The heart of the design is pretty normal I think. It's only the power supply that is a bit "more" than usual. I think Klaus' amp has good opportunities to work good. but as always compactness is good, less electrical and mageitical fields to pick up.
I can’t say I agree with the circuit concept but I am amused by the idea of “tweaking” the implementation – for such an “extreme” design many tricks have been missed
Starting with the cascode, bipolar is good, Baxandall/Super pair compound cascode could be better by reducing bc550 hfe variation effect on signal – continuing the idea of improved cascodes, the tail current sink cascode fet gate should be attached to the bc550 emitter (ok some rf stabilizing R-C snubbers may be required but the af performance gain is worth it)
Referring the RIAA network and output amp to gnd seems to miss an opportunity that xfmr output allows – the signal from the input pre stage is referenced to the positive rail, either use folded cascode to refer it to gnd or change the reference for the composite op amp to the positive rail and let the xfmr’s isolation refer the output to the system/output gnd – there is less need to for heroic supply noise reduction if you can remove a major supply coupling mechanism (of course this option adds a additional V++ supply for the op amps, but the improved compound cascode performance may not require 35 Vcb on the cascode Q)
The composite op amp certainly can be improved with a better output amp – time has moved on, many superior dsl amps are available now (ths6012/tpa6120 if you can solder it!)
Also the designer apparently misunderstood op amp +,- input impedance matching for distortion reduction; don’t increase impedance! Lose the 1.5k Rs, change the 2k to 4.7k and parallel them with ~400nF to more closely match the + input RIAA network impedance vs frequency and to reduce noise/phase shift in the feedback path at high frequencies
Another potential place for improvement is using the xfmr pri R canceling negative impedance “mixed feedback” technique that Walt Jung published in “Op Amp Applications” to greatly reduce low frequency distortion common to xmfr coupled circuits – implementing this technique around a Av 500 stage may be impractical tolerance wise but the inverting unity gain balanced version Walt shows can’t add much distortion and allows the much needed AC coupling between the high gain (= high DC offset) stage and the xfmr pri to reduce distortion producing DC offset/primary current
1) Could you point out what you don't like from a conceptual point of view?
2) Baxandall: made a quick browse through diyaudio, and it seems to be difficult to make it stable. Could the fet stay in that case?
You mentioned have a several trials to turn out a satisified result. Did you alreadly refer to the last version that the designer posted ? If so I want to know where I should be care and somewhere that I will missing.
Where to buy the output LLxxxx output transformer ? May do few modification on circuit to skip it ...
Yes, latest design. (not that that jcx refers to)
try to get it look like on the photos on that web page. Where possible, I snipped the resistorlegs and soldered resistor bodies at the underside of the board, doing sort of diy-smd-ing between op-amp pins.
Ultrashort signal- and decoupling pathes. Place the electrolytes *between* both balanced circuitparts per channel.
Mind low impedance for current returns ('ground').
Begin with psu (test), then output stage (test ) and then input stage.