'Audio Lies'

[lineup]

... is that it's a pain to set up
... and do properly,
... and that's not as entertaining

SY

this is excuse number 2, 13 and 41
from the Audiophile listener's handbook

How to Avoid Taking Blindtests

 

[ctviggen]

I try to do blind testing on myself whenever I can. I'll have my fiancee change between two items and have me tell her which is best.

 

[rdf]

No lineup, the problem with a great many double-blind tests is they're structured like the AES-sponsored one described below:

http://stereophile.com/asweseeit/107/

That was the 91 AES Convention in New York. My brother, who at the time was a lead audio designer for Nortel and made a living from structured audio testing, attended a similar demonstration at a late-Eighties AES conference in Toronto with one of his peers. They found the methodology so unprofessional and flawed as to leave mid-point to go harmonica shopping instead.
This discussion would be better served if posts like yours dealt more with the test proof you so adamantly believe exists instead of ad hominem and unprofessional characterizations of personalities. Unless of course you want to present proof it's your field of speciality and show the research.
Frankly I'm dismayed the moderators didn't step in during the last five pages or so.

 

[phase_accurate]

To distract from validity-of-testing-issues I made a SPICE model for the bi-wiring case and it does indeed show differences in nonlinear distortion.

Although they are very small with my model. Furthermore I don't know how much the typical signal-dependant impedance fluctuation of a driver is in reality.

I will post the schematic and FFT as soon as I have time. Maybe in the meantime someone comes up with real-life figures of aforementioned nonlinearity so I can set the parameters of my model accordingly.

Regards

Charles

 

[gamma]

Originally posted by Christer
So, do you mean this test was done in the way I suggested earlier in this thread?

Do you mean this?

Take two identical, brand new pieces of equipment, that are supposed to benefit from burn in. Listen to them and confirm you can't hear any difference. Then burn in one of the pieces, but not the other one. After that, perform a DBT.

honestly: No it wasnt a blind test at all - but "to my defence" I can say I were VERY sceptical to "burn-in-effect" before the test.

 

[jneutron]

To distract from validity-of-testing-issues I made a SPICE model for the bi-wiring case and it does indeed show differences in nonlinear distortion.

Although they are very small with my model. Furthermore I don't know how much the typical signal-dependant impedance fluctuation of a driver is in reality.

I will post the schematic and FFT as soon as I have time. Maybe in the meantime someone comes up with real-life figures of aforementioned nonlinearity so I can set the parameters of my model accordingly.

Regards

Charles

Assume the wire run is #12, the length is ten feet, using two neutrik speakons.

The round trip wire resistance is 2.02 milliohms per foot times 20, or 40.40 milliohms.

The connector has 3 milliohm contact resistance guaranteed, total of 12 milliohms.

Assume the woofer driver impedance has a minimum of 4 ohms, the tweeter 8.

Assume the power required of the woofer is about ten times that of the tweeter. For a 1 amp peak tweeter signal, the woofer will draw up to 4.(close to the square root of 10).

So you will have a cable resistance total of 52 milliohms, and the simulation will look for the difference between one ampere sine across 52 milliohms, and that one ampere riding on a 4 ampere DC current across that 52 milliohms.

I do not have typical binding post resistance values..

Cheers, John

 

[Christer]

Gamma,

Yes, that is what I meant. So OK, you didn't do a blind test. However, the other important thing is if you did listening tests before to convince yourselves that you couldn't hear any difference before burning in one of the units. That is necessary to esatblish that the perceived difference can be attributed to the burning in with reasonably certainty.

 

[phase_accurate]

Hi John

I can take these of course. I actually took 50 mOhms for a "common" cable and two times 100 mOhms for the bi-wiring case.
I took same signal levels for both LF and HF. Your choice is more life-like and would even show more influence on the tweeter.

What I am still unsure about is the amount of driver non-linearity involved.

As you might have recognized I am not trying to prove your theory but mine (I am that selfish sometimes).

Regards

Charles

 

[jneutron]

Hi John

I can take these of course. I actually took 50 mOhms for a "common" cable and two times 100 mOhms for the bi-wiring case.
I took same signal levels for both LF and HF. Your choice is more life-like and would even show more influence on the tweeter.

What I am still unsure about is the amount of driver non-linearity involved.

As you might have recognized I am not trying to prove your theory but mine (I am that selfish sometimes).

Regards

Charles

That's just fine..all the info is useful.

I used 52 mohm as the "best case" setup.

Try with a #16, 5.15 mohm per foot, total 100 milliohms.

I recommend this simulation:

Two branches fed from a zero impedance voltage source, the source a 5K sine at 8 volts.

output 1...a 100 milli resistor feeding an 8 ohm load through a capacitor of appropriate value.

output 2...an identical circuit to the first, with a small adder. Pull a DC current from the cap-resistor node to ground using an ideal infinite impedance current sink. (on second thought, I gotta think about that a little, there may be something funny in doing it this way, you may need to include an arbitrarily large inductor to keep the hf out of the sink.).

