High Voltage Power Supply Regulation

[Stuggi]

I'm in the beginnings of designing a high voltage (<600VDC) power supply to measure leakage of capacitors in tube radios. The idea is that you apply a high voltage to the capacitor and measure the leakage in microA.

I've found a suitable transformer for the project that's 230VAC/400VAC and 50 W, which should give me around approximately 100 mA on the 400VAC winding. If I got the math right I'd be looking at 560ish VDC after rectification and filtering.

Now, I'm trying to figure out how to do the voltage regulation of the output. I don't really need it to be variable, I can live with around 10 steps just below the common ratings of capacitors, as I don't really need to test them exactly at the rating (a cap that leaks probably is going to leak at 400V even though it's rated at 450V). Also, the exact voltage isn't that critical either, 5 volts precision is probably fine.

What I've seen online some people seem to use a N-channel MOSFET for the arrangement, like this one by Mr. Caldeira



I'm thinking of replacing the pot with a rotary switch with different resistors, which begs the question what will happen in a brake-before-make scenario, if the gate doesn't see any voltage it should just stop conducting right? Any recommendations on suitable MOSFETS?

[magic]

50W is way overkill for testing leakage or breakdown voltage of components. Current output of a few mA or less would be sufficient and much safer to work with.

I have built a 100V generator for similar purposes, it consists of a small 12V transformer and a capacitive voltage multiplier (1N4148 and 10µF elcos), followed by 22kΩ series resistance and a 100V zener. I think it delivers some 0.2mA at 100V and a few mA into a short. I can grab its output and just feel a bit of tingling.

[exe]

I always wanted to make a high-voltage (100V+) supply, but I'm concerned about safety. Not just about current limiting of output, but also to have less high voltage inside so when I'm poking around it wouldn't zap me too hard. In a charge pump like here https://circuits-diy.com/charge-pump-circuit-getting-higher-voltage-from-low-voltage-source/ the voltage is gradually build up. So, not sure how to make it idiot-proof. I can only think of adding current-limiting resistor at each stage, but I didn't check if this gonna degrade performance.

As of stabilizing it, I saw high-side regulation using low-voltage regulator. Probably, something like this: https://www.ti.com/lit/an/snoa648/snoa648.pdf . Not sure if this approach is going to work for 600V.

[magic]

Don't use Schottky diodes. Their leakage isn't worth the 300mV reduction in voltage drop per stage. I tried and ditched them for the good old 4148.

This circuit is already quite current-limited, because the amount of charge it transfers per cycle is limited. However, the capacitors deliver high surge current when charged up. My version still felt kinda OK touch without the limiting resistor, but it tended to destroy Zener diodes immediately, so I added the resistor.

Capacitance can be reduced if frequency is increased.

[jdutky]

You can get high voltage transistors (rated for 500 V to 1000 V) and build a current mirror to limit the current output of the supply. Similarly, you can build a discrete voltage regulator either from the same high voltage transistors, or from lower voltage parts (and a lower voltage reference) and use a divider network to cut down the supplied voltage to match the reference voltage.

Based on what I've seen of high voltage supplies in old CRT oscilloscopes you do the regulation on a divided fraction of the output voltage and regulate an oscillator into a step up transformer. These days you would probably do that with a boost mode switching supply, and you can probably get off the shelf SMPS regulators to do just that.

(speaking without experience actually building such a supply)

-- Jeff Dutky

[David Hess]

Based on what I've seen of high voltage supplies in old CRT oscilloscopes you do the regulation on a divided fraction of the output voltage and regulate an oscillator into a step up transformer. These days you would probably do that with a boost mode switching supply, and you can probably get off the shelf SMPS regulators to do just that.

The various methods that oscilloscopes use to regulate high voltages are what occurred to me also.

When high voltage pass devices are not available or desired, a variable low voltage regulator can have a high voltage stacked on top allowing precise control of the high voltage output.  This can even work by adding a small DC offset voltage into the first stage of a voltage multiplier, and DC restorer circuits work this way also while producing nanosecond edges.

The other common method, especially when a wide control range is required, is to implement regulation on the primary side of an inverter.

