Building an Inexpensive HV Feedthrough: The Complete Saga

For posts specifically relating to fusor design, construction, and operation.
Post Reply
User avatar
Liam David
Posts: 151
Joined: Sat Jan 25, 2014 10:30 pm
Real name: Liam David
Location: Arizona

Building an Inexpensive HV Feedthrough: The Complete Saga

Post by Liam David » Wed Jan 29, 2020 4:57 am

Obtaining a suitable HV feedthrough was one of the main challenges I faced while building my fusor. I can remember multiple instances of 30kV feedthroughs showing up on Ebay over the last ~7 years, but those moments are few and far between. This post will chronicle my attempts at building an HV feedthrough for cheap, though with all the failures and redesigns, in the end it would have been far more economical to buy one outright. The final revision has been tested to 50kV without issue and should easily withstand 70kV or more. And yes, I have simulations to back that claim. I'm putting it forward as an economical alternative to purchasing a fluted ceramic feedthrough. Hopefully those interested will be spared the expensive learning curve I endured.

I started with a spark plug, which worked just fine for the demo fusor days where voltages were always below 15kV. It was of the resistive kind and heated up quickly from the short-circuit NST current. While special spark plugs with elongated ceramic have been used to achieve fusion a couple times, notably on one of Richard Hull's earlier devices, they are certainly not recommended for anything past the demo stage.


The next feedthrough set the generic design for those to follow. A 1/4" alumina ceramic tube was passed into the chamber using a 1/4"x2.75" conflat compression adapter and the air side of the ceramic was closed with a Swagelok cap. A piece of wire connected the grid to the cap and that was it. The only other notable point is that I used an o-ring and not plastic ferrules to seal the cap to the ceramic.


Like I explained in my "Don't Build This HV Feedthrough" post several years ago, the HV arced through at a measly 20 or so kV. The ceramic wasn't nearly thick enough to withstand the needed voltage and a simple one-line calculation can show that. Per that same oversimplified one-liner, I built another feedthrough but with larger alumina tubes. A 3/4" conflat compression adapter sealed to an alumina tube which was capped with another Swagelok fitting. Inside the 1/2" ID tube were telescoped successively smaller tubes, all friction-fit together. Another piece of wire connected the grid to the cap, and that was that, so I thought. Here it is without the compression fitting.


It had a major problem, far subtler than the last. At low pressures, high voltages often do not take the geometrically shortest path to ground due to Paschen's Law. When I applied 25-30kV to the feedthrough, with or without plasma established, the current tracked from the central conductor along the insulator surface, passed through the sub-50um gap between the friction-fit alumina tubes, and shorted itself to the chamber. This is illustrated by the red path in the following image.

feedthrough track.PNG

Moreover, there appeared to be significant static buildup between the layers which would periodically discharge through the ceramic. This caused my power supply to drop the voltage to quench the apparent arc, though even with an iron-core x-ray transformer, I could not push the voltage higher. I first achieved fusion with this feedthrough, but it was clear that another revision was needed. The voltage didn't blow holes in the ceramic like in the first version, so I still use it as a secondary feedthrough on my fusor.

This time I took far more care in the feedthrough design. I wanted to replace the telescoping tubes with a single piece of material to prevent both the tracking and charge buildup. I simulated the old version in ANSYS Maxwell to fully understand the reasons for its failure, and then used the same methods to verify new design. Here is a plot of the electric field strength where red denotes it exceeds the breakdown threshold for alumina, even at just 35kV. Furthermore, the field at the triple junction points (alumina, stainless, air) was strong and helped the static discharge.

