IEEE

Oral-History:Wolfgang Mecklenbrauker

SHARE |

From GHN

(Difference between revisions)
Jump to: navigation, search
Line 665: Line 665:
 
<p>Yes. It started then at the beginning of the '80s, but then I had already left Philips Research because I had an offer to go to Vienna. Maybe I come back to certain things which happened at Philips at the time I was at MIT. Theo Claasen had at that time considered certain structures called transmultiplexers. The French Philips people headed by [[Oral-History:Maurice Bellanger|Maurice Bellanger]] in Paris, they had the idea. They published also a lot on building a transmultiplexer. This is a machine which translates frequency division multiplex telephony into time division multiplex telephony. And because you could already see that time division multiplex would be the future in the telephone business of course you must link these two areas together. That is done by a transmultiplexer. Theo Claasen had developed a certain structure for a transmultiplexer and he told me that while I was still at MIT. I was able to contribute in certain stages of this work. We were able to publish these results. Also I think some good patents came out of it which were maybe useful to Philips. </p>
 
<p>Yes. It started then at the beginning of the '80s, but then I had already left Philips Research because I had an offer to go to Vienna. Maybe I come back to certain things which happened at Philips at the time I was at MIT. Theo Claasen had at that time considered certain structures called transmultiplexers. The French Philips people headed by [[Oral-History:Maurice Bellanger|Maurice Bellanger]] in Paris, they had the idea. They published also a lot on building a transmultiplexer. This is a machine which translates frequency division multiplex telephony into time division multiplex telephony. And because you could already see that time division multiplex would be the future in the telephone business of course you must link these two areas together. That is done by a transmultiplexer. Theo Claasen had developed a certain structure for a transmultiplexer and he told me that while I was still at MIT. I was able to contribute in certain stages of this work. We were able to publish these results. Also I think some good patents came out of it which were maybe useful to Philips. </p>
  
<p>This was really the key consideration to think more deeply about linear time varying systems. We were able to show that decimation and interpolation, which was already developed at that time by Ron Crochiere, [[Oral-History:Lawrence Rabiner|Larry Rabiner]] and [[Ron Schafer Oral History|Ron Schafer]], were systems which you should call linear but time varying. We found general statements about these systems, and because there was at that time a very curious statement that the complexity of one of these operations, I'm not sure at this moment which it was, but an interpolator was more complex than a decimator. This was puzzling us very much, and we could show that in fact both systems had the same complexity. The linkage between these two structures was a circuit transformation called transposition. It is known from normal filter design that you can apply transposition to a filter structure and you get the same filter function with the same specification but with a different filter structure. And so you for example speak about a transversal filter and the transpose of a transversal filter. In one case you use a digital filter and make the sampling rate decrease after the arithmetical operations of the filter. Then you get for example a decimator. The complexity is then reduced drastically if you interchange the sampling rate decrease with the arithmetical operations. But doing the same operation with the transposed filter structure and doing the increase of the sampling rate after the numerical operations of the filter you get an interpolator with the same complexity as the decimator. And this is documented in the publication which is published here. </p>
+
<p>This was really the key consideration to think more deeply about linear time varying systems. We were able to show that decimation and interpolation, which was already developed at that time by Ron Crochiere, [[Oral-History:Lawrence Rabiner|Larry Rabiner]] and [[Oral-History:Ron Schafer|Ron Schafer]], were systems which you should call linear but time varying. We found general statements about these systems, and because there was at that time a very curious statement that the complexity of one of these operations, I'm not sure at this moment which it was, but an interpolator was more complex than a decimator. This was puzzling us very much, and we could show that in fact both systems had the same complexity. The linkage between these two structures was a circuit transformation called transposition. It is known from normal filter design that you can apply transposition to a filter structure and you get the same filter function with the same specification but with a different filter structure. And so you for example speak about a transversal filter and the transpose of a transversal filter. In one case you use a digital filter and make the sampling rate decrease after the arithmetical operations of the filter. Then you get for example a decimator. The complexity is then reduced drastically if you interchange the sampling rate decrease with the arithmetical operations. But doing the same operation with the transposed filter structure and doing the increase of the sampling rate after the numerical operations of the filter you get an interpolator with the same complexity as the decimator. And this is documented in the publication which is published here. </p>
  
 
<p>'''Nebeker:''' </p>
 
<p>'''Nebeker:''' </p>
Line 996: Line 996:
  
 
<p>Well, thank you very much. I appreciate it very much. </p>
 
<p>Well, thank you very much. I appreciate it very much. </p>
 +
 +
<p></p>
  
 
<p></p>
 
<p></p>
  
 
<p>[[Category:People_and_organizations|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Engineers|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Universities|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Signals|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Signal_processing|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Digital_signal_processing|Oral-History:Wolfgang Mecklenbrauker]] [[Category:IEEE|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Engineering_profession|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Culture_and_society|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Workplace|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Communications|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Multiplexing|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Fields,_waves_&_electromagnetics|Category:Fields,_waves_&amp;_electromagnetics]] [[Category:Microwave_technology|Oral-History:Wolfgang Mecklenbrauker]] [[Category:News|Oral-History:Wolfgang Mecklenbrauker]]</p>
 
<p>[[Category:People_and_organizations|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Engineers|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Universities|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Signals|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Signal_processing|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Digital_signal_processing|Oral-History:Wolfgang Mecklenbrauker]] [[Category:IEEE|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Engineering_profession|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Culture_and_society|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Workplace|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Communications|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Multiplexing|Oral-History:Wolfgang Mecklenbrauker]] [[Category:Fields,_waves_&_electromagnetics|Category:Fields,_waves_&amp;_electromagnetics]] [[Category:Microwave_technology|Oral-History:Wolfgang Mecklenbrauker]] [[Category:News|Oral-History:Wolfgang Mecklenbrauker]]</p>

Revision as of 19:27, 16 September 2010

Contents

About Wolfgang Mecklenbräuker

Mecklenbräuker received his engineering degree and PhD at the technical university in Aachen; he specialized in network theory, with relevance to filters and circuit theory. He started working at Philips in 1971, working on digital signal processing and digital filters with J. B. H. Peek and Dr. Claasen. Some of his particular work was on limit cycles, digital filters and data transmission, two-dimensional filtering, reactance frequency transformation, the transmultiplexer, transposition, and his discovery of the Wigner distribution. Ca. 1980 he went to the Technical University of Vienna. He has continued to work on applications of the Wigner distribution and coding for digital signal processors, but has spent more time acting, in essence, as a research manager for his graduate students. He has been part of the Institute for Communications Engineering and High Frequency Technology, and active in the European Signal Processing Society. He compares European and American signal processing; noting military support in America, and more university-industry collaboration in America; and in Europe more of a mathematical, university background.

About the Interview

WOLFGANG MECKLENBRÄUKER: An Interview Conducted by Frederik (Rik) Nebeker, IEEE History Center, 7 December 1998

Interview # 349 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:

Wolfgang Mecklenbräuker, an oral history conducted in 1998 by Frederik (Rik) Nebeker, IEEE History Center, New Brunswick, NJ, USA.

