As I have explained many times, the problem with "scientific" A-B amp testing is that the amplifiers are connected to a common source and then into a speaker switcher, which can switch either amplifier into the speaker load.
What the ABXers fail to understand is that the distortion characteristics of an amplifier are reflected back into its input jacks because the feedback loop contains these distortions. Since the inputs of the amps are being fed from a common source, the sum of the distortions of both amplifiers are present at the input of either amp being used. Thus essentially both amplifier under test under these conditions will sound the same, and generally the overall sense of the listening session is unsatisfactory, for good reason.
When we do circuit design listening tests, we build one channel new, one channel old, (matched gain of course), set a pair of matched speakers side by side, randomize the connections, and swap back and forth with the balance control. Nobody, including the tester, knows which channel is which until the tests are done.
We ask ourselves first the question, is there any difference? Second, is there a better-worse difference?
Almost always we can hear the differences, but the results are not always what we would have predicted.
This process eliminates the ABX common input issue which has clouded so called "scientific" listening tests for years. Of course the ABXers don't believe us, but they are too pig headed to make the simple measurements to the amplifiers to show themselves that the distortion is indeed appearing at the input jacks. It is.
Frank Van Alsitne