4-125a amp for fusor ion trap

[Doug Coulter]

I wound up with a 300 ohm 2w metal film R because I had them in stock - I quit collecting 1/2 watt ones when I stopped getting old TVs out of dumpsters...sigh. The original builder had done such a sweet job bussing the two tube sockets together with #10 wire and bypassing the bypass caps I didn't want to take that apart, I'm kind of using whatever is there when I can if it looked decent.

Here's the schiz for the driver stage. Those resistors were "there" and I left them since they seem to "do no evil". I added all those bypass caps. This is all under-chassis, to which I've added a 33 cfm fan, so that the air blows over the big filament transformers first, then exits past the 6146, where an above chassis fan takes over the cools the big tubes. I hope, anyway. Its big enough to move stuff around on my bench if placed so it blows downwards while I work on the chassis.

Hopefully close to final design

Sorry this is taking so long, I'm trying to be sure things are nearly nailed down first, and other things have been eating into my fun time. This seems pretty resilient, handles no-drive gracefully (around 30 ma plate current in that case), and well, the transformers I built do seem to do the right thing for the frequency range of interest. Nothing appears unstable at this point. The extra resistors seem to make it all a little bit "squishy" so it has kind of a sloppy form of ALC if you think of it that way. I'm trying to make it as idiot (me) resistant as I can here....and flipping the wrong switch or turning some control out of range is not too uncommon in the lab.

BTW, Bob, if you love NOS you should get in a truck and drive here - I have 2 buildings full of goodies, most of which I want to shed. A tube collection started by one of my grandfathers, added to by me, my Dad, and my brother - even Tek CRTs or those weird ones that had a connection in the center of the screen for weird old radars. I did sell off the ones audiophools will pay huge bucks for, but if you ever needed stuff from pre wwII and on up - old fat pin tubes, loctals, octals, minies...or a box of matched 6au6a with tektronix serial numbers, and so on...
Around 6000 or so just tubes. UTC linear standard transformers....lots of cool old stuff, and some worthless truly boat anchors. Mostly free to whoever will bother to carry them off. What I don't have is the larger power tubes though there are a few oldie but goodies.

[Bob Reite]

The two watt resistor may still be "too big" to act as a screen fuse, even with two tubes. Just be careful not to lose the plate supply without the screen supply turning off.

[Doug Coulter]

Very much a case of crossed fingers, yup. I based it on this handy webpage to avoid pencil-os or calculatoros. Shows 2w at around 40ma/grid, with more voltage drop than you'd get with 240 ohm per tube (which would effectively be 120 for two).

Web calculator for this junk

Yet another online calculator: http://www.sengpielaudio.com/calculator-ohm.htm
I might make a linux program to do this locally, it's kind of cool. Type in any two knowns and it figures out the rest.
It's in some nasty javascript, but I think I can figure out ohms law and P=IE in perl and GTK.

More scary for me is I don't know the burnup characteristics of any of the pieces involved here. In the really old days I would deliberately burn up half watt resistors of 2.2 k with 120v AC (as the guy who maintained all the AV gear in my school, it was a way to guarantee callbacks that got me out of class, shhh). They'd go in a few minutes, but that's at ~6 watts...so even that is a little scary. I doubt those ohmite resistors I got way back then were all that great compared to the slicker allen-bradley ones we used for good stuff.
And I know nothing about how metal oxide film R's go up in smoke - time vs power and all that.

Partial mitigation is the squishiness - that power supply isn't going to put out a lot more than needed, and the drop in the screen voltage shouldn't hurt.

At the bias I'm temporarily putting on the output tubes (-68v for now)..the total screen current is < 10 ma with no plate voltage (yeah, I need to draw that circuit and put it up too).

It's a DC feed into a 5k bypassed R to ground on the lower end of the grid drive xfrmr output. It's such that grid current will increase not only the output tube bias as it'll drive up the supply itself ( a voltage multiplier) but also on the 6146 which uses the same source divided down. This setup may not allow full power, I haven't tested that yet, getting ready to do that soon. I expect screen current will initially go up with drive and plate volts to pull more space charge off the filament, but that's why we test.

I have a relay in series with the on-chassis 330 v supply (it can unground the center tap of the winding) and it has both a front panel switch, and a wire to the octal socket in the back. When I get the big HV supply made, it'll open that interlock when the big HV input is off....and I may need that as I may need to reduce the "protective" bias to get to power.

I just finished getting two MOT's to match to simulate a center tapped plate transformer. Not ideal, but...you work with what you have. Took 4 extra turns on the primary of one of them, now both will get me to 2300vdc under light load at 105v AC input before the core losses go up too badly. I also have a fat 28v center-tapped transformer I plan to use to buck the line down to 120-14 or 120-28, with a switch on that box's front panel. Right now all that is just a pile of not completely tested parts on the bench....I'm looking for a chassis to re purpose in my junkpile. I have some old 2500v, 20uf caps I got from Richard Hull, I think they were originally used for very large xenon tubes at airports...they make quite a bang when shorted out at full charge...One should be too much for this.

