For the VAS, 2SC3648T and 2SA1418T will perform about the same if not better than 3503/1381 but they are NOT TO-126 so the PCB will have to be redesigned.
I hope these aren't EOL yet as I plan to use them.
For the differential 2nd stage only, KSC1845-FB/KSA922-FB are better than anything else I have run in a sim - provided source and sink currents are kept below 4mA otherwise they will get hot as they are TO-92 and only 500mW and 50mA max but they have the lowest capacitance and very high Hfe (>500) which is awesome for a BJT that can handle 120v.
TTC004B/TTA004B have excellent Hfe linearity but they don't sim as well as 3503/1381, I suspect because of the higher capacitance.
If you can get your hands on KTC3206-Y and KTA1024-Y from KEC, these come pretty close to the 3503/1381 but I haven't run them in a sim yet.
If running at less than +-40V rails, BCX56T-16 and BCX53T-16 might be good too.
Last edited:
They are all the same except they are binned for voltage offset. BFQ10 is the best with 5mV offset and higher CMRR due to tighter matching but these are ancient unobtanium, I wouldn't use them unless I had heaps of them.
Any of the SST/U401-404 as well as LS844 series will be better and Jfe2140 are better yet but watch out as the pinouts/package aren't the same and you will need to adjust the compensation and or degeneration around the LTP especially if using Jfe2140 as it has higher capacitance.
Last edited:
You can refer to the excellent simulation by sandrohv on the honeybadger thread. I think many would have seen this, its a really good and detailed one on lot of stability and loop gain measurement tricks for LT spice.
https://www.diyaudio.com/community/threads/sw-vfa-01-audio-power-amplifier-video-series.347673/
https://www.diyaudio.com/community/threads/sw-vfa-01-audio-power-amplifier-video-series.347673/
I simulated with MJE15030/31 with 75pf compensation cap.The performance isn't great. The phase margin at 20Khz is around 35 degrees which is below the nominally acceptable 45 degrees, although gain margin is around 40dB. With slightly smaller compensation capacitors it improves to barely 40 degrees of phase margin. Have to simulate with TTA/C 004
Sasmit, Cob is what's hurting your phase margin, the TTA/C will be a lot easier to compensate, Cob of the MJE15030/31 parts is around 10 times that of the TTA/C and 35-40 times that of the A1381/C3503.
You might need to play with the feed forward cap and also try more tail degeneration in the LTP and 2nd diff to get more phase margin.
You might need to play with the feed forward cap and also try more tail degeneration in the LTP and 2nd diff to get more phase margin.
Sorry to say that but you don't seem to understand what the gain and phase margin are, nor why they are important, so your conclusions about stability are wrong. What happens at 20kHz or 30kHz is of no interest from the stability viewpoint. What is important is the phase at the frequencies where the loop gain is around unity, say -10dB to +10dB - you want at least 30° phase margin there.
I have tried to create a simulation of the original AEM6000 (attached). I could not find a model for the BC639, so I have substituted 2N3904/2N3906 for the LTP current sources and the cascode transistors. At 1kHz the simulation reports distortion close to that reported in the original article, but it also shows an order of magnitude more distortion at 20kHz. That is, the distortion does not stay flat out to 20kHz as the original article reports. My assumption is that the MJE340/350 really don't spin fast enough for excellent high frequency distortion performance.
M
My simulation with the MJE15030/15031 was not very encouraging, the loop gain measurements were much worse compared to faster devices like the Toshiba TTA/C 004B I tried later. Cib/ob are a major factor in loop gain performance I think . With the TTA/C 004B at 20Khz distortion in simulation measured 0.018 % at 1khz it was 0.006%. This was with +/- 43 V supplies which would equate to a ~30v secondary transformer. It improves with +/- 50V marginally.