HV power supply build - circuit safety components.

Things at the limits.

HV power supply build - circuit safety components.

Postby chrismb » Fri Jun 24, 2011 11:49 am

I am building an HV supply consisting of several stages with buck-boost CCFL converters.

I'll be posting this in detail quite soon so you can see progress - but first I need to order the HV components (from China/ebay)!!!

What I am building looks like this, with 10 stage;

HV_multi_supply_stack.jpg


Two issues I am facing, for safety and circuit protection;

1) Each stage mustn't exceed 3kV else the total chain will exceed the values of the coupling capacitors, yet could reach 5kV when unloaded. I can simply add a resistive drain on it to rein back the voltage, but are there any straightforward less lossy ways to limit each stage to 3kV? - e.g. is there such a thing as a 3kV Zener diodes or transorbs!?! Or would regular diodes work?

2) The stages are each fed separately but capacitively isolated from one power inlet. The risk is that if a capacitor fails in the chain then it could dump the stage voltage to the power inlet. What are the best means to couple the power inlets together such that if they diverge by more than a hundred volts or so (an HV transient short trying to get out) then it'll be shorted within the HV supply? e.g. are regular transorbs quick enough to stop an HV spike getting on to external lines, should I be thinking about using some form of spark gap, or are there recommended TVS components in the face of potential multi kV pulses? Or all three together - or more!?
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: HV power supply build - circuit safety components.

Postby Doug Coulter » Fri Jun 24, 2011 12:55 pm

Before getting into the other fine engineering details, I'd like to ask "why do it this way"? Is is just to break up the big AC supply into lots of little ones which are presumably easier to make?
(that assumption may or may not be a true one)

This does need more in the way of AC coupling caps (in terms of volt ratings) than a standard CW multiplier where you don't need increased voltage standoff for each stage. I'd guess if you made them all big (in capacity) you'd have a stiffer supply, but I've not noticed any big problems with that in CW multipliers run at decently high frequencies, and with HF, the need for stored energy in any cap or combination of them is less -- so the thing becomes less intrinsically dangerous to itself and the load. I'm not saying your idea is bad, just that I don't see the advantage of it so far -- seems like more work and more parts to get the same job done. It can be hard to make "more parts" more reliable then "less parts", but as usual "it depends". I know the big inverters that power my house are very reliable, and that they use a lot of little fets rather than a few large ones...but from looking at how they did it, that was a lot of work to tweak into working right.

I'd suppose you could optoisolate (though you might need a fiber to get the standoff volts) to regulate each supply based on the output of that stage. I'm just not yet understanding why this topology is one you favor. No one else seems to be using it at this point, though back in the day, there were a lot of floating batteries used in stacks, or various strange chainings of transformers.
In another approach, a single spinning shaft (ala the thing JohnF made recently) could power any number of isolated little generators, which output could be used directly or stepped up simply, and get rid of the need for the AC coupling caps and give improved robustness (other than mechanical).

I'm right now looking for fibers and transceivers myself for opto isolating things that will have to float way up high in my beam device. I'm hoping to find something cheap and standard, like what is used for digital audio -- I don't need that much bandwidth, but I want cheap and reliable for certain. And if a gel cell sitting up there turns out to be not enough power, I will definately build a smaller version of that mechanical energy transfer thing JohnF made, but probably with smaller servo motor as a drive and one or more small steppers as the generator(s) -- both are amazingly efficient and I won't need kw of power anyway for ion sources and focus supplies. You don't have to spin a 24v stepper very fast at all to get a couple hundred volts output, and it's nicely split phase so easy to rectify and filter with small parts. It will certainly be photogenic if I build that!

It has only proved difficult, but not that hard or expensive to find decent main HV transformers here -- or make them, so I guess I'm coming from a somewhat different direction. Is that what you are trying to eliminate the need for?
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 3515
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: HV power supply build - circuit safety components.

Postby chrismb » Fri Jun 24, 2011 1:10 pm

Doug Coulter wrote:I'm not saying your idea is bad, just that I don't see the advantage of it so far -- seems like more work and more parts to get the same job done. ...

I'd suppose you could optoisolate (though you might need a fiber to get the standoff volts) to regulate each supply based on the output of that stage. I'm just not yet understanding why this topology is one you favor.

It comes down to several reasons!

