New HV stuff, questions

Things at the limits.

New HV stuff, questions

Postby Doug Coulter » Sun Sep 19, 2010 1:43 pm

Bill, our master scrounger, just came up with the pictured stuff (among other cool things), and I have some questions for anyone (Cliff? John?) who might know more than diddly about this stuff.
Here's a pic of some of the new goodies.
NewHVParts.jpg
New HV parts


Here's a detail of a stack disk:
Stack.jpg
Stack disk


And here's what my sleuthing says is inside one of the disks:
stackstage.gif
Schiz of a disk


Now for the questions. The seller says each disk is rated at 20kv output -- does this sound reasonable, or can anyone find data on these? Seems Wallis High volt was absorbed by another company that has a not working web page. As these are full wave doublers, that implies I need on the close order of 20kv pk-pk on each drive phase.

The smaller transformer appears to be a Spellman, identical to the one CarlW has been talking about on fusor.net, and mine resonates at 11.5 khz, quite high Q.

The huge Wallis transformer is about 85::1 turns ratio each side, and resonates at 127khz, also high Q. It's all FAT litz wire, even the secondary is fatter than number 20 solid.
Ah, perhaps I got faked by a parastic (leakage L) resonance on the Wallis. It has another resonance at 28khz or so, which acts like a 50::1 turns ratio too.

Starting from here....and assuming I really need 20kv pk pk -- what would y'all do? I can pretty much make anything electronic, no troubles. I see Carl has used an inductor across his transformer to raise the resonant frequency some -- a good idea, but it would seem to have some limits due to increased circulating current. I'd guess you'd do the opposite on that Wallis transformer to lower the resonance some, I kind of doubt the stack diodes are all that fast. (or see above, maybe run it at the lower resonance)

What's everybody's favorite topology for this sort of thing? I can see where the circuit that Carl wound up with is good, and solves the problem of eating the harmonics on the drive waveform so these big resonances don't draw tons of current during switching harmonics, for example -- possibly an added feature of having that extra series resonant circuit in series with the drive?

Having looked inside a couple of Spellmans, I know they like the lm3525 kind of pwm chip (and so do I, I've been using them for other things for a lotta years).

I am not looking for super output power here -- just don't need it, my Spellman SL2KW is way more than enough watts, just not enough volts for some things I want to do here, and sadly, Cliff, I can't afford to buy a higher volt one new right now (maybe next year). I'd have to take my wife to Hawaii to calm her down after that and I can't afford both ;)

It's interesting that these stack disks have a separate cap in them. The fiberglass board sitting on top seems to be a choke with 2 47k resistors and 2 spark gap arresters series-ed across it, which would seem to be intended as an output filter using that extra set of caps. (the spark arrestors appear to have taken some heavy abuse).

I'd prefer to use the Spellman transformer if possible -- it looks better-made, and is obviously a lot smaller, which will simplify packaging -- if I can run it fast enough to get to voltage before I get to saturation. From the size, it will do plenty of watts for my uses.

Here's a dirt simple driver that I like when I can live with its limits....I did up a PCB for this guy and have used it for testing things. Amazingly, the FETs rarely need heatsinking if things are right otherwise. Click the pix as usual to get a larger version you can actually see. I only add some proximate stiff power supply filters and an RC snubber to their example here, and it works great -- you can add an open collector to the RC timing to get current limit/shutdown and that works well too. Of course, you can't change duty cycle, and deadtime is a function of the fet and gate R (it works well if you get the R right), so to change output, you change the input volts. I run mine on a 7815 from a separate floating xfrmr. Hard to beat for simple, but not flexible.

IRS2453.gif
IRS sample circuit


I've used this one with great success to drive a Glassman stack and transformers with, just straight in, no fancy L's and C's outside their little TV flyback-appearing xfrmrs. Those have two primary windings, one on the opposite side of the core, I'm guessing to get some amount of leakage L. They used a half bridge with 6.8 uf in series in their (super overly complicated) circuit.
That thing must have had 40 trimpots! So I ditched it and just kept the "good parts".


