Setting current limit of a linear regulator?

In the attached linear regulator schem, why are they setting the current limit of Q7 by varying POT1? Isn't that a dodgy way to do it?

BTW, xtor P/Ns are incorrect...all i know is that Q7 is an MJ15022G...

MJ15022G datasheet

(The schem is a copy of the schem that i had (and its a real product, in use in industry)...but the schem had crossing traces, and it was impossible to know which traces joined at the crossings, and which didnt....so my version of it, attached, my not be a perfect copy)

BTW, actual product features nine Q7's in parallel.

BTW, this is a real product in use in industry

They are in effect sampling the voltage across the emitter resistors, and that should be proportional to load current.

It will have one diode drop added (Q7 B-E), which will approximately cancel the base-emitter voltage drop of Q5, so it will allow adjustment down to much lower levels, near zero, instead of having a 0.6V threshold before it could limit, as it would be if the sample voltage was from directly across R29.

Looks fine to me, a bit of lateral thinking to give a better control range!

If there are multiple Q7/R29 combinations it also avoids the situation where if one transistor failed (so no current in its emitter resistor) and that was the one being directly sampled, the current limit would be inoperative.

Agree. Nice twist on a very common technique.

This is called high-side current sensing. To make it adjustable, it often is necessary to run all of the output current through a high-power pot. At a much lower power level, this is the method shown on the LM317 datasheet. Messy at 1 A, and extremely difficult at higher currents.

With two diode drops, and one of them at widely varying temperatures, the current regulation "knee" might be relatively soft compared to other techniques. Still, a nice compromise that doesn't require a rail-to-rail opamp or instrumentation amp. I'm definitely gonna use this somewhere.

And . . .

There are only a small number of "standard" ways to regulate a constant voltage linear power supply, and most of them are not this. Notice that the GND for the control circuit (the bottom end of V1) and the GND for the power output (the bottom end of V2) are not tied together. The control circuit rides on top of R7 and R31, fed by constant current source Q2. At the concept level, this regulator is a high-power version of an LM317 (as opposed to a the more traditional approach of a 78xx type).

ak

Thanks, do you agree that however POT1 is adjusted to give the current limit level....when Q7 gets hotter, that current limit level will rise?

Yes, but the solution is simple, an dates back to the early 1960's with the first solid-state audio power amps. Locate Q5 off-board, physically and thermally near Q7. The two Vbe's never will be equal because of the difference in structure between a power transistor and a small signal transistor, but that is part of what R19 is for. With nearly identical thermal coefficients, the two transistors will track and the setpoint should be relatively stable. In an LM317, both transistors are on the same die, and track very well.

This never will be as good as a "lab" supply because the two transistors will not track perfectly. A lab supply gets around this by not using unmatched transistors as part of the current sensor; it uses perfectly matched transistors in a differential pair: a single current shunt resistor that is not a part of the output transistor balancing and has a very low temperature coefficient, and am opamp or instrumentation amp turning that signal into something that can be compared to a stable voltage reference.

ak