Designing a regulated HV supply

[Hiemal]

I've been doing a great deal of research, poking around, and exploring different ideas and whatnot for designing a HV regulated power supply.

A few basic parameters;

1. Needs to be battery operated.
2. Supports at least 1 mA output at around 50-70kV.
3. Regulated/regulation would be nice.

This is part of my ongoing adventure in making a low powered X-ray box, to irradiate plants/seeds. Safety is paramount, and anything HV or x-ray related will be in a thick lead box. It'll have safety interlocks (if you open the box, power will be disconnected to the inverter) and a microcontroller to control/time everything. The microcontroller will also have a watchdog of some kind to monitor if it locks up, and will shut everything down if that happens.


The microcontroller stuff is fairly self explanatory. The biggest problem that I'm facing is figuring out how to make the inverter itself... there's quite a few ways I've gleamed that could potentially work... but figuring out what way to go with is a problem in of itself.


1. Flyback converter - Flybacks are one of the "simplest" inverters, in that they have just one switch. Problems are that compensating such an inverter can be problematic due to it being a boost derived converter with that lovely RHPZ.

2. ZVS Based Buck Fed Converter - Resonant, a plus for efficiency. Figuring out compensation would theoretically be simpler... but I'm not sure what type of controller I would use or what sort of feedback system. The buck converter part would be constant current... but needing to regulate the output current based off the output voltage of the secondary side, makes me think that I would likely have to roll my own since most, if not all, buck converter controllers don't expose the non-inv pin of the error amp.

3. A hysteretic ZVS? - Resonant, but with a basic "pow pow" comparator circuit with hysteresis to measure the output. If the output goes high, shut the ZVS off. Once it reaches the lower set point, turn back on. Rinse repeat. Probably one of the simpler control ideas since it inherently doesn't require any compensation, but figuring out the hysteresis and whatnot might be problematic.

Given that my output is so high, I know I'll need a voltage multiplier. I've already made a ... 5 or 6 stage voltage multiplier which I've encased in resin for this purpose. (I can't remember the exact number of stages off the top of my head). How would using a voltage multiplier change the compensation for the controllers? It would presumably slow down their response time, since it takes multiple cycles to actually charge each stage. Can I still use a voltage multiplier with a flyback converter, since it's not "true" AC in a sense?

I also have a small transformer I bought off of AliExpress. It has 3200 turns on the secondary, and 9 turns on the primary, bringing the ratio to 356. Obvs the primary can be changed if needed. 


There's also a good chance I'm overthinking this a bit, since it's fairly low power requirement but I still want to comb anyone's brain here who might have some ideas/thoughts to help. Would running it open loop be wise? Would...any of these ideas work? Am I being silly? Etc etc.

(ALSO! To note, I used to frequent 4HV.org a bit, my username over there was Inducktion.)

Thanks everyone!

[davekni]

Several questions come to mind before making any useful suggestions:
1) What is the use duty cycle?  On for 1 second and off for 60?  In other words, how important is full-load (1mA) efficiency vs unloaded efficiency?
2) Is there actually any unloaded time?  (Does the X-Ray tube have any control grid, or is the supply turned on and off to time exposure?  I don't know what is typical for X-Ray tubes.)
3) Roughly what capacitor values are in your 5 or 6-stage multiplier?  Would have a major impact on time constant and needed drive frequency.  Also, are the capacitors film or NP0/C0G (minimal change in capacitance with voltage)?  Or common cheap HV ceramic disk caps that have ~10% capacitance at full voltage?
4) What output ripple and accuracy is acceptable (ie for hysteresis feedback)?
If full-load time is short and unloaded time is long, then a flyback is worth considering.  Inefficient at full load, but efficiency doesn't drop much at light load.

Sounds like a fun project.  Good luck with it!

[Hiemal]

1) What is the use duty cycle?  On for 1 second and off for 60?  In other words, how important is full-load (1mA) efficiency vs unloaded efficiency?
 It'll only ever be used at short duty cycles I believe. It'll be a point of experimentation however, since I'm not 100% certain what sort of doses seeds/plants require to get any sort of mutations. Regardless, the amount of off time will almost always be longer than the on time.

2) Is there actually any unloaded time?  (Does the X-Ray tube have any control grid, or is the supply turned on and off to time exposure?  I don't know what is typical for X-Ray tubes.)
No, no control grid; the tube I have is from an old x-ray head, and from what I can tell it's basic. The supply will be turned off/on for exposure control.

3) Roughly what capacitor values are in your 5 or 6-stage multiplier?  Would have a major impact on time constant and needed drive frequency.  Also, are the capacitors film or NP0/C0G (minimal change in capacitance with voltage)?  Or common cheap HV ceramic disk caps that have ~10% capacitance at full voltage?
 They're the standard blue ceramic HV caps; they're fairly chunky ones, if I remember right. Either 10 nF or 100 nF, rated to 20kV. or 30kV. I know, it's dumb that I don't remember precisely but I'll try to take a look at them when I get back home tonight.