Interview

Interviewee: Wolfgang Mecklenbräuker

Interviewer: Frederik Nebeker

Date: 7 December 1998

Place: New York City, New York

Family and childhood; post-World War II Germany

Nebeker:

So, you were born June 16th, 1938 in Dortmund.

Mecklenbräuker:

Right.

Nebeker:

Do you remember anything yourself of the war years?

Mecklenbräuker:

Yes. I remember when I was with my family in Ober-Franken, Upper-Franconia, near Kronach. My father was a soldier in the war, but he had a heart attack and I think it was 1943 when he was dismissed. He was a teacher, and after receiving medical treatment he taught in the place where we were staying in Upper-Franconia, nearby Kronach.

Nebeker:

What was he teaching?

Mecklenbräuker:

He was teaching in elementary school, children from six to fourteen. I can remember that the war ended and Americans came with tanks into this little village. We were just looking down from the hill at what was happening in the valley. But I think people were very cautious and thought that it was useless to do any dangerous things, so no bad things happened. After the war Americans stayed in the village, but then I can't remember too much because after some three months my father organized a convoy with two …

Nebeker:

Trailers.

Mecklenbräuker:

Trailers. And we put all our belongings on it. It was not just my family. There were many of us, say twelve or fifteen people, including my mother's sister. And then we moved from the Bavarian area where the Americans were, through I think the French part, and finally to Dortmund in the British part. It was a three-day journey, which today you can do by car in about five hours.

Nebeker:

What was the reason for going to the Bavarian area?

Mecklenbräuker:

Oh, the reason was fairly simple, because Dortmund as you know is situated in the Ruhrgebiet, and the Ruhrgebiet was under heavy attack, especially the area where I lived. There were chemical works, possibly producing explosives, and these were really very badly attacked. And therefore because one of our family members was also a teacher and lived in this village in Bavaria, we decided to go there.

Nebeker:

I see. And your father could get work there?

Mecklenbräuker:

Yes. I was later told of another alternative. There was a massive deportation of school children to the eastern part of Germany. My parents were not very happy to go there, because with a little bit of imagination they could see that this would mean trouble. And therefore they moved to the southern part of Germany.

Nebeker:

I have talked to others who have told me about the turmoil and hardship in many parts of Germany after the war. Do you remember those times?

Mecklenbräuker:

A little bit, but we were lucky because our house was not too damaged. We could get in after we returned from Franconia. And by my mother´s brother, he had a farm that was very nearby, we were well supported, so I can't remember starvation or similar things.

Education

Nebeker:

Were you interested in science and technology as a boy?

Mecklenbräuker:

Not so much as a schoolboy. The high school or Gymnasium, as it is called, starts in Germany when you are ten years old and lasts for nine years. This change from the normal basic school to high school was for me a sort of shock since I was not a good schoolboy in the first years. But this different type of teaching introduced a new way of thinking for me, and from that time on I was interested in mathematics, in physics, in biology, in all kinds of science- oriented things. Chemistry was not so much a favorite of mine, but I concentrated on the sciences during my 9 years of high school. I finished when I was nineteen, nearly twenty years old -- the war had delayed my education for one year.

After finishing high school and taking the final examination -- the Abitur, I was determined to attend a technical University. I remember three alternatives that I considered. The first was becoming what is in German called a Markscheider. This refers to geodesic people who work in coal mines. Because I came from the Ruhrgebiet, I was interested in what was going on there. When I realized that Germany at that time had about twenty open positions for Markscheiders per year, I decided that this was not my business. Then I had a good interview with people who give advice to new students in which ways they should direct their studies. They recommended geophysics as something that would be very close to my own interests. But my father had died when I was 17, and my mother was caring for me. She was really upset when she realized where geophysicists work, and she thought maybe I would disappear into the African woods.

Nebeker:

Or Antarctica, or …

Mecklenbräuker:

Yes. She was not very happy about the possibility becoming a geophysicist. But then I had also two colleagues who were strongly determined to study electrical engineering.

Nebeker:

Is this still in high school, or is this the university years?

Mecklenbräuker:

This decision is just on the border, having done the Abitur, but having not yet decided how to proceed further. So there were these two colleagues of mine, they were determined to study electrical engineering, and having this business with the geophysics I said to myself, "Electrical engineering might be very nice for you, too," and "Why not try that?" And we all went together to the technical university in Aachen. That was the nearest location of a technical university, some 250 kilometers from my home.

Nebeker:

So you got the diploma engineer at Aachen.

Mecklenbräuker:

Yes. And then in Aachen during the first two years the teaching in mathematics was excellent. I was really excited by the type of lectures I heard in this field. The lecturer was a very young and exciting professor.

Nebeker:

Who was that?

Mecklenbräuker:

Professor Claus Müller. He confessed sometime later that he too had played with the thought of becoming an electrical engineer. Therefore I can only say he taught me the mathematics which was really necessary for the study of electrical engineering. And…

Nebeker:

Did you come across Professor H. Döring at Aachen?

Mecklenbräuker:

Yes. I will come to that.

Nebeker:

Good.

Mecklenbräuker:

Electrical engineering courses started in the second year. I was not really very happy with the type of presentation, because I couldn't understand the essentials, and I was playing with the thought that maybe becoming a physicist would be better for me. But Aachen had organized the studies that I had to take all examinations after the first year and to take all examinations after the second year, and then the first part of my Diploma studies was finished. And having done these examinations I could see that the results were quite supportive and I decided to stay with electrical engineering. And then of course you mentioned already Professor Döring. Professor Döring was giving courses in radio engineering. He became a friend of mine and I'm still in contact with him.

Nebeker:

I had the privilege of interviewing him a few years ago and seeing his radio tube collection spanning the whole history of electronic tubes from begin of the century until the most recent developments.

Mecklenbräuker:

Yes. Because he is a Viennese, he visits Vienna once a year, and at these occasions we usually meet. But maybe I continue at this time. I was really fascinated by high frequency engineering, and I did my diploma work at his institute. The supervision was done by a young lecturer, this position was called Docent. He was having the privilege to give courses, and this was Dr. Butterweck.

Nebeker:

Butterweck?

Mecklenbräuker:

Butterweck.

Graduate studies, dissertation

Mecklenbräuker:

Having finished my studies and also my diploma work, Professor Döring took an interest in me, but at that time he could offer me only a half of a position as a university assistant (a combination of teaching / research assistant).

Nebeker:

This is after you completed your degree.

Mecklenbräuker:

I completed my Diploma degree in 1964, and he offered me a half of such a position. But I was already engaged to a girl. Maybe I'm jumping too fast, but after the third year of my study I had the privilege of taking a three-month language course in England. It was during that time that I made her acquaintance in Bournemouth at the south coast of England.

Nebeker:

Oh. Is she English or?

Mecklenbräuker:

No, she is German.

Nebeker:

I see.

