Getting high voltage into a vacuum
Posted: Sun Aug 08, 2010 11:24 am
Edit - I moved this topic here because it fits better here - this design series "almost works" and doesn't give a long life due to reduction of the glass or quartz from hot D ions, the unequal division of E field in a stack of varied (or identical) insulators with non-zero spacing (so you can fit them at all - none of this tubing is precisely round) and so on. These work, for awhile, and are easy to fix, but now the new BN one I did, which I will soon document better, is the "way to go" and no mistake about that, even though it cost a lot more - it cost less in the long run...literally. It's funny how you spend all this time documenting attempts and what turns out to be temporary success, then forget to give the real answer that solved it for good and all - and it's not purely any sort of especially amorality. It's just that once a problem is solved, it's solved, and you move on. And yeah, no one likes showing off their failures, sure. But the point of this site is to instruct, to write the book on how it is really done - right. And also to show how that idea that seemed good at the time - and was - was supplanted by a better one. This saves everyone effort - no need to try (or suggest) what others have tried and found the failure mechanisms in, thinking you've thought of something new. Nope, it's probably been done, but not documented due to human and other failings. I'm hoping to correct that here, at least a little bit.
I had enough expensive commercial HV feedthroughs fail that I finally bit the bullet and started making my own.
There are some pictures and words here.
This design works well for DC and RF, but has a problem similar to the commercial ones -- it doesn't like hot plasma much, and in the case of hot hydrogen, the quartz gets reduced to metallic silicon after some running. It does solve some of the problems with the commercial ceramic designs around the fact that the conductor is exposed far back into the open volume in those.
But I've recently had to cut back the quartz/pyrex end inside the tank and add a ceramic insert that is easy to replace -- it gets reduced too, but not so quick, and if I can find some pure alumina, is supposed to reduce into something volatile that the pump will just take out over time. The thing is, I've not found a good source of the pure stuff as decent size tubing yet. The stuff at McMaster is mostly 84 Al, which has a binder that reduces to conductivity. Specifications of things are a weak point with them -- most don't care, and it gives them more flexibility to change suppliers on the fly.
Quartz is by far the best insulator you can use for most of this, but on further analysis, that's not always what you want. I've gone to pyrex for the outer tube in that design, as the higher bulk conductivity keeps most of the glass nearer to ground -- so it doesn't attract ions so fiercely, with a quartz inner lining so it's not touching the high voltage electrode.
The tubing couplers I use are surprisingly good for high vacuum -- they don't seem to have enough linear dimension to have much permeation through them, at least compared to the 7" viton O ring around my big door/window. I just didn't notice a change when I went from none, to one, to now, 3. They pass a He leak test with flying colors.
A major feature, or bug, is that they don't hold the tubing very well along the in-out axis. This means you have to keep the tube from creeping into the tank under vacuum, but also that you can adjust that as you want without breaking vacuum, or even making much of a leak while you do it. Since we have been playing with what happens when the end of our cylinder grid sees the field gradient near the end of the side arm it's running in, this is a "feature" for us, not a bug. Having the grid simultaneously occupy two places where Paschen's law works out differently seems to help us run better under some conditions as the ionization is easier to get going in the big part of the tank than in the sidearm. Things seem to happen quick when just a little of the grid end is exposed the the larger tank, so I machined a clamp for the glass tubing, and some shims to place in there so I can position it precisely and repeatably.
I had enough expensive commercial HV feedthroughs fail that I finally bit the bullet and started making my own.
There are some pictures and words here.
This design works well for DC and RF, but has a problem similar to the commercial ones -- it doesn't like hot plasma much, and in the case of hot hydrogen, the quartz gets reduced to metallic silicon after some running. It does solve some of the problems with the commercial ceramic designs around the fact that the conductor is exposed far back into the open volume in those.
But I've recently had to cut back the quartz/pyrex end inside the tank and add a ceramic insert that is easy to replace -- it gets reduced too, but not so quick, and if I can find some pure alumina, is supposed to reduce into something volatile that the pump will just take out over time. The thing is, I've not found a good source of the pure stuff as decent size tubing yet. The stuff at McMaster is mostly 84 Al, which has a binder that reduces to conductivity. Specifications of things are a weak point with them -- most don't care, and it gives them more flexibility to change suppliers on the fly.
Quartz is by far the best insulator you can use for most of this, but on further analysis, that's not always what you want. I've gone to pyrex for the outer tube in that design, as the higher bulk conductivity keeps most of the glass nearer to ground -- so it doesn't attract ions so fiercely, with a quartz inner lining so it's not touching the high voltage electrode.
The tubing couplers I use are surprisingly good for high vacuum -- they don't seem to have enough linear dimension to have much permeation through them, at least compared to the 7" viton O ring around my big door/window. I just didn't notice a change when I went from none, to one, to now, 3. They pass a He leak test with flying colors.
A major feature, or bug, is that they don't hold the tubing very well along the in-out axis. This means you have to keep the tube from creeping into the tank under vacuum, but also that you can adjust that as you want without breaking vacuum, or even making much of a leak while you do it. Since we have been playing with what happens when the end of our cylinder grid sees the field gradient near the end of the side arm it's running in, this is a "feature" for us, not a bug. Having the grid simultaneously occupy two places where Paschen's law works out differently seems to help us run better under some conditions as the ionization is easier to get going in the big part of the tank than in the sidearm. Things seem to happen quick when just a little of the grid end is exposed the the larger tank, so I machined a clamp for the glass tubing, and some shims to place in there so I can position it precisely and repeatably.