Oral-History:Karl Ulrich Stein
About Karl Ulrich Stein
Karl Ulrich Stein graduated with a doctorate in electrical engineering from Stuttgart University, writing his dissertation on magnetics. While still at Stuttgart, Stein began working at the Siemens research lab, later taking a position at the company after graduating. At the beginning of his career, Stein worked in memory performance and thin magnetic film, and in the 1960s with integrated circuits. Stein was also involved with using a symmetrical amplifier and symmetrical design of memory array, leading to patents in Germany and the United States. He later moved to working with microelectronics and microcomputers in the early 1970s in the PC market and automotive industry in areas like engine control, dashboard and fuel consumption. Stein also took part in the basic development of Siemens’ components group, was general manager of Heimann, a small Siemens affiliate, Werk für Diskrete Halbleiter of Seimens and Central Lab.
In this interview, Stein discusses his education and career at Siemens. He talks about his German university training and the differences between engineering training in Germany and United States when he was in school, along with the importance of learning English and reading American textbooks. Stein covers the various areas he worked in from memory performance to semiconductors to asynchronous transfer mode (ATM). He also talks about the ways market and corporations affect product development, and the difficulties of working within the European market. Stein also discusses being in management and the issues associated with restructuring and manufacturing, along with his ideas about the move to broadband during the original interview. At the end of the interview is appended an epilogue written by Stein in December 2009.
About the Interview
KARL ULRICH STEIN: An Interview Conducted by William Aspray, IEEE History Center, 30 June 1993
Interview #169 for the IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.
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, IEEE History Center, 445 Hoes Lane, Piscataway, NJ 08854 USA or firstname.lastname@example.org. 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:
Karl Ulrich Stein, an oral history conducted in 1993 by William Aspray, IEEE History Center, Piscataway, NJ, USA.
Interview: Karl Ulrich Stein
Interviewer: William Aspray
Date: 30 June 1993
Location: Munich, Germany
Education and Siemens
This is an interview on the 30th of June, 1993, with Dr. Karl Ulrich Stein in his office at the Siemens Werke Hofmannstrasse in Munich. The interviewer is William Aspray. Since you've been so kind as to provide me with a resume, I won't have to ask some of my questions. Your education at Stuttgart was in electrical engineering?
It was in electrical engineering. During the studies, you could specialize in what we call Elektrische Nachrichtentechnik, electrical communications engineering. At this time the mainstream of communication technologies was frequency-focused, not time domain-focused as digital technology.
It was a classical education with a large share of mathematics and physics. One of my famous professors in electrical engineering was Richard Feldtkeller. He recommended the use of American textbooks.
He recommended e.g. one of Frederick Terman's textbooks.
So in an early stage, he opened our view to the U.S. literature and the U.S. theory in communications as a key to the world. I liked it very much. As an engineer it's necessary at a very early stage to be open minded to what's going on in the world. He also recommended the study of English, because English turned out to be [a] common language in electrical engineering.
How common was this, do you think, in the education of electrical engineers around Germany, that other people would get the same kind of advice?
At this time, this was quite new. It was not that common at all. Europe came out of a very strong national understanding, and it was not common to open the door and say: "Look around the world at what's going on in engineering." It was quite new in the fifties. But I think it was very good advice because shortly thereafter I became aware that our education, for example, at a German university in solid-state physics, compared to the education you could have at a leading American university, was not adequate. German research was still lagging due to the war and the restrictions after it, e.g. in microwave and radar. So I turned to American textbooks to learn. In German you got e.g. in solid-state physics for electrical engineers only [in] Eberhard Spenke's book. Spenke, Walter Schottky's partner for some time, was a very good basic semiconductor book, but it was not written as textbook for EE's.
I noticed that when you were still working on your engineering degree you were already working for Siemens.
