Oral-History:Paul K. Weimer

About Paul K. Weimer

Paul Weimer was educated in the Midwestern part of the United States in the late 1930s and early 1940s. He was recruited as a graduate student by Radio Corporation of America and employed as a vacuum tube researcher. His early work on the famous Image Orthicon television camera tube led to military experiments on guided bombs employing television technology. After 1945, he continued to improve the Image Orthicon, and it became one of the most commercially successful camera tubes. By the later 1950s, Weimer and his colleagues had shifted their attention to semiconductor devices, and eventually produced a solid-state camera tube employing thin-film technology.

Weimer discusses the various technologies, experimental and commercial, used for television work. He shows how television camera technology interacted with and contributed to the technology of semiconductors and integrated circuits. He concludes the interview with a discussion of the commercialization of this technology and its impact on RCA and the television industry.

About the Interview

PAUL K. WEIMER: An Interview Conducted by Mark Heyer/Al Pinsky, IEEE History Center, July 8, 1975

Interview # 022 for the IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.

Copyright Statement

This manuscript is being made available for research purposes only. All literary rights in the manuscript, including the right to publish, are reserved to the IEEE History Center. No part of the manuscript may be quoted for publication without the written permission of the Director of IEEE History Center.

Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, 39 Union Street, New Brunswick, NJ 08901-8538 USA. It should include identification of the specific passages to be quoted, anticipated use of the passages, and identification of the user.

It is recommended that this oral history be cited as follows:

Paul K. Weimer, an oral history conducted in 1975 by Mark Heyer/Al Pinsky, IEEE History Center, New Brunswick, NJ, USA.

Interview

Interview: Paul K. Weimer

Interviewer: Mark Heyer/Al Pinsky

Date: July 8, 1975

Education

Weimer:

I graduated from a teacher's college out in Indiana, by the name of Manchester College in 1936. The Depression was still going on, at that time I was aiming to teach physics and math in high school in Indiana. But I drove around all over the state, looking for a job and I didn't find one. I heard of an opportunity to get into the physics department at the University of Kansas. I got a graduate assistantship out there, and I went out there to start graduate work. Of course, it was one of the luckiest breaks of my life that I didn't get into high school teaching because I would have probably stayed with it. I enjoyed graduate school and I got a masters degree the first year out there. I was still in debt for some of my undergraduate, so I took two years off to do some teaching. I taught in a little old junior college, out in the middle part of Kansas, called Taber College. I taught physics, math, and chemistry and directed the orchestra because it was that small of a place. I enjoyed that and I really planned at that point to make college teaching as my career. Of course, I wanted to do it at higher level so I got an assistantship at Ohio State in 1939 and I was there for three years then and I got a Ph.D. in physics in 1942. Of course, the war was on at that time and it seemed like a good idea to get into industry instead of looking for college teaching. I heard of RCA through Charles Burril, he was around making callings at Ohio State and he picked up several of us fellows who were contemporaries there: Harold Law, Stan Thorpe and Guy Woodward were all at Ohio State at the same time I was. So, we all came out here. I had had all physics, and never had a course in engineering, so I was a little bit concerned about coming to work for RCA without any engineering background. I landed down in Camden, where the labs were just being constructed at that time. So 1942 I was in Camden for about four or five months and then the labs were finished and we all were moved up here.

Arrival at RCA

Heyer:

When you all came down here, you didn't have jobs or job offers?

Weimer:

No, we were all made offers at Ohio State. They were good offers at the time, although when I look back now, it wouldn't look so good. I recall I started out at $225 a month plus overtime that was with a Ph.D. in physics. The overtime was for Saturday and it was a good enough salary, I was still a little bit in debt for school, but I was pretty close to being even. On the basis of the job I got married with practically no money in the bank.

Heyer:

So in 1942, you joined RCA, and you were working in the Camden labs. What were the big projects, how did you get into that?

Weimer:

I was hired to come work in Dr. Zworykin’s group at Camden under George Morton. So I started out working in some early forms of camera tubes there. And then we were transferred up here, of course the Harrison people brought together with the Camden people, a new group was formed with Dr. Al Rose as head of the camera tube work. So Harold Law and I were both assigned to work with Al Rose. I always considered that one of the lucky breaks in my life because Al Rose was one of the [inaudible passage]

Image Orthicon

Heyer:

Was he already at RCA?

Weimer:

He was already famous, he had been at RCA since about 1935, I guess. He had invented the Orthicon, which was a great improvement over Zworykin's original Iconoscope. And he had invented the basics of the Image Orthicon and then we were brought together to develop this tube. My particular assignment first was to develop an electron multiplier to go with the Image Orthicon. Harold Law's assignment was to develop a way of making the thin glass target that went with it. The multiplier that we came up with was the so-called pinwheel multiplier and it was a thing that I made out of thin silver magnesium metal and I remember cutting out the pinwheels with scissors myself and bending them up. We tried it and it wasn't the first thing we tried. We did try and it worked very well, so actually it is still used in the new Orthicons that are still sold. Although the original Image Orthicon itself has now been superseded by other things.

