Microwaves have some uses around the place, particularly in our microwave ion source. They might be used to impart some spin on ions too in a fusor - there are a lot of possible applications in the fusor with or without other fields. The cheapest way to get microwaves at power - assuming you want 2.45 ghz or thereabouts, is to use an oven magnetron, in this case a panasonic 2m167 I had scavenged (these are common and available from microwave oven repair parts suppliers). 2.45 ghz is right in the region where most are going to waveguides, rather than coax, but other than scaling issues, you can do either if you have low loss coax for the frequency involved, and by golly, it sure can be a nice thing to be able to remote the tube itself a little from the apparatus without having to make a big waveguide. As also pointed out in the stuff for the uWave ion source, in a waveguide you are generally stuck with lower power density than you can get in a more coax-like situation (there it was a tuned quarter wave cavity), and sometimes you need or want that.
As this sort of thing is "rocket surgery" for many, I thought I'd show what I did with a side of why I did it this way vs some other way. Here's what the result looks like:
Not pretty (yet), but it works fine as shown. I used a computer type chassis fan with some wire through the magnetron heatsink area to hold it on (baling wire!) to keep the maggie3 cool. Even at low powers, you're putting around 36 watts into the filament alone and that will heat things up to the point of messing up the magnets if you're not paying attention. This little fan gets it done with minimal fuss, and my magnetron power supply has a 12-14v wall wart incorporated to drive the fan.
Here's the rocket science part. Normally, that stub that sticks out of a magnetron is actually a quarter wave antenna over a ground plane, and stuffed into a waveguide to radiate power into it. The nominal impedance at the base of such an antenna is about 72 ohms, which is one reason you can easily buy coax with that basic impedance. Now, quarter wave things with a low impedance at one end show a very high impedance at the other end - if one end is shorted, the other acts like an open circuit (infinite impedance) 1/4 wave away, sans any losses that exist to make the math not perfect. This means that at the top of that antenna, the voltage is "infinite" again assuming no losses. In real life, losses are fairly negligible, so this isn't a very bad assumption. What this all means is that at the tip of that antenna, the RF voltage will be quite high - enough to make most RF connectors arc, for example, if the antenna has low losses (is prevented from radiating). Arcing is very much not what's wanted here. Well, it turns out that at 1/2 (or any even multiple of 1/4 wave) the impedance is the same at both ends. Presumably, they've done their homework inside the magnetron so that the impedance at the base of that antenna (under the ceramic insulator) is about 72 ohms, so we want a pickoff spot 1/2 wave from the base, meaning we have to add another 1/4 wavelength to the assembly so that 1/2 wave happens right inside the connector where it would be most likely to arc at high powers. Assuming we're going to use 72 ohm coax, everything should match, and things like antenna tuners or exact lengths of coax aren't required if we load the other end of the coax correctly (say, with another 1/4 wave antenna).
Thus, what's needed here is just enough additional length so that when included in the connector effective length, we get another quarter wave from the tip of the maggie. We also need to make all this live inside a ground shield, so the whole thing is more or less coax. We have some limit there to how perfectly we can make the impedance of this short piece match 72 ohms, as the stub on the tube is the diameter it is, and the ring of RF gasket is the size it is - but a small discontinuity here doesn't seem to hurt anything much. It turns out some magnetrons are more forgiving on the range of outer piece diameters that can engage that RF gasket than others, but I found the ideal match in a piece of plumbing coupling that was meant to be female to our standard 1" ID copper pipe, and even the length is correct right off the hardware store shelf. The actual ID of this piece is larger than 1", more like 1.125", and it just fits over the tighter magnetron shown below:
The sharp eyed will notice I cut a 45 deg chamfer on the inside of the Cu fitting to make it self center a little better, and bite the gasket harder. This isn't required, but when you have a lathe, there's no reason not to make it nice.
So, how do you make all the connections, and still not have a hole in the side somewhere that will leak microwaves out into your environment - particularly the center conductor? I tried a few things that turned out bogus, and then came up with this rather slick idea that does indeed work. But first, the ground. The female 1" pipe to pipe coupling does this part - one end punches into the gasket on the tube, the other end gets a piece of copper welded over it, with a 5/8" hole punched in it to clear the little protrusion most type N female chassis mount connectors have. But before you solder that endplate on, you first make a piece of thin sheet steel 2" by 5" and make some holes in it - 4 for the standard microwave bolt pattern for this type tube, and one big one in the center, 1.225" or so, to clear the plumbing fitting. This sheet is silver soldered onto the pipe fitting about 3/8" up from the bottom, so it can act like a spring when you bolt the piece to the magnetron, forcing the copper into the gasket. (in truth, this is the belt, the suspenders are elsewhere).
