Then you also do not agree with Toole and Olive and a whole range of others.
Please investigate it, but do so in a scientific manner such that the results can be confirmed and believed.
But take it from me, these kind of studies are exceedingly hard to do, which is why they aren't done. What is done instead are simple studies where large assumptions are made and biases enter from all sides. Then false conclusions are drawn. This is why audio, unlike say video is in such "dark ages". Video is highly quantified and well understood, but massive amounts of money were spent to get to that level.
I am not sure that I understand, but yes, I believe that if the original is distorted to the point that the listener is offended then further distortion is unlikely to be perceived. We actually saw this in our first perception tests where at very high levels of nonlinearity peoples ratings diverged (in the sense of stability) drastically. It's as if once something gets "bad" people find it very difficult to rate it - "How bad is bad!" When we dropped the very high nonlinearities from our data set the correlation went way up.
Am I such an incompetent researcher that I would have made such an abominable error - I don't think so.
Last edited:
Electronic music may have an emphasis on LF content though without nearly as high levels in the mids/highs.
At low frequencies 120dB is not very extreme, and especially not below 20Hz (again, some electronic music, movie sound). There's a huge difference between listen at high levels in the lows compared to mids. OSHA, IEC etc. uses A-weighting when setting the exposure limits based on the effect of frequency vs amplitude in our hearing apparatus (love that word). Same with IEC EN-50332 for headphones.
This seems somewhat humorous and nihilistic, but its not actually correct I think.
1) there is always a level below which NLD is not an issue. This level will depend a great deal on the specific design. It is possible to create a system with no objectionable NLD for any usable SPL level in a small room.
2) I don't understand point 2, but I would say that other than levels beyond which a certain speaker was designed to attain, it is the linear distortions that dominate the perception.
3)I don't know that there is that wide of disagreement, but yes, there is some. Horns have progressed substantially over the last couple of decades and the newer ones make the older ones sound horrible. These trends are widely accepted as are the reasons for them. Its been the understanding of them that has led to this situation. You can't improve that which you do not understand.
I can agree on everything in this post, but would add that IMO even the tests performed by the persons named have been biased. I have not seen, yet, any group listening test that I could agree with, regarding its methodology and/or technical realization. Unbiased and completely correct listening test with a larger group of people is impossible, IMO.
This discussion is starting to make me uncomfortably confused.. 🙂
Saying that nonlinear distortion is not a major issue in audio is like saying knifes and guns is not dangerous.. as long as you stay away from them or wear a bullet proof west.
Saying that nonlinear distortion is not a major issue in audio is like saying knifes and guns is not dangerous.. as long as you stay away from them or wear a bullet proof west.
Yeah, that is excessive. 112dB at the mix position is often called for on riders. Of course that's average and includes a lot of bass.
When I heard the driver breakup, I seriously doubt it was anywhere near 120dB peak. Maybe 105 peak. Loud, but not crazy. It was certainly sooner than I expected a 1" CD to get raspy. No one else seemed to be bothered by it though, so there you go. 😉
I still wonder if the shallow profile of the waveguide has something to do with it. The round waveguides I've gotten from Parts Express where similar, tho they did not have the same throat shape nor a block of foam in them. That ought to count for something.
I find this almost humorous. I am not sure that such tests are "impossible", but I would place a large bet on there not ever being such tests done.
Regarding bias in my tests, if there was any bias, and I tried to remove all such aspects in the design, it would have been for finding something actionable in NLD. The null result that we obtained in the CD paper was hardly what the client was looking for! It would have been far better for us if we had found some significance in the test.
I'm just telling you what I found after searching for decades. Go ahead and search for yourself. Good luck with that!
Last edited:
Well I measured it at 120 dB, so maybe it was just a lot cleaner than you were expecting and so you under-estimated the level.
A fair bit of research exists on the audibility of distortion.
Agreeing with the caveat that sine wave tests are absolute worst case, I personally don't advocate harming frequency response for moderate improvements in distortion, but I think sine-based targets still set a useful absolute limit and IMO something we should strive to achieve, even if linear distortion (frequency response) is far more important to get right
Following are from my personal notes, some of this from internet postings pre-web (so Olive's position may have evolved). Hope this is useful.
Dave
--------------------------------------------
Olson, H. F. Elements of Acoustic Engineering, D. Van
Nostrand Co. In., New York, 2nd edit. 1947.