4) What output ripple and accuracy is acceptable (ie for hysteresis feedback)?
If full-load time is short and unloaded time is long, then a flyback is worth considering.  Inefficient at full load, but efficiency doesn't drop much at light load.
I'd like it to be fairly accurate as far as regulation goes... 5%? 1%? I'm not sure what's considered accurate or inaccurate, but it doesn't need to be laboratory grade. Basically the preference is for me to be able to dial in something, and for me to have some level of confidence in knowing that's what the tube is actually receiving without having to fiddle with anything.

[davekni]

They're the standard blue ceramic HV caps; they're fairly chunky ones, if I remember right. Either 10 nF or 100 nF, rated to 20kV. or 30kV. I know, it's dumb that I don't remember precisely but I'll try to take a look at them when I get back home tonight.

Are they something like this example:
https://www.ebay.com/itm/393028082633
If so, their capacitance will likely drop by 10x (to 10% of spec'ed value) at full rated voltage.  They may still be useful for your needs.  The capacitance reduction needs to be taken into account in design.  Also, these caps have high loss (dissipation factor).  I've used such in a HV multiplier.  The bottom ones died due to overheating when driven at 30kHz.

Thank you for the other answers.  Also, do you have any specifications for your transformer?  Is the core gapped?  Volt-second rating, or inductance and voltage and frequency, or DC current capability?

[Hiemal]

https://www.aliexpress.com/item/2255800905006413.html?spm=a2g0o.order_list.0.0.21ef18021iabbS&gatewayAdapt=4itemAdapt

is the listing for the transformer. MP-2057. UY10 core material. Supposedly rated to 60W. Secondary is 9H or greater in inductance. Core can be gapped if need be!

And, yes, those are indeed the style of caps I have. Would there be a better suited cap for this? Alternatively, the X-ray head I took apart for this actually has a voltage multiplier on a PCB.... I could likely repurpose that, though it's quite a few stages. I'll take some photos of it when I get home.

[davekni]

Supposedly rated to 60W.

You need 70W (1mA at 70kV), although intermittently.  Such Chinese parts are often overrated.  It may be tricky to get 70W even intermittently.

Would there be a better suited cap for this?

Yes.  They will be larger and more expensive.

Alternatively, the X-ray head I took apart for this actually has a voltage multiplier on a PCB.... I could likely repurpose that, though it's quite a few stages. I'll take some photos of it when I get home.

Unless it is damaged, an existing multiplier designed for this exact purpose is likely better than anything home-built.  Any transformer included?

[Hiemal]

Yes! Okay so I remembered when I got home, here's some photos of everything.






 [ Invalid Attachment ]

To note... The x-ray head I dismantled was full of oil. I kept it all, and put all of it in a glass container which is currently residing in my closet. I'd like to encase as much of the HV stuff in resin as opposed to oil, since oil is ...very messy. Not to mention, I'm not 100% certain what kind of oil that it is. It's yellowish. Kind of unpleasant odor. I did a specific gravity test, and it's quite a bit lighter/less dense than water, and very readily floats to the top. I'd like to avoid using any oil, and if I could somehow clean off the parts of it I feel like I would feel overall better about the entire thing. 

There is a transformer with the unit. 8 stage multiplier, but I can't decipher the ratings (either voltage or capacitance) of the caps.




Edit;; bought a PCBs testing kit just to be 100% positive, since I don't want to mess around or make assumptions.

[davekni]

The x-ray head I dismantled was full of oil. I kept it all, and put all of it in a glass container which is currently residing in my closet. I'd like to encase as much of the HV stuff in resin as opposed to oil, since oil is ...very messy.

Oil is cheaper than resin, insulates better, and aids in heat dissipation due to convection.  However, I avoid it too for the same reason, very messy.
If you do decide to use the original transformer and multiplier, it likely needs potting too, since it was originally in oil.

Given that you already have a new transformer and multiplier, probably makes most sense to start with those.  Do you have a meter to measure capacitance of your multiplier?  Or, can you read capacitance value and voltage?  Since you will not be at full voltage, perhaps 10nF caps would be around 2nF during operation.  That is likely enough for the frequencies your transformer can handle.  Power dissipation will just be a gamble, but may be fine given low on-times.

What is the transformer core diameter?  I think it is too small to handle flyback's net DC field at any reasonable frequency.  I recommend some form of resonant drive.  You could adjust output voltage by adjusting input DC voltage using an off-the-shelf buck (or boost) regulator between battery and inverter input.  That avoids any issue with loop stability.

