I have been using my [HPSDR setup] (https://rkrishnan.org/posts/2020-03-30-back-on-the-air.html) for the past few months. I had to clear the desk one week to de-dust the table and when I put everything back, the Amplifier won't transmit.

After much head scratching and poking with oscilloscope, I figured out that the Transmitter board (code named Penelope) itself is not transmitting. The output broadband amplifier OPA2674 is known to fail and indeed it had failed. So, I replaced it from another dead penelope I had and also ordered a few OPA2674 in the meantime as a backup. OK, so that worked. I even made some FT8 transmissions with ~300mW power.

Even with that fixed, the 20W PA wasn't working. Graham, the designer of the PA told me that there is no need to apply RF to test whether the MOSFETs are working. Just DC bias would do. So, I did the bias calibration again, i.e. keep the bias pots to the minimum and use an ammeter to measure the current to adjust ~450mA current per device. Reminder that MOSFETs are voltage controlled device. i.e. Gate voltage determines the V_DS. At some point though, the ammeter shows that the current jumps from 300mA to 1.5A. The transformers also get hot. That is strange because transformer secondaries are isolated with capacitors, so only AC can pass through them. So, this means, PA was sort of "oscillating" at some frequency because something strange was happening in the MOSFETs. I also erroneously removed one of the transformers wound on a binocular core, thinking that it was getting hot because of a short between primary and secondary. That wasn't necessary.

So, I ordered some RD16HHF1 devices from RF Parts, $5.75 per piece. They finally arrived yesterday, after one month wait.

Desoldering the transistors wasn't easy. Initially I tried to just unscrew every screws to the heatsink except the one holding one of the transistors and use a hot air gun to heat those legs and pull things down along with the heatsink. That did not work. I finally cut the device which left only the legs and then used soldering iron to heat the legs and got them out quickly. But that left a lot of solder on the holes.

Per Graham's advice, I used a de-soldering wick and generous amounts of flux paste to clean the holes. They look very good now. Between today and tomorrow in my free time, I am going to put back the transformer and put back the MOSFETs and check if the PA is working fine. EDIT: I put the MOSFETs in and calibrated the bias and it works great. Made a CW contact with 9M63MR.

A lot of websites and online "maker" community websites/spaces encourage and teach soldering skills. They are extremely important. However, de-soldering techniques are under rated. If you solder something, one day you need to de-solder it too, for repairs and experiments. So, learning those techniques are very important as well.