100kV isolation transformers, high precision voltage and current sensing

[Spirit532]

So here's a few interesting questions.

I'm designing a power supply for a very low power(2W), but rather high voltage microfocus x-ray setup. However, due to how the tube is constructed, and partially what the datasheet says, I have to run the tube in a grounded anode configuration, where the cathode sits at -100kV, rather than supplying it with the full positive potential.
While the electrostatic focusing mechanism only benefits from this, being able to simply run through a pre-calculated resistor, I am facing an interesting problem, and that problem is filament supply. I need to somehow deliver 3.2A at 1.8V to the filament and monitor that current(in a feedback loop), all while it's floating at -100kV.

Question #1: Is there an easy, intuitive solution, or do I just have to wind a physically huge isolation transformer and place the entire monitoring circuit, including logic and ADC, at -100kV, all while feeding the data back via something like fiber optics?

Question #2: Since I'm going to be using a CW multiplier to generate the high voltage required, if I try to measure the output voltage at the transformer or one of the first legs of the multiplier, I'm going to get an inaccurate estimate of the output, and for my application the output voltage must be within +-0.5kV of the desired output throughout the entire range of -40 to -100kV.
What's the best way to measure the full output voltage directly without affecting it too much? Just a massive resistor divider with a filter, built out of long through-hole parts in oil or silicone?

Question #3: Since this is an ultra low power(2W) tube, I'm going to also need relatively precise current measurement from ~8 to 20uA with a resolution of at least 0.1uA and ~+-5% accuracy. Since the anode is grounded, I should be able to just do a low side(relative to the rest of the system) measurement, but are there general good practices to measuring tiny currents in high(er) noise environments, other than lots of shielding?

To calm the grumpy people that are terrified of x-rays: I'm very well educated in this field and I know exactly what I'm doing. Nobody will be exposed, it's all tightly controlled and measured.

[jbb]

I haven’t done X-Ray work but I have done high V converter work which has similar auxiliary supply issues.  Am I right that 100KV is quite high and will produce fairly high energy X rays?

I’m guessing you’re going to have a (silicone) oil tank in any case which makes things easier. Do you have some kind of handbook on material compatibility?

On the filament supply: I think you’ll end up with a physically large transformer in oil.  The big question is how precise do you need to be? Does it need AC or DC excitation?  What comes to mind is a smart driver on the primary side and dumb transformer on the HV side. This will be simpler but offer less accurate control of the filament current (especially at low currents).

Another option to the conventional transformer Is kind of ‘inductive power transfer’ scheme (ie air cored low coupling factor transformer) which could consist of 2 pancake coils in oil. That means you’ll only need one clearance gap rather than several as in a conventional transformer. This would then require brains on the HV side for control.  Maybe it will need some attention to magnetic leakage so it doesn’t perturb the electron beam?

If you have brains on the HV I guess you could measure the cathode current itself and do the beam and filament current control on the same board. Communications could be via fibre, but the thought occurs that the oil tank is likely opaque and oil is fairly transparent, so maybe you could just use IR / red LEDs and photodiodes and beam the signal through the bulk oil? Does the communication need to be very fast?

On V sensing: I expect you’ll end up with a big resistor divider. You’ll need to look out for stability and temperature coefficients, and may need to add caps across each resistor to improve the frequency response. I think your V divider could wind up eating more current than the tube, so you’ll have to take that into account.

[Spirit532]

Am I right that 100KV is quite high and will produce fairly high energy X rays?

Yes - the peak(!) x-ray photon energy is directly proportional to the voltage, so I'll get a normal distribution curve with 100keV photons at the top. It's relatively uncommon in a microfocus application, but the tube choice was driven mainly by price, not by exact specifications

I’m guessing you’re going to have a (silicone) oil tank in any case which makes things easier. Do you have some kind of handbook on material compatibility?

The tube has a beryllium window and isn't designed for submersion, it's going to operate in air with very large silicone doughnuts hugging it for insulation. The rest of the HV circuit will likely be either in oil or filled with silicone.

