Oral-History:Takao Nishitani
About Takao Nishitani
A 1998 IEEE fellow "for contributions to the development of LSI chip architectures on digital signal processors," Takao Nishitani produced pioneering DSP (digital signal processing) and ADPCM (Adaptive Differential Pulse Code Modulation) technology as an employee at NEC during the 1970s and 1980s. The DSP chip he developed revolutionized microprocessing and the computer industry in the 1980s and 1990s. As professor at Tokyo Metropolitan University, IEEE volunteer, and leader in the Signal Processing Society Tokyo Chapter, Professor Dr. Takao Nishitani has bridged the academy and industry to promote industrial signal processing research.
This interview covers Takao Nishitani's education, industrial career, academic career, and institutional leadership. Dr. Nishitani describes his childhood and University education, considering the impact of Japanese student protests on academic computer science in the late 1960s. The interview then details the roles that corporate objectives, the Japanese economy, politics, and international trade played in DSP research during the 1970s and 1980s. IEEE trade publication Electronics and the IEEE International Conference on Acoustics, Speech, and Signal Processing promoted DSP advancements in the international computer science community. Participation in international networks influenced Nishitani's application of nonlinear theory to ADPCM, and he served as the Japanese representative for the International Telecommunications Union's development of ADCPM standards. Nishitani discusses applications of DSP and ADCPM technologies, concluding with discussion of his transition from industry to academia and of his work with the IEEE and the International Conference on Acoustics, Speech, and Signal Processing.
For additional information on signal processing and DSP applications in the 1980s and 1990s, see Frederik Nebeker, Signal Processing: The Emergence of a Discipline, 1948-1998 (Piscataway, NJ: IEEE Press, 1998).
About the Interview
NISHITANI: An Interview Conducted by Michael Geselowitz, IEEE History Center, Piscataway, NJ, USA, 13 May 2008
Interview #477 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, IEEE History Center, 445 Hoes Lane, Piscataway, NJ 08854 USA or ieee-history@ieee.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:
Nishitani, an oral history conducted in 2008 by Michael Geselowitz, IEEE History Center, Piscataway, NJ, USA.
Interview
INTERVIEW: Professor Dr. Takao Nishitani
INTERVIEWER: Michael Geselowitz
DATE: 13 May 2008
PLACE: TEPCO Electric Power History Museum in Yokohama City, Japan
Childhood interest in engineering and electronics
Geselowitz:
Professor Nishitani, would you tell us about your early years and how you got interested in engineering?
Nishitani:
I was born as the son of an engineer in my grandfather’s manufacturing company. When my father was in his university hood, he got pneumonia and his health had been not very good since then, so he and my grandfather started a Japanese antique shop. When I was a boy, I often visited their shop. I always went to the back of the shop where there was a workshop. I liked to make toy ships, there. This was the start of my engineering. My father got out of the antique shop business when my grandfather died and started tutorial services in mathematics for high school.
Geselowitz:
About what year was that?
Nishitani:
That was when I was around fifteen years old. I was an example for my father's tutorial services, so I had to learn a lot of mathematics in order to help my father. I had to study hard and therefore became good at mathematics. Those who were good at mathematics were highly welcome to become engineers in Japan at that time and I also had great interests in electronics technology in my high school hood. That's the reason I took an entrance exam to join the electronics course in Engineering Department of Osaka University, because only Osaka University and Tohoku university in Japan had the electronic course at that time.
Geselowitz:
What was your father's background that he became a mathematics tutor?
Nishitani:
His background was as a designer of diesel engines.
University education in signal processing and computer science
Nishitani:
When I joined the university, I was not a very good student. Especially in my third year at the university, we had long vacations because of campus disputes. We didn't study any subjects. Even if we would like to hear some lectures we could not have them.
Geselowitz:
Was the dispute between the faculty and the administration or were the students involved in it?
Nishitani:
It was between the students and faculty.
Geselowitz:
What year was that?
Nishitani:
It was about 1968 or so.
Geselowitz:
That was the same year that the students in France were striking and the students in the United States were protesting.
Nishitani:
Yes, that's right.
Geselowitz:
It was that international student problem in 1968.
Nishitani:
Yes, that's right.
