This is an odd one 
At times it seems I can differentiate these and then the difference suddenly eludes me. This was with earbuds.
Even the generator tone doesn't sound particularly pure, possibly a result of non linearity in the tiny drivers although I am not driving them hard.
At times it seems I can differentiate these and then the difference suddenly eludes me. This was with earbuds.
Even the generator tone doesn't sound particularly pure, possibly a result of non linearity in the tiny drivers although I am not driving them hard.
Code:
foo_abx 2.0.4 report
foobar2000 v1.3.16
2017-11-16 08:11:24
File A: 100Hz_gen.wav
SHA1: df5df3f13b1395c129f43889d1bd1e2edf531334
File B: 100Hz_rec.wav
SHA1: 64f3885a1a96b87ead64c7c73ba7c2ba47674726
Output:
DS : Primary Sound Driver
Crossfading: NO
08:11:24 : Test started.
08:11:39 : 01/01
08:11:49 : 02/02
08:11:59 : 03/03
08:12:12 : 04/04
08:12:33 : 05/05
08:12:47 : 05/06
08:12:59 : 06/07
08:13:12 : 07/08
08:13:12 : Test finished.
----------
Total: 7/8
Probability that you were guessing: 3.5%
-- signature --
412136bb9a08326ab8b6dde62e727bb5d2724418
Wow
Didn't the question of a replay chain with higher intrinsic distortion crop up in another of these threads ? Someone found it easier to differentiate files when the replay chain wasn't top flight performance.
I have just tried again with Sony MDR-V7 headphones that have much better bass response and scored just 2/8.
Didn't the question of a replay chain with higher intrinsic distortion crop up in another of these threads ? Someone found it easier to differentiate files when the replay chain wasn't top flight performance.
I have just tried again with Sony MDR-V7 headphones that have much better bass response and scored just 2/8.
I managed to more 6/8 's in succession but couldn't do any better this time around. Again with the earbuds. Haven't tried speakers tbh.
Code:
foo_abx 2.0.4 report
foobar2000 v1.3.16
2017-11-16 12:14:48
File A: 100Hz_gen.wav
SHA1: df5df3f13b1395c129f43889d1bd1e2edf531334
File B: 100Hz_rec.wav
SHA1: 64f3885a1a96b87ead64c7c73ba7c2ba47674726
Output:
DS : Primary Sound Driver
Crossfading: NO
12:14:48 : Test started.
12:15:15 : 01/01
12:15:25 : 01/02
12:15:34 : 02/03
12:15:46 : 03/04
12:15:56 : 04/05
12:16:07 : 04/06
12:16:17 : 05/07
12:16:27 : 06/08
12:16:27 : Test finished.
----------
Total: 6/8
Probability that you were guessing: 14.5%
-- signature --
77c3e3afaa482c1b034f66c472e2d1183fe891e0
Code:
foo_abx 2.0.4 report
foobar2000 v1.3.16
2017-11-16 12:22:17
File A: 100Hz_gen.wav
SHA1: df5df3f13b1395c129f43889d1bd1e2edf531334
File B: 100Hz_rec.wav
SHA1: 64f3885a1a96b87ead64c7c73ba7c2ba47674726
Output:
DS : Primary Sound Driver
Crossfading: NO
12:22:17 : Test started.
12:22:57 : 00/01
12:23:09 : 01/02
12:23:24 : 02/03
12:23:36 : 03/04
12:23:48 : 04/05
12:23:57 : 04/06
12:24:09 : 05/07
12:24:19 : 06/08
12:24:19 : Test finished.
----------
Total: 6/8
Probability that you were guessing: 14.5%
-- signature --
4da8e3ba55ed8b06dfcc6d8c9cd20b1c88931d1fsorry if this seems a distraction to the test but i was trying to wrap my head around monotonic functions and from what i can see or glean with my mathematically deficient mind i see as a function that roughly describes the operation of all loudspeakers.
if this description is wrong please enlighten me.
the first portion of the function shows the inability to accurately reproduce the frequency with the given input power, not "getting off the peg" the mass and mechanical resistance are not yet overcome, sometimes referred to as "stiction", then we arrive at the "flat middle" portion of the function which represents the linear range where the ability to reproduce the frequency of interest is unaffected or changed by input power until we arrive at the other extreme at what most would describe as the onset of distortion due to thermal and mechanical limits being reached.
problem i have is i can't make sense of how this function would translate if power and frequency are varying in proportion?
if this description is wrong please enlighten me.
the first portion of the function shows the inability to accurately reproduce the frequency with the given input power, not "getting off the peg" the mass and mechanical resistance are not yet overcome, sometimes referred to as "stiction", then we arrive at the "flat middle" portion of the function which represents the linear range where the ability to reproduce the frequency of interest is unaffected or changed by input power until we arrive at the other extreme at what most would describe as the onset of distortion due to thermal and mechanical limits being reached.
problem i have is i can't make sense of how this function would translate if power and frequency are varying in proportion?
I add this as it seems appropriate...
Im engaged in comparison testing between 4 ribbon diaphragm designs at the same time this thred is running. I too seem to have experiences similar to Mooly"s. One moment I clearly hear differences. The next I do not.
I find that when I cannot discern differences well I have to shut it down and go do something else for a bit, then come back. Heck sometimes even getting up to goto the bathroom seems to reset the brain or whatever the issue is.