On the filament supply: I think you’ll end up with a physically large transformer in oil.  The big question is how precise do you need to be? Does it need AC or DC excitation?  What comes to mind is a smart driver on the primary side and dumb transformer on the HV side. This will be simpler but offer less accurate control of the filament current (especially at low currents).

Another option to the conventional transformer Is kind of ‘inductive power transfer’ scheme (ie air cored low coupling factor transformer) which could consist of 2 pancake coils in oil. That means you’ll only need one clearance gap rather than several as in a conventional transformer. This would then require brains on the HV side for control.  Maybe it will need some attention to magnetic leakage so it doesn’t perturb the electron beam?

If you have brains on the HV I guess you could measure the cathode current itself and do the beam and filament current control on the same board. Communications could be via fibre, but the thought occurs that the oil tank is likely opaque and oil is fairly transparent, so maybe you could just use IR / red LEDs and photodiodes and beam the signal through the bulk oil? Does the communication need to be very fast?

I need to be quite precise with the filament current, and the datasheet specifies that it has to be driven with AC. The electron beam is crushed electrostatically, then driven into a thin tube that has a ring magnet around it, to get the 20um focal spot. Fiber should work fine, but I've never really designed something that would float at such a high voltage relative to ground, so I'm not entirely sure on what kind of transformer I should be designing. I was thinking about stacking a couple Qi wireless chargers side by side, since I only need around 10-12W delivered through ~20mm of silicone rubber isolation for 100kV.
The cathode current can be measured on the other(grounded) end, so I could move the circuit away from the HV drive system.

On V sensing: I expect you’ll end up with a big resistor divider. You’ll need to look out for stability and temperature coefficients, and may need to add caps across each resistor to improve the frequency response. I think your V divider could wind up eating more current than the tube, so you’ll have to take that into account.

My original idea for the divider was to tap the first leg of the 8-10 stage CW multiplier and cut the AC component, then scale the voltage appropriately, but I'm not sure whether the output voltage will scale linearly with load(in theory it shouldn't), so I don't know if calibrating it that way would be possible. Another solution could be building a 1GOhm divider to reduce the load.

[T3sl4co1l]

Most generally, you're doing isolation by converting static (DC) or quasi-static (AC, low frequency?) electrical power into E&M fields/waves, then back again.

Anything which can do this will work, and to this high-level view, is equivalent.

So, transformer: dPhi/dt == EMF.  Voila, we have E&M, and so we can say a transformer is a (near field) antenna.  We could go the other way, using dE/dt to induce displacement current -- coupling energy through a capacitor.  Both of these have the downside that they need to be fairly nearby, which puts a lot of pressure* on the insulation system.

*Electrical pressure, voltage.

If we open up the spacing in the transformer or capacitor, we can use RF tuning methods to cancel out some of the reactances thus created.  Instead of a nearly ideal transformer with undesirable leakage, we have a leaky transformer better described as a poorly-coupled pair of inductors (or capacitors).  This is very demanding on the Q factor of the components, and requires carefully balanced design to avoid generating interference, but can be done.

If we open it up even farther, we leave the near field entirely (depending on frequency), and move into a radiating mode.  Now we need good antennas to ensure low power leakage and losses, and we need efficient transmitters and receivers to avoid huge losses there as well.  Antenna size is acceptable in the microwave range, but the electronics aren't so great.  (A transmitter over 80% efficiency isn't unreasonable, but I'm not sure you can make a rectenna as good?)

You can continue going up in frequency, which doesn't really do you much good at first.  I guess inbetween, you could use, like, radiant heat on Peltier modules or something (or any suitably insulating fluid also carrying heat, like pumped and heated transformer oil?).

Once you get into significant quantum energy levels (namely, a few eV -- visible light), it gets interesting again. There, your transmitters and rectennas suck pretty bad (~20% each), but they are at least very cheap and available.  That is to say -- you may consider a stack of LEDs and solar cells for this application.   (Or if that's a bit too bulky, a laser, fiber and cell can be smaller, if maybe not any more efficient, and definitely not cheaper.  Some isolated scope probes use this!)

In short, you can look at options literally anywhere across the spectrum.  Figure out a basic design -- and therefore capability and cost tradeoff -- for each typical case, then pick the best one.

