An HV stack that sums volts and watts from multiple sources.

Things at the limits.

An HV stack that sums volts and watts from multiple sources.

Postby chrismb » Sat Aug 25, 2012 1:06 pm

OK, so I posted this on fusor.net. Doug has commented negatively on it in another thread.

If there is prior art which shows a serial voltage stack with multiple supplies feeding its different stages, then please let me know so I can withdraw this (it is currently a patent application with the UK IPO).

I grant you, it 'looks similar' to a lot of stuff, and also looks pretty obvious as soon as you see it, but show me someone who's built a stack like this before.

If you ask 'why bother with this design', then I have 30kV supplies in regular use right now that put out 5mA each which I built for less than 25 quid each. I've never even seen supplies of those parameters advertised for sale in UK before, let alone what I imagine they would have likely cost me.

I've also never seen stack-based HV supplies that have DC isolation such that you can put them in series to further add their voltages up.

But if you know otherwise and someone has built such supplies before, then please do let me know.

hv_psu_pat.pdf
(138.15 KiB) Downloaded 469 times
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: An HV stack that sums volts and watts from multiple sour

Postby chrismb » Sat Aug 25, 2012 1:21 pm

In regards the question of whether this amounts to a big spend on the isolation capacitors at the top of the stack, this is what I put in the patent application, and also serves to show how to reduce the capacitor BOM costs in 'a regular CW stack' too;

In prototyping work, it has been established that the coupling capacitors, C1 to C6, need only be
quite small in comparison with the output stack capacitors. In one set of prototypes, 10 nF
capacitors were used for the output stack capacitors, CA to CD, whilst the coupling capacitors, C1
to C6, were varied from 50 pF to 1 nF, with the power supplies operating at 40 kHz. Little
advantage was found in using coupling capacitor much above ~470 pF in this example when the
output stack capacitors were delivering 5 mA at 25 kV from a 6 staged topology.
Using 100 pF capacitors provided efficiencies better than 90% of the 470 pF capacitors (which
were, in turn, practically indistinguishable from using larger capacitances). Using such relatively
small capacitances for the coupling capacitors therefore provided a significant cost advantage in
prototype assembly because these capacitors need to withstand the full output voltage of the stage
they are feeding, yet higher capacitances at high voltages are costly, so there is a strong commercial
advantage to minimising the capacitance specification of the coupling capacitors.

In a further prototype design, two 100 pF 30 kV capacitors were connected in serial to provide an
isolation of 60kV. The 50 pF effective capacitance was still satisfactory to drive a 2.5 mA current at
55 kV into a stack composed of 12 stages of 10 nF capacitors with only a further loss of efficiency
not measurably greater than 5% over the variant using 100 pF coupling capacitors. The ripple
current was measured under load and found to be related principally to the output stack capacitors'
values, not the coupling capacitors values. Lower coupling capacitor capacitance only appeared to
reduce efficiency, rather than affecting ripple current.

This may be useful information to practitioners in the field if diagram 7(ii) is reconsidered. This
depicts the prior art of a 'standard' multiplier stack in common use. The above tests suggest that the
capacitors to the left of the diode chain may be of significantly smaller values than the output stack
capacitors to the right of the diode chain, without increase in ripple current and only a minor loss of
efficiency, if any*. It is common practice in these stacks for all the capacitors to be of a given value,
or, otherwise, for the lower capacitors to be given higher values so as to 'support' the upper stages.
However, the work on prototypes of this present invention show that the capacitors in series with
the power supply can be significantly smaller; employing coupling capacitors with less than 10% of
the capacitance of those in series with the output load has not been significantly detrimental on
performance in the prototypes built.

*(Once it is recognised that, in a Villard design, the 'coupling' capacitors that feed the central diode
node of the pair that rectify across each 'output stack' capacitors (serially connected to the load)
serve different functions, then it can be appreciated that the coupling capacitor will tend to swing
through at least twice the voltage as may be present on the output stack capacitors, therefore the
capacitance value needs be only one quarter (as the energy content that the capacitors can deliver
per AC cycle is the square of the voltage). In practice, when load is being drawn from the output
stack capacitors then the voltages on the coupling capacitors may swing through an even larger
voltage, such that the capacitors would still likely be suitable with a lower capacitance still.)
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: An HV stack that sums volts and watts from multiple sour

Postby Doug Coulter » Sat Aug 25, 2012 3:36 pm

I won't argue that this isn't good for the money for 30kv at a few mills - way not enough for a fusor, of course. I've done that with a single large CCFL and a standard stack too. Obviously, having more input power is nice. Obviously, if you can't just build a bigger transformer and LV driver, this is applicable. But there IS prior art, I've seen schematics of old Spellman and other HV supplies that do similar things, but they tend to use a single LV driver to drive multiple outputs, since once you're making all that yourself, it's cheaper yet.