1) I am 'parts limited'. I have kept my eye out for big ferrite transformers here in the UK for 5 years and have never seen one I can buy.
2) Price. For euro 1.50 each I have bought 50 CCFLs that can put out 3kV and 5mA (simultaneously - albeit running them at 30V instead of 12V, but they seem to be good enough for that. Fully potted in Aluminum shielded enclosure.
3) The reliability of a zip is not usually a function of its length - get many parts all the same and you tend to get advantages of construction accuracy and reproducability. Hgh volume manufactured parts are always more reliable that hand-build (nothwithstanding design flaws!)
4) A stiff supply, as you mention
5) A less rippled supply - as each one can run at its own frequency and phase, you get the averaged sum of ripple at any one moment.
6) Reversing polarity - you just tie the supplies' common connection to wherever you want the 'ground' to be on the stack. Just one single wire swap gives you reverse polarity or +V /-V with the ground in the middle, or any such output configuration. As long as you use the same rating of cap throughout, of course, which costs a little extra, but I am planning on it, on account of the increased flexibility I get.

Is that enough reasons for now? ;)



It can be hard to make "more parts" more reliable then "less parts", but as usual "it depends". I know the big inverters that power my house are very reliable, and that they use a lot of little fets rather than a few large ones...but from looking at how they did it, that was a lot of work to tweak into working right.

As above with the zip, but also if one of the supplies goes down, you just replace that one single part, not the whole transformer. Plus - your power keeps on powering, doesn't shut down suddenly if you lose only one stage.

The other thing about power shutoff - you can switch each supply individually and get a discrete voltage control without a variable input voltage. Just switches give you the option to cause stages to drop out, simplifying voltage control (so long as you are not after precise control).
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: HV power supply build - circuit safety components.

Postby Doug Coulter » Fri Jun 24, 2011 1:14 pm

To get a little closer to your actual question -- here's how Spellman does things. We bought some parts from a fellow HV experimenter, and he'd de-potted a Spellman stack. These stages are rated 10kv each, but only usable to about 5-6 kv outside the potting.
BacksideVM.jpg
Backside view

The blue thing is a resistor across the stage output. The inductors separately decouple two sets of rectifiers that are otherwise in parallel. The red things are surge absorbers, presumably standard units. Here's the other side of the board.
FrontsideVM.jpg
Frontside view

This one shows a homebrew transformer, good for roughly a kilowatt, and the caps and diodes. Note how large the diodes are...over-design for safety in impulse (arc) conditions. The caps are .022 at 10kv here, polypropylene types with good pulse characteristics.

Other parts of this supply (not currently easy to get to and not in this building) include protection from an arc across the output, which is more or less a series resistor made of a whole lot of 2w carbon resistors in series-parallel and also potted, as once this is built, the main risk to it all is an external arc. The fact that every stage uses the same value caps means that a quick discharge happens fairly equally across all stages, so eliminating the risk of seriously overvolting any one stage -- The diodes sill have to carry the peak current from the AC coupling caps in that case, but the series ballast limits that peak to the couple of amp range, well inside what these diodes can take. There is further a big resistor across the entire thing for voltage monitoring (also potted), with a floating ground end for sticking a little current measuring resistor into.

As you know, I really like their designs, and they've been utterly rock-solid here in deep abuse, which might be why I like them so much (and CliffS is a hero anyway).
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 3515
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: HV power supply build - circuit safety components.

Postby Doug Coulter » Fri Jun 24, 2011 1:25 pm

Gee, you're forgetting the mail does work between our locations. I've mailed a few things to your country, no problems. But you have to ask me (or anyone else here). I like the idea of using a bunch of CCFL's in series actually, though with the ones I get I'd do it slightly differently, skip the coupling caps, and just run each off nicads or something -- you could make a rig to mechanically connect to a bunch of battery pack contacts to charge them all between runs. If you ran each off a small generator (think dirt cheap surplus stepper motor or something like) then no batteries at all. Though Jon tells me you guys don't see things like that commonly there. Here, they cost more to ship than to buy and they are all over the place.

Ripple added at random might actually be worse, but more likely they'll phase lock if they are all oscillators. Else you could get some fairly horrible "beat notes" at low frequencies that are *harder* to get rid of. Though with the CCFL's I use, even small caps make ripple utterly nil -- hard to even measure at a millivolt/scope div.

Reversing is not going to be hard either way if you build it yourself. For the huge one I'm working on (stalled out because I didn't need it yet), all you need to swap polarity is a single extra pair of AC caps and to mechanically turn the whole thing upside down.

