New HV stuff, questions

Things at the limits.

Re: New HV stuff, questions

Postby Doug Coulter » Sun Sep 26, 2010 12:24 pm

JohnF,
I got that book you recommended, fairly cheap used, and it's one of the better ones on the shelf, I agree.
(now I can't find where you mentioned it, so I'll mention it again here)

Switching Power Supply Design
Abraham L. Pressman
McGraw-Hill
ISBN 0-07-052236-7

Found it for about 50 bucks ($273 new). I already knew most of the content, but it's still a big winner. One of those that are more expensive to not-own than to own.

I see where people are using fancy LC's in things to try and get around the inevitable low self resonance of a high turns ratio transformer, and then deal with the fast edges without burning switches up (like Carl just did on fusor.net), but I also find that usually there's enough leakage L in things that I can tune that very simple H bridge driver for least quiescent current and it's fine if the frequency is high enough for the use intended. I tend to cheat a little and just brute force the core itself, more volts/turn, and keep the resonance high that way. After all, I'm not CliffS who has to make money at this game, just a guy who needs onesies. This looks like a winner as is. That weird transformer pictured above has a ton of interesting peaks in the impedance spectra, so it doesn't seem to mind the rather cavalier treatment.

If not, I may "graduate" to adding some L across the xfrmr primary for starters. The FETs I'm using (irfp264 in this one) don't mind having the diodes "hit" before the other FET turns
on, it seems (that driver has a lot of dead time), and a minimal RC snubber keeps the heat down in the semis and actually is what eats most of the power at no load. As the load goes up, things actually cool off....so far.

I've now tested this with one stack disk up to 20kv, and it is eating about 60w waste power, a bunch in the snubber, the rest in the fets and xfrmr. Not so bad. I've found
this usually gets better with some load on things -- eats the energy stored in the xfrmr so there's less kicking back.

I'd love to figure out a way to use one of those power factor correctors to pre-regulate the rails to the H bridge to control power, but as they are boost-only, and I need
down to nil and up to about 150v, I guess I'll have to make a buck mode switcher for that. No big deal if I can put my hands on the right core material for the L for that.
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 27, 2010 1:43 am

Doug
Pages 186 -208 detail current fed current mode operation

enjoy the read
johnf
 
Posts: 433
Joined: Sun Aug 08, 2010 3:51 pm
Location: Wellington New Zealand

Re: New HV stuff, questions

Postby Doug Coulter » Tue Sep 28, 2010 8:08 am

I am reading that and am struck by something. So we put this switch on the front (shown in the book as your usual buck converter with all the parts, L and diode). Since I am or can contrive to be feeding and inductive load (see waveform switch at the beginning, not the end of the dead time), maybe I don't need another L at'all. Suppose I put a half-bridge there (with built in diodes) driven off a timer that looks at the low side outputs of the main H bridge driver? Just an OR gate plus variable pulse shortening. At any rate, that's where my thinking is taking me right now.

The full bridge chip gets its own 15v power -- and you can do pretty much anything with the fet rails without it noticing, and it is very good at never turning things on it shouldn't -- always keeps that top bootstrap cap up or it won't go at all. But of course the low side bridge drive is a signal that isn't flying through space -- it's ground reffed, so easy to use for something else too. So you simply let that main bridge driver run all the time and modulate its fet rails with shorter pulses. Crap, a 555 might do that job, with a half bridge driver chip! Though I can probably come up with something better. It's what's in my mind to try next. It should give me decently fast loop time for various regulation tasks.

Now, I like the current transformer idea vs a series R in the bottom of the DC for example. I gotta figure out how to pull that off, it's just a better way to get that info. Minus a constant, or a linear slope, what goes in is what's going out, current wise. Get the DC FB signal off the usual many-R HV probe built for this.
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 Doug Coulter » Sun Oct 03, 2010 3:48 pm

I've gone silent on this thread because I'm actually doing stuff on it right now. I've now tested the drive and one disk to full output (off a bench supply driving the H bridge) and it all works fine, nothing is getting too hot, etc. While I've done a heck of a lot of switchers, from insanely tiny to very large, I have never done an off-line one, this will be my first, and I'm trying to locate a nicely stiff isolation transformer so I can put the scope on things while I get any bugs worked out. I have designed and built the buck magnetics (I'm going to try it like the book at first anyway) and am working on how to drive that switch, thinking of using something kind of dumb like a 555 timer (using the control input to vary width), triggered off the H bridge drive signals to the bottom fets, OR'd, and having the buck frontend be a half H bridge, as I have some nice floating fet drivers for that and it solves some problems. The H bridge driver chip runs fine with no volts on the H bridge fet rails, so it can start up OK, else we'd have chicken/egg priority issues. I am still mulling how I am going to close the control loop -- I'll have to take some feedforward delay/phase measurements so I don't make an oscillator by accident.