The second assignment I had in connection with the Image Orthicon was to study the electron optics of the tube and find out why it was that in this early tube we were getting poor landing of the electron beam at the edge of the target. In other words when we look at the transmitted picture it would look as though you were kind of looking through a porthole with the picture being good only in the center and then shading off into the dark regions around the corners.

Heyer:

So it's like looking through the early optical system of the whole lens?

Weimer:

Right. And of course the reason for that was that the low velocity beam was receiving transverse velocity somewhere along in its path from the gun to the target. In the Orthicon and in the Image Orthicon, the beam is sped up when it leaves the gun and then it slows down when it reaches the target. And if it does not approach the target exactly normally it's not going to have enough energy to reach the target. It's going to turn around before it gets there. So the objective was to see what can we do to make that beam have its energy directed in any other direction than directly normally into the target. So we found that we could modify the electron lens in front of the target, we could modify the magnetic field that deflection coil produced. We found that both of these effects produced transverse energy, in other words the electrons would start in large helical motion, which would prevent them from landing. By then we were able to balance one effect against the other, and we could make it land very nicely. So that was the second thing. We worked on the Image Orthicon throughout the entire war period. In the latter part of the war period we were working on the smaller more compact forms of Image Orthicons which were to be used in bombs for guiding the bomb as it was dropped from an airplane. I don't know whether any of these were dropped on the enemy, but we did see some movies where they were dropped on targets and we could see the target growing bigger and bigger and you could see the corrections being made and then it would be over and that would be the end of it.

Heyer:

That's coming back now in cruise missiles I guess. So this development was going on during the war. What was seen as the importance of the size of the bomb?

Weimer:

Well of course, television was considered useful to the military as a form of recognizance. I mean just for looking at the ground and televising a picture of the ground and so it was considered important, or at least it was important enough that Harold Law and I always got our deferments every six months. The project was continued and felt we were making some contribution and in hindsight I don't know how much it was.

Heyer:

Was the Image Orthicon used in commercial applications after the war?

Weimer:

As soon as the war was over, the Image Orthicon had been fully developed to the point and it had been picked up at Lancaster where it was commercialized. When broadcast television was revived in a big way after the war, well then the Image Orthicon became the workhorse of television. I recall RCA had a big press conference when they announced the Image Orthicon. They had Ben Grower there, who was the top NBC announcer at that time, and I remember him saying that the Image Orthicon was the atomic bomb of television. I think that was a typical saying at the time. Then they did the experiment of lighting a match to show how sensitive it was. It was very much more sensitive than any previous camera tube. It had a lot of good features, but of course it wasn't the last word.

Heyer:

So then it was put into use as a production item for commercial television?

Weimer:

Yes, it was used in practically all television up until the 1960s, very nearly twenty years. It was the broadcast tube.

Image Isocon

Heyer:

It was during that time that you were moving onto other projects?

Weimer:

Well we were moving on to other things; one of the first things that I worked on after the Image Orthicon was to develop some kind of a low-noise tube. Although the Image Orthicon was very sensitive, it gave a very noisy picture and part of the reason for that was that the signal was taken from the return beam. And we used the total return beam, which was not very well modulated by the target, it was rather poorly modulated by the target and it happened that it had the most electrons on the return beam it the dark areas of the picture because that was the areas where the beam wasn't needed to discharge the target.

Heyer:

So the signal was inverted.

Weimer:

The signal was inverted, so I was asked to see what we could do about that. Well to invert the polarity of a signal, usually means a whole new device because it has to work by an entirely different principle. We kind of stumbled onto an approach, which was resulted in a tube called an Image Isocon. From the outside it looks almost like the Image Orthicon, the photo cathode and the target were all the same, but what it amounted to was instead of using the polar return beam and letting that polar return beam go into the multiplier, we found that one could distinguish the various kinds of electrons in that return beam. There were two kinds: there was the reflective portion of the return beam and there was the scattered portion. If you build your electron optics so that you admitted only the scattered portion into the multiplier, you would get the other polarity signal and very much better signal to noise ratio. Because you see the scattered electrons were only produced if the beam touched the target. If the beam didn't reach the target they would all come back as reflected and we would not let any of that fraction into the electron multiplier.

Speaker 2:

You said that you stumbled onto that?

Weimer:

Well I stumbled in the sense of it would have never occurred to me that this kind of thing would be possible. I was actually looking for other ties. I was building modified electron guns and multipliers, thinking about trying to do some velocity selection in the return beam, quite a different approach. And I found that when I built these special structures they were along the edges at certain points where the return beam would strike the edge of this structure. I noticed that I was beginning to get an inversion of the polarity of the signal. And that might be occurring not over the whole picture but only in little piece of the picture, off in one corner. At first we didn't know what that was. We gradually developed the theories to what was going on and then we developed the electron optical system for separating out those electrons. Actually it is a development that I am quite proud of, but I don't take credit for theorizing the principle to begin with. I theorized what I had after I saw it and then of course after you saw it in a little piece you had to develop electron optics to make-work. Now the Isocon was rather critical to operate, but we did demonstrate its operation and got good quality pictures, which operated the whole area of the target. It was operated here by Al Schroder who experimented with [inaudible] and it showed that it indeed gives lower noise in the dark portions of the picture.

So it looked as though we had something. Lancaster picked it up gradually but it turned out to be rather difficult to operate, to get everything working just right, and so it didn't sweep them off their feet. But the need still existed ten or fifteen years later. By that time advances had been made in electron optics and Lancaster was able to improve the electron optics, so it became much simpler to operate. Although it was operating on exactly the same principles before, it was just developed to the point where it could be operated more easily and more reproducibly. So the Image Isocon is now sold as one of the types of very low-light tubes. And it has certain advantages of some other low-light level tubes, although it is one of a group now. There have been other approaches, which have worked at very low-light levels. By very low-light level, I mean tubes that can look at starlight, or look at the night scene with no moon, just starlight. Now the Image Isocon, to perform at the level, requires an image intensifier be put in front of it, but all of the ultra-sensitive tubes require an image intensifier. Now the intensifier tube for sensitive very low light level tubes is called the Intensifier Isocon. But the Isocon got its sensitivity by this trick of producing the video signal in a very efficient fashion. Some of the other tubes get their sensitivity by means of extremely high-gain before you stored the target, so you have an awful a lot of gain then it doesn't matter much how you discharge it, that's another approach.

Heyer:

What kind of general use do those tubes see?

Weimer:

Well the Isocon has never been a big moneymaker for RCA, because it's a specialized type of application. It's for scientific applications, astronomers have used them and the military has made numerous tests of the intensifier Isocon versus other kinds of very low-light level tubes. They still have a great need for television cameras that operate at very low light levels. It wasn't intended to be the "end all" of tubes.

Selenium Photo-Conductors

Weimer:

So soon after developing the Isocon, a bunch of us started working on trying to develop a photoconductive camera tube. There appeared to be a possibility of building a photoconductive tube that might be a more simple tube than the Isocon and Image Orthicon, which were relatively complex and expensive tubes. Stan Thorpe, Bob Goodrich and I were all working on examining various kinds of photoconductors to see if we could build a photoconductive tube. The concept of the photoconductive tube was not new, people have tried to build photoconductive tubes in the past no one had ever gotten one that had produced a decent television picture like what we could get with the photoemissive tubes. So the first type of photoconductive material that we found that would work in a camera tube was amorphous selenium, which as kind of interesting in that we had not been working with the amorphous selenium, we had been working with zinc selenide. Evaporating zinc selenide which was known to be a photoconductor but of course to use it in a camera tube it has to have special properties, it has to have very high resistively and of course you have to be able to put it down very thin and very uniform. And so we got a zinc selenide that worked a little better than some of the previous things. We got to feeling that zinc selenide might be dissociating when we evaporated it. When we put the material in a heated boat and heated it up in a vacuum and it spews off in the vacuum and it forms a coating on the target. So we thought one of those constituents of zinc selenide ought to be selenium, so we ought to try selenium.

However, when you evaporate selenium it goes down in a form that is called an amorphous form, it's not crystalline. Now crystalline selenium was well known as a photoconductor, it was the photoconductor used in the early days. But an amorphous selenium according to the texts was not photosensitive, it was an insulator. But it turned out that when we evaporated it and got the amorphous form and put it on a transparent conducting signal plate it gave a beautiful television picture just by scanning it at low velocity. In other words, the low velocity method of scanning was a more sensitive way of detecting photoconductivity than the previous physical measurements that had been used in looking for it. So we thought at that time, gee we have discovered a whole new photoconductor. But about that same time, the people at Batelle were doing the very early work on the xerography process. And it turned out that they were evaporating selenium and also using amorphous selenium. Although at the time we discovered it and we didn't know about their work and at the time they discovered it, they didn't know about our work. I think in actual timing they probably did it a few months before we did. We published it thinking that it was the beginning of years and years of new photo conducting technology, but we realized afterwards that they had also discovered that it was a photoconductor. Well that didn't really happen so much because it did prevent us from building a new type of television and camera tube, which was our objective. However, the first Vidicon that was demonstrated at the IEEE had amorphous selenium as the photoconductor.

Heyer:

When was that?

Weimer:

When was the demonstration?

Heyer:

Yeah.

Weimer:

Without looking it up, I would guess it was 1949. That sort of comes to my mind. But one of the problems with the amorphous selenium was that it is in a metastable state that if it is heated up very much it gradually converts to the crystalline form. And when the conversion occurred in the camera tube it would tend to become more conducting in these spots that were converted. So you would start out with a beautiful picture, but it would gradually develop stars in it. The stars would get bigger and bigger and so after a hundred and fifty hours of operating; why it was no longer useful. So we really needed a better material then the amorphous selenium. Stan Thorpe was working with anomaly sulfide and he found that anomaly sulfide, if it evaporated in a poor vacuum it made a good Vidicon target also, and it was longer lodged. So, the commercial forms of the Vidicon that came out used [inaudible] sulfide. The porous form of [inaudible] sulfide is still used in many of the standard Vidicons that are sold.

