3metrejim wrote: ↑Thu Sep 06, 2018 1:32 amCould I get the instability problem if the circuit ground is not too good?
Yes, definitely. It's really important to have a good ground, on PCBs and throughout the whole rig. These devices these days have stupid gain, right up to GHz in some cases, so any stray reactive components due to the layout of the PCB need stamping on, and that's why you'll see so much "stitching" on RF PCBs. I've done simple boards - one in particular involved an RD06 being driven by a "BFG" bipolar and and lost count of the number of iterations I had to do of the layout to get the thing unconditionally stable!
I don't like the way your board is connected to ground only at the corners, and from what you've said I'm guessing that the PCB around the FET isn't grounded properly as just the FET source is bolted to the heatsink in this area. If you look at some proper pallet amps, they usually come with the FETs bolted to a spreader under the PCB and then the PCB also bolted to the spreader in quite a few places, and loads of stitching all around. Google "WRF-350H" for an example. There's a good ground from the FET source to the rest of the PCB via the 4 screws holding the board to the spreader closest to the ends of the FET.
That Chinese board has a relatively poor top ground plane, and only 6 plated through holes for bolting down to the heatsink, none of which are that near the FET. So, what I did was to ensure a decent connection from the FET source to the PCB by passing at least one of the FET screws through a solder tag, and soldering the other end of the tag to the top ground plane of the PCB. On the chinese boards I've got here, there's a stitched through piece of ground plane at one end of the FET (bottom of the board in your picture) which is ideal and possibly there just for this purpose. If you don't get what I mean here let me know and I'll post a picture later on.
Looking at the second photo you're right, the ground could be better overall, he's used the case but it sounds like there's insufficient earthing around the PA. I always have a fat wire from the PA PCB back to the PSU to guarantee a proper DC path to ground - in this case, the pad right next to Vcc which it looks like he hasn't used. I don't like using RF earths for DC returns. I'd tidy up that positive supply wire away from the output transformer too.
So - sort out your ground, definitely make sure the FET source is connected to that ground plane on the PCB and try to bolt the board down through more of the plated through holes. I don't know if you've got the circuit for that PA, I think there's one on the net somewhere but check it and in particular check that all caps to earth are present and working - supply decoupling and decoupling at various points in the bias circuit. See how you get on with that lot.