Geselowitz:
Okay. I didn't know it was in Japan too.
Nishitani:
It turned out to be a very nice chance for me to learn a lot of things, because when I went back to the university we were asked to study a lot to catch up the level of other students in other grades. Otherwise we would become dropouts in our engineering life. I joined the electronics applications laboratory. Professor Kiichi Kitamura and associate professors gave me two books. One was Signal Theory written by Lewis E. Franks and the other one was Microprogramming written by Maurice V. Wilkes who engaged in the development of the IBM 360. At that time our professor wanted to change the research area from switching transistors to computers. In terms of switching transistors, they were engaged in electrical switching systems in transmission. Therefore, they gave me the book on signal theory for transmission research purpose. Since he also wanted to change their research to computers, they also give me Wilkes' Microprogramming. Microprogramming is basic ideas for making large computer systems. Microprogramming is kind of a computer within a computer. Both books were very hard to read for me.
Geselowitz:
Were they translated or in English?
Nishitani:
They were in English. Franks' Signal Theory was very difficult for me, because he used a lot of advanced mathematics, which I have never learned, due to the 3rd grade absence. Wilkes' approach was another big problem, because he always compared his ideas with conventional computers. I didn't study about conventional computers, so everything was new. I could not understand why it was important but I still learned a lot of things from his book. Franks' theories gave me a lot of interest, because they were very hard to read, and this hardness gave me some hope for a bright future. I believed that when I could understand those difficult sections of the book, I could get a lot of new information. Therefore I was eager to read such books, and as a result my background was in both signal processing and computer science.
I joined NEC in 1973, after receiving my Master’s degree.
Geselowitz:
When you graduated with your bachelor’s degree, were you given offers to go into Japanese industry? In other words, why did you decide to stay and get a master's course rather than going straight into industry?
Nishitani:
In Japan when someone got a Ph.D. degree – and even today it is true in some sense – then not so many companies accept him. If a company wants to do research in specific directions then it will get a smart student or Ph.D. However if the Ph.D. student's interests are not in the same specific direction the company wants to go, they never want to hire the Ph.D. student, because he would be too specialized. Japanese companies could not fire them easily, due to strong labor unions. In the 1970’s, however, some Japanese companies hired some master's students for the possibility of developing their companies by utilizing those with high educations but not so specialized as the Ph.D. I wanted to be an engineer of new fields.
Geselowitz:
So, you thought if you stayed in school for two more years and got a master's degree it would help you in your career?
Nishitani:
Yes, definitely. So, I stayed for the master's course and then I joined the NEC Corporation. My background was in signal processing given by Franks and computer science given by Wilkes.
Geselowitz:
Was your master's thesis more on signals or more on computers?
NEC and digital signal processor development
Nishitani:
It was somewhere between or a combination of the two. It was a kind of image processing. NEC Laboratories' bosses considered that I was a researcher in pattern recognition. Not in communication, not in computers. However I insisted that I would like to do something in signal processing, so they finally decided that I should join the communication laboratories. The communication laboratories did a lot of things. At that time NEC changed its focus from analog communication to digital communications. Everyone in the laboratories was engaged in digital communications at that time. However, in my case, since I knew something about computer science, I was assigned to develop digital filters.
Geselowitz:
What were NEC's main products in those days that they thought that digital filters would be important?
Nishitani:
Most of NEC's big business at that time was transmission and communication equipment such as FDM and PCM channel banks and so on.
Geselowitz:
They were manufacturing those for large telecom companies like AT&T?
Nishitani:
That's right. Also some other companies overseas. Then, they changed their mind and wanted to expand their activity to large scale computer systems and semiconductors. There was some activity on speech recognition or digital modems for data transmissions on joint projects between company laboratory and some manufacturing divisions in NEC. They needed some processing of digital signals. Initially they had very little activity on digital filters, but they wanted to realize FDM/PCM channel bank conversion in digital domain, called trans-multiplexer later. So I joined that area at the laboratories. At that moment in the digital filters we needed too much amount of multiplications for real-time processing. Also multipliers required a lot of logic IC chips, and I had to make a multiplier by using a lot of chips. One multiplier became so large that nobody wanted to use it. Therefore, making digital filter hardware was one research area. Remember, I had joined NEC in 1973 and in 1971 Intel sold the first microprocessors. The microprocessors were the initially the 4040 and the 8080. They did not have the capability to do digital signal processing. But soon Intel introduced bit-slice microprocessors, which were designed for making 16bit or 32 bit minicomputers by using 8 or 16 bit-slice microprocessors in parallel for realizing one word processing. My boss changed his mind, and said, "Hey, Nishitani, you should try to make some digital filter by using bit-slice microprocessors." NEC wanted to have some business in the digital filter area.