Heyer:

Are those carbon materials released if you focus an intense light like at the sun is it possible, I imagine the selenium would?

Weimer:

Well with selenium that is certainly true. You could burn it right there. With the [inaudible] sulfide if you focused light from the sun, why yes you could ruin it. So it was not perfect in that respect, but it was very much better than selenium. And it is normally considered a relatively good target. But it could be damaged if a person operating it made a mistake or if, say a scanning failed so that the device put the whole beam in just one line. That could burn a line into it. And so for many applications that was available and that was one of the incentives for developing the silicate Vidicon. Because the Silicate Vidicon used an element, silicon, which was a single crystal element and was very much more rugged than the two-six conductor.

Silicon Vidicon

Heyer:

You worked on the development of the Silicon Vidicon tube?

Weimer:

No, I was not working on the Silicon Vidicon. I got into other directions of work. The Silicon Vidicon was first developed at Bell Laboratories by a man named Crowl. Under Gene Gordon, he developed the first silicon diode target. But it looked so good and interesting that RCA soon picked it up, but it was not so picked up so much here at the laboratory. It was picked up in our development shop at Lancaster and they developed the Silicon Vidicon further to the point where RCA now sells a very good Silicon Vidicon that is better than anyone else's, and certainly better than the experimental ones.

Heyer:

There seems to be two lines of work in video tubes. One is the mechanical development of the device and the electron beam work, and other is the material. Both types of research are done here?

Weimer:

Yes, they are both done here. Sometimes researchers may be involved in only one aspect of it almost entirely but sometimes they may be involved with another aspect of it too. The Isocon work was electron optics and electron beams. The Vidicon work was mainly the development of suitable target material. We knew how to build the electron beam and it didn't take us long at all when we had a proper material to put it into a tube, which we used low ballistic scanning of course because that was well known at that time. So very quickly we built the Vidicon gun, and it looked very much like the guns that are sold now. But the critical point in the Vidicon was the fulcrum of the material. There has been lots of work on new photoconductive materials for the Vidicon that has gone on even until this day.

Forty-Five Degree Deflection Tube

Heyer:

To just sum up a little bit to this point. You were involved, primarily with development of the mechanics of the electron beam and scanning and so on. I see that in 1959, according to your biography, you went to France study for a year.

Weimer:

Yes.

Heyer:

It looks like later you came back and got into semiconductors?

Weimer:

Well, that's true. That opportunity to go to France occurred when RCA awarded me the European Fellowship which was a very nice thing to have, and which has been given to a number of people here. We could choose the place that we wanted to go and the field that we wanted to study and we were paid enough that we could take our families and enjoy ourselves for a year. So I felt semiconductors were a coming thing and I had an opportunity under Pierre Grant at the University of Paris, former Mount Superior. I chose to go there.

However, I was involved in some other things in the 1950s prior to this trip to Europe, if I could tell you a bit about that. While the Vidicon was still a very hot project there was an even a hotter project that came along which was very important to RCA which was to develop a color Kinescope. Sarnoff asked the laboratories to make a crash program to develop a color Kinescope. Now Harold Law was already started on the shadow mask approach before this crash program came along and as it turned out the shadow mask did turn out to be the best approach. It is used in the color Kinescopes today. It wasn't that clear at that time that this was going to be the approach. Many of us were asked to just rack our brains and find every kind of approach that you can think of, and if it looked good we would get some people together to work on. So of course right then the Vidicon was hot and it was with mixed feelings that I got started on something else, but this was a hot subject too.

We did come up with a different approach, which we called the forty-five degree deflection tube. The shadow mask tube was a three gun approach, it had problems of registration, of making these three beams track each other. So we thought that we could do it with a single beam. By applying a voltage to the target, which would swing the beam from one color to another. So we tried it out. We built a screen that was this big and with the help of Matt Rin who was a circuit man, we fixed this in demountable systems and we scanned it. The way it worked was that the beam came up at an angle through slots in a metal mesh on which the phosphor was, and turned around to hit the rows of phosphors. It gave a very nice picture in that size dimension.

Heyer:

What size is this? Six inches?

Weimer:

It was about a seven inch diameter screen, and so we were very happy with our results. The shadow mask was also getting the big push and it really blossomed. So all the other approaches sort of fell back in favor of the shadow mask and we got color pictures and roll of paper and that ran it out. That gave us about a six month interlude.

Color TV

Heyer:

That was an interesting time in the 1950s. Suddenly there was the decision that color was the thing to have.

Weimer:

Well then after that, color of course being the center of everyone's attention, and then it seemed reasonable that I should work on a single tube pick up which would give a color signal. The color cameras at that time still consisted of three cameras, of three tubes. One for each of the primary colors. So we developed a very early tri-color Vidicon. It would give the three color signals. It had three output lead one corresponding each of the primary colors and it didn't have quite as much resolution as one could get from three well-registered tubes but they were also in register and you couldn't throw them out by defocusing. But that had some significance to me personally because in order to do that I had to develop techniques of producing very fine patterns. These involved making colored filters, which were going to be put right in the tube we made. We made red, green and blue color filters. And the filters were relatively tiny, they were much finer than what people were doing at that time. It was as fine as what people do now in integrated circuits. The center-to-center spacing of these colors was about six-tenths of a millimeter.

