IEEE
You are not logged in, please sign in to edit > Log in / create account  

Oral-History:David Middleton (2000)

From GHN

(Difference between revisions)
Jump to: navigation, search
Line 33: Line 33:
 
<br>  
 
<br>  
  
=== Interest in communication theory influenced by World War II, Harvard's Radio Research Laboratory ===
+
=== Interest in communication theory influenced by WWII, Radio Research Lab ===
  
 
'''Geselowitz:'''  
 
'''Geselowitz:'''  

Revision as of 15:18, 8 May 2009

Contents

About David Middleton

David Middleton went to Harvard for a Bachelors in Physics (1942). He then worked at the Radio Research Laboratory at Harvard during World War II, working on radar counter-measures, passive and active jamming. This work flowed into work on communication theory, the study of the transfer of information, as statistically based applied physics. He did his graduate work at Harvard, then taught there, but in 1955 shifted to a career in consulting. His career has centered on noise and signal communication theory, including work lately on scattered channels and models of interference. He has been involved in the IRE and IEEE, particularly with the Information Theory Society, Signal Processing Society, Oceanic Engineering Society, Aerospace and Electronic Systems Society, and the Electromagnetic Compatibility Group.


This interview covers Middleton's academic and consulting careers, as well as his involvement in the IEEE.  Middleton describes some of his research approaches to communication theory and considers the impact of computers and the Internet on his consulting work. See David Middleton Oral History (2007) for a later interview updating Middleton's predictions on the future of communications work and discussing his development of a physics reference text.

About the Interview

DAVID MIDDLETON: An Interview Conducted by Michael N. Geselowitz, IEEE History Center, 14 February 2000

Interview # 388 for the IEEE History Center, The Institute of Electrical and Electronics Engineering, Inc., and Rutgers, The State University of New Jersey

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, Rutgers - the State University, 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:
David Middleton, an oral history conducted in 2000 by Michael N. Geselowitz, IEEE History Center, Rutgers University, New Brunswick, NJ, USA.


Interview

Interviewee: David Middleton
Interviewer: Michael Geselowitz
Date: 14 February 2000
Place: New York City


Interest in communication theory influenced by WWII, Radio Research Lab

Geselowitz:

Dr. Middleton, maybe we can start talking about your early years, your education, and how you got interested in signal and communication theory


Middleton:

Well, my early education is a matter of record. I went to Harvard College, and in due course did my graduate work there and got my Ph.D. I taught there as a special instructor in electronics and then as an assistant professor. I branched out into consulting, and eventually began consulting full time. That’s the quick version.


How I got interested in communication theory—which I’ve been doing for the last fifty years now—came out of World War II. I got my bachelor’s degree in Physics in June of 1942. I immediately joined a number of my associates my age under the direction of various senior professors at Harvard to teach the first half of a six-month course in radar for Army and Navy Signal Corps officers. Then in December of 1942 I joined the newly formed Radio Research Laboratory at Harvard. It was under the direction of Dr. Frederick E. Terman, brought in from Stanford University. I was part of a very small research group and was an assistant to Professor John H. Van Vleck. I wrote a little article on it in the IEEE Communication Magazine in 1977. It gives a little flavor of the atmosphere there in that particular group.


We were a counterpart to the MIT Radiation Laboratory—known as Rad lab—that was working on radar while we were working on countermeasures. We were in a friendly rivalry. The senior people at our laboratory communicated freely
with the senior people visiting from MIT down in the river, and vice versa. Both labs did important work. Let’s put it this way: We didn’t lose the war, and we might just have helped win it!


I was in Group G, the theoretical group. Now when you’re dealing with research, there’s always the question of where does it end? That theme will recur during this interview. I might remind you that I am trained as a physicist, not an engineer. Engineering is too difficult for me [laughs]. Actually, there are different motivations for the engineer and the scientist, and I’m in-between. In a sense, I’m an “amphibian.” This makes me a little hard to classify, and I’ve had that problem all along.


Geselowitz:

Would you call yourself an applied physicist, then?


Middleton:

Basically yes. I’ve studied quantum mechanics and so forth in graduate school, but I’m not using it. I’m primarily a classical applied physicist. My other area of interest is statistics, not the conventional statistics of most statistician, but time series analysis. I was interested in physics and statistics of norms.

These theoretical interests came out of my work at the Radio Research Laboratory, which was focused on two major problems: First, the active electronic jamming of radar; and, second, the passive jamming of radar. I had to do a lot of applied work, however. For passive techniques, everybody’s heard of “window” and the related tin foil chaff drops in Europe and Japan during the war. Our laboratory was the primary designer of this technique, which we first made as a prototype. Because of the vast quantities of material that were to be involved, it turned out that it was I, a physicist, who had to order the equipment that subsequently went into manufacturing!


I remember one time we had a ten-centimeter radar on our loop to test and we were working on jamming it. I was on an airplane going around with jamming equipment, looking at the scope. I remember feeling trapped. This was my closest exposure to the experimental side. I remember another incident when Winfield W. Salisbury was in charge of something called TUBA, which was a very high-powered microwave noise broadcaster. They were testing it. It was in the fourth or fifth floor of the laboratory building. There was a board with tools on it next to the device, which was shooting microwave radiation. The emitter was not pointed at the board, but there was enough energy coming out of the side so that the board caught fire. The iron in the tools just heated up like mad—a little dramatic incident.


