LR8 circuits with pass transistor - how is it supposed to regulate?

[Agent24]

I am interested in building an adjustable high voltage supply for doing some experiments with long LED strings.

It seems that the LR8 regulator + a pass transistor for extra current is popular in the valve audio space. I wonder if the same design would suit my purposes also.

However, I noticed something odd about all the schematics that people have drawn - the pass transistor is driven by the output of the LR8, and there seems to be no way for the LR8 to actually regulate the output voltage, given that it appears to receive no feedback from the output.

This is completely different to say, an LM317 with pass transistor(s), which I've always seen to have the transistor in front of the regulator, and the output of the transistor and the output of the LM317 tied together, so it can sense the output voltage as normal.

I just can't see how these LR8 with pass transistor circuits are suppose to regulate. At first glance it seems to me that it has no way to tell what the output voltage is, but, since these circuits seem popular and must work to some extent, I am surely missing something stupidly obvious.

I have attached some example schematics that I've found. Some use MOSFETs, some use Darlingtons or even a single BJT (e.g. https://www.nutsvolts.com/magazine/article/a-semiconductor-based-high-voltage-utility-power-supply), but it's all mostly the same thing, and none of it makes much sense to me.

[Kim Christensen]

I just can't see how these LR8 with pass transistor circuits are suppose to regulate.

They simply rely on the fact that the base-emitter voltage drop of the bipolar pass transistor is more or less constant at around 0.7V. So they just set the LR8 0.7V higher than what they want the output to be.
Voltage difference would be higher for a MOSFET, but the same idea of setting the LR8 higher than the output voltage still applies.
Yea, the load regulation will suffer a bit, but probably doesn't matter for many circuits being powered.

[Agent24]

So it works as I was thinking, the regulator sets a desired output voltage but has no real way to verify it - there is no feedback? In that case I suppose it'd work for driving LEDs or another fixed load just fine. But not so great for a more complex load that has varying power draw.

I found a single example of this circuit done the other way, I wonder why nobody else uses that design? The loads are always constant, so regulation is not so important? And the higher cost of some good high-voltage PNP transistors with enough gain and power just isn't worth it?

[Kim Christensen]

Well, that one says it requires a heatsink on the LR8 whereas the other circuits may not.
This is because to turn on the transistors, apx 2.7mA (1.4V/510) must flow through the LR8 first. Then you've got the base current on top of that. So if there is a large voltage drop across the LR8, then it's power dissipation could raise it's temperature enough that it goes into thermal shutdown.
You could reduce the dissipation in the LR8 by raising the value of the 510 ohm resistor to something higher, but that might effect stability. (I haven't analyzed that)