4-125a amp for fusor ion trap

High frequency, antennas

4-125a amp for fusor ion trap

Postby Doug Coulter » Tue Feb 19, 2019 12:50 pm

Hopefully, I'll get the design details in here too (at least for my own good!).
Bill found us an old homebrew ham project sometime back, most of a CW transmitter for I think 6 meters. It had two "new" 4-125a output tubes, an under-chassis 6146 driver, and a real primitive pcb in a box with a 6cl6 that used a home etched xtal in some kind of multiplier to get to frequency. A 5u4 and a couple of 6b4's rounded it out. It clearly never worked, the soldering was amazingly bad, and when I got it there was no tank coil, but a 75k resistor in series with the plates (maybe it was arcing inside those tubes, it happens with that style).

I'm working to redesign it and bring it back to usefulness it never had before. The 5u4 went and became a couple of TV flyback damper diodes (it had an L input filter with an 8uf oil cap!). Now it has a 500v 82uf panasonic cap a fraction of the physical size, and so forth. The 6b4s (2a3 type tubes more or less) are gone and a VR 150 and 105 now regulate screen volts for the 6146. I tore out the PCB and put in a perf board that doubles and triples some available AC into 70 and 100v (roughly) to bias things.

The 6146 stage now will make a few hundred volts pk-pk into the 4-125 grids and drive them nicely positive despite starting from -95v bias supply.
Input and output of the 6146 are now "modern" toroid transformers I wound that work well at the range of interest (1.5-3 mhz, but pushing it at the top end, too much Q). Based on other measurements I haven't posted yet, I think I want to be down around 1.7 mhz for now.

I did do a semi successful ion trap experiment, which seems to be showing me that like with DC, things are mushier in these conditions, and the ions see a net field a lot less than the applied one for both AC and DC.
Which is no surprise and I may wind up going lower yet (And higher volts, but it's dodgy now with 8kv AC + 25kv DC on the thing - RF in the shack). My other amp was kinda pushing it at 150w and bad matching, this should do 500 or so and this time I have matching designed a lot better.
Net with that first test was it was getting better as I ran off the edge of the graph - which is defined by how low density I can run. I'm trying to get my peak onto the paper here at a pressure and ionization percentage I can hit reliably. Initial results were encouraging with 25kvdc + 7kvrf giving similar Q to 50kv DC...though lots less neutrons at the lower voltages, it also wasn't eating power either - most went into losses in my pi network! (Glad I used pyrex for coil form...)
I need one more run with that setup to see stray ions out in the tank with the faraday probes and see how fast they're going, but it's obvious I have to go lower in F and probably somewhat higher in RF volts (which will still leave me in a good area in the mass spec equations).

The rest to come. I wish I knew if I could float the secondary of a MOT to make a bridge rectified HV supply for this, and also I'm working on how to protect the output tube screens (which would now just be tied to the 350v plate supply shared with the 6146) - they'll cook if there are various other faults, and these aren't all that cheap, even though most hams seem to think they're not that great. (I do have a 3cx1500, but overkill...and limits on my house power...)
DS0027.PNG
input and grid drive to the final

20190210-4-125 amp-2.jpg
the underside in progress (I've added things since)

20190210-4-125 amp-3.jpg
Fire in the hole! They at least light up.


I've been "conditioning" these to hopefully let them self-getter for awhile. Procedure is to light them up for an hour or so, then tie all the grids and plate together and put enough volts to draw rated plate current (around 50v in this case) and run that all day. This may or may not be enough, I've heard the best way is to run enough power input (somehow protecting the screen, maybe just run the grids at or near ground) to light up the plates red hot - where they will act as getters. I do plan to try seriously current-limited HV on this to see what happens...maybe with a big series R in the screen supply too. Meanwhile, a lot of tedium, adding fans and power supply control and so on. And getting my little coupling transformers right...
It's been both cold and wet here...so I'm not getting in many shop hours for awhile, but spring is on the way.
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: 4-125a amp for fusor ion trap

Postby Doug Coulter » Tue Feb 19, 2019 1:28 pm

I am planning to use transformer output coupling for this. I need to isolate between the ~~ +2.5kv amp plate voltage, and the -25 or so kvDC I'm adding in.
So a transformer will drive the pi network, the RF ground end of which will get the DC feed. Right now this pi is ~1035 pf at the low impedance end, and grid + 25 pf at the high impedance end, with around 115 uh L (needs a little increase to hit where I want, to around 130 uh, which is only 2-3 more turns).