Difference the outputs. That is the distortion.

You could just as easily use an inductor with 4 ohm load in circuit 2, and add the dc voltage to the voltage source.

I agree that driver non linearity is of concern...but who cares about that if the signal being presented to it is already non linear??

I learned long ago, if a large complex analog system is not performing correctly, the first thing done is to make sure the power supply is operating within tolerance...otherwise, troubleshooting the circuitry that uses that supply is useless.

Cheers, John

 

[serengetiplains]

As I've said over and over, we routinely do just this kind of controlled tasting in the wine business. No excuses about blind tests, you either can get a difference or you can't.

The main issue with controlled testing in audio is that it's a pain to set up and do properly, and that's not as entertaining as speculating about phonon-meson resonances in cable dielectrics.

SY, am I to take from this that you think our taste and hearing senses respond similarly to blind testing? It would seem to me almost an evolutionary imperative that the taste sense be able very quickly to differentiate subtle differences, as those differences could carry ramifications for one's health---no small matter in the evolutionary mix. I cannot think of a similar rationale for hearing. In my own personal experience of listening for audible differences, I require as much as several hours to hear some of the more subtle differences, which sometimes appear to me as a subtle grating that might hit me on musical peaks etc. Give me two wines and I can tell A from B within 5 seconds. I don't see these processes as closely comparable.

 

[jneutron]

SY, am I to take from this that you think our taste and hearing senses respond similarly to blind testing? It would seem to me almost an evolutionary imperative that the taste sense be able very quickly to differentiate subtle differences, as those differences could carry ramifications for one's health---no small matter in the evolutionary mix. I cannot think of a similar rationale for hearing. In my own personal experience of listening for audible differences, I require as much as several hours to hear some of the more subtle differences, which sometimes appear to me as a subtle grating that might hit me on musical peaks etc. Give me two wines and I can tell A from B within 5 seconds. I don't see these processes as closely comparable.

Differentiate the same wine from the same vineyard from two consecutive years, or from the same year using wine from new oak or 5 year old oak, or made from different vendors using oak from different locals.

Heck, even I can see through that argument.

Cheers, John

 

[poobah]

Hey Neutron,

Your theory re biwiring has as much to do with fidelity over a single wire as it does the merits of bi-wiring. How 'about a new thread as this thread contains many other (9?) contentious and distracting issues?

There is something weird about all this... it seems to foul up superposition.

 

[serengetiplains]

John, I'm not simply throwing up arguments, as you perhaps imply---I'm looking for an adult conversation---but giving my experience. My wine tasting sensitivity (nothing you would know about) is at a high level so, yes, new oak, 5 year oak, easy to discern; wine from two consecutive years, even easier ... and all and any within the first taste.

Essential to one's taste or hearing sensitivity is the length of time one has given attention to developing that sensitivity---no overnight Robert Parkers. For my part, I have noticed a difference in my taste and hearing sensitivities and have perceived, as I intimated previously, a difference between the time required to discern audible subtleties. I say "audible subtleties" well aware that, with increasing time between listening sessions, factors enter in to confound the objectivity of the comparison such that "subtleties" I might claim to hear might not exist. I'm personally unaware of a solution for that problem, but I'm far from convinced that every audible difference that can be heard can be perceived within the same time frame. Video is of course a different animal altogether as not only are our eyes very fast to perceive very small differences, a true A/B test can be performed simultaneously. Wine and audio comparisons are unlike video comparisons in this regard.

 

[jneutron]

Hey Neutron,

Your theory re biwiring has as much to do with fidelity over a single wire as it does the merits of bi-wiring. How 'about a new thread as this thread contains many other (9?) contentious and distracting issues?

There is something weird about all this... it seems to foul up superposition.

Yes, a new thread is warranted. There is too much noise on this channel as the origional poster covered so many different topics..

And, yes, it would appear that if my theory and analysis holds up, it does in fact demostrate a violation of superposition. If that is the case, then the theory of superposition needs to be either corrected or abandoned. I believe it would most likely be modified to fit branch issues. But I'm discussing this with a few of my collegues, they are pretty smart. One in particular is used to my goofball ideas, but listens. As on occasion, I am correct..

I have broken laws before, but never with intent to do so.

I ran the analysis using 4 amps dc and 1 amp ac into both a 50 milliohm (10 feet of #12) and a 300 milliohm wire run (30 feet of #16)..the numbers are staggering in the 300 milliohm run.

If my analysis is correct...I'm gonna have fun..

Cheers, John

 

[jneutron]

John, I'm not simply throwing up arguments, as you perhaps imply....

I was not implying that, it was more like an unstated accusation..it would seem that I owe you an apology...Sorry..

