umm.. Am I missing something here about the null test. But what about the accuracy of the amp being used to amplify the difference signal??  Surely it's own distortion and frequency  characteristics are going to have an impact on the amplitude of the difference signal. EG if the actual level produced by the difference between input/output of the test amp is -60db at a particular frequency and the amp being used has a +10db frequency response at this point the level will now become -50db. God knows what random affects distortion and phase characteristics of the "amplifying" ampifier will have.

Am I just being stupid here?

GTM