If you just want an easy suggestion, do lights and panels if you have the space, or figure out the insulation requirements of a probably resonant transformer system.  For data over the isolation barrier, consider putting ADCs local and clocking it out over fiber optic.

Tim

[Spirit532]

In-air HV isolation would be quite painful, so I'm going to have to encase it in silicone. LEDs and solar panels are a no-go, especially because I need ~10-12W for the whole thing.
As for pushing data back and forth, fiber is easy and really cheap - the data rate needs to be maybe a few kBit/s tops, just to receive current settings and report back on the actual current once in a while.
S/PDIF modules are cheap and work with basically anything that transmits light - hell, I could probably just point them at each other through the clear silicone, but a 50mm piece of plastic fiber is peanuts.

Seems like there's plenty of existing solutions(like Qi 1.2, rated up to 45mm) that could provide the power I need for the filament and still be able to sit in silicone at the full isolation distance.
I've found a few rather large(9cm dia.) devices on Aliexpress that claim they can transmit up to 30W and 50W peak - if it can do 30W when the coils are stacked together, it might just be capable of delivering the ~12W I need for the filament and feedback circuitry at ~20mm separation.

[Spirit532]

After thoroughly sanity checking myself, I think I'll just be going with a slightly larger ferrite transformer to deliver the power, but I'm placing the current control circuitry on the "hot" end.

[David Hess]

Question #1: Is there an easy, intuitive solution, or do I just have to wind a physically huge isolation transformer and place the entire monitoring circuit, including logic and ADC, at -100kV, all while feeding the data back via something like fiber optics?

I would just monitor the current on the primary side of the transformer.  I might also use several transformer cores in series with balancing to get the required high voltage isolation.  Were you planning on vacuum potting?  I sure would.

Question #2: Since I'm going to be using a CW multiplier to generate the high voltage required, if I try to measure the output voltage at the transformer or one of the first legs of the multiplier, I'm going to get an inaccurate estimate of the output, and for my application the output voltage must be within +-0.5kV of the desired output throughout the entire range of -40 to -100kV.

What's the best way to measure the full output voltage directly without affecting it too much? Just a massive resistor divider with a filter, built out of long through-hole parts in oil or silicone?

A precision high voltage resistor divider is not that hard to make but the usual high voltage construction methods will be required.

Question #3: Since this is an ultra low power(2W) tube, I'm going to also need relatively precise current measurement from ~8 to 20uA with a resolution of at least 0.1uA and ~+-5% accuracy. Since the anode is grounded, I should be able to just do a low side(relative to the rest of the system) measurement, but are there general good practices to measuring tiny currents in high(er) noise environments, other than lots of shielding?

Those current levels are not that low.  Are you anticipating coupling from some other noisy source?  Limiting measurement bandwidth to only that which is required atones for a lot of sin.

Current measurement can also be made on the ground side of the high voltage multiplier.  Tektronix oscilloscopes did this to measure the beam current to protect from burning the phosphor.

[Spirit532]

I would just monitor the current on the primary side of the transformer.  I might also use several transformer cores in series with balancing to get the required high voltage isolation.  Were you planning on vacuum potting?  I sure would.

I'm definitely doing vacuum potting in high purity silicone designed for isolation. It has a minimum guaranteed dielectric strength of 25kV/mm at 20C and 60% RH, I'm going with 15-18 mm on all sides of the HV winding - though obviously it's not going to be good for 400kV, but it should be fine below 100kV.

Those current levels are not that low.  Are you anticipating coupling from some other noisy source?  Limiting measurement bandwidth to only that which is required atones for a lot of sin.

20uA is quite a low current still, compared to several mA - I haven't designed a low current measurement device before, I'm more on the high power side of electronics.
Adding a filter is probably a good idea. I'm expecting some noise to come from the high frequency transformer generating the CW input voltage, but I was planning on putting it in a can with everything up the digital side shielded, just to be sure.

[duak]

I would look at Color TVs for prior art, especially those with a vacuum tube HV rectifier as they had to provide the heater voltage for the rectifier floated at the HV DC output voltage.  This was done with a secondary winding of a few turns of highly insulated wire wound on the HV (flyback) transformer.  In some sets the HV DC could be as high as 30 kV at quite a few uA.  It's not 100 kV, but it could be instructive.