One issue is that the H bridge driver has built in deadtime, and this topology wants overlap there, which makes sense in the text. I may have to add a diode (at least) to keep main FET rails from flying above the power input on turnoffs, or figure out some way to get closer to overlap on that bridge. The only obvious way involves slowing down the gate drives (big series Rs) but that sounds even worse to me than having to catch some flyback energy. The main HV transformer has a pretty complex impedance vs F plot -- I'll just have to wing that, hard to predict what will actually happen there. Too many poles and zeros to predict easily -- several of each, it seems.

I note that in the book, the guy doesn't like the fet intrinsic diodes much. From the prices he mentions, that's about when they did truly stink. The ones I'm planning to use aren't great (270ns reverse recovery in an IRFP-264) but won't be hit much -- in a bridge, the fet that has a conducting diode right after turnoff of the other fet is about to turn on itself which will turn the diode off again anyway, and from just one forward drop, not much energy involved. Yes, Rds is still low in reverse D-S polarity at least for low volts.

Of course the other consideration looms -- how in heck am I going to handle getting 180 or so kv from this thing to inside a vacuum tank, without putting the kill in kilowatt?
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 Doug Coulter » Tue Oct 12, 2010 7:57 pm

Ok, here we go, a little bit of progress. Other things have been much too hectic.

HVMockup.jpg
Mockup of most of the parts



This is at the point where I know it works. I've full-power tested it with one disc of the multiplier on the bench, but to fool with it much more in reasonable safety, it was just time to get the box going on, get the HV parts mechanically stable and so on. All this stuff is heavy, and just propping it up doesn't get it for the fiddly parts of the work, like driver optimizing. I scrounged the mains bulk supply and heatsink out of a battery charger (Thanks BillF), to use with one of my full bridge drivers, with perhaps a switcher in between to take care of providing the main bridge in current mode, so as to be able to vary the output easily. Other schemes are possible, but I've also made the buck inductor for the front end supply, so that's probably how it will go.

This stuff is surprisingly heavy -- the xfrmr or one disk weighs nearly 20 lbs. So it seemed a solid mount was in order before progressing too much further.
Since I'm shooting for the 100-120 kv range for this, we will probably use only 6 of our 9 stack disks at first, which means lower current at the bottom of the stack for a given output current, nice and conservative.

Not coincidentally, the resulting height will be just right to mate to the main fusor HV feedthrough.
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 » Wed Oct 13, 2010 12:52 am

Doug
When you start on the 100kV beast keep in mind some corona rings just outside the footprint of the stack.
These will stop any arc overs from sharp edges around your stack stages


remember no field within a faraday cage this is what the rings are for so the field variations are controlled by the rings

What you are suggesting sounds about right corona becomes a pain at over 20kV a stage
johnf
 
Posts: 433
Joined: Sun Aug 08, 2010 3:51 pm
Location: Wellington New Zealand

Re: New HV stuff, questions

Postby Doug Coulter » Wed Oct 13, 2010 12:37 pm

I will, John. The thing came with a big ring/hat for the top, but my experience is like yours. At 20kv losses start getting serious. At above 50kv or so, it seems any air gap can be breached via the streamer phenomenon -- unlimited length sparks more or less. So if you can force a situation where the air never sees more than about 20kv across any gap, you're way ahead.

I am wondering if I could bend some 1/2" or 3/4" annealed Cu pipe into rings that would mount at the interdisk gaps without wrinkling it too badly. Can't go much bigger than that or they'd be touching -- the disks aren't all that thick. They do have little round corona nuts on all the connecting bolts as is, but I suspect you're right (again) that they won't quite get it done. The good copper screen in there is to prevent emi coming out of the drivers messing up other stuff, and to make sure that arcs do have a place to go -- other than to me, but again, I saw a 100kv system toss a spark 6 feet out, then back to a couple inches from the beginning. Once was enough for that, scared the .... out of me. I'll get a pic of the stack up here in a minute or few.
Stack.jpg
Stack. One more disk will be added to this in first use.


