New H bridge driver for AC fusor drive

Things at the limits.

New H bridge driver for AC fusor drive

Postby Doug Coulter » Sat Dec 30, 2017 3:38 pm

Since initial tests of almost-RF drive for ion source and main grid were pretty successful and interesting... I thought I'd do some "product improvement" on the power supplies for these.
Here I'll just talk about the one for the main grid, the coil shown in "recirculation". Based on what I saw there, I need more speed - and more volts. Sadly, more volts also means needing more speed if any of the math is right...but we are just going to creep up on this as I can put the required tools together - this stuff ain't found on street corners.

To get to the hundreds kHz, and hopefully a good bit more voltage, I've acquired a couple of bigger bulk DC power supplies - 30v @ 10 amp each, planned to be in series, and computer controlled at some point, but for now, I just added reverse polarity protection diodes and a path to ground for the lower one in case of arcs.

To use this ~ 60v at ~10 amps - which is getting back to real power...I need a better H bridge and driver than the ones based on the IRS2453 simple circuit with it's own built-in high side drivers. Not only is it limited to around 90kHz - those fet drivers are much too wimpy for the IRFP-264 fets I want to use here.

So, I'm going with a good old LM 3524 to create push pull square wave drives at variable frequency up to almost 500khz and some horsey fet drivers and using transformer coupling for the high side, since these drivers can't do that without some help. I copped a pretty neat trick from an old Glassman power supply I parted out, which I'll have to draw and attach - it uses a pnp darlington transistor for turnoff of the high side FET which can get it done a lot faster than the current that will come through the transformer - the "power supply" in this case is the gate capacity itself, charged up when the thing is on (and the miller capacity too).
As you can see, it's kind of a mess on my bench here - I just fixed up one half of the drivers while I worked on drive waveform quality - I did do some short tests with full volts on the fets to see the miller capacity issues - they're there, but not too bad, this drive scheme is a bulldozer.
20171230-1602-RF-4.jpg
The kludge for testing

I tried two transformer types - the toroidal one is parted out of an old Glassman power supply, and is ~ 1.5mh windings. The other one is ~ 2mh windings, a little bit thinner wire. I went with the former, for the slightly fatter wire and less need of insulation anyway in this voltage range - the other ones will be used with *real* high voltage fets later. It wasn't a huge difference, but at an amp or more peak current, you could see the series R.

DS0019.PNG
The waveforms it produces...could be worse.

The blue and green traces are the outputs from the LM3524 chip (in this case, the emitters as we're using non inverting gate drivers).
Yellow and violet are the low side fet gate and the high side one respectively. There's a little attenuation and slowing in the high side one, but it'll do. Part of the attenuation is those diodes forward drop. The transistor shown below *dramatically* cuts the turn off time, even more so when there's actual supply for the fet D<>S and there's miller capacity to deal with.

Sadly, if I want to go a lot faster, it looks like I'm in the "make your own drive source with deadtime to drive the driver" land. Unless someone knows of an IC a lot faster than an LM3524. This one's at its rated max, or close.

The final drive for the high side:
FetDrive.jpg
Cute driver circuit for fast turnoff


I have a parallel project going with an ESP8266 and an MCP4631 dual i2c potentiometer to control this remotely. More on that when there's more. For now, knobs.

Some data sheets:
LM3524D.pdf
The old standby used by Spellman and almost everyone else.
(1.14 MiB) Downloaded 14 times

ucc37322.pdf
9 amp gate driver, noninverting. The inverting flavor would work with minor circuit changes.
(1.44 MiB) Downloaded 10 times

irfp264.pdf
A honkin big NFET
(147.04 KiB) Downloaded 19 times

MPSA75-D.PDF
Darlington bipolar family, I used an MPSA64
(136.32 KiB) Downloaded 12 times
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: 3056
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: New H bridge driver for AC fusor drive

Postby Doug Coulter » Thu Jan 11, 2018 9:28 pm

A tiny bit of progress in between fighting with the elements and infrastructure - it's been a rough week or two here with weather the coldest its ever been the whole time I've lived in the mountains. Our record low set some years ago was 6F with 10 mph wind, one night. We just had two solid weeks worse than that. I still can't get water to flow through the buried pipe from the rain collector, whose 100 gallon ice cube has already melted in a few days above freezing. Gheesh, I'm not gonna bathe in an ice cube.