Geselowitz:
What was your boss's name?
Nishitani:
Mr. Atsushi Tomozawa who had been sent to INTELSAT in the States by NEC and returned to Japan, when I joined NEC.
Geselowitz:
He thought that you could use this new microprocessor chip to make digital filters that could be sold to AT&T or whatever?
Nishitani:
Yes, but his imagination was not correct. This is because even if we used such microprocessors the system becomes very huge. It's not so easy to sell. I'm talking about DSP—digital signal processor—which can be found now in every sort of electronic device.
Geselowitz:
That was a big revolutionary area for which you are known.
Nishitani:
Yes. And I'd like to talk about why NEC joined the development of DSP. There are two backgrounds. One is that Japan had a heavy consequence from the Iran-Iraq war in the 1970s, a severe shortage of petroleum. Every economic activity was slowed down. The research laboratory's was not a profit center, and its budget came from the profits on the sales side. Therefore the NEC wanted to shrink the budget for laboratories, and we had a very hard time. At that moment one manager in the manufacturing division visited us. He said, "The COMSAT Laboratory in the United States published a “call for a study proposal”, because they want to halve their satellite channel costs by reducing the bit rate of the speech information.” The request for proposals asked that in addition to the main study contract of low bit rate speech compression, any bids had to include hardware implementation. The manager said, "In terms of hardware, we are already manufacturing wide variety of communication equipment, but in the study area we don't have any idea." At that moment nobody in NEC was engaged in speech compression. Then Mr. Tomozawa and other managers of the laboratories considered his presentation and their need for funding, and created a small group for studying speech compression to participate in the COMSAT proposal. In studying compression techniques, we decided to have some programmable devices because there were several algorithms and they are not fixed. Then I was told I became a member of this group and that I should make some signal processing equipment for such purposes. I designed a huge speech compression system to implement speech compression techniques. Speech compression technique was very interesting for me. For one thing, there was no field of speech compression in our labs at that moment. I was a member of one of the first such groups. We got our supplies from NEC’s echo cancellation technique guys.
Trade publications and the IEEE
Nishitani:
My specific project had only three active members. We had to learn a lot of things from IEEE Transactions. Especially, as you know, the signal processing and communication Transactions. Finally we realized a 32-bit per second ADPCM(Adaptive Differential Pulse Code Modulation) system. Our system was unique because it was very robust to transmission bit errors. I incorporated the appropriate algorithms and we made some very huge hardware and sent it to COMSAT, who seemed interested. However, Mr. Tomozawa came to me one day and said, "We should change our mind. The programmable device is no longer interesting, because the size is huge and in consideration of a mass market we should not make such a huge system."
Geselowitz:
For COMSAT it was okay for a one-shot deal.
Nishitani:
Yes, for experimental purposes only.
Geselowitz:
But if you want to commercialize it, then you can't have a system that's this big.
Nishitani:
That's right. I had to agree with him in terms of size. Therefore I told my boss that I'd like to publish on the hardware aspects. I wrote up a single page presentation paper for the annual National Convention of The Institute of Electronics, Information and Communication Engineers (IEICE) which is a Japanese counterpart of IEEE. Then a longer article in the journal Electronics. Electronics was appeared through an interview by an article writer. I understand that the Electronics journal in the United States and the Japanese journal have some relationship. One article writer thought this was very interesting and he wrote an article on my work.
Geselowitz:
Electronics is a trade journal, as we say.
Nishitani:
Yes, a trade journal. Reporters attended the conferences of IEICE. This one gentleman says, "This is interesting," he wrote it up, and then he told Japanese readers that it was a new kind of engineering. As Electronics is a trade journal, it is easy for many people with average engineering backgrounds to read. Many people in NEC noticed my activities, thinking, "There was one curious guy who made a very huge system for speech processing." However, it was still seen as a curiosity, too large to be commercially successful.
