So it raises the question in listening tests where more than one frequency is used. Is it the HD or the IMD you hear when you reach the limit. For me it will be IMD 100 out of 100 times.
And if you have HD you have also IMD in that case
And if you have HD you have also IMD in that case
I don't know that either one is a "mechanism" - just that a nonlinear circuit creates these things. Also, there's no such thing as a circuit that creates a single harmonic from nonlinearity. A "curve" input-output response such as a 12AX7 tube may come close in making only H2. but it's just substantially larger than other harmonics - they're there and measurable (and calcuable given the transfer function, though I don't know how offhand, it takes some extra mathematics). Various transfer curves give various amounts of different harmonics. A common example is crossover distortion which makes lots of higher harmonics which are easily audible.
These two different frequencies appear on the OUTPUT of the nonlinear circuit (and as I just said above, they're not the only ones, but that's not important here). If you run that output into the input of another nonlinear circuit, I'd think you'd get the sum and difference of H1 and H2 as well, though depending on the nature of the nonlinearity, perhaps at a much lower level.
I see. You left yourself enough plausible deniability that you can move the goal posts.
Tom
Because the two tones don't both exist at the input.
Let's take an example. Let's say we have a black box with an exponential transfer function, that black box will cause harmonic distortion when presented with a single tone on the input. It will also cause IMD when two or more pure tones are presented at its input. The amount of IMD will be some function of the harmonic distortion of the black box, i.e., higher HD -> higher IMD.
But presented with a pure tone at the input the harmonics present on the output of the black box will not cause IMD as the multiple harmonics aren't present at the input. For the black box to cause IMD from a single tone input the black box would have to have some knowledge of its output before the output happens, which isn't the case. Not even if there's feedback within the black box.
Tom
No, I have talked about thick tailed bell curves for a long time. Nothing new there.
Feedback can be laggy, but a 180° phase shift is often only reached past 1MHz or so. So for audio frequencies it's near intantaneous.
If all the components were ideal, I'd build a power op-amp with tonnes of gain and be done with it. But pretty much every component and groups of components are being modulated in some way.
I think the assumption for the black box example is that there is only one distortion mecanism.
If you have two black boxes in series there will be IMD because of the HD in earlier stage. (Even with i single frequency input to first stage)
So a power amp has several stages in series. And HD in first stage wil generate IMD in last stage.
Oh, sure. You've been talking about those for a long time. I wholeheartedly agree on that. But a categorical dependent variable can't be described by a thick tailed bell curve, so I don't understand why you keep bringing them up.
Tom
Just for the record most of the distortion and noise you hear is not produced by the system per se. Not electronics, not cables, not house AC, although those things are not faultless. Most noise and distortion is produced by the interaction of the Listener‘s mind with the things in the room that are not (rpt not) related at all to the system. OK, what things you ask. Things like cellphones, TVs, watches and clocks and books and CDs and LPs. This can be easily proven by slowly removing those items from the room. Wild, huh?
Last edited:
Because real world data sometimes has a thick tailed distribution. It wouldn't preclude the existence of an "average ear" for the purpose of establishing auditory thresholds.
Moreover, its tail thickness that distinguishes different types of white noise, which is always a matter of concern in audio electronics. Low noise on a FFT may not be so low if its an infrequent loud pop. It can still be white and low level as viewed in the FD.
Last edited:
Tom, I get your point. However, if one wanted to determine if a test subject could hear quantizing distortion at 16-bits, the experimenter might want to determine whether the test subject was guessing or really hearing something. Isn't that usually how it works? So they use ABX DBT, which scrambles the brain of the test subject after a few trials. But the experimenter wants to make really sure, 95% sure, that no guessing is going on. That sort of thing sound familiar? So, I think you know what comes next with more test subjects and how statistics plays into it.
What do you mean by "scrambles the brain of the test subject after a few trials"?
ABX DBT is not scrambling anything. Either the participant can correctly identify A and B or they cannot.
We don't have to turn audio into a science. I still find value in the science of audio, though. Especially when it comes to manufacturers' product claims.
Tom
It is well known that ABX with insufficient user training is biased towards false negatives. Nothing new there, its a well known fact in perceptual testing science, it just means that test subject training to a level of comfort and reliable performance is mandatory. This is something often unknown to amateurs at perceptual testing.
Regarding scrambling the brain, I know of no one other than Howie Hoyt who has been subjected to more ABX DBT than anyone else in the forum, due to his line of work.
This is what he had to say:
..for better or worse I have been dragged into hundreds of A/B, ABX/ Random long selection repeats, etc. and in my experience Mark is right, they show excessive negative results, due to lack of training in hearing the difference, as well as the mental confusion of having things switch up. They do show decent correlation for a large number of people for gross differences. Also, IMHO none of these tests do much to eliminate pre-existing biases towards certain SQ contours (the "I'm used to my own speakers" thing).