Now you have your basic outer conductor for your first part of the device, the short "adapter coax". Since you silver soldered or brazed the joint that holds these pieces together, you can then solder on the top piece without worrying about re melting that joint. I used plain old copper flashing for this - that's on the thin side, about .020", but for this it's adequate.
Now, assuming the connector is about 1/2" more or less to the "arc danger area", and knowing that 1/4 wave, the amount we want to add, is about 1.1", we need a bit of extension to make that happen, and to match the serendipitous correct length of the plumbing piece. This turns out to be about .488" we need to extend the top of the center conductor on the magnetron. Here I cut, drilled and tapped a 1/4" copper rod (3/8" or even 1/2" might have been better) to 4-40, a standard screw size, and soldered the bottom onto the tube. Since the tube top is stainless steel, you need that special flux you can get from McMaster to do this well - no matter how you sand, file, clean, soft solder won't stick on this without that flux (has some HCl and some Hf in it). You could, in theory, also braze or silver solder this, but since that's also right where the vacuum seal pinch-off is, I wouldn't try that unless I had a few spare tubes to maybe make into garbage learning how to do it very quickly before I did heat damage. The particular tube I used for the prototype would allow me to use a mere 1/4" diameter rod for this, but the one shown next to it would have let it fall inside - so for that one I'd have needed 3/8" or more - no big deal, just something to look out for - the tube antenna is about .6" so you've got some some fudge to play with there.
It turns out the solder end of the type N connector sticks down a little too far, it has a lip, so I simply sanded that off on my belt sander. There's still plenty of meat there, and by golly, it's an almost perfect sized hole to tap 4-40 too - a little large, but tight enough, so I did that. I simply cut the head off a short 4-40 screw, cleaned up the threads, and screwed it into the type N.
So, now it's time to put it together. You put the plate/plumbing down first, and finger tight the magnetron nuts to give it some clamp action - that's the reason we used thin steel as a spring, and used slotted holes to accommodate some spring here. The chamfer on the fitting makes it self-center on the tube nicely.
Then you go ahead and screw the type N connector center screw into the copper threaded rod you soldered on top of the magnetron, till it's fairly tight (don't over or under do this one and strip or break things, but you want some tension on those threads to make good contact). Once you are satisfied you've got it right, you then mark and drill 4 holes in the flashing end to put screws in to make contact, and ensure it won't turn in use and get unscrewed (and then leak microwaves). You can get away with such thin a top as the flashing, because those screws are in sheer, not tension, in use - they need no pull out strength to speak of, it's that one in the center conductor doing that, and to be honest, that alone might get the job done and you not need the steel spring piece. I like having both, myself. Tests show no microwave leakage from this, but low quality coax (for this frequency) will get warm at a couple hundred watts microwave power, so choose that wisely. Teflon is nice...and you want the big stuff (.4" OD), not the little stuff (.2" OD) for this job.
If warming of this due to skin losses is an issue for you - first make sure the copper is very clean, or you can silver plate it, as I plan to do just to make it really nice. You lose about a watt or two in skin effect losses with dull copper, enough to notice and get a little warm.
I've run this both CW and pulsed, using a modification of the same power supply we used for the microwave ion source. Works a champ! Microwave away! Remember, if you're going to run a lotta power, you've got a microwave oven, so make sure you put a screen over any viewports and pay attention to microwaves leaking out through other feedthroughs, as they will if you let them. In my case, all the FT's are faraday caged and lead sheilded for X ray and EMI purposes anyway, but be aware you don't cook your corneas looking into an unshielded view-port!
I'll edit this with links to the original microwave ion source and power supply when I find that again.
And here they are - this is on this board and shows some others who have done the original design and had it work well.
And here's my original write up on the old web page.
This is all of a piece - the idea here is being able to dupe the effects of the original without having to force the ions to traverse a long skinny tube on the way into the fusor, but to have them there already and let the main fusor field attract them better than could be done with the original design.
A half wave at 2.45 ghz is just about the grid to wall spacing in my fusor. This *may* allow me to orient some of the D's coming in from one side the opposite of those coming in from the other side, which in theory should increase the chances of not trying to violate spin conservation (Pauli) and get the fusion reaction I want to favor - D+D->He, versus the neutron or proton making more-common ones.