Geselowitz:

I know that the Rad Lab was in “Building 20” at MIT. Where on the Harvard campus was your lab located?


Middleton:

There is a Zoology Building on the east side of campus that is shaped like an inverted “U.” It was converted for wartime use, and we had the main cross part.


Geselowitz:

Is that the part where the doors have famous relief sculptures in bronze of insects and so forth?


Middleton:

Right. And there are two famous bronze statues of rhinoceri right in front. As you faced that front part we had the left wing, and they added two frame stories on the top of it. We were not nearly as huge an operation as Rad Lab. We did have a portion of the laboratory overseas with the British. We and MIT jointly shared this British facility.

End of WWII, impact of WWII on postwar research

Middleton:

These stories have been an attempt to answer how I got started. What I’ve been doing subsequently came very directly out of this war experience. The work I was doing didn’t end when the war ended. As you know, the development of the cathode ray oscilloscope was accelerated by the war. That development was one of the founding elements of the old TV business.


There is a great deal of connection between the technical projects of World War II, for example noise reduction, and the subsequent developments. When I say reduce noise, it sounds like “get rid of the pesky mosquito.” But it is a fundamental problem in the communication process, whether you’re talking about acoustical noise or electronic noise. Now, one person’s music is someone else’s noise. The basic problem is finding the maximum, most effective use of a given communication channel. This kind of work is not concerned with what goes over the channel—the various information—only its electro-magnetic “shape,” for example, the propagation of modulated wave. To put it succinctly, as I did my book, communication theory is the study of the transfer of information from one point in space-time to another.
Now, there would be nothing to it if there weren’t a lot of factors in-between the two points in space-time which act to interfere with the transfer. It’s a never-ending battle: How can we improve our communication? It is today a hotter topic than ever, because of wireless telephony. Just getting it to work was a model of ingenuity. I’m concerned with the most basic theory of the communication link. That field has enough challenges in it for a lifetime of research, from examining the most basic things, to analyzing the much more elaborate systems that we have today.


As I said, I’m trained as a classical physicist, really sort of a statistical physicist. There’s no good word or even pair of words that accurately describe my interests. On social occasions, people sometimes ask me “What do you do?” I’ve learned just to say, “I’m a scientist of sorts—an applied physicist and mathematician.” Even then I can tell from their expression that they have no idea what I’m talking about. So we go on to other topics, such as politics, God, and sex. This is a well-known phenomenon for scientists of all sorts.


Geselowitz:

Interesting. During the war, did you continue to teach in addition to doing research?


Middleton:

No, I was teaching up until the end of 1942 and also going to school. Then for three years—1943, 1944, and 1945— I was doing the research at the Radio Research Laboratory. As I said, it was fascinating time, because I got to meet a lot of the senior colleagues like Van Vleck and others who eventually won Nobel Prizes. We were a small group. Some people had come from Los Alamos Laboratory to join us in Group G. Many of them went back to Los Alamos after the war—the career government employees.


Geselowitz:

Why? Was Group G disbanded immediately upon the end of the war?


Middleton:

Yes. The war ended in the middle of August in 1945, and we folded by the end of December. We all still had jobs, but essentially we were getting reports finished up and cleaning up the accumulated documentation of three and a half years of fairly intensive effort. Harvard still has in its collection the reports of the Radio Research Laboratory, in some cellar somewhere. I think they have actually been declassified.


Geselowitz:

Right. I know that the Rad Lab material has been declassified, so I imagine that the Radio Research Lab material has been as well.


Middleton:

Rad Lab actually publicly published a set of books; the series has twenty-four volumes. I don’t have the whole set, but I have the one on threshold signals – Volume 24, edited by James L. Lawson and George E. Uhlenbeck.


Of course, I met them and other important physicists and engineers many times. For example, we had close contact with the staff at Bell Labs. In fact, I remember meeting and going shopping with Walter Brattain. Right after the war, he came to spend a year at Harvard to discuss his work on the transistor; I don’t remember the exact year, but it was about the time that he and William Shockley and John Bardeen got the Nobel Prize in Physics for their transistor work.


Geselowitz:

They invented the transistor in 1947, but they didn’t get the Nobel Prize until 1956.


Employment and research at Harvard

Geselowitz:

So, from this anecdote I see that you must have stayed on at Harvard.


Middleton:

Yes. I had the not unreasonable thought—or so it seemed at that time—of being an academic. That was the atmosphere in which I grew up in college and graduate school. That was the expectation, so that was my ambition at the time. I went from being a teaching fellow, which was a part-time position for graduate students, to being an assistant professor. This was, in fact, not a major improvement—an assistant professor is the lowest form of academic life.


For a variety of reasons, some of them even perhaps my fault, I wasn’t given tenure. I was upset at the time, but in retrospect it was a very good thing for me. I spent the rest of my life so far being a consultant, with off and on stints as an adjunct professor.