Memory chips commerce in the U.S. and Japan
Nishitani:
But then there was another big political event. It was a trade war between the United States and Japan. Japanese economy had recovered at that time. Many memory chips were shipped to the United States from Japan while the United States shipped to Japan only microprocessors such as 8086 or so. One microprocessor needs a lot of memory chips, so Japanese trade to the United States was very huge while American trade to Japan was very low. Therefore there was a big gap in the balance of trade.
Geselowitz:
Right. From the U.S. perspective it was a trade deficit.
Nishitani:
Yes. Therefore there were some governmental discussions. Every time in such governmental discussions the United States government said, "This is unfair, because United States always makes clever and smart chips, but Japan is only expanding the memory size. This is a kind of trick.” At that moment in time NEC especially shipped a lot of memory chips to the United States. One guy in NEC semiconductor group considered that we should make a single smart chip every year. By making such a chip the United States government will have to be stop a complaint. The project was called "Chip of the Year." Every year NEC would make one clever chip and ship it to the market. In this project, the first one was not mine; the first chip was a floppy disk controller, which was used in every IBM PC.
Geselowitz:
Floppy disk controllers. Okay.
Nishitani:
Every IBM PC contained such a chip. NEC earned a lot of money by producing that chip. As the article of my huge system from Electronics appeared in the project planning phase, the project head person rushed to our laboratory to the feasibility investigation. I prepared several possible approaches under their implementation limitations.
The digital signal processing (DSP) chip
The first DSP chip
Nishitani:
Fortunately, the second chip in the Chip-of-the-year project became a DSP of mine. At that moment the project was carried out by the collaboration between myself and engineers in the microprocessor division in semiconductor group who were familiar with microprocessors. The background difference between the two was that I knew about the field of application—signal processing and hardware features based on ADPCM—and they knew a lot about microprocessor implementation. The microprocessor division people might look for several other candidates such as more advanced microprocessors, but the computer division people would consider only huge systems. If asked, "How much memory do you need?" they would say “a megabyte or so.” However I understood the speech compression system, especially in the case of ADPCM, had different system parameters. Also, in terms of architecture, my architecture was more flexible for processing, like a microprogramming written by Wilkes. The microprocessor division therefore chose my ideas for making a chip, and did it in less than two years. This became the first DSP in the world. We cooperated to make some good DSPs.
Geselowitz:
Roughly how large was the first DSP chip physically?
Nishitani:
It was a 5 mm x 5 mm chip.
Geselowitz:
Wow.
Nishitani:
I cut off a lot of nonessential functions. I added a lot of important functions for communications in addition to a built-in multiplier followed by an accumulator for continuous dot product operations. For example, a high speed serial connection for a single-chip PCM codec was introduced as a replacement of A-to-D and D-to-A converters, and a look-up table for linear/non-linear PCM conversion was also introduced for signal processing in a linear domain. The chip had a lot of new ideas in its architecture.
When we disclosed our approaches in l980, it was the same time period that Bell Labs produced a DSP. However due to the organization of AT&T, they could not sell such chips. Therefore, for maybe three years NEC was the single company that produced the smallest and most powerful DSP in the world.
ADPCM algorithm and the DSP chip
Soon, we decided to program our robust ADPCM algorithm onto the DSP chip. I presented our design to ICASSP (IEEE International Conference on Acoustics, Speech, and Signal Processing) in 1982. At that time ICASSP had one session which described some ITU-T (the standards arm of the International Telecommunications Union) activities for speech compression. My paper was assigned to this session even though I didn't know about such activities. Several speakers from Bell Labs and from companies from France that were working on ADPCM joined this session. Our ADPCM seemed to be the best one in the session. The ITU-T expert group chairman asked NEC to join in standardization effort for ADPCM. Although ITU-T is organized as several committees for international standards, discussing among researchers from government institutes, I was assigned as Japan representative. One company from France, Bell Laboratories and NEC joined together to cooperate and make a world standard.
