Oral-History:Laurence Milstein

About Laurence Milstein

Milstein received his Masters (1966) and PhD (1968) from Brooklyn Polytechnic. He worked at Hughes Aircraft (1968-74), RPI (1974-76), and UCSD (1976 to present). His research has largely been on spread spectrum communication and Code Division Multiple Access (CDMA). These technologies largely had military application (preventing jamming, hiding signals in noise) until about 1990, but have since acquired greater use in commercial wireless technology. CDMA is one of two digital cellular telephone technologies in widespread use in the United States. Milstein has largely been a theoretician.

Milstein has had heavy involvement in IEEE’s Communications Society and Information Theory Society. He believes the transformation from analog to digital communication is the biggest change in communications technology. He explains coding theory and digital modulation. Milstein believes the future of communications engineering will involve a continuing shift from circuit switching to packet switching, integration of communications and computers, preventing different devices from interfering with each other, increased data transmission, and above all the switch from wired to wireless communications.

About the Interview

LAURENCE MILSTEIN: An Interview Conducted by David Hochfelder, IEEE History Center, 2 November 1999

Interview # 381for the IEEE History Center, The Institute of Electrical and Electronics Engineers, Inc.

Copyright Statement

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

Request for permission to quote for publication should be addressed to the IEEE History Center Oral History Program, IEEE History Center at Stevens Institute of Technology, Castle Point on Hudson, Hoboken, NJ 07030 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:

Laurence Milstein, an oral history conducted in 1999 by David Hochfelder, IEEE History Center, Hoboken, NJ, USA.

Interview

Interview: Laurence Milstein

Interviewer: David Hochfelder

Date: 2 November 1999

Place: Atlantic City, New Jersey

Overview of communications technology innovations

Hochfelder:

Perhaps you could give a brief account of what you feel the significant innovations have been in communications technology since your involvement in the field?

Milstein:

<flashmp3>381 - milstein - clip 1.mp3</flashmp3>

The single biggest change was the transformation from analog communication to digital, and that transformation was due to advances at both the system level and at the level of basic technology. It led to the convergence of the communication and computer industries, and was propelled by phenomenal advances in digital signal processing and VLSI technology. On the system side, the conceptual advances that took place were due to such things as Shannon’s work in information theory, various fundamental advances in coding theory, and new results in signal and modulation design.

One of the classes that I teach at UCSD is a senior-level course on communication systems. It’s a three-quarter sequence: the first is on analog modulation, the second is digital modulation, and the third is information theory and coding. When I teach the analog modulation, the students can very easily relate to it because they’ve been relating to analog AM and FM since they were old enough to turn their radio on; their television works this way—they have total familiarity with it from an intuitive point of view, so they enjoy learning what it’s all about and how it’s actually functioning. Yet, they realize that’s not what’s exciting today. What’s exciting today is what they see in the second quarter of the sequence. The interesting thing, however, is when I get to digital communications in the second quarter (which they’ve been looking forward to), many of them nevertheless have a tremendous problem with it technically because it’s not as intuitive to them as is analog communications. However, they know that all the jobs today are in digital communications. Many aspects of digital communications are becoming buzzwords: CDMA, TDMA. But this is what they want to learn about. So it’s not lack of motivation, but they often have a much more difficult time in the second quarter than in the first quarter. In other words, perhaps the single biggest change is a transformation of communications from basically an analog communication-dominated field to one that is totally dominated, at least in terms of innovations, to a digital communications-dominated field. Also, in a slightly broader sense, the convergence of the communication and computer industries has been very dramatic.

Coding theory and digital modulation

Hochfelder:

Could you talk a little bit about conceptual shifts in the hardware a little bit? Can you explain what some of the milestones were?

Milstein:

The fundamental work of Shannon showed that perfectly reliable communication, at least in principle, can be achieved at any rate less than a quantity known as channel capacity. Shannon's work proved the existence of a code that can do this for certain types of channels, but he never showed how to actually find the code. What that result, in turn, did was to motivate the entire discipline of coding theory. Coding theory is the attempt to find the Holy Grail, so to speak; to find a practical technique that approaches what Shannon promised could be achieved. Since coding, by nature, is a digital technique, it made clear that there was something to be gained if you take the signal and sample it, quantize it, and encode the quantization levels and transmit that information across the channel. Note that, even though in the quantization process, you introduce an irreducible error, if you appropriately sampled the signal at a high enough rate and you've done the proper encoding, you can bound the overall distortion on the signal, and, in general, you're going to do much better than you would with an analog modulation format.

Hochfelder:

Through this error correcting, is there a trade-off between error correcting and bandwidth?