I made the diploma at the university. After the diploma I was engaged in Munich by Siemens. But during the first years there in the research lab, I made use of the cooperation between Siemens Research and Stuttgart University. By this I could finish my Ph.D. thesis in the research lab as an employee of Siemens. This is very similar to what you can do with a master's degree in the U.S., but typically in Germany you terminate your Ph.D. thesis at a university first, and then you get engaged to work in industry. Naturally, some students who had to earn money went under pressure to a part time job and completed their diploma during their stay with a company. The universities did not like that, due to budget pressure. The system did not support the part-time student. The whole situation to study EE was even more difficult because of the two levels of education we had: we had the Fachhochschule terminated by an engineering degree, and we had the university terminated by the diploma or doctorate in engineering. It was very hard at this time -- to continue studies at university after terminating the Fachhochschule. Thirty years ago, you had to completely redo your university studies, from the beginning, from the first semester. Today this is easier. It was changed to have more compatibility in the EU.
For your doctorate, what was your research on?
For my doctorate degree, the research was in magnetics. At this time the state of the art for computer memories was the ferrite core memory matrix, according to the Forrester principle, with three wires. The ferrite rings have been made smaller and smaller to increase the memory capacity of a matrix. The trend was to use less and less magnetic volume to store one bit. One interesting alternative to have less volume, to miniaturize to bring more bits in one matrix, and to allow the batch fabrication instead of a complicated wiring cores by hand was the thin magnetic film. The switching of the thin magnetic film was one of the basics for the memory performance of the thin magnetic films. By theory, this switching could be a kind of coherent rotation of magnetization after exceeding a magnetic threshold field strength. But in practice, other processes happen, one of them is a non-coherent rotation. I studied the transition at low magnetic fields, where non-coherent rotation happens; and at higher fields, where coherent rotation happens and explained it in detail. This was the core of my Ph.D. work. It was applied solid-state physics, applied magnetics.
How was the research received?
The research was published in German and American journals and conferences and received a lot of citations worldwide. However the research achievements in magnetics for random access memories have been overtaken by semiconductors. The basic question of all memories is the signal-to-noise ratio at a sense-amplifier input. To detect a very weak signal from magnetic switching we made a real fully working memory. I also wrote a theoretical paper about how small this magnetic film could be and what signal-to-noise ratio is required, bringing traditional communication engineering to the subject. This was not common in computing, as people believed at this time in faultless computing. As sensitive, low noise semiconductor amplifiers are the crucial point, we started to study these amplifiers.
Integrated Circuits and Memory
At this time the integrated circuit came up. It was the second half of the sixties. Then we made the first IC's for reading out and for control of the semiconductor memories utilizing the Siemens semiconductor factory. Getting deeper involved in IC's, I personally soon found that was easier to make a semiconductor memory.
This was around 1968. At this time, in the U.S. most IC work seemed to be driven by the military and even some technological generations had names given by the military systems. Activities outside the US have been measured if they were 3, 2 or only one year behind.
What were the applications intended here in Germany?
In Germany, one of the first applications was a three-stage, bipolar transistor amplifier. It was first intended to be used for a hearing aid. But its first volume application later on was an audio amplifier used by Grundig in the entertainment industry. For my colleagues working in the semiconductor business group, their driving force was entertainment, especially Grundig. Before the IC, they also made innovations with Grundig on germanium transistors for tuners. They also pioneered infrared remote control by a GaAs infrared transmitter and Si receiver diode.
In contrary to entertainment it was very hard for us to implement innovations in computers because the lead of the U.S. was so dominant at this time. However, as before for my early work in magnetics, the request of the Siemens computer group was still strong on Random Access memories. With this aim we started to make our own p-channel MOS process in the research lab, first to investigate transistors then to use it for ICs. Soon we realized that our MOS process allowed us to integrate a RAM with 256 bits on one chip. Within half a year we had the first working silicon chips and were able to proof an access time of seventy nanoseconds. This was the research goal for our previous program with magnetic film memories where we finally needed five years to achieve it. As we needed half a year only with MOS we were convinced that MOS was the way to go. The circuit used was a static RAM-cell with eight transistors, then we went to six transistors. Sometime later on a three-transistor memory cell was used by the start-up Intel for the 1103, the first 1K memory in the volume-market. To move to a 4K chip we needed another circuit. To use a one-transistor memory cell was the consequent next step, despite a lot of skepticism against the dynamic principle. Another difficulty was to implement a small cell in a silicon gate process which became the forefront technology at this time. To achieve high speed we made the first n-channel silicon gate. But the most important thing was to have the right sense amplifier for a small dynamic memory cell. I proposed a switched flip-flop as amplifier. This kind [of] sense amplifier was not in general use in memories at this time; instead, a single-ended amplifier was used.