Geselowitz:

This interview need not necessarily be chronological or linear, but I would like to keep track for the sake of future listeners. So before we go on to your consulting career I’d like to ask what you were doing at Harvard doing this stint in academia. What was your research on?


Middleton:

Basically, on the noise and signal communication theory problem. It was then that I started my book, An Introduction to Statistical Communication Theory—a massive tome!—which eventually did get published after I left Harvard, in 1960. I’m happy to say that it seems to have become in the end perhaps a classic in the field. In 1996, IEEE Press republished it in their “Classic Reissue” series.


Geselowitz:

That’s quite an honor.


Middleton:

It came out in 1960 from McGraw-Hill, and lived until 1972. Then it was out of print for a while, but I talked to a smaller publisher and had it published by them until 1996. In 1996, as I said, it was republished as a “classic” by IEEE, and that got me thinking about a sequel. I got in touch with the American Institute of Physics, and had a contract with them to do a small book. I either write a very small book or a very large one. This one is turning out to be very large one. There is going to be another nine hundred or thousand page “monster,” bringing the field up to date. It will summarize much of the work I’ve done, but of course it will not just focus on my work, but will cover much of the work in the field.


Geselowitz:

According to your curriculum vitae, this year at Harvard after the war went through 1954.


Middleton:

I left in the summer of 1955.


Geselowitz:

And the whole while you were there you were working on this signal and noise theoretical problem, which had important applications.


Middleton:

Yes—signals and noise--communication theory, as it developed. It wasn’t a full field yet.


Geselowitz:

During that period, in 1948, Claude Shannon had published his information theory paper, which is considered one of the classic papers of all time. Did it have an impact on your work at the time?


Middleton:

Yes, and this is a good point for me to make a distinction. What people call communication theory is often centered on the coding problems, which is a very significant area. It occupies lots of talent and lots of money, and has lots and lots of uses. Although I’m not a coding expert and I don’t intend to be, I appreciate coding—it may be the most important part of communication theory.


However, what I’m working on is also important—I’m dealing with the signals as they get injected in to the real world, having been encoded or encrypted or both, and then decoded at the other end. A person in coding theory is trying to encode a signal in such a way that, despite the various competing constraints of the channel, which is noisy, he can extract the message at the other end without error. Of course, a key component of Shannon’s famous theory is that if you have enough time—if you don’t send the signal too rapidly—you can always send a signal without error. From this idea has come all kinds of important applications.


So, to answer your question, Shannon’s theory did influence my work, and I have a chapter in my book about it, which is appropriate. But it’s not a coding book, and I haven’t been involved directly with the coding problem. What I’m doing I believe has impact simply because it’s dealing with the physical challenge. So for what I do, I use the expression “transmission of information from one space-time location to another.” The words “space-time” indicate the fact that you’re dealing with the propagation of acoustical or electro-magnetic waves in the physical world.


Geselowitz:

Theoretically, the information people don’t care about the channel; but in practice we know that presumably the actual, physical limitations in the types of physical noise are going to effect how one has to code and decode. The nature of the signal also impacts the exact effect that the channel has.


Middleton:

A basic question is, can we send signals where the information content isn’t in the major noise region—this is where the famous matched filter idea comes from. We worked on it during World War II. Van Vleck and I published several papers on it together. North’s also did import work, but it wasn’t published until 1963, and we didn’t know about it until after we’d tried our approach. I have a little historical note that I’m putting in my new book about that. It’s an example of an innovation or discovery taking place in two different places independently and almost simultaneously. Our paper was published in 1946, and we later with great pleasure gave North his credit. Actually, he approached the problem from a somewhat different technical point of view. We used the Schwartz Inequality approach, and he used a calculus of variation approach. In those days these two approaches looked quite different; only later you’d find out that they give exactly the same results.


The matched filter concept

Geselowitz:

Two sides of the same coin. Could you explain the historical record and the interested lay person, the matched filter concept?


Middleton:

I’ll do my best. The term matching filter has to do with the filter at the receiving end, which makes more favorable use of the channel then a filter at the transmitting end. Take the case of white Gaussian noise. Of course, this noise is “colored” in frequency, and so is much of the signal energy. As much as you can, you try to put more of the energy in the signal where the noise is weakest, and a less energy into the signal where the noise is strongest. This practical approach comes right out of the theory. So the idea of matching is to get an optimum processor. A matched filter has a certain structure that is determined by the signal and by the noise. In the Gaussian case, it’s easy to do.
Now, we extended this idea, still based on fundamental analysis. First we maximized signal to noise ratio at the output, subject to some reasonable constraints. Signal to noise ratio is an interesting concept, and it’s a very important parameter in describing performance, but it’s incomplete because it doesn’t give the probabilities of correct and incorrect decisions. This is where, for example, the work of Siebert at MIT with his Ideal Observer statistics, comes in. It was developed for people looking at the radar scopes, and he used the work by the statisticians Niemann and Pearson did to introduce the concept of probability of error and probability of correct decision into communication theory. The radar work was during World War II, but the completion of the theory in an intellectual sense came with Wald’s work on statistical decision theory, and with the work of Van Meter—one of my students—on Bayesian statistics.