Milstein:

Yes. Error correction coding uses excess bandwidth because it introduces controlled redundancy. By way of analogy with analog communications, FM performs better than AM, but it typically uses a higher bandwidth. If you wanted to listen to, say, high quality classical music, you wouldn't put your AM radio station on; you'd put your FM station on. Commercial FM starts out with a larger baseband bandwidth, which any system can do, but then it uses a much wider IF bandwidth. What that larger bandwidth effectively buys you is a larger signal-to-noise ratio. Similarly, what the error correcting coding is doing is trading off the bandwidth for robustness against noise. That is, it tries to make the signal more tolerant to the errors that are caused by the noise.

What further aids you is the fact that where you use a digital modulation format, you can employ data compression techniques. Just think of HDTV. In general, what data compression does is try and remove redundancy as much as possible from the message you're going to transmit. The error correction coding then inserts controlled redundancy to combat the errors that you know you will encounter in going across the channel.

Career summary

Hochfelder:

Could you give a brief summary of your career?

Milstein:

Yes. When I got out of school I went to work at Hughes Aircraft Company and worked on satellite communications for about six years. Probably two-thirds was analog voice transmission over satellite links; the other one-third was on the development of new modulation formats or receiver designs for digital communications through satellites. I left Hughes in '74, I taught at RPI for two years, I joined UCSD in '76, and I've been there ever since.

Spread spectrum communications and digital communications theory

Milstein:

<flashmp3>381 - milstein - clip 2.mp3</flashmp3>

At Hughes, I got into an area called spread spectrum communication. Spread spectrum had been around at that point for several decades. Up until roughly a decade ago, it was primarily the domain of the military. What spread spectrum does, going back to your bandwidth issue, is impose a very high rate spreading sequence on top of the information. Thus, you transmit the information at a bandwidth much wider than what the information itself will demand. In turn, what you get is a robustness to interference. Not the noise, which is what error correction coding is designed to fight, but interference. The military has tremendous interest in spread spectrum for anti-jam communications, and also because it tends to hide your signal in the noise. Regarding the latter point, you're transmitting a given amount of energy in a much wider bandwidth, so the power spectral density is correspondingly lowered. If you're trying to hide the presence of your signal from an unintended receiver, as you certainly want to do in a military context, this is called low probability of detection, or low probability of intercept. Spread spectrum is an obvious choice of a signaling format for this application.

While in the broader sense my research area is digital communications theory, the vast majority of my work is in spread spectrum communications, and most of that has been in either interference suppression or what's called Code Division Multiple Access. Note that just like spread spectrum in general, CDMA has been around for decades. It was used not simply by the military, but also by the commercial and government sector. For example, NASA was very interested in one study I worked on at Hughes. It concerned a geosynchronous satellite, which NASA called the TDRS, or Tracking Data Relay Satellite. The TDRS communicated with a bunch of low altitude satellites. Of course, the geosynchronous satellite appears stationary to a single point on the Earth. The low altitude satellites are constantly moving. Anytime a low altitude satellite comes in range of a TDRS, it would transmit to the TDRS data it was collecting from the Earth. The problem was that it was using omni-directional antennas, and so some of the energy that it was transmitting directly to the TDRS was bouncing off the Earth and back up to the satellite. This created multipath. One of the key virtues of a specific type of spread spectrum signal design, called direct sequence spread spectrum, is that you can resolve individual multipaths and constructively use them. It's a technique to make a signal robust to multipath fading. Thus, because there was the need both to allow multiple signals to be transmitted simultaneously and to combat multipath, CDMA was chosen as the signal design.

Code Division Multiple Access

Hochfelder:

Can you explain a little bit what CDMA is?

Milstein:

Your radio and your TV work in a mode called FDMA, which is Frequency Division Multiple Access. In FDMA, each signal gets a specific portion of the frequency band totally dedicated to it, and each band is disjoint from the other bands, so all signals can be on simultaneously, without interfering with one another.

Probably the simplest dual to FDMA is called TDMA, or Time Division Multiple Access. In TDMA, each user occupies the entire frequency band, but is on sequentially in time. Let's say you're the first user. You're assigned a particular time slot and you come on. You only have a small time segment initially, but you have the entire frequency band to use so you can transmit at a much higher data rate because you have that much more bandwidth. Then the second user comes on, the third user, the nth user, and eventually the cycle repeats. That is, the first user comes on again and it continues its message. So whereas in FDMA all users are on all of the time but they occupy an orthogonal frequency band, in TDMA, users come on sequentially in time, but they occupy the entire frequency band, so they're orthogonal time instead of in frequency.

