|
Ken interviewed me for a job on a new team he was
assembling at Bell Laboratories in Murray Hill N.J. in September 1964. Our
department head was Bob Ryder. Our task was to design
“Beam-lead Planar Tunnel Diodes.” The purpose of the Tunnel
Diodes was to act as a very low noise pre-amplifier in the microwave
towers. The first amplifying device the faint signal sees in the horn must
be low noise. It is not important to have high amplification. The devices
had to be small to handle the high frequency and the goal was a small
square of Germanium 5 mils by 5 mils (0.005 inch.) and 1.5 mil thick. The
beams would stick out 10 mils on each end and were 3 mils wide. The total
length of the device with two beams was 25 mils. For size comparison a
three by five card is 8 mils thick and the thickness of three cards
equaled the length of the diodes. The were barely visible with the naked
eye.
I was then working for Microstate in Murray Hill
building and calibrating test instruments used to measure tunnel diodes.
At Microstate each diode was hand made and chemically etched in KOH to get
the right characteristics. They were expensive, up to $500 for a matched
pair of diodes. We were to make them smaller, with many units on a single
slice. Bell could then produce them cheaply in large quantities.
During the interview Ken said: “I’ll hire you,
but with your skills you could be kingpin in a small company. Why do you
want to be a small fish in a big pond?” I had worked for two small
companies and they didn’t pay that well, had no pension or medical
insurance plans. Characteristically for Ken he asked me a probing
question: “What would you do if you had an idea and I told you it
wouldn’t work, so forget it?”
I thought carefully: “ If I thought it would work,
I would pursue it at lunch, after hours and any extra time I might have
during working hours.”
The team was made up of Ron Davis, a Canadian
Physicist and PH. D., George Gibson, a Scot working on his PH. D., Steve
Kovel, a technician from The Bronx in New York and me, a Danish
technician.
Tunnel diodes were originally called Esaki diodes.
Professor Esaki in Japan had discovered the tunnel effect. For some reason
the US didn’t like the name. Curiously, when Herman Gummel, a German who
was my department head in the 1970’s had discovered the “early
voltage” for transistors, Bell Labs named it “the Gummel number” to
Herman’s embarrassment.
Tunnel diodes were made on Germanium crystal slices
one centimeter square and 15 mils thick. So they were very fragile to
handle. The tunneling effect is achieved by doping the Germanium very
heavily with arsenic. The heavy doping changed the characteristics of the
diodes. As the voltage across the diode was increased the current would
increase also, until it reached a peak and then the tunneling effect would
decrease the current back down towards zero. Before it reached zero it
turned around again and went up. The down slope is called “the negative
resistance region” and this makes the diode capable of amplifying very
small signals. The peak to valley ratio had to be about 10 to 1 for the
diode to amplify sufficiently.
Martin Lepselter in Murray Hill had just invented the
beam-lead technology. Once the devices had been made in a pattern on the
slice several metals were deposited in turn connecting to the devices. The
last step in the process was etching the slice so only a small amount of
Germanium was left and the “beams” stuck out of the sides. The top
layer was gold and enabled the device to be mounted face down and bonded
to a circuit board. Marty later became laboratory director in
Allentown’s semiconductor fabrication line, the one Ken had helped set
up earlier.
There were many problems to overcome. Germanium
crystals can be cut on three axis, 111, 110 and 100 and we had to try all
three. Only one axis worked. The arsenic had to be diffused into the
germanium and we tried several interesting ways. This was before diffusion
ovens were commonly in use in use. At one time we used timed Helium
flashes. Once the devices were etched apart the problem was to handle and
measure them. I used bamboo chopsticks scraped down to a single strand. By
touching my hair I could generate enough static electricity to make a
diode or two stick to the end of the chopstick. This was one of Ken’s
suggestions.
We drew a glass pipette down to a thin tube, polished
the end flat and could pick up a diode by sucking on a rubber tube
connected to the pipette. To measure the diode it had to be held down
against a test jig deposited on a glass slide and it had to be oriented
correctly. I designed and built a test rig where I could handle and
manipulate the diodes.
