I've been looking at the PSRR of the EQ stage, and simulated some different variants, about 8 so far.
To get a big improvement, you have to go to a much, much more complex design, and I am very wary about that - the opportunities for trouble outweigh the gains. The stock design has an adequate PSRR, and is nicely simple. If anybody wants the gory details, I can explain what I looked at, but the upshot is that the transistor count goes way up to get a big win.
So my engineering choice is to keep the signal path simplish, and accept that the power supply has to be vaguely quite; it needs noise in the audio band to be below a few mV. You can't quite do that practically with a purely passive design, but even very simple regulation can deliver it; something like an LM317/337 would be way more than good enough, or a simple VBE or whatever.
If the +/-24V supply is not completely horrible, a VBE style divider (amplified capacitor) simulates as doing very well for the head-amps, in all 3 variants. Depending on which head board(s) you have, the total draw for a stereo setup is round about 0.1A per rail, say 5W in total.
My suggestion for power supply (unless you have another device running on +/-24 V that you want to piggy back off) is as follows:
1) Remote box containing moderate size (say 30VA) toroidal transformer with 18-0-18 secondaries, rectifiers and a CRCRC filter made up of 2,200uF caps (35V or higher please) and 4.7Ohm resistors. The resistors will drop about 0.5V each, giving a smoothed output at about +/-28V, with perhaps 50mV peak to peak of ripple. This gets fed, via a 3 wire cable, to the actual phono unit.
2) The EQ board has the 317/337 pair on it, and a couple of holes to allow people to feed the +/- 24V to the head board(s).
3) The head boards contain their own local VBE circuits.
What do people think of this as a plan?