The amp arrived with one of the two IRFB4227 power transistor cut out of the board and the holes for it desoldered. I ordered a few of that replacement as well as some of the main switch-mode power supply transistors and the main bridge rectifier, just in case. Some folks on forums like talkbass have suggested that the IRFB4127 could be used in place of the 4227, as it has higher RMS and peak current specs, but the turn-off delay is longer by quite a bit. Either way, the replacement transistors arrived and I set about replacing the known-bad one and figured I'd go ahead and desolder the other one while I was in there.
This is not a good picture of the power MOSFETs on the far left side of the chassis. They're a bit of a pain to get to in situ, so the easiest way to deal with it turns out to be trying to loosen the hold-down bracket and removing the heat sink. That may require removing the big heat-shrinked capacitor right in front of the middle screw; the amp came for me without the middle screw in place, so I don't know how the PO or his service tech dealt with that problem, as it also didn't look as though the cap had ever been desoldered. Maybe needle-nose pliers? An about 5mm deep 90-degree PH1 bit? I don't know.
Here's a shot of the switch-mode MOSFETs. The bridge rectifier is behind that. These are hard to get to for unscrewing, but a 10mm right-angle driver should be fine. These are actually working well in my board; the power amp disaster did not affect the power supply's outputs.
Worth noting - my transistor tester reported that the still-installed output MOSFET was still good! I'm not sure how that happened, but I saved the part just in case.
After installing the new parts, the amp powered on and passed its self-test. No sound output, though. Next step came testing to make sure the power supply was actually working. I got a case of the apprentice shakes and blew out both replacement output MOSFETs and let the smoke out of one of the diodes nearby.
Next comes trying to trace the signal through the head. I did NOT feel like trying to tear out the input board, so I started at the beginning of the main board.
Input of C33: couldn't find it.
Pin 7 of U18: success! 2kHz right there.
Pin 1 of U18: 2kHz once again.
Output of L7: a small DC voltage offfset. No sine wave.
Okay great, that's a big swath of components that could have failed, but at least the two dual op-amps were good. In reviewing the schematic for potential test points, I noticed that there is only one more place that the signal actually passes through on its way to the output MOSFETs, and that's an IRS20957 audio IC.
That's where this repair job is right at the moment. I suspect that people who have PF-500 heads that have sound that "cuts out" but no fault light to go along with it are probably dealing with this IRS20957 IC failing intermittently. What's interesting about this amplifier's self testing is that it isn't failing on whatever is broken in that chip, at least not consistently. About half the time I had my bodged test lead connected to either the main instrument input, the preamp output, or the power amp input, the amp would fail to start. I never had any problems with it starting without the test plug inserted, and the preamp board was clearly okay with the 0.2V input amplitude, as it passed it undistorted to both the main board pin 7 input and the tuner out jack on the output board. Is it possible it was faulting because of additional load on the IRS20957? Who knows. I do know that that heat sink is hilarious. I also know that, at least according to the product example picture at Full Compass, the revision J board (mine is a revision C) doesn't even have a heat sink on U1.
I ordered a replacement IRS20957 and a few more IRFB4227s just in case I pooch up this board again in testing. The solder pads for the outputs are getting very tired, and the last time I tried to do a SMD (installing a pair of THS3092 opamp ICs into an MHS-5200A signal generator in place of the AD812s that came in it), I managed to destroy three of the sixteen traces on the board and had to do some very careful scrape-and-bodge work on the board; thus, I don't have a whole lot of faith in getting to where I can call this rev C board a finished, working, and perhaps most importantly, reliable project. Fortunately, that Full Compass link up there exists, so I have a (hopefully rev J) replacement main board on order to put in once I've finished screwing up the one currently installed.