Rev. 2 Electric Field 35kV [0, 8.86e6] Alumina Dielectric Strength.png

Quartz glass has a high dielectric strength and is far cheaper than alumina. I obtained a 14" length of 25x22mm quartz for <$10 off Ebay, a far cry from the ~$70 total for the alumina tubes. Per the increased diameter, I purchased a 1" compression to 2.75" conflat adapter from BMI Surplus for $50. The diameter was reduced inside for the purpose of mounting ion gauges and other equipment, so I drilled it out with a hole saw due to a lack of lathe access. A 1" Swagelok cap (~$20) again seals the air side using an o-ring and acts as the HV connection point. Washers and a flanged nut are pinched between the end of the quartz and the cap, making an electrically secure connection to the threaded central conductor. On the grid side, the conductor is held concentric inside the quartz with a loosely-fitted ceramic bushing. To reduce the sharp edges of the thread as seen by HV, I sleeved all but the section connecting to the nut and grid itself with 1/4" OD stainless tubing. One final design choice was adding a 1.25" OD ceramic washer between the quartz and grid to eliminate the line-of-sight for ions (hydrogen) impacting the glass, which can reduce it to silicon and cause arcing. To prevent the tube from creeping into the chamber, I added a few layers of kapton tape to increase its effective OD.


Before purchasing any components, I simulated various assemblies in ANSYS to confirm that the electric fields did not come close to material breakdown. I considered using macor or boron nitride either inside the quartz or as the sole insulator, but due to their prohibitive cost and minimal benefit (and often issues with exceeding breakdown fields in these materials), I opted to pass. A couple observations based on the simulation results.

1) The vacuum gap between the conductor and quartz is about 1/4". Per Paschen's Law, this short distance will have great difficulty arcing over at high vacuum even without quartz. In all practical chamber sizes including 2.75", plasma will form long before arcs.

2) There are few surfaces across which the voltage can track (and quartz is much easier to keep clean that alumina)

3) Fields at the triple junction points are greatly reduced by the rounded metal edges and the presence of a larger gap between the quartz and compression adapter.

4) At no point does the field in the quartz come anywhere close to its breakdown strength, even at 70kV

Rev. 3 Electric Field 70kV [0, 2.5e7] Quartz Dielectric Strength.png

The overall field with a different scale is this:

Rev. 2 Refined Potential 70kV.png

After running dozens of simulations, I was convinced that the design would withstand the full voltage of my power supply (70kV) and purchased the materials. Barring the painful process of drilling and polishing the compression adapter, assembly took a couple hours over two days. The cost of the feedthrough was about $150 and could be reduced to some $120 since the McMaster ceramic washers I used are rather expensive for what they do.

So how does it hold up? Very well, actually. So far I've tested it at 50kV both with and without plasma. The limiting factors are my interconnect cables and plasma stability at those voltages. Here is the business end after some hours of plasma bombardment. One ceramic washer cracked due to the thermal expansion of the graphite grid. The tolerances have since been adjusted and the ceramic runs red-hot without issue. To give credit where it is due, the design borrows heavily from Andrew Seltzman's and Doug Coulter's feedthroughs, so I'm not claiming originality here.

Quartz still completely clean.

Sputtering apparent on ceramic, but it causes minimal issues and certainly no more than on standard alumina stalks.

Grid at some 40-50kV in a deuterium plasma.
droidcam-20200125-155835.jpg (59.92 KiB) Viewed 289 times

So what's the takeaway? Mostly for new member of the forums: First, spend some time doing the requisite research and at the very least some basic math before investing in some unproven design. The simulations demonstrated failure modes rather quickly and upended my basic calculation of breakdown simply using the alumina thickness. HV in vacuum also does not behave intuitively. Also, building a feedthrough is a viable and dare I say the preferable method forward, if not for the knowledge gained then certainly because of the reduced cost over commercial units. One could think of this post as one of many recent attempts to reduce the cost of fusor construction, among the increasing number of 2.75" conflat chambers, use of cheap precipitator supplies, and Mark Rowley's recent post about plug-and-play neutron detection with a BF3 tube and Ludlum Model 3, a setup I and others have paralleled with a 3He tube and Ludlum Model 12.

Comments and questions welcome.

Liam David

User avatar
Mark Rowley
Posts: 508
Joined: Sun Dec 29, 2002 5:20 am
Real name: Mark Rowley
Location: Sacramento California

Re: Building an Inexpensive HV Feedthrough: The Complete Saga

Post by Mark Rowley » Wed Jan 29, 2020 8:38 pm

Impressive work Liam.

There are five main factors that contribute to the failure of a fledgling Fusor enthusiast...