Now Bayes’ ideas, as you know, incorporate a priori probabilities, and this is a great area of philosophical discussion. I think it’s getting more and more settled in favor of the Bayesian view. I’m prejudiced, of course. I have here a manuscript that is going to be published. Not by me; by Professor Jaynes (who is now deceased, alas, but it’s being prepared by his colleagues), discussing the impact of Bayesian concepts on maximum entropy principles, decision theory, and other practical areas.


I may be getting maybe a little technical here, but this is an important, fundamental problem. In a full communication theory, you have to have a priori probabilities; in other words, you have to anchor your theories about the signal in the real world. What are those a priori probabilities? In many cases, you’re going to have to deal with something that hasn’t happened yet. In the frequency theory of probability, you’ve got a long string of events, coin tosses if you wish, and you have the record, and as far as you can tell you know that the coin is an honest coin. But what is the probability of an airplane crashing right over here? You may have some air safety records, but how relevant are they? Or maybe you’re going to launch a missile, and no one has launched a missile before. What do you do? You have to do an experiment.
For example, there’s something called CFAR, “constant false alarm radar.” In CFAR there are two probabilities. You have the probability of false alarm, the conditional probability that if the signal is not there, you still think you detect it. Such false positives are common in medical tests, and this is a big problem. If you get a lot of false alarms in the test, so you have do it over and over, and try to work as many statistical angles as you can.


The other probability is the probability of correctly detecting that something is there. That’s one minus the conditional miss-probability. We can calculate this in many cases. So, there’s a probability that the data contains the signal which you’re looking for, and there’s the probability of one minus the other probability that it doesn’t. We’re really forced to face up to the fact that it’s a probabilistic question. So where do we put the threshold? . We want the false alarm probabilities to be low, but we can’t be too low or we’re going to miss things. Experience generally tells us where to put it. In statistics, you have something called H-one (H1), the alternative hypothesis.


Geselowitz:

H-naught (H0) is the basic hypothesis; H1 would be the alternative.


Middleton:

Right. The problem with most applications is people don’t know the distribution of H1’s signal and noise. In theoretical physics, we do. This is one of the great challenges that the engineer has—he knows that if there is going to be a signal there, it’s going to have a certain at least statistical structure that he can assign to it. For example, he might say, “Well, I know the noise is Gaussian, because I’ve measured it.” There’s always a measurement involved.
What our work on matched filters did is clearly state on paper that you’ve got to measure the parameters of the system of the real world you’re in. It’s a physical problem, not a mathematical one. You have to go beyond the level of just
paperwork. I’m a paperwork guy, but even I have to face these questions. The subject is fraught with nags in philosophical circles, but I think the engineers have, on the whole, been very practical about it.


Engineering experience

Geselowitz:

Let’s move on to some of your more recent work. Your early background was in physics and mathematics, but the vagaries of the war put you into an applied group that was really doing engineering. Do you feel that this prepared you for your subsequent career?


Middleton:

Definitely. It sharpened me, having to get real world answers, or at least provide the tools for somebody else to make the measurements to get those answers. What are the things we have to know? That’s still my job when I consult for different organizations. I can play a numbers game. I can say, “Here are the assumptions. Here’s the probability you’ll detect signal, here’s the probability that you’re estimates will be such and such good. Here are the factors. If you twiddle them that will change it a little.” That’s the virtue of the theoretical approach. But it has to be coupled at some point to what the engineer laughingly calls the real world.


So the answer to your question is certainly, my applied work had a definite influence. I’m not really a mathematician by inclination. I use mathematics. That’s generally what physicists do. And that’s what engineers do also; there are theoretical engineers doing work similar to my “applied physics”. Always bear in mind that at some point you have to be able to make a measurement. Even physicists say this. In other words, you have a nice little model, and it has parameters. If you can’t measure the parameters, it’s not real.


Geselowitz:

I have an engineering background. We used to say that everything was the boundary conditions. The formula looked really easy until you sat down to do the problem set and you had to figure out the boundary conditions and you realized that was really the rub.


Middleton:

That’s correct. For the mathematically inclined, the boundary conditions are also abstract constructs. They’re approximations. Then you have to try it out, and maybe you have to twiddle those a bit. So there is this game that we have to play with nature.


Geselowitz:

A game we have to play if we’re going to produce useful products.


Middleton:

Well, the level of usefulness is a philosophical problem. I don’t want to go off into arm waving, but each discipline obviously has its uses, although sometimes they are not appreciated. So the word “applied” very strongly connotes making that connection with the chaps who are going to build things or operate them, and they in turn have to think about some of these things too. So there’s a mutuality required there.


Consulting career

Geselowitz:

Would you like to say a little something about some of the high points of your consulting career? Some of the specific problems. My prejudice would be to imagine that once you left the bosom of academe, the Ivory Tower, and became a consultant, that the applied side would have loomed larger. Because presumably the needs of your clients were analogous to the military needs of the U.S. during World War II. Even though they may have brought you as part of a team where you were the theory person, at the end of the day, they wanted something that worked. Likewise, your consulting clients have a problem they have to solve or a product they have to bring to market, or something like that.


