Observations of a controlled Cable Test - Page 9

No, a "control group" is something different than "controls".

A control is, among other things, a trial known to be audible at a given level of test sensitivity. You should include them at several levels, to measure how accurate your subject is.

In some tests, a negative control is also included.

Been meaning to ask, why is Signal To Noise Ratio irrelevant?
Bob, W7ETA
A novice

This is a cable test. It's very hard to have meaningfull controls. The null hypothesis is that two somewhat similiar cables are audibly indistinguishable. What controls would you propose? Let's say you include controls and your subject doesn't identify them, is he "guessing" or does he have some hearing loss? What did you prove? Why would one even want to test subjects that say they don't hear a difference, when you can't even show that people who claim to hear a difference can do so?

I think it is nonsense to test people who deny having a trait to see if they are wrong. It makes far more sense to select individuals may have the trait and test them.

Nonsense. Got any #20 zip cord?

And with the #20 zip cord, you have something that is audibly distinguishable.

You've shown the test sensitivity wasn't that high. Next question?

When you are responsible and include both positive and negative controls, you test the test. If you're tired of arguing about how good tests work, it seems to me that might be a good idea, yes?

And the way to do that is to provide some positive control so you can tell who those people are. So what's your gripe?

The answer to that is over in the LP vs. CD knockdown thread. Don't have the cite offhand, but you can peruse http://www.aes.org/sections/pnw/ppt.htm for the ppt's on the loudness tutorial, fundamentals of hearing tutorial, and perceptual coding tutorial, and you'll get a good start.

How do you know?

With a long enough length, the resistance becomes a factor, no?

Because it's been documented in a DBT.

Goodness. It is long enough and small enough you get a perceptable different in quality due to loudness mismatch.

That sounds good, but for two problems.

One, I don't recall ever establishing the length---hence the question. With lengths short enough I'll bet even #20 would not be audibly different. Some people place their amps rather close to their speakers.

Two, a test of this nature usually calls for a level match. Indeed, I called for one to take place before this test was conducted, and ChrisWiggles said they did one. Thus the bulk volume differences won't occur. There was some argument over whether level differences ought to be removed, but it's simply noncontroversial (as you know) that level differences are audible. So if we're trying to judge differences with sonic quality, we have to level match.

Honestly, I appreciate where you're trying to go, but I think you're as guilty of some of the same oversimplification you're accusing others of.

OK.
Let's presume that 20 gauge is audibly distinguishable.
Yet . . .
some humans are deaf.
and others . . . have above average hearing . . .

thus . . .
it's a variable, not a control
So, now we have a multivariate rather than univariate analysis.
How did this simplify things?

Well, if the person in question isn't going to hear that difference, he's likely SOL with any other claims.

You are assuming the 20 gauge is a screening test. What is being proposed is adding a 20 gauge wire arm to an A vs B trial (as a presumed control). It's a different experimental design, and changes the statistical analysis in a major way.

So far as I know, no blind test has ever shown audible differences in cables when level matched. And I think it's safe to say everyone agrees that differences due to level are totally uninteresting (most hifi systems come equipped with a volume control, after all).

It's clear that with a ridiculous enough cable the differences would be audible even with some kind of level matching since there is a change in the frequency response from LCR in the cable. It would be interesting to find out how ridiculous it needs to be.

Right. As the cable's impedance starts to approach that of the speaker itself, it does alter the frequency response of the system---frequencies where the speaker has low impedance will be attenuated more than frequencies where the speaker has high impedance.

Obviously there is some point where the differences will be audible, but one cannot be as confident as jj_0001 would like to be.

With a known control, you can test the TEST.

And thereby show a given level of validity for a given test.

Please stay on subject.

Excuse me, does "proven in a DBT" mean anything beyond "proven in a dbt"?

The much-reviled Noisaine test proved that small speaker cable had audible effects. Ergo, it does. It's been demonstrated. Therefore you have to accept the fact.

Either you accept the real data or you don't. Which will it be?

If one is not confident, one rejects actual data.

First, all differences are at their heart differences in level, when considered in the germane domain, that of "partial loudness", which is to say the same thing that Greenwood and Zwislocki demonstrated maps to the firings of neurons in the auditory nerve. Now, you mean "broadband differences in level" perhaps, which are uninteresting UNLESS THEY HAVE ANOTHER POINT, which is that THEY TEST THE TEST. Oops, please stay in context, will you, we're testing the test, that's the context right now.

So, now, tell me, does the fact that different cables CAN cause a level difference (ohms law here, you don't even need to enter the realm of complex impedence to show this compared to known human loudness DL's) show that you can use a too-small cable to show that the test is in fact responding to small audible differences?

Yes, it does show this, and has, definitively, in a very well reported (if somewhat less than wonderfully described thanks to editorial contributions) test. It is shown. It is. Fiat est. It takes one test, reports, and repeats to show an affirmative, and that is far exceeded in this case. So either you can accept the fact (and yes, it is a pure level difference, which manifests to the human as a change in quality, which shows that the test, in fact, does have reasonable sensitivity) or you can reject the actual data.