My wine tasting sensitivity (nothing you would know about) is at a high level

Your statement is accurate at two levels..I know nothing about your abilities, and do not have that ability..I can, however, tell a red wine from a white...eyes open, of course.

Cheers, John

 

[poobah]

I am with ya John,

I have followed along with the math and see the weirdness myself... I started running sims but decided to persue a purely math approach. We should be able to prove it all one way or the other...

 

[SY]

SY, am I to take from this that you think our taste and hearing senses respond similarly to blind testing? ...

Give me two wines and I can tell A from B within 5 seconds. I don't see these processes as closely comparable.

It's not so easy when you're trying to determine the potential organoleptic consequences of printing the cork with ink from Vendor A versus Vendor B... Yet these sorts of things are life and death decisions for my company, and there's no way we or our customers would accept anything less than proper blind tests.

There are analogies between different sorts of sensory testing. In my last job, I did haptics, now I do organoleptics. Different senses, yet all of the human biases and need to eliminate them to get to what's real are still the same. My friends who do optical stuff tell me the same thing.

BTW, I've tasted with Parker and I'm a lot less impressed than you are!

 

[serengetiplains]

SY, I agree, and Parker's not to me the be-all of wine tasters; I used him as an accessible example. I prefer the Brits, myself. Have you seen Mondo Vino? Interesting perspective on Parker and his Mondavi/Napa links etc. Ah, the world is corrupt.

Wine tasting seems to me more amenable to A/B testing, and for that reason alone should be used in the wine industry. I can listen to music for hours at one sitting. I typically can't drink wine for hours at one sitting. Nor can I compare what ears and taste buds might perceive (assuming, as regards the latter, I'm still conscious) after several hours' respective indulgence.

I'm suggesting that, in music, that which grates on one's hearing (or brain) after two days/months/years might not be (is not IME) perceptible in 10-20 minutes' A/B. But 10-20 minutes is a more than adequate amount of time to perceive the true quality of a wine. If this is indeed true, and if long-term irritations do figure in audio satisfaction, as IME they importantly do, A/B testing shows its limitations.

 

[serengetiplains]

I was not implying that, it was more like an unstated accusation..it would seem that I owe you an apology...Sorry..

Accepted and appreciated, John. I wish I had engineering skills, but alas do not, and I appreciate what an engineering perspective brings to the audio arts. I am not, however, equally crippled as regards a scientific understanding generally, as science is about the only thing I read in my spare time.

 

[AcidOrangeJuice]

No lineup, the problem with a great many double-blind tests is they're structured like the AES-sponsored one described below:

http://stereophile.com/asweseeit/107/

That was the 91 AES Convention in New York. My brother, who at the time was a lead audio designer for Nortel and made a living from structured audio testing, attended a similar demonstration at a late-Eighties AES conference in Toronto with one of his peers. They found the methodology so unprofessional and flawed as to leave mid-point to go harmonica shopping instead.
This discussion would be better served if posts like yours dealt more with the test proof you so adamantly believe exists instead of ad hominem and unprofessional characterizations of personalities. Unless of course you want to present proof it's your field of speciality and show the research.
Frankly I'm dismayed the moderators didn't step in during the last five pages or so.

I disagree with you at all.
Stereophile is a magazine which obtains many money with the sponsors of speaker cables and power cables. The doctrine the Stereophile is: the most expensive equipment is the best!!.
As the DBT are a latent danger for the vendors of snake oil; I doubt very seriously that Stereophile (who sponsors many of these mystics products of doubtful credibility) could have a position in favor of the DBT.

That if is a fact proven and verified is that DBT, when are done well, are a powerful tool in order to verify if really an AUDIBLE difference exists.

As a proof, DBT are used almost daily in order to improve and to fix perceptual lossy audio codecs (as mp3 or vorbis for example) and this has worked successful. For example, LAME mp3 encoder (which is one of the best mp3 encoders available in this moment) has been improved (with extensive DBT tests with many people) even the point that with specific settings, recommended by the developers, you can obtain transparency (transparency is the term used to define the point which doesn't exist perceivable audible differences between the original source audio file and the mp3 encoded file).

Due to the nature nonpredictable of lossy audio compresion algorithms is not possible to use measuring instruments commonly used in electronic design. These are perceptual audio codecs algorithms, and, that I know, the only effective way in order to improve is with DBT tests.
Even the same creators of the mp3 format (Fraunhofer scientist institute in Germany) have said this. Without the DBT tests used extensively there, had been impossible to develop efficient audio lossy encoders.

In my own experience, I have used DBT tests with the intention to improve mp3 encoders too, and I have been able to see that this really works, until finding very subtle differences.

It's true that to develop a DBT tests in hardware is more difficult, but nonimpossible. If your brother could not develop a hardware DBT test is because the circumstances did not allow it; but this is not a fault of the DBT test methodology. Another people, as for example Bob Carver, have been able to do this successful.