Thanks Sebastian, I was beginning to feel that I was talking to myself. Odd that there's so little apparent interest, this is potentially a very high-performance design at a very reasonable cost.
The first thing I've done is measure the DC offset. I say measure, both my digital meters have a lower limit of 1mV and both read zero. I can't find any unshrouded probes to plug into my old analog meter.
<1mV is good, it means the servo is working OK, and the rest of the amplifier obviously sounds OK
Anyway that meant I could plug in my sensitive IEMs, Shure e2c's. Now I can hear a very quiet 100Hz hum from the fullwave voltage doubler with nothing playing, it's totally inaudible even in the quietest passages with music playing, but I may be able to do something about it anyway. All I had in SMT diodes to build it with was ultra-fast Shottkys which are more likely to cause radiation than slower ones and if conventional rectifiers don't do the trick I can try strapping a cap across them.
Now there's more volume than you can really use. I have a couple of jumpers in there which double the gain, they can be removed. I'm mostly driving it using a Sony NW-E003 Walkman anyway, they are notorious for their weedy output. I may wire the jumpers to a panel switch on the rear panel, there's room.
OK, take it all out of the box again and try to figure out the relay transistor...
w
Edit: found the relay transistor fault, problem with the footprint. I had to skew the transistor to make it work, but no big deal. About 4 seconds delay before switch-on, no pops on switch-on or switch-off.