So, you can use such a strategy to test the test. Perhaps you are trying to avoid testing the test, I don't know, but the facts are absolute in this case, it's been proven. What's more, the results are right directly in line with the known sensitivities of humans to small level changes.

So, what's your problem?

As to "is this a reasonable speaker cable" as far as I know, nobody technical thinks so, although perhaps you ought to check out the effective gauge of some of those interesting cables.

It changes nothing.

It is not a "presumed control" 20 feet of #20 wire using pretty normal speakers is known to cause audible level differences. It IS a control.

It is no MORE than a control, and you do not have to change your statistical analysis of the actual probe situation in any fashion whatsoever.

Of course, you do have an interesting situation if your probe situation shows a positive, and your control a negative. Such is life.

There's no oversimplification here. Go back to the Noisaine test. It details the smallest wire used, how it was in fact detected by listeners, and exactly what the setup was.

So you have a control demonstrated right then and there. There's nothing hypothetical or anything else involved.

They may have not included it as a control. That is irrelevant.

It's very simple - either the cables make a difference in frequency response that is audible after level matching (say at 1kHz) or they don't. No need to discuss partial loudness or neurons.


If you want a control of that sort, using a cable is a very bad idea. It would be better to adjust the level a little using the preamp and determine the threshold for hearing volume differences.


See above. There are much better ways to do that.


Link please.

A control can not vary, it must always be poitive or always be negative..
An example is: "testing a micro-organisms for sensitivity to a drug". You soak small disks of paper in an antibiotic. You but them on a petri dish. As a control you place small disks of paper soaked in the same solution as other disks minus the antibiotic. Thus, you have reduced the number of variables, not increased them.

I think you need to study standard psychoacoustic testing. You're making an authoritive statement based on another discipline. Furthermore, you're asserting that the positive control for the listening test adds variables, when in fact it obviously, clearly reduces them.

Look I'm happy to accept the fact that cotton twine sounds different than copper cable, too. It's all about degrees.
Are you saying they might not have tested the test? Oh my goodness!

Look, Mike L. and Chris didn't need to test the test for one simple reason: the test conditions were part of the hypothesis. Mike L. was confident he could discern the differences between the exact two cables chosen, under the exact conditions selected. He maintained this confidence throughout the test itself, until it was terminated prematurely due to negative results.

So while it might not provide the results you were looking for, it provided precisely the confirmation that I and many objectivists would hope for. That is, it confirmed that Mike L.'s brain fools him.

Now, as a followup to this test, Mike might want to do what you are recommending, and introduce a test setup that virtually anyone ought to discern.

You're avoiding the context of this discussion. First, level matching is NOT included in this control arrangement, because it produces, deliberately, a small, audible change in loudness.

So your insistance on modifying the control is simply making the control no longer a control.

As to partial loudness, it is entirely germane, because it shows the connection between an overall change in loudness and the way that any change in audible results from another cable must be expressed. It completes the thought, and as such is a necessary, material statement.

Please do not confuse the issue, please remain in context, which is a cable with a specified size and length showing an audible difference, and thus being a valid control. That is the only issue at hand.

Nobody doubts that you should make level-matched tests, UNLESS, say, you're trying to find out (yes, there are other ways to do so, which is irrelevant, although useful) if the cable creates an audible level change.

If so, you're compensating out the issue you're trying to test. That's simply a mistake.

And, again, you ignore the context. It is literally perfect in this regard, as it is something that the listener here, IN CONTEXT, could accomplish.

Certainly there are other methods, methods that require calibration, specialied equipment, etc, that were NOT AVAILABLE. What is available to anyone within reach of Home Depot is zip cord that's too small.

You seem determined to reject the obvious, utter, absolute superiority of using a too small cable IN A SITUATION WHERE NOTHING ELSE IS AVAILABLE.

If you want to loan the OP a lab, a listening room, and equipment, then maybe you and he can arrange to do a better job.

Until then, please STAY IN CONTEXT, and help, rather than raising methods and measurements NOT AVAILBLE TO THE PERSON WHO RAN THIS TEST.

Goodness!

Really? The context here is:

Somebody's living room.
The ability to change cables.

There are no meters of sufficient resolution. There is no calibrated attenuator.
So, in other words, your claimed methods are IMPOSSIBLE in the context we are discussing.

In short, you're not helping a bit, here, you're talking pie in the sky.

Yes, if I was doing this in my lab, I'd use different methods. Those methods and systems are not available in this case, and insisting there are better methods to the fellow who started this thread, who doesn't have the equipment, is a waste of time.

I am not going to go digging for a link to the Noisaine test, it was published before the net existed.

And I'm not going to find it and scan it, and violate somebody's copyright. If you can't find it, ask Tom, he's certainly going to know where you can find it, and he's hardly difficult to find.

