Blind Listening Tests & Amplifiers

[nw_avphile]

Originally posted by SY
There are a number of amps that would fare horribly on a null test that are, nonetheless, regarded by some as having an excellent sound quality. Not everyone is designing with the idea that an amp's output voltage should faithfully replicate the input voltage times a constant.

The null test (when front-to-back time is accounted for) is very useful to compare the degree of fidelity of amplifiers. It does not, however, tell you which errors are audible and which errors aren't. Only valid listening tests can tell you that.

I have to agree with Tube_Dude here. You can both listen to the difference signal AND you can measure it. Further, if you minimize ALL forms of errors, aren't you improving the amplifier's performance?

Remember, the sine wave/resistive load argument usually used against "perfect amplifiers" with lots of NFB, etc., cannot be used here. We're talking about minimizing errors and distortion while driving real speakers (yours even) with real music.

I further agree it would seem this form of testing has been quietly swept under the rug by the same folks who refuse to conduct blind tests of amplifiers. As Tube_Dude said, the XL-280 and Hafler test caused quite a stir in its day, and that was BEFORE most of the mainstream audio press started discussing the "sound" of amplifiers like they do today.

On a more technical note, my experience with null testing is much of the error is from phase shift at high frequencies. Is this what you're referring to by "front-to-back time" SY? Amplifiers that have flat response for a few octaves beyond 20khz tend to do better with null testing unless you bandwidth limit the null signal. It also helps to use a CD as your source material as they, as we all know, have little output above 20khz.

It's also interesting that the overall levels of distortion obtained from null testing (even corrected for any HF phase shift) with music and speakers are usually considerably higher than those obtained using sine waves and resistive loads. This supports all of you who believe sine waves and resistors don't paint the complete picture.

So, in my opinion, we have a sensitive test for audible subjective differences between amplifiers (the blind test) and a sensitive analytical test for all kinds of distortion and errors using music driving real speakers (the null test). It's surprising they're not used more by people evaluating amplifiers (at least that more people are willing to talk about).

One has to wonder what goes on behind closed doors at companies Harmon International which has invested millions building some of the best audio technology R&D centers in the world? It's hard to believe that someone there hasn't done blind comparisons or null tests to compare/evaluate their products across multiple lines?

Harmon sells power amps from Mark Levinson, HK, JBL, Madrigal, Lexicon, Crown, Proceed and probably a few I'm missing. It wouldn't surprise me if some of the HK or Crown power amps, for example, are indistinguishable from those that cost ten or twenty times as much and are sold under the high-end Harmon names like Levinson.

I know that a lot of the differences in the above brands are the market they sell into, brand name, packaging, status, etc. But wouldn't it be an interesting exercise to arrange for blind and null testing of representative products from the above brands and have the results widely published?

[analog_sa]

Whoever started this topic found a truly ingenious, politically correct and moderator-approved way of sharing his opinion of the deluded idiots who are interested in this hobby because they hear a difference. Well done indeed! No doubt this person has made lots of audiophile friends (some of them, of course, customers).

Tube_Dude

I am a bit dusty on the Hafler test but it does use op-amps in the setup, doesn't it? Which kind of accepts op-amps as being perfect. Most 'tube-dudes' i know are higly allergic to op-amp sound which may simply mean they're deluded enough to hear it. Just one reason the test may be meaningless.

cheers

peter

[planet10]

I wish i could find the emails... but at one time a thread went thru one of my mail lists where two Statisticians proved beyond any doubt that i had that the only valid result that can be garnered from the ABX test is that 2 DUT are different. If you can't tell a difference the test does not prove or disprove anything.

dave

[Steve Eddy]

Originally posted by planet10
I wish i could find the emails... but at one time a thread went thru one of my mail lists where two Statisticians proved beyond any doubt that i had that the only valid result that can be garnered from the ABX test is that 2 DUT are different.

Uh, why does it take two statisticians to "prove" that? I'm not aware of anyone claiming that ABX testing does anything more than detect differences between two components. Who's saying otherwise?

Sure you're not confusing that with an AES paper regarding the minimum requirements for statistically significant results?

If you can't tell a difference the test does not prove or disprove anything.

Yes. That's called a null result.

[analog_sa]

Maybe it doesn't prove, but it sure hints at deficient hearing

[nw_avphile]

Originally posted by analog_sa
I am a bit dusty on the Hafler test but it does use op-amps in the setup, doesn't it? Which kind of accepts op-amps as being perfect.

No, the Hafler circuit was completely passive (no op amps or even transistors to blame anything on).

And, FWIW, I don't think folks here are "deluded idiots" for enjoying their hobbies. As I said, it's hard to argue with the concept of spending time and money on something that makes you feel good (for whatever reasons). Further, by your definition, I used to be a "deluded idiot" myself.

But I believe there are also people here with limited budgets looking for the best performance they can get for a given amount of money. Those folks, in my opionion, might be better off spending their limited funds on things that can be demonstrated to make an audible difference like bigger power supplies, more output devices, low distortion circuit designs, etc. Spending $8 each for "audiophile approved" poly capacitors and the like is better left to those who have other priorities.

I guess it's the difference between art and science. If you view DIY amplifiers as art, this thread probably isn't for you. If you view it more as science, there's some good information here.

[planet10]

Originally posted by Steve Eddy
Yes. That's called a null result.

I wanted to mention it because i am sure there are some reading who would erroneously conclude that if you heard no differences in this test, that there were no differences.

dave

[Rob M]

Originally posted by planet10
If you can't tell a difference the test does not prove or disprove anything.

Statistical tests don't prove or disprove anything. Based on a listening test, they can tell you that you'd have to have been astronomically lucky to have guessed right as many times as you did if the devices actually sounded the same. Or, they can tell you that based on the number of correct picks, there is no evidence that you aren't randomly guessing. What you can conclude from that depends on the experimental setup, the listeners involved, possible confounding factors, etc., and how much statistical evidence in one direction or the other constitutes "proof" is a matter of expert opinion.

I would think that if you take enough people who ought to be able to tell the difference between two amplifiers, and put them through enough carefully controlled listening tests, and still are left with no evidence that the two devices can be distinguished by sound alone, that would start to look to me like proof. And, if in many experiments carried out by many different labs, no one had ever been able to show that anyone, however skilled, could tell the difference between two particular amps when all other factors are controlled for, then that would look a lot like proof.

[analog_sa]

better off spending their limited funds on things that can be demonstrated to make an audible difference like bigger power supplies, more output devices, low distortion circuit designs, etc

I was under the impression that nothing made a difference. Why would more output devices make a difference if the amp is not being overdriven? Or a bigger PS? Such excess can possibly be justified only by a headbanger, right?

$8 for a cap? you're surely kidding if you think you can get anything decent for that. Maybe in the seventies?

[Tube_Dude]

Why would more output devices make a difference if the amp is not being overdriven?

They may make a diference if the loudspeaker where you are doing the nulling test have very low impedance in some frequencyes...and for a indistorted signal the amp must put some amps in...so more transistores may help!
Nothing unexpected!!!

Or a bigger PS?

But that's also expected...
In a power amplifier we have 3 "inputs"...
1- the input signal
2- the power suplly modulation in function of the output current
3-the output from the loudspeaker(EMF)
Of this only 1 is the wanted input
2and 3 are distorted and delayed inputs that the amp must correct if not distortion(specialy intermodulation) is added to the output
So no wonder that a stiff power suplly may cause a diference...

All in all nothyng fancy or esoteric here...only going a litlle deeper in the way a amp work in real world loads...and not in the proverbial 8 Ohms resistor load!!!