In CDMA, or Code Division Multiple Access, they can all be on simultaneously and they can all occupy the entire frequency demand, so you can't distinguish one from the other on the basis of either time or frequency. So how do you do it? Well, you superimpose upon each data symbol the so-called spreading sequence that I was talking about before. By picking a judicious set of spreading sequences, you can give each user a different spreading sequence and you attempt to make these signals orthogonal—not in time, not in frequency, but by virtue of the code itself.

In most CDMA systems, you can't really make the sequences truly orthogonal because these systems are what are known as asynchronous. This means there's no coordination of the timing as to when a given signal starts relative to the time when any other signal would start. If the spreading sequences are properly designed, by virtue of a cross-correlation operation at each receiver, the unwanted signals at those receivers are significantly attenuated. In other words, you take all the signals that are coming in, multiply by a replica of the spreading sequence of the waveform you want, and then integrate the product. If you attenuate all the other signals to a sufficient level, they tend to look like, in many cases, a higher noise level. Thus, you get this semi-orthogonalization, not by having disjoint time slots, not by having disjoint frequency slots, but by using these semi-orthogonal spreading sequences.

Hochfelder:

Each receiver has to have it's own particular code sequence?

Milstein:

Each receiver has to know the spreading sequence of the waveform it first wants to despread. Despreading means multiplying by a replica of the incoming spreading sequence. That then collapses the bandwidth down to the information bandwidth. Simultaneously it allows all the other interfering users to stay at the spread bandwidth because they're using different spreading sequences. If you envision this product of the incoming waveform and the locally generated spreading sequence then going through a low pass filter at the information bandwidth, you pass all or most of your information and filter out most of the interference. This is overly simplified, but it's the essence of what is going on.

Education and employment history

Hochfelder:

If we could go back and get some dates and milestones. When did you graduate with your bachelor's degree?

Milstein:

I graduated from CCNY in '64, and got my master's at Brooklyn Poly in '66 and Ph.D. at Brooklyn Poly in '68. Brooklyn Poly became New York Poly when it merged with the engineering school of NYU, and later they changed the name to simply Polytechnic University, which is what it is now.

Hochfelder:

And you went to Hughes Aircraft in '68?

Milstein:

Correct, and I was there until '74. From '74 to '76 I was at RPI and from '76 to now at UCSD.

Evolution of research interests

Hochfelder:

How have your research interests changed over time? Obviously, digital communications and spread spectrum have been constant.

Milstein:

I would say up until about a decade ago, the vast majority of my research was for military communications systems. This may not be exact, but, say, for the first four decades of spread spectrum, it was in very large part the domain of the military just because of the anti-jam and LPI characteristics of the waveform. That does not mean that it wasn't used in the civilian sector. I gave an example of NASA's interest earlier. However, the majority of the progress and strides that were made in spread spectrum came from the military community, and that was the focus of my work.What happened in '89 or so was two things. One was the Berlin Wall fell down (i.e., the Cold War ended), so suddenly there was much less interest in military communications. Secondly, wireless communications became a hot commercial, consumer-type of a topic, and there was interest, in particular, in CDMA for wireless communication. That transformed a large portion of my research for military spread spectrum communications into research for commercial spread spectrum systems. Probably half of my funding still comes from the military these days. The military still has the same interest it always had in spread spectrum, and in particular, in CDMA. What has built up, though, is the commercial interest in spread spectrum. That's probably the single biggest change. I've always been essentially a theoretician. I don't build things; I do things like performance analysis and conceptual design of communication systems, and that hasn't changed at all.

IEEE; Communications Society, Information Theory Society

Hochfelder:

Could you describe and outline your involvement with the IEEE and the Communications Society?

Milstein:

Around 1967, when was a doctoral student, I joined both the Communications Society the Information Theory Society. I've been members of both of those societies ever since. I'm now also a member of the Vehicular Technology Society. But CommSoc and IT were my prime allegiance. When I was at Hughes, I don't think I had very much interaction. I became quite active when I left Hughes and joined RPI. For the past two decades I've been extremely involved with both societies. I was on the Board of Governors, for two terms, of the Information Theory Society. I was also on the Board of Directors of CommSoc. I think I filled out one term of someone else the first time I was on the Board, and then I was on for two more terms. I was also Vice President of the Technical Activities for two years, '90 to '91. I was an Associate Editor for the Communications Transactions for seven years, and for the Communication Magazine for five of those seven years. I've been on the JSAC (Journal for Selected Areas in Communications) Board many years. I'm currently the Editor in Chief of JSAC. I was also the book review editor for the Information Theory Society for three years. Then there were lots of conference activities. I've been heavily involved in MilCom in particular, but not solely. I was on the GICB, the GlobeCom ICC conference board, which was the overseeing board for GlobeCom and ICC. I've been on the MilCom conference board, and one year I was the chair of the MCB. I was the co-technical chair of the first MilCom back in '82. In fact, for a while there were only four of us involved in the very first MilCom. Outside of ComSoc, I've been on the IEEE Fellows Committee, which I'll chair next near.