Thoughts?
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: New HV stuff, questions

Postby johnf » Sun Sep 19, 2010 3:37 pm

Doug
I've got a similar stack in a big box thats been sitting in the carpark for the last year
Mine says Advance Hivolt similar or the same capacitor discs.
got a funny feeling that the 20kV is the peak rating on the caps so you should use the preadsheet on the VM website to get a handle on what voltage is going on the caps for what drive voltage and pick some nice safe value below this. 8 -10kV would probably be a good start.

Also the cap discs being 20kV gives us a clue as corona loss per stage rises rapidly 20kV upwards. lots of Ozone isn't a good thing for electronics or humans

My unit came with the power supply and transformer so there is a 3 unit high full depth box with the drive electronics a five unit high box about 1 foot deep with the High freq transformer in it then the massive box with the stack in it.
Someone had dropped the transformer box and one of the ferrite cores is broken. From memory the disc stack in mine is two groups of 5, one group being the multiplier the other an output smoothing filter cap

Now its time to measure your transformers exactly for L, Fres, Coupling factor (leakage inductance) and R from this you can calculate inter turns C, interwinding C,

Use LTspice if you haven't got a spice simulator, blowing up virtual silicon is much cheaper than the real stuff

over on fusor.net I gave Carl a reference to a good book on topologies for SMPSU design


Edit
The resistors and spark gaps are for output arc protection to protect the diodes and capacitors in the stack from dV/dt, the spark gaps do not last forever --replace them.
johnf
 
Posts: 433
Joined: Sun Aug 08, 2010 3:51 pm
Location: Wellington New Zealand

Re: New HV stuff, questions

Postby Doug Coulter » Sun Sep 19, 2010 4:13 pm

Yes, the 180 kv "rating" on this seems a little optimistic -- it's only a little over knee high, not much to stand off that much voltage. It did come with a large corona donut.
But the disks aren't that thick (about 1.5") and those bolts and sockets imply that the spacing internally between inputs and outputs is pretty darn small. I mean, OK they are potted and maybe use PCB material or better for insulation there, but....20kv sounds a bit much -- and it's more than I actually need anyway. 100kv would be a huge step up from the 50 I run now, and require a lot more work on keeping us safe and getting it to the inside of the tank.

I'll see about finding that book. Edit:: just found it used for 50 bucks and ordered it. In an edit of the above, I mentioned that there is something a little fishy with the Wallis transformer -- I then did a wider sweep and found a resonance at 28khz (a lot more believable) so that first one must have been some kind of resonance with the parasitic L's and C's in the thing. Actually, that might be good if I just wanted to drive it with a plain old sq wave, as the 3rd and 5th harmonics wouldn't draw a lot of current due to that leakage L. But I'd rather use that Spellman transformer if I can, they are just higher quality and a heck of a lot smaller. I only need 500w at most here -- which that dumb simple driver circuit easily makes with cheap fets and 2 sq in of PCB mount heatsink each....it's cool as long as you don't need fancy features, but of course current limit and voltage regulation like a "real" supply is always nice to have, and kinda hard to add in to that unless you make another switcher for the fet rail power and use that to get there, yuck. It does at least solve the problems of driving a full H bridge nicely -- that thing really works and has a ton of sequencing stuff in it to make sure it never does something stupid. One wonders if it wouldn't be OK to run in "stutter mode" just shutting off the timing RC at some duty cycle to regulate the output. Ugly, but it might work at almost zero cost. You'd probably hear it "squegging" in that mode -- maybe a useful eyes-off indicator, actually? I like that kind of thing, and have all my counters on a multichannel audio as well as into the computers so I can hear what's going on instantly -- helps a ton when tuning this and that and far quicker feedback than looking at a bunch of gages and scope traces.