Middleton:

Well, I’ve been fortunate in a number of ways. From my point of view it’s very important to be dealing with someone who understands where I’m coming from, who is a good, I like to say, “mitten snatcher.” Sometimes it’s a third party. But that’s an essential item. I’m not going to get very far with a lecture on photo emission physics—which won Einstein his Nobel Prize—without glazing the engineer’s eye. The client’s engineers will say, “We are paying you X dollars an hour—what’s in it for us?” This happens fairly naturally. I have an area of expertise with which the client is concerned, and they invite me in. We talk about some problems, and I say I can do something about them and I tell them what I can do. I don’t mean solve the problem right there, but I tell them how we can go about solving the problem—and I show them that there is a connection to my area.


I’ve never advertised, for example. My business is word of mouth. It’s who you know. In the context of the spectrum of problems in which I have some expertise, I’m a little leery about defining it in a few words. What is that saying, “The expert is a damned fool who happens to be from out of town”? My thrust is academic, and then I try to get out of the academic mode at the appropriate point. In regard to what I’m doing as research, then that raises the question of the terms “pure” and “applied.” I don’t think anything is pure anymore. British mathematician Godfrey Hardy who wrote a book “The Mathematician’s Apology.” In it is a quote I like to use: “Anything that I do in mathematics that has a practical use, I disavow its effect.” The joke’s on him, actually—he has a number of theories, in coding theory and in number theories, and many of them have applications. So, okay, you’d better be careful.


Geselowitz:

Be careful what you do and what you say, right?


Middleton:

Well, Hardy doesn’t have to worry about it anymore. He died shortly after World War II. He was a brilliant person. He was the chap who discovered the Indian genius Ramanujan and got him to come to England after the World War I.


Anyway, there is no more separation, and if you think about it, perhaps there never was. This fact was demonstrated during World War II. In a sense it was a physicists’ war. We had both experimentalists and theoretical types, but it was the engineer who really came of age. Many engineering disciplines began at that time.


Most of the physicists wanted to get back to chasing theoretical particles and esoteric things like that. The engineers wanted to stay in the real world, but that is one of the problems I see. I feel the engineer needs the theoretical background and much more physical background. Look at the schedule an engineer has to go through to get a degree. I don’t see how, in a good school, you can have an engineering course that’s under five years long.


Geselowitz:

Well, that’s another issue that’s being debated.


Middleton:

It’s a problem of money and time, of course.


Shifts in communication theory

Geselowitz:

Right. Getting back to your case, can you give me a few examples of things that you came up with, either on the theoretical or applied side?


Middleton:

Well, as you know, science and technology go together, and the technological advances are very gradual. It’s almost continuous process. Of course, in science you sometimes get a shift in “paradigm,” to use the phrase of my classmate Thomas Kuhn.


Geselowitz:

That’s right, you would have been with the famous historian of science Thomas Kuhn at Harvard


Middleton:

Yes. He was a class act. We took courses in physics together and argued about science and its history. We had a good time.


Geselowitz:

Kuhn suggested that these great paradigm shifts were rare. How do you see them playing out in your field?


Middleton:

Well, when you look back over larger hunks of time, ten or twenty year intervals, there’s been radical change, although if I stop to look back, some of the things I wrote twenty years ago are still salient today. Of course, other things I wrote didn’t stand the test of time. I didn’t do what some musicians do. Brahms put out a lot of music under pseudonyms extensively and then disavowed it if it wasn’t well received. People do that. Which is a shame in way, particularly from your point of view—from the historians point of view. It would be very interesting and important to see all the intermediate developments.


I can’t help getting in another footnote. The composer Ludwig von Beethoven is one of my idols. Beethoven’s deafness was a godsend to his audience and to his music colleagues because he wrote things in notebooks. We can see the evolution of his musical ideas. The second movement of the Fifth Symphony had fifty-seven stabs at it before he got it the way he wanted it. But we don’t have any idea how Mozart’s mind worked. He’d sit down and he’d write the thing.


Geselowitz:

I don’t think it was the deafness. I’m not a musical historian, but you hear about Mozart. He was some kind of genius. Beethoven was also, but there’s something about Mozart. He would sit down and somehow into his head would pop the entire piece, then he would just transcribe it and be done. It was a marvel to the other musicians at the time.


Middleton:

Yes, he would just write it out, only very rarely changing it. With Beethoven, you’d see the struggle. From my point of view. The deafness in a sense heightened this ability. I’m not enough of a musician to hear the music without music, though I suppose if I’d spent more time with it, I’d improve. Brahms, for example, would sit there and train by reading scores.


Getting back to my work, everyone in my field works differently. With me it’s an incremental process like that of Beethoven. I really consider myself not terribly bright. Once in a while I have an insight after the fact—a certain intuition which I don’t really articulate verbally. I do something and it just looks right. Then I think about it, and I find some reasons that show that I’m right. It’s very hard for me to describe this process. Everything I do suggests something. I just hope keep my customers happy with what I’m doing


Geselowitz:

Could you please give me some examples to illustrate this process?


