Double Blind tests *did* show amplifiers to sound different

In many online debates, position taken by some that when amplifiers are used that have flat frequency response and low distortion that no double blind tests have shown them to sound different.

Well, I managed to dig up a 32 year old test that says otherwise. What is fascinating is that one of the article/testers was no other than Arny Krueger, the most vocal champion of such arguments in forums!

The story is that it occurred to me one day that I had never, ever seen Arny document any blind tests he had run. Yet he said he had run 100,000 of these (no joke). After pressing him for an example, he said there was only one test that was ever documented(!) and it was in this 1982 Amateur Audio magazine. While that is pre-Internet era, one would think that someone, some place would have scanned a copy and put online. This has been done for other articles that said no difference was found but this article, was never, ever put on the net for anyone to read. Arny to his credit had made references to it here and elsewhere but never discussing the details of it.

My challenge then begun in trying to locate a copy of that magazine issue or another that was printed in UK. I came awfully close and missed it by one month on the UK publication. But came up empty other than that despite searching everywhere. I mentioned this to Arny and to my surprise he linked me to an auction site that I had never heard of with someone having that issue. With reluctance I signed up to their service and after paying my $11, the magazine arrived.

The article is actually quite short. What I am about to show you is about half of it and contains all pertinent portions. I am happy to field and answer questions that may be in the rest of the article.

So let's start at the top. Here is the title of the 2-page article:



What a headline, no? It says they have "proven" that amplifiers sound different! You almost don't need to read any more but for completeness let's go into it.

Here are the amplifiers in question:



So these are very powerful amplifiers. We are not talking a tube SET amplifier or some other amp that runs out of juice at 20 watts. We have very powerful amplifiers clocking at over 200 watts at 4 ohms.

This being the pre-CD era, a turntable was used. Here are the details of that and the speakers used:



So how did they do? Here are the votes for the three tracks tested:



We see why they say "prove" as the two tracks show statistically that the results are not due to chance.

So the conclusion is as expected:



Not the message we see broadcast on forums.

So what caused the difference? I think it is what I believe to be behind such audibility differences namely what happens at clipping point which occurs far more often than we may think:



No data is presented on what would happen if the volume was turned down.

So there you have it. "Proof" that amplifiers do sound different in double blind tests .

And that has been their argument. But they have no answer for why they didn't present this test for 32 years. And that even 200 watts was not enough in that situation.

microstrip said:

Amir,

Unfortunately I have offered my vintage Hi-Fi News & Record Review collection to a good friend long ago, keeping only the few issues I considered of historical interest and do not have this article anymore. However I would like to understand what means exactly "clipping occurred 1% of the time". Why not tapping the amplifiers violently with an hammer 1% of the time?

Click to expand...

. They don't say much more on that than this:



I don't know how when playing music on an analog scope they managed to detect 1% distortion. I know how to instrument the amp internally but even then, it would be hard to come up with a percentage as they did.

A couple of points:

1. The test shows the importance of selecting the right clips that reveal the problem at hand. Two of the clips were highly revealing but the other was not at all. In this situation the selection was likely made at random. So it was a luck of the draw as the saying goes that two clips were revealing. They could have picked two other and arrived at the conclusion of no difference.

This is a problem. When we go through the trouble of setting up a test like this, we need to look at the hypothesis at hand. That hypothesis should not just be a layman one of "let's see if we can make this point or that." Instead one should think through what "1% clipping" could do and find content that makes it revealing. Without a hypothesis, you wind up using random files and getting potentially random results.

2. I think the protection circuits in amplifiers need to be examined more. They are complex beasts, performing dynamic current limiting for milliseconds at a time. The nature of negative feedback system is that it will fight this mechanism. The current limiter puts the brakes on the output stage of the amplifier. That creates distortion which the negative feedback sends to the input of the amp. That circuit then tries to force the output stage to do what it is demanding, i.e. push it harder. That in turn causes the protection circuit to current limit more.

As if that were not enough, majority of amplifiers use unregulated power supplies for the output stage of the amplifier. That voltage sags under load creating yet another indirect feedback loop. Now we have three feedback loops none of which is considering the other in the system. Such behavior simply cannot be simulated or predicted. Testing is also very hard since as noted, speaker loads are different and cause the current limiting to act differently. And of course that factor is also content dependent.

In the case of severe limiting, none of this matters. But when limiting is undetectable as such, then there can be coloration.

Whatmore said:

I'm not sure what's surprising here. They drove a difficult load with amps that were not up to the task (they clipped). Unless we expect amps to clip identically then of course they would sound different.

Click to expand...

They tested the amps and they produced over 200 watts/channel each. The argument has been that just about any amount of power is sufficient for everyone.