The only reason to buy a bunch of CCFL's to do this is if you can't manage the other stuff on your own, and you happen to get lucky with cheap off the shelf CCFL's, which you did.
It IS nice that they don't have to be in sync or even at the same frequency for it to work. But you don't NEED that with one big LV square wave driver and either multiple transformers, or just the plain old CW stack.
It's going to be a nightmare to regulate and get to low ripple...but possible.

I have no idea if anyone has patented this particular approach, but I'm dead sure plenty of people have used it, because I've seen it before. Usually in really extreme high voltage supplies - like 160 to 250kv and at substantial power, using a single driver but multiple transformers/capacitors. RCA had almost this exact design shown for using a 100v ac source to make all the taps voltages for phototubes, back in the early '50's or so.
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: An HV stack that sums volts and watts from multiple sour

Postby chrismb » Sat Aug 25, 2012 5:08 pm

(Copied in from other thread)


Doug Coulter wrote:Price has nothing to do with patent-ability, and Spellman among others has published this design for a few of their supplies going back decades. Nuff said?

Feel free to find one. If you do, I'll paypal you $20.

But suit yourself, now you can go sue Spellman, Glassman, and others who used this trick way long ago

Doug, there's no need to be like this. They either have done it, or they haven't. If they have, it is prior art, unpatentable. I did the searches and nothing came up. I did more than that and hunted further literature. Nothing that I found. Maybe there is. But I searched for a long time, and came up with no such previous use of multiple supplies feeding one stack.

It's obvious once you see it, but nothing quite like this in such a simple form has been done that I have found.

Feel free to prove me wrong and get your $20 off of me. Show me an example of a single stack of stages being fed by multiple capacitively coupled power sources that share a common ground, and come collect your $20!...



that last cap having to be very high voltage (and thus higher joules)

I run 100pF/30kV link capacitors up to 30kV, then two of those (50pF/60kV) on further stages up to 60kV. Gee. 0.1 J. Big joules.... :?


And you do need a decent sized ac feed-in cap if you want any regulation worth talking about - with a non-stiff supply, feedback regulation is a very tough loop to close without it hunting or oscillating.

I've put the detail into the patent. If you do not wish to read or believe the results, nothing much I can do about that.

If you bother to try this out for yourself, rather than simply dismiss it without generating any numbers, you'll discover that the link caps only impact efficiency, not ripple current and regulation. You can disagree all you like, but I've done it, measured it, written it up. Try it, you'll see for yourself.

The reason is quite simple if you think about it - as the input AC oscillates, the cap follows the phase. As the cap lags behind the input AC, say in the negative going stroke of the input cycle, so the voltage across it goes up considerably, because it is undergoing a 'differential' charge. As it reverses and goes back the other way, however much it was on the 'wrong' side of the lower potential of the stage in the negative going cycle, so it will be equally effective on the top of that stage's potential in the positive going cycle.

So the cap can lag on each stroke, but the potential by which it 'lags' adds up constructively in the next half-cycle.

Basically, the link capacitor can see double the peak to peak volts or more in the applied AC, and that's fine because it makes no difference to the ripple voltage so it makes them more effective if they undergo more voltage deviation and the full pk-pk volts during a cycle, whereas the 'storage' caps, in series with the load, you want those to undergo as little voltage deviation as possible. As the energy content is the voltage squared, you can see immediately why the link caps can be really small in comparison with the storage caps.

Using smaller caps for the 'coupling' merely causes them to undergo much higher duty cycles. Clearly you'd be able to take that too far and cause them to fail. But the ceramic disc types I use (100pF/30kV, retail purchase price of $0.40 each) do fine and do not appear to overheat at all under a 5mA load. (100pF at 40kHz is ~40kOhm impedance... I guess that means about 0.3W per cap when carrying 3mA?)

If this was well-known and understood, I'd expect to see the building and recommendations for CW stacks to include this detail, because, as you say, HV caps are expensive. In the tests I have done, it suggests you can substantially reduce the capacitance of half the caps (the caps not in series with the load) in a CW stack without ANY effect on the output at all. You might have to turn up the input a little to accommodate work losses in those link caps, but if they are of sufficient quality and can handle the RMS current they are exposed to, then any increase in input is marginal. I have found it to be essentially unmeasurable (maybe a little loss noticed, but measurements are below statistical significance to report).

IF course, had you done the legally required searching for prior art the USPTO requires (but most ignore) you'd already know all this. It's up there.
$20 in it for you if you show me a real document showing this.