I added the 25 pf in the belief that it would help swamp the few pf the ions add, and to get the F0 down a little with less modification on the pi L, as I'm kind of out of room for more turns there.

The impedance at the parallel resonance is near 500 ohms, and the ARRL tells me my plate impedance is going to be ~ 2500 ohms, so 2.3:1 turns ratio more or less, at least that is going to be easy to twiddle. I'm using a bunch (and this stuff really ain't cheap) of type 61 ferrite rod as a core, and wound a test winding on it.
Isolation is going to be provided by this rather thick wall chunk of teflon. In low power testing with the winding shown, things look pretty good, but of course I'll need better than that skinny teflon wire, the core wrapped in insulating tape and so on for the real deal.
20190219-4-125 amp-1.jpg
Some of the pieces...


The output of the VNA looking into the low impedance end of the pi network (here we are stepping up rather than the usual direction).
L+25pf+Pi.jpg
What this transformer will be driving


All of course, roughly speaking, this is still only "on paper" and no plan survives the first hit...
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: 4-125a amp for fusor ion trap

Postby Bob Reite » Thu Feb 21, 2019 2:02 pm

I'd make the screen supply variable. From zero to 300 volts. Don't go above 400 on the screen (Maximum rating for a 4-125A). Max screen disappation is 20 watts. Maybe put a 240 ohm 1/2 watt resistor in series to current limit. Oh. I've seen people run these tubes grounded grid, with the screen grounded as well though a one ohm resistor.
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: 142
Joined: Wed Nov 11, 2015 1:02 pm
Location: Wilkes Barre / Scranton PA

Re: 4-125a amp for fusor ion trap

Postby Doug Coulter » Fri Feb 22, 2019 11:01 am

Thanks for the advice, Bob. Turns out I'd taken some of it already! Making screen volts variable would be pretty painful here, but I'm watching screen current closely and have what amounts to a fuse inline, as well a a meter monitor. So far so good, using control grid bias for now, and currently have a bit too much of that, so things are all staying too cool as it were.

////
For anyone following along who lacks the resources/info that Bob and I have, here's the datasheet for the output tubes, and maybe more importantly, a link to where such things can be found.
4-125A.pdf
Data sheet for the output tubes used, here I have two in parallel.
(849.29 KiB) Downloaded 372 times

Good place to get vacuum tube datasheets.
Get 'em while they're hot. I've downloaded sheets for most of the tubes in the huge collection I inherited.
////

Well, I've done a tiny amount of testing now - but had to stop till I make some shielding for this, as there's enough RFI generated to screw up even wired keyboards and scopes on that bench....I'm using a test primary on those ferrite rods for a plate load (with no other load, I should fix that) - it's about 300uh and I'm testing at ~ 1900khz.
Real early times for this box.

I'm running the screens at 340v now, metered and with 240 ohm 2w R in series. They seem to be in the range of < 20 ma in tests with fairly low plate voltage (around 1kv). I plan higher plate volts, at least 2.5kv when I get a proper supply going.

