DIY HV Power Supply for Fusor. An Experimental Approach.

[Linda Haile]

This thread is basically continuing from the 'Signal Generator Problems' thread, as these problems have now been resolved.

The idea here is to develop a HV power supply for a fusor which is as simple as possible.

The power supply is basically divided into three parts, the signal source, the transistor driven ferrite cored transformer, which will probably comprise of more than one transistor/ ferrite cored transformer, and the rectification/smoothing circuit, which will probably include a multiplier stage or stages.

Total power will be less than three kW, which should allow for 50kV at 50mA, with some losses. (this seems to be a reasonable place to start.)

Initial experiments will be done using what I can find lying around in order to get a basic 'feel' for what I'm attempting. ( I have built some low power HV supplies in the past, but nothing like the power or efficiency required from this unit.)

The first step, as I see it is to get the signal generator to drive a transistor/ferrite cored transformer circuit and then modify this circuit to get the characteristics that I require.

I'll include a series capacitor between the signal generator and transistor to protect the generator should the transistor blow during destructive testing. As the value of this capacitor relies on several factors (frequency, current, etc) I'm just going to use a 'trial and error' approach, starting with a 1.16uF 2.1kV item from an old microwave oven and see what happens.

I'll include a diode in parallel with the coil to protect the transistor from reverse voltage spikes.(some circuits say this should be across the coil, some say it should be across the transistor, I might try two if I can find another one) As I have a 10kV 0.5A diode from the same microwave oven I'll try this first and see what happens.

I've found what appears to be a reasonably fast transistor from the horizontal deflection circuit of a large TV. It's in a TO3 case. It's rated for 1500V and 50W. The data sheet mentions 3MHz, as the maximum frequency of the signal generator is only 1MHz, I'm assuming this will be ok to start with.

The next step is to 'knock up' a heatsink for the transistor to a design that I've used successfully in the past. This will take a few days as I don't have much equipment here at the moment.

I'll then dig out some ferrites and wind some coils, and sort out a DC supply to power it.

The idea of this project is to learn as I go, and is based on trial and error, as this is a bit of a 'black art', however, any comments will be appreciated.

[Doug Coulter]

Here is the simplest circuit I've ever used that also works well. You tune the deadtime via the gate resistors. I used 27 ohms for IRFP264 fets. For higher current ones, reduce that, for higher voltage ones, increase it.

I've since added the RC snubber to the PCB design, and a large filter cap for the FET rails voltage, which is required to keep that noise out of the thing's built in oscillator.

Hard to get simpler than this.

I run the fet rails off a transfomer/variac and a bridge rectifier, and you start low until you get the frequency right for lowest quiescent power drain -- which will differ based on what topology you wind up using (see the thread Carl has going now for examples).

[Linda Haile]

How much power will that circuit handle Doug?

I'm looking for around 3kW for my fusor design.

[John Futter]

Lyn
I admire your fortitude
A homebrew high voltage PSU is not for the faint hearted. I do this for a living in my present employ.
While myself, Carl,and Doug make it look easy ( only recent posters named-- there are many more on this board with skill in this area) -and as the power goes up so generally do the problems at a square to the voltage output / power.

Ie bigger FETs more gate charge to shift
Stray circuit inductance especially in the magnetics ( this changes with how far your cores are magnetised)
Circuit trace cross coupling becomes a major factor when dealing with peak currents in the 10's to 100's of amps in the primary circuits, add to this voltage coupling in the high voltage stages getting into control circuits on the primary side ie unexpected modes of operation

My fusor effort is using a Bertan 1kW unit got from the USA all the way to New Zealand, Ok it cost me 350US$ ( not telling how much shipping cost but the purchase price is small by comparison) but thats a bargain compared to me buying the bits to build one with no blow ups on the way.

If you are using this as a learning exercise all the better. I suggest with starting at the 100 -300watt output level.
At these levels problems are similar to higher powers --but the silicon explosions are cheaper.

You will need a very good scope, home made current transformers to measure FET source current, transformer primary current, high speed voltage divider to check HV AC waveform before rectification, and more importantly a good library on SWPSU design.

