1KV buck converter

Anther one of John Larkin's pencil sketches - no resistances, no inductances or capacitances and part numbers only for the transistors.

At least with an LTSpice .asc file you get that stuff automatically.

And you get some hint at the switching spikes, which can be nasty.

Uniball Vision ink pen, not pencil.

It's an idea. If you want a finished design, write me a purchase order.

The guy who contacted me already had an idea the Baxandall inverter would work. I'm still waiting for Infineon to come back to me with the Spice model of their IMWH170R450M1 so that I can send him a simulation.

I won't charge him for it. If Infineon doesn't come through soon I'll see if I can bodge an existing MOSFET model to match the SiC datasheet.

Whacky ideas including a Diac aren't going to open anybody sensible's wallet. The kind of gullible sucker who would vote for Trump might go for it.

Interesting. Gotta think about that one.

Diacs are cool. I don't understand how they work. I've tried simulating one with transistors but they don't unlatch. Must be some silicon trick.

I once needed to indicate the presense of high voltage without dissipating a lot of power. An RC and a diac and an LED made a nice bright 0.5 Hz flash with low average current.

CCFL transformers usually have two windings with about a 100:1 ratio, so it should be possible to make a sort of non-saturating blocking oscillator, my buck topology but without the diac. Maybe even rectify a winding, a forward converter.

Diacs come up to 380 volts and one amp, so one might use a diac as the actual power switch in a buck or forward converter. Maybe even use two in series.

Before there were IGBTs, there were megawatt power switchers, converters and cycloconverters, that used enormous SCRs. That was terrifying. At least diacs can turn themselves off.

Gotta go... the donuts have arrived.

The highest diac rep rate will likely be in the tens of hertz range. (think cicada)

RL

Why? Diacs and sidacs turn on in under a microsecond. I can't imagine them taking milliseconds to turn off.

Interestingly, sidacs have been used in series strings up to many kilovolts, which suggets a kilovolt step-down converter with diacs/sidacs as the only active components.

Diac and thyristors aren't new ideas. Diacs have been around since 1957, but nobody I've worked with has ever used them.

They don't exactly turn themselves off - they just stop conducting when the current through them fall below the holding current - I could only find one data sheet that specified that and it gave a range from 10mA to

Step recovery diodes do turn off a lot faster than that, but it does take while for the stored charge to get swept out of the diode.

Step recovery diode are great for getting a fast edge out of relatively slowly changing current, but the fast edge only shows up after the stored charge has been cleared from the diode, so it's hard get it to show up exactly when you want it.

Probably not a step-down converter trying to convert 10uA from a 1kV source into about a 3mA at 3.3V.

You are right. A diac isn't the best choice to make a 99% efficient converter.

Yes. TBH at that power level and ratio I would expect 20% to be challenging and 50% miraculous.

I've been thinking of a version on a RM14 core where I could fit 6000 turns of AWG38 wire. Centre-tapped, that would give me two 61.2H primaries, and 245H for the tank circuit. If the parallel capacitance of the windings can be held down to 25pF that's a resonant frequency of

The resistance of windings is 474 ohm, so I^2.R is 500uW. 10uA at 1kV is

It's strictly a back of the envelope calculation, and 25pF is an ambitious target. You might need to bank the windings to get there

- perhaps as flat pancakes wound one after the other with self-bonding wire, bonded by brief self-heating and stacked onto a cut-down coil former.

Punch some numbers.

Cap size - charge rate - dV.

Turn-0ff time is only a small part of it.

Diacs turn off when discharge current < hold current. They also have significant internal impedance.

RL

Numbers are cheap. 2 nF charged to 1 KV stores a millijoule. Transfer that at 10 Hz and you get the required 10 mW.

The problem is the circuit.

Really? That's amazing.

Compared to a kilovolt, their ON voltage drop won't matter.

The idea of using diacs for this application was just a speculation. Transistors make more sense.

Diac section is pretty simple to breadboard.

RL

I needs a circuit first.

One could charge an RC and use diacs to dump that into an inductor, basically make a pulse-driven buck regulator. But charging the capacitor would be inefficient. I'd expect overall efficiency to be maybe 25%.

Bill - there was some talk here about a power supply I was working on. It used a pnp transistor driven by 2 capacitors and a diode to switch a solenoid style inductor. Efficiency was about 80%. This was two or three years ago. Initial starting powrer can be a puzzler.

Hul

Bill Sloman snipped-for-privacy@ieee.org wrote:

Hi Hul - getting Baxandall style inverters to start-up cleanly can be a problem. The amplitude can overshoot. The 1959 original would "squeg" is the feed inductor was too big, and you used bipolar transistor switches, which is something I've never been able to simulate. MOSFET switches apparently aren't susceptible to this problem.

Bill - the best approach I found for initial power was to tie a capacitor between the (most) positive rail and the other lead through a diode to the + rail of the control circuitry. How effective that was I don't remember though.

Hul

Bill Sloman snipped-for-privacy@ieee.org wrote: