Page 1 of 1

Coulter Pattern HV Feedthrough

Posted: Wed Jan 09, 2013 6:43 am
by JakeJHecla
Hello all,

- I've been making some progress on my fusor, and I'd like to share some updates about a new feedthrough I built. After dealing with a dodgy oil potted AC44 sparkplug feedthrough for a while, my new HVPS killed it in a rather dramatic fashion. The insulator failed on the vacuum side near where it emerges from the metal body of the plug. It let go at about 40kV with a nice “pop” and some porcelain fragments. The nasty EMI off of the arc was sufficient to shut off my webcam and wreak havoc with my laptop! Because I'm on a shoestring budget, an off the shelf feedthrough was off the table as a replacement. In light of this, I decided to try my hand at making a more robust design based on glass tubing. It didn't take long after starting down this road to discover that Doug Coulter had already built and tested one! As a result, I modified my plans and borrowed some tips from his lab report. Below is a summary of what I did so as to assist anyone else in need of a cheap feedthrough.


1”x15/16” SS tube (1’) (Online metals) - $7
Swagelok UltraTorr Tubing coupler (1”) - $30 (eBay)
1” Thick wall borosilicate tubing (Hotglass Color and Supply, Seattle) $7.75
.375”x .25” boro tube (supplied in 5ft lengths, also Hotglass) $4.00
1/4” Aluminum rod (McLendon’s) $2.00
Harbor freight diamond cutting wheels (x3, they’re crap) $4.00
Misc. viton o-rings (unknown provenance, found in the bottom of the toolbox)
1.25” x 3” Aluminum rod (toolbox find)


Metalwork, welding: I started off with a 6" conflat blank that had previously had a sparkplug feedthrough installed. I removed the old feedthrough, and then used the school’s lathe and a boring bar to cut a centered 1" hole though the flange. Next, a friend of mind TIG welded a 2” long piece of 1” x 15/16” SS tube into the through hole (flush, weld seam on the vacuum side!).

Glasswork: Using several harbor freight diamond cutoff wheels, I cut two 6” lengths of 1” borosilicate tubing. The ends were then flattened and smoothed with smaller diamond deburring tools until they were no longer a mortal danger to the o-rings. I then cut a 15” length of the .375” boro tube to use as a stalk insulator. These were then set aside to await modification.

Machining: I turned down a 1.25” x 3” piece of aluminum to 1” OD over its whole length. I used my lathe (a little Unimat I got for my 18th last year) and cut a 1” section of the aluminum down to ~.8” so as to fit inside the boro tube. Next, I made two evenly spaced o-ring grooves to fit two fortunately sized o-rings I found in my toolkit. Finally, I drilled a .25” x .5” hole in both ends to create an interference fit with a .25” OD aluminum rod. A 16” section of ¼ rod was then cut and pressed into the vacuum side of the plug. A .75” nub was pressed into the opposite side as a powersupply clip point.


The aluminum rod and plug assembly was first slipped into the 1”x 6” boro tube. I then covered the rod with the 15” x .375” OD boro tube, and a grid was attached. The rod and grid were then inserted through the SS tube in the flange. The two 1” tubes (glass and SS) were then sealed together with the ultratorr coupler and pumped down.


Even before bakeout, the leak/outgassing rate was fairly small (a micron or two a minute). I have run my HVPS up to its maximum voltage (50kV), and no arcing was present. Lots of corona though! When running the fusor, however, there was very significant x-ray leakage. The feedthrough has since been nestled inside a much larger PVC pipe externally lined and capped with 1/16 lead, which seems to have solved the problem.


Sorry for the poor quality, but my only camera other than my cellphone was recently run over by a jeep. These are the best I can do.

1) Fully assembled feedthrough
2) Closeup of the plug
3) Assembled feedthrough in action on the fusor

NOTE: Should you choose to follow these instructions, you agree to hold me harmless for any idiotic thing you may do to injure yourself. This project is not safe or advisable. If you die while attempting it, blame Darwin.

Re: Coulter Pattern HV Feedthrough

Posted: Wed Jan 09, 2013 1:01 pm
by Jack Puntawong
Great work! Wow, thats a neat looking assemble HV feedthrough you've got. By the way, how do you know when it will "let go". You mention that it'll let go at 40kV but is there some sort of indication? I haven't operate my fusor yet because of the unfinished chamber. I want to try crank my stock 30kV feedthrough to about 40+ kV. How do you know when it's at it's limit?

Re: Coulter Pattern HV Feedthrough

Posted: Wed Jan 09, 2013 4:53 pm
by Doug Coulter
Nice job, and thanks for the credit. I'm still fooling with design details myself, as in things like how much glass (you can get telescoping tubing in pyrex or quartz) thickness, and how far the glass part sticks inside, and whether to put a grounded shield over it so hot D ions don't reduce the SiO2 to silicon, or metal from the grid or stalk sputter onto the glass and make a short.

It's real obvious when one of these "lets go", no worry you won't notice - you will. The big advantage of the basic design is not that it never fails - they do. The advantage is how cheap and easy they are to fix if they do - the metal parts never fail, it's just the glass/quartz, which is just a straight piece of tubing cut to length - easy. I have found that leaving the system under hard vacuum, and bringing up the HV (after the D fill) very slowly until you no longer see little flashes on the grid is a good way to make them live longer. I have also found that if you let pyrex get good and hot, it becomes a much worse insulator. I like pyrex for at least the outside pipe, though, as it's slight conductivity helps distribute the HV from the hot end to ground evenly and reduces the chance of arcing outside the tank quite a bit. But it's high dielectric constant is bad other ways, so right now I'm doing pyrex outer and quartz inner pipes.

This is an ongoing work over here - I'm always improving mine or trying something to make it live longer, with somewhat mixed results - I plan to report when I come up with any certain improvement, as some of the things I've tried seem to help, some not, and mostly help some aspect while making another worse. For example, the copper pipe/expander I used to shield the glass on the inside worked pretty well grounded, but a nicer machined and anodized Al piece there seems to have made things worse, and I'm working on figuring out why at the moment.

Re: Coulter Pattern HV Feedthrough

Posted: Wed Jan 09, 2013 10:32 pm
by Peter Schmelcher
I also have given up on scrounging a HV feed through and have started down the build it path.

Doug, do you have a feeling for the creep distance down the Pyrex tube surface? Just wondering how short the surface can be for a 50 to 70kV feed through. BTW Parker has an o-ring design manual with a vacuum leak versus squeeze graph on page 3-19 ... ndbook.pdf .
In the past you have mentioned housekeeper seals. They seem like a very doable seal if you have a lathe. My uncle was in the neon glass business and I was bending glass before I was a teen, so familiarity probably distorts my perspective.

Jake, I purchased an inexpensive 6” diameter diamond lapidary blade via ebay from China for $10 including shipping. I used it to cut 1” off the bottom edge of a dangerously chipped and abused Pyrex bell jar. The trick is to keep the diamonds cool while having a fast surface cutting speed. The fastest edge speed I could comfortably muster was 30 miles per hour. I used standard machining coolant, typical biodegradable water soluble oil, plus of course eye protection and a heavy jacket should things break badly. Glass can have internal stresses that cause it to shatter unexpectedly. In the end the 32” perimeter cut did not measurably wear the blade, but took 6 hours. The internal stresses only caused a small and unimportant flaw in the finished edge. In hindsight I should have first made a shallow depth cut from the outside then cut all the way through from the inside.

Re: Coulter Pattern HV Feedthrough

Posted: Wed Jan 09, 2013 10:58 pm
by Doug Coulter
Above 50k or so in air, all bets are off, spark lengths tend to become unlimited due to the "streamer field emission" effects. So it's not so much the length of the pyrex, as far as I can tell here - when I have arcs, say off the feed end of my ballast R, which is a long straight resistor with the other end at the FT, the sparks off the back are longer in air than the glass length on the FT to ground! This is probably because I took a lot of care not to have any sharp edges at the FT, but not so much at the other end.

When I go to higher volts, I plan to put the whole mess there under oil, or maybe diggable potting compound, ugh.

I once fired up a 125kv supply with just a wire to a FT. I got a spark almost 10 feet long - directly *away* from the nearest ground (the tank), it made a U turn at my mill without hitting it, and came back to ground on my tank - 8' away. That'll make a believer out of anyone. I'm just glad I was standing on the other side of the tank at the time.

A housekeeper seal could be used since we're not fooling with a ton of current here, but then you'd have to re-make the whole thing when (not if) the glass fails somewhere. I've had pretty decent luck with straight seals of tungsten rod into pyrex though, as long as the tungsten was .040" dia or less. The procedure there is fairly easy, but it pays to weld nickel or something to both ends of the tungsten rod first - some of them leak lengthwise. If anyone asks, I can detail the procedure, I've been thinking about making a movie and putting it up on youtube for that one - once you see it done, it's easy.

