Hi kleinstein, nice to read you again and thanks for your feedback!
However the transistor T4 looks a little odd. I kind of understand that it helps to provide a little extra current from the OP if the voltage is high. So it might be OK, but it's unusual. I don't think R24 parallel to the transistor is optional.
I used T4 previously but now I'm not populating it since 2 revisions of the board. In fact, T4 is optional and R24 is mandatory. Until now, I didn't see any issues with this.
The most critical would be likely a low output current and with a low ESR cap (e.g. 2-20 µF film type) at the output.
Will take look if I have one, but more likely I'll find a ceramic type or low ESR electrolytic one in my parts collection... Nice idea!
The maximum output voltage is rather limited due to the low supply to the OPs. Why not give at least 18 V to at least the critical OPs ?
That's due to the specs I decided to go for. First I wanted 0-30V but soon realized that this would need a quite larger amount of effort to build. So I finally decided to go with 0-12V and that's why everything is how it is. I use a 2x9V toroidal transformer which gives me exactly the voltages I need to keep this working at high currents and even if mains voltage drops within a certain margin.
If I make another PSU at some point, I'll probably go with 0-24V but then I'll have to overthink some of the parts and probably also do automatic transformer tap switching or implement a tracking (switching) pre-regulator.
I guess I could have specced everything up to make it go to let's say 18V or something quite easily but I didn't see any point in this (at least for my use cases) so I decided to go for plain 12V (however it goes up to about 13V even under load anyway).
However the current regulation part looks like it could be rather slow (because of the may4080) and possibly could oscillate. Here the critical load would be an inductive load, like a transformer winding at low current.
Yeah that's true, I wouldn't try to use CC mode for powering a single LED. The LED probably would be dead until the current limitation kicks in. It reacts in somewhere about 2ms which is rather slow. I was wondering why the MAX4080 is so slow and whether there are faster ICs or whether this would be faster with descrete opamps. Also the output cap slows CC mode down. I'll happily take suggestions on how to make this faster (however I'd love suggestions which don't involve very big changes even more :-D).
I'll try inductive loads, didn't really do that yet.
With the 2 extra minimum current sinks, the current regulation is also not really precise, as these currents are also included and especially the sink with T8 is not very stable and the sink with IC9 (LM334) may not work well at low voltage.
I only use the one with the transistor and the zener because the LM334 didn't work well at low voltages as you correctly identified. I did leave the footprint in the design to be able to choose but it is unpopulated as of now. The zener gets quite hot, around 70 °C but it is stable there.
Yes, it is not very precise. CC mode or current measuring is a few mA off always, but most of the time about 10mA. Suggestions very welcome.
The slow current limit might not be a sufficient short circuit protection.
I'm not sure whether I tested this properly but shorting the outputs at various voltages didn't seem to do any harm to the PSU. It went into CC mode, at lower voltages obviously tended to heat up more and at low voltages eventually turned off the outputs due to overtemperature condition (controlled via software, measured via NTC) which is currently set to 80 °C at one of the TIP3055's.
At higher voltages (like 4-5V), where the NPNs don't have to burn so much power, the PSU stays happily for hours with a short circuit on the output.
0.1 Ohms emitter resistors for the 2N3055 may be a little on the low side and could require matched transistors to get good load sharing. One might still get away with it at the low voltage. The usual value is more like 0.22-0.33 ohms.
Maybe I was lucky but in the last 3 iterations of this, the NPNs got nearly identically warm. Sure this will also be due to the shared heatsink so they heat each other up but during fast temperature risings (faster than the heat can spread through the heatsink), they were nearly identically warm.
In the latest iteration I even used 82mOhm resistors because the 100mOhm ones weren't available. Until now, no issues as it seems.
I didn't want them to heat up too much because of temperature drift and even more waste of energy and also because it is hard to get high power resistors with a relatively small footprint.
But I'll keep that in mind, if I ever encounter issues with load sharing.