If I understand the literature correctly, the big risk to screens is having no plate voltage so all the electrons go there and melt them, or in some tubes (no info on these), too little load on the plate causing secondary emission from the plate to land back on the screens and cause the screen volts to rise too much (haven't seen that). Seems as if the main thing (other than that resistor-fuse) I should be doing to protect the screens is prevent that low voltage power supply from coming on (it has a relay in series with the winding that makes the 350) unless the real HV is there already. I've seen that in other designs, this thing doesn't yet have provision for that. I guess I could detect the big HV with a divider, a neon and a CDS photocell or something like...dunno. It seems an easy mistake to make and probably causes quick death of the output tubes. I could of course source the 120v for this from the HV supply, or better, just the 120 for the DC control relay (so the filaments have warmup time when I do it in order). I want to keep the complexity down and the idiot resistance high, a challenge of course.

I was real gratified that up to 1500v on the plates, and near full drive/swing, no arcs...it might be OK under real life conditions once I move this rig to a place where the stray RF doesn't fry all my other toys.

Yes, I've seen these grounded-grid too, and have an old highly modified Drake amp with 4-400's in it run that way. I've been avoiding grounded grid due to the need for all that drive power and getting away with it so far. If this isn't enough (I doubt it) the Drake is next...I don't think I should need that much, I was almost there as far as I could tell with the 4 sweep tube amp even with some mismatch..this is a pretty good step up from there. This setup not only has twice the rated plate dissipation, but is also class C vs the sweep tube amp being class B. And I think I'm going to have a lot better match, now that I have a better matching network and measurements. It's not quite rocket surgery...close but no cigar so far. I added a little 6 turn inductor on a form that fits in the main pi coil (the beaker) and can be pulled in and out to adjust the L, works very sweetly. Looks like this when set where I think it's going to be:
L+6tdeep+25pf+Pi.pdf
With adjustable L and additional 25pf across the grid
(146.5 KiB) Downloaded 333 times

Oops, PITA pdf. Here's the relevant jpeg.
L+6tdeep+25pf+Pi.jpg
Visible without download...

So, a few hundred ohms load looking into the final pi network at the parallel resonance, should be fairly easy to hit that number. The pi net is 1015 pf input, the L which is around 130uh (adjustable a little) and 25pf+grid, which is around 55pf total (~ 20:1 step-up!) The fusor ions aren't much load themselves, they look like a few more pf and some parallel R (around 750k) in earlier tests. I want to push this to higher ion density, and if I understand right, this will increase the load (decrease the net R at the fusor end of the pi) which should in turn raise the no load impedance at the pi network input. Still should be in bounds for the range of interest. I might have to add a few turns to the output transformer secondary...

I was very gratified by the results of the last run with the sweep tube amp. But things like Q were going up as I went off the page as it were - this will get us further in that direction I think. I can't really run much lower pressure (the edge of the page) so I'm trying more drive and a lower F since it seems the ions kinda smear things out - we measured that at DC and they're seeing a lot lower field than what we think we put on. Seems true for mixed drive as well, so more power and lower frequency to match (to make the equations still predict it'll work, since they seemed correct last time).

To get the bandwidth I want - the 6146 is running out of oomph at pretty high plate current and input drive at the edges of the ~ octave I'm trying for having enough to get to grid current (at all) on the final. Since (so far) the 96v bias on the final grids results in 0 plate current for zero drive, it might need an adjustment so I could get by with less drive RF voltage - I don't think I need a zero ma resting current with these. The previous builder left me tons of holes in the box, so I can add a pot for that....or once I know more, just a series R from the source (it has a 5k to ground at the bottom of the grid transformer so grid current doesn't just increase the bias without limit). I'll have to draw up the schiz...it's a little oddball, but seems like a pretty reasonable and conservative approach with low failure probability - designed to fail resulting in nothing drawing current after only cheap smoke (R used as fuse and what not).