So we had filters to produce in those dimensions, and we had the transparent strips to produce in even finer dimensions, and they had to be insulated from each other. There were three sets of signal plates, which were registered with these color filters, and they had to be connected in all three of them. Two would have been easy because you could have done it like this, but with three of them you had to have lots of crossovers. So we had to develop integrated circuit techniques to do this. We did that successfully but not enough to really make a commercial tri-color Vidicon. It never made the grade, the three-tube cameras were able to get a picture if one was willing to spend the time to get them tuned up. After all what you pay a cameraman is nothing compared to what you are paying the performer. So the need for the signal tube was not nearly as great we thought it might be. So that never became a product but it did enable us to develop techniques, which were very useful in the work that we did on semiconductors when we came back.

Educational Technology

Heyer:

You came back in and started in thin-film devices?

Weimer:

Yes, let me tell you just one other thing that I did in the 1950s and is kind of interesting to me. I haven't thought about it in a long time. I had an idea around the middle 1950s which sort of dated back to my training in teacher's college. I thought it would be nice if one could build a teaching device which would have feedback which is now a very old principle, it's called programmed learning but it was new at that time. So, I wrote a paper, which was published in the IEEE educational journal and was so well thought of that it was reprinted in a book that somebody got out. From the teaching point of view it was an extremely exciting thing. I got all excited about it. Actually that was another thing that I invented. I think that the person really invented this ahead of me was B.F. Skinner, who was one of the top psychologists. But I didn't know about his work until after I had written mine write up and so it was kind off the subject but one that Ed Harold who was director of our laboratory at that time urged that I write it up as a paper and send it in and so I did. It had this much of an impact. It was mentioned various times in the early days.

Heyer:

I was just talking to Ed this morning and he was talking about teaching machines. In the 1950s he was interested in but nothing much ever came of it. He never got very far.

Weimer:

Ed retained an interest in it much longer than I did. I was interested in it. I had this idea of carrying out the idea in terms of boards that would light up. The student would push a button and get a reward if it was right and no reward if it was wrong. Then the next question would be based on how the student had answered the previous questions. So there would be continuous feedback. If you built a program into the machine, you could make the machine really adapt itself to the student. It's an interesting idea but that was a side issue.

Thin-Film Semiconductor Devices

Weimer:

<flashmp3>022 - weimer - clip 1.mp3</flashmp3>

When I came back from Europe I actually wanted to do something in solid-state. I was involved in work at that time in trying to build silicon transistors. Wallmark had thoughts about putting together integrated circuits using silicon. But in competition with the idea of integrating circuits using silicon there was a thought that one might be able to do it with thin films. Everyone felt that if you could build a device by thin films the chances are you could make very complex patterns and build them cheaper than with silicon. You could make thin film capacitors and thin film resistors, but the missing link in that thin film integrated circuit work was the transistor. Since I had this experience with fine patterns in building this tri-color Vidicon it seemed a natural for me to look into the possibility in building a thin film transistor. There was already some work on a film transistor going on. Jim Dresner was doing some work. He was thinking somewhat along the lines of another man by the name of G.T. Wright, who had written it up and gotten an awful lot of publicity. He was going to build a solid-state transistor based on space charge limited currents. Sort of like the analog of the tube. He would have an emitter, he'd embed a grid in it, and then he would have a plate up on top. Those were the film transistors that were being worked on here. Joe had built something like that.

But it seemed to me that our fine pattern work lent itself to a much different type of geometry: a planar geometry. So we built a thin film transistor, which had a source and an insulated gate and then a drain. Well that's another example of where accident helps you because the first ones that we built we didn't insulate the gate. We had just built it with a block in contact and then we realized that when we were just beginning to get evidence of transistor-like characteristics that actually we had an insulator in there. So we realized that's what we should be doing. We moved immediately in trying to build an insulated gate thin-film transistor and the one we built we gave an article on it in 1961. It attracted a lot of attention and we wrote the paper on it in 1962. It was the first thin film transistor that gave nice pentode-like characteristics, nicely saturated characteristics. And that paper was reprinted in one of these books that are just collections of papers. And he asked me if he could use my paper in it and I was very happy for the company that it had because he was quoting Shockley a lot in it. Mine is last in the book but I was glad to be in it.

Heyer:

Now what is the difference between a thin-film circuit and a silicon circuit?

Weimer:

The silicon device which is most like the insulated gate in thin-film transistor is the MOS transistor. As it happened it was basically the same thing, it was a source and it had an insulated gate and a drain. As it happened we got ours to work before they did because we got our characteristics. I remember Steven Hoffstein saying to me that that's exactly what we are looking for in silicon. Well they eventually found it, and Hoffstein published the paper and that was the first MOS transistor in silicon. Because it was in silicon and silicon turned out to be the dominant technology, their development was the more important one. Well actually we had preceded it with our thin film.