I have experienced this phenom many times in past development work of this kind. If it wasn't for the consistency in the perceptions when I can hear the differences I would be left to assume it all in my head and I cannot trust any of it Ha.
Im engaged in comparison testing between 4 ribbon diaphragm designs at the same time this thred is running. I too seem to have experiences similar to Mooly"s. One moment I clearly hear differences. The next I do not.
I find that when I cannot discern differences well I have to shut it down and go do something else for a bit, then come back. Heck sometimes even getting up to goto the bathroom seems to reset the brain or whatever the issue is.
I have experienced this phenom many times in past development work of this kind. If it wasn't for the consistency in the perceptions when I can hear the differences I would be left to assume it all in my head and I cannot trust any of it Ha.
Last edited:
I think hat you guys are missing the whole point. It is not that a loudspeakers nonlinear distortion is never audible, that is ridiculous, it's that it can be made to be insignificant. Hence, do that.
Further one cannot relate the audibility to a metric as one does not exist that works well. THD and IMD do not work as metric because they do not correlate to perception. So we cannot compare graphs of THD or IMD and say that one is "better" than the other.
Further, our tests used musical signals not pure tones. A steady pure tone distortion is going to be many many times more audible than real musical signals.
Further one cannot relate the audibility to a metric as one does not exist that works well. THD and IMD do not work as metric because they do not correlate to perception. So we cannot compare graphs of THD or IMD and say that one is "better" than the other.
Further, our tests used musical signals not pure tones. A steady pure tone distortion is going to be many many times more audible than real musical signals.
Last edited:
The exact opposite would be true. A complex signal will mask the distortion. The simpler the signal the more audible nonlinear distortion will be. We tried to use a median signal, just a few instruments. For orchestral recordings nonlinear distortion would need to be very high for it to be audible.
that's what i call the "puppy under the blanket"
just because the distortion is masked it does not "cease" to exist.
the puppy is the "distortion"( in this instance i would prefer a term like "sonic signature" to describe the lumped parameters) the blanket is the "masking" if you will.
changing the properties of the blanket can make detection of the "puppy" easier or harder but does nothing to the puppy or the fact that it under the blanket
(no animals where harmed in the making of this post)
just because the distortion is masked it does not "cease" to exist.
the puppy is the "distortion"( in this instance i would prefer a term like "sonic signature" to describe the lumped parameters) the blanket is the "masking" if you will.
changing the properties of the blanket can make detection of the "puppy" easier or harder but does nothing to the puppy or the fact that it under the blanket
(no animals where harmed in the making of this post)
Last edited:
1) well, I do not think we are missing the whole point. The point would be if you brought the proof that the speaker distortion was insignificant. It is almost sure less significant than linear issues, but not proven to be insignificant.
2) at least I am not interested neither in THD or IMD NUMBER, nor in Gm GedLee metric number. On the other hand, I am interested in distortion spectra at various amplitudes and frequencies.
I know your AES publications and newer presentations and I have reservations to accuracy of nonlinearity definitions that you are providing. You are speaking about 21 examples of nonlinearities that you have tested with audience, but none of them is exactly described neither by maths nor by detailed plots, so it is impossible to verify what you were testing. From THD, IMD and Gm numbers we can see nothing. We can only see 4 oversimplified plots of nonlinearities that are said to be an example of what was tested.
3) we are not testing only steady pure tones here. We have a music sample here as well
Audibility of speaker nonlinear distortion - test
with original file and the mathematically added nonlinearity, that was based on midrange measurements. Yes the nonlinearity is not frequency dependent, which is a pitty, but I am not able to do it.
Last but not least, speaker nonlinearity may create quite ugly intermodulation products and AM products. This is well shown by W. Klippel, the leader in exploration of speaker distortions of all kind.
There is no "proof" that it is insignificant, just as there is no proof that it is. But most loudspeaker designers that I know of, certainly the entire JBL staff accept what I am saying is true. Believe what you want.
There was no point and no room to have published the "exact" details of what was done. You will either have to accept - or reject - that a PhD physicist can accurately define such things. I would be happy to get you more detail if that helps.
Hence you are criticizing me for doing exactly the same thing that you did; simplifying a complex problem down to something simpler in order to be able to actually perform the tests.
You can't just throw out our results because you don't like them. You have to "prove" them wrong, not just claim suspicion. That you clearly are suspect of THD and IMD means that you got the point of the papers. The extrapolation of those papers and my experience with designing speakers is where I claim my "insiginificance of nonlinearity" in speakers comes from. There is no audible nonlinear distortion in my speakers, so I have to assume that it can be done. That it is not always the case is also surely true as well.
Last edited:
I am not throwing out your results of your tests, I am providing my tests and the results we can see are the results of the participants here. It is a web based, files based ABX test and as such it has its limits, as every test methodology, on the other hand it is available to anyone and the ABX protocol can be verified, provided we do not suspect that everything is cheated.
I am only saying that you have not proven that speaker nonlinear distortion is insignificant and inaudible on music. You have not brought any valid proof of this claim.
And your tests do not disprove my position either. There is no proof either way. But I will "hang" with the authorities like Toole and Olive - they don't have any proof either, but we all believe the same thing. It's based on our experience and the fact that valid tests "indicate" that we are right.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.