I did this with a MOT on a variac (shunts driven out), a half wave rectifier and a 1 uf filter, so the HV was both on the low side and very ripple-y. I started with a 100w 35k resistor in series with that. I never cranked the variac up to max, where I saw 2600v no load. I'd really like to do something a lot smarter than that....I do have
some 20uf@2500v oil caps, but that half wave thing grates on my sensibility, even though it'll probably work OK (with extra heat and peak current in things). It's after all a 1200w transformer. Which needs the usual 20 or so extra turns on the primary to not be in full saturation at 120v - that's all a work in progress.
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: 4-125a amp for fusor ion trap - impedance matching

Postby Doug Coulter » Sat Feb 23, 2019 4:59 pm

Gotta start somewhere, I'll try to get the schematic for the amp (at least so far) up soon too. This is what it drives, it was hard to describe in just text.
I'm using a 6" rod (4 pieces of .5" type 66 ferrite "in parallel) for the primary. The secondary will be on the outside of a 1.5" ID by 2.25" OD teflon pipe for voltage isolation between the negative HV from the Spellman and the positive voltage from the amp's supply.

This resonates as shown above in the VNA plot with the parts shown.
FusorRFMatchingNetwork.gif
Initial matchining network, voltage and impedance step up from amplifier plates.
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: 4-125a amp for fusor ion trap

Postby Bob Reite » Sun Feb 24, 2019 12:26 am

The good BC rigs had some sort of screen overload protection, even if just a fuse. If you do lose the plate supply the screen overload will operate when the screen tries to grab all of the electrons that were destined for the plate.

Yeah you generally only get 10 dB of gain in grounded grid, but it's so nice not having to neutralize. While in theory a grounded cathode tetrode should be stable, I've had some go into business for themselves and have required neutralization as well as parasitic supression.
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: 142
Joined: Wed Nov 11, 2015 1:02 pm
Location: Wilkes Barre / Scranton PA

Re: 4-125a amp for fusor ion trap

Postby johnf » Sun Feb 24, 2019 2:32 am

I have a stack of NOS 4-125A's
I used them in a student radio station on AM years back two in the RF deck two in the modulator good for 300watts carrier power 24/7/365
At work we used them in a Wein Filter that switched 0, 7, and 10 thousand volts to the capacitor plates in the Wein on a duty cycle basis that matched the C12 C13 C14 ratios that exist in all recently living matter. So it dwelled on c12 for only a few microsecs but c14 for two seconds the capacitor plates came in at around 4000 pf so quite some current flowed through the tubes shifting capacitor charge. Two tubes one on top of the other used as fast switches.
Yep fuse or fusable resistor for the screen.
I found that the usual 30 turns around the carbon resistor anode stopper mostly had these things tamed without further complication unlike later tetrodes of higher gain and frequency limits
johnf
 
Posts: 433
Joined: Sun Aug 08, 2010 3:51 pm
Location: Wellington New Zealand

Re: 4-125a amp for fusor ion trap

Postby Doug Coulter » Sun Feb 24, 2019 10:39 am

So far it hasn't taken off in parasitics as far as I know (there's a ghz scope nearby), but yeah, I know the trap trick. This is pretty well bypassed and tight - all the driver stuff is under-chassis. No hint of smoke so far, I have plate current at zero with no drive (around -100 on the control grid and 340 in the screen) and showing "hard to measure that small" screen current, always well under 10 ma in tests so far. Worked well enough to drive everything on my bench nuts with half the drive it should take to get to class C and only 1kv or so on the plates, so I'm having to set it up so the RF coming off the plate circuit doesn't get into the room so much....prior to more tests - I think I need a little less control grid bias to get off zero, but then I haven't tested it with full plate volts yet (still working up a good plate supply).
So that's where it sits for now.

The fusor.net board isn't letting me log in, or use my real name and email address to set up again....or I'd comment on what I'm doing over there. There is misunderstanding on what I'm up to here.
While I'm using more or less the same math as a quadrupole mass spec - it is after all just the equations of motion of charged particles in a field, I'm no way doing the same thing a quadrupole mass spec does. I'm just trying to bounce the stuff back and forth through the grid (cylindrical symmetry) moving fast enough to get fusion, and not so fast they wind up hitting the tank walls on each pass - a balancing act. To that end, I'm not using the same operating line in the Matheiu equations as a quadrupole, nor am I using the 4 rod setup at all. A mass spec runs just *outside* the stable area, so ions not the right e/m will be lost, and even the right ones will be lost if you make the rods too long. I run in the middle of the utterly stable, pretty non-e/m-selective region of the numbers, as I don't need good rejection of the wrong stuff - everything is way different than D, I don't need to be picky here about accidentally trapping the wrong thing (H is 1...He is 3 or 4, and so on - that's ratios, not little differences!).