Mecklenbräuker:

So we met while I was a student. At that time she was still going to school. We are married now and have three children. But when completing my studies I had the plans to marry, therefore I said to Professor Döring that one half of a position would not be acceptable for me. Fortunately he had just met Professor Bosse who had taken a position at the technical university in Darmstadt. He had worked at Siemens for about ten years, he was a very excellent scientist in network theory and filter design. So in the first stage as a high frequency engineer I went from Aachen to Darmstadt. And under the supervision of Professor Bosse I did my Ph.D. work in network theory. By accident and by a hint from the filter design group at Siemens in Munich there was an open problem in the literature, and that was my problem that I solved in my dissertation.

Nebeker:

I read here the translation of your dissertation titled "A Contribution to the Theory of Lossy Two Ports."

Mecklenbräuker:

That is correct.

Nebeker:

Can you, in a few sentences, say what that work was?

Mecklenbräuker:

High performance analog filters are usually built from inductances and capacitors. These are usually considered to be lossless, but as practice shows, each coil has its resistance and each capacitance has its leakage, a certain conductance. And so it was already an idea of Wilhelm Cauer to develop a theory that takes these element losses into account. There were already many attempts to get a solution of this problem, but they were mainly obtained in an approximate sense. And suddenly by two people, Anderson and Lee -- I think Anderson was a scholar of Guillemin at MIT -- published a paper in IEEE Transactions on Circuit Theory on the design of one ports from lossy elements. I studied this paper intensively and it was rather difficult for me to get a grasp of the paper. Then I discovered that at some point these authors were in error. I wrote my first contribution to the IEEE Transactions on Circuit Theory correcting this error. After having done that I considered that this same technique should be able to work also for the design of two ports, which is really the most important application. You want to built analog filters, and being able to design these filters by using lossy elements, that would be nice.

Nebeker:

How independent of your advisor Bosse was this work?

Mecklenbräuker:

He suggested to me the problem, and I mainly came up with the solution. I was not sure whether this solution was okay. Therefore I gave him a rough sketch of my work, and after consideration he said to me, "Okay, we will take it, and this will be your thesis work." But at that time I personally was not happy with the approach. It was rather difficult to explain and in a certain way it was clumsy. I had to distinguish between different cases which then were solved separately. That stimulated me to study the literature further, and taking up the same thing after one year of study I could give the total solution in a complete and unified way.

Nebeker:

Was that this paper here, the sixth on your list?

Mecklenbräuker:

Correct.

Nebeker:

It has essentially the same title, but it appeared in the International Journal of Electronics and Communications.

Mecklenbräuker:

It has the title of my thesis, and is maybe the summary, a readable summary of the thesis, but all proofs are not in there.

Nebeker:

I see that this received the NTG Prize.

Post-doctoral employment options; expansion of dissertation research

Mecklenbräuker:

Yes. At this stage, having completed my Ph.D. at TH Darmstadt, I was looking all over Germany from the south to the north for a possible research / development position.

Nebeker:

Were you thinking of academic positions?

Mecklenbräuker:

No, only in industry. And having applied for different laboratories, from Philips to Telefunken, I already had an offer from AEG at their research laboratory in Frankfurt, but I was not really happy with that. It was more or less a position concerned with problems in graphical layout for printed circuits.

Nebeker:

Now when one received the doctorate engineering degree in Germany at that time, was that something that only the people going into research typically earned?

Mecklenbräuker:

Yes. I think that is still the case, people doing a Ph.D. are inclined to go mainly into research.

Nebeker:

So that was the expectation when you decided to earn the doctorate.

Mecklenbräuker:

Correct.

Nebeker:

That you would go into some research position.

Mecklenbräuker:

Indeed. Being unhappy with this offer from AEG, I met by chance Professor Butterweck in Darmstadt. He was at that time already in Eindhoven at the technical university and had worked I think for four or five years in the research laboratory of Philips in Eindhoven. He came by in Darmstadt and asked, "Did you not consider a position in Eindhoven with Philips?' Until then I had not thought about that. He quickly arranged for me to have an interview in Eindhoven and I got an offer. Actually I was happy and very lucky because Philips had decided to make two offers to me. One was concerned with computer research and the other was doing research in digital signal processing. At that time I still had contact with Professor Schüßler from the University of Aachen. I had the opportunity when I was at the university in Darmstadt to intensify all these contacts. In the meantime Professor Schüßler had a position at the University in Karlsruhe, and then at the end of the '60s he moved to Erlangen. First this contact was rather loose, because the Darmstadt group was still working in analog network theory and system theory. Schüßler was determined to work in digital signal processing on the basic problems from the design of hardware to the design of filters with certain types of specifications. And so Schüßler was really doing hardware and theory together, and he built a very powerful group, as you certainly know.

Nebeker:

Now what was his connection with Eindhoven, with Philips?

Mecklenbräuker:

Schüßler had no connection to Eindhoven.

Nebeker:

Okay.

Mecklenbräuker:

But of course doing digital signal processing in Eindhoven, this helped to intensify the contacts with Professor Schüßler. After working in DSP for two years there was one of the main conferences on this field in Erlangen, but we were not able to present our results because we were a little bit too late for publication at this conference. But we attended it, and so in this sense the contact with Professor Schüßler, which also had originated at the technical university in Aachen, intensified.

Nebeker:

I see. Just to for the record, it was in January of 1971 that you started with Philips.

Mecklenbräuker:

This is correct. But I have to make a comment on my dissertation work. Philips had the nice procedure that new people applying for a position in the research laboratory had to have several interviews with other researchers, and in my case one of them was Hans Voorman. Of course I was rather proud of my dissertation work, but he was, on the contrary, he was not impressed at all. He was very well aware of some important drawbacks of filter design using lossy elements.

Orchard had published a very important relationship between losses in filters and the sensitivity to build them. It turned out that building these filters with lossy elements led to very sensitive filter properties. So filter elements had to be specified with high accuracy, and normally you have to work with say tolerances of ten percent or less. But that was far too big to make good filters with lossy elements. Learning about this result was really a big and astonishing moment for me. Having concentrated on the dissertation work and having not followed to intensively the literature belonging to the whole field, I totally had missed this very important relationship. Coming back to digital signal processing, in '71 Alfred Fettweis has published his very important publication on wave digital filters. In this concept he used the same principle, building lossless filters leads to very insensitive structures. He transferred this idea into the digital domain, building lossless structures, and had the result that for example the multipliers in these filter structures could be specified with very low accuracy. At that time that was very important. For example now, being in Vienna I very closely work together with Siemens. Siemens built a special signal processor as an integrated circuit, a signal processor for the implementation of wave digital filters, and this processor contains no multiplier at all. Because the multiplication coefficients can be implemented by shifts and a few additions, because you have only very few bits to specify the multiplier coefficients, you don't need a special integrated multiplier circuit on the processor chip.

Nebeker:

When was this built?

Mecklenbräuker:

This chip was built I presume in the beginning of the '80s, but I’m not quite sure.

Philips research group; filter design and digital signal processing

Nebeker:

Can you tell me generally something about the Philips research group? I mean this is a famous research institute. Was there any language problem for you for example?

Mecklenbräuker:

No, for the technical part there was no language problem, because typically the working language is English. But very astonishing for me, when I moved my small family to Eindhoven, they offered to me and my wife a language course in Dutch. It turned out that this was very important to us. Dutch is very similar to German, and without a formal training for Germans this results in a very broken Dutch. So this language course really helped. And of course it was also my intention to be able to discuss all technical problems in Dutch with my colleagues.