Then provide a reference.

JJ, at risk of drawing some snark, I am seriously confused here. Why are we compensating out the issue we're trying to test? I don't see that at all. We're not trying to test the notion that bulk level differences are audible. That's a given. But such differences alone have nothing to do with sonic quality; because once the comparative test is done, the volume level will be adjusted based on the selected cable alone.

Yes, I understand that. Now what's your point? My point is that his test would be more secure to the audio fantasy brigade if he could show that it had at least a decent sensitivity. Not that that will help the argument much, but it's a start.
Now that is quite enough.

I require that you retract your claim of "the results you were looking for". I am a professional scientist, I do audio research for a living, and I have been reviled so far by two generations of SUBJECTIVISTS for pointing out that their emperor has no clothes. Your suggestion that I "hope for" anything is simply wrong and insulting. I am an objectivist who has been insisting, demanding, arguing for, etc, DBT's, for at least 30 years, and I am full well on the *public* record as arguing for such since at least 1979, in AES papers, on AES committees (I chair two AES Technical Committees), in IEEE publications, in IEEE committtes, while reviewing papers for both AES and IEEE, on netnews, at the (groan) audio asylum, here, and elsewhere. I have contributed to both BS1116 and BS1116.1 protocols, I was one of the designers of the MPEG-Audio triple-stimulus testing method, and so on.

How dare you call me a subjectivist. That is literally fighting words!

I am not the least bit suprised by the outcome of the test, it's what I would expect, unless the cables contained a network, were too small, or some kind of pathological speaker, say like a Scintilla, or other speaker with an absurd impedence curve, was involved (which can provoke all sorts of problems both with any change in cable size and cause lots of amplifier misconduct as well). Such evidence, as far as I'm concerned, only shows that there is a problem with the impedence curve of the product. Speakers do not need to dip to ~1 ohm reactive impedence inside the audio band (or anywhere else).

My point here is simply that he could have tested the test, as well. That way (duh) when an audiophile claims (yet again) that "the test was bad" we could show conclusively that it WAS NOT BAD.

My goodness. I suspect you have little experience (I have 20+ years) in arguing with the fringe audiophiles, otherwise you'd expect them to raise that issue within a minute. And, you will note, they already have argued, above, that due to this or that, the test was invalid. A positive control rejects this premise. Hello?

We still see arguments about "stress", I haven't seen anyone raise the invalid objection involving the "range rule", but I expect we may. Audiophiles are fond of claiming that 'DBT's don't work for music" despite the reams and reams of evidence otherwise, and when some of those reams of publications are forced down their throats, they retreat to the special pleading of "but we're more sensitive" and the moving goalposts of "but I meant for high-end audio", when pointed out that Floyd Toole, Sean Olive, and Harman (Floyd has retired, btw) use them for high-quality mainstream audio testing, and have built a very nice facility for exactly that.

Of course, there is a simple reason that DBT's don't work for many high-end tweaks, it's because THE TWEAKS CAUSE NO AUDIBLE DIFFERENCE.

D'oh.

If red cable makes someone happy (for example), so be it, but please, let's not have the "but it makes the audio waaarrmmmeeerrr" protest. Accept it as part of the sensory gestalt, and be done with it.

It's not a question of audibility.

You know, cut the crap, JJ. You have been purposefully provocative and hostile from the start. I know damn well you're an objectivist. But if you can't take even a little snark than quit dishing it out. Your opinions are welcome but your hostility is not. Know who your allies are on this thread, and work with them, or get the hell out.

No.

If you have even entered this discussion, you should have heard of this test. It is not a "new" or "special" thing, it's well known, and I'm not going to a library and look through old copies of audio magazines for you.

I do believe it's been mentioned up-thread. You can look or not, as you like, but the evidence is plain and cold. Tom Noisaine is one of the authors. That's all I recall of the citation, and I don't care to help out somebody who is rude and insulting, and furthermore clearly hasn't read the literature themselves.

What's more, it matches the results one would expect from Allens and Hall's (two separate papers on the same issue) work on level roving, etc. You can find those under "Joseph L. Hall" and "Jont Allen" in JASA, if you are determined to find something on the issue. You may find it a bit tricky to relate, but that's not my problem, and I'm out of patience with repeated demands for well known, scientifically validated, issues.

In short, there's nothing unexpected, use a cable that's too small, and it causes an audible level difference. I have no idea why you find this so surprising.

What a load of crap. JJ, this is not an academic conference, this is an enthusiast forum. It is not your place to impose any sort of prerequisites for having this discussion. We all here to learn, except perhaps for you. Indeed, we objectivists can even learn from each other, if it is possible to cut through the hostility. Just today on another thread I indicated that Krabapple's efforts to point us to some useful references has caused me to modify my position (indeed, to further cement my objectivism.)

I'll say it again: know who your allies are on this thread, and work with them, or get the hell out. I don't care how much you know if you're going to insist on being an a$$hole.