Middleton:

Because it’s been an evolutionary rather than revolutionary process, it’s hard to remember any particular light bulbs going off. Okay, I’ll give you one example. I just finished some work, published this last July, on a very common, very important problem. The channels that we talk about in communication theory are scattered channels—they scatter the radiation. This is very hard phenomenon to describe quantitatively. Clearly we have to use some kind of statistical technique to do so.


There are two essential problems, to my mind, within scattering: multiple scatter and boundary scatter. The classical mathematics of these is very complex and there’s been a lot of work done by extremely able people. These problems also rise in quantum mechanics. De Broglie got his Nobel Prize way back in 1929 in part for his efforts to diagram scattering effects that work has continued up until this day. You’re trying to count the interactions, but the radiation is a scattered and each wave is on the edge of another one. It gets to be a classical mess. There have been techniques for handling it, but in my mind none of these has been terribly satisfactory.


All the techniques boil down to monstrous computational problems. You can set up the mathematics, but it doesn’t mean anything ultimately unless you get some numbers out of them. That’s the iron rule of the theory, I think. Understanding of qualitative phenomena is important too, but science is measurement.


I’ve devised a new way of solving this problem. For the essential idea, let’s subsume them into a statistical frame where we have an input and an output and a procedure. Classical theory, which is developed in terms of operators, is very handy for thinking but not very good for calculating. Statistical theory is good for counting, but difficult to conceptualize. I’ve been able to show an isomorphism of one to one between the two. This in principle allows us to calculate different scatters—triple, quadruple—in a statistical way.


Of course what we’re interested in is the statistics of the whole set of scatters, which I’ve succeeded in doing. One has to make certain assumptions, which you then test against the reality, just as you would in classical physics. You have to do an experiment and measure the parameters. You want a theory which is parsimonious in numbers. You don’t want to have forty-six parameters and forty-six terms. It wouldn’t make sense.


I’m really very pleased with my method. I think one or two people may understand it now. I’m trying to beat the drums, to say, “Here’s what you do,” because it removes the difficulty of dealing with the multiple scattering problem in a classical way. For example, scattering off the ocean’s surface is a classical problem, and a very important one. How do we handle it? There are breaking waves, there are spectral reflections coming from each wave crest, there are shadow effects. It’s a monstrous problem—I know, I’ve worked on it as well. My technique will allow consideration of multiple scatter. The limitations are that you don’t get the micro-structure, but that’s a small price to pay for getting the right answer!


I’ll stop on that. I don’t want to get too technical. But I consider this to be very important, and I’m trying to promote it. It does involve some statistical ideas with which some people in the field with a classical approach are not familiar, so I see promoting my idea as a communication problem.


Geselowitz:

So, you have published this?


Middleton:

Yes, I said, this past July. I’ll send you a copy if you want.


Relationship between consulting work and academia

Geselowitz:

That would be great, thank you. I won’t make heads or tails of it personally, but it will be good to have in our files. This episode actually raises an interesting question about the relationship between consulting work and those proprietary products, and your visiting academic positions and those publications. Would you say that the consulting work allowed you do the small increments, and that you could then in your academic stints reflect on a little more time depth to integrate?


Middleton:

In a sense, it’s all academic. Don’t tell that to my clients! What counts for them, and for me, is that I can communicate possible solutions to them. Actually, it’s a self-adjusting process, in a way. I don’t get the contract if they don’t understand me. It has to be somebody who understands the idea, and I have to be able to communicate it and explain the basic thinking. They’re not usually concerned about whether the five basal functions here are two, but they are concerned about the payoff—what is the thinking behind it, and does it fit our kind of work? So there’s always that problem. I’ve been working to be able to communicate it. That’s a must.


Now, the academic part literally would be talking to people who are in academe, or who also do consulting but work primarily speaking at meetings and symposia. That type of consulting is not what I do, but it’s also not academic. It’s really the same talk that academics give, but their interests are different—they overlap, but they’re different. I can talk to both groups. The industrial types and the lecture-types have all been through the academy; the professors stayed in it.


Now the career requirements are different. Publication is very important for academics because it’s their track record. The manager trying to get his group to work with an academic can say this guy has done this and this and this, and his scientists and engineers can appreciate it. It’s a communication link; it’s a credential. It has a slightly different—or maybe sometimes radically different—meaning in academia than it does in the industrial world, but it’s something tangible there for people to look at.


Geselowitz:

How about the teaching side of academia? Does having students have an impact on you and your work?


Middleton:

Well, I’ve had about sixteen or seventeen Ph.D. students. Seven at Harvard and the others at various other places where I was in a visiting capacity. I tried for a while to keep up my academic connections. This is sort of a touchy business—the faculty doesn’t ordinarily welcome you with open arms, but if they feel they can get something out of it they’ll go along. I haven’t done that now for almost ten years. I’m not really interested in it at this point. I think I have established my academic credentials.


There is one slight caveat to that. When I work with younger people, my knees tell me that I’m eighty, and these guys are thirty, thirty-five, and sometimes forty, but I don’t consider myself much older than the people I work with. Therefore, it surprises me, but I realize they are not as prejudiced as I am. I use the word prejudice without prejudice, I just mean they are even more open to new ideas, so it keeps me aware of alternatives. I hadn’t thought of it that way, but now that you’ve asked the question, I would say that in a sense there is that, and it’s very good for me. There was also always some money involved in these academic appointments, but that became less important over time. The primary reason was to keep on the cutting edge of the profession.


