Well, this is an ongoing issue. With the 1" OD original design, the failure rate was too high for my taste. I have noticed that getting it really hot doesn't help, and being sloppy about super-cleaning all the glass/quartz parts is a fairly big deal. But I want to not only use all the SL2KW (50kv/40ma) supply can crank out, but go to a newer stack that can make more voltage - 80-125 or thereabouts. I am hoping I won't have to encase the "in-air" parts in oil as Carl W cleverly did, as that would reduce the adjustability and convenience of the thing a good bit. Pursuant to that goal, I've redesigned the thing to go to 1.5" OD, and with a very radiused end that is also the end connector for the large cartridge type ballast resistor I use (50k/225w, forced air cooled). Further, I've bought some much thicker wall PVC to wrap around it all and direct the cooling air, inside the 7" PVC outer piece that is copper covered (for my safety and EMI reasons).
So, here are some shots and movies of the build. I'll of course report further once this is "in action" as to how it works (or fails). This is a very brutal envirronment for a feedthrough - charge accumulation (we have a capacitor in every piece of glass or quartz whether we like it or not), chemical reduction to conductive stuff - silicon etc, and sputtering of conductive stuff onto the FT. So this hues true to the original FT design idea - I can't make it so it can't break - so I make it so it's easy and cheap to fix.
Yes, pure BN (not the stuff with a binder) or pure alumina would not be reduced, and have some other advantages. But from experience, regular alumina (AD-84) is not pure enough by far - the binders reduce, and pure stuff is so expensive and hard to machine - as well as not gas tight to an O ring seal - it's not worth it. And when I asked Accuratus about pure BN (comes in several grades and of course we'd want the non-hygroscopic, no-binder version) - we're talking multiple thousands of dollars. What if it fails? That's a real big hit to the wallet. We might have to go there - the other advantages, like huge thermal conductivity - are compelling, but that price...
Nicely polished with the big coupler. You can also see the lathe tool I made to cut O ring grooves. Just a chunk of A2 drill rod ground up with a support for my tool holder. I run this slowly and delicately of course, and have to resharpen it after every job, but since I don't have to do this much, that's no big deal. Also shown in the little jar is "Doug's magic vacuum lube" which is about 85% beeswax mixed with a little Diffoil-20 to make it spreadable at room or warm temps. You can't put this together without this. After it's heated a few times, it all disappears out of the system according to the mass spectrometer.
I've found through experience that Al makes about the best conductor. It's good and heat conductive (well...long and skinny it's not perfect like a heat pipe, and we might go there later) but the real deal is - after anodizing it, it just doesn't arc or draw current in the exposed plasma anywhere like copper does. My longest lived example of this (still living) has a zillion hours on it and little to no sputtering to the insulators, it just works. I've even heated the end of that one to very close to the melting point of Al, and it didn't even sieze threads on the grid that was on it, or show any damage. Aluminum oxide, the eventual result of anodizing and sealing, is impervious to reduction by hot hydrogen, and somewhat self-healing it seems. Just an FYI - for whatever reason, this works better than all other things tried, and there have been a few.
Anodizing need not be hard. I use one part battery acid (H2SO4) to 3 parts water (distilled) for the solution. Do a little work on a junk piece, as the solution works better once "aged" with a little aluminum sulfate in it. You can use Al flashing for the cathode, or lead as I show here, and I use roughly 7 amps per sq foot. If you do this with a constant current capable supply, you go for peak anodic resistance - the voltage rises, then drops - you stop right then to get the maximum thickness. This proceedure is somewhat complicated by the fact that as the solution heats up - it has a huge tempco and will make the voltage start to drop before you get there...try 45-60 minutes for ideal results, then seal by dropping the piece in already-boiling (de-oxygenated thereby) water for half an hour to seal the anodizing (close up all the honeycomb pores).
I did this with all the metal pieces after polishing them to mirror smoothness. Anodizing makes the surface look hazy, but...it's worth it. It still seals to O rings fine.
Here is a mockup without the telescoping quartz inner insulating tubing. This shows the big Lesker tubing coupler, the metal pieces and the custom plastic clamp I made to prevent the thing from creeping into the vacuum tank - it's a problem even with 3/4 and 1" ones, so I assume this is going to be even worse. I didn't have PVC of the right size (at least not without a change of lathe chuck to chuck the too-large stuff I have) to do this, which is what I prefer. Instead, I cut out a round from HDPE with a hole saw, turned it to exact OD and bored it to exact ID dimensions, then added 3 setscrews with inner tube rubber on the glass side so the screws wouldn't shatter the glass with point-force. This is because, hey, HDPE is slippery - the opposite of what's wanted. At least it's a great electrical insulator. This piece also serves to block airflow from that end of the 3" ID PVC clear pipe we got to help insulate this whole mess, so the air I inject into there with a regenerative blower will flow out, along the Al end, along the ballast resistor, and out the back efficiently.
I decided to use quartz as inner insulator. For a lot of reasons. Pyrex is ideal for the outer part as you can get it in inch sizes (quartz is all mm) that fit, and its slight conductivity has the same effect as corona rings/resistors in equalizing the voltage drop along it - but without all that big junk and wasted heat. But pyrex has a high D (around 8) and is not ideal as an insulator. So quartz, with a lower D, higher Q, and better standoff volts/inch is the bulk of the insulation.
Here's some vids on how I do the quartz, and another showing of the mockup.
http://www.youtube.com/watch?v=_1d1seqp ... jA&index=3
Cutting off the chipped end - cats are dangerous to quartz if you just lean it on the wall(!).
Finishing the cut on one of the pieces:
http://www.youtube.com/watch?v=d2_zZubM ... jA&index=2
And the result:
http://www.youtube.com/watch?v=xGiyLmxp ... jA&index=1
All of this will be "super cleaned" before final assembly - I've found that even a trace of skin oil makes for trouble. There is a bit of carbon stuck inside the quartz from flame annealing and soot coating, but I'm thinking that will make it work better, actually. The main failure is charge concentration at a spot - which then makes an arc through. A little conductivity along the length of each piece should (in my theory) help equalize that and keep "hot spots" down a bit. We'll find out in use, of course, and I'll report how it works out. This is somewhat making a virtue out of a bug - that carbon is darn hard to get off even with abrasion and ultrasonic cleaner.
Hopefully, I'll get all this installed and fired up in a week or so - there is a lot to do as ancillary work for this teardown. We plan to also wind some water-cooling pipe around the sidearm, insulated so we can also use it to make an axial magnetic field of a few hundred gauss, and a couple "ring electrodes" off each end of the grid to play with making the electrons dance around as I think we want to do. More on that perhaps in the theory section here, but the basic idea is to make the electrons take a much longer path from the grid to the tank sides - getting more ions per electron, and maybe make them oscillate axially near the sidearm walls for the same reason with some RF on the ring electrodes, same idea. This is one case there the hugely larger e/m ratio of electrons vs ions is on our side - so I'm going to try and take advantage of that. The tiny fields that will really affect electrons won't be noticed much by the ions, which are lumbering giants by comparison, particularly if we are talking the effects of RF fields. The amount that will move electrons back and forth 6" or so at some frequency won't make the ions dither more than a tiny amount. Ditto with the magnetic field. That will be its own post, of course.