The only way I have been able to get reliable results and consensus is by weeks of training with specific exaggerated SQ problems (missing bits, odd noise floor contour, mechanical noises on magnetic recordings, etc.) and then reducing them to near-inaudibility. Done this way people train their brains to identify specific sounds. In a similar way we hired and trained pre-press graphics people by placing cards in front of them with small color variations and asked them to pick out the outlier. Some show an ability immediately, but many can learn by repeated testing, resulting in a crew of artists who could pick out microscopic color variations I personally could not see.
https://www.diyaudio.com/community/threads/the-black-hole.349926/post-7310119
Moreover, what Howie described is pretty much in keeping with what others have described. There are other people's findings I could point you to, but I think Howie sums it up reasonably well.
Regarding scrambling the brain, I know of no one other than Howie Hoyt who has been subjected to more ABX DBT than anyone else in the forum, due to his line of work.
This is what he had to say:
..for better or worse I have been dragged into hundreds of A/B, ABX/ Random long selection repeats, etc. and in my experience Mark is right, they show excessive negative results, due to lack of training in hearing the difference, as well as the mental confusion of having things switch up. They do show decent correlation for a large number of people for gross differences. Also, IMHO none of these tests do much to eliminate pre-existing biases towards certain SQ contours (the "I'm used to my own speakers" thing).
The only way I have been able to get reliable results and consensus is by weeks of training with specific exaggerated SQ problems (missing bits, odd noise floor contour, mechanical noises on magnetic recordings, etc.) and then reducing them to near-inaudibility. Done this way people train their brains to identify specific sounds. In a similar way we hired and trained pre-press graphics people by placing cards in front of them with small color variations and asked them to pick out the outlier. Some show an ability immediately, but many can learn by repeated testing, resulting in a crew of artists who could pick out microscopic color variations I personally could not see.
https://www.diyaudio.com/community/threads/the-black-hole.349926/post-7310119
Moreover, what Howie described is pretty much in keeping with what others have described. There are other people's findings I could point you to, but I think Howie sums it up reasonably well.
Last edited:
ABX is close to useless. You were stressed out because ABX was evaluating you instead of dut.
I would love to have a reference to a peer reviewed study on this. That someone agrees with you on a forum is nice, but not scientifically sound. I'm not knocking Hoyt or anything he has to say. I'm just saying that there is a difference between opinion and science.
If ABX testing is so controversial, why is even recent research using it? Here's an example: Pawlak, Alan; Lee, Hyunkook; 2020; The influence of loudspeaker-listener distance on the detection of low-bitrate audio coding artefacts [PDF]; Applied Psychoacoustics Lab (APL), University of Huddersfield; Paper 576; Available from: https://aes2.org/publications/elibrary-page/?id=20814.
I also searched for "ABX controversy in audio" on various search engines. It appears the only controversy is that audiophiles don't like ABX testing because the results don't jibe with their preconceived opinion of what they can and cannot hear.
I also don't understand what this has to do with thick tailed bell curves and categorical data.
Tom
"There is considerable debate regarding preferred methodologies for high resolution audio perceptual evaluation. Authors have noted that ABX tests have a high cognitive load [11], which might lead to false negatives (Type II errors)."
https://www.researchgate.net/public...f_High_Resolution_Audio_Perceptual_Evaluation
"...forced-choice tests such as ABX tend to favor negative outcomes when differences are small if proper protocols are not used to guard against this problem.
https://en.wikipedia.org/wiki/ABX_test"
There is more out there. Do I have keep looking them up for you, or can you do it yourself?
Anyway, to my understanding it is noncontroversial that ABX tends towards false negatives for test subjects untrained in working with the protocol. Once the test subjects are used to ABX protocol then it tends to compare equally well with other more naturally comfortable-for-the-test-subject protocols. Thus adequate training with the ABX protocol is necessary for good results.
What it has to do with all audio perceptual testing, including for thresholds with thick tailed distributions, is that proper testing is often not done. Engineers thinking they are smart enough to do ABX DBT without consultation with experts in perceptual testing should have their results taken with a grain of salt.
https://www.researchgate.net/public...f_High_Resolution_Audio_Perceptual_Evaluation
"...forced-choice tests such as ABX tend to favor negative outcomes when differences are small if proper protocols are not used to guard against this problem.
https://en.wikipedia.org/wiki/ABX_test"
There is more out there. Do I have keep looking them up for you, or can you do it yourself?
Anyway, to my understanding it is noncontroversial that ABX tends towards false negatives for test subjects untrained in working with the protocol. Once the test subjects are used to ABX protocol then it tends to compare equally well with other more naturally comfortable-for-the-test-subject protocols. Thus adequate training with the ABX protocol is necessary for good results.
What it has to do with all audio perceptual testing, including for thresholds with thick tailed distributions, is that proper testing is often not done. Engineers thinking they are smart enough to do ABX DBT without consultation with experts in perceptual testing should have their results taken with a grain of salt.
Last edited:
Jeez. Take a chill pill would ya?! It's hard to tell fact from fiction when you make general statements such as "some people..." and "it's commonly known..." without providing a reference. Sorry buddy. I'm not willing to take your word as gospel.
I hadn't considered the forced choice aspect of ABX. Thanks for pointing that out. I'm curious for your thoughts on how to address this shortcoming, but we can take that conversation at another time or in a new thread.
Tom