May 21, 2002 at 6:04 PM Thread Starter Post #1 of 34


Member of the Trade: Audiogeek: The "E" in META42
Jun 23, 2001
What does DBT stand for?

(it's related to the FORUM, not the Fourum Topic, so I didn't think the Member's Lounge would be appropriate)
May 21, 2002 at 7:31 PM Post #3 of 34
The link to that thread is wierd... Here is the link straight to the Audio Asylum section about it:

Basically, DBT stands for Double Blind Testing. A way to do this would be to have a switch box that compares two components. Press the (A) button to select the first component, press the (B) button to select the second component, and press the (X) button to randomly select either A or B. You could listen to both, and then try the random button, and see if you can guess which one has been selected. Using that method is called ABX (named for the three buttons).

This all sounds well and good, I mean, its good to test things, double blind testing can be very effective and legit. So why make certain areas a 'No DBT/ABX Zone'? Basically, the problem is that Double Blind Testing is INCREDIBLY DIFFICULT TO DO CORRECTLY!! Not like 'kind of difficult', but INCREDIBLY DIFFICULT. There are people who go to school for 10 years just studying how to properly give this sort of test. One of the biggest problems that prevents us from doing any DBT or ABX testing is the design of the switchbox. How do you build a switch box that lets you do instant switching, but that doesn't affect the sound AT ALL? To test cables, you would need something that isn't active OR passive. If it was electrically connected, then it would garble the test by changing the auditory signature. Having someone switch the cables manually is the only way, BUT, that way, it takes too long for your auditory memory, and you can see what they are doing, which changes your preceptions (don't try to claim that it doesn't! We are all human beings, and are prone to bias, one way or the other). The only thing I've been able to think of is a robotic system who's entire control and motor system is on an independent power source. Super high precision griping systems would replicate a human grip on the cable, but could rapidly remove one cable, and plug in a second set. The switch would have to take less than 2 seconds. Then the first cable would have to be moved out of the way, and the robot would have to let go of the newly plugged in cable, move back, and then shut down. This would have to be done to prevent any effect of having a metal hand gripping the cable, and to prevent any electromagnetic interferece from a robot affecting the sound. You start to see how this gets incredibly complex. I think maybe NASA could build a robot that does that.. But even then, it would be hard to make it so that Button A and Button B's results would be identical IN EVERY WAY to each other except the cable, and that each position would be identical to a system that had only that cable, and no other cables. For a double blind test to work, ESPECIALLY with subtle things like cables, you need to control for EVERYTHING for it to be statistically valid. You might say "oh, if i make a good switchbox, and use quality parts to do electrical switching, then it'll be fine" and then someone can't tell the difference between two cables, you might laugh triumphantly and say, "See? I've proved that they are the same!" But it wasn't scientific, because you didn't control your experiment! If you tried to pull some crap like that at a research institution, they'd laugh you right out. Or at least look at you disapprovingly.

So basically, arguments about Double Blind Testing and ABX testing are doomed to go nowhere, and will always result in flame wars and unhappiness. It is NOT a cop-out for people who don't want to admit that their cables don't make any difference. It is based in the science of double blind testing, and its inherent difficulties.

When some professional academic institution develops a complete, perfectly controlled mechanism for doing ABX and double blind testing with audio components that takes into account ALL possible control aspects, then I'm sure Jude will lift the DBT ban so they can tell us about it. But until that happens, its best to just let people come to their own conclusions. I can use little tests to convince myself, but I can't go boasting that i've proved anything until I do a REAL DBT test, with lots of subjects.

People try to 'prove' that certain cables are better than others, or they try to 'prove' that cables don't make a difference, but none of that is true proof, its all just hearsay, conjecture, and unscientific testing. You can give your opinions, either that you didn't notice a difference, or that you did. No one can tell you conclusively that you are 'right' or 'wrong'. I can shout PLACEBO at you all day long for claiming that a 500$ cable was better than a 20$ cable, but i'm just blowing wind because I can't offer any proof besides what I happened to notice (that is assuming, of course, that I actually listened to the expensive cable
). Likewise, I can call you "Mr. Tin Ears" till i'm blue in the face because you didn't notice a difference between the radio shack cable and the 1000$ wonder-cable, but I can't prove jack, I can only offer my own experience, once again assuming that I actually listened to them both myself!

Anyway, hope that clears things up. Things like this are very touchy in the audio world, and other worlds too. In scientific communities, DBT and ABX testing about other things is regarded VERY carefully... If you are doing tests like that, you are expected to document EVERYTHING, and to analyze the effect that EVERY variable has on the situation. Very difficult. But the scientific community realizes that those things have to be done to really get any true value out of a DBT test. What we are looking at here is a very real application of the Heisenberg Uncertainty principle, where the act of measuring something changes the outcome in an unpredictable way. This isn't a limit of technology, its a fact of the universe's structrure (if you want, we can talk more about Physics and its implications on the world, and our daily lives).

