Originally Posted by JWolf
Sorry, but you people are wrong. Different types of cables can sound different. You may not be able to measure this, but you can hear it.
Then you should be able to prove it with a reproducible test.
Science is hard, but we've had several hundred years of practice doing it well, so we have a whole list of things required to pull off a good test of this sort:
- Specify the products to be tested precisely, so that others can build or buy identical replacements, within a reasonable amount of measurement error. Specify those measurement bounds. Example: Cable A shall be made of 6.0 feet of such-and-such cable with a Brand X Type Y outer jacket, soldered to a thus-and-so connector with type Whatsit solder, having 0.0032 Ω of resistance and 123 pF of capacitance +/- 0.1% as measured on a calibrated 5 ½ digit HP 34401 DMM. Cable B shall be identical to Cable A except that it uses 6.0 feet of this other inner cable type, giving 0.0033 Ω of resistance and 234 pF of capacitance. Test to be done with a Brand Q Model P headphone amplifier with the Bizzabong upgrade.
- An adequate sample size. "I made my aunt listen to both cables, and she liked the pink one better" is not good enough.
- A double-blinded test. This means you cannot experiment on yourself!
- Publish your data. "My buddies all sat around the amp smoking unspecified herbal products and decided cable A sounded better than cable B" is not "published data." How many participants did you use? How many separate tests did you subject each participant to? What music did you use? What was the volume level of the amplifier? (Which must be given in some reproducible measure, such as Vac RMS of a full-scale test signal at the test amp's volume setting.) Most importantly, what were each participant's responses? Without the raw data, we cannot know the data's p value.
- Select and specify a randomization method. For a psychoacoustic test such as this, I'd suggest ABX testing. If you like some other method better, specify it, and be prepared to defend your choice rationally.
- Independently replicate the test multiple times. If two groups come up with different results after following the test protocol you get from following the above points precisely, the test is not reproducible, and is therefore scientifically invalid. There are many reasons this can happen, all of which mean the original test did not prove what it claims to have done.
- Should you manage to get completely through that gauntlet and still have an independently reproducible test difference, vary one detail of the test, then go back to step 2. For instance, if the difference disappears when you don't use the Bizzabong headphone amplifier upgrade, that suggests some interaction between the three items under test (one amplifier upgrade, plus two cables) which is responsible for the difference, which calls into question the difference in the cable being the reason for the difference in the test results.
I'm probably missing something important here which someone with expertise designing scientific test methodologies could point out. The important thing here is not to make the test pointlessly difficult, but to increase its statistical power sufficiently to overcome the psychoacoustic difficulties inherent in the test. With insufficient statistical power, we must disregard the test results.
On point 3: If your only aim is to please yourself, that's perfectly fine. I am a committed DIYer, which means I cannot honorably rail against any person who decides that they need to build a $500 headphone cable in order to achieve happiness. Just don't try to make me believe, should that cable indeed induce happiness, that your "test result" is a scientifically-useful data point. Go ye forth and listen to your headphone cable. I sincerely wish you all the enjoyment you can squeeze out of it.