Anyway,,,
20180111-2218-Bridge-1.jpg
The board on the left is nominally done


Got a few holes drilled, tapped, custom heatsink insulators punched and so on. It occurs to me that with the right ttl inputs, and an analog one for modulation, along with a decent tuned circuit, I just made a multi-kilowatt AM band transmitter here....
The stuff on the kludge boards - still to be done and mounted up. I'm going to have a switch between an ESP8266 and digital pot to do duty cycle and frequency, and just some plain old pots for when I'm up close and personal with it.

I found out the hard way the other week that it's a good idea to have a local fast-kill switch. When that ion grid stuff started arcing, I couldn't find the window on the gui and click off fast enough, and it was dangerous to go near where I could pull the plug at the moment....no more of 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: 3056
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: New H bridge driver for AC fusor drive

Postby Doug Coulter » Mon Jan 15, 2018 1:24 pm

Step by step, inch by inch...
It's looking pretty good, actually. I have the manual option ready for real testing now. After some futzing around, I decided to stay with the minimum recommended values for Rt and Ct which are real close to the minimums that work right (LM3524 - 1000pf and 1k for reference). This limits my top frequency to 284.5 kHz - not real bad. Pushing it, I got to around 350, but had some issues with shoot through current and weird ringing probably as a result of the shoot through (though nothing got too hot and if needed, I know I can go there again).

I put some cheezy old heatsinks on the front of the panel and did a little testing. At around 8 amps at around 7.5volts input to this little coil, it DESTROYED that CCFL lamp and let the air in (and the Hg out, no doubt, OMG micrograms are gonna kill us all!) Nothing but the ccfl got warm (it shattered).
20180115-1411-Bridge-1.jpg
Test setup


It looks like this from behind. The empty board is room for the ESP 8266 remote if I get to the point of needing to be elsewhere when I run. Almost the same code and digital pot will do for those 30v 10 amp power supplies I'm using for this - they're different than the 5 amp ones I did before, no uP in them at all (probably a good thing in this case) and appear to be more or less redesigned PC power supplies.
20180115-1412-Bridge-2.jpg
Lousy photography but I didn't disturb the wiring either.


This makes very pretty HF arcs no sweat...with the small stepup coil. It shows BTW several interesting self-resonances, some of which look like series resonances where the step up is far greater than the turns ratio would indicate. This may be an interesting take on that one:
self-res.pdf
Someone's odd theory that seems relevant to the observations here.
(1.31 MiB) Downloaded 46 times


So, about ready for a real smoke test. I DID try to make this easy to repair if in fact I get smoke, but we'll use the current limit feature and hope for the best.
That weird group of resonances, some parallel and some series...is pretty interesting on its own. Now, with a fusor load, which isn't resistive (at least we hope not), who knows? This is not the easiest thing to measure directly. For example the HV probes I have are grossly influenced in their AC divide ratio by a person's hand a couple feet away (or anything else with a signal or ground on it). Not helpful if you want to KNOW what's going on.

I'll try and get a full schematic drawn up and scanned in at some point. I'm still a luddite as regards not having any schematic capture software here. I did buy a USB drawing tablet and the verdict is - use paper and a scanner, it's faster, looks better, easier, and you see what you drew at the point of the drawing tool, not on a screen somewhere else...but then I could never draw those things on the backs of matchbook covers 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: 3056
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: New H bridge driver for AC fusor drive

Postby Doug Coulter » Tue Jan 16, 2018 8:00 pm

Well, I got some smoke, but it was only my flesh...I tested this HARD with a Spellman stepup transformer worth 3kw ratings...and got my hand into the wrong place - lucky I had current limits set for around 300w. The only thing that fried was me. The word for this is: OW.

