I understand how to do a bypass test with preamp, but what's the method for a power amplifier?
You load the output with power resistors. Connect with a couple of voltage divider resistors so the output thru these dividers is the same as input. In the case of the Spectral I think it was 20 db of gain. So the dividers reduce amplifier output by a factor of 10x. So one volt in would get one volt out. I at one time had a matrix of switched resistors so I could quickly change this precisely with a few switches.
I had a pair of interconnects made to connect to these and feed another power amp. So you could use your usual amplifier and put another one between source and power amp. You could compare the effect of two different amps or an amp vs wire. Though unknown to me when I first did this the Swedish AES uses the same kind of a test. They've gone further and load the amplifier with a simulated loudspeaker load instead of just resistance. They also do sighted and then blind comparisons this way. I don't know the latest results, but at one time not long ago they'd only had either two or three amps pass the sighted test, and one (Bryston) pass the blind test as being transparent. They never tested a Spectral to my knowledge.
This was the most revealing listening test method for amps I've done or know about. The Spectral had no sound I could detect (mine was sighted only). Most amps had a little coloration. Some had lots. I used Quad ESL 63s at the time I've done these tests. Its been awhile since doing any more recent ones.
One of the most interesting things to come out of it to me were the sound of triode amps. They seemed to be more 3D, more airy, more real than say the Spectral. And tube lovers of course believed the Spectral incapable of making this sound. I could use the Spectrals on the speakers, and place a triode tube amp between source and Spectral. The result sounded just like the triode amp. Indicating it was all a coloration. When I initially did this my expectations were that inserting the Spectral would kill those extra qualities of the triode. I was trying to determine whether an SS amp lost 50% of the resolution or 75% or maybe only 20%. Turns out nothing was lost and all that extra "resolution" was actually an additive coloration. With the added coloration at the input the Spectral could create exactly the same sound.
People who like SET's or other low powered tube amps could use this approach to get the sound they like and feed it into beefy amps so that they aren't restricted to efficient speakers. I've had a hard time convincing them the qualities they like are colorations that a good amp can faithfully reproduce.