Okay, made some progress, I remembered that I bought a bunch of pre-made variable LM2596 boards off fleabay a while back. I fished one out of my parts bin, desoldered the trimpot off of it and modified the board to match my feedback circuit. Even went to the trouble of replacing the diode with the one off my own board. What I wasn't able to pull across from my board were the BC857 (used a BC557 since they're nominally the same thing) and the switching components as they're in a different footprint.
Sure enough, the damned thing does what its designed to do, right up to 21V or so
.
I fear, that leads me back to one of a few possibilities either:
- My board layout is naff (it could be better, but I didn't think it was that bad)
- My selection of switching components is off
- I've got some damaged silicon, either the transistor or probably more likely the regulator
- There's some weird behavior happening due to the rest of the circuit (should have been ruled out by lifting R3, but it's still possible
I do have enough parts to hand to quickly throw the switching section of another board together which will give me an isolated test. I did test this regulator when I first built it but admittedly not over its full range. Think that's probably my wisest course of action at the minute.
What lies beyond Q3 and R3?
The output goes to the input pin of a LM317 and R3 goes off to the output of the same. Don't think the subsequent circuit is to blame as lifting R3 and driving the base with a voltage yields the same results.