We introduced a differential sense-principle with the flip-flop as an amplifier. This was a great move to have symmetrical memory. The symmetrical amplifier was connected to 2 bit lines. This offered the advantage to sense the difference of a cell in an active field, and a dummy line on the other side. From the dummy side you get word line noise only; from the other side you have noise and the signal meaning the binary information according to the charge of the capacitor in the one-transistor cell. This allowed us to step up to the 4K memory. That's a basic principle that's still used today up to 16 megabit now.
Today all these developments are, of course, group efforts. But was this mainly your idea?
My idea was to use a symmetrical amplifier and the symmetrical design of the memory array. Naturally, it was implemented by a group for circuit and design, which reported to me, and a neighboring semiconductor technology group, led by my cooperative colleague Manfred Zerbst, well known to all MOS-people by his fundamental findings in MOS, the Zerbst-plot. Back to the symmetrical amplifier. You need the symmetrical sensing if you are in a noisy environment as e.g. in magnetic memories before. We got German patents for the amplifier and the memory array and later on the U.S. patents.
Patents and Licensing, European Market
Was this put into use immediately by Siemens?
This was a problem. Innovation is driven by volume business, which for computers was in the U.S. The researchers from competing companies understood your ideas better on IEEE conferences as ISSCC than the colleagues in your own company. Siemens semiconductors at this time had no special effort in memories, the big Mega-Project for a 1-M-Ram started in 1984. Intel published on ISSCC 1970 a 1-K-memory with a 3-transistor cell. Then two years later everybody expected their 4-k-chip. I presented my paper on the storage array and sense/refresh circuit mentioned above on ISSCC 1972. Sometime later Intel continued their lead with a 4-k one-transistor cell SRAM.
Did Siemens gain from this financially? Was there licensing to Intel?
I think so. Big companies do a lot of cross-licensing. For cross-licensing, each party must bring in some key patents. By this, Siemens gained soon from our work as they got access to the IPR of cross-licensing partners leading in semiconductors. But I learned from this that even the engineer must consider for which segment of the growth-share matrix or portfolio of his employer he works. Do you understand what I mean?
Okay. Here is the idea. If you work for a cash cow business of your employer your ideas are very likely to be used to maintain the profitability and relative market share of the business. However if you work for a question mark business where the market growth is high however your company does only hold a minor market share and therefore a lot of money is needed, or in other words, this business is deep in the red, your inventions also have question marks concerning their use. If you have a choice, you must look where you put your own efforts. From my understanding it is very important that you deliver your ideas to businesses that are so innovative and strategically oriented, that they can use your ideas. At this time nearly all Europe-based computer businesses have been in the question mark segment, due to different reasons. One reason is quite simple. The European market was split in national markets. So if you did any development in Europe you did it for a small market only. If you make any product development in the U.S., your market was ten times the market you had compared to the market in this field e. g. in Germany. There were some attempts to build a bigger European computer company, Unidata, to access better to the segmented market, between Siemens and the French and Dutch.
Philips, yes. But it didn't work.
So I learned as an engineer it's better to invent for businesses that have the power, that have the force and the strategy to implement the innovation. This was not the computer and the memory field. And the invention and design of a 2-chip- MOS-microcomputer by some colleagues in the Data Lab, terminated without response in the business, confirmed my opinion.
Microelectronics and Microcomputers, Automotive Industry
As a consequence, I took the chance in corporate research to do more for the application of microelectronics and later the rising microcomputers for the stars, cash cows and fostered question marks of Siemens businesses. This move was strongly supported by the head of Siemens Research to whom I reported, Heinrich Welker, a well known theoretical physicist who invented the compound semiconductors, the Silicon-gate field-effect transistor and together with Herbert F. Mataré the junction transistor, at about the same time as the invention of Bardeen and coworkers.
