RF driver for drift-tube accelerator

High frequency, antennas

RF driver for drift-tube accelerator

Postby Doug Coulter » Mon Aug 29, 2011 6:43 pm

Well, I've decided to make the leap into RF driven particle acceleration for the new beam system. Not necessarily less hassle than DC, but different in kind, and at least I don't need the full potential in any one place that way. Not sure whether I prefer shocks or RF burns, probably the former, but I'll report on that when I've had more experience.

OK, some really brown estimates here, and I know they're not real good, but you gotta start somewhere. Lets suppose I want about 90kv per beam, and about 1 ma per beam. That implies I need about 180 watts of drive if everything else is 100% efficient. OK, so I need 500 watts or thereabouts. I plan to have 3 stages of RF gaps per beam, so I need 30kv peak voltage (60kv peak to peak). I will ground the alternate drift tubes to avoid needing a two phase center-tapped output from the RF driver, and also avoid some loss and voltage standoff issues where one of the tubes is right inside the grounded tubing coupler into the vacuum system -- it'll be nice to be able to have that one grounded.

On top of the presumably mostly-resistive load from the particles, this is going to have to be well shielded and will create a capacity to ground as part of the load. I'm guestimating about 500pf total for all the tubes to shield capacity (assuming I use quartz and pyrex vs something with higher D). Add a few pf for the short coax and some tuning fudge, we need to make this 30kv peak across about 700pf.

To make the tubes decently long at these voltage levels (too short gives you other issues, like can I make it and can things stand off the volts/inch), I chose a nominal frequency of 5 mhz. Obviously I have to stay off precisely 5.0 mhz or I'll be jamming WWV's time service with any leaked RF and bring trouble on my head. But close will be good enough. It's the nature of the device that you can tune the frequency around a little as long as you tune the voltage to match, so that the transit time through a drift tube matches up with the RF phase as the ions hit the gaps.

So, in some perfect world, which I don't inhabit, what I want is a super frequency stable high power, low impedance RF supply, where the voltage wont' change with changing loads (because we all know that no ion source is perfect and all vary a bit). This, quite simply, ain't gonna happen, though it's possible. Yes, I could do a full gallon amateur radio lashup with a VFO (or xtal), 100w or so exciter, and a linear amp with ALC. I even have all that junk, though I have doubts the ALC in it is anywhere near good enough for this. What I don't have is a spare few KW of mains power, or the floor space and cooling for all that, so I propose to scratch build one for this.

Looking around the shack, and keeping simplicity in mind, I'm looking at using a 3-500z triode in a Hartley oscillator setup, much like a vacuum tube tesla coil setup. I might just be able to get it to sync to an xtal oscillator that only puts out a few watts, rather than needing 100w or so to drive it as a linear amplifier. In this case, I want class C operation for efficiency, and to run the power tube at the highest possible plate voltage so the stepup to 30kv peak is easier (far easier than stepping up from a FET for certain). Looking around, I also have a 1400W microwave oven transformer that doubled, will give the rated 4kv needed for the max plate voltage recommended for a 3-500z tube. Looks like I can either add a few turns to the existing filament winding for the 5v, 14 amp filament requirement, or I might use a 5v 20 amp switcher from Marlin for this -- price is right after all. I'd like to cram this whole mess into about a 12" rackmount box, which looks a little tight, but possible if I keep this simple, which is plan A for certain. In this kludge, the idea would be to regulate the output voltage via sensing it off the tank coil, and adjusting the effective grid-leak current to keep the output volts constant. If I'm lucky, the oscillator will do the right frequency change with output power change -- but I don't feel that lucky, so I'll probably build up a 5-10w output xtal oscillator to try and keep it synced, and then just adjust the output power via grid current in the main power tube. In theory, it should take a lot less RF to sync up another oscillator than it takes to drive the same stage as an amplifier.