In my experience, the worst risk of things getting fried is diodes and that in an arc across the whole thing - big peak discharge currents can pass in even small caps if their quality is good -- you need a current limit/ballast for certain to keep peak currents down in that case, and to limit how much of that junk you need, you need to make sure other arcs just can't happen internally.

I do worry that for example your coupling cap on AC8 has to store a lot more energy to be the same uF and have a higher volt rating, which makes things have more stored energy and much higher chance of failure should the thing arc.

With a straight CW design -- you can hook the transformer and ground anyplace in the stack too, BTW -- and not only that, do the same thing with the transformer stuck on either end, if the transformer insulation will hack it. I've not yet managed to fry a transformer even in extreme abuse. The push pull nature of a full wave stack seems to protect them well, and the driver behind a lot better than you can do in a single eneded stack where the discharge currents don't cancel.

Edit, note if your potted supplies have the usual series ballast cap in them, usually about 10-20pf, it's not going to be very "stiff" no matter what you do unless you can un-pot them and put in a decent one.
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 3515
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: HV power supply build - circuit safety components.

Postby chrismb » Fri Jun 24, 2011 2:03 pm

Doug Coulter wrote:Edit, note if your potted supplies have the usual series ballast cap in them, usually about 10-20pf, it's not going to be very "stiff" no matter what you do unless you can un-pot them and put in a decent one.

I've hooked them up feeding one stage (voltage doubling) and put across it a 500kOhm 12W resistor (pack - 4 x 3W) and it will push 6mA/2300V/11W into the resistors with a 25V/0.75A input into the CCFLs. Ran them like that for 30 mins and the inverter did not get over 30C. (The resistors were smoking, though!!!).

10 of those stacked up would therefore give 23kV/6mA, and it would've cost me (in approx US) $15 for the supplies, $40 for the capacitors and $20 for the diodes = $75 for all the parts needed to build a 23kV 6mA supply. Should I presume I can do better than this by some other means? Dead easy to assemble, too!

I've also cranked one through to max on my 30V supply, and it was pulling 1A with 3100V reading across the resistors. I didn't leave it for long as the resistors were overheating - the CCFL wasn't. I am beginning to feel maybe that's to high a voltage, but I will build it as if it is a 30kV supply and operate only intermittently there (if I ever feel a need to, at all).

Left unloaded, the stage runs up to around 4750V, but no breakdown so far has occured left at that voltage.
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: HV power supply build - circuit safety components.

Postby chrismb » Fri Jun 24, 2011 2:04 pm

Doug Coulter wrote:Ripple added at random might actually be worse, but more likely they'll phase lock if they are all oscillators..

Why do you think they might phase lock?
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: HV power supply build - circuit safety components.

Postby Doug Coulter » Fri Jun 24, 2011 2:13 pm

Because they do phase lock in practice given only the slightest hint of the signal from another one here. In a Royer oscillator, it's natural to have that happen. I use unpotted one (admittedly, I pay more for them, new) and it takes very little to make it happen, sometimes just being nearby is enough. When I want to be sure (or I want to control whether they lock in or 180 out of phase), a few pf between them on the transformer primary gets it done...
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.
User avatar
Doug Coulter
 
Posts: 3515
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: HV power supply build - circuit safety components.

Postby chrismb » Fri Jun 24, 2011 3:06 pm

Doug Coulter wrote:Because they do phase lock in practice given only the slightest hint of the signal from another one here. In a Royer oscillator, it's natural to have that happen. I use unpotted one (admittedly, I pay more for them, new) and it takes very little to make it happen, sometimes just being nearby is enough. When I want to be sure (or I want to control whether they lock in or 180 out of phase), a few pf between them on the transformer primary gets it done...

Mine will be spaced 50mm apart, or so (to make space for the diodes/caps) and will be fully shielded.

I guess we may get to see, because I am planning on a current monitor port for the PSU (this being one of the reasons for making a custom supply) and can look at the behaviour of the ripple then - assuming I get that far in the project.
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: HV power supply build - circuit safety components.

Postby chrismb » Fri Jun 24, 2011 3:06 pm

So, back to the questions; what are the best ways to protect against fault transients getting back to the supply, and are there any 3kV Zeners?
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Next

Return to High Voltage, High Power

Who is online

Users browsing this forum: No registered users and 2 guests