Fets are fairly robust -- I avoid frying them by simply making the rail supply kinda on the wimpy side and it gives up first. Nice big SOA on those guys. Of course you tune for lowest quiescent current at low levels first...and keep an eye on case temperatures. I've not yet had the courage to run off a bulk rectified mains supply....Remember I'm running off PV solar power and my inverter (4 grand plus!) is not something I want to fry by accident either. Of course, it's smart enough to shut down if you short it, but it has to be a good short -- .1 ohm it will just burn up the load. But it doesn't like HV spikes being pushed into it backwards.

I can believe the original was huge. That transformer box is about 10" on a side and has a big whiney fan on it. Probably many times the power rating I need. Here, with fusors, we scale things down as we get better Q for a lot of reasons, one being operator survivability. We are right now at a Q of 3x what Richard was getting at his record run at HEAS in stable DC mode, and about 100x more than that in a pulse mode we're playing with. At that point, you run the power input way-way down unless you want to die! We are finding that fusors run better with a pulse into an initially empty inner grid....by quite a lot -- seems like we get full speed fast-on-fast collisions that way, rather than slow-on-neutral types that you get from space charge in the grid center. It would account for the about 300x best case difference we have from Richard's results quite nicely. I think that first 3x is just precision of our grid and the better geometry, along with better gas purity (we tested that last and it is for sure a big deal). But it's still speculation till I can get in there and measure some things better -- hours in the day and all that.
More wiggle sticks and probes need to go in there next time I let the tank up to STP. Which will be soon -- I have a ton of new neutron detectors to test tomorrow, then I can tear it down for upgrades. Bill really scored big on that last trip around the country -- we're rolling in cool stuff at this point.
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: New HV stuff, questions

Postby johnf » Mon Sep 20, 2010 1:24 am

HVstack.jpg
Doug

hears a pic inside the HV stack I've got

is it similar??

I took the Al corona shields off to show internals of plastic blocks

left most rectifier stack right most capacitor stack through big series resistor

Note I got it wrong only 4 caps per stack

allegedly a 50kV 100mA supply
johnf
 
Posts: 433
Joined: Sun Aug 08, 2010 3:51 pm
Location: Wellington New Zealand

Re: New HV stuff, questions

Postby Doug Coulter » Mon Sep 20, 2010 6:26 am

Similar but my stacks have more connections - 4 innies and 4 outies -- see schiz -- that second cap stack is also in each disk and the center hole is round and smaller. I think all these of this sort of design were meant for serious power. In my case, I'm interested in more volts, don't need too much power, but do need the push. I am going to turn these upside down (have to add two initial caps to do that) for negative output, as I have an 11 stage Glassman stack rated for 125kv at 50 ma (zowie) that works already, and would be a lot harder to "turn over" than these. So if I wind up with plus and minus 125 kv, that's power for a beam-on-target machine that should really sing. Of course, handling those voltages in air is going to be a trick -- I'm going to try and avoid oil, but may do some encapsulating and use some thick plastic pipe etc. If I wind up shooting at silicon wafers from both sides one of those supplies should do it, but I also want to play with "harder" fusion fuels and will need the volts for them to "go". P-Li (17 Mev output) has been a favorite of mine, but the instrumentation to see hot alphas in a sea of hot protons is going to be "interesting" at best. Nice thing about DD is the signature neutrons (and also, the thing you hate once you've had enough of them for long enough). So I'm using DD to learn as the metrology is easier for that, but planning to move on if/when I get to a really good place on that. If others want enough neutrons to activate the whole periodic table, I have no argument with that, but it's not my focus --
I'm after power gain, with or without them.

I'm wondering what on earth I'm going to wind up using for bleeder and voltage monitor. I don't want to eat much power in that, but....a zillion 10 meg R's in series? I built some HV probes (freq compensated, neato) with those, some applied distributed capacity, and pyrex/oil to contain it all.
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: New HV stuff, questions

Postby Cliff S » Mon Sep 20, 2010 9:22 am

Hey Doug,
I am familiar with these stacks. They are used as the accelerator supplies for ion beam implanters for wafer fab. 160kV is a very common voltage. 20kV per stage is very typical. We have stages that run up to 25kV per. From the date codes on the stack “wafer stages” I would say these stacks are no more than 10mA, as that was near the limit for implanting at these voltages back then. Even today, still the power for high energy implantation does exceed a few kilowatts. Low current implanters can reach up to 40kW, and some specialty apps, even higher. See pic attached. This is a 120kV, 1A, 120kW implanter power supply we recently developed using our new ST 12kW rack line.

