HV power supply maybe

Yes, I first read about phase shifted bridge 1990 in ED or maybe EDN. Because all the gates are driven 50:50 it lends itself easily to gate transformer drive.

Going into an LLC is very workable. I am thinking of using that topology myself. I thought it was novel until I found a TRW paper from the early

1960s !

If I did a 2-channel HV supply, that would need, I guess, three square waves. On as the "reference" side of both supplies, and then a phase shiftable square wave for each supply.

The frequency might be, say, 50 to 100 KHz and I'd want to shift the phases with 10s of ns resolution. Probably needs an FPGA.

It is more elegant than PWM.

I hate it when my brilliant inventions turn out to be 50 years old.

Series resonating gives a free voltage boost and takes out the transformer parasitics and probably reduces spike noise.

Yeah, the RP2040 has some fancy counter-timers. Maybe they can do it, possibly with a little help.

But it's nice when a neat, but impractical, idea you had in 1975 turns out to be useful in 1993 when the technology had moved on a bit.

Most inventions get invented in several places at much the same time, but this one wasn't.

Some of the parasitics, and as long as there is switching there is always some switching noise.

And you can do some fine phase shifting on clock edges with the MC100EP195/6 part - 10psec steps on an up to 10.6nsec delay, and the MC100EO196 lets you (slowly) analog tweak below that.

On a sunny day (Wed, 03 Jan 2024 11:29:02 -0800) it happened john larkin snipped-for-privacy@650pot.com wrote in snipped-for-privacy@4ax.com:

Is it not much simpler to drive a capacitor with L to ground from a complementary output stage?

• | T1 | C1 D1 |------||-------|>|---------- | | | T2 L1 === | | --- C2 : | | /// /// ///

Now Q of C1 L1 and load sets the voltage, no transformer needed. Simple PIC micro can drive that.

PIC micro. Has hardware PWM generator and comparators that can directly switch the PWM off. Used many times.

Fine control of the mark-to-space ratio, as should have been obvious.

The MC100EP196 is a clock timing control part - not "glue logic" - and Mouser have 493 in stock so it doesn't seem to be obsolete.

But they can drive properly terminated transmission lines, and if you are relying on fine control of mark-to-space ratio you don't want reflections moving the switching edges around.

But even at 130MHz a counter-based timer has a granularity of 7.7nsec. The MC100EP195 gets that down to 10psec.

If you're driving a load, you'll be dealing with large common-mode components.

If you're driving an isolation transformer, this won't be a major issue.

RL

Sloman never jokes. Never.

If a timer channel can run at 125 MHz, and my resonant drive is, say,

50 KHz, that ratio is 2500:1, which gives me pretty good resolution to control the HV output. Some cute tricks could improve that too, like some delta-sigma twiddling with the quantized phase shifts.

Should work.

With feedack from the HV out and some luck on time constants, it might just dither itself to better resolution.

My 24 volt supply will be grounded, but I want the HV output channels to float.

Are you commenting on the quality of my hand-drawn schematic?

Not a lot, so I may be able to use that cute little IXYS driver.

It will probably need a C-W voltage multiplier after the transformer, depending on the transformer I can find. I'm thinking 1500 volts out per channel, maybe 2KV. Diode drops won't matter at HV, low current.

A C-W multiplier lets me use cheap SOT-23 diodes and reasonable caps, to.

I'm planning a new product line and was just toying with possible boxes. A dual HV supply might be fun.

I'm thinking maybe 1.5 or 2 KV at a couple of watts, per isolated channel.

High praise!

Rube Goldberg and Heath Robinson were cartoonists, not electronic engineers.

John Larkin never gets my jokes, even the unsubtle ones,

You keep telling us about exotic switching transistors (SiC etc) where the switching losses don't catch up with dissipation in the on state below a couple of MHz. Why limit yourself to 50kHz?

Sigma-delta relies on the digital filtering to move most of the switching noise up to the high frequency end. You can do it systematically, as I discussed in my 1996 paper.

Sloman A.W., Buggs P., Molloy J., and Stewart D. “A microcontroller-based driver to stabilise the temperature of an optical stage to 1mK in the range 4C to 38C, using a Peltier heat pump and a thermistor sensor” Measurement Science and Technology, 7 1653-64 (1996)

Circuits don't dither themselves. You have to design it in, as also discussed in my 1996 paper.

Be adventurous. Get one wound, or printed, these days.

And it's going to float? That much insulation will be rare in an off-the shelf high-frequency transformer.

We always bought them in. The woods were full of photomultiplier power supply specialists. They have probably branched out into providing power supplies for high-voltage electro-optic modulators as well these days.

Yeah, something like that but switchmode with voltage and current sense ADCs and a digital control loop. It's still just an idea. I have a couple of nice boxes and need stuff to go in them.

Does it really make 400 KV?

I design stuff to be manufractured and sold, preferably surface mount, so I only use ebay or amazon parts for breadboards.

I'd expect low loaded Q, 5ish or even less maybe.

A C-W multiplier uses low voltage diodes and caps. That's nice. Maybe one HV cap at the output.

We have a 22nf 500v 1206 cap in stock, 13 cents. 22nF 1500V is $1.06, not awful.

With low Q, I would use a fixed frequency and use the phase shift trick to regulate the HV output, with ADCs to provide voltage and current feedback of course.