VVR issue - voltage drops under load

[R.G.]

Yep. And building on that, one could protect the pass device with a modest heat sink that had a thermal switch on it that opened the high voltage to the whole mess when it got too hot.

That's about as far as the design goes with only stone-axe tools and techniques. It's possible to do it much better in the technical performance sense, but that probably exceeds the typical amp-hacker's experience and patience.

[Zozobra]

Some interesting points thrtr. I have access to a CNC mill so I was planning on making a few generic PSU PCBs so I may well add in the full kitchen sink voltage scaling into them, or at least make a seperate board for it.

The pot was bad & I've replaced it with a 24mm Alpha. This has cured the loss at the wiper but I'm still seeing the reduced output at 100% rotation so I guess the MOSFET is damaged

[J M Fahey]

I'm still seeing the reduced output at 100% rotation

Numbers, please

Even better: repost your drawing with new voltages and pot `set to 10.

Even much better: repeat with pot`set to 5 . Post both sets of values.

[Zozobra]

I should have a bit of time tonight to meter the voltages
The output was still ~295V at 100% rotation from memory.

[defaced]

Ah. I just misconsidered it. That is indeed a very crude way to do current limiting. There are much better ways that don't destroy things or rely on the imponderables of the zener and MOSFET.

A more predictable way to do this is to use a current sampling resistor there where the 20R resistor is, and to use an ordinary NPN transistor with its base to the MOSFET source and its emitter to the outside end of the sampling resistor. The collector connects to the gate circuit of the MOSFET to steal gate voltage. The NPN turns on somewhere between 0.45 and 0.6V reliably, and it eats all the gate drive by shunting the current on the MOSFET gate to the output. With any given NPN, the threshold is predictable and there's a lot less wobble in guessing the MOSFET turn-on voltage. This makes the 20 ohm resistor value change to something like 1-2 ohms, and the threshold, once adjusted for the exact NPN, is quite reliable. I used this for current limiting to solve overcurrent issues when switching out of standby.

I didn't originate this, I adapted it from a circuit in the ARRL handbook from about 197x? and it was old then.

This addition just makes it predictable, it doesn't keep the output device from frying. But there are other ways to do that.

Simplicity is good - however, as attributed to Albert Einstein, everything should be as simple as possible - but no simpler.

I had this floating around on my desktop. It looks like it contains what you're describing with the addition of a pot labeled "sag". I submit is for consideration, though aside from saving it from some site (probably this one), I have no connection to the design.

[Zozobra]

Here is the updated schematic with the 100%/50% rotation voltages noted. My apologies for the delay - I've been away flying with work and I haven't been around along enough to make the measurements.

[Zozobra]

Replaced Q1, still lose 100V across R6

[J M Fahey]

1) Are you sure the Zener is connected the right way?

In any case, take it out of the circuit (we'll worry about current limiting later) and remeasure at 100%/50% rotation.

2) are you using a Log/Audio 1M pot?
50% rotation *should* give you about 50% voltage on a linear pot.

3) there's about 1mA passing through R6 ... why?
It can't be going into Q1's gate.

[Zozobra]

The circuit behaves sensibly with the Zener removed. Out of curiosity I swapped the Zener orientation and this dropped the output of the circuit to ~25VDC.
I did notice that B+ out of the main filter cap was much higher than expected and it turns out that my outlet is reading 293VAC. I'll take the requested readings once I get a sensible outlet voltage again.