3rd degree burns even though I yanked it outa there really fast, you bet. Oh well, I grow back, but FETs don't so much. (I'll maybe get a pic once it stops stinking and swells up impressively).
So, I tried KiCAD to get y'all a schiz, but it's so broken right now that the component editor won't let me make a box with 4 pins (the 12v switcher that this runs on internally) and I'm not into joining half a dozen mailing lists to report obscure bugs (eg giving the world my email, it's bad enough already), so I'm trying gEDA, which looks more to my liking, even though the java side-issue programs for converting other components to it's format are also utterly broken. I'll try and get a schiz...at least this one lets me just draw what I want - with only 50 times the effort a drafting board would take (it does, however, save space).
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: 3056
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: New H bridge driver for AC fusor drive

Postby Bob Reite » Wed Jan 17, 2018 10:20 pm

RF burns are NOT Fun! Last time was on a 1600 watt AM rig. You were supposed to be able to change the J plugs hot... well it reaches out and "grabs" you. So after that. Power it down, change the J plug and power it back up.
The more reactive the materials, the more spectacular the failures.
The testing isn't over until the prototype is destroyed.
User avatar
Bob Reite
 
Posts: 48
Joined: Wed Nov 11, 2015 1:02 pm

Re: New H bridge driver for AC fusor drive

Postby Doug Coulter » Thu Jan 18, 2018 9:28 am

Tell it, brother! I'll live, but gheesh. It's too cold in my shop right now to test on the real stuff...so I got sloppy and did what I could on my cramped little tinker bench. Live and learn (if I live!).

It's looking pretty good and operates really smoothly re frequency and duty cycle, with an odd thing that happens with a stiff load - that huge spellman transformer for example. At some frequencies and duty cycles, the scope goes nuts (won't trigger right as there's a lot of noise), the transformer makes a scritching kinda sound (this is all theoreticly pretty supersonic) but...nothing odd shows up in current draw or on the voltage probe on the 1 stage doubler I'm using to make it into DC to measure - having found my HV probes really STINK with HF AC.
I don't *think* I have any kind of ground loop - the low side fet ground go right to the separate power supply for the main bridge, and only get a thin wire back to the logic ground for drive reference purposes, and I don't know if I bother to fix it since otherwise, it ain't broke. But interesting. The LM3524 is not a shaky-flaky chip, it's one I've been using for decades for all kinds of things, it just works. I don't *think* it's picking up noise that'd mess up that oscillator. You'll also see this chip or it's close brother that doesn't have open collector outputs, in most Spellman and Glassman supplies as the main signal generator. It's one of those things that the whole industry decided was the going to be the one for switchers.

In that paper linked a couple posts above, the guy has an interesting theory that can't be all correct (if it was, certain funny things would happen with base-loaded antennas that we don't see happening if I understand it right)...but I'm seeing some of the same stuff he predicts and measures here. I'm going to have to put a discharge tube near one of my big coils and see if I can see the peaks and nodes like he did, that looks like fun. But of course, I have to be more careful at these power levels, just being a little too near shattered a CCFL when one of those fuzzy RF arcs hit it.

What I AM seeing, and he predicts, is a family, almost a comb (they are't evenly spaced!!!!!), of various self resonances. It's not "this coil is a capacitive reactance above self-resonance" like in the ARRL handbooks for decades, not even close. And - I don't think he said this - some of them look like *series* resonances, in that the impedance goes to near-zero, and all of a sudden you get a vastly bigger step up ratio than the turns ratio would indicate. At these (fairly narrow) frequencies, the output goes to the sky (or something arcs because of that) - for the same input voltage.

This is going to make it hard to measure exactly what I'm putting on the fusor when I get over there to do that - it'll be a little embarrassing if it works and I can't get the numbers - that even I can believe in. I can't use a feed-forward guess from the turns ratio due to this. I don't think there is such a thing as a rectifier that will handle 50 or so KV (100 PIV) at 280+khz...