Nebeker:

Were the vast majority of your colleagues Dutch?

Mecklenbräuker:

This was indeed the case. There were Germans, there were Austrians, there were English, there were Swiss people working in the laboratory. But of course in the majority there were Dutch engineers and scientists.

Nebeker:

Can you tell me something about how it was organized and what group you worked with?

Mecklenbräuker:

This group I came into was led by Dr. [J.B.H.] Peek. He had already worked in digital signal processing and had completed his thesis work. He built a digital correlator out of very basic building blocks. By the way he did his dissertation work with Fettweis when he was a professor in Eindhoven. So, if you look into this business you will see many chance links. Not intentional, but mainly by the work which you are doing. I was engaged with the general idea to do research work in digital signal processing. Having worked for ten years in this field I must confess that I was very free in choosing relevant problems. People said to me, "Do something interesting."

Nebeker:

Oh, so you weren't given a more specific assignment?

Mecklenbräuker:

No, not at all. I started on the already mentioned work of Fettweis, because this was really maybe the closest connection to my earlier work in analog filter design. I visited Fettweis in Bochum and had a vivid discussion with him. But he explained to me that it would be impossible for him to discuss with me his latest results, because he had very close links with Siemens. And so I was with Philips and he was with Siemens. It was a very unhappy situation, and that was enough for me to decide not to work any longer in this specific area of DSP, because I considered myself not to be able enough to compete with him. He had a group in Bochum and I was working for myself, and in this sense I concluded, "I will always be second in this field, and this is not enough to get very good results."

Nebeker:

Why was Philips interested in digital signal processing at that point, in 1970? Why would they have you work on it?

Mecklenbräuker:

Philips was interested because the main application field at that time was telephony. And because in the 70´s Philips really looked very many years ahead.

Nebeker:

And saw that digital coding of telephone signals would become an important issue.

Mecklenbräuker:

Correct. Having very close connections to Bell Laboratories, and Bell Labs having already started realization of digital filters, mainly Lee Jackson working together with McDonald and Jim Kaiser. We knew about their work, and that was one of the main reasons that they decided to go into this field.

Nebeker:

But for you, they just wanted you to explore digital techniques.

Mecklenbräuker:

There were already other groups also working on digital signal processing, but the lucky situation occurred when Dr. Peek realized that the finite-word-length effects in digital filters were an important problem. At that time there were already first results in this respect, but as always you start with approximate considerations. Really good proven results were not yet available and having observed this, we went into this whole problem area of finite-word-length effects.

Limit cycles pubications

Nebeker:

Yes. And I see that in the early '70s you and Peek and Claasen published some very important papers on these limit cycles.

Mecklenbräuker:

The starting point of this investigation was, as I said, the idea to go into this business. Some time earlier there was published a paper by Sid Parker and a student of him, Hess who was a German. I am not able to trace his biography anymore. I knew only that he did his dissertation work with Sid Parker and they published a paper on limit cycles. I'm not aware of the correct title, but this is certainly in the papers which we have published. Again we found that at certain points their theoretical considerations were not quite fully based, and as always if you look into a business more carefully you find that not all problems are solved. There are open problems and suddenly you see that a whole new world is opening and you have the opportunity to dig deeper at this point. And lucky for me, when I was still doing my dissertation work in Darmstadt I had attended a course on the stability of nonlinear control systems. It was concerned about all these methods described by Popov criterion, Ljapunov functions to get a stability of difficult nonlinear systems. And having realized that these finite-word-length effects are leading quite in the same direction, I was also lucky to find a publication by a Russian named Barkin. Looking into the business a little bit more deeply, there was a connection with the earlier mentioned Popov criterion which I knew from the stability theory of nonlinear control systems. That led to this publication in IEEE Transactions on Circuits and Systems on frequency domain criteria for the absence of the zero input limit cycles.

Nebeker:

Okay.

Mecklenbräuker:

So we could really give necessary and sufficient conditions -- no, only sufficient conditions for the absence. Necessary conditions are very difficult. Sufficient conditions. We could show that certain types of nonlinear structures, for example having one quantizer or two quantizers in the very basic second order section to build a recursive digital filter, with certain coefficient values are unable to sustain limit cycles at all. That led to this paper, and we were able to generalize many approaches from Mr. Barkin and were able to make really very good estimations on the coefficients where limit cycles are possible and where limit cycles are impossible.

Nebeker:

Now I know that these papers were very important, because they were reprinted in important places.

Research collaborations at Philips

Nebeker:

And how did you and Peek and Claasen get onto this limit cycle question?

Mecklenbräuker:

I mentioned already earlier, I believe it was Dr. Peek´s idea to look into this business, and in industry of course competition between researchers can be quite high. After some while I think we realized that all of us could work together quite intensely when this competition could be elevated somehow. So we decided all papers which came out of our common work would always have an alphabetical order of our names. So you see all the papers are by Claasen, Mecklenbräuker and Peek. This has nothing to do with less important work or ideas of Dr. Peek. This was always our type of publication.

Nebeker:

Now you said earlier that Peek was your supervisor at Philips?

Mecklenbräuker:

That is correct.

Nebeker:

What was Claasen's position?

Mecklenbräuker:

Theo Claasen was a colleague of mine. He started at Philips nearly at the same time as myself. During the first year he had to do his military service in the Dutch army, and after one year he did work in low distortion amplifiers for home TV systems. Also in this type of work nonlinear effects within these amplifiers are very important, and he was also very successful in this field. He built a good basis.

Nebeker:

Okay.

Mecklenbräuker:

He built a good basis for a description of the nonlinear effects in these amplifier circuits by a Volterra series approach

Nebeker:

Volterra series?

Mecklenbräuker:

This originated already from Wiener. He had applied this type of approach to nonlinear systems, and as I said already, Mr. Claasen was able to develop a systematic description and methodology to do this, to connect this with measurements. As I can judge from the outside, I had the impression that this was very successful. Mr. Claasen spent half a year in Sweden as leave of absence from his work at Philips in Eindhoven. He met in Stockholm Professor Zetterberg. Professor Zetterberg offered Mr. Claasen to move his thesis work to this nonlinear description of amplifiers as it was developed by Mr. Claasen. But I think having nearly completed his dissertation, a quite identical article of Bussgang was published in the Proceedings of IEEE. This occurred in the middle of the '70s, and it meant that Mr. Claasen could no longer use his results for a thesis work. This fact was quite ...enttäuschend.

Nebeker:

Disappointing?

Mecklenbräuker:

Disappointing for him. Thank you very much. After review of the whole work which Mr. Claasen had done so far, they decided that he should write a thesis on the nonlinear methods to analyze finite-word-length effects in digital filters. Although the common work was done by Claasen, by Dr. Peek and myself, we said that Mr. Claasen´s own part of this total field is so big that he should be able to write up a dissertation on his part. That was done. The thesis had to be defended publicly at Stockholm KTH, and one of the opponents was Jim Kaiser.