The basic question for our new research was. "What's going on with the microcomputer, and what's the importance of the microcomputer for Siemens's businesses?" This was in the first half of the seventies. We invested to learn the microprocessor's importance for factory automation, the medical area, the new automotive business and (naturally) computing. At the same time, the first ideas on a PC came out. We made the first Siemens-PC utilizing an American motherboard, but we found out it's too hard for our small market to make or adapt software. If you have only one tenth of the turnover with your PC's software, development is ten times as expensive relative to your potential revenues and furthermore most of the costs are at the beginning.
Yes. The production is nothing.
This was the time when the PC business started in the US and Apple took off. We had to be very cautious because the market was not here.
So what was the strategy?
Siemens decided to look for some areas where the German and European market was good. In fact, it was the automotive field we chose because in the automotive field the German and European industries had a leadership, especially if you look at cars such as Audi, BMW, Mercedes, and Porsche. They are located very close to us. They are all within 200 km around Munich. Volkswagen is 500 km, also not too far away. At this time we approached the automotive industry to innovate by microelectronics and the microcomputer. This approach was split up in two areas: one was engine control, the other dashboard. At the beginning of the seventies, the aim of engine control development was primarily to reduce the emissions. Then in 1974 with the first oil-crisis it was necessary to reduce fuel consumption. I remember we started in 1972 to make a complete digital engine control -- the first experimental car utilizing an advanced compact computer, a General Data Eclipse, to control the injection of fuel. At this time Bosch had already done the pioneering up to production with an analog engine control system, according to its leading position in engine control, not only in diesel, but also in the Otto engine. We utilized some Bosch sensors and some Bosch parts because they were well adapted to our needs. This included the fuel injection and all these things. So in principle we only removed the Bosch analog control and replaced it by a digital control with the computer, gaining all benefits of the digital world in specific accuracy and stability. But at this time Siemens and its potential customers were not yet ready to use the results. Siemens started this engine control business some years later. This business is today in Regensburg, and it's growing well. By our research work we only did the first steps. You need more than just the ideas and the basic patents of the engineers to make a business. This was one application of computers in automotive. Another field we also entered was the dashboard. We developed the first BMW trip computer, as BMW called it.
Today it's quite common to use this. Despite the fact that BMW was one of the smaller companies, it was and is very innovative. This was in short the research chapter of my microcomputer story.
Later I became responsible for the basic development for the Siemens components group, from 1976 to 1980. Then I was asked to be general manager of Heimann GmbH, which was a small Siemens affiliate. This was a very, very interesting period because it was the first time I was responsible for a business
I see. What's their business?
They had two strong fields, X-ray baggage inspection equipment and flash tubes. Just from the beginning I was asked to go to this place as a troubleshooter. At this time about 25 percent of the shares was held by the founder, Professor Heimann. He was retired, but still had a strong personal influence. So it was a personal experience for me to have close contact with him and maintain a good relationship. He was a successful founder, rare at this time in Germany. He was a physicist, and he started his business with flash tubes and TV-pickup tubes. But the flash tube business was very suffering at this time because they had three technologies and a shrinking turnover, too small for three technologies. They had development in Wiesbaden and manufacturing in Singapore. We restructured the products, qualified one of the technologies to become number one worldwide after one year. With Singapore as technical base it was easy to get access to the Asian market.
The other branch of Heimann was baggage inspection. Beside TV-pickup tubes, Heimann developed tubes for special applications: high resolution tubes for medical and lowest-level light pickup tubes. The last one enabled Heimann to enter successfully the emerging baggage inspection equipment. This was at the end of the seventies and the beginning of the eighties. During my time we developed the first fully solid-state inspection machine. It's the principle you have today with a belt going through and a line of detectors.
It was a very challenging technology change from the tube-age to semiconductors to make this first machine.