Now, does anyone have any practical experience with moving voltages this big around and getting them through vacuum tank walls? I'm obviously going to need to do better than what I am doing now, which is borderline at 50-60kv.
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 » Wed Oct 13, 2010 1:48 pm

Doug
all of our systems use aluminium tube for the corona rings but I cannot see why copper wouldn't be acceptable although its a lot heavier.

Now getting those big voltages into the vac are done just the same way glass metal stacks with the corona rings and balance resistors on the outside giving well defined voltage gradients inside and out, again around 20kV max per stage.
Glass metal join is synthetic latex and the whole stack is assembled and cured in an oven with slow ramp up and down with temp to effect a good cure
johnf
 
Posts: 433
Joined: Sun Aug 08, 2010 3:51 pm
Location: Wellington New Zealand

Re: New HV stuff, questions

Postby Doug Coulter » Mon Oct 18, 2010 1:38 pm

Ok, here's a video of the thing with another load. I built up a new driver board for this one, using fets better suited to the speeds and feeds I expect, in this case lower voltage but larger current ones. The interesting thing is with some loads, the fast drain transition of the fet is enough to overwhelm the 25.5 ohm series resistor in the gate and turn the fet back on momentarily! Gotta fix that, and I'll embed another video of the fixed version when I do. I'd used the R to try and slow the fet down some and hide some of that deadtime, but obviously that is NOT the right answer here. For one thing, the evil current spikes are causing an on board bypass capacitor of 82 uf to get hot.



Sorry about the lousy focus, I'll dig up another camera and see if it has a macro mode.

Driver is an IRS2453, fets are IRFP 264's in a full H bridge. I am basically using the schematic in the chip data sheet, with a .002uf timing cap, and a 2k + 2k pot for the timing R, and 25.5 ohm gate R's. That last is the parameter I'm tuning on right now. The chip has what I'd call fairly extreme dead time, and it's not known at this point whether I can make the R's large enough to eliminate monkey motion there without having what we see in the movie -- the miller capacity of a fet can turn an "off" one back on when the drain waveform moves too fast, which will only be worse at the higher rails voltage in real life, or so I think. Hopefully I won't have to add some npn/pnp emitter followers (complimentary symmetry) to stiffen the outputs...I am using an RC snubber across the bridge outputs of 100 ohm and 1k pf, swiped from that battery charger supply -- it had about the same speeds and feeds...I may try increasing the C to the point the R starts getting hot and see if that helps any, but I think the load will kind of swamp that unless I get ridiculous there.

irs2453d.pdf
Bridge driver data sheet
(371.73 KiB) Downloaded 359 times

irfp264.pdf
Fet data sheet
(147.04 KiB) Downloaded 344 times


The Wallis transformer hits rated output for the stack disks with 125v input to the bridge rails, so I went to lower voltage fets that can hack the higher current here.
The "real" transformer isn't as tough a load (at no load) as the one I'm using on the bench is, but the one pictured is more representative of someone's homebrew transformer (because that's what it is). It lights up gas tubes nicely from a decent distance, particularly if you're holding one end of the gas tube.
HVDriver.jpg
Pic of the electronics and transformer test load


Here's the electronics involved. The perfboard is the "housekeeping supply" to run various things like the H bridge logic and fans, a simple building block that emits 15v, 5v, and unregulated 20v on one winding, and 15v and unregulated on the other -- the output float from the line, and each other for convenience. It turns out this driver circuit doesn't mind one bit about what order the supplies come up in -- so you can have it running, then apply main rails voltage, or the other way around, nice. I added the series C in the transformer primary to see if that had an effect, re walking into saturation as mentioned in that excellent book JohnF recommended (and now I recommend it too) but in this case no change at all occurred -- the driver is very symmetrical it seems, even in the funky modes shown in the movie.
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 chabekostc » Mon Oct 25, 2010 9:55 am

Doug, Cliff or?

I got a Wallis High-Volt Stack, schematic attached. I would like to use the two sense output wires that come out the bottom of the stack. This is a 120KV, 100 Ma air insulated stack, with a red rubber (of some ilk) potting of the transformer that feeds it. Any info on volts/Kv, Ma, etc is what I am looking for.

Any help will be greatly appreciated.

Charlie
Attachments
schematic of 8 stage CW Stack.pdf
(366.45 KiB) Downloaded 344 times
chabekostc
 
Posts: 6
Joined: Sun Oct 24, 2010 8:42 am

PreviousNext

Return to High Voltage, High Power

Who is online

Users browsing this forum: No registered users and 3 guests

cron