Nebeker:

Oh.

Mecklenbräuker:

It was quite a heavy discussion, but Mr. Claasen.

Nebeker:

Passed.

Mecklenbräuker:

Passed the examination with highest degrees.

Nebeker:

What was the working relationship at Philips with the three of you? Did you work largely independently and then.

Mecklenbräuker:

No. Mainly in discussions. The group had their rooms very close together, and of course sometimes we parted and every one of us considered certain problems on his own. Then after a while we had a joint discussion, putting as in a puzzle the different results together and observing that certain problems might be not yet solved completely, we tried to make a whole picture of it.

Nebeker:

Was this work tied to application?

Mecklenbräuker:

Yes.

Nebeker:

I mean certainly in a general sense, but in particular applications?

Mecklenbräuker:

We did no hardware. All our work on limit cycles was done theoretically or by simulation. There was another group which realized in effect digital filters in hardware, and this resulted in the publication which is here, "A New Type of Digital Filter for Data Transmission." This is not a very good title, because if you read this title you will not get an idea what is happening in this paper. I have to explain that. It was the idea of the first author, Piet van Gerwen, to code the impulse response of a digital filter as in delta modulation just by using ones and zeroes. We enlarged the possible values which could be used to encode the impulse response by using also powers of two, say the first power of two, the second power of two and so on. This type of digital filter worked very well, because in data transmission also the data are binary, and so you have nearly no multiplications; you just shift the data for one or more steps. This is the working principle of this type of digital filter. This group was, as I said, really also doing hardware. They continued to build also a recursive digital filter with strange nonlinear effects, and we were able to show that our results on nonlinear effects in recursive digital filters could explain these effects in their filter realization.

Nebeker:

In that actual filter.

Mecklenbräuker:

Indeed.

Nebeker:

Now, the title there says it is for data transmission.

Mecklenbräuker:

Correct.

Nebeker:

As opposed to telephony.

Mecklenbräuker:

Yes.

Nebeker:

So that certainly was an interest of Philips at the time.

Mecklenbräuker:

This was a group concerned mainly with problems in data transmission, and in this sense there was a certain collaboration in the field because these people were interested in our work and we were able to apply our results to their filter structures.

Nebeker:

How often did it happen at Philips that you would consult with other groups there?

Mecklenbräuker:

It depends of course on your own intentions. For some groups it was not very easy to collaborate because these groups were closed units, the leaders of these groups intended to form closed groups. If that happens then all results of the group of course come from this group, and this is also important if you present the results for a director or a group of other people. But on the other hand, especially with Mr. van Gerwen, the discussions were always very open, and also with other colleagues of the same group, so in fact we had no problems with that type of policy.

Nebeker:

One of the things that I've heard about the research group at Bell Labs, is the way that most of them knew many others doing other things there and there was a great deal of, "Oh, I've got some problem now with data transmission. Okay, I'll talk to that person."

Mecklenbräuker:

I agree.

Nebeker:

And there were a great deal of these informal contacts all the time.

Mecklenbräuker:

That is very important.

Nebeker:

Was that also the case at Philips?

Mecklenbräuker:

Indeed, that was also the case within Philips Research. They had a very nice series of presentations once a week. Every Thursday morning starting at nine and lasting until ten o'clock there were always several talks for all researchers of the laboratory. For example having stayed for one year at the laboratory, usually these newcomers were asked to give a presentation of their work. The same happened with me. After these presentations there was always a more or less intensive discussion or as in my case, colleagues suggested a certain collaboration. That happened also in my collaboration with Dr. Tullio Rozzi. He was working in the microwave group, and following his presentation a very intensive collaboration resulted. He worked on the microwave part of the problem and I could make the connection to specific results of network theory.

Microwave theory and network modeling

Nebeker:

Since that paper, "Wide Band Network Modeling of Interactive Inductive Irises and Steps", received the Microwave Prize, could you expand just a bit on that work?

Mecklenbräuker:

Dr. Rozzi had the idea that the propagation of the different modes which exist in a wave guide can be modeled by different transmission lines, each working for one transmission mode. Considering irises and other types of discontinuities in these wave guides he had the idea to model these as an infinite network consisting of lossless concentrated elements. Having set up this picture which was physically very correct, of course there was a problem how to approximate the network. I could contribute to this type of work, and the total result seemingly has impressed others so much that they awarded us this prize.

This was my last contribution to microwave theory. It had a connection to the high frequency work which I was doing at the university in Aachen. There I had done my diploma thesis work in microwave networks, and this helped me to a large extent to understand Dr. Rozzi´s language. And on the other hand having studied network theory for my Ph.D. thesis work in Darmstadt, this all fitted together in this type of wave guide problem.

Nebeker:

Well that is a good example of having the right background for a particular problem.

Mecklenbräuker:

Precisely.

Visiting professorship at MIT digital signal processing group

Nebeker:

Right in the middle of your ten years or so at Philips, you spent a year at MIT with the signal processing group. Will you tell me a bit about that?

Mecklenbräuker:

That was a quite interesting opportunity, because having stayed at Philips already for three years, I was still hesitant whether Philips liked my presence or not. I had no direct feedback on my work, although I had the impression that it was quite successful. And having already started to make contacts going back to a university, in fact return to Darmstadt, the department of EE offered me a position while I was still with Philips. Just before taking a decision in this respect I got the offer from Philips to go for one year to MIT. So for me the picture was very clear. Of course I would accept this offer. It was a really exciting year to be able to work in the digital signal processing group of Professor Alan Oppenheim. At that occasion I made the acquaintance with Russ Mersereau. He was already working in two-dimensional signal processing at that time. Jim McClellan had a very brilliant idea to transform the design of a nonrecursive linear phase one-dimensional digital filter into a design of a two-dimensional filter. For example the frequency behavior of a one-dimensional filter is transformed into the frequency behavior of a two-dimensional filter. This step is so important because generally two-dimensional nonrecursive filters have quadratic complexity. Their impulse response is a quadratic array, and therefore the number of parameters is growing quite radically with the degree of the filter. Being able to reduce the design step into a one-dimensional problem helped very much.

Nebeker:

Had you worked with two-dimensional filters before you went to MIT?

Mecklenbräuker:

No, not at all.

Nebeker:

And that was something they were working on there.

Mecklenbräuker:

Yes.

Nebeker:

So you joined them.

Mecklenbräuker:

It was my intention to work on problems which were interesting to them. And that was for me the most important point, being able to work with them closely together. Again at that time it was important for me to have this background of lossless filter theory because in filter theory you are able to transform for example low pass filters into other types of filters. A specific type of transformations called reactance transformations not only transforms the frequency behavior but simultaneously it can be used also as a circuit transformation. This is a transformation of the frequency axis, and having additionally the effect that coils and capacitors can directly be replaced by parallel or by series resonant circuits. So for me the fact that you design a frequency transformation was immediately linked with the fact that there should exist also a structure that realizes this frequency transformation. Jim McClellan transformed the design of one-dimensional filter into one of a two-dimensional filter, and the two-dimensional filter with its quadratic array of its pulse response was then realized in hardware or in software. So the complexity of the filter structure was still of the N-squared type. Only the design had low complexity, not the filter structure.