Werk für Diskrete Halbleiter
But after one year with Heimann I got the offer to [be] head of the Werk für Diskrete Halbleiter of Siemens, responsible for all technical activities of the discrete semiconductor business. Again, starting as a troubleshooter. This was a very, very hard task because I had to repeat what I had done at Heimann for a business of ten times [the] size and about 6,000 people worldwide. We had factories in Germany -- in Regensburg, two factories in Munich, Pretzfeld, where the pure silicon and the power-semiconductors have been pioneered. We had one in Italy, two in Malaysia and two affiliated companies in the U.S. We had to restructure [every]thing, from basic Si and compound-semiconductor crystal-growth to manufacturing and semiconductor packaging. The hardest was to reduce work force and maybe to sell or close factories to be successful with the main activities.
Was it clear to know which ones you should close?
At the beginning it was not clear. But there is a set of rules to assist this decision from access to markets, [to] available skills, to competitive costs. At the same time we had to restructure technologies and the product lines. We had too many different products. The product range for the discrete semiconductor at Siemens at this time comprised Power, Small signal and Opto, from low-noise gallium arsenide for microwave, standard discretes such as the TO-92 package and as small (TO-23) surface-mounted transistors and diodes, special semiconductor components for entertainment electronics (for tuners for example), and a full range of opto components from visible LED displays to the infrareds. Some products had a leading position in the market share and technology as the transmitter and receiver diodes for infrared remote control; many were in a less favorable position. We had to restructure the product and technology spectrum and at [the] same time foster new families as power-MOS. Naturally, in such cases you have to have close cooperation with the commercial people. Siemens had a principle, which [w]e helped quite a lot, that you have a one-on-one relationship with one technical person and one financial person. We were able to regain ground with some help from the market, and come to profit after two years.
Our major factories were in Malaysia, in Malacca and Penang; and in Germany in Munich and in Regensburg. In the U.S. we had one affiliate for optoelectronics and one for microwave gallium-arsenide. For me, the task to manage the bundle of different mature and emerging technologies from development to volume manufacturing was a huge challenge. To find and coordinate the right people for the different requirements was maybe the most important key to success. Let me give you some examples of the most challenging areas. In optoelectronics we had to cover the range from gallium arsenide for infrared-LEDs to heterostructures for fiber-optic lasers. Packages ranged from inline LED to the emerging surface mounted devices and to micromechanics of lasers. The mass-production of small-signal transistors and diodes was since the beginning for the consumer-electronics market. At my time the first satellites for TV-broadcast have been launched. We produced the required low-noise gallium-arsenide transistors for the receiver front ends in competition with Japanese companies. You can't imagine what happened after one of the first satellites didn't work properly. After huge efforts to start production we then had immediately some100,000 transistors in stock.
Another emerging product line was in power semiconductors: the power-MOS. For me personally, another case study of the long way from research to the world market. We started research in my department on this technology in the beginning of the seventies at the same time when we worked on memories. When I moved to the components division I completely forgot this project. But when we searched out of component-application requirements for what we called a "microcomputer-compatible power semiconductor" we came back to this work on power-MOS. This power semiconductor should have an input compatible to MOS-ICs, and an output that could replace a relay. This was the basic idea. Our first products have been called SIPMOS; Siemens Power MOS. Technology started with a planar short channel made with a special Ion- implantation, a some micrometer short channel, some 10 mm long length MOS transistor. The beginning of applications of these components was as difficult as their manufacturing. One of the strangest applications I remember was for control of golf carts manufactured in Japan, so there was soon interest for licensing from our partner in Japan, Fuji Electronic Components, with roots back to the Siemens activities in Japan before the First World War.
Now to the high end power semiconductors. If you travel within Germany today, you use the ICE, the Intercity-Express, which holds since its first presentation in 1985 some speed records for high-speed trains competing with the TGV and the Shinkansen. The development of this train was one of the major challenges for the high power semiconductors as it required for the control of its AC motors Gate Turn-Off Thyristors (GTO) for 4,000 volts as key-component. To avoid the dependence from Japanese competitors the Siemens Transportation Systems required an in-house supplier. This opened the gate for a family of high-speed trains today successful worldwide, the most recent step in the pioneering work of Siemens in rolling stock which started with the demonstration of the first electric train in 1879.
