with a Master of Engineering Project
In 1997 I was finishing off my M.Eng at Simon Fraser University
and I decided to build micro vacuum tubes into a ceramic substrate
as my project/thesis. After many months of effort I sadly discovered
several major limitations with my approach. These limitations made
the project impractical. It did get me an M.Eng (here comes the
rationalization) and research is still valuable even if tells you
what don't work.
The first obstacle was getting enough energy into
the device to boil off thermionic electrons and make the vacuum
tube work. I needed to heat a small piece of BaO/SrO coated metal
over 800 centigrade to get a reasonable current flow (900 or higher
would have been better). Pumping nearly 100 watts of power into
the little piece of ceramic I could only get 760 centigrade and
that only generated a few microamps of current. It was enough to
show the thing worked but not enough to make it commercial.
The problem is that heat loss from an object (black
body radiation) is (Tobject - T
So if you want to increase the temperature of something you have
to use a lot of electrical power (or thermal insulation) to overcome
that "temperature to the fourth power" factor.
The second obstacle was my electrons kept impacting
into the ceramic walls and building up charge islands that screwed
up the amplifying characteristics of the vacuum tube.
Eureka (I think)
In the years after my M.Eng I couldn't give up on the problem
of how to create an emitter with enough electrons to make the thing work.
There is lots of research in this area using micro machined silicon
needles, carbon filaments and other approaches but none of them
solved my problem.
Finally it occurred to me that old fashioned neon tubes create
lots of electrons at room temperature using a couple hundred volts
and a few milliwatts of power. The problem is they also create lots
of Ne+ so the plasma column is (mostly) electrically neutral and
cannot be easily controlled like a vacuum tube. In some very crude
experiments I verified that there are regions in the plasma column that have
a dominant charge (both + and -) so it was possible to make a
plasma tube amplifier that works something like a vacuum tube amplifier.
In 2008 I decided I had enough information to start buying and building equipment
for the project. This is multifaceted problem with puzzles to solve in many realms.
I started with building a vacuum chamber and a system to fill it with noble gases
because I had the least confidence it could be done in a home lab. I needed to pull
a vacuum in the a milliTorr and I wanted to see what the plasma was during the
experiments (and take pictures).
That done I moved on hyperstable high voltages supplies and a means to measure
plasma properties. I'm not completely happy with the results
but I decided if I got stuck on that stage I'll never complete the project.
The next step was to build an complete package that could instrument a vacuum
chamber for controlled plasma experiments. Putting dozens of wires through the chamber
walls would never work so I needed to get a microcontroller inside. It had to be small
to fit in a small (safer) chamber. It had to be managed via a two wire
interface to minimize the number of holes. And it had to be cheap because I was likely to
fry a few with stray high voltages. Progress on this is on the
The vacuuum chamber I borrowed for my M.Eng could pull
a "hard vacuum" (less than a microTorr) but it was worth $30K used, the roughing/diffusion
pump needed its own 20A breaker, and "strip down and clean" would take a full week. Like a British
sports car it needed maintance all the time. Worst of all it was metal can so I
couldn't see inside and metal walls and high voltages don't get along.
For my M.Eng I needed a good pump because vacuum tubes need hard vacuums.
To get that you need an oil filled diffusion pump backed
(in series with) a good roughing pump and a liquid nitrogen (or
helium) cold trap to capture the last molecules of air. A little
dirt on the walls, a little contamination of the oil, a tiny problem
in a seal and you get a crappy vacuum.
For plasmas (like in a neon sign) you need a few hundred milliTorr. You can find
plans to convert compressors from old refrigerators or air conditioners
into pumps but my research suggested this would not give me a reliable
The good news is there are pumps costing a few hundred dollars used
to repair automobile air conditioners that will get down below 50
milliTorr. I used one of these and you can read more on my pump
Vacuum chambers for high vacuum (sub-microTorr) experiments cost
many thousands of dollars and require careful maintenance. I wanted to see the way
the plasma behaved in the experiments but good bell jar will set you back $1000, you
can't drill holes in it and glass has a habit of imploding (glass handgrenade) if you
It turns out you can build an excellent chambers out
of plumbing supplies for a couple of bucks that will easily operate
well below into the milliTorr range needed for plasma experiments. See
my page on chamber
construction for details.
