Cable testing protocols

[mike1127]

I'm hoping to do some single-blind cable testing. I don't want to use an ABX box. When I'm doing sighted listening comparisons, I like to listen to a particular setup for about five minutes, with several music selections, to get a feel for it---so I want to use that type of listening during a blind test. Because I won't use an ABX box, I will have a friend hook up a cable that is unknown to me, and do this several times while varying the cable choice according to the following protocol:

- I will select two cables to be compared to each other
- one will be called A, the other B. The assignment will be made by coin toss and will be unknown to me
- the friend will, by coin toss, select the ordering ABAB or ABBA
- on four consecutive runs, the friend will hook up the cable as indicated by the selected ordering
- I will listen for as long as necessary to each cable. In my recent sighted tests, I tend to feel I need about 5-10 minutes with various music selections
- I will be required to identify whether the protocol was ABAB or ABBA

In addition, but this is not necessary, I could comment on the identity of A and B. I could say "A was clearly the better cable", but I am only required to prove I knew the selected ordering ABAB or ABBA.

We will only do one trial to start, to get a feel for the situation. Doing 20 or 30 trials will be time-consuming, especially because I feel that I'm sharpest if I limit the number of critical listening sessions I do in a day.

-Mike

 

[Pio2001]

Good idea.

The choice of the ABAB and ABBA patterns allows you to eliminate the need to detect identical samples, and the need to identify which is which.

In ABX, when you compare X to A and X to B, one of the two comparisons is between identical stimuli, and you don't know which one.
With AB and BA, you need to characterize the sound of A and B in order to know which is which.

Keep in mind that for the test to be valid, you don't need to answer after the first presentation is over.
If you don't answer, and run the same trial the day after, or if you ask for listening to the mystery sequence over and over, until finally deciding if it is ABBA or ABAB, the result will be the same. So don't feel forced to give an answer if you are not sure about it. If you need time, take it.

 

[nick_charles]

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif - the friend will, by coin toss, select the ordering ABAB or ABBA - on four consecutive runs, the friend will hook up the cable as indicated by the selected ordering -Mike

I think you are on the right track but I have some concerns about your approach.

In both variations you know that there is a change between the first two stumuli and the last two stimuli, there are 3 transitions and you know 2 of them are changes. It would be better to have A and B appear wholly randomly thus removing the expecations effect.

You have a 50% chance of getting the order right by guessing so if you follow your model you need to do at least 10 sets and have a very patient friend.

Good Luck

PS nice hamster !

 

[mike1127]

Quote:

Originally Posted by nick_charles /img/forum/go_quote.gif In both variations you know that there is a change between the first two stumuli and the last two stimuli, there are 3 transitions and you know 2 of them are changes. It would be better to have A and B appear wholly randomly thus removing the expecations effect.

I'm not sure what the basis of your objection is. There's no statistical reason to do this---all that matters is we agree what constitutes one trial, and that there's a 50% chance of getting it right by guessing. In this case four "listens" constitutes one trial.

Even a cable enthusiast like me would like some help by putting contrasting cables within a few minutes of each other.

I'm not using an ABX box because I don't want to put the extra contacts in the way and I don't believe quick-switch is the way to go. Therefore I need a friend to connect/disconnect cables. I need to listen to each configuration up to 10 minutes. As you mentioned, I need a patient friend even to do four listens. If I were to ask my friend "Okay, hook up A again... okay hook up B again... okay hook up X... okay A again..." it would take forever. So I am trying to find some way to order the trials so that in the fewest possible we set up something that has some guaranteed back-to-back presentations.

Note that this problem of "expectation when switching from A to B" is actually present in an ABX test! (every time you switch from A to B). An ABX test also guarantees that you can hear X and something different than X back-to-back (by switching A-X-B-X).

Think of this as having some of the same goals as ABX, only attempting to compress the necessary number of listening runs.

Unfortunately, if I'm not sure of an ID, I am not forced to answer, which means the test could go on longer than planned.

 

[mike1127]

Quote:

Originally Posted by nick_charles /img/forum/go_quote.gif PS nice hamster !

Thanks! It's actually a rat. Rats make great pets. They are bred to love humans and have good temperament and cleanliness. This rat, named Jacy, has dumbo-type ears.

 

[nick_charles]

Quote:

Originally Posted by mike1127 /img/forum/go_quote.gif Thanks! It's actually a rat. Rats make great pets. They are bred to love humans and have good temperament and cleanliness. This rat, named Jacy, has dumbo-type ears.

Nice rat !

 

[xnor]

Hmm, I'm also not sure if this really works as expected.

When doing an ABX listen, you have a probability of 0.5 to guess right/wrong.
So you'd need 6 listens for the probablity of guessing it all right to sink to about 1,5%.
Listening to A, then X, finally B should be enough actually to decide if X was equal to A or B.
That would be 30 mins per listen, and 3 hours for 6 listens.
With AXBX: 40 mins or 4 hours respectively.

With your protocol you have to listen to ABAB or ABBA = 40 mins
Assuming that p = 0.5 for one "trial", you'd still need 4 hours.

Switching wise, your friend needs to switch as often for ABAB as for AXBX.
In the ABBA case, your friend still would need to switch as often (unplugging the first B, then plugging it in again), else there would be a severe flaw in the test.