Cheers,

[David Hess]

I would just monitor the current on the primary side of the transformer.  I might also use several transformer cores in series with balancing to get the required high voltage isolation.  Were you planning on vacuum potting?  I sure would.

I'm definitely doing vacuum potting in high purity silicone designed for isolation. It has a minimum guaranteed dielectric strength of 25kV/mm at 20C and 60% RH, I'm going with 15-18 mm on all sides of the HV winding - though obviously it's not going to be good for 400kV, but it should be fine below 100kV.

One thing I learned messing around with high voltages is that it is amazing what becomes conductive if the voltage is high enough.  Do not forget to isolate the windings from the transformer core or cores.

[Spirit532]

One thing I learned messing around with high voltages is that it is amazing what becomes conductive if the voltage is high enough.  Do not forget to isolate the windings from the transformer core or cores.

As mentioned, isolation is going to happen on both sides. Below and above the winding, so that the spacing to the ferrite core is the same as it is to everything else. The low voltage coil will be wound directly on the ferrite.

I would look at Color TVs for prior art, especially those with a vacuum tube HV rectifier as they had to provide the heater voltage for the rectifier floated at the HV DC output voltage.

Sadly the way isolation done there is not going to work here - it was usually potted alongside the main transformer, and I'm keeping the HV section very low power(~7W total input accounting for all losses).

[jbb]

Hmm
Well, a great big core and appropriate lots of potting should see you right.

It’s unusual, but have you considered PCB windings for the transformer secondary? You can keep very good control of the windings which might help with the potting. It may also be possible to cut some spokes and slots to help locate the windings in the core (but this might cause other issues...).

[Berni]

What you need in my opinion is a massive ferrite C core.

You can then turn a custom massive thick bobbin on a lathe to hold the secondary coil with lots of clearance to all parts of the core. Ferrite is not conductive at low voltage but certainly is at high voltage since its essentially very fine iron filings held together with a non conductive binder. Then to get power into the transformer you just wind a primary onto the other leg of the C core, without any special clearance requirements since both the primary and core are sitting at ground potential. To be sure the core stays at ground potential it might also be a good idea to stick some copper foil to both core halves using conductive glue and ground it.

The ferrite also lets you operate the transformer at high frequency (20 to 500KHz) so you only need a handful of turns on the coils to get the few volts you need. Such a large core would have a good bit of energy storage capacity so switchmode topologies that make use of this are proabobly a good idea such as flyback. Tho these topologies tend to be fairly electrically noisy and winding the secondary so far away from the core causes extra leakage inductance that makes it worse. A good alternative might be driving it using a Royer oscillator or a ZVS oscillator, these provide a more sinusoid like drive current and don't care about leakage inductance

As for the HV side sensing circuitry you can simply have an extra winding to create a supply voltage for it and treat the -100kV as being ground, this lets you use normal opamps up there to sense all those parameters you want. Tho i would recommend putting the elctronics in a shielded box that's also connected to -100kV to protect the circuitry from any potentially very strong fields drawn to anything grounded.

[Doctorandus_P]

Some time ago I saw a few pictures of very specialised high voltage transformers.
These are used in big antenna systems (100kW or more, etc) where some auxilary power supply has to be coupled into the antenna.
These do not only need high voltage separatrion, but also very low capacitive coupling.

I tried an image search, but that did not work for me this time.
It was a toroidal transformer with a diameter of around 40cm and a relatively small cross dection of the ferrite / iron core (Few square cm).
The primary winding was wound tightly around the core around the whole circumference.
The secundary winding was wound around a spool with the same diameter as the toroidal core.
The two toroids were both mounted individually in an external frame to keep the separation at a maximum.

So if the diameter of the 2 toroids is 40cm you can have an air gap of 18cm.

[NiHaoMike]

Have you considered a BLDC motor driving another through a plastic rod?

[PartialDischarge]

I would just monitor the current on the primary side of the transformer.  I might also use several transformer cores in series with balancing to get the required high voltage isolation.  Were you planning on vacuum potting?  I sure would.