Geselowitz:

Just to keep the chronology straight, since this interview is partly biographical, when did you leave Cambridge and how did you end up in New York? Your clients seem to be all over the world and your visiting positions were all over country, so you wouldn’t seem to need to be in any particular location.


Middleton:

My then wife and I left Cambridge and bought some land in Concord in about 1959. Then in 1971 we got divorced and I got remarried and moved to New York. That’s thirty years now. I kept an office in Cambridge until 1989, and I had one in Concord as well. It is important to be able to see people in Cambridge much of the time, but if you live there it is hard to separate your work from your personal life. That’s not a problem in New York, with its vertical living. Vertical living is different than horizontal living. There’s more isolation vertically than there is horizontally. I think people are less apt to run up and down stairs with me. I have an intercom and the mail and the telephone and things like that.

Impact of computers on consulting work

Geselowitz:

Have you noticed now a change in the way one does consulting business with the spread of the Internet and wireless telecommunications?


Middleton:

I must confess I’m a dinosaur. When I get my book finished, I will probably buy a laptop and get hooked up, but I don’t have one now. Fax I find invaluable; it really speeds things up. I fax to Tel Aviv, to Paris, to Crete and so forth. It saves me endless time. I carry a cell phone. That’s as far as I am technologically.


Geselowitz:

You know, the fax is quite amazing, but e-mail to me is an extension of a fax. The idea that you compose a bunch of information and send it off, and just like a fax you have pictures and whatever. You send it off and you’re done, and the person picks it up when they pick it up. With the phone you either get there right away, or they’re not there and they don’t get the information. Fax and e-mail solve the Japan problem that they are twelve hours different from us, so that you can almost never get them on the phone. But if you send them the fax at a good time for you, you go to sleep, and they pick it up, read it, and they respond, and it’s waiting for you when you come into the office in the morning.


Middleton:

Well, I think for business that makes sense. I debated this. I think I am a fairly intelligent dinosaur. This question comes up and my colleagues rib me occasionally about where’s your computer? Well, a computer can be used in a variety of ways. I’ve been thinking about it. It can be used as a word processor; it can be used as an e-mail device, a communicator in that sense; and it can be used as a technology processor in fact for mathematics and things of this sort. That’s why I’m primarily interested, in drawing pretty pictures and having the computer compute my formulas. My point is it would take me a little while, and I don’t feel I have the time while I’m working on the book. This is an excuse but it’s also an explanation. I guess I’m a little defensive.



I don’t at this moment have a need for a computer in the ultimate sense. I work with some colleagues who will compute for me. I have two excellent young ladies whom I employ to do technical typing. It’s very difficult work. My stuff has footnotes and subscripts and those sorts of things. I suppose if I really mastered a computer, I could do it, but it works well the way I’m operating. They can type and I can proof read. Here’s chapter one, and it’s almost done. Chapter two is even worse. They’re loaded with mathematics.


Geselowitz:

I’m sure that the two young ladies are grateful that they live in the age of the computer processor. This is not ideal material for a typewriter. So, you’re using the technology indirectly.


Middleton:

That’s fine. I say make a disk for yourself and give me one for the publisher. That’s simple. So that, at the moment, handles my needs all right.


There’s another factor, too. I’m not sitting in a laboratory. I know outfits like Dell and a few others have good back up, but there’s nothing like having the computer nerd next to you who can unscramble the fact that you’ve accidentally leaned on the wrong button and blanked out your hard drive, or you didn’t have the floppy in and you’ve lost the data and stuff of this sort. So I’m leery at spending time on that. Quite often I take out the fax instructions, and I try to re-program it. These things with multiple operations, they’re the ones that get you. This is just pressing one button. You press one, then you do that, and there’s a sequence. There are combinations that are just impossible.
So those are all my excuses, and I think reasonable ones. I feel that incumbent upon me as a scientist to do something about this eventually. But that’s my first project after the book. As I said, I’m well enough equipped to make pretty good use of my time.


IEEE

Geselowitz:

Before we let the interview run out, because of my position I need to ask you about your IEEE membership: when you joined and why you joined and which societies. Presumably you joined when IEEE was the IRE. The IRE structure had “professional groups” which eventually became the IEEE Societies.


Middleton:

I joined IRE as a Student Member in 1942 and then as an Associate Member in 1944 and a full Member in 1945. God, that’s fifty-five years ago. I was made a Fellow in 1959 and a Life Fellow in 1986.


Geselowitz:

Right. The fellowship you earned by virtue of your work and the life designation you earned by survival.


Middleton:

Right. So now I take a very direct interest in the IEEE, though not in its administrative affairs. I feel my energy and efforts should be directed to what I do the best, so I saddled myself with this book. It’s my own fault, but it’s a project that will probably keep me alive longer than if I did nothing. But I feel the urge to communicate, and IEEE can help there.