What is the voltage rating of the resistors in your measurement string?  Looks like only 5 resistors.  Are they rated over 10kV each?  I have some 10kV resistors that look somewhat similar.  Haven't seen higher voltage ones in such size.

[Hiemal]

The capacitors in my homemade multiplier are 1 nF 20 kV each.

The measurement string resistors are 1 Gohm per, and are rated 8kV a piece. I figured adding resin might help increase their standoff voltage as well, but please correct me if I'm wrong. I do have some "better" HV resistors, I think I have two (or three?) 2G ohm ones that are rated to 30 kV, 5 watts per.

As for cleaning off the old parts, can I just...wash them in soap and water? Is that even okay to do? 

The transformer core diameter is 11 mm.

I did also find this app note which I find kind of interesting https://www.analog.com/media/en/technical-documentation/application-notes/an55fa.pdf for how they handle both feedback and control. It looks very similar/if not the same as how some buck converter drivers work for LED's... I'm curious if it would be possible to change it to a Mazelli/ZVS driver, changing the switcher, and making it constant voltage... It's an interesting idea that I might have to play around with on some protoboard.

[davekni]

The capacitors in my homemade multiplier are 1 nF 20 kV each.

Given likely steep capacitance loss with voltage, this may not be sufficient for 70kV at 1mA.  Perhaps workable if drive frequency is high enough, say 30-50kHz.  Will need more than +-7kV input to account for capacitor voltage swing, perhaps as much as +-10kV input.  Somewhere on the forum I saw a link to a multiplier calculation web page.  Or you could run a simulation (LTSpice or other).

The measurement string resistors are 1 Gohm per, and are rated 8kV a piece. I figured adding resin might help increase their standoff voltage as well, but please correct me if I'm wrong. I do have some "better" HV resistors, I think I have two (or three?) 2G ohm ones that are rated to 30 kV, 5 watts per.

I believe voltage rating is for internal resistance element.  HV resistors are often rated based on being under oil or potted.  Might work for your short 1-second use, but I'd recommend making a new string good for at least 70kV rating.

As for cleaning off the old parts, can I just...wash them in soap and water? Is that even okay to do? 

Yes, soap (ie. dish detergent, not literal soap) and water is generally fine.  Rinse very thoroughly.  If transformer windings are not potted, rinsing those will be hardest.  Might finish that with an alcohol rinse step.  Of course, dry thoroughly too.
Has anyone tried ammonia for washing?  Being volatile, that would make rinsing perfection less critical.

The transformer core diameter is 11 mm.

Thank you.  A bit larger than I was guessing from pictures.  I don't see any specification for the core material.  PC40 would be most likely, most common for Chinese cores.  PC40 would be a good material for your use.  With sine-wave drive, 20kHz or above should be fine to avoid saturation at +-10kV output (+-3.1V per turn on secondary, a bit higher on primary due to coupling factor).

I did also find this app note which I find kind of interesting https://www.analog.com/media/en/technical-documentation/application-notes/an55fa.pdf for how they handle both feedback and control. It looks very similar/if not the same as how some buck converter drivers work for LED's... I'm curious if it would be possible to change it to a Mazelli/ZVS driver, changing the switcher, and making it constant voltage... It's an interesting idea that I might have to play around with on some protoboard.

Yes, I think this is a great idea, excepting one issue I've ran into.  I have a similar build for a smaller core feeding a small home-made plasma globe.  The transformer becomes a dual-resonant system.  One is the normal intended ZVS resonance of primary inductance and primary capacitor(s) on ZVS driver board.  The other is the secondary resonance caused by intra-winding capacitance and load capacitance (capacitance between multiplier halves).  ZVS oscillators have a tendency to lock to the upper pole where primary and secondary are out-of-phase.  Upper pole will not generate much output voltage.  I solved (patched) this issue with added R/L/C components on the primary to damp the upper pole.  Added complexity and power loss, however.

BTW, you will need to gap the core for any resonant design where transformer forms the resonant inductor.

[MRMILSTAR]

Here is a link to an on-line voltage multiplier calculator. I found it to be very useful when I was designing my 14-stage unit.

https://www.extremeelectronics.co.uk/calculators/cw-voltage-calculator/

[abstruse1]

Small comment -- household ammonia is really ammonium hydroxide and is mostly water.  Ammonia gas will cone out of solution easily but the water will be left behind.

I've  used dielectric gel with good results, but it's slightly tedious to dig out the potted components in case they need changing.  This stuff:

https://raytechgels.com/products/raytech-magic-power-gel?currency=USD&variant=30742981247053&utm_medium=cpc&utm_source=google&utm_campaign=Google%20Shopping&gclid=CjwKCAjw_ISWBhBkEiwAdqxb9qaCvUhPrC5tqlNHAHXhB25c87i8PRsLEIbxZH-WF0jEXYWan3DbfBoCxXgQAvD_BwE 

[davekni]

Small comment -- household ammonia is really ammonium hydroxide and is mostly water.  Ammonia gas will cone out of solution easily but the water will be left behind.