I'm definitely doing vacuum potting in high purity silicone designed for isolation. It has a minimum guaranteed dielectric strength of 25kV/mm at 20C and 60% RH, I'm going with 15-18 mm on all sides of the HV winding - though obviously it's not going to be good for 400kV, but it should be fine below 100kV.

That is going to be a mess, silicone is not free flowing so its not that easy to vacuum pot, especially with weirded shapes like transformers. Anyway why would you want to pot a research module like that, you will not be able to make repairs, change things or study potential problems. Use transformer oil, just need to be dry and by dry I mean less than 0.01% water content. Good transformer oil withstands more than 35kv/mm. Paraffin oil is a good alternative, from the environmental and safety points of view, but again needs to be very dry.

But dielectric strength doesn’t mean that much. You need to study the relative conductivities and permittivities of the materials inside the tank. At dc conductivity is what matters, and permittivities at ac. The electric field and equipotential lines will distribute accordingly and in a very savvy way. Thats why in practice mixed insulating materials are used in high voltage tanks, solid insulators and oil, one is good where the other isn’t. Their distribution and width matters cause you are creating capacitor dividers (and resistive dividers at the same time) inside. For example if you place a sheet of insulating plastic around the xray crystal tube with high er, you will probably create arcing since now the tube ampule forms a capacitor with smaller value than the sheet

[djacobow]

I have nothing to add to this thread, but it has been interesting to read. I have never dealt with these sorts of problems.

I do have one question, though: how does the tube know what potential it's at?

[mjs]

How much do you have to pay for the modules and how much power do you get through ? I did a proof of concept by just plugging a 1W laser diode to 1mm PMMA fiber end and easily got the 10mW/3.3V I needed. That's not suitable to any kind of real use, so I'm looking for commercial components to focus on the actual measurement electronics next.

[jbb]

It’s totally out of scope, but now I’m imagining an X Ray tube that replaces the filament with fibre fed laser heating. After all, converting laser light to electricity to run a heater is very inefficient.

But for practical use I’d stick with isolating transfomer.

[Yansi]

I do not see much issues here. Just grab a large enough ferrite core,  wind a secondary with high voltage cable to obtain those few volts to run the filament and measurement circuitry.

Use a cheap multimode optical fiber for the feedback.

I don't think you could do much simpler than that. Not sure why everyone is recommending some ultra expensive complex solutions as some laser powered modules.

No special transformers required, no thick bobins to be turned on a lathe. Just grab an off the shelf ferrite core, and a length of a high voltage insulated cable.

//EDIT: Just thoughts for the filament supply plus the measurement circuitry isolated supply. Generating the -100kV is another matter I have unofortnately not much experience with.

[duak]

This fellow has done some HV work with similar voltages: http://boginjr.com/electronics/hv/multiplier/  I can't vouch for the long term reliability but the concepts are sound.  His website has various other tidbits.

A lot of the nuclear research facilities such as CERN, Fermilab and SLAC work with EHV and have published papers on the equipment designed and built there.

Cheers,

[Spirit532]

Whoa, lots of replies.

It’s unusual, but have you considered PCB windings for the transformer secondary?

Usable up to ~20kV tops from my experience, and a bit of a mess to make, especially for higher power.

What you need in my opinion is a massive ferrite C core.

Doing exactly this - 20mm separation on either side, 40mm ID C-core.

As for the HV side sensing circuitry you can simply have an extra winding to create a supply voltage for it and treat the -100kV as being ground, this lets you use normal opamps up there to sense all those parameters you want. Tho i would recommend putting the elctronics in a shielded box that's also connected to -100kV to protect the circuitry from any potentially very strong fields drawn to anything grounded.

The high voltage PSU is very low power(calculated to be 8W input), and the HV transformer itself is only good to ~15kV. The rest is going through a 5-stage CW multiplier. This is why I'm doing a separate -HV filament supply.

a toroidal transformer with a diameter of around 40cm

That is gigantic for my application - the entire system will probably be thinner than that, and definitely shorter.