1) Motivation (most important)
2) Cost of obtaining a suitable power supply
3) Cost of obtaining a neutron detection system
4) Cost of obtaining a suitable high voltage feedthru
5) Cost of deuterium

There are other lesser factors as well (chamber, gas handling equip, etc) but those can now be easily purchased on the cheap.

With your work on designing an easy to build HV feedthru, new folks stand a much better chance at reaching the end goal. Add the recent successes of the $150 neutron detection system, precipitator supply, and PEM cell deuterium, there’s little doubt we will have more successful attempts at amateur fusion, and beyond.

Again, your work is much appreciated.

Mark Rowley
Last edited by Mark Rowley on Wed Jan 29, 2020 9:32 pm, edited 1 time in total.

User avatar
Richard Hull
Posts: 11810
Joined: Fri Jun 15, 2001 1:44 pm
Real name: Richard Hull

Re: Building an Inexpensive HV Feedthrough: The Complete Saga

Post by Richard Hull » Wed Jan 29, 2020 9:32 pm

This was a great pictorial followup posting on failure and success. This kind of thing is what makes a great return value post in any forum. Well done.

Richard Hull
Progress may have been a good thing once, but it just went on too long. - Yogi Berra
Fusion is the energy of the future....and it always will be
Retired now...Doing only what I want and not what I should...every day is a saturday.

User avatar
Liam David
Posts: 151
Joined: Sat Jan 25, 2014 10:30 pm
Real name: Liam David
Location: Arizona

Re: Building an Inexpensive HV Feedthrough: The Complete Saga

Post by Liam David » Thu Jan 30, 2020 5:01 am

Thanks for the compliments. Hopefully the design will get someone out of a rut.

I absolutely agree that motivation is the key here. My journey from this (late 2013, early 2014):

Yeah, don't look too close. It's pretty ugly.

to achieving fusion (summer 2019) was quite a rocky experience.

Failures included:
- Noise pickup by the neutron detection system. Heck, had it not been for this, I would likely have entered the neutron club back in 2016 (viewtopic.php?f=18&t=11221&p=73947#p73947). Statistically speaking I made a few neutrons, but nothing remotely distinguishable from the noise. Then I hit senior year in high school, moved cross-country, started college, and finally got around to reassembling everything just under a year ago.

- Turbo pump magnetic bearing cracked. My attempted repair was doomed from the start. I ended up buying a new one. The <100l/s units seem to have gotten much rarer on Ebay in the last few years (or at least the ones that aren't a metal salad).

- X-ray transformer internal arcing and rebuild (x2). I finally got it working after removing an entire secondary.

- Feedthrough redesign (x3)

- A non-working, filthy, and completely un-openable diffusion pump gave me a false sense of progress for a while.

- Power supply arc sense and over-current shutdown. I ended up reprogramming it with the help of Cliff S. and other engineers at Spellman.

- PEM cell and Hydrostik storage failure. Mostly due inexperience, and there were too many variables in play at the time.

- Obtaining deuterium. I finally found a distributor in Ohio willing to sell to an individual.

- HVAC roughing pump oil discharge into the chamber. Turns out these cheap pumps don't have an anti-suckback valve, but the one I got for $150 new still pulls below 30 microns despite 7 years of constant use.

- The viewport I cracked with an electron beam.

- Both turbos nearly crashing due to air inrush caused by a power outage. The lack of a valve in the roughing pump allowed the roughing line to hit atmosphere within seconds. A solenoid valve has alleviated most concerns.

- The lack of fine deuterium flow control with a cheap needle valve. I replaced it with a $20 mass flow controller and pressure control is much easier.

- Plasma instabilities and operational challenges.

- Finding a good table/frame to mount stuff on.

- Making space in my dad's already-crowded workshop around the central HVAC system and washer/dryer.

- Getting shocked with HV. Thankfully it was just a few kV but it sure did put a dent in my enthusiasm. Got lucky once, may never again.

- All the leaky gaskets, failed stainless solder attempts, x-rays bouncing around corners, and a thousand other minor things you never think of till they arise.

You just have to stick with it.

Post Reply