Testing audiophile claims and myths

[castleofargh]

I've come across stuff that took years later to be measured or resolved that was audible. Sometimes the test gear needs to catch up.

In this case it is the chassis material that makes the difference, but steady state signal tests do not show it up? I've seen that countless times. Not in a DAP, but in devices with magnetics, like toroidal transformers and output inductors placed next to steel or other ferrous material chassis. It does not affect power, distortion and only in grossly incompetent designs: noise or crosstalk. However it is distinctly audible. Not many find this out as it doesn't occur to them to make the chassis out of 2 or three different material to see what happens , or take other mitigating measures, as the conventional measurements do not show any issues. Once you've heard it, it is something you know to avoid.

The issue I believe is we do not have a reliable way of measuring non steady states in the analogue domain. In an ideal world we could put dynamic signal, even music in the test system and get a readout of the error. In digital we can subtract one from the other and look at the difference in ideal circumstances. However in analogue things like the phase shifts cause by even the out of band frequency limits dominate any subtraction a surprising amount. Attempts to compensate for them are extremely complex and usually result in failure. I have seen a few try this with varying results. Brunel University was demonstrating a prototype in the 1980's, but nothing seemed to come of it. I have not seen any AES papers etc. A major hi-fi company tried twice, once getting very involved by brick-wall filtering the bass and treble out to examine the mid-band subtraction FFT. I cannot recall much other than they found some interesting thing, but it was generally a failure. If some one has this working, they will likely not share it as they could have a significant advantage over the competition. Like the early laser infrarometers Wharfedale and B&W had in the 70s & 80s. But once Klipple came along, everyone can measure this stuff.

So my previous experience does not discount the fact that a chassis made an audible change, and that change did not result conventional measurements finding the effect.

Sits awaiting the flame war...

Right, I should have stuck to that particular case and not generalize to any 2 DAPs. Even I have been troubled by phase or even overall time shifts between gears making inaudible differences look huge in a null test, so we end up with the reversed issue of measured difference we can’t easily judge in term of audibility. But would that be a problem here where you can just swap the casing on the same DAP, record a song twice and subtract them?

 

[bigshot]

Responding to Gregorio's post...

Most audible differences that can't be measured haven't even been proven conclusively to be audible differences yet. I think that is much more common than pathological cases as you call them. Audiophiles tend to look to the unlikely exceptions to the rule when they haven't even disproven the most likely reason (bias) yet. Then they justify the uncontrolled bias by pointing to the effect of psychology on perception and questioning the validity of controlled listening tests in general.

 

[Dogmatrix]

Would not at least in the case of recorded sound everything audible be measurable since the act of recording is measuring sound in the recording space . So anything not measurable is infidelity

 

[SoundAndMotion]

Would not at least in the case of recorded sound everything audible be measurable since the act of recording is measuring sound in the recording space.

There’s an appealing logic to that. But it’s not true… at least if “measurable” means the measurements are doable. If it means it’s possible, even with currently unknown methods, then the logic stands.

Think of 2 examples that we do know about. (I don’t know of examples we don’t know about – duh!)

Think of the frequency response. Although analog circuits exist in spectrum analyzers, currently we typically digitize the time-domain voltage and then perform a calculation (Fourier transform - FT). If we didn’t know how to do a FT, we would have a recording of the signal, but no frequency response.

Similar with TIM (transient intermodulation distortion), which was unknown before… somewhere around the 70’s. You could have recorded it, but without specifically looking for it, not known of its existence.

 

[Dogmatrix]

I am thinking along the lines of microphones and the subsequent equipment being measuring devices
But I could comprehend of things like beat frequencies happening as the result of reproduction and therefore not strictly speaking present in the recording/measurement

 

[gregorio]

Even I have been troubled by phase or even overall time shifts between gears making inaudible differences look huge in a null test, so we end up with the reversed issue of measured difference we can’t easily judge in term of audibility.

Yes but as you say, that’s the opposite of the claim of audible differences that can’t be measured. BTW, have you tried DeltaWave? It seems very good at auto alignment when nulling two different recordings. Down to an accuracy of 1/1000th of a sample apparently.

Most audible differences that can't be measured haven't even been proven conclusively to be audible differences yet.