You can form plasmas and arcs in any gas but the process tears apart
gas molecules and creates highly reactive ions. Certain plasma cutting
processes use these reactive ions to cut materials like ceramics
that cannot be cut any other way.
For my project I do not want my ionized gas to corrode
my electronic device so I am forced to use noble gases like: Helium
(He), Neon (Ne), Argon (Ar), or Krypton (Kr). These gases are all
available at local industrial and scientific suppliers but I chose
to use helium. See my page on
noble gas supplies for details.
The striking voltage for a plasma is given by Paschen's Law which
is dependant on the gas type, the gas pressure, and the gap distance.
Generally you will need several variable voltage high voltage power
supplies for your experiments. If the electrode gap is small (like
a millimeter) the striking voltage is under 200 volts but as you
increase the gap the voltage increases so standard neon signs need
20,000 to 40,000 volts to operate.
In my project the electrode gaps will typically be
a few millimeters but for special configurations like Pirani gauge
calibration the distance could be a centimeter. To meet these
requirements I have constructed several variable voltage power supplies
that have maximum ranges of 300, 600 and 1200 volts. For designs
and operation go to my power
A note on high voltages: 50 ma on a path that crosses
your heart (like hand to hand or right hand to left foot) will kill
most humans. All you need is 200 volts to overcome your skin resistance
and you will get that 50 ma easily. These power supplies will kill
you quick if you are not extremely careful. This message will be
repeated on the power supplies page.
The development of plasma amplifiers requires careful measurement
and control of gas pressure. Commercial Pirani gauges are very good
at measuring pressure in the region used for the plasmas in this
project but at $300 for the sensor and $2000 for the meter/control
circuit I decided to try and make my own. There are a few internet
sites that describe making Pirani gauges but I was not convinced
they would be as reliable as I needed so I decided to do some analysis
on the sensor selection from basic physics and design a my own measurement
On my Vacuum Gauges page I take you
through the process of selecting ordinary
low power incandescent lamps and calibrating them using my design
of a measurement circuit. The sensor/circuit combination can be
build in a few hours for a few dollars. The resulting gauge is close
to the results I would expect from a commercial unit but better
aligned to the budget and self reliant nature of the hobbyist/inventor.
The page documents the tests that were used to verify understanding
that is not directly part of the plasma amplifier project. For example
it includes the experimental design and results for the tests required
the calibrate the Pirani gauge. It also includes tests to confirm
the accuracy of the predictions from the theory and simulations.
As I encounter other interesting phenomenon I will also park results
in Plasma Tests.
The project is primarily a hardware application but (shudder) I
may actually have to learn some theory to make the thing work or
explain to others how it works. I will use the
Theory path to capture
my analysis and explanations. The topics may be interesting to visitors
but its primary goal is to explain things to myself six months or
six years from now after I have cleared my medium term memory so
I can learn other things.
The Plasma Theory and Simulation Group at UC Berkeley developed
a plasma simulation package called XOOPIC
in 1997. This only runs on Linux but has been converted to Windows
and commercialized by X-Tech. This
uses are "Particle In Cell" strategy to simulate plasmas and I will
be attempting to use the free version in parallel with my lab work
to make sure I can explain and predict the behavior of my plasma
Nvidia has developed a remarkable series of video
cards and a software library (called CUDA)
that allows general purpose problems to be run on the cards as if
they were massively parallel processor arrays. Their 8800, 9800
and 280 series cards can process an incredible 500 gigaFlops each
when properly loaded and up to three cards can be run in a single
standard personal computer. Since the cards need over 300 watts
each the computer is not exactly standard ... it would be
more like a high end gamer's water cooled PC but it is still in
the price range ($4000) of a power user.
Lower end Nvidia cards can also run CUDA and I plan
to convert OOPIC to run inside a GeForce 8400 ($60) and if it works
find someone to fund a larger (perhaps an Nvidia
You can follow my efforts on simulations and hardware
acceleration on my Simulators page.