The nice thing about my design is I use a thick, heat conductive rod and it pulls heat out of the vacuum (eg your grid) better than most. Anodizing the Al if you use that seems to help a lot more than you'd think at first. It seems the high surface resistance (even though it's not enough to stand off the full HV by far) limits arcing and sputtering quite well, and hot D does not reduce aluminum oxide.

I cut my glass with 1" diamond wheels in a toolpost grinder I made for my lathe, works great - I put the glass in the chuck and spin it at about 50 rpm, the toolpost cutter going about 10k rpm, and use water from an eyedropper to keep the dust down and the diamonds on the tool. Then I use other cheapo diamond tools to chamfer the end the O ring is going to be inserted into, and flame anneal the whole thing, seems to make them live longer, and you don't get cuts in your O ring.

Re: Coulter Pattern HV Feedthrough

Posted: Thu Jan 10, 2013 1:13 am
by Jack Puntawong
Sorry Doug, one more thing. When it "let go", is the feedthrough going to be broken or can it be reuse afterwards ? My feedthrough is made of alumina and stainless steel. Am I right to think of a "let go" being a electrical breakdown ? I was afraid that my feedthrough will turn to something like the link below =(. ... =136&ty=33

Re: Coulter Pattern HV Feedthrough

Posted: Thu Jan 10, 2013 4:10 am
by Peter Schmelcher
Thanks for the heads up.
I vacillate between going along on your feed through ride and attempting some of the heavy lifting by evolving the design. Sometimes it’s a building day while others are reading and thinking days. I only have a 1mA 50kV supply so grid cooling is not an issue; however, I have always wanted a good application for heat pipe cooling. If I ever scrounge a high wattage supply this would be it.

I just tested lead free solder in a vacuum and it is fine. I will start a new thread about it. My thinking is that rebuilding the feed through could be un-soldering the copper tube part of the housekeeper seal.

I have an old unimat DB-200 lathe that I use for surface grinding it is not really rigid enough for turning steel let alone stainless. I clamp the unimat to the table and the work goes into a proper lathe chuck that hangs vertically from the spindle like a bat. I also made a tool post that clamps to the table, everything is very rigid, and stainless turns great. The catch is the axes are wonky. I have to think more than twice before making any cut and I always look into a hand mirror lying on the table to see the action.

Re: Coulter Pattern HV Feedthrough

Posted: Thu Jan 10, 2013 5:22 pm
by Doug Coulter
Nope, no lichtenburgs seen here. I've got pics but I'll have to go fishing through tons of them to find a good example.

What most often happens is some breakdown, probably due to excessive surface charge, on the end of the glass inside the tank, or near the end. Taken to extremes, a piece might fall off. I've had no breakdowns outside the tank or at the tubing coupler, though I did need to make a custom clamp to keep the thing from creeping into the tank over time.

Which is BTW, a "feature" rather than a bug. I find it very valuable in my setup to be able to push-pull or twist to get my grid positioned just so (it really matters how far the end of the grid is from the tank proper, it wants to be near the end of the sidearm it mostly is in). Once its right, I can shove in a shim between the stop clamp and the tubing coupler to keep it perfect.

Don't use stainless or brass - use either OFHC pure copper or 6061 aluminum for the metal parts, I prefer the latter these days for a bunch of reasons, one of which is it cuts like butter and leaves a very nice finish for the O rings to seal with. You do make the plug a bit smaller than the glass so the greater expansion with temp of the metal never causes it to break the glass, though. 20-30 mils smaller than the ID of the glass seems to be enough with 1" nominal OD glass. Anodizing the aluminum seems to make things work a lot better - my longest lasting one, still in use, is done that way with quartz-only in a 3/4" tubing coupler. The end of the quartz is getting slightly dark after many many hours of operation on that one, and that's it. The anodizing seems to help keep the exposed part of the FT from drawing too much current on its own.

Ceramics break down differently. Unless you have pure alumina (expensive and difficult to get), the binders are chemically reduced by the hot D to metals. There's no cure at that point - it's toast.
Having played with both, I like pyrex for one layer, due to its slight conductivity acting like a voltage gradient equalizer, and quartz inner if you need more insulation due to its lower dielectric constant (less energy stored in stray capacity). Ceramics lose on both counts, at least with what I've done so far.