Cut-off freqency 3750 5000 7500 10000 15000 Hz
_Objectionable distortion_
Music Triode 14.0 8.8 4.8 3.4 2.5 %
Pentode 10.8 6.0 4.0 2.8 2.0 %
Speech Triode 14.4 10.8 6.8 5.6 4.4 %
Pentode 12.8 8.8 6.4 4.4 3.0 %
_Tolerable distortion_
Music Triode 6.8 5.6 4.4 3.4 1.8 %
Pentode 5.6 4.0 3.2 2.3 1.35%
Speech Triode 8.8 7.2 4.8 3.6 2.8 %
Pentode 8.8 5.2 4.0 3.0 1.9 %
_Perceptable distortion_
Music Triode 1.2 --- 0.95 --- 0.75%
Pentode 1.1 --- 0.95 --- 0.7 %
Speech Triode 1.4 --- 1.15 --- 0.9 %
Pentode 1.5 --- 1.2 --- 0.9 %
------------------------------------
From: (SeanOlive)
Subject: Re: What is THD??
Date: 16 Feb 1995 13:10:11 -0500
Someone has already challenged my above statement on the audibility of THD
so let me clarify the numbers & my reasoning.
The audibility of distortion depends on several factors which include
(from Cabot):
1) Characteristics of the nonideality in the equipment under consideration
2) instantaneous level of reproduced sound
3) type & complexity of signal being used
4) spatial effects in reproduced sound field
5) characteristics (spatial & background noise) of listening room
6) linear & nonlinear distortion present in the rest of the equipment &
program material
7) listeners' ability, experience, training, etc.
All of these factors influence audibility, and the discrepancies in
thresholds quoted from different studies are largely due to differences
in how these variables were controlled. As Cabot points out in (1), most
of the studies are flawed in some way due to lack of control in these 7
variables, and others. The number I quoted was very ballpark and might
represent what an untrained person might hear under typical conditions. I
admit it is not very scientific & I apologize.
The best study -according to Cabot- on THD is by Bryan and Parbook
(1960). Using a 357 Hz sine wave they plotted thresholds of individual
harmonics as a function of the fundamental's level. Their thresholds were
much lower than what is generally expected, although Cabot points out
that unlike other studies, they kept the distortion of their system below
their measured thresholds. Their lowest threshold measured was for the
fourth harmonic: .05% at a listening level of 70 dB. The thresholds of
the harmonics closely follow the equal-loudness contours.
In practical listening situations, it would be difficult to duplicate this
result because the variables listed above could not be controlled to the
same extent as experimental conditions. That is why I said 1% OR LESS
would be audible, and that the thresholds would much higher on music. I
apologize for any misunderstanding. Hopefully, this additional
information will exonerate my temporary lapse from being totally
scientific. Thanks for keeping me honest !
The reference is given below, and I highly recommend it for interesting
reading on this subject.
Reference:
(1) Richard Cabot, "Audible Effects Vs. Objective Measurements in the
Electrical Signal Path", Proceedings from the AES 8th International
Conference,Washington, DC, (May 3-6 1990)
-----------------------------------
From: J. Robert Stuart (Meridian Audio), Digital Audio for the Future, Audio, 3/98 pp 30-37. (see Attached)
A harmonic's audiblity depends on the frequency and amplitude of the fundamental, as well as the percentage of distortion. The ear is more sensitive to harmonics of fundamentals at medium frequencies (1-2 kHz) and medium levels (80 dB). In the plot shown just above, a second harmonic smaller or equal to a given percentage is audible if it is inside the corresponding contour curve of constant distortion (i.e., 0.1%)
---------------------------------------
From "Hearing, Its Psychology and Physiology" by the
Acoustical Society of America (my fave reference) on the
audibility of 2nd harmonic distortion with a pure 370
Hz fundamental:
"The masking of an added harmonic is negligible below
a sensation level of 40 - 50 dB. From 50 to 80 dB,
the amount of harmonic necessary for an audible change
increases rapidly, first in absolute magnitude, and later
in relative magnitude as well...."
"...The qualitative character of the audible change produced
by adding this harmonic was different at the various sensation
levels of the fundamental. At low levels the harmonic was usually
heard as a separate tone. In the middle region [50 - 80 dBSPL]
it was heard as a sharpening or brightening of the timbre of the
tone, whereas at high levels the changes were so complex and so
dependent upon differences of phase that any generalization about
their character would be misleading."
Most sensitive threshold for the 2nd harmonic was at 60 dBSPL, 0.3%. Two listeners were tested.
------------------------------------------
I took the next attachment from a paper in my collection, "Subwoofer Performance for Accurate reproduction of Music" (JAES, June 1988) by Fielder and Benjamin. They studied the literature and ran their own distortion audibility tests. The study was done explicitly in the context of audio reproduction and they even looked at the distortion performance of some subwoofers to baseline their tests with some semblance of reality. Their general recommendations, to ensure inaudibility in the most sensitive cases are:
- second harmonic should be less than 3%
- third harmonic less than 1%
- higher harmonics less than 0.3 to 0.1 %
- room responses had a significant impact on the acceptable limits for a given situation
- as expected, acceptable limits were level dependent, and higher levels tolerated higher distortion.
- distortion audibility was phase dependent. Shown are the most sensitive results.
