little power supply

Aug 17, 2025 Last reply: 10 months ago 10 Replies

My summer intern Enzo is designing a pulse generator and he doesn't have much time before school starts, so I volunteered to do the power supply for him.



This is an ordinary slope-compensated autotransformer-flyback Cockcroft-Walton-voltage-multiplier programmable high-voltage power supply.


formatting link
formatting link
I had a hard time getting the loop compensation right, but that turned out to be mostly a dumb mistake and a mediocre data sheet. It sims crazy slow so some of the oddities are hacks to speed it up.


Ignoring the hyperbolic text, I don't see a great advantage to that. Here's my latest version:

formatting link
It has nice dynamics and can't make more than 650 volts no matter what's programmed; that protects the GaN fets it will drive.

Here's an isolated (transformer coupled) output pulse into 50 ohms:

formatting link
It's little ratty, good enough, but we think we can pretty it up in a production version.

Those pulses are 8.5 KW but low energy and at 1 KHz, low power. But we keep blowing out attenuators, even 20 watt ones.

Low rep rates help, but 100 Hz into a sampling scope is really tedious.

My intern loves my old Tek 11802. He's amazed that it has a real glass CRT *and* touch screen.

Could be internal arcing in the cermet. That'll have a lower energy threshold than thermal.

Yup. Back in my antenna-coupled tunnel junction days, I was using a

20-Hz pulsed laser with lousy pulse-to-pulse energy stability, so it took the best part of an hour to get one trace (512 points) with 100 averages.

I love mine too. Have you noticed SD-24s dying lately? I have three out of four showing "Illegal channel number".

Fortunately I have these little 60-ps TDR modules that we've made, but they aren't as flexible as SD-24s.

Cheers

Phil Hobbs

Mine seems to be holding up. I think one can move the eprom from a dead SD24 into another one, maybe even copy one.

How do they interface?

We're designing our own PCB planar transformers for our high-voltage GaN pulsers. I'm testing some Coilcraft planars, to learn more about things. Even their giant 300 watt planar power trannies are amazingly clean and fast. I use the SD24 TDR as the pulse generator and the other channel as the secondary snooper.

formatting link

Plain UART for the OEM version, with USB serial and RS485 as options.

There’s also a super simple, self-timed analog mode—it puts out a 1E7:1 time stretched replica of the TDR trace, i.e. 1 ps becomes 10 us. Dead useful.

There’s a configurable delay line that sets the Rx pulse width. That lets us trade off time resolution vs. scan speed. (In order for the sampling loop to converge at each delay value, if the pulses are narrow, you need more of them.)

I’ll send you one to play with if you’re interested.

I do that for diode lasers. Works great.

Cheers

Phil Hobbs

I ordered two new 20-watt attenuators on Tuesday and my guys blew them out on Wednesday, hours after the UPS guy delivered. He grabbed a couple fistfulls of candy on his way out.

We got a 1 GHz Rigol scope so we don't have to use the sampler, so we were running at 20 Hz or something like that. Tiny average power and microjoules per shot.

I'm going to have to make my own attenuators. Not pi or tee but probably something else. Needn't be symmetric. I'm thinking 15 or maybe 16 resistors.

Time to seed some Feen-A-Mint gum into the candy dish next time the truck pulls up?

What can brown do for you? ... to you?

No, we encourage all our service people to take treats. Candy, cookies, chips.

OK, sorry. I thought you were complaining about getting cleaned out of lagniappe meant for customers.

Great word, lagniappe. When I was a kid we'd buy some candy and ask for lagniappe, which was roughly a free 10% bonus, or the 13th donut out of a dozen.

Join the Discussion

Have something to add? Share your thoughts — no account required.

Didn't find your answer?

Ask the community — no account required