Have you considered a BLDC motor driving another through a plastic rod?

That is a crutch even I won't consider, and I've done some pretty hacky stuff. Fairly easy to implement, but a bit nuts to add mechanical transmission of power. That, and you can't pot it.

That is going to be a mess, silicone is not free flowing so its not that easy to vacuum pot, especially with weirded shapes like transformers. Anyway why would you want to pot a research module like that, you will not be able to make repairs, change things or study potential problems.

I have some low viscosity 2-component HV silicone that will work great for potting. It's not going to be an extremely tight assembly, so penetration won't be an issue. I might also be potting under vacuum, with air ingest to push it as far as possible.
This isn't a research module - I'm intending this as a part to go into an assembly that will sit inside an x-ray microtomograph. It will never be taken apart or serviced.
Oil is really messy to work with, and potting gives me standalone modules I can mount in air, rather than having to design and build enclosures and passthroughs for every cable. That, and as you've mentioned, it needs to be very dry, which is hard to achieve without using drying agents or buying brand new oil.

There is a Taiwanese company making up to 0.5W (recovered electrical power) per channel laser fiber power delivery systems.

Nope. The filament needs ~5.5W peak, + regulation losses.

I do have one question, though: how does the tube know what potential it's at?

It doesn't. You need a regulated HV power supply to get the x-ray energy you want. In my case it's a beryllium window tube, so anything from ~10kV and up will output x-rays, though 10keV won't do much other than photoionize surfaces.

How much do you have to pay for the modules and how much power do you get through?

Not a lot, and see above for power.

It’s totally out of scope, but now I’m imagining an X Ray tube that replaces the filament with fibre fed laser heating. After all, converting laser light to electricity to run a heater is very inefficient.

It could work - just have to set up a high vacuum glass-to-metal and beryllium window production line first. Quick, patent it!
Filaments burning out is a very large part of why X-ray tubes die in general, so if you eliminate that, they could last a very long time.

Designing a compact HV source with isolated filament supply is quite a weird combination, considering basically no modern microfocus x-ray tube uses a grounded anode setup, because as you can see, complications arise.
Grounding the anode lets you leave the focusing voltage regulation to the tube itself(via a resistor to the "grid"), but modern electronics have gotten good enough to generate the focusing voltages externally. Kevex tubes are a nice example of that - the tubes are ridiculously long and there's a whole braid of cables coming off the anode.

[Yansi]

Not sure why you have not reacted on my suggestion of winding a couple turns secondary to get the floating filament supply.

But, you should give it a try, I am pretty sure it might satisfy your needs of being quite compact in size. (bulkness defined mostly by how thin of a high voltage insulated wire you can get).

Not sure about 100kV+ rated wires, but those readily available 40kV ones are pretty thin and couple of turns would not be much of a challenge within a small core, such as ETD34 or a 3cm toroid.

[Spirit532]

Not sure why you have not reacted on my suggestion of winding a couple turns secondary to get the floating filament supply.

It's mentioned in my previous reply - because I'm winding a separate transformer for the HV supply. The main HV transformer will be very low power, and I'd rather not triple the HV system's input power and make the driver circuit more complex.
That, and the HV transformer is going to be at +9kV(max, accounting for CW losses), while the CW output will be at -100kV.

[jbb]

It’s unusual, but have you considered PCB windings for the transformer secondary?

Usable up to ~20kV tops from my experience, and a bit of a mess to make, especially for higher power.

Sorry, wasn't clear; I meant just doing the primary and secondary as PCB windings on separate boards, then potting the void between them.
I'm impressed that ~20kV is possible with both windings on one PCB!

It’s totally out of scope, but now I’m imagining an X Ray tube that replaces the filament with fibre fed laser heating. After all, converting laser light to electricity to run a heater is very inefficient.

It could work - just have to set up a high vacuum glass-to-metal and beryllium window production line first. Quick, patent it!
Filaments burning out is a very large part of why X-ray tubes die in general, so if you eliminate that, they could last a very long time.

Forgive my ignorance; I though X-Ray tubes are largely made of glass.  I've never worked with them myself (and might be too chicken given high V and radiation hazards I'm not trained for).