I also am familiar with Wallis HV, as one of their founders was a close colleague of mine. After he left Wallis, he went on to start a new HV company in the UK, “Start Electronics”. He partnered with Spellman in the mid 90’s, and when he retired, we bought him out 100%, and now that facility is “Spellman UK”. Wallis is now known as “Hitek”. (http://www.hitekpower.com). They are a smaller version of Spellman, and still make these types of stacks, as does Spellman. (Our SLS series, http://www.spellmanhv.com/en/Products/R ... s/SLS.aspx ).

The small inductor/R/SG’s are not for arc limiting as John thought (sorry John). They are the LR for the filter stage, working with that “extra” cap you see on each stage. The L needs to be clamped so it doesn’t see high voltage across it during arcing in the implanter, (and boy do these babies arc!). Often there would be an external arc protection resistor, external to the stack. There may also be “distributed limiting”, R’s in each stage of the stack to limit discharge current.

The Spellman DXR XFMR may not drive to the full voltage you will need to get 160kV out of the stack. This would require about twice the voltage drive. I don’t think the secondary layer to layer insulation will hold that off for long…but maybe… You will need to drive the primary with 300V, since the DXR drives to 150V. You will need to double the frequency the DXR runs at, typically up to 25kHz, (go to 50-60kHz), in order to avoid saturation. If you don’t have any luck with XFMR’s, let me know. I may be able to scrounge a set of coils from our “25kV per stage” drives.

Best Regards,
Cliff
Attachments
ST.jpg
Cliff S
 
Posts: 2
Joined: Thu Aug 12, 2010 9:27 am

Re: New HV stuff, questions

Postby Doug Coulter » Mon Sep 20, 2010 9:37 am

Thanks Cliff! I may take you up on the scrounging at some point, but for now it's probably easier to blow up stuff I have in hand already ;) When I run into the limits of that, I'll get real interested in your spare pile.

That's one heck of a power supply stack you picture there -- I'd bet John's outfit would be drooling over it. I probably couldn't even light it off with the mains power I have here.

So that Spellman transformer is maybe not good for this. I'm kind of wondering if putting a small enough inductance across it to get near 50khz won't have its own set of issues. Would you not then run into some good sized losses in that inductor, and in the series R of the transformer with the high circulating current due to its C? I'd rather not trash that one -- if it's not ideal for this, there's always next month's project.

The Wallis Xfrmr had some insulation skinned off the litz on the secondary, which I was able to pry apart and re-dope, and it seems good, if on the huge side for the power level I want. Luckily the litz conductors are only shorted together for a couple of turns near the end, so it's probably still fine. (no shorts between the turns, just within a single litz)
It seems to resonate at 28khz or so, with a bunch of peaks farther up the spectrum, I'd guess due to parasitics in it. The glue joints on the core pieces were broken some too, but pushing things tight together there or pulling them apart a bit doesn't seem to make much difference on the bench oscillator, so maybe it's usable. It was surely made for high current output, that thing is fat wire indeed. I'm of a mind to just try a plain old square wave on it, as the impedance would be high for the fast edges due to the existing series L it has.

So when you give a current output, would that be for the entire stack, or one stage? That low number would seem to have to be for the entire shebang (with the bottom disk taking a lot more, obviously). 10 ma is plenty for what I want to do for now if that's the net output at voltage. Except for the usual experimental glitches and arc issues, I plan to go with about a 1 ma beam current, which is plenty to fry a target in my case. And actually a bit of challenge to get if you insist on nearly all monatomic ions, as I do. Hence some of the discussion with John about pre-selecting them out of a beam of mixed type ions, pre acceleration. In this case wasting power means overheating a fairly temperature sensitive target, and molecular ions are mainly a waste of power for me.