Now to calculate the circulating current in the tank, given the above assumptions. My gut tells me it's going to be fierce. X = 1/(2*pi*F*C). For .7 nf, 5 mhz, this gives me 4.547284088e+1 ohms. Uh oh, that will never fly at 30kv! That works out to roughly 600 amps! Back to the drawing board! I either need to cut that capacity WAY down, or refigure other things here. So this little post is instructive if nothing else. 60 amps RF circulating current would be at least feasible, though it would require the very best parts in existence to get that level of Q. The question of course is, can I cut it down that far with a physically realizable build of the basic accelerator? Shortening the tubes by going up in frequency is a wash -- cut the capacity in half, run twice the frequency, same problem, no gain there in the tradeoff space. Same thing happens if you double voltage. Seems like cutting the voltage in half and going to twice the stages doesn't buy anything either, because now each tube gets longer to accommodate that, and therefore has more capacity to ground. Hopefully my guestimate of the capacity in the load is way off! It might be wise to build a mockup, complete with shielding and see about that, as this can't be done with reasonably gettable parts. I doubt I could get tank coil losses down below 1% of 45 ohms, after all, that'd be quite a trick with skin effect and shielding losses though not quite impossible. It now becomes apparent why the guys back in the day limited out at 40-50kv for cyclotrons! And I've seen some of their matching devices, they were huge (a couple foot tank diameter by 10 feet long), in a tank (vacuum or pressure) with a lot of distance between the fat tank coil and the tank walls to reduce skin effect losses in the shield, not to mention they were running 10's of kw and had spare power to burn in losses. While I can get about 800w out of a single 3-500z (and I have two), that's kinda pushing what I want in my room anyway.

And heck, I had a neat circuit all drawn up in my mind, but now I have to drop back and see how low I can make the drift tube capacity by clever design work. I was not wanting to have to use tiny diameters, as I figure the focus issues will be not so easy to work out here as with DC -- depending on when particles hit each gap, the effective lens focal length will vary with phase. I'd planned to put a little phosphor screen on the wobble stick I described elsewhere on the board and move it around to look at focus, but it will need a bit more finesse than that to see if the focus varies at 5 mhz! I'll need an RF probe to work with that, movable in the beam.

Sigh, but I thrive on a challenge, and heck, have already gathered the box and parts to make this thing. Those who won't be stopped, can't be stopped! More later. Now I have to rassle up a good cap checker for the range in question. The multimeters don't go that low, and neither do the war-surplus LCR bridges -- I'll have to make and calibrate a known inductor and see where it resonates with my mockup test load. Maybe I can get the load C way down and make this not so hard? Stay tuned. I won't be stopped.
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: 2968
Joined: Wed Jul 14, 2010 8:05 pm
Location: Floyd county, VA, USA

Re: RF driver for drift-tube accelerator

Postby vmike » Tue Aug 30, 2011 9:20 am

Seems like you'd be better off in the long run to run a balanced output rather than to ground one drift tube. The matching is no more trouble than matching to a coax and you will have less loss due to heating in the transfer cabling. The termination impedance of your drift tubes will doubtless be high anyway. All in all, less matching, less loss, less work. If you had two 3-500s you could build a push/pull osc/driver, set your frequency with your tank coil/cap or even make it tunable over a small range. The entire osc/amp is about winding the coil anyway.

Just a couple of thoughts.

mike
I tried to contain myself, but I escaped.
User avatar
vmike
 
Posts: 30
Joined: Wed Sep 15, 2010 5:20 pm
Location: Stillwater, Oklahoma

Re: RF driver for drift-tube accelerator

Postby Doug Coulter » Tue Aug 30, 2011 12:57 pm

I thought about going balanced and really still could (with any number of tubes). A lot of the magic is going to be where the impedance matching from about 3-4k at the tube plate to whatever the accelerator is. Assuming the "DC" load is 2ma/90kv (as far as produced beam energy is concerned) we get about 45 megs for that. Volts look like about a 10:1 stepup ratio. Neither will match with any coax that can physically be made of course (about 1.3k is the limit with near-infinite D/d ratio for any center conductor that will take the current), but at merely 5mhz, any coax is going to look like a capacitor to ground, not a flat impedance anyway. It's almost looking like you make the matching part integral with the accelerator device to keep output capacity as low as possible - or burn up things with the internal circulating current and I2R loss in the conductor to the load. I'd probably put the matching tank right at the accelerator (dealing with 30kv rf without radiating it or making too much capacity to ground is the real bear it seems) and use coax from the tube plate to there - in a pi network situation, you need some extra capacity to ground on the low impedance end, which here is the tube plate rather than the feedline to the device. EG our pi network (if that's how we configure the match) is a step up rather than the usual step down.
In this case, an extra ~25 pf from a foot of coax added to the capacity of the drift tubes is enough to kill it, practically speaking. That length will want to be as near zero as possible, and handbuilt hardline with large D/d required for the final wiring at the high side.