In addition, at that time the chip was sold in the U.S. by NEC Microcomputers Inc., which was located near Boston. Researchers at The MIT Lincoln Laboratories, who were engaged in some work with military equipment, visited NEC Microcomputers Inc. and learned our chip a lot. Then, one guy in Lincoln Laboratory started to give some lectures at an international conference on how to use the NEC DSP. Then, many people around the world became even more familiar with my DSP.
Geselowitz:
Do you know who that was?
Nishitani:
It was Joel A. Feldman, who wrote the first application paper on DSP in IEEE Trans, on ASSP..
Geselowitz:
Was that also an IEEE conference?
Nishitani:
Yes, I believe so.
Geselowitz:
Was there any organization overseeing this standardization or did the three companies just get together themselves? Was IEEE or any other international organization involved in the standards process?
Nishitani:
The activities were managed by ITU-T, which was a very huge organization on international telecommunication standards under the United Nations. It had study group for each standard. In this study group, there was heavy ADPCM technical discussion on which way to go, they made a special expert group, and in the expert group I understand that there were two competitors: one from France and one from the United States.
Geselowitz:
From Bell Labs?
Nishitani:
Yes, but they are representatives from USA and France. I also became a representative researcher from Japan. The Bell Labs approach, to send a 4.8 to 9.6 Kbps data modem signal through telephone lines, was very timely proposal on expanding data transmission. The French approach was very interesting, because even if we connect ADPCM/PCM and PCM/ADPCM conversions in tandem, there was no degradation of the signal-to-noise ratio. That was a very interesting and practical function for actual network use, because telephone switching is carried out in PCM only and therefore at every switching office ADPCM is first converted to PCM and after switching the PCM is converted back to ADPCM again. To ITU-T this function was very important. Therefore, because of my knowledge of robust ADPCM to transmission bit-errors, caused by satellite communication paths, the ITU-T expert group asked me to join them. The NEC system was, of course, by DSP. Bell Labs also made their system by their DSP. The French system, however, used a huge system made with off-the-shelf chips. France could not easily modify their system, while our system and Bell Labs' system were both software programmable, so it was very easy to make some compression algorithm changes. Both sides compromised, and we made a single world standard. At that time our system was robustly improved version of the algorithm proposed by Professor James L. Melsa at Notre Dame University. I visited the University for the purpose of meeting him. And at that moment, Professor Melsa introduced me to a nonlinear theory professor Ruey Wen Liu. Prof. Liu gave me a lecture for one hour on how better to use a nonlinear theory to systems such as ADPCM. By using such a nonlinear system approach, I have succeeded to realize ADPCM/PCM tandem connections without loss of signal-to-noise ratio which works on many different ADPCM systems. At that point the French approach was no longer a single unique ADPCM algorithm with lossless tandem coding. Then the 32-kbps ADPCM became the world standard by the cooperation of NEC and Bell Laboratories and supports from French and Canadian representatives. Especially the Bell Laboratory people backed this approach, because until that time no world standard proposed by them was directly accepted. This is because many European countries often support each other, and therefore were favoring the French approach. In the case of ADPCM, the NEC algorithm and the Bell Labs algorithm became a world standard without European input. I believe 66% of the algorithm comes from my idea and 33% comes from Bell Laboratory, but they say their contribution is more than 50%-50%. I don't know exactly. But anyway, DSP and ADPCM were most interesting challenges I met in my life.
Marketing of the DSP chip
Geselowitz:
NEC marketed the chip.
Nishitani:
Yes.
Geselowitz:
But Bell Labs probably didn't – AT&T was probably a user of the standard but they probably didn't market it.
Nishitani:
That's right.
Geselowitz:
Therefore from a commercial point of view it was okay to share the credit because your company got the commercial benefit.
Nishitani:
There is another story. At that moment we had DSP available in the market, but Bell Laboratories had DSP though it was not available on the market. Therefore in the final phase Bell Laboratories researchers asked us to change some precision requirements that would require a much longer processing word than our DSP, in order not to use our DSP in the ADPCM market. Our boss accepted their approval and we started to make another chip for ADPCM by using programmable DSP approaches. It became a popular chip in the market too, but the chip was designed as a custom chip of the transmission division. So at that time we had two different successful DSP chips on the market, one for general purpose and the other for custom chip in ADPCM channel bank systems.