I do have meters on all this - they were in the original attempt at a 6 meter thing, and no way I'd remove them. They're recessed from the front panel as they float up at high voltages - it looks cool.

The plate supply will be where I can hit switches easily. The RF amp itself will be on the other side of the setup, so you'd have to reach over the ?? kv DC part to get to it...I plan to drive the thing from that old GR sig gen, which has way more output than needed, or the Feeltech generator, which does 25v peak to peak into 50 ohms (or nearly). With the transformer input I dodge the fact that the Chinese thingie puts out 15v DC at full snot while it boots internally...since the other amp was DC coupled that was a disaster in the making. The Chinese thing is USB programmable, clunky but it might come in handy.

[Doug Coulter]

It's always slower than you wish it was. Are we there yet? BillF found us a power supply chassis from I'm guessing a hamfest. It appears to be a power conditioning unit (part of one) for gasoline generators - which is why the variac is there and one of those meters is for frequency. I may swap in a voltmeter there for output volts.

Mechanical mockup

I've matched two 800w MOTs for effective turns ratio and will use them to simulate a center-tapped plate transformer with the CT grounded. Not idea, but I don't think I'll need anything like their full ratings. I've driven the magnetic shunts out. They "like" about 105v rms, past that they're in saturation at least at no load, and they make right around 2500v no load there. I figure it'll drop down somewhat with loading...there will be enough meters to see what's going on.

That big cap is an oil cap once used, I'm told, for huge xenon strobe lamps at airport runways. 30 uF at 2500v. A little scary. The perf board there will hold two 5 diode strings of 1.5kv 6 amp TV flyback diodes in TO-220 cases (shown in a tube balanced on the meters - I bought several tubes of things like this when it became evident there would be no more CRTs or support parts for them).

Not shown is a thumb sized little 12v switcher to run that fan, which will use an existing hole in the case, and a relay that will actuate the interlock on the amp chassis so that when this is off, that supply is also off other than filaments and fans.

I didn't really plan to need a variac, but since it's there, it should make testing at reduced power levels a lot easier, and maybe running that way. My guess is that this will all be more than enough, RF wise, and as I learn more, the most I might have to do is swap the various impedance matching and isolation transformers in the amplifier itself. Gee, maybe I should have built that for plug in coils like the old days!

[Bob Reite]

Should have sent you that 2 KW continuous duty plate transformer out of a gates BC-1G, but the freight would be more than the transformer is worth. Hope the MOT can deliver enough current.

[Doug Coulter]

Heck, I might have road-tripped for that! But I think 2 (count them!) MOTs rated 800w each will be enough...as is, you have to lift that chassis very carefully or the bottom plate will buckle under their weight - so if you go by pounds...
Glad I have a top for this that stiffens the box quite a lot. And glad I won't be moving it much.

Now of course, those ratings were with serious cooling, a full wave doubler, and here, each will think it's driving a half-wave thing, but the ratings were 400 ma apiece - after doubling, when in the oven.
You know, the old fingers-crossed routine. I do have a fan blowing on them - probably too small. We'll keep the IR thermometer on them at first.

I figure the copper losses at 4x normal due to I2R, but two transformers to spread it around, and core loss is really the big one with these, at least at light loads...It will be interesting to look at the efficiency and total heating vs load. With just one, and a half wave test supply and resistor load, the power loss in the transformer goes *down* with load, at 105vac input (where they start really saturating with no load - eating about 50w of input power each). Total power of course went up, but most of the total went to the load R instead (a 225w resistor).

I'm hoping I don't need all the power this can make (and probably it won't draw 4-125a tube ratings without some more tweaking to my combo grid-leak + DC bias etc).
I was really close to "enough" with the other amplifier - but it was getting to the "pushing it hard" range as configured - and in this wacky setup, well, you want a little extra before smoke.

With some other new parts - a much better (I hope) output matching transformer (the original bit the dust by a nasty arc through the isolating quartz tube)...I'll just have to make some runs once this is all wired and fired. I did get a relay onboard the power supply to operate the interlock on the RF amp that kills all the 330v stuff - but not the bias supply.

Well, I hope I was close...the function of "how fast do the ions move vs the field I put on" is a pretty nasty slope at any pressure I can make it stay lit at - I don't want to go to extra ion source and differential pumping! At 50kv DC it's almost 1/10th - ions only going around 5 keV. Seems better with RF but I made smoke before getting a definitive transit time measurement, all I know is the Q was going straight up - faster than exponential - going off the plot to the left, with dinky input voltages - perhaps 25kv DC and 8kv RF. (The output pi network steps up quite a bit...). This lashup should be able to double those numbers, and at least with DC, it goes up at least exponentially with drive level.
Here it was still a curve-up even after logging the Q axis, with the AC added...something fun is definitely going on there.