I'm open to more scientific investigation into the properties of audio equipment, but people must realize the depth and complexity. A sound card and a cheap osilloscope are not enough for me to 'prove' a cable is the same as another cable. You aren't controlling the experiment. Its not scientific.

Now, anyway, to wrap up this little rant/explanation, I'd like to note that I'm still agnostic when it comes to cables. The engineer in me says 'bah, they are the same' and yet the audiophile in me says 'no, there is a difference, its just subtle.' I've heard differences, but i'm not sure if its placebo or not. I don't think someone who says their 500$ cable sounds way better than their 20$ cable is full of crap, but I also don't think that someone who feels that its placebo should be laughed out of the room. I think we should be open minded, listen to people's intelligent reviews and experiences, and then try it for ourselves. Then we can decide what we want, and what makes us happy.

Group Hug! Cable naysayers, and yaysayers alike!

May 21, 2002 at 7:42 PM Post #4 of 34
Bah, that's what I get for copying the url before signing in. Thanks for the catch, phidauex, I fixed the link. Also, nice discussion of DBT intricacies.
May 21, 2002 at 7:50 PM Post #5 of 34
Thanks dhwilkin and phidauex!
That sums it up pretty much...

Of course, one does have to wonder about CD players and amps with more than one input/output which would enable swapping cables pretty quick, or designing a switchbox with really short wires to minimize the effects on the sound and having a listen to it using reference-grade stuff to make sure the difference is audible, but that kind of thought isn't allowed here
May 21, 2002 at 10:14 PM Post #6 of 34

WOW! that explained a lot to me. I was just about to post a thread where I was going to question the no-DBT thing. But you've convinced me.
May 21, 2002 at 10:40 PM Post #7 of 34
Good post, phidauex.

But I think there is a kind of blind test not so difficult to set up. A sort of "digital" blind test. That test would consist in the comparison of an original piece of music digitally error-free extracted from a cd, and this same piece of music "modified", I mean, played from a good source, passed through a cheap cable and then recorded with a good sound card. These pieces could be recorded into a cd-r and listened to with anybody's favourite equipment. The actual test would consist in being able to identify the original piece from the modified one. To make it serious, one original piece would be the reference, and some original and modified pieces would be recorded randomly at the cd. The challenge would be to correctly identify which pieces are original, and which modified. Of course, to make it blind, the person randomizing the pieces before burning them to cd should be different from the person trying to identify the pieces.

You can say that I'm adding the effect of the sound card to the comparison. I agree, but I think it could be sort of an "advantage" to make this test easier.

In fact, I could provide such a test for you people. You would send me the original file that you wanted, I would modify the piece, and then I would put at a web page some modified and original copies of the files without telling which is which, except for the original reference. You would only have to download that files, burn them to a cd-r, and try to tell the originals from the modified ones.

An easier, less scientific test could be set up, with a reference original file, a modified file, a more modified file (say an additional MP3-encode-decode step), and the original again. You would have to tell which of the 3 is which, knowing only that the other file is the original reference.

What do you think? If you are interested, we could discuss it more thoroughly, I'd be open to any additional suggestions and/or modifications.
May 22, 2002 at 3:29 AM Post #8 of 34
Ricky: I think a test like that could be interesting, but I wouldn't necessarily call it scientific. Once again, there are too many variables. With the mp3 idea, I don't think it would be very good of a situation at all, with the fact that a file will be encoded slightly differently every time anyway, mp3 is 'semi-analog' in format.

I think a test like that could fall into the catagory of 'tests that are good enough to convince yourself' but that still aren't real DBT tests. Its sort of like putting one cable on the right channel, and the other cable on the left channel, then going back and forth. Its good enough to get an idea for yourself, but it still isn't scientific enough for 'proof'.

But your plan might be an interesting way for people to sort of 'audition' cables, though you'd need some pretty drastically different cables to notice the differences I think, what with all the other variables. You would be testing the effect of the cable, but that effect would be muffled by all the other effects going on. Its like, in a Source -> preamp -> amp -> speaker situation, changing one set of interconnects is like changing 1/3 of the interconnects. Changing one set of interconnects on a Source -> A/D -> source -> preamp -> amp -> speaker is like changing 1/5 of the total interconnects. The differences might be much harder to discerne.

So anyway, I think its a good idea, and I might even want to participate, just to give it a whirl, but even that isn't sufficient for scientific testing. Just more 'well, its good enough to convince myself of my position, but..' Its that big unscientific BUT that makes it hard to defend a position based on a DBT-like test.

I'm glad you guys liked my comments, sometimes I sure can fire off a diatribe, and sometimes they come out hitting the spot, and other times they kind of land a bit askew.

May 22, 2002 at 5:59 AM Post #9 of 34

Originally posted by phidauex
...Basically, DBT stands for Double Blind Testing. A way to do this would be to have a switch box that compares two components. Press the (A) button to select the first component, press the (B) button to select the second component, and press the (X) button to randomly select either A or B. You could listen to both, and then try the random button, and see if you can guess which one has been selected. Using that method is called ABX (named for the three buttons).