Ph.D. studies and research management position at NEC
Geselowitz:
At what point when you were doing all this work did you manage to earn your Ph.D.?
Nishitani:
For the Ph.D., the first step was the DSP and the ADPCM. The final one was image processing or video processors. We combined DSP chips in such a way that we could process video signals. By 1990 we could process video compression by using our DSP. However DSP itself became a little bit modified to have a special function for video signal processing. Therefore our system could realize software-programmable video compression by using maybe thirty different chips – huge system again. And at that time the video compression market was not big. Most people said if you sell a hundred units of video compression equipment then it was a bestseller. Our system became very popular and sold more than two thousand units. However, after that the ITU-T decided a new world standard of video compression, so we gave up our approach. But I wrote up my DSP, ADPCM and video processor work, and was awarded my doctorate. These were the three developments.
Geselowitz:
And the three together were enough for the university to award you your doctorate?
Nishitani:
Yes. But they counted on several published papers which were not directly connected to the above three and they also counted some international conference papers
Geselowitz:
When you finally got your doctorate what was your position at NEC?
Nishitani:
Research Manager, I believe. Yes.
Geselowitz:
You had moved up into management.
Nishitani:
Yes.
Geselowitz:
Mr. Tomozawa didn't come to you anymore to say, "I changed my mind." You went to researchers and said, "I changed my mind." Is that right?
Nishitani:
Mr. Tomozawa stayed at our research laboratory only one or two years. He soon moved to the transmission division. However, when I was engaged in the ADPCM DSP, he called me and told me cheerfully that this is your first valuable results for our company, although other research results are all valueless. Therefore, I believe that he changed his mind again. My boss had been changed from Mr. Atsushi Tomozawa to Dr. Rikio Maruta, who was once engaged in the FDM/PCM trans-multiplexer implementation. He was another Research Manager.
Geselowitz:
Then you became Research Manager.
Nishitani:
Yes.
Transition to academia
Geselowitz:
Why did you decide to move from industry to academia?
Nishitani:
I was promoted to Research Manager, then Manager and finally General Manager of NEC’s multimedia laboratories. In the multimedia laboratories we had more than two hundred researchers. However, NEC met very severe recession, due to the aftereffect of bubble economy on rocket high land price. All over Japan, companies had to make some tough decisions because of small income and deficit. The Japanese economy was suffering. As I have already told you, the budget for the research laboratory mainly came from profits of NEC. Our president at that time recently wrote an article in a popular business journal that he sent a senior vice president to our research laboratories to close down some laboratories. We met severe time. At that time a lot of people moved to the academic area. Although my laboratories had succeeded to survive, I spent very hard time for management during that period, and I finally decided to move to the academic area.
Geselowitz:
Was that in 2004?
Nishitani:
That's right for my case, but the move to academic areas had started a little bit earlier.
Geselowitz:
Have you enjoyed being in academia? How does being a professor compare to being a researcher or research manager?
Nishitani:
I was very happy as a researcher. However from Research Manager to General Manager, most of the work I did was management. When I went into the academic area, many people assumed that I had a very good technique with actual engineering skill. At first, however, I didn't have much skilled technique. I had to catch up to today's technology. Now it is becoming very interesting.
Geselowitz:
Good. You are at Tokyo Metropolitan University?
Nishitani:
Yes. Another point is that I first went to a private University in Shikoku Island for the first time. Most of student there stayed only undergraduate period. I tried hard to teach them, but the students in my labs were quickly graduated. Finally I moved to Tokyo Metropolitan University. Here, I had some graduate students.
Oh, and a very important point is that although I made a lot of inventions on the DSP and ADPCM, I don't receive any money for those patents. In the case of the DSP, it was in the 1980s and there was no practice of delivering money to the researchers. As I am now a professor, maybe I will make money to write patents now.
Geselowitz:
The company gets all the profits.
Nishitani:
Yes. The ADPCM case was a very tricky one. I said about 66% of the ADPCM comes from NEC. As we gave up the patent fee, the system could become a world standard. If not, they would consider some other way. Therefore my boss gave up the patent fees. That's a very big regret for me.
IEEE
Geselowitz:
Interesting. Your bosses may not have given you anything, but in 1998 IEEE recognized your work on DSP by making you a Fellow.