I also scanned in their tabulated results, attached. These things are always situational (listener dependent etc) but these numbers can probably be used with decent confidence
------------------------------------------------
Distortion audibility generally follows the equal loudness contours. This was shown by Bryan and Parbook (1960).
I used to believe that odd order harmonic distortion was the most audible. A few years ago I took a test of different distortion products, 2nd through 5th. To me, the 4th was the most obnoxious, qualitatively, sounding very rough. It was worse than 5th. Interestingly, the Bryan/Parbook study found that the 4th harmonic product had the most sensitive detection threshold at 70 dB, and could be reliably detected at 0.05% (-66 dB). This was with tones over a very clean system, and headphones. In a real room with music, the threshold would probably be higher.
------------------------------------------
Agreeing with the caveat that sine wave tests are absolute worst case, I personally don't advocate harming frequency response for moderate improvements in distortion, but I think sine-based targets still set a useful absolute limit and IMO something we should strive to achieve, even if linear distortion (frequency response) is far more important to get right
Following are from my personal notes, some of this from internet postings pre-web (so Olive's position may have evolved). Hope this is useful.
Dave
--------------------------------------------
Olson, H. F. Elements of Acoustic Engineering, D. Van
Nostrand Co. In., New York, 2nd edit. 1947.
Cut-off freqency 3750 5000 7500 10000 15000 Hz
_Objectionable distortion_
Music Triode 14.0 8.8 4.8 3.4 2.5 %
Pentode 10.8 6.0 4.0 2.8 2.0 %
Speech Triode 14.4 10.8 6.8 5.6 4.4 %
Pentode 12.8 8.8 6.4 4.4 3.0 %
_Tolerable distortion_
Music Triode 6.8 5.6 4.4 3.4 1.8 %
Pentode 5.6 4.0 3.2 2.3 1.35%
Speech Triode 8.8 7.2 4.8 3.6 2.8 %
Pentode 8.8 5.2 4.0 3.0 1.9 %
_Perceptable distortion_
Music Triode 1.2 --- 0.95 --- 0.75%
Pentode 1.1 --- 0.95 --- 0.7 %
Speech Triode 1.4 --- 1.15 --- 0.9 %
Pentode 1.5 --- 1.2 --- 0.9 %
------------------------------------
From: (SeanOlive)
Subject: Re: What is THD??
Date: 16 Feb 1995 13:10:11 -0500
Someone has already challenged my above statement on the audibility of THD
so let me clarify the numbers & my reasoning.
The audibility of distortion depends on several factors which include
(from Cabot):
1) Characteristics of the nonideality in the equipment under consideration
2) instantaneous level of reproduced sound
3) type & complexity of signal being used
4) spatial effects in reproduced sound field
5) characteristics (spatial & background noise) of listening room
6) linear & nonlinear distortion present in the rest of the equipment &
program material
7) listeners' ability, experience, training, etc.
All of these factors influence audibility, and the discrepancies in
thresholds quoted from different studies are largely due to differences
in how these variables were controlled. As Cabot points out in (1), most
of the studies are flawed in some way due to lack of control in these 7
variables, and others. The number I quoted was very ballpark and might
represent what an untrained person might hear under typical conditions. I
admit it is not very scientific & I apologize.
The best study -according to Cabot- on THD is by Bryan and Parbook
(1960). Using a 357 Hz sine wave they plotted thresholds of individual
harmonics as a function of the fundamental's level. Their thresholds were
much lower than what is generally expected, although Cabot points out
that unlike other studies, they kept the distortion of their system below
their measured thresholds. Their lowest threshold measured was for the
fourth harmonic: .05% at a listening level of 70 dB. The thresholds of
the harmonics closely follow the equal-loudness contours.
In practical listening situations, it would be difficult to duplicate this
result because the variables listed above could not be controlled to the
same extent as experimental conditions. That is why I said 1% OR LESS
would be audible, and that the thresholds would much higher on music. I
apologize for any misunderstanding. Hopefully, this additional
information will exonerate my temporary lapse from being totally
scientific. Thanks for keeping me honest !
The reference is given below, and I highly recommend it for interesting
reading on this subject.
Reference:
(1) Richard Cabot, "Audible Effects Vs. Objective Measurements in the
Electrical Signal Path", Proceedings from the AES 8th International
Conference,Washington, DC, (May 3-6 1990)
-----------------------------------
From: J. Robert Stuart (Meridian Audio), Digital Audio for the Future, Audio, 3/98 pp 30-37. (see Attached)
A harmonic's audiblity depends on the frequency and amplitude of the fundamental, as well as the percentage of distortion. The ear is more sensitive to harmonics of fundamentals at medium frequencies (1-2 kHz) and medium levels (80 dB). In the plot shown just above, a second harmonic smaller or equal to a given percentage is audible if it is inside the corresponding contour curve of constant distortion (i.e., 0.1%)
---------------------------------------
From "Hearing, Its Psychology and Physiology" by the
Acoustical Society of America (my fave reference) on the
audibility of 2nd harmonic distortion with a pure 370
Hz fundamental:
"The masking of an added harmonic is negligible below
a sensation level of 40 - 50 dB. From 50 to 80 dB,
the amount of harmonic necessary for an audible change
increases rapidly, first in absolute magnitude, and later
in relative magnitude as well...."