This time I have an isolating tube made out of thick wall teflon for the output xfrmr - 1.5" ID but 2.3" OD; > 3/8" walls. I don't think anything I have is going to blow through that!
The primary is 60t wrapped around 4 .5" by 6" type 61 ferrite rods, with 4 layers of glass tape on them. That goes inside the pipe. Secondary is to be determined, but
it looks like I'll be trying to hit a few hundred ohms Z at resonance, depending on a lot of things - and open circuit if it goes out. In bench tests at signal generator power levels it seemed like it coupled OK despite the big separation, and would give me decent efficiency.

[Roberto Ferrari]

Doug

Please comment how do you get out the magnetic shunt from MW transformers.
Here are almost Chinese and the shunt is integral part of the lamination.

[Doug Coulter]

Maybe I should make a video, it's not like I'm short of MOTs. Just watch the dumpsters, the crop is good every season! I probably have 10 spares or more. Nice source of self-tapping screws, relays and so on, as a bonus. I can't resist...

They seem like part of the lams, but they are not in any xfrmr I've yet encountered - and nearly all have Chinese heritage.
They are for certain, glued in so tight...and often interference fit with say, the filament winding or primary or secondary windings - or all of the above. That varnish they use, it's tough stuff.

I make a custom punch that just fits this transformer (they're all a little different) because if your punch spreads the shunt lams, you're stuck for good. I use some rectangular steel bar ground to size or milled (usually just ground as I have lots of scraps and look for one real close to start with).

I hold the transformer in a big vise, usually not quite flat as they often have welded on big bases, but so the base is on top of one jaw, and the top of the xfrmr is on the other - windings always in midair, not touching anything. A good whack or two with a 2 lb persuader in the hard ones, or little 10oz short handle tappie for the easy ones gets things started. As soon as something moves, I drip in some Kroil, followed by sewing machine or gun grade oil, then leave it for an hour. Some grease would probably work better if you could dissolve it in something thin, I just use what's on the bench. If it doesn't swell the lacquer, you're good. Whatever they use, it's impervious to any solvent I'd have in the air with me. Haven't tried the really toxic stuff.

At that point, they usually just tap out - it takes a lot of taps - maybe .020" at a time - but they come. You need to be sure your punch fits and is long enough to get them all the way out without the punch doing damage to the other windings.

More often than not I have to strip out the filament winding first, by the "whatever it takes" method. Usually cut it and pull out half-turns at a time till it's easy (usually only 3 or so turns total).
That gives a little more room for the shunt to come out before killing something else with it or your punch.

I've never seen a shunt that was actually one with the rest, they're all paper wrapped and glued in (like the very devil, but "just" glue). if you've got one like that, you're out of luck - or you have a valuable collector's item. But remember, they have to be able to make these things, and slipping the already-wound bobbins over the core is key to cheapness, so I doubt anyone did that much, if ever. And that thinking is key - back in the day when I made my money by fixing consumer electronics, (1960s) - you know they had to have a way to make it....just work it out. Sometimes that takes a lot of close observation, and maybe wrecking one or more. Usually you can get it with one.

Here's a couple pix. the first one is just after the first smoke test - 2100vdc at 95v on the primaries. Had to quit because that dummy load R on the left there was getting toasty - it would have lived but the thing had proved its point. FWIW, I can't turn it on with the variac turned up. The peak current trips that weird switch+builtin breaker the original builder of this put in. Never seen one quite like that - it was some military special, but responds really fast (it's rated 12 amps!). No problem if I turn it on, then turn up the variac.

Smoke testing. Looks like I hit 2200+v at top snot, more tests to come (where is the power going now? I know the 50k load gets hot quick!) 2100 at 95v ac rms input

Light showing through where a shunt used to be. The transormer on the right needed 3 more turns on the primary to have the same effective turns ratio under load.

These transformers were Korean, for what it's worth. I didn't pick them for country, I picked them as a nearly matching pair. Both pretty old (microwave ovens tend to be old when someone throws them away).

[Roberto Ferrari]

Thanks Doug

Clever approach and clear explanation.
Waiting for next steps.
Best,

[Doug Coulter]

More to come of course, as I've not really beat on this enough, and haven't buttoned it up and arranged for connection to the amp, but here it is cooking the dummy load = a 50k ohm big power resistor. The light and IR images are offset in this inexpensive new IR cam, I'll live. (you can have pure one or the other as well).

Nothing gets hot but the load...At least not quickly.

IMG12.png (35.18 KiB) Viewed 4222 times

In about 10 minutes one of the transformers had gained around 6 C, the core in this case - all windings, the diodes and capacitor stayed cool. \
This was at 2.2 kv output, roughly. Not sure I trust that HV probe to be on the button. At one point I had 3 different meters hooked up at the same time, and all diverged differently near the desired operating point of around 2.2kv (well, I think that's where this will be happiest, can't speak to the actual amplifier yet).

It now has an almost accurate DC voltmeter on the front as well, scavenged from an old CD ionization chamber rad detector. Was kind of a surprise those are some really decent meters - 50 uA full scale. If I'd had a cheezier one to stuff in there (like a 1 ma FS meter) I'd have done that instead. This one took 50 megs of resistors to read right with 2500v full scale. I put them inside a bit of quartz tubing, at those impedances, a stray breeze is going to mess things up a little.