakers can play as loudly as they need to play. Perhaps this is true, but when interpreted in this reasonable way, it is apparent that the statement is sort of silly. Distortion is definitely an issue if a speaker is cranked up higher than the SPL it can handle, and measurements of non-linear distortion do a pretty decent job of finding where this threshold occurs for a given speaker.
It is almost certainly true that with the great majority of speakers, the anomalies in the frequency response are much more audible than the non-linear distortion. It is for this reason that Dr. Toole and many others have made comments similar to the comments he made. This is understandable, and it is appropriate to say something along the lines of what he said, but in saying it, I would try to be careful not to suggest that speakers don't distort badly if you turn them up too loud, or to suggest that measurements aren't useful for finding the threshold where this occurs.
OP
- Thread Starter
- #34
Or, Erin could just turn on auto captions, and I believe you can export the transcript.
Or, you can just watch the damn video without complaining about it.
Re. distortion: context is very important.
OP
- Thread Starter
- #36
amen
I do find it interesting/comforting that two of the leading researchers in psychoacoustics essentially say “meh” about distortion. Especially since Geddes spent so many years researching it and coming up with a very high correlating metric for it. (Geddes’ chat is below for this who missed it)
Over the years I have found that compression is more important and tends to be what I hear when I hear problems that aren’t otherwise linked to. Take, for example, the Klipsch Fives powered speaker with a built in limiter that I have reviewed. It has relatively medium distortion in the LMF. But it also has a lot of limiting. 2dB or more even in the lower midrange to Midbass. What did I hear? The limiting, by far. And this was well before I had a clue that such a limiter was built in.
IOW, compression is far more important to me than distortion. Namely because it directly shows up in frequency response as a non-linear linearity. We all know the impact of FR. And that jives with what Geddes told me when we had our chat.
that said, I will continue to provide HD results since it’s so easy to generate as well as IMD/Multitone for transducers in hopes that one day it can be useful in some way. But it isn’t anything I feel is a necessity as I do about compression/limiting.
Chat with Dr. Earl Geddes of GedLee Audio
Last edited:
- Joined
- May 21, 2019
- Messages
- 4,036
- Likes
- 6,827
In the research I've seen for very small transducers in cellphones, tablets, etc., they know that the transducers will be pushed beyond what they can handle. So an attempt was made to create a clipping profile that was least objectionable. Given how good small devices sound, with a surprising amount of bass, it's fair to say this was successful. This is despite them distorting like crazy. So it's clear that the full story is not at all understood.
I don't have the paper on hand but I and another member discussed it in the early days of Amir's testing.
- Joined
- May 21, 2019
- Messages
- 4,036
- Likes
- 6,827
IIRC during one of his talks @j_j (pretty sure it was him, apologies if not) shared a story about setting up a PA for a party in his youth, and the crowd was asking to turn up the volume despite the system being at its limits. So he inserted a germanium diode into the circuit to increase the effective distortion. Although the output power didn't change, the spectrum did, due to nonlinearity, and the crowd was satisfied.
From my perspective, from reading and thinking about what I've heard myself, nonlinear distortion seems to have its primary effect on perceived loudness, with secondary effects on pitch and roughness when gross.
Geddes' comments on group delay were interesting but the only good research I know of is by Blauert from the late 70s, which gives audibility thresholds at 2 cycles or so, with some variation across the audible range. How this becomes an increase in perceived loudness I don't know.
From my perspective, from reading and thinking about what I've heard myself, nonlinear distortion seems to have its primary effect on perceived loudness, with secondary effects on pitch and roughness when gross.
Geddes' comments on group delay were interesting but the only good research I know of is by Blauert from the late 70s, which gives audibility thresholds at 2 cycles or so, with some variation across the audible range. How this becomes an increase in perceived loudness I don't know.
They are not different if they are measurably indistinguishable, which is to say, if the balance of the individual harmonics is the same with both. It is conceivable that the balance of the individual harmonics would be different between the two sources such that Amir's ability to hear one does not translate directly to the ability to hear the other. I doubt if this is the case though. Amir likely has thought about this and likely has good insight into this particular question.
The other question you raised is interesting. It occurs to me that signal processing could be used to nullify distortion, if the distortion characteristics of the drivers are very will understood. But only up to a limit, and the limit is probably fairly tight. The problem is that the idea suggests that it would be possible to get unlimited volume from an arbitrarily small speaker by nullifying the distortion. Intuition is sufficient to realize that this is not possible, and from this it follows that the ability to do this kind of thing successfully would be limited to a very small increase in the volume level that can be achieved before the distortion becomes intolerable. The question is only what that small increase in volume would be exactly. 3 dB? 6 dB? When you consider how much greater the excursion of a speaker diaphragm has to be, for a doubling of power (+3 db), I think that the improvement that is inherently possible is probably not greater than 3 dB.
What is potentially realistic and potentially done in some sound bars is the use of DSP to implement different phase shifts for the individual drivers, to compensate for the varying distances from the listener to the individual drivers. Said differently, by varying the phase for the individual drivers, it may be possible to make the forward-radiating beam spread out wider. I wrote "may be possible" because I'm not at all certain this is possible. If you think about phased-array radar, differential phase shifts applied individually to the individual radiating elements makes it possible to steer the beam. But the narrow directivity of the beam is due inherently to the overall dimension of the array compared to the wavelength, and also due to the nature of electromagnetic radiation. There are speakers nowadays that use multiple tweeters in an array, the implication being that it is possible to control the beam width by controlling the phase offsets among the individual drivers. I have decided whether I believe this is believable. Even the ability to steer the beam in a phased-array radar is possible because there is just one wavelength. For a given, specific wavelength, it is not difficult to see how relative phase of different radiating elements could be used to steer the beam, and if this can be done to steer the beam, maybe it is possible to steer the beam to the right from the elements on the right and to the left from the elements on the left. Maybe. And maybe it is possible to do this sort of thing even for wavelengths covering a range of an order of magnitude or greater.
Years ago "hi-fi" was 1% or less distortion. I think distortion correlates reasonably well between measured and audible, but the audibly objectionable level is much higher than most people expect for most signals, especially with real music or movies as the source. We've gotten used to seeing infinitesimal numbers and most folk have not done any sort of listening to see what x% distortion really does sound like. I used to have some short files of single- and two-tone tests wherein I added none, 0.1%, 0.5%, 1%, 3%, 5%, and 10% distortion so people could see when they could pick it out. That was years ago in another forum and the results were consistent with prior research: most could not hear 0.1%, most could hear 3% or more, and the median seemed to be around 1% or so. For actual musical content the percentages for detectability would increase but I did not try that test (did not have a way to process large music files). And of course for deep bass it takes much more distortion, and in fact often enough people liked the added distortion because it made the bass more audible (the second or third harmonic of a very LF signal is much more audible to us -- see equal-loudness curves).
Corruption of frequency response and the image or sound field due to room modes, reflections, bad speakers, etc. seem much more common and audibly objectionable IME/IMO. And of course the ones who use a high-power (often pro with modest SNR) amp to drive highly-sensitive speakers (e.g. compression drivers) and complain about a high hiss level...
Corruption of frequency response and the image or sound field due to room modes, reflections, bad speakers, etc. seem much more common and audibly objectionable IME/IMO. And of course the ones who use a high-power (often pro with modest SNR) amp to drive highly-sensitive speakers (e.g. compression drivers) and complain about a high hiss level...
Last edited:
Similar threads