Yes, but water evaporates too.  Normal detergent solutions will not evaporate completely.  They contain Sodium Laurel Sulphate or other such surfactants that are solids at room temperature.  Other surfactants are non-volatile liquids at room temperature.  Also, most detergent solutions include small amounts of either sodium chloride or magnesium chloride.  That's why thorough rinsing is important.

[Hiemal]

Did some washy washy. Caps on the pre-made multiplier are 1nF Y5P caps. Tube itself is a trophy TRX708, and I found an approximate schematic (I say approximate since I think there's minor/small differences, but it's essentially the same) of the head insides https://www.repairfaq.org/sam/trx78sch.pdf <- here.








As for the inverter, I'm doing it in two stages I think. First stage is a 24 to 100v regulated quasi-resonant active clamp flyback, and then the second stage will be the true HV step up system. I also bought some new, different 1G ohm 1% resistors that are rated to 10kV, at 1.4W. Also re-evaluated my use case and realized that I *probably* ought to go with softer x-rays, and as such want to design the power supply for 30 kV to 50kV, which should hopefully make things a little easier.

Additionally, using the first step up system will, theoretically, let me use most AC-DC converter chips that are normally designed for mains input... and should help open up some more design choices. Potentially.

Wonder how the original supply worked; the transformer that came with the tube isn't center tapped. From what I read it seems to have been some sort of LLC resonant converter I think...

[davekni]

Caps on the pre-made multiplier are 1nF Y5P caps.

Can't find any capacitance vs. voltage curves for Y5P.  Likely better than Z5U and worse than X7R.  Can you read the temperature coefficient code on your new multiplier caps?

Tube itself is a trophy TRX708, and I found an approximate schematic

Interesting to see the "approximate" schematic.  Not at all clear what function the relay serves.  Perhaps the schematic isn't quite accurate there, or I'm not dreaming up a useful external connection.
Both the low transformer turn counts and low multiplier capacitance suggest this circuit operated at a relatively high frequency (20kHz or higher).  Yes, some form of resonant converter makes sense, especially if the transformer core is gapped.

As for the inverter, I'm doing it in two stages I think. First stage is a 24 to 100v regulated quasi-resonant active clamp flyback, and then the second stage will be the true HV step up system.

Total efficiency will likely be lower with two stages.  Since efficiency isn't critical for low duty cycle use, might as well do what is easiest.
BTW, no center-tap is required for many ZVS circuits, but there is still the issue of running at upper pole.

Edit:  Thinking about your needs a bit more.  A UD2.7 or similar DRSSTC gate driver board should work well as a single stage.  They generally run on 24Vdc input.  Gate drive power for larger IGBT bricks is similar to the ~50W you need.  UD2.7 outputs intended to feed GDT would instead feed through a resonant cap to your HV transformer primary.  (Bypass output caps on UD2.7.)  UD2.7 enable input (interrupter input) would come from HV regulation feedback.  Enabled when HV is low and disabled when high.  No intentional hysteresis.  No actual GDT or IGBTs.  Single-stage current transformers (CTs) would monitor transformer current (UD2.7 output current through resonant capacitor and transformer primary).  CT secondaries feed UD2.7 feedback and OCD inputs.  CT burden resistor values within UD2.7 would need to go up some since monitoring much lower current than typical DRSSTC primaries.

I also bought some new, different 1G ohm 1% resistors that are rated to 10kV, at 1.4W. Also re-evaluated my use case and realized that I *probably* ought to go with softer x-rays, and as such want to design the power supply for 30 kV to 50kV, which should hopefully make things a little easier.

Yes, lower voltage will be easier in several ways.  Caps will be closer to rated value at lower voltage, and insulation is less stressed.

[Anders Mikkelsen]

If you want to use the original transformer from that X-ray head (as used in the Kodak/Trophy 2200 system, possibly others), it's designed to run with series loaded resonant regulation. With 10 uH in series with the primary (for 110 V mode if I recall correctly, that's both primaries in parallel). Resonant frequency ends up being around 160 kHz, and you run above this to maintain ZVS, changing the frequency to regulate the voltage. It's made to be run by a full bridge from rectified mains, but you can run it from a half bridge with a 320 V bus in 110 V configuration for example.

The original driver used an MC34067, my own driver for these was based on a FAN7621 which includes the half bridge driver.

The relay in the original circuit is used to select the biasing resistor for the filament cup voltage, which is a few hundred volts negative in relation to the cathode. This voltage serves to focus the beam and make the distribution more uniform. The relay selects between 4 mA and 7 mA mode, with 4 mA selected by default if I recall right.