I’m talking about audible differences and there are no “audible differences that can’t be measured”. Audible differences that are inaudible are obviously not “audible differences”. What you’re talking about is entirely different, the misidentification of a psychological difference as an audible difference.

There’s an appealing logic to that. But it’s not true… at least if “measurable” means the measurements are doable.

Logically then, surely the opposite must be true; Dogmatrix’s statement IS true if the measurements are doable. So let’s look at your two examples:

If we didn’t know how to do a FT, we would have a recording of the signal, but no frequency response.

Firstly, obviously that’s a hypothetical, we do know how to do a FT, it certainly is “doable”. Secondly, I don’t claim to have sufficient math ability to fully understand Shannon’s proof of the sampling theorem but doesn’t it rely (at least indirectly) on Fourier? In other words, if we didn’t know how to do a FT, we wouldn’t have the sampling theorem, digital audio wouldn’t exist and therefore we would not “have a [digital] recording of the signal”?

Similar with TIM (transient intermodulation distortion), which was unknown before… somewhere around the 70’s. You could have recorded it, but without specifically looking for it, not known of its existence.

As I understand it, TIM shouldn’t be an issue with any reasonably competent design and arises in response to a test signal that probably never exists in music recordings. In other words, it’s existence wasn’t known because it probably didn’t exist (in real world music/sound recording and reproduction)! It was discovered by accident about 50 years ago when someone cross wired an amp. I assume you are referring to the fact that TIM is not revealed in a typical THD measurement and before the 1970’s TIM was not specifically measured. However, although the TIM test/signal was not done before the 1970’s it was certainly “doable” and, a null test would reveal TIM (if present). This brings us back to: What is published as specs and what is typically measured as opposed to what can be measured. Obviously, if something isn’t typically measured, that does NOT mean it cannot be measured.

So, as both your examples were/are in fact “doable”, then Dogmatrix’s assertion was true!

I could comprehend of things like beat frequencies happening as the result of reproduction and therefore not strictly speaking present in the recording/measurement

The beat frequencies are in the recording, except in the case of false/perceptual error “beats”, such as “binaural beats”. In which case, it’s not in the recording or the result of reproduction, it’s all in the mind of the listener.

G

 

[SoundAndMotion]

So, as both your examples were/are in fact “doable”, then Dogmatrix’s assertion was true!

But it’s not true… at least if “measurable” means the measurements are doable. If it means it’s possible, even with currently unknown methods, then the logic stands.

Think of 2 examples that we do know about. (I don’t know of examples we don’t know about – duh!)

Would not at least in the case of recorded sound everything audible be measurable since the act of recording is measuring sound in the recording space .

(I added bold)
God, you love to argue!

I said "if" to say not everything is "doable", because if we don't know how to do it, we can't do it. (duh!) I said "I don't know of examples" because "we don't know about" them. Then I gave 2 examples that we "do" know about, to show that "if" we didn't know about them, we couldn't do the measurements, because we don't know.... oh, come on!

Is that clear yet? No math involved. (BTW, yes Shannon depends directly on Fourier). Without Shannon, digital audio wouldn't exist in its present form, but it certainly can exist without Shannon (perhaps less efficiently, or perhaps with empirically derived methods that are as efficient).

Dogmatrix's original sweeping statement can be made correct, either by stating the assumption that even if we don't know how yet, we eventually will, or by being less sweeping.

 

[gregorio]

I said "if" to say not everything is "doable", because if we don't know how to do it, we can't do it. (duh!)

You seem to have missed my point, that everything is effectively doable, for example:

Then I gave 2 examples that we "do" know about, to show that "if" we didn't know about them, we couldn't do the measurements, because we don't know.... oh, come on!

The example you gave of TIM disproves your assertion. We didn’t know about TIM before the 1970’s, we didn’t know the test signals to use to generate TIM and we didn’t know there was a form of IMD that a standard THD measurement wouldn’t reveal BUT STILL we would have been able to reveal/measure it with a Null Test (which had been around for many decades before TIM was discovered)!

A null test will reveal any and all (“everything”) differences between 2 audio signals, regardless of whether we know what’s causing it or whether or not we have a way of measuring it in isolation.

Without Shannon, digital audio wouldn't exit in its present form …

Dogmatrix and I were talking about the digital audio that does exist, not some hypothetical digital audio that does not exist.