"...The qualitative character of the audible change produced
by adding this harmonic was different at the various sensation
levels of the fundamental. At low levels the harmonic was usually
heard as a separate tone. In the middle region [50 - 80 dBSPL]
it was heard as a sharpening or brightening of the timbre of the
tone, whereas at high levels the changes were so complex and so
dependent upon differences of phase that any generalization about
their character would be misleading."
Most sensitive threshold for the 2nd harmonic was at 60 dBSPL, 0.3%. Two listeners were tested.
------------------------------------------
I took the next attachment from a paper in my collection, "Subwoofer Performance for Accurate reproduction of Music" (JAES, June 1988) by Fielder and Benjamin. They studied the literature and ran their own distortion audibility tests. The study was done explicitly in the context of audio reproduction and they even looked at the distortion performance of some subwoofers to baseline their tests with some semblance of reality. Their general recommendations, to ensure inaudibility in the most sensitive cases are:
- second harmonic should be less than 3%
- third harmonic less than 1%
- higher harmonics less than 0.3 to 0.1 %
- room responses had a significant impact on the acceptable limits for a given situation
- as expected, acceptable limits were level dependent, and higher levels tolerated higher distortion.
- distortion audibility was phase dependent. Shown are the most sensitive results.
I also scanned in their tabulated results, attached. These things are always situational (listener dependent etc) but these numbers can probably be used with decent confidence
------------------------------------------------
Distortion audibility generally follows the equal loudness contours. This was shown by Bryan and Parbook (1960).
I used to believe that odd order harmonic distortion was the most audible. A few years ago I took a test of different distortion products, 2nd through 5th. To me, the 4th was the most obnoxious, qualitatively, sounding very rough. It was worse than 5th. Interestingly, the Bryan/Parbook study found that the 4th harmonic product had the most sensitive detection threshold at 70 dB, and could be reliably detected at 0.05% (-66 dB). This was with tones over a very clean system, and headphones. In a real room with music, the threshold would probably be higher.
------------------------------------------
Attachments
That is possible. Or maybe it's just one of the those Catch 22 things of clean systems. The fewer the problems, the more the remaining ones stick out. Sometimes, you just can't win.
Thank you for your interesting post. At least, testing with sine tones may be controlled to a high degree. We are able to measure and define generator distortion, amplifier distortion, we can use headphones with low (and measurable) distortion.
"In a real room with music", everything becomes debatable, the choice of music for test, quality of recording, speakers used, listeners position in the room, room acoustical properties, background noise, etc.
Anyone who might be interested in audibility of harmonic distortion on sine tone, please try these 2 files:
http://pmacura.cz/signal1.wav
http://pmacura.cz/signal2.wav
Spectra are attached. The best way is too use good headphones and lower volume. Foobar ABX is good way in case you like to check your result.
http://pmacura.cz/signal1.wav
http://pmacura.cz/signal2.wav
Spectra are attached. The best way is too use good headphones and lower volume. Foobar ABX is good way in case you like to check your result.
Attachments
thanks DDF
interesting reading. well i think fletcher/munson is the biggest part of the min/max phenomenon i observed.
interesting reading. well i think fletcher/munson is the biggest part of the min/max phenomenon i observed.
The first waveguides I built were Dayton 8" plastic units with conical extensions at 45 degrees then a radius/baffle. I had assumed they were spherical. They weren't that bad TBH, they had an open and clean quality.
ISO 226:2003 is the latest and most up to date equal loudness curves and I think it's important to mention that those should be read as an indication, not absolute values.
ISO curves are based on several studies which naturally have slightly different results, also naturally thresholds differ between test subjects.
Some have worse hearing than the curves suggest, some have better. Also the ears ability to detect sound does not stop abruptly at the high and low range as the curves may suggest.
No, I was just referring to the point made by Dr Geddes about amplifiers :
"So there is an audible difference between amplifiers that measure the same"
There are no 2 different amplifiers that would measure the same. Such case does not exist. It is only a question of measurement methods and resolution to find differences. Even spectra of harmonic distortions measured at all levels from minimum and maximum and at wide range of frequencies will also show differences. Noise spectrum and level will be different as well. Time responses will differ. Speaking about "two amplifiers that measure same" is a myth and oversimplification, in case that amplifiers are not identical.
- Status
- Not open for further replies.