There were two ways to go. One was to design and draw
the whole thing up on multiple drawings and submit it to Ken and Bob. They
would then send it to the machine shop that would make a bid for the work
and fix a time for completion. This could take weeks. I opted to design it
so I could hand build it in our lab using a vice and hand tools. This
earned me the nickname “The Danish milling machine” from Ken, but it
was easy to see he was pleased. All told I measured 2500 individual tiny
diodes over the years.
One problem with research is time. It took three
weeks to finish a test run from bare slice to finished diodes and we
didn’t know if the new devices would work until we measured one at the
end of three weeks. We found that capacitance was a measure of how the
peak to valley ratio would turn out. A standard capacitance bridge from
Hewlett-Packard would not work because the capacitance was only 1 pF (picofarad),
a very small amount, and the cables to the bridge added much more
capacitance than that.
I used small pieces of printed circuit board cut into
little squares to make 1,2 and 3 pF capacitors. These I could measure
accurately on a Hewlett-Packard Bridge. A 30 Megahertz generator provided
a signal and a tiny laboratory jack held a sharpened tungsten wire. By
poking the tungsten wired down in the diode I sent a signal into it. The
bottom of the slice was the other side of the capacitor and was connected
to a meter. The output was related to the diodes capacitance. The small
chips were used to calibrate the meter. Probing a number of diodes across
the slice gave us an accurate measure of capacitance and doping
uniformity. The more capacitance, the higher the reading was on the meter.
This test told us early in the cycle if a particular slice was worth
processing all the way to the end or should be abandoned. This saved us
many weeks of time.
Ken and Bob respected our ideas and encouraged us to
try new ways. Never before or since have I worked in such a pleasant
environment and it was due to Ken and Bob’s light way of management.
Everybody could suggest solutions and try them.
After four and a half years we could produce the
diodes reliably with good yields. Only one thing remained to be tested:
How did the diodes react to temperature changes. Microwave towers stretch
from hot deserts to icy mountains and the diodes were meant to be
installed near the horns. Sad to say, the characteristics changed
drastically with heat and cold. The diodes wouldn’t work.
The next morning Ken walked in and said: “Finish
the technical memoranda’s and clean off the tables. This project is
closed!” I couldn’t
believe my ears? Four and a half years work for four people and all was a
waste?
But it wasn’t a waste; we had discovered a number
of things during this time that would help semiconductor research later in
other ways.
I was 27 when Ken hired me and had worked in the US
for only two years. Ken was like a father to me, guiding me and pointing
gently to behavior that would be detrimental to me. He also got me
promoted in less than two years. Bob Ryder taught me to play GO, the
Japanese board game. Bob was a master at the game. At lunch I often went
to the game room and found Bob playing Sam (Hin Shu) Poon, a Chinese
scientist. Bob told me to go get a small board and play him on the side. I
would be thinking hard and he would glance up from his bigger game now and
then and look at my board. When he noticed I had made a move he would
quickly put a stone down on my board and go back to the big game. Bob was
so expert that I never managed to win a single game against him.
I worked directly for Ron Davis and shared a
laboratory with him. Ron was ten years older than I and treated me like he
was an older, wiser brother. His English and his manners were flawless and
I owe my mastery of the language to Ron who gently corrected my use of the
language.
Bell labs was a marvelous place to work. One day we
needed to know something about epoxies. We went to “Mister Epoxy”
whose name was Fred Keimel. In a large laboratory he had samples of every
make of epoxy on the market and he knew the characteristics of each.
In 1969 our laboratory director Willard Boyle went to
lunch with George Smith, one of our department heads and while they were
eating they invented a new device. They called it “charge coupled
devices” and applied for a patent. The patent was granted in 1972. The
problem was that nobody could make it at the time, the yield was way too
low. A charge coupled device is a long string of semiconductors that pass
bits of information in a chain from one to the next. This can be used to
represent a string of pixels like the pixels in a picture on TV. It
wasn’t until the 1980’s that the yields were good enough to produce
perfect slices and today every digital camera and video camera use this
technology.
I have included some pictures of tunnel diodes and
the test gear. Also included is a timeline describing the steps needed to
produce a tunnel diode.
I hope you can use some of this. You made delete what
isn’t of interest. I would appreciate a copy of what you chose to
include.
Best of luck,
Per Biorn
|
