I really appeciate the idea. In the short time that I am readig this forum, I have seen that much concern is taken for THD, then less for offset, but very few for noise. If I remember well the noise should be indeed of concern specially with FET transistors that should (correct me if I am wrong) be affected of 1/f noise. Also I never have seen explicitely (maybe it was done anyway) dimensioning resistors for noise avoiding. I mean trading a higher dissipation with a smaller R. This is done always when you design measurement instruments. Sometimes I had also the impression ( i should research on that) that sophisticated schematics (i.e. cascoded and double CCSs, especially Zener ones) besides of the linearity (read THD) advantages may bring in additional noise. Very interesting post
R1 = 11K, R2 = 5.6K for 1ma at +-12V, caps about 10uF, R3 about 100, R4 I’d just use a 1M pot and adjust it to a good level since you are just comparing transistors and don’t need to measure actual noise levels. Measure with a scope or a volt meter. Scope is better.
* There is voltage and current noise. You want to test for both unless you're sure one is negligible for your application.
* To measure noise you want a well-defined bandwidth which is meaningful for your application (e.g. 20 kHz for audio). In other words: you need a low-pass filter somewhere.