Heyer:

Thin film devices were built on substrate of...?

Weimer:

Yes they were made on an insulated substrate. We took a glass plate and evaporated the very first ones putting down the metal source and the metal drain. At the time we were using gold to make good contact to the semiconductor. The first semiconductor that we used was cadmium sulfide. We would evaporate cadmium sulfate down which would connect the source and the drain and then we would evaporate to silicon monoxide as an insulator and then we would evaporate the gate. And the gate was aluminum. If you applied the voltages to the gate well of course that caused the current between source and drain to be modulated. We got very nice characteristics, and we went on from there. Frank Shallcross, who was working as part of my group here, found that you could actually make thin film transistors using cadmium selenide. Cadmium selenide seemed nicer than cadmium sulfide so we all switched to cadmium selenide for our N-type transistors. But then also I had found that [inaudible] could be evaporated and could produce a P-type transistor which just was the inverted characteristics using holes instead of electrons. Well having now an N-type of transistor and a P-type transistor of course one tries to invent and see what he could do with it. I submitted a disclosure on the complementary inverter. This was a complementary storage element in the form of a flip-flop with P and N-type transistors, which draws current in neither state. So it's a very low-power drain type of thing and it has really become the basis of the solid-state memory element. That was a bi-product of thin-film transistor work and we were able to get this basic patent on that type of device before those two devices were well known in silicon. We had an early start.

Solid-State Image Sensors

Heyer:

I noticed in some of the thin film stuff here that you seemed to come back to image tubes or image forming tubes.

Weimer:

Well, yes. It was my first love so to speak. When we had a thin-film transistor and a technique for building thin-film integrated circuits it seemed to us the best way for us to develop this technique was to build an integrated circuit using the thin-film transistor. It seemed to us that a nice way to do it would be to combine and try to build a solid-state image sensor, which would really, be a very complex integrated circuit using the new integrated circuit techniques. Well the logic of using the thin film techniques at that particular time around 1963 or 1964, was that silicon integrated circuits weren't advanced far enough to build complex things. So in about 1964 we had built a 180- stage scan generator which was a really shift register that contained hundreds of transistors and capacitors. We put a pulse in one end of it and we would drive it with a clock and the pulse would progress down the line just like a shift register was supposed to. And we were going to use the pulse for scanning purposes with its movement down the line. We use it for our helmets. And so we published that and it attracted attention both because it was a solid-state sensor and because it was a new type of technique based on thin films. Well then as time went on the government was interested in the development of solid-state image sensors. For quite a number of years, we had contracts from right field to develop solid-state sensors based on the thin-film technique. We kept making them bigger and bigger until finally we got to the point where we had built one which had five to twelve elements in a crossed array based on 2 mil centers. So it was an array, which was about this big, in other words just about the same size as the picture area of TV camera tube like the Image Orthicon.

Heyer:

An inch square would you say?

Weimer:

About an inch square. There were five hundred and they were 2 mils apart. They wanted us to keep working on this to see if we couldn't make a workable device. We got pictures from them, the pictures were rather crude but they were workable and progress was being made. However, this work was being done from the period up until our last contract with them in about 1971. Well of course over that period silicon had made a lot of progress so we began to feel that this battle between thin films and silicon was long since won by silicon. By that time we were particularly interested in developing a solid-state image sensor so we felt that with our group here which included physical chemists like Frank Shallcross and mechanical and electrical engineers — when Pike was working with us at that time — we ought to have work going on in parallel on silicon type sensing arrays. So we were doing silicon in parallel with the government work that we were doing on films for Wright Field. And we came up with some interesting new types of silicon devices. One of which was a charge-injection type of silicon sensor which has recently been developed independently by General Electric and they are making very good progress with it.

That type of approach was a useful approach, based on the "X-Y" method of scanning which we had been using earlier with our photoconductive arrays. The X-Y method of scanning was essentially having an array of photosensitive elements with a coordinate strip, an X-coordinate strip and Y-coordinate strip connecting to every one of those elements. And you would scan the element by using digital shift registers or "scan generators" as we called them at the periphery of the array to apply pulses to those X-Y strips. The coincidence of those two pulse would cause the particular element to be discharged and you would look at the video signal appearing from that element. That approach was the one that we had used with the photoconductive arrays. They had rather poor uniformity and because it was difficult to make arrays. It was difficult to get the array perfect and it was difficult to build these scan generators in such a way as to have every pulse identical. Some would be a little different from others and they would overlap in all different fashions and you get a kind of non-uniformity or streakiness in the final picture. About June of 1969, a very important paper was published by Songstrom from Philips. He called his device "a bucket brigade". Actually what it was a series of transistors which he connected up in a rather clever way and if he sent a pulse down that series of transistors the pulse would come out the other end but also if you sent an analog signal down that series of transistor it would come out the other end, delayed but still an analog signal. Well that is a whole new kettle of fish compared to just a digital shift register. A digital shift register just sends down a series of zeros and ones. He suggested that this kind of thing had a lot of applications for delay lines and for image sensors. He didn't say how to do it in image sensors.