Trying to run even a short balanced line in air with those numbers on it scares hell out of me in a shop full of semiconductors and expensive gear. RF in the shack! I want a real solid shield regardless. It only takes a tiny imbalance to radiate quite a lot near-field with these numbers, even though the frequency is pretty low. Just the coupling of a pf stray capacity carries some real nasty current to things around the line.

I do have more than one 3-500z (and several other big xmitting tubes) but the constraint is more space and power input. I'm running this entire mess (houses and shops included) on one trace SW 4024 inverter...I'm not sure either of the tubes I have is good. I hear they get gassy sitting around, and that you have to contrive a way to run the plates red hot at low voltage to re-activate the zirconium getter they're coated with to put them back in use after sitting awhile. The 700-800 w I ought to be able to get out of one should do the job fine with some left over.

They are said not to like "loafing" as you really want to keep the anode hot to keep the getter working, which takes a couple hundred watts of plate dissipation per tube (on top of the 72 watts for the filament alone). I suppose we could buy new or new old stock ones if that seemed worth it, I think they go for a couple hundred bucks these days. Bad, but considering the effort and cost of the rest, not that bad perhaps. It could be almost required if I go with colliding beams as they're going to have to be real snug up against the accelerator tubes for this to work, it seems. You'd really have two separate complete RF things in that case, and lock them together some way at some lower power point - if they get out of phase your collision point moves around too much in beam-beam work. To make that a little easier, I could do the beams at 90 degrees instead of 180 to put the tubes closer together. You then get .707 of the interaction energy, which could reasonably be made up elsewhere.

This looks like it will take a few iterations to get going, but that's kind of normal for anything this out of the mainstream. I was thinking about that in the mechanical design so that's good - I made it easy to fiddle with on purpose. AT that, this will be easier in a lot of ways than dealing with 100's kv DC which has its own bunch of issues when small size is in the mix.

Not much has been done in this energy range with RF. Mostly they were shooting for more sub-atomic physics energies (megavolts and up), and to get there with this type accelerator, using heavier ions, lots of tubes and lots of length. This is a bit of a twist on anything yet made that I have info on. I see no point in pushing to higher energies than the cross section peaks for the stuff involved in the reactions I want to do. In fact, according to the literature, I'd have a ton of trouble using this design to get relativistic with protons or deuterons or tritons.

Seems to me the next step really is making a mockup (or more than one) to get a better feel for the actual capacitive load, rather than pull some big number outa my butt, and scheme ways to cut that down so I don't need huge center conductors to carry the circulating current between the L of the tank and the C of the load. This exercises main payoff so far was to show how important that is in the overall design.

Studying the target device literature, I may need to go to more stages to get the bunching (luminosity during the pulse) I want. That would allow lower volts per stage to get to the target energy in this case, so driving it balanced, the net volts required goes down even more. That might take it to the point where it won't have "issues" where the beam pipe has to go though a grounded coupler, dunno - that was one of the motivations for grounding half the drift tubes. You don't want beam pipes to shatter from dielectric losses....At these kinds of power densities, even quartz will get warm and maybe even have punch through issues.

For those wondering what in the world we're talking about, a few pages from Haliday here on it. I will probably use my uwave ion source, and handle radial focusing like Alvarez did via lens aperature design (taken from Terman, of course). Just a couple books from back in the day that don't seem to have lost relevance.
LinAc1.gif
Haliday on resonant linac drift tube design

LinAc2.gif
More on drift tube accel issues


This is spread out over a few threads here. I'm using speed of particle calculations from another thread to get tube lengths and speeds/feeds, and detailing the physical construction of the device in yet another one (or a few, if I've not posted the design for the magnetic wiggle stick, that gets a thread too). That thread needs updating, as I've made a target/probe holder and some targets for it. Funny how one project really won't fit into any one thread and keep true to the title and organization, but that's just how science, and especially research is -- a little from here, a little there and then tying it all together.
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: 2968
Joined: Wed Jul 14, 2010 8:05 pm
Location: Floyd county, VA, USA

Re: RF driver for drift-tube accelerator

Postby chrismb » Tue Aug 30, 2011 5:11 pm

(PS.. you did ask me to generalise that little formula I gave, which is thus, and also added to the other thread...)

The time t, in microseconds, taken for a particle to cross a potential of V volts between electrodes D centimeters apart, is;

t = D.(SQRT[ 2A/Vq])

A = Particle's atomic mass
q = Particle's unit charge
chrismb
 
Posts: 620
Joined: Thu Aug 05, 2010 7:32 pm


Return to RF

Who is online

Users browsing this forum: No registered users and 1 guest

cron