Help solve a small controversy - research design help

There's an interesting controversy in audiophile circles.

I want to compare two groups of subjects. The object is to see if subjects can distinguish between two almost identical headphones - one new, the other "burned-in." (There's currently no empirical research on "burn-in" or "break-in" of stereo headphones, but a blizzard of conjecture.)

Some posters to the audiophile forum have demanded an "ABX" design, with a double-blind. This cannot, however, be a costly study. It looks like the purchase of two or three headphones will be necessary, however.

Which kind of design is most desirable? Which statistical test? How many subjects would be needed to test it fairly?

Thanks in advance for your expertise.
hm..sounds familiar, what textbook do you use? ( hint:sometimes you can find answers in solutions).
If it's a real life problem I would do as follows:
1. Get 2 sets of heaphones new/old.
2. Get at least 30 people in the study
3. Without telling them which one is which test their perception ( or whatever you want to test).
4. Record data ( I'd use simple yes/no format, my questions would be-is it new or old set)
5. Set Ho( there is no difference in perception) and rejecton rule.
6. Use z test for differences in two proportions ( some excel add-ins have it, or if you have more advance stat package)
7. Draw conclusion.
Good luck
Refinement or corruption?

Thanks for the feedback so far.

what textbook do you use?
Actually, there's no textbook. After reading this procedure, I wondered if another option might be possible.

1. Give all subjects a heariing test using freeware ToneGen.
2 Give all 30 subjects a "control" run between two identical headphones, YES/NO format. (7 different audio samples, 10 seconds each)
3. Inform subjects that their previous run was for control purposes. Now, have subjects try to distinguish between headphones for real. (same audio samples)
4. Statistically test.

Is there a problem that the subjects are first used as controls?