Proposed Test: Cable Differences - Page 2

That gave me a lot to think about. 2 x 19K is only 38K, which means 44.1 would be enough! Maybe it is the 16 levels of quantization that hurt redbook; we should try 24/44.1. Of course we've also been discussing the published claim that no one can hear the difference between 16/44.1 and 24/96 ... let's save that for another day.

The speaker cable debate bores me -- it is analog interconnects where the action is IMHO. And digital cables, especially the widespread (and IMO wrong) opinion around here that cheap digital cables are OK since "the bits get there". (Actually they don't, at least not at the right moment in time anyway).

The "bits are bits" statement reminds me of the famous book "sharks don't get cancer". 'Cause sharks do get cancer. I love flat, bold statements that are just wrong. Very clever way to win arguments. "We will eliminate our depenence on middle eastern oil in ten years".

But let's have a good Labor Day weekend ... come back and re-join the hi-res and digital cable fight later.

Kind of metallic.

You mean 2 ^ 16 . But that isnt really the issue here. It does not mattter whether you use 16 bits or 24 bits the pattern of relative energy levels will be exactly the same from 20 - 20K viz they do drop off markedly. Perhaps this is why the incessant focus on 20K + begins to seem a bit like misapplied energy.

I did a search and couldn't find anything.

Why is it nobody recables Electrostats? Recabling dynamics is almost standard procedure, but you never see a recabled stax. Why is that?

I assume it's because finding the plugs/terminations is hard?

nick, is your avatar from Hear no Evil See no Evil?

/ot

Man I did not write clearly, I see that. Yep, by 16 quantization I meant 2^16 steps, of course. I really do understand binary, I promise.

But my real question was not about the high freq drop off ... you explained that. Here's what I meant:

1. Assume (some people disagree) that 24/96 sounds better than 16/44.1.

2. Assume that the increase from 96 to 44.1 effects only the Nyquist frequency, raising the high-end threshold.

3. But if there's no energy up there, then that increase will not lead to audible improvements.

4. Therefore the only benefit would be the 24-quantization vs the 16-quantization (in that part of the spectrum that HAS energy!).

5. Hence that's the real advantage of 24/96, in which case 24/44.1 if it existed would sound as good.

I am not sure #2 is correct -- that was my question.

It would be great if someone could set up an epic cable DBT.

I think that the differences in cables is very subtle, and it might be difficult to discern in a few minutes -- this is where double-blind testing loses its validity. However, over a period of time, at least a couple of days, those differences would become apparent.

1. We can call this a testable hypothesis rather than an assumption

2. Changing the sampling rate to 96K raises the bandwidth from < 22050 to < 48000 - there are other pragmatic effects of raising the sampling rate but to keep it simple for now the major effect in terms of audio quality is to raise the bandwidth i.e the highest frequencies that can be captured goes from slightly under 22050 to slightly under 48000.

3. There is energy above 22050 but its relative importance is under question and there is far less of it compared to the 20hz - 5000hz range. Tests on the effect of filters suggest that filters above 16k , certainly 18k have little if any audible effect - so raising the sampling rate just to get higher frequencies in play may not be have audible benefits.

4. To pursue this hypothesis you first have to provide non anecdotal evidence that hypothesis 1 can be suppported otherwise 4 is conjecture.

5. ???????????????

Testable under what conditions?

Hear a difference vs. prefer a difference?

Pls. explain.

Reliably (Probability of guessing < 5%) detect a difference under blind testing.

Create a setup that allows both 24/96 and 16/44.1 to be played back under conditions that mean that the 16//44.1 and 24/96 are presented in random order and the subject (and experimenter) does not know which they are listening to and must judge whch they are listening to.

The listener only has to prefer high resolution to redbook the majority of the times he expresses a preference (i.e., more than can be explained by chance, and it does not have to be as high as 95%, there is no basis for that high a signficance level as I have explained in other posts). He does not have to judge which he is listening to, that is not part of the proper protocol IMO. We ask only: do you prefer A to B, B to A, or no preference?

The listener does not have to take an X sample and say it is A or B -- ABX testing is just wrong for this type of thing (I know I am flying in the face of what most other people say here, but I stand my ground, and have credentials in statistics).

It matters only that -- when A and B are randomized and blind to me, and the person conducting the experiment cannot influence me (either because he has been blinded too, or more simply because he is not in the room) -- that I prefer high res to redbook rather often.

We can easily compute the threshold for a given number of tests; I would use a nice moderate alpha level -- actually I would use lots of test, so that the conclusion is not in doubt statistically.

If out of every 10 trials I express no difference 9 times, and prefer hi res to redbook one time, and I can always do this, in every 10 trials -- I repeat this for 50 sets of 10 trials, say, then guess what -- I am using hi res. 'Cause 10% of my listening will be more enjoyable, and that's good enough for me.

Nick -- that's what I meant by Nyquist. I totally understand folding frequencies. And yes, this was about a testable hypothesis. We are miscommunicating, and I'll take the blame for not writing clearly, but it's not worth any more back and forth ... I was trying to separate the two aspects of the difference between 24/96 and 16/44.1 -- following from your (important and worthwhile) observation that there is little energy above the Nyquist freq for 44.1. I was just looking for your opinion of the 24 bit vs 16 bit quatization, and whether you felt that alone could lead (some) listeners to prefer 24/96, that's all.

As to whether there is an audible difference in the first place, I agree that remains to be tested well. I do not consider the famous published study to be valid. It used ABX testing (which I consider incorrect for the hypotheses being tested, and there are papers in food sensory testing journals that agree with me btw; see previous post for a general discussion), and did not focus enough on the few listeners with 8 or 7 scores for further testing.

Some day this will be tested correctly. Until then the jury is out. At least IMO.

Preference vs difference and why it matters

So where do you place the threshold 80% , 75% , 55% , if you want 75% or less you really do need a big sample to rule out random guessing. Some of the old AES filter tests used /20 as a yardstick and saw 15/20 as good enough, though nobody managed that - 8/10 is too weak especially with really big numbers of subjects where you would expect some to get it by random guessing - in the same way that if you have 100 variables you will get significant correlations between some of them by chance.


To express a preference you must perceive a difference, so you are saying that being able to tell a difference 1/10 times is sufficient ?

I think the idea is to grab the 8/10 guy, and keep testing. That will smoke out whether he is the expected chance lucky guesser, or the real deal. I would even do that with the 7/10 guys.

Since we have no love for the null hypothesis, I am indeed suggesting that we can set the threshold low, say 60%, but really only to pick people to continue with. Test them further, lots and lots and lots.

And yes, if I express no difference 900 out of 1000 times, but I prefer hi-res to redbook 100 out of 1000 times, and never redbook to hi-res ever ... then that says something very meaningful: it is a small difference, easy to miss, but real. That's why I want to use A vs B, with A and B assigned randomly, and sometimes the same, and sometimes with ringers ... and ask the listener only to pick one of the four each time:

1. I hear no difference
2. I hear a difference but have no preference
3. I prefer A to B
4. I prefer B to A

This is very different, and much better, than A/B/X testing. Boy, will we learn a lot when we make both A and B the same (without telling the listener, of course), or make A 24/96 and B a 64 KB MP3. All the while telling the listener we are randomly assigning one type of signal to A, and another to B.

We don't even tell him what we are testing! Could be resolution, could be cables, could be CD players ... he doesn't see and doesn't know!! And he thinks that the two things we are randomly assigning to A and B stay the same. Ha!

This is testing. And the analysis is very revealing. One subject at a time.