Dogmatrix's original sweeping statement can be made correct …

Dogmatrix’s original sweeping statement is already correct, although it only covers our real world. To make it correct, does he/we really have to add the caveat: “This assertion may not be true for intelligent life from other planets or in parallel universes where digital audio may have a different form”?

G

EDIT: Incidentally, without Shannon, digital audio almost certainly would exist in it’s present form. The proof was published independently by others and in some cases prior to Shannon (Raabe, Kotelnikov, Whitteker, Gabor, Ogura and arguably the first, Borel in 1897). The difference is where it was published and it’s influence on the communication industry.

 

[SoundAndMotion]

You seem to have missed my point, that everything is effectively doable, for example:

You missed my point: if you don't know how to do something, you can't do it. No, everything is not effectively doable.

God, you love to argue!

I learned a few years ago when I went round for round with you for many cycles (on Shannon, in fact), that you don't try to counter with facts and logic, you use attrition. I gave up in that exchange when you basically said (I'm paraphrasing from memory, but I can dig it up): a perfect signal destroyed by noise is still perfect; it's just in the noise.

I can shred your post (e.g. using a null test), but why bother, you'll just come back with more inane stuff you don't understand. Your comment about life on other planets and parallel universes was so stupid, it's offensive. BTW, Dogmatrix mentions recording, not digital audio.

I'll give you this: I used "Shannon" as a placeholder for the Sampling Theorem (which you also often do, e.g. yesterday). So you are right that we're both technically wrong when we do so, because others came up with the same proofs separate from Shannon. Should we both stop, or since you know exactly what I meant, can you use facts, not nit-pick-to-death.

So without the Sampling Theorem, digital audio could still exist (contrary to what you often say), but would be in a different form.

 

[bigshot]

Points are being missed left and right around here!

 

[bigshot]

So you are right that we're both technically wrong.

That should be the motto for this group!

 

[gregorio]

if you don't know how to do something, you can't do it.

you don't try to counter with facts and logic, you use attrition.

Isn’t the bottom quote ironic and hypocritical? Your top quote is just a repeat of the same mantra which is not a “counter” with facts/logic. Your own example, we didn’t know how to measure TIM but still we would have been able to reveal/detect it, with a null test. Of course, one can argue that if we know how to do a null test then we do “know how to do something” and your top quote is obviously correct but then your claim that we didn’t know how to detect/measure TIM was false.

You claimed that Dogmatrix was wrong but the two examples you gave to support your assertion were incorrect. TIM would have been detectable and (paraphrasing) “what if the sampling theorem didn’t exist” is inapplicable because it does exist. So apart from attrition, do you have any examples which actually support your assertion.

I gave up in that exchange when you basically said (I'm paraphrasing from memory, but I can dig it up): a perfect signal destroyed by noise is still perfect; it's just in the noise.

We can revisit that if you wish, although you’d have to define “destroyed”.

I can shred your post (e.g. using a null test), but why bother …

Then why bother having a sound science subforum in the first place and why do you bother posting anything in it? Please give a valid example of an audible difference between 2 audio files that would not be picked up by a null test.

BTW, Dogmatrix mentions recording, not digital audio.

True but then what percentage of recordings do not involve digital audio?

Your comment about life on other planets and parallel universes was so stupid, it's offensive.

But then if you give an example that would only exist in a parallel universe or on another planet, how else should one state the caveat? And, isn’t giving such an example to claim someone else is wrong stupid and offensive to start with?

So without the Sampling Theorem, digital audio could still exist (contrary to what you often say), but would be in a different form.

Even if Shannon and everyone else who can be credited for the Sampling Theorem had never lived, then others would have probably discovered essentially the same “form”. However, both your and my opinion on this is just speculation and irrelevant anyway, because on this planet we do have the Sampling Theorem.

G

 

[bigshot]

Whew! The last paragraph had to do with sound. I thought I’d get all the way to the end with no sound.

 

[gregorio]

The last paragraph had to do with sound. I thought I’d get all the way to the end with no sound.

True, much of it was to do with audio rather than sound; TIM, null tests, sampling theory, digital audio, etc. Are we not allowed to discuss those things?

G

 

[bigshot]

It looked like just general argumentativeness to me, but I admit, I didn’t read it carefully. We’re in the part of the cycle where it’s more fire than facts.