Hochfelder:

Could you talk about the overlap, both with regard to membership and research interest between the Information Theory Society and Communications Society?

Milstein:

The Information Theory Society is more academically oriented, whereas CommSoc has, in general, been more dominated by industrial people, in particular from the Bell system. I think that's maybe the simplest distinction to make. If you look at the research oriented technical committees within CommSoc, in particular the Communications Theory committee (which is the one I have had most of interaction with; I've been a member of that for ages—I chaired it for a number of years), there is a large overlap with the Information Theory Society. The Information Theory Society tends to focus on the mathematics of Information Theory, which they call Shannon Theory, and the more abstract aspects of coding theory and detection theory. Whatever the topic is, their prime focus is on a mathematical approach to the work. In the Communication Society, in particular the Communications Theory Technical Committee, there are some very theoretical members as well, but the problems they tackle typically are more system-oriented. The more realistic you make a problem, oftentimes the less ability you have to get a nice closed-form solution for it. But if you took just those regions where there's overlap, I would say there's almost a continuum from one to the other. There are many people that are heavily involved in both and they publish heavily in the Information Theory Transactions and also heavily in the Comm. Transactions and in JSAC.

Predictions for the communications engineering field

Hochfelder:

What do you think would be some of the technological transactions in communications engineering in the next ten to twenty years?

Milstein:

Everything seems to be going away from circuit switching, which is what we have now, to packet switching. Seemingly people have an insatiable demand for more and more, as well as easier and easier, access to communication systems. What I've found somewhat surprising about; for example, wireless communications is that while I could understand it catching on rapidly in underdeveloped countries (because they have no infrastructure), it has also caught on very rapidly in developed countries where an infrastructure does exist. I see people walking around with handsets and there's a wired-line telephone two feet away. Not me. I've never been totally convinced there might not be some kind of health hazard involved.

Hochfelder:

What sort of health hazard?

Milstein:

<flashmp3>381 - milstein - clip 3.mp3</flashmp3>

Radiation from the antenna. I know there have been studies that said there are no hazards. Of course, cigarette companies have lots of studies also. There's no point in even dwelling on that one. In any event, I never found the need that many other people seem to have regarding wireless communications. Nevertheless, it has caught on and, no matter how much capacity and whatever data rates you give people, they want more. Whatever the size of a handset, they want it smaller. They want it lighter, and they want the battery to last longer. This is in large part technology, but there are system-level considerations involved as well.

My belief is that you're going to see even more convergence between communication and computers. It's very difficult to tell the difference now, and I think it will be more so in the future. I think interference between systems will continually be a serious consideration. One of the common problems in the military is what they call self-site interference. A platform that is on a ship, or on an airplane, or on a jeep, has multiple antennas, so there's multiple, almost non-interoperable systems all radiating simultaneously, and they self-jam one another. It's a non-trivial problem. Interference in general I think is always going to be a problem, and it will become more and more of a problem the more ubiquitous communications become.

Lack of sufficient bandwidth is always going to be a problem. The FCC is always going to be badgered by many constituents. To obtain additional bandwidth, system designers are continually trying to go to higher frequency bands. However, as you go to higher and higher frequency bands, you get into propagation problems. So you may be able to communicate over small distances in certain cases, or you may be able to communicate only in certain localized areas under certain atmospheric conditions, but you want this to be as general as possible.

Hochfelder:

What do you think will be the division between wireless and wired communications in the next ten or twenty years?

Milstein:

Things are going more and more wireless. Certainly with a zillion teenagers walking around with handsets, wireless voice communications is commonplace. I may be one of the last holdouts going to a public telephone.

Hochfelder:

Me, too.

Milstein:

So we are the last two of the holdouts. Seemingly data is the next big transition. I think there is a sizable demand for wireless laptops. You can get the service now in a limited form from, say, Metrocom, a company in the Bay Area. My suspicion is that this will become much more ubiquitous.

Maybe in part this is tied to a related question. I think the distinction between voice and non-voice will shift tremendously in the direction of non-voice. For example, you see this right now with both e-mail and web browsing.The question you asked me is the distinction between wireless and wired-line. My suspicion is that as much as possible we'll ultimately go wireless. That seemingly is what people want, in their car, while they're flying; wherever they are, they want to communicate untethered.

Hochfelder:

Okay. I think we've pretty much hit the high points. I know you're pressed for time. Do you have any last remarks or issues you want to raise?

Milstein:

No, I'm really happy to respond to your questions.

Hochfelder:

Well, sounds good. I appreciate your time.

Milstein:

No problem. Appreciate you coming down. You did it right in time, also.