Of course that was very intriguing to us and so we got to thinking along those lines. Actually I didn't see the paper until six months after it was out and so early in 1970, I suddenly convinced that this business of charge transfer was the way to build image sensors. So we sat down to try and figure out how to do it and we came up with methods, which had not been mentioned in Songster’s paper. In fact what he mentioned didn't seem like a workable scheme. So we submitted disclosures on this and got started working on the bucket brigade approach for image sensors. As we were working on bucket brigade sensors, meanwhile there was a different type of charge transfer device, which was invented at Bell Laboratories. That was the charge coupled device and is relatively similar to the bucket brigade. If you are going to build the bucket brigade using MOS transistors, it looks very like much a charged coupled device, the only difference being in the bucket brigade you do some diffusions at the interface between the semiconductor and the insulator. But in the charge coupled device you don't have to do those diffusions. You simply move the charge along the interface, along the surface. So the charge coupled device is actually the better device and physically it was a more original device than the bucket brigade. As far as image sensors are concerned, the principle by which it would be used in an image sensor is exactly the same. So we felt that our early work on bucket brigade image sensors was a help to us in getting moving on a charge coupled image.

The Bell people of course had a great big group working on charge coupled image sensors and they moved very fast and were able to demonstrate that it could be used as a charge coupled sensor. They built arrays, which were small to begin with. They built one with 108 elements by 64 or something like that. It had enough elements to get an image but not enough to be a really high-resolution image. The charge coupled approach which of course was pioneered at Bell Laboratories began to look to all of us as the most promising approach for image sensors. So we switched our bucket brigade work to charge coupled work. However, we did have some interesting developments with the bucket brigade. We built one of the first miniature cameras using the principle of charge transfer, in this case a bucket brigade.

Heyer:

I guess you were involved in that.

Weimer:

This solid-state camera that we built used the bucket brigade silicon sensor. It was really the first tiny silicon camera that was built based on charge transfer. Now the big advantage of the charge transfer was that you could get very much better uniformity with charge transfer than anyone had been able to get with the X-Y addressing up to that point. So we built this little camera which was just a few inches on the side, this was with Windpike, Frank Shallcross and Mike Coback, who was a member of the group at that time. Al had already arranged the press conference in New York and we demonstrated this thing. One of the science editors of CBS was there, what was his name?

Heyer:

Earl U. Bell.

Weimer:

Earl U. Bell, he got all excited about it and so he took the camera onto the evening show. He was holding it in his hand, taking pictures of Jim Jenson with this little camera. Although it had relatively little elements, it only had 32x44 elements you might be amazed to know that with that few number of elements you can get pictures of faces, which are entirely recognizable. So that demonstration was a success. It was reproduced in a lot of places and we beat Bell Laboratories to it.

Heyer:

Beat them at their own game.

RCA SID

Weimer:

I know the camera was around here for a year, so at every occasion they were showing this camera. But progress marches on and so it got to the point where we needed more elements to get high resolution pictures. So by the time we gave this demonstration we were already working on charge coupled cameras. We built a 45x60 charge coupled camera. At the time we built one that was smaller than Bell’s but we were still behind Bell. Then we built a 256x160 camera. That’s if you interlaced it, if you didn’t interlace it then it’s 1.8x1.6. And that one at the time it was built was the biggest integrated circuit, which had ever been built. So we were in the swing there with charge coupled sensors. Meanwhile at Lancaster they built a solid-state sensor, which had 512x320. In other words twice as many elements in each direction as the one I had just quoted as being the biggest one that had been built in either integrated circuits or in solid-state sensors. The Lancaster device, the 512x320 element thing, has now became a commercial product. Lancaster is selling them, actually it was the first one, the first solid-state sensor which became a commercial product which had all five hundred lines so that it could be run compatibly with standard broadcast television. All of the others had fewer lines and so they had to be used for just special applications. They could not be incorporated with a standard broadcast signal. So this 512x320 camera is now beginning to be sold, it is called the SID, the RCA SID. RCA is in a good position now with their solid-state camera, however recently Bell Laboratories has announced a 512x512 sensor which would have higher resolution than RCA’s but theirs is just an experimental device it is not a commercial product yet.

Heyer:

So they have a couple years to go?

Weimer:

They have a few years to go before that would be a commercial product.

Heyer:

Are the SID image devices being used in color cameras?

Weimer:

Yes, that is the objective. That they can be incorporated into a portable camera and they have built some models of small cameras using three SIDs. They work out pretty well in all respects except for their sensitivity because the CCD is basically a low-noise type of device. Part of the light is absorbed in the polysilicon gates and the samples that are being sold at present are illuminated from the gate side. The light has to go through the polysilicon gate, it absorbs some of the blue light and so to keep your color balance you have to cut down the sensitivity in the other colors the same way. In other words you can’t use all the sensitivity that they have and so that three-sensor color camera requires a lot more light than would be desirable. On the other hand it is possible in principle to thin the silicon sheet and bring the light in from the backside in which case very little of the light would be lost and it would be a very sensitive device.