Nebeker:

Right.

Mecklenbräuker:

Complexity.

Nebeker:

Yes.

Mecklenbräuker:

Having realized that that was a frequency transformation, I said to myself, "There must be a structure, and we must be able to find a structure which has this low complexity." Having told and explained this to Russ Mersereau, indeed we found the structure and our first publications went into this direction. But we were not the best in this business because some time later Jim McClellan found a still better way to use our idea and to find even better structures.

Nebeker:

Now these articles that you and Russ Mersereau have authored, there were a couple of them, "The McClellan Transformation for Two-Dimensional Digital Filtering," these were very well received.

Mecklenbräuker:

Indeed. Tom Quatieri, he was at that time I think a masters student working with Russ Mersereau, and he was working on this problem. In a short time in this collaboration -- Russ Mersereau, Tom Quatieri and myself -- we got very interesting results published in these papers. There was a certain period that Tom was a little bit embarrassed because he said, "After all it is my masters work, and you guys, you have taken all these results from my work." But that was really not true. I'm not quite sure on the problem area of Tom Quatieri´s Ph.D. thesis, but he is still working at MIT Lincoln Laboratories and is doing successfully very, very advanced work in speech processing and coding.

Nebeker:

Perhaps we should move a little bit more swiftly in order to get up to the present.

Mecklenbräuker:

Okay.

Nebeker:

So you had this very fruitful collaboration coming out of your year at MIT.

Mecklenbräuker:

Right.

Philips assessment of digital television applications

Nebeker:

Is there more you'd like to comment on about the MIT?

Mecklenbräuker:

Yes. Coming from MIT, having this experience in doing two-dimensional filtering, there was immediately the idea that this should have applications in television, because this is the main business of Philips. There were already people doing digital television at that time, but I had the idea what we proposed at that time was too advanced. So we had for some time intensive discussions, but because there was no direct or intensive response from the television people we just left it. We said we do not see any possibility to pursue this two-dimensional filter business and we do not want to work here while for example television people could help us with certain problems much easier and much quicker than we could go through all these different considerations which are necessary for applications in television. So we left.

Nebeker:

What you're saying is that it wasn't the right climate for digital television there in the mid-'70s.

Mecklenbräuker:

Right.

Nebeker:

And because you weren't getting the response from the television people.

Transmultiplexer research at Philips

Mecklenbräuker:

Yes. It started then at the beginning of the '80s, but then I had already left Philips Research because I had an offer to go to Vienna. Maybe I come back to certain things which happened at Philips at the time I was at MIT. Theo Claasen had at that time considered certain structures called transmultiplexers. The French Philips people headed by Maurice Bellanger in Paris, they had the idea. They published also a lot on building a transmultiplexer. This is a machine which translates frequency division multiplex telephony into time division multiplex telephony. And because you could already see that time division multiplex would be the future in the telephone business of course you must link these two areas together. That is done by a transmultiplexer. Theo Claasen had developed a certain structure for a transmultiplexer and he told me that while I was still at MIT. I was able to contribute in certain stages of this work. We were able to publish these results. Also I think some good patents came out of it which were maybe useful to Philips.

This was really the key consideration to think more deeply about linear time varying systems. We were able to show that decimation and interpolation, which was already developed at that time by Ron Crochiere, Larry Rabiner and Ron Schafer, were systems which you should call linear but time varying. We found general statements about these systems, and because there was at that time a very curious statement that the complexity of one of these operations, I'm not sure at this moment which it was, but an interpolator was more complex than a decimator. This was puzzling us very much, and we could show that in fact both systems had the same complexity. The linkage between these two structures was a circuit transformation called transposition. It is known from normal filter design that you can apply transposition to a filter structure and you get the same filter function with the same specification but with a different filter structure. And so you for example speak about a transversal filter and the transpose of a transversal filter. In one case you use a digital filter and make the sampling rate decrease after the arithmetical operations of the filter. Then you get for example a decimator. The complexity is then reduced drastically if you interchange the sampling rate decrease with the arithmetical operations. But doing the same operation with the transposed filter structure and doing the increase of the sampling rate after the numerical operations of the filter you get an interpolator with the same complexity as the decimator. And this is documented in the publication which is published here.

Nebeker:

Moving along here.

Mecklenbräuker:

Yes. This is…

Nebeker:

Number thirty-six in your Veröffentlichungen.

Mecklenbräuker:

Right. "On the Transposition of Linear Time-Varying Discrete-Time Networks and its Application to Multirate Digital Systems." This is a conference paper, and we published this also in Philips Journal of Research.

Nebeker:

Number thirty-one on your list.

Mecklenbräuker:

Right. And this came out of the work on the transmultiplexer which had started while I was at MIT. The paper is published in '77 and this gave us the idea that you should consider multirate digital systems from this type of view of transposition.

Nebeker:

I see. And has that been a very fruitful way of new approach?

Mecklenbräuker:

Yes, yes. In fact one of the first books in this field was written by Larry Rabiner and Ron Crochiere, and they used this paper as one of the fundamental chapters of their book.

Wigner distribution discovery, applications in signal processing

Nebeker:

That's a tribute. What else would you care to comment on in those years at Philips in the late '70s?

Mecklenbräuker:

Maybe the most exciting thing that happened during my Philips time was the personal discovery of the Wigner distribution. The Wigner distribution has met me three times while I was at Philips. In 1973 Professor de Bruijn, a well known mathematician of the University in Eindhoven, gave a talk at the Philips Laboratory on the Wigner distribution. I attended this talk and being not prepared for this type of considerations I could not understand what he was talking about. Therefore for me it was a problem where I said maybe the mathematicians can understand that but as an engineer I am unable to grasp it. When I came back from America, from this one year stay at MIT, there was within Philips Research organized a grand tour for the mathematics advisors which the laboratory had from the universities. Mr. de Bruijn was one of them. And they visited also the group of digital signal processing of Dr. Peek, Dr. Claasen, and myself. Already during this discussion de Bruijn mentioned that the Wigner distribution might be important for us. When the discussion was over I said to my colleagues I tried already five years ago. I have the impression that I do not understand that material, so we did not look deeper into these mathematical methods.

By chance, thereafter I discovered in the library of the lab the just finished Ph.D. thesis of a young mathematician. He had done his thesis work with Professor de Bruijn. His name was Guido Janssen. And I said, "I must be able to understand what a young man is writing on the Wigner distribution." Of course I was wrong, because this was a mathematical thesis, and the mathematical part was unclear to me. But he had written an engineer's view as an introduction to the Wigner distribution. This was for me the key to open the whole field and I discovered that the Wigner distribution was a time frequency description of signals having many relations to different types of short time descriptions like the spectrogram and so on. Out of this whole work came what is published in three papers in the Philips Journal of Research. I remember that we said, "We have two possibilities for this publication." We could send it to the IEEE Transactions on Signal Processing, but this might take quite a long time, because the type of work was quite advanced, was new, and we had the fear that the review process would slow down the publication, and so we decided to publish it in Philips Journal of Research. Of course this journal has also an internal review process, so mathematicians and physicists and engineers are reading these papers before publication, but the review is done quite...