/////

I noticed you mentioned variable frequency on fusor.net -- do you care to share any thoughts on that here? That's one that never occurred to me. Do you work around the slope of the series resonance in that type circuit to get variable output with that? Interesting idea if so, and maybe not that hard to implement either.
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: New HV stuff, questions

Postby Cliff S » Mon Sep 20, 2010 12:29 pm

Doug,
I wouldn't worry too much about winding losses at 50Khz in the DXR, assuming your load power would be well under 3kW. But if in doubt, take the finger/temp probe test to the coil and core, (after HV is turned OFF of course). 50-60 on the outside coil is acceptable. 70+, it’s getting too hot. Core can run 80C-90C...it likes it there.

The 10mA is both...total and each stage. At least as far as the DC average current goes. The lower stages charge transfer is higher of course. 9 times or so for this 9 stage beast. But the average current in the diodes is pretty much the DC output,(more or less...OK more).

Variable frequency on the series resonant: The series resonant topology, operating below resonance, is a really efficient and low EMI topology. The only (simple) way to keep it such, is to frequency control it, below resonance. Once you start going near the resonant frequency, approaching it from the below resonant side, you will start to go into continuous conduction, chopping and slamming diodes off. It too works, but you have now don't have the true ZCS effects like you have operating far below resonance. The big negative operating below resonance is high rms currents, and low frequency operation and instabilities at light loads. Operating near, at, or above resonance has some good advantages on delivering high power, especially over a narrow range of output conditions, with some tough tradeoffs when you have a large load variations, that could range from 0 volts, full current, to full volts, no current.
As I have said in some of the other posts, there are countless ways to control the countless resonant topologies. Spellman has two patents to control series resonant topologies in ZCS mode, while maintaining high frequency. I spent years of my life getting them to work in practical applications.

Best Regards,
Cliff
Cliff S
 
Posts: 2
Joined: Thu Aug 12, 2010 9:27 am

Re: New HV stuff, questions

Postby Doug Coulter » Mon Sep 20, 2010 1:33 pm

Thanks Cliff.

If I can get 10 ma out, that's plenty for my plans....no problems other than making a nice current limit to a lot less than that. When I was building my own stacks, I did notice a lot more current in the diodes at the bottom being the limit to how many stages I could run (I went all the way up to 24 at one point -- you can laugh now -- but I learned better). I should be getting first smoke out of this today, using my super simple driver (I built a few up awhile back) on that Wallis transformer, just straight in which I predict will be non-ideal, but it will work well enough to make sure I understand all the other issues. It looks like about 50::1 too, so in theory I need 400v peak to peak to drive that....I will of course start closer to 24v on the H bridge...with a current limited bench supply for that.

So you're kind of doing what hams call "slope detection" with the variable frequency. Interesting indeed. I guess in this case the series L also acts like a high impedance for all those odd harmonics of the square wave (or the more complex ones in pwm). I will try that trick at some point, probably with that Spellman transformer, which seems like a really good thing, if not ideal for driving that particular stack. Once I make a good enough shunt inductor to raise the resonance some -- I don't like to run things down where I can hear the fundamental, and the caps get too big, I like that low stored energy thing a lot.

I guess I should look up your patents? It would seem the feedback loop gain would vary depending on where on the slope you were, which could make things a bit more interesting?
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: New HV stuff, first light

Postby Doug Coulter » Mon Sep 20, 2010 5:42 pm

I hooked up the repaired Wallis transformer to my simple driver circuit and set that for resonance of the transformer, which was 24 khz. Nothing fried, and the quiescent current was under 100 ma with 24v in (which is 48 pk pk on the xfrmr). So, what the heck, I wired in a single stack disk, and tried that. Quiescent current went up a good bit, but with retuning to a higher frequency came on back down even lower than before. I wound up around 34 khz for lowest draw with no DC load. With 30v input, I got 3.68kv output from a single stage, implying that I need 163v input to get full spec output -- suspiciously close to what you get if you just rectify the power line. At any rate, I pulled some nice DC arcs to a resistor without burning up anything, depending on the current limit in my bench supply (but with 4700 uf on the fet rail supply -- fets are tough!).