Heyer:

It would be turning the whole business around?

Weimer:

It would just be turning it over and bringing the light to the backside. That is still underdeveloped and it’s not a product yet.

Heyer:

You have to have a very thin substrate?

Weimer:

It has to be a very thin substrate. But of course we have that technique already because the Silicon Vidicon has a thin substrate. You have to scan it from one side and bring the light in from the other side. So we think possible to do it but it requires some time.

Heyer:

These were some of the things that you were working on?

Weimer:

Well Lancaster was working on that.

Heyer:

Lancaster was primarily in manufacturing?

Weimer:

They are a manufacturing company.

Pinsky:

They used to be an industrial tube division. They just turned to electric optics I think.

Weimer:

As a matter of fact they are now part of the solid-state division. Of course they are still making the tubes but a good fraction of their work is aimed at the solid-state type of camera. So I guess it is logical as far as that group is concerned.

Heyer:

They had a group for growing their silicon crystals?

Weimer:

Well, they don’t grow their crystals. They buy their silicon but they have facilities for processing it and making the integrated circuit. The fact Lancaster has developed some very fine techniques for building the masks that are required for building these kinds of silicates. Of course you see this kind of circuit is far larger than any of the standard integrated circuits even including all of the semiconductor memories. I mean they are tiny compared to this.

Heyer:

They are less than an inch?

Weimer:

Well I think the actual dimension of the active area is something like 500 mils by 700 mils: it is large.

Heyer:

That is very interesting. Well we are pretty up to date I think.

Pinsky:

The level and the time. Well he was a teacher once.

Weimer:

The tough assignment was being restricted to this general area we’ve been talking about. Well I think that the solid-state devices are here to stay as exciting new devices. There still have not quite met our objective of building a really low cost device. And you see that has been one of the objectives all along. One of the objectives was that a solid-state sensor because it is made by the processes that one used in making integrated circuits could be extremely low cost, so it could be used in consumer type applications where TV has never been used before. The other thing about the solid-state sensors is that because they work by entirely different principles, many of the aspects of solid-state scanning are entirely different then what tubes were. We might be able to build devices that can do new tricks or can do things that tubes couldn’t do. It is more than just trying to do it as good as a tube but cheaper. It is maybe a whole new device. From the standpoint of cost, we are not there yet. I think the device as it stands right now, the cheapest camera tube that can be built is the Vidicon: the small two inch Vidicon can be built for relatively few dollars. I have heard prices like twenty five dollars quoted for a tiny Vidicon camera. It’s a pretty small tube too, it is a tube as big as my finger and actually our solid-state cameras are a whole lot smaller than that. They will be ultimately but right now they are about the same size as this kind of a Vidicon, and they produces a picture which loses quality, about like a Vidicon. The cameras at present cost them thousands of dollars, not a hundred dollars. It is also that the solid-state devices haven’t been developed far enough. We are not absolutely sure, but we have found a way of building a good solid-state camera and by we, I mean we as the whole industry including Bell and GE. We found a way of building a solid-state camera, but we are not sure that these ways are necessarily the ultimate. I mean they work but they are still complex devices. I am not sure that these prices will be down to the range of these cheap Vidicons. And we wanted to make it still cheaper.

Heyer:

It sounds like that you started with the stage when you had the Image Orthicon?

Weimer:

That is a very good analogy. And we had a very fine tube but we needed something more like the Vidicon. Now you see if one could build this kind of device and build it cheap, one has the possibility of going on to new types of consumer applications. The one type of consumer application that everybody thinks of, and Al Rose in particular around here has mentioned time and time again, if we have these kinds of cheap cameras and if we have a cheap method of storing the videos, then you see we are running with Eastman Kodak. We would then have a product, which would really outsell any of the present kind of cameras, where you have to develop the film and send it off to be developed. It would even have advantages over the Polaroid type thing where you use chemical photography. You get only a few pictures and they are relatively expensive whereas you do all this electronically and you have it on tape or a tiny disk.

Heyer:

One of the questions that always come up in thinking about electronic recordings like those is the display. Like the difference between having a tape recorder to listen to or a book, which you can just, open up and look at.

Weimer:

Everyone assumes that once you have this that every time you wanted to look at it, you plug it into your color TV set and you would see the thing in color. Then of course RCA has their way, the color TV in every room.

Heyer:

Any other consumer applications that you can think of? I am just going through my head: the surveillance one?

Weimer:

Unfortunately the way things are in the cities that may become more and more important. It may be more than surveying holes in the prisons, it is surveying your own home. If you live in an apartment house.

Heyer:

I wonder what the possibilities are of combining image sensors like this with the processors that showing up everywhere now?

Weimer:

Well, I think that is certainly an area of future application. Whether or not it’s necessary or best to do a lot of signal processing on the image sensor chip or have an image sensor on one chip and another chip right beside which does all the microprocessing that’s something else. Sometimes it is not always best to have different combinations. But I think they will go hand in hand and that will be a future application for imaging devices.

Heyer:

OK.