This all sounds well and good, I mean, its good to test things, double blind testing can be...

Phidauex, I think you are making the problem far more complicated than it needs to be.

The purpose of a double-blind test is to prevent the subject from being influenced by "experimenter bias". This means that it is important for the human experimenter (who interacts with the human subject) to not know how the independent variable(s), which in this case are the specific audio sources, are structured with respect to the subject. I can think, off the top of my head, of a very simple way to structure this type of experiment in a manner consistent with that objective.

Use two human experimenters. One interacts with the test subject one does not. Put your audio sources behind a wall, and put the first experimenter behind the wall as well. The second experimenter interacts with the human subject, and explains and monitors the test. Run the headphone cable through a small hole in the wall.

The first experimenter flips a coin to determine which audio source is "A" and which is "B". He plugs the headphone cable, which came through the hole in the wall, into source A. The experimenter that is in the same room as the subject provides some queue to the other experimenter (knock on the wall, flip a switch to light a light, or whatever) when it is time for the experimenter on the other side of the wall to change the input from one source to the other. When the queue comes, the experimenter quickly connects and disconnects the headphone cable from one source and plugs it into the other.

There is no experimenter bias, because the only person who knows which source is which has no face-to-face contact with the subject.

Controlling the independent variables in an experiment like this is not complicated. Analyzing the data to infer causality is also not complicated...anyone that has done statistical analysis in the social sciences would be able to do it.

Not as cool or complicated as your robot proposal, but a lot cheaper to build and equally effective.

The Headroom guys even set something like this up on their tour.

It's been over a dozen years since I got my degree in experimental psychology. It's nice to put it to use


May 22, 2002 at 6:42 AM Post #13 of 34
Here's a link to an article in LA Weekly you might find interesting.

The relevant paragraph for this thread is:

"...the sound quality of consumer digital and analog recording equipment in recent years has reached such a high point that, for human ears -- excluding those of our dear self-professed audiophiles, a self-deluding bunch -- the difference between an analog and a digital reproduction of an analog source has become, if not entirely indiscernible, then certainly superior to MP3 downloads."
May 22, 2002 at 2:58 PM Post #14 of 34
Zend: For simply going back and forth between sources with a headphone, I think that could be fast enough. The situations I was talking about are of the type that often occur in other hifi arguments, where someone wants to double blind test with cabling, but shutting off all the equipment, making 4 disconnections, and then 4 reconnections, possibly with locking RCA plugs, the process takes too long.

And yes, i think the robot idea is cool.

Jeff: I hear this kind of thing a lot, someone says something to the effect of, "With modern technology, no one can distinguish between <insert audio component here> and <insert audio component here>." They say it like its a scientific fact, yet they generally have no basis. Some magazine did a test where they played a cd audio track, and one that had been encoded at 128kbs MP3 to people, and people couldn't tell the difference, and so they 'proved' that 128kbs MP3s were the same as CDs. They didn't take into account the fact that they used a cheap soundcard with crappy headphones in a loud office to do the tests. In fact, that probably didn't even cross their minds.

Once again, I'm forced into an agnostic viewpoint. I 'think' records sound better than CDs. Maybe its just my sources. Maybe its all in my head. Maybe the records really do sound better. All the time we are learning that our bodies are more amazing than we thought. While that doesn't necessarily MEAN our ears are more sensitive than we thought, it certainly suggests its possible. One thing I like to remember is that jitter was identified by the audiophile community before it was demonstrated to be a real effect. Now we can measure it, and predict it, and correct it. But back in the early days of digital, when audiophiles claimed they could hear this wierd effect, people said they were self-delusional. Now, once again, this doesn't mean that everything an 'audiophile' hears is a real effect, but it does demonstrate that sometimes the ear picks up something that we don't have a concrete explanation for yet. I don't put much stock in people who claim to have 'proved' this or that with respect to what you can and cannot hear.

I have seen lots of hearsay, conjecture and technobabble jibba-jabba on both sides of the fence, and I'm at the point where I'm willing to just trust my ears, and go with my instincts until someone decides to do some real, verifiable, repeatable, scientific experimentation.

Oh, and to those of you who are bored by this thread, remember that NO ONE IS FORCING YOU TO READ THIS. if you are bored, get up off your computer, and go DO SOMETHING, don't just make snide little comments. mkay? mmkay.

May 22, 2002 at 3:50 PM Post #15 of 34
It usually takes me weeks to months to get a handle on any new component in my system. The first step is to get past burn-in and novelty issues. Novelty can make some things sound good that will become fatiguing over the long run (something known and used by less than scrupulous audio salesmen).

I have yet to see any methodology used in audio that takes into account long-term effects, such as the discriminatory learning that is part of the perceptual process.

Users who are viewing this thread