Nebeker:

Relatively fast.

Mecklenbräuker:

Fast. And so the papers were published in this journal. It is a pity that the journal itself doesn't exist anymore. Because of certain reasons certainly linked to economics Philips has decided to stop this publication. And if you now look into the literature these papers are really not mentioned very often. It was not the mainstream publication journal in the field, and today it is not, cannot be found in any library. And so I think the most rewarding statement that I can make is the Wigner distribution is now accepted in the signal processing.

Nebeker:

Community.

Mecklenbräuker:

Community, and many people are working with it and applications are coming slowly. But in certain areas it is proven that this type of presentation of signals is very fruitful.

Nebeker:

And you and the others at Philips were the first ones to bring this to signal processing?

Mecklenbräuker:

I would not say so, because really the application in the signal processing field was already in the mind of this mathematician, Professor de Bruijn. He had considered this in one of his papers from 1973. It is a pity for me that I did not look at that time into his publication. That was one of the biggest mistakes I made. In this paper he gives a view that the Wigner distribution is a type of mathematical representation which is similar to that used by composers if they note their music in bars for the different notes, and this is also a kind of time frequency representation. So this view was very clear to him, and as I said already, it is one of the really big mistakes that I did not look into his paper. Because at that time it would have become clear to me what it meant for signal processing. I would have been able to look into this business five years earlier. But having discovered this at the end of the '70s and published in 1980, I think it was still a big contribution.

Nebeker:

And I see that just last year you and…

Mecklenbräuker:

Mr. Hlawatsch [pronounced LAH-watch].

Nebeker:

Hlawatsch put out a book on the theory and applications of the Wigner distribution.

Mecklenbräuker:

Right.

Nebeker:

So you say that it's becoming more widely used in signal processing?

Mecklenbräuker:

That is correct.

Nebeker:

Is this a good point to ask about your move to Vienna?

Mecklenbräuker:

Yes. The move to Vienna had nothing to do with this Wigner distribution business, because when I gave my presentation in Vienna I talked about this transposition type of view to linear time varying systems. And while I was still at Philips, the Wigner distribution came up and it was my intention to follow or to initiate this type of research in Vienna. And I must say that this group, led by Professor Hlawatsch, is very successful.

Nebeker:

Was he there at the time, at Vienna?

Mecklenbräuker:

Yes. He was my first diploma thesis student.

Nebeker:

Oh, I see.

Mecklenbräuker:

At the beginning of the '90s I think he spent a year at the University of Rhode Island working closely with Professor Faye Boudreaux-Bartels. They still have a close relationship and produce publications in the field from this collaboration.

Transition to academic employment, Technical Univ. of Vienna

Nebeker:

Had you been looking for an academic position in the '70s?

Mecklenbräuker:

I was looking for an academic position, that's true, but I must say I never pursued this with much energy. So at that time it was not usual that these positions were open, published or advertised. It was usually so that somebody puts a question to a multiple of candidates. Then they would apply, and already this type of approach was an indication that such a person was a serious candidate for the position.

Nebeker:

Yes.

Mecklenbräuker:

In this sense I was approached in Germany, in the Netherlands, and also in Vienna, and after some while you get used to this type of procedure. Professor Döring gave me advice at the beginning of my industrial career, he said, "You have to go there for each possibility which is offered to you, you should go and look and try to be one of these candidates, and it will happen to you. You never know where, you never know when, but I'm sure that it will happen." And so this was in fact the case with Vienna. I am quite aware there is always a certain stress in this situation. When I gave my talk in Vienna I had already ordered in the hotel where I was staying a card for the opera. The performance started at half past seven, and at that time the commission was still discussing with me different aspects from my talk or my personal views on certain types of problems which they had. And I was really under stress, because the opera performance was beginning. And I thought, "Maybe after one half of an hour I can go there after the break," but they wouldn't let me go, and so I missed the whole opera performance. I had to pay the card, but I had not the pleasure to listen to the opera. And so as it goes after some while I got the information by telephone that the offer was given to me, and then you know that a certain stage of consulting or...

Nebeker:

Negotiation?

Mecklenbräuker:

Negotiations, thank you, will start. To begin with you have to make up your mind based of course on the research money, on the number of positions which you get, on the rooms which you will be offered, and this takes quite a while. And one of the most important reasons that I took the position was that I turned forty-three years old at Philips, and I saw the group growing older. So I had the impression that while we were doing quite important things, the connection to young people is not in a sense as possible as it would be as a professor at a university. And that was for me the most exciting aspect of the offer of this position. I must say and now for eighteen years in Vienna that is really the most rewarding aspect of this position, being able to find always interested, well-focused students, being able to work hard, they are motivated, and after some while they get their own results and that is one of the best personal experiences which I have in this position.

Nebeker:

So you've had many students at Vienna?

Mecklenbräuker:

Right.

Nebeker:

Becoming a university professor there are a lot of other obligations, and at Philips you were a full time researcher.

Mecklenbräuker:

Correct.

Nebeker:

And now you have all these other things.

Mecklenbräuker:

Yes.

Nebeker:

What has that meant for your own research?

Mecklenbräuker:

Of course I have to admit that my own research has slowed down. I am more or less at this moment a research manager. But being able, and that is a great experience from my research work at Philips, to have a type of feeling in certain situations that I can lead people in the right directions. Maybe these directions are not so open for these young people. And of course you personally are also not sure what comes out of your advice, but in many cases you see that you felt certain things that these young people are able to prove. And in this sense I think it's a type of multiplication effect which happens in this position. Also if you are not doing bits and bytes at your desk or with a computer, you have these continuous conversations, these continuous discussions with young people. And you have the impression that in this sense you really can lead people who are not at this moment so self-assured and they need this support to become personalities which find their own way.

Institute for Communications Engineering and High Frequency Technology

Nebeker:

Right. And you have also been part of this group, the Institute for Information Technology and High Frequency Technology. Is that how it's translated?

Mecklenbräuker:

Yes. Communications Engineering is the first part.

Nebeker:

Communications Engineering and High Frequency Technology?

Mecklenbräuker:

Correct.

Nebeker:

Could you tell me a little about that group?

Mecklenbräuker:

Yes. This in fact is an institute which has come out of a fusion of two separate institutes, namely one for communication engineering and one for high frequency engineering, a very huge field. And this was put together because the predecessors of both positions, they left nearly at the same time their institutes and the government was interested in building larger institutes, not these one-man institutes which are usual at German universities. So these two institutes were put together and at the beginning it was merely an organizational advantage. There was not such a big collaboration between these different parts of the new institute. I started digital signal processing from my side as one of the main activities in this institute and my colleague, Ernst Bonek, he got his position in 1984, and he pushed at the radio engineering part of the institute the mobile telephony.

Nebeker:

Good choice.