After all that with what you are doing now is a good start!!.
Understand all those waveforms and find a way of measuring all the important parameters of your magnetics and to a much lesser degree capacitors and circuit layout (maybe I should not have put circuit layout this low down the order depends on your skill level as to whether this should go up the order)

Have fun
Learn
and above all stay safe
Remember
One flash you're ash
only one hand doing anything while measuring the other in your back pocket

[Linda Haile]

Thanks for the tips and encouragment John.

The main point of this is to learn. Also to build a power supply suitable for the modified fusor that I'm building. (I've described my ideas on this elsewhere)

It will be a modular design, I'm starting with a 300W dc supply for the reasons you point out.

I appreciate there is a lot of expertise in this area on this site. While I have plenty of experience with vacuum systems/general engineering my knowledge of electronics and RF is limited. It is for these reasons that I'm adopting this approach.

Once again, thanks for the tips, I'll research the points you raise in depth in due course.

[Doug Coulter]

John has it right, as usual. Here I have "unlimited power" for most intents and purposes, and wind up limiting things to 500w or rarely 1kw for *very* short times. Remember, even after you jump all hurdles to making that much power, you have to be able to use it -- I have 2.5kw available for example, in just one 50kv Spellman supply (and there's more supplies around).

Long before you (or I) can get anywhere near really using that, there are a host of other issues to handle.

Grids don't have a good way to lose heat in a near-vacuum, so grid/field design so they don't get so hot is a big one. The tank walls have air on one side, so that part is easier, but -- how hot are you willing to let them get? At 500w input, mine get really hot, such that I had to space my neutron oven away from them or melt it. And that's with a computer case fan blowing on it all through carefully made shield/channels to direct the air and stop the gammas.

Remember Richard's record run was ~400w total input. That's one of the "happy places" on all the curves. Our work so far seems to be telling us that with higher voltage and lower current and gas pressure, things get better -- and the power input can go down for the same amount of fusion.

The circuit I showed easily does a couple kW with bigger heatsinks (I used a big thick rack panel), when everything *else* is right.
At the few hundred watts level, it doesn't make so much difference if other things are right -- the fets can take a licking and keep on ticking. But you have to pay attention to the fact that you are basically driving a reactive load that has resonances, and there are ways to make that really hard on your driver, or easy -- and that's where most of the gain at the margin happens.

See Carl's thread on that -- when the fets switch matters compared to what the load phase angle is doing right then, and making appropriate changes to the load behavior is one of the keys to success.
You'd have to know Cliff -- but he's telling no lies whatsoever. Listen up when he has something to say and you'll learn faster with less pain and wasted bucks. You need hands-on too, but no point wasting that kind of experience and wisdom.

viewtopic.php?f=11&t=4789#p27627

Shows what he did, and it was a real smart thing to do. If you go through that and understand it, you'll be a lot closer to "getting" all this.

I personally like fets a lot more than igbts, but that's a question too complex for here. Fets tend to have a better SOA at the ratings they have, igbts have higher sat voltage so lose more power, and are usually slower switching but can have bigger other ratings -- so for them it matters more to have "everything else" right -- including heatsinks. Since I run on solar power (my whole 4 building campus) I became kind of an efficiency freak, not everyone has to be, but it's a nice place to be anyway --

[Linda Haile]

Thanks again for the advice, Doug.

I do read Cliff's posts with great interest. I've also been following Carl's present thread with interest and have read a lot of other threads on this subject.

While I appreciate that around 500W is generally sufficient for the average fusor, 3kW is about the maximum that can be driven from a normal 240V wall socket here in the UK. This therefore has to be the practical upper limit. (5kW supplies are generally hard wired here)

The reason for the extra power above the usual 500W or so is that, as previously posted in other threads, I'm planning to attemt to significantly reduce/eliminate neutrals as far as possible from my fusor.This will lead to increased currents for the same fusor operating pressure.

I believe the results you've posted regarding impurities in the system increasing fusion rates can be accounted for due to these larger ions ionizing more of the neutrals in the chamber, thus increasing the ratio of ions to neutrals.

As you point out, the limiting factor will be grid cooling. There have also been some interesting posts on liquid cooling recently using fluoroinerts. As posted elsewhere, I'm already experienced at connecting swagelok fittings to ceramic tube and I'll use this experience when designing the grids/feedthroughs.

viewtopic.php?f=6&t=3077&hilit=fluoroinert#p19230

Initial experiments will use lower power. As previously stated, the final design will probably consist of several modules in parallel.