Since I don't really need 180kv, I pulled the stack apart into two pieces and will start off using 5 disks for 100kv max. Now I need to build a big R array for bleed and measuring the output volts, and do the current sense and limit stuff....and build a box for it all. I am guessing it would be wise to copy the Spellman PTV plan and have the controls and meters in a remote box, I'd not like to walk up close to this to turn a knob! More to come, but it's looking good as of now, even with a simple pure sq wave drive from that silly simple circuit above. Now I have to figure out whether I should just use the simple guy (for which I have a PCB layout) or something slicker based on an lm3524 and separate fet drivers...but much more complex re interlocking, which the simple circuit has built in so it never does stupid things that would mess up an H bridge.

Edit:
Here are some pix of the thing on one of my benches.
HV_Bench.jpg
Benchwork

Note scope probe near the right hand secondary winding to pick up the "blue" waveform via cap coupling.

H_Bridge.jpg
H bridges


This is a little platform I built to run a couple H bridge boards off a common large main power diode/cap setup mostly swiped out of a large commercial supply. I can enable either H bridge, the other one on top is tuned for my 35kv homebrew stack. I retuned the lower one for this range, it was originally set for 2khz to drive an old iron-starved X ray supply.

ScopeHV_inout.jpg
Scope traces


OK, you can see me being lazy here -- that scope can provide a perfect bitmap to put here (see the SD card? Well, my helper Dave has also coded that so if I hook up a cable, I can see it realtime here on the PC --in linux), but....I just took a pic. This was at 30 something volts in, about 5kv out of the one disk.
Top wave is half the H bridge output (see probe in previous pic) and the other is a probe near one of the secondary windings. As I go up in volts (to 125 so far) those edges get "faster" and my snubber starts to moan (only a 3w 1k R). Further, the distortion you see in the secondary wave becomes a bit more pronounced, a little peakier with the shoulders lower in voltage. Remember the differentiation in that second waveform due to the unique coupling method.

Also, the waveform rise and fall times on the drive get a lot faster, and hit the other phase during the dead time, bouncing off the FET diodes. Doesn't seem to hurt anything, and if experience is a clue, will largely go away under load. I tuned F for lowest no load loss about twice this level, and half max. Seems to be an OK place to be without fancy tricks.

I'll post another scope trace when I get another snubber made with more than a 3w 1k R for that part....you can see the difference in effective impedance in the load at higher oersteds (though the no load current merely scales linearly more or less), and it's kind of cool. In the trace above, it's hard to see the dead-time, but it's obvious when you cross that point and the load flips back on its own when let free. That usually has a lot less energy in it if there is a good load.

Edit:
Changed snubber to 1k 10w WW (wow, that seems pretty inductive even at these speeds...the wave changed a lot for same conditions) and good .01 good cap (I'll have to pull the other one off the board to find it, but I think it was a different value -- I just lifted the burnt R to add the new one in instead).
First trace is after some retuning (note F diff on scope) for min smoke, with about 55v rail (about 110 pk pk out).
55vrail.gif
55v rails, new snubber.

Second is at full snot for my bench supply about 125v.
125vrail.gif
Full output


I note much faster risetimes, and some scope probe and ground loop ringing off the catch diodes. Draws a little more now when tuned (it's a touchy adjustment with a too wide range trimpot), but it seems mostly in the snubber (based on what gets hot, I'm blowing about 1/4 to half the power into that snubber now, probably need to reduce both C and R). Seems to always be flipping during the deadtime with this snubber.That slight shelf before full turnon is a bit worrying, it's probably where the rest of the power is going. It's been there throughout. But the fets aren't getting hot at roughly 70 w input here (no load but the voltmeter on the output with HV probe).
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

Next

Return to High Voltage, High Power

Who is online

Users browsing this forum: No registered users and 1 guest

cron