Central Lab, Broadband
After about eight years' responsibility for semiconductor business I decided to return to R&D-management, triggered by [a] situation in my family after my first daughter and later on my wife died. I became head of technology planning in Siemens corporate R&D. Two years later I accepted the offer of the Public Networks Group to join and sometime later to head the Central Lab for communication networks. The main task here in the Central Lab is to make the advanced systems development including prototyping for public networks. For an engineering manager, this is a very interesting and challenging job, because the Public Communication Networks Group belongs to the three biggest players worldwide in this field: AT&T, Alcatel, and Siemens. I expected the opportunity to develop innovative ideas for the world market, in a group with a business which has the power to make innovations happen. For the time being, we are transferring our prototypes of ATM-switches proved by [such] customers as Deutsche Bundespost Telekom and NTT in Japan to product development. Asynchronous transfer mode, ATM, is the technology selected as standard by the public network operators for the next generation broad-band networks, a packet switching with a fixed packet length and priority features for time sensitive applications. The ATM uses many technologies familiar to me from ASICs, to fiber-optic interfaces, to software. In such an ATM-system, even in prototypes, we have a family of different ASICs. The biggest one of it has the same number of transistors as Intel's Pentium. We designed these chips by ourselves from scratch. We utilize either Siemens or other silicon foundries. Then we implement all this, including the software. Naturally, we also use a lot of standard components, memories and microcomputers. For fiber interfaces we utilize semiconductor lasers and bi-directional (transceiver) fiber-optic modules. So the Central Lab is a very interesting and innovative, sought-after prototype field.
On the other hand, what we have today in networks [is] the move to broadband as the big challenge, a big competition. For nearly 100 years all this worldwide communication was done by telephone. The question now is how fast it will turn to broadband and which technology will take over. For the technology, I'm sure it's ATM standard. But we also know that even the basic standards of ATM are fixed at a terminal, which means requir[ing] a terminal with a specific network interface. But, there's still a lot of interaction with a fast-developing market. We have the chance as an industry to continue with the success we had in the telephone business. It's a chance. However, It could have some similarity with the computer business, where it was very hard or even impossible for some companies to leave their mainframe success story and live in an environment with distributed computing and all the PCs. So we are well aware that it's necessary to be open-minded and flexible and responsive to the market as such a change happens.
Epilogue, December 2009
What happened to....
Magnetic films of my first years in research? The search engines such as Google Scholar still preserve plenty of publications and patents, including my final publication on the narrow limits of magnetics in a matrix, not considering the perspectives of the moved magnetics to today's terabyte hard-disk drives.
MOS-DRAMs made their way according to Gordon Moore's law, based on an exponential growth of the volume. Siemens started late in DRAMs, played after the IPO of its semiconductor business in Infineon in the league of the first ten manufacturers, struggled in the 2009 worldwide economic crisis with its affiliate Qimonda.
The MOS-Power components survived and Infineon has been successful to develop it into a star in its portfolio. Infineon claims to be No. 1 worldwide in power electronics.
The opto-semiconductors flourish within Siemens in OSRAM, the lighting Group. It's stated to be the world’s second largest manufacturer of optoelectronic for the illumination, sensing and visualization sectors. The CEO of OSRAM Opto Semiconductors, Rüdiger Müller, responsible for this product since my days there at the end of the 1980s recently stated: “the rise of the LED, which has only just begun, is set to sky-rocket in the coming years.”
I am sad that the traditional core business of Siemens, the reason for its foundation, telecommunications, could not survive within Siemens. It happened in the worst case as I indicated in my interview in 1993 in the closing sentences.
- Research and development management
- Communication switching
- Packet switching
- Computing and electronics
- Integrated circuits
- Solid state circuits
- Electron devices
- Semiconductor devices
- Electronic equipment manufacture
- Magnetic memory
- Digital systems
- Computer classes
- Engineering education
- Fields, waves & electromagnetics
- Conference activities
- Lasers, lighting & electrooptics
- Optoelectronic devices
- People and organizations
- Manufacturing & production
- Road vehicles