The simplest method, and what I would consider first, is a step-up converter driving a voltage multiplier.  Output current is limited but plenty for leakage measurements.

[jdutky]

The simplest method, and what I would consider first, is a step-up converter driving a voltage multiplier.  Output current is limited but plenty for leakage measurements.

This is exactly the method that I am familiar with from the Tek 475/A oscilloscopes. There is a high value resistor array that drops the HV down to a "reasonable" level and the regulation circuitry drives the input to a transformer and then a voltage multiplier (said voltage multiplier being now difficult to obtain except by salvage from another scope of similar make). My impression, from the allowed divergence from spec of the HV rail is that the regulation is pretty loose, but at over 2400 V it's got some room for play.

The trick for mass production appears to be getting good tolerances on the HV resistor array, but I'm betting you could tune a one-off build even without the expensive resistor array.

-- Jeff Dutky

[Stuggi]

50W is way overkill for testing leakage or breakdown voltage of components. Current output of a few mA or less would be sufficient and much safer to work with.

Oh, it most certainly is, and I will probably limit the supply to around 25mA or something along those lines, but having a larger than required power transformer is usually just a plus, and I have to work with what's available, and at 25€ I decided that I'll probably not find anything cheaper. 

[Stuggi]

As of stabilizing it, I saw high-side regulation using low-voltage regulator. Probably, something like this: https://www.ti.com/lit/an/snoa648/snoa648.pdf . Not sure if this approach is going to work for 600V.

That's quite an interesting concept, and should be a bit more stable than just a potentiometer and a MOSFET. Here's AN-103 for TI/NatSemi that goes more into the same subject, with a bit of a different take on how to go about it. 

[Ian.M]

For low current regulated HV PSU applications, its significantly more difficult to regulate down from a higher voltage than it is to boost up from a lower voltage, as high voltage pass transistors with SOAs suitable for linear DC operation are few and far between since the demise of CRT monitors and displays.

Consider using a boost converter feeding a Cockcroft-Waton voltage multiplier.  See https://www.analog.com/en/analog-dialogue/articles/high-voltage-boost-and-inverting-converters-for-communications.html for details.  A variable output voltage can be achieved without having high voltage across any front panel potentiometers by injecting a bias current to the feedback pin, which reduces the output voltage from that set by the feedback potential divider. 

[David Hess]

The simplest method, and what I would consider first, is a step-up converter driving a voltage multiplier.  Output current is limited but plenty for leakage measurements.

This is exactly the method that I am familiar with from the Tek 475/A oscilloscopes. There is a high value resistor array that drops the HV down to a "reasonable" level and the regulation circuitry drives the input to a transformer and then a voltage multiplier (said voltage multiplier being now difficult to obtain except by salvage from another scope of similar make). My impression, from the allowed divergence from spec of the HV rail is that the regulation is pretty loose, but at over 2400 V it's got some room for play.

The trick for mass production appears to be getting good tolerances on the HV resistor array, but I'm betting you could tune a one-off build even without the expensive resistor array.

With some exceptions, Tektronix oscilloscopes of that generation use primary side regulation of an inverter.  Feedback is taken from the secondary side because the cathode voltage must be closely regulated for accurate CRT deflection.

An example of a step-up converter driving a voltage multiplier can be found on page 18 of Linear Technology application note 45.  More voltage multiplier stages will be required to keep the switch voltage below maximum at higher output voltages.  The flyback configuration can be used to get higher voltages with fewer multiplier stages.

[Andree Henkel]

I built a supply for biasing APD (avalanche photo detectors).
I used resonant royer topology. Suitable Trafos availabe are usually used for CCFL backlighting.
While it was 300V, it was limited to very low output current using high value resistors, which with Cs formed lowpass to filter the rectified AC voltage, and using high resistance also direkt feedback was done for controlling the output voltage.
With pot the output setpoint was adjustable for the individual spezimen of APD (each has a certificate with Voltage Value at which m-Factor is 50 - one photoelectron -> 50 avalanche electrons) so the voltage was very stable as required in this application.

[David Hess]

The resonant Royer topology is more suitable where higher power and low noise is desirable.  I think the only thing which makes it economical now is that CCFL transformers for it are still available.