Sense pin functionality with LT3080/LT3083

[gxti]

I'm currently working on a beefed up bench version of Dave's uSupply. Well, the only thing it really has in common with the uSupply is the analog stage, but I digress... I'd like to see if I can add a separate sense return i.e. a Kelvin connection as part of the feedback loop. In particular, I'd love to move the current shunt downstream of the supply so that the dummy load doesn't conflate its readings, and to do that the voltage sense has to happen after the shunt to get precise line regulation.

It seems like the obvious way to do it is to make the voltage op-amp's feedback come from the sense wire, instead of the set pin. Is this a good idea? Does it harm stability? If it works, does it end up making the LT3083 just an expensive pass transistor?

The first picture is the original schematic which is more or less just a higher-current version of the uSupply output stage. The second one is the "improved" version. More illustrative than anything, I have no idea if it'll work...

[gxti]

Just today it dawned on me that I could plop this in LTspice. I rarely use LT parts so I tend to assume whatever IC I'm using won't be available, so when it is it doesn't even occur to me. Anyway.

I've set up a crude simulation and the good news is there doesn't seem to be a difference with driving the set pin statically vs. incorporating feedback from the output. The bad news is either way it oscillates when current-limited! I'm guessing this might just be a limitation of the LT3083 model, but here's what I've got. I seem to be analog-challenged so if anyone has tips on stabilizing the circuit (or the simulation), please share!

I'm beginning to think the solution with discrete pass transistor instead of LT3083 would be much better value (1/4th the cost probably) but my simulations on that front are even more dismal.