Nishitani:
Yes, that's right.
Geselowitz:
You already mentioned writing a paper for IEEE and attending ICASSP. And IEEE is one of the sponsors of ICASSP. Do you remember when you first joined or became aware of IEEE? Do you remember?
Nishitani:
Yes. I joined in ICASSP first in 1980, and then joined the IEEE, but before the DSP presentation in ICASSP’1980, I learned a lot of signal processing architectures from the publications of the IEEE Signal Processing Society. For example, Prof. Bede Liu and Dr. A. Peled of Princeton University presented a very nice architecture of digital filters using read-only memory. There were some activities at some other universities or companies on how to make huge digital signal processors. It was very interesting for me, so I subscribed to the Signal Processing Society's Transactions. After the 1980s when I made the DSP chip I wrote to present such a paper in the Signal Processing Society's conference. I also became a member of IEEE, when I participated in ICASSP82 for presenting our ADPCM system.
Geselowitz:
You joined in order to present at the conference.
Nishitani:
Yes.
Geselowitz:
How did you move from just presenting at conferences to being a volunteer in IEEE? I know that you were chairman of the Signal Processing Society Tokyo Chapter and chaired some conferences yourself. I also understand that you are now involved in planning the ICASSP.
Nishitani:
Yes.
Geselowitz:
How did that come about, that you went from being a writer publishing in conferences to actually participating?
Nishitani:
It's a very difficult question to answer. In the case of being the chairman for the Tokyo Chapter of the Signal Processing Society, many professors asked me to become a vice chairman in order to encourage industrial researchers. This is because we did not have so many signal processing researchers except speech processing area in Japan. And most of professors were engaged in theoretical speech processing. Therefore, they said that my becoming the vice chairman surely expand signal processing activities in industrial world. I agreed to become a vice president, then automatically I became the chairman after two year periods. Planning for an IEEE conference to be held in Japan also arose that same period, and since I had become the chairman of Tokyo section, I invited many people from all over the world. A lot of Koreans and Taiwanese come to visit– not only for the presentations but also for the workshops. It was a very successful, and we made a lot of profit and sent it back to IEEE.
Geselowitz:
And that made IEEE happy.
Nishitani:
Yes. I hope so. In Japan there are some retirement ages. In my case I have to retire around the year of 2012. I decided to invite many Signal Processing Societies' activities to Japan. That's the reason why I became one of the general co-chairman at the retirement age.
Geselowitz:
So ICASSP is going to be in Japan soon?
Nishitani:
Yes, it will be in Kyoto in 2012.
Geselowitz:
And you are the chair?
Nishitani:
I am the co-chair with Professor Sakai at Kyoto University.
Geselowitz:
Right. Someone has to be organizing locally, and he's down in the Kansai while you are up in Tokyo.
Nishitani:
We didn’t want to insist on having it in Tokyo, because nobody wants to come to Tokyo. It's almost like New York.
Geselowitz:
Right. I have not been to Kyoto yet, but I'm going on Thursday or Friday. I hear it's very beautiful.
Nishitani:
Yes. That's right.
Development of DSP chips and the DSP market
Geselowitz:
You have a very, very interesting story and career. Is there anything else you'd like to tell us for the tape before we finish?
Nishitani:
Yes. I was involved in the development of several DSP chips during NEC. For example the first one was the 7720. Then soon we developed the 32-bit floating point DSP. After that we had the K3 Risk like DSP, and now we have VLIW K5 DSP. Although the DSP area became a part of my management, I am very happy to become a witness of the series of NEC DSPs. In terms of DSP business, Texas Instruments becomes the DSP major player in the world. I regret I could not recover the DSP market, but the general manager of the laboratories was rather difficult position to do so. However, before I left NEC, we were working on putting a video compression engine on a single chip, as a system-on-chip approach. This became another huge market for NEC. I am proud of such an expanding market, and my role in creating it.
Geselowitz:
Like you said, it's necessarily your chip but you had the first one and now DSP is in everything.
Nishitani:
Yes.
Geselowitz:
There's DSP in this room, in this museum, really everywhere. It completely transformed the world.
Nishitani:
Yes. I believe so.
Geselowitz:
Well, that's great. Thank you very much. I really appreciate your time.