Having said that, double the volts and current and you might have something you could run a fusor from and it'd be a boon to amateurs, who don't have to pay a bit of attention to patents on things for their own use anyway.

Doug, do you really have an understanding of this patent? Just read it. There's clearly stuff you haven't read, maybe there might be something that will surprise you!

You can double the volts and current easily, simply by having 4 of these and putting them together. There's even a diagram to make this plain-and-simple to understand. They can be serialed (with sufficient isolation in the caps and inverters) and paralleled, no problem. This is a key part of the advantage they bring. Just have a small pile of them, and set them up in any order and configuration you like to achieve your HV electrical requirements of the day. High current - parallel them all. High volts, serial them. A bit of both - use them individually, or in pairs. Whatever.

This idea is all about flexibility to achieve a specific HV load requirement, and the capacity to make use of mass-produced low cost components.
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm

Re: An HV stack that sums volts and watts from multiple sour

Postby Doug Coulter » Tue Aug 28, 2012 4:19 pm

Sure $20 for research that is normally charged out at around $400/hour, yeah, I'll go for that. Not. I'd have to ask Cliff if I can make public something he swore me to secrecy on.
There are advantages to multiple feed ins, to keep the ripple current at the low end of the stack lower - which is why some earlier CW designs used bigger caps at the bottom, going down in value on the way up.
Then they found out what that does in an arc situation. You might find that the saturation of the royer oscillator cores is the only thing saving your little supplies in an arc situation.

If your AC supply is the limit to regulation, then yeah, the ac cap size isn't as important. But taking reductio ad absurdium, why not then just use a value of zero uF if it doesn't matter - you're not adding up, and yes, I've tried this. Yes, for HF square waves, you don't need much cap, but you do need some, and any effective reactance shows up as effectively a series R in the output when it comes to voltage vs load. You can't cheat that one, and plenty of old out of date patents (like the CW one itself) mention this. I did see you mention that this supply has high output impedance (it's squishy) and the small caps are one possible reason why, the other being that your AC sources are too.

No, I've not had time to read your patent yet - been busy with other things, including the latest egregious misuse of the patent law (apple vs world). Rounded rectangles? Prior art goes back to things that fit in shirt pockets, but Samsun wasn't allowed to have evidence admitted that their shape actually preceded the iphone by a month to market - damn swift "copying" if you can to it a month before the "real" thing, eh?
Yet they lose in a court with a stacked jury and dishonest jugde that favors "the home team" even though Apple is Chinese made and keeps their profits overseas and doesn't pay as much US tax as Samsung!

Which is why patents, like Samsung's which are required to run a 3g phone - they're pretty basic - don't matter. Apple has more money and just bought a billion buck award with only about 10 million on lawyers so far.
An Samsung were only asking 400k for all the phones Apple's made that violate those patents (which every other phonemaker pays (and we customers all pay indirectly), while apple wants $40 per unit for everything Samsung ever made - and got half that. The system is so fucked I just don't even like discussing it that much till we get out the guillotines and eliminate crony capitalism the only way that's ever worked in all history.
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: An HV stack that sums volts and watts from multiple sour

Postby chrismb » Tue Aug 28, 2012 4:53 pm

Doug Coulter wrote:If your AC supply is the limit to regulation, then yeah, the ac cap size isn't as important. But taking reductio ad absurdium, why not then just use a value of zero uF if it doesn't matter - you're not adding up, and yes, I've tried this.

I've already explained - it reduces efficiency, but it does not reduce ripple. A good engineer would pick an optimum, wouldn't he? Where the loss of efficiency is minimised for the maximum cost saving.

Would it reduce output impedance to have smaller coupling capacitors? Wouldn't that be dominated by the much larger output capacitors in series with the load? Can't say I've seen any change of behaviour in load dependency between using different link capacitors.

As I have explained, there are two types of cap in a stack and they do fundamentally different things. So what's the reason the same values are used, either side of the diodes, in a typical CW stack? Seems to me either the designer can't be bothered to pick two different values of cap to save money (seems like a poor commercial outcome in that case), or they don't realise that the voltages that cycle on those caps are different and that the link caps undergo bigger voltage differentials (thus can be smaller in value for a given power throughput). Well, I don't know, maybe the cost saving is minimal and not worth picking out different values. There I was thinking every penny counts in business.

In any case, do you really have HV power supplies that you can simply string together in series for more volts, like your typical bench power supply? I've seen a few floating supplies before, but typically only rated for isolation to their own working voltage.

Did you see the embodiment for reversing a high voltage? Effectively instantaneous polarity reversal. I can't say I've seen that before.
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 6:32 pm


Return to High Voltage, High Power

Who is online

Users browsing this forum: No registered users and 6 guests

cron