The issue is obvious on reflection. We have some huge series resistor chain in the probe, followed by a smaller resistor to ground. Say a gig ohm series, a meg to ground for a 1000:1 probe.
Now, a scope probe is set up so that the ratio of the C across the big R and the little R also produce the same divide ratio and it works broadband.
Now...for one thing, there are no available caps of the right pF you'd hang across the 1 gig - that wouldn't have corona and arcing. But even that's not it. So to get rid of the proximity of the rest of the world effect (this is literally hand-waving - waving to it changes the output 2::1 or more), you have to shield that gig impedance tip - but if you get close, things arc to the shield....I can't see how you "win" this one. Any ideas?

I've been using a fairly good scope probe some inches away to get a qualitative look at the output - it's just a fractional pF capacitive coupled thing since the feedthrough is a pretty big chunk of stuff - about 100pF capacity to the surroundings if I belief the MF meter. This is obviously differentiated, suffers from the same hand-wave, and gives me no "real" numbers, just a look at the waveform and timing. That won't do for the eventual paper that will have to be written on this when it works.... 8-) If it works. So far, it's encouraging re getting solidly out of the noise at far less drive than ever before (around half what the old threshold was) - scaling up is gonna be exciting I bet. With DC anyway, it went up a heck of a lot faster than linear - about the 3rd power of the increase over the threshold...so, this should be on the order of 8 times better at the same 50kv if I get there and it all adds up the same. But it could be better or worse, which of course is why we test.
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: 3056
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: New H bridge driver for AC fusor drive

Postby Doug Coulter » Thu Jan 18, 2018 9:42 am

That's one huge hunka chunka ferrite transformer, from a 3kw/60kv supply:
20180118-1049-Bridge-1.jpg
Test rig

You might recognize some of this from earlier posts on fusor.org - this is from a DXR-3000 [sic?] supply we've scored one complete HV part and a few of the spares, and the doubler is one Brian un-potted from something which works pretty well as an around-the-shop rectifier-filter to around 10kv for testing stuff.
At those series resonances, I was getting 3kv out of this with 4 volts (count them) into the H bridge...that's WAY too much for the turns ratio. Off peak, it took more like 30v into the bridge to almost get to 3kv. As I said above, this is a little weird. It doesn't disappear at least under mild load (like a huge 500w 1 meg resistor we have) so it's not just a little spike getting the caps all charged up.
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: 3056
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: New H bridge driver for AC fusor drive

Postby Paul Fontana » Thu Jan 18, 2018 5:42 pm

Sounds like you're broadcasting, all right! One thought - perhaps instead of measuring voltage, measure current. This could be done with a loop around your RF solenoid, or with a Rogowski coil around the lead. This should give at least an estimate/upper bound on grid voltage, since high-side IV should approximately equal input IV (less radiated/broadcast power and heating power in the transformer).

BTW, there are fast HV rectifiers, like http://hvstuff.com/3a-50kv-100ns-high-voltage-diode-hv-rectifier-high-frequency, even in Si, which makes me think there must be HV/RF snubbers you could use like a zener to regulate the output voltage. The folks who developed RADAR must have made some kind of tubes to do this job.

I hope you got your hand taken care of properly! Not worth losing any limbs over. Be safe!

-- pwf
Paul Fontana
 
Posts: 5
Joined: Tue Feb 23, 2016 1:17 am

Re: New H bridge driver for AC fusor drive

Postby Doug Coulter » Fri Jan 19, 2018 2:14 pm

Yup, this is definitely a spark-gap sort of transmitter when we're in the corona range, and it often is in testing. "In real life" it'll be inside a copper screen faraday cage...

Yes, we'll be measuring current fer sure, as at these speeds and feeds, on the ground side, it's pretty easy - don't even need a frequency-dependent coil, a simple small value resistor will do for that.
When we were blowing and smoking hardcore - we almost never went over 20ma, and usually it was more like 10. Here, if I get the improvement I'm looking for, I might want to hold it to far less, at least when I'm in the room with it...so it might be a 1 or 10k resistor for current sensing.