Mecklenbräuker:

Yes. And as it is, now it is really lucky for both of us, because if you today look at mobile telephony, I always say of course the radio engineering part, it gives you the connection through the base station, but what you have in the mobile telephone handset, this is a special computer. And digital signal processing is one of the main components which you have in this system. So in this sense the institute is now in a situation that we, in the communications part of the institute, are moving our interests into this time varying stuff. Because the mobile telephone channel is one of the most difficult channels which you can imagine. Because of all these moving things, if you move fast you have Doppler effects and these different reflections in the urban area or the suburban area and so on. This is really a very difficult communications engineering problem area. On the other hand, the radio engineering people considering now space division multiplex or smart antennas and so that the antenna can focus on the different users in a radio cell, they are doing a lot of digital signal processing. Because all the algorithms which are necessary for the steering of these antennas is digital signal processing. So, as you already said, the choice had not been better in this sense, and this field is really one of the best fields for digital signal processing and for radio engineering. So we are really at this moment in a very happy situation.

Nebeker:

And came into it at a good point also.

Mecklenbräuker:

Yes.

Technical University of Vienna career highlights

Nebeker:

In your eighteen years or so at the Technical University of Vienna, what are some highlights that you would mention?

Mecklenbräuker:

Highlights are being able to see applications of the Wigner distribution. That is certainly true. Highlights are also what we are doing in automatic code optimization, mainly this is a quite different business. We have not talked about it here. For more than ten years, we are producing automatically optimized code for digital signal processors. If you have the task to make an optimized code, which means it should be optimal in time and memory, and the code should be as efficient as possible, then at the present time you have really only one possibility which is to code by hand. You have to write this whole program in assembly code by hand. And in industry this is a very time-consuming activity. We are considering tasks where industry says they are doing this in three or four months. And having a bright idea in this time means that you have to change the whole code, and it takes another month to have the final result. A bright student of the institute had the idea that a certain type of optimization should be applicable. He pursued this, made it also for his Ph.D. thesis work in this field. This was 1993.

Nebeker:

What was his name?

Mecklenbräuker:

Dr. Wess.

Nebeker:

Wess?

Mecklenbräuker:

Yes. And it happened that at that time Siemens visited the institute with quite different questions. I think they came to us because they knew that we were engaged in speech coding. We gave an overview of what we did, and only by chance they told us about their problems in producing these optimized codes. And we said we can do it. They were very skeptical. These were the Siemens people in Austria. There's a big factory for integrated circuits in the southern part of Austria in Villach, in Carinthia, and there is also a quite large development group which is called Entwicklungszentrum Mikroelektronik, EZM. We were able to attract a new student. He was able to write within three years an optimizing compiler for their purposes. It was a very interesting time, because not many companies are willing to spend three years until the first results come out.

Nebeker:

Did they sponsor this research? Siemens?

Mecklenbräuker:

Yes, yes. They sponsored this research. They still do it, and I'm happy to say at this time they have launched a new type of processor which is the Carmel DSP core. It was launched quite recently, and has a very complicated hardware structure. But we are engaged now to produce also an optimizing compiler for this processor.

European Association for Signal Processing, IEEE Signal Processing Society

Nebeker:

I know you're very active in the European Association for Signal Processing, president of the Administrative Committee, is that right?

Mecklenbräuker:

Yes.

Nebeker:

Can you tell me the importance of that organization which of course might be seen as a rival to the IEEE Signal Processing Society? Although I know they have collaborated on many things over the years.

Mecklenbräuker:

Yes. It was founded in the late '70s by Murat Kunt [pronounced koont]. He started with a signal processing journal published by Elsevier, and of course to support this journal the European Association for Signal Processing was founded. When I heard about this I was really very skeptical at that time. But some twenty years later I think Signal Processing is a well known journal, and has established its position in the field. It is certainly a difficult undertaking at this time. It is a competitor to IEEE, that is true, but on the other hand being able to publish this journal and having the intention to strive for high quality, I think this is the best proof that this position can be filled and Signal Processing was able to do this.

Nebeker:

I know you have an annual conference and other meetings.

Mecklenbräuker:

Right. We have a biannual conference, and we try to have it so it's not in direct competition with ICASSP if possible, because ICASSP is I think every sixth year in Europe, and because it is an even number it has turned out that ICASSP is always in the odd numbered years in Europe, and we with the EUSIPCO are always in the even years. So in this sense there is not a direct competition, that would not be very promising. This conference is of course mainly European, but also it attracts people from other countries, of course Americans, Far East people, but from the number of attendees it is approximately fifty percent of ICASSP.

Signal processing communities, industrial research support in Europe

Nebeker:

From your perspective of decades, several number of decades in signal processing on the European side, except for the one year at MIT, and you were at Santa Barbara also I know.

Mecklenbräuker:

Yes.

Nebeker:

But, most of the time in Europe. Has there been a different flavor to signal processing in Europe, a different style, or has it been very a single community would you say?

Mecklenbräuker:

No. I think until the end of the '80s of course America, the Department of Defense was one of the main financial supporters of microelectronics and signal processing. To mention only these two most important areas which have to be mentioned in this connection.

Nebeker:

Yes.

Mecklenbräuker:

In Europe of course there were university groups, two of them I have already mentioned, Schüßler in Erlangen, and Fettweis in Bochum. There are many more groups now specializing in different fields at the universities, and industry I think has also seen that digital signal processing is one of the main driving forces for many, many applications. I think as always, Europe takes its time, is in certain things not so advanced, which is sometimes a pity, sometimes it can be an advantage. But in general I have the feeling that in this area, at least at this moment, Europe can compete with America. And coming back to mobile telephony, I think the worldwide success of GSM is a good indication that having united Europe from the technical point of view gave a basis for a system which is now applied in many, many countries.

Nebeker:

I've heard one characterization of European engineering that the people typically have stronger mathematics backgrounds than American engineers. And I heard for example that the Fettweis wave filter idea had a better reception in Europe because so few people in the States were really prepared to deal with it. Would you agree with that characterization?

Mecklenbräuker:

Yes, yes. That might be one of the difficulties to really grasp the ingenious way to look into these problems, because many people in signal processing, at least at the beginning of the '70s, were glad to get rid of the whole difficult mathematical structure of the analog network theory. Considering quite different structures and having developed quite different filter structures which had also to be optimized by the Chebyshev criterion for example, they developed the Parks-McClellan algorithm for filter design. I think that was one of the reasons that these people maybe were not yet able to see this.

Nebeker:

Are there any other general statements one could make? I mean you pointed out the military research in the United States was a big part of the picture. How about the ties to industry or the amount of industry support?

Mecklenbräuker:

I think American universities are working in an easier way together with industry. I do not know exactly the reason, but from the times when I was in industry, I had the feeling that having spent so much company money into its own  research activity, it is really difficult to argue or to convince people to say what about putting some money of our research also into universities. And so this is my experience in Austria. In certain aspects it is much easier to work with smaller companies which don't have their own large research activities and which are able to work closely with the university and transfer the results into their business as with bigger companies.

Nebeker:

Yes, I can see that.

Mecklenbräuker:

And from this point of view, I must say I am really proud that now for more than six years this code generation group is able to collaborate very closely with Siemens because it is really also in my eyes a type of competence that this group has that it is able to get this money from industry for their support.

Nebeker:

Well, thank you very much. I appreciate it very much.