Geselowitz:

Members often communicate through their Society. Were you active in any IEEE societies at any point, or did you just read the publications?


Middleton:

I was not really active. Of course, besides reading the journals I did review papers for them. My main IEEE function, in that sense, has been to review papers. I’ve written supporting letters and done sponsoring for Fellowships. But that’s really the limit of my administrative interaction in the IEEE. The rest is professional, reading and publishing in the journals.


Geselowitz:

Did you attend IEEE conferences?


Middleton:

Yes, that’s another question. There are conferences all over the place. I’ve made it a rule—though I may have violated it from time to time—not to go to a conference unless I’m presenting a paper. There is an IEEE Oceanic Imaging conference in Newport, Rhode Island, coming up in May, which I will attend. There’s another one at MIT on electro-magnetics. In Tel Aviv there’s a three-day symposium on higher statistics, and I was invited to speak. Those also have some involvement with the IEEE. IEEE is really the most important sponsor of technical conferences.


Geselowitz:

What Society would you say is closest to your own interest in terms of reading and writing publications, given your broad background? Ones that occur to me are the IEEE Communication Society, the IEEE Information Theory Society, and the IEEE Electromagnetic Compatibility Society.


Middleton:

I belonged to all of them! I’m not doing as much of that now. If I had to pick one for most important, I would say Information Theory. I also just joined the IEEE Signal Processing Society, which is publishing work in my area now. They have a different emphasis than I do, but they do have Venn diagrams in the lab! Of course I’ve also had research published by the IEEE Oceanic Engineering Society. I’m also in the IEEE Aerospace and Electronic Systems Society.


Research: models of interference, scattered channels

Geselowitz:

A lot of the research effort during World War II was in what we would call aerospace engineering. A lot of your colleagues at the Rad Lab and the Radio Research Lab stayed in that sort of military and aviation application of radar. What is it that you do in the Electro Magnetic Compatibility group? That seems a little different than your other interests.


Middleton:

That came out of the work I did for the Department of Commerce, I believe for the Institute of Telecommunication Sciences in Boulder. I published papers with one of my students, who sadly has since died. A lot of the work in that Institute had to do with them looking at the electromagnetic compatibility problem for e-commerce. They advised the State Department on the standards, for example, and that came right out of my work. The EMC recommended compatibility is very close to ECM electromagnetic countermeasures. It’s the old problem of how do you detect a signal in interference, and of course you need to model the interference.


A lot of what I considered my more important work was trying to make what I call canonical models of interference we’ve actually encountered. Gaussian interference, which is certainly in there, but also impulses that come from various manmade and organic sources: snapping shrimp in the ocean, lightning discharges from the CB radio, vibrations from trucks. Later the Japanese began taking interest in this problem, and then the Russians did. I met some of the Russians in Tel Aviv, and they told me they had students working on this. It’s very flattering, I suppose. Of course, I know they were using it in military, just as we were, but some of it was in the public domain. It’s a little upsetting, but that’s the way these things are.


Geselowitz:

Does your work have any application on the micro scale? Because I’ve heard it said that a big problem we face in the future now is that they’re not training enough engineers. Everyone basically wants to join a dot-com and just do programming. It’s assumed that Texas Instruments will continue to print the circuits or whatever they need to put in their machines, and that not enough attention has been paid to the interferences on micro chips themselves.


Middleton:

I can see that this is a problem, although I haven’t been concerned with it myself. For my work I use the word canonical—not in the ecclesiastical sense, but in the sense used in quantum mechanics. In one sense the description is independent of any particular physical process or structure. In other words, it’s a form. For example, we’re doing the same things in detecting underwater acoustic signals that we did in detecting radar signals, except we’re dealing with a pressure wave in the acoustic case and an electromagnetic wave in the radar case, and the numbers are different. But the formulae and the concepts are exactly the same. So we can develop canonical models that can then be applied locally.


Geselowitz:

So really you can almost summarize your work as being the search for canonical statistical descriptions of a channel communications.


Middleton:

Exactly. A channel includes the signal injected into it and the way it’s done—an aperture or an array—and it includes what happens in the channel, such as inverting, and it includes the noise. It includes the receiving apertures and the arrays, and then the way you process the outputs of those arrays to obtain the signal.


Geselowitz:

And any loss in the channel also has to be accommodated and modeled.


Middleton:

Oh yes. And then the energy problem and the scattering problem


Geselowitz:

You mentioned your scattering work earlier.


Middleton:

Yes. So, that’s the description, and that’s exactly what my book is about. I’ll give you a copy of my notes. That’s the story of my intellectual life in a nutshell.


Geselowitz:

In a very small nutshell. We can end the formal interview and just look at some of your materials. I’m wondering if there’s anything you’d like to add? Are any questions, or is there anything you want to make sure gets on the record?


Middleton:

I have a few vignettes which are in the materials I’m giving you.


Geselowitz:

Okay, are they any you’d like to put on the tape?


Middleton:

Maybe you could get back to me with that at some point.


Geselowitz:

Okay, I’ll read the material, and we can do as a short phone interview or something like that. In that case, I’m going to end the formal interview and thank you very much.


Middleton:

Well, thank you.