Speaking of "in real life" - with the real stuff I plan to start with, at one of those exceedingly weird low impedance points (looks like a series resonance, but with no series cap I see) - I got the following results - I'm using an old TV HV rectifier the shop I worked in as a teen used to replace the 1b3 types of tubes here, and a 750pf filter cap. A three turn primary was used, #16 wire, which didn't get warm, but the stranded speaker wire (supposedly same gage) leading up to it did...
20180119-1509-Bridge-1.jpg
If I use enough light to get depth of focus, you can't read the numbers anyway...life.

6.45kv peak (half wave rectified) with 4.5v into the H bridge at 9.5 amps or so.

In other words, that'd duplicate what I was running the other day, easily, but at 91 khz.

I feel like I need to do a bit more mechanically here as well as insulating, this is pretty coarse work (but it was sure nice to have a milling machine to just mill out a pocket for the lower core leg, eh?).
This might not be where we run, of course, and for reference, with ~ 5v in at 150 ma - quite a difference in current - I saw 3.3 kv out, at a different frequency.
This arrangement is not a place I want to futz with putting my hand just anywhere in either - I'll prototype harder soon.
I note that it must be some kind of inductive feedback that makes that scritching noise as I still get it with this rig, something must be getting coupled back into the oscillator. It doesn't seem to affect much else, but it's distracting and I like to keep things so my normal human senses can detect anything *accidental* odd going on - most of our discovery has resulted from that.

I can of course use more turns on the primary, and still be fine even outside these oddball series resonances (if that's what they are - they act like that anyway). I have two of these 30v 10 amp supplies...so up to 60v/10amp in.

For grins, with this setup and my IR thermometer, I want to find out where that nearly 50 watts went. The speaker wire got warm - but not 50 watts worth (more like < 1/10th that). Nothing else got warm right off, other than the power supply turning its fan on. Energy has to go somewhere, and intellectual curiosity wants to know where it went - it might be a clue into this otherwise mystifying behavior.
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: 3056
Joined: Wed Jul 14, 2010 7:05 pm
Location: Floyd county, VA, USA

Re: New H bridge driver for AC fusor drive

Postby Doug Coulter » Fri Jan 19, 2018 2:21 pm

AFAIK there is no such thing as a snubber at the high KV short of a spark gap kind of thing. The radar guys did use spark gap tubes (and I have some) to generate the pulse to the magnetron or klystron by shorting a capacitor through a pulse forming network (the TL;DR version - it's more complicated than that, and I'm studying it for when we know what kind of pulse we really want, as that's about the cheapest way to get there from here).
Thing about spark gaps and HF HV - everything changes with the phase of the moon. A small change in frequency changes the spark over voltage (which once an arc starts, is a short to ground). There is no such thing as a zener at these levels - and in fact, the Si diodes are really a zillion little diodes in series - over a few kv, you can't make a one junction diode. You can verify this by (attempting) measuring the forward drop of a HV one. 30-200v is common - at ~ .6v/diode at low currents.

So the big boys just catch some stuff on fire while they learn, and figure out ways to limit stuff at the manageable voltage end of things....

There is still some art to this, it's not all hammered down. Maybe, part of the attraction!

Edit - I didn't realize till I looked at your link that I already have a bunch of those same rectifiers, or at least identical looking ones from China that cost the same. I bought them and put 4 of them in that full wave voltage doubled 60hz 100kv supply I built but never put into service, as the safety issues with 1 uF maxwell capacitors and the 8kw *continuous* rated stepup transformer were kind of daunting...Maybe we'll take it to HEAS and find a "greater fool" to buy the thing - we had a lot of bucks in it. Big tank, all in oil. If it ever arcs across the 2 seriesed 50k 200w ballast resistors, all heck is going to break loose with plasma at both ends of the cable and some gallons of atomized oil sprayed into the lab from the stored energy in 3 of those caps...it might not be a survivable oops.
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: 3056
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