So until they get me back on the fusor.net board, this is the best I can do about all that.

While OT for this thread, I want to mention that with 25kv DC + 8kv peak AC I got better Q than I did with 50kv DC (which was already around the best reported on fusor.net), and it's going up as I run off the edge of the plot! That's why I'm building this more powerful source, so I can run the conditions where the peak might lie. It's already working at 100x (at least) the pressure where a quadrupole craps out - probably because of where in the a and q map of those equations I'm running, and it doesn't seem super picky subjectively at least.
I'm guessing this is because at some ion density *not handled by the ion trap math which assumes no interactions* that things are smeared out - we have after all, even electrons and other species, and enough space charge to create some repulsion in the D+s. With DC we measure the particle velocities having around 1/10th the speed they "should" have for the applied field due to things like this. I'm betting the same is true with the ion trap class of drive signals, and this seems to be borne out by the results.

I'm a little hesitant to show this plot, but not for the reasons one might suspect - this is me spinning all the knobs and looking for any sweet spots (which I found) and even some time with no power, where cosmic ray triggering of the neutron detector makes things look "too good to be true" - because divide by zero kinds of issues.

But on the 25kv line, you can see what happens with RF off, and RF on - the lower flat line is with it off, and lower current is lower pressure, which the color map doesn't show very well. Notice it gets better fast as the pressure goes down to the point it's not drawing much power at all - it's barely staying lit. This all agrees with any reasonable theory one might make, except perhaps that it probably shouldn't work this well at again, around 100x the pressure a normal mass spec craps out at (but I'm using a more stable portion of where the math says it'll fly at all).
Further, for fun last night, I tried logging the Z axis (which I have mapped to Q in arbitrary units) and if it'd been going up exponentially, there'd be a straight line as a result. Nope, it's still curved going up, so it's better than exponential. I'm enthused.
Qsweep.png
Note the stuff above 25kv is probably errors, things started to arc and make lots of RFI....I don't edit my raw data here - honesty in reporting and the possibility you'll see something new later on.

This is of course better here where gnuplot lets me rotate this around any axis and other neat tricks.

And here's the version with the log(Q) plotted instead. It's still concave-up. That's pretty encouraging.
Logged.png
Logged Q


If it turns out that stuff with the higher DC bias is "good data" - and it might be, I haven't re-run my scripts that do the math on it yet - then I'm going to need some better bypass capacitors for the DC feed end - that's what arcs now (they're rated at 30kv). Those are to get an RF ground at that end of the drive network, and keep from back feeding RF into the spellman and messing it up - that was a pricey investment!
It seems really hard to find things like 2500pf 50kv capacitors....or thereabouts. Bill found the ones we're using and it's the closest thing he's found so far.
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: 4-125a amp for fusor ion trap

Postby johnf » Sun Feb 24, 2019 1:32 pm

Doug
two 30kv 4700pf doorknob caps in series padded with an RG213 gimmick cap to get exact value. Rg213 101pf /metre
johnf
 
Posts: 433
Joined: Sun Aug 08, 2010 3:51 pm
Location: Wellington New Zealand

Re: 4-125a amp for fusor ion trap

Postby Bob Reite » Sun Feb 24, 2019 1:41 pm

The 2500 pF capacitor probably doesn't have to be that exact value. Probably the bigger the better. It's purpose in life is to provide a ground return for the RF side of things and keep the RF out of the Spellman HV supply.
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: 142
Joined: Wed Nov 11, 2015 1:02 pm
Location: Wilkes Barre / Scranton PA

Next

Return to RF

Who is online

Users browsing this forum: No registered users and 3 guests

cron