Somewhere in the early 1980s, when I was a graduate student at Stanford, I took a class on VLSI design. Truth be known, I dropped that class midway through the quarter, but in recent years as I've seen Professor Wolf speaking at various technical conferences, I've often wondered if perhaps he'd been the TA in that VLSI class, way back when.

When I asked him that question, at the outset of our phone call in November 2004, to my delight he said, "I did TA the VLSI design course for several years, starting in 1979. If you took the course in that timeframe, then I was your TA because I was the only TA for the course." The rest of our conversation is as follows:

********************************

Wolf: It's actually better here because there's less traffic.

Wolf: About 2 years ago, I uprooted our VLSI course – which was more of a traditional Mead-Conway course – and redid it as an FPGA course. It's not a standard design course, in that we first talk about VLSI, which is sort of digital circuits; we look at what an SRAM is; we look at interconnect delay; and then, we look at FPGA architecture and use that to understand VLSI design. The students [come to] understand why FPGAs are designed the way they are. Then we go on, and do logic design with an FPGA twist. [Through all of this], the students learn how it is that logic design is fairly standard.

There are a couple of reasons why [this teaching strategy] makes sense, at least for a lot of schools. It's actually a lot easier to talk about some of these concepts when you've got something real in front of you. So, if you use an FPGA as the basic source material for a course, you can say to the student, "Here's a chip design for you to understand."

In reality, most Princeton students are not going to go on and become custom VLSI designers. A few of them might go out and do microprocessor design projects, but most of them who are going to do logic design, are going to do FPGAs. So, I decided to redo our design course in this way after listening to various colleagues talk about the decline in custom chip starts over the last couple of years. Certainly, there are custom chips being designed today – large and interesting chips – but not as many as there were in the 1980s. Teaching FPGAs in a VLSI course is a natural conclusion here.

Wolf: First of all, FPGAs aren't the solution to all of the world's problems, but they are a huge business that's growing like crazy. So, it makes sense for students to understand FPGAs.

They definitely offer enough [technical complexity] to satisfy the needs of a basic course like ours. FPGAs are the right medium for teaching today.

There are probably some schools where it makes sense to teach a [more traditional design] course – schools like Berkeley, UT Austin, or Stanford where they have enough faculty to teach VLSI design and test, and they've got a large number of students who have a chance to really be working in this area. But for a school like Princeton, teaching [with FPGAs] makes more sense right now.

Speaking globally, there is a link between shifts in the technology and the shifting labor [patterns in engineering]. An increasing number of back-end chip designs are being done overseas. However, our using FPGAs in our courses has nothing to do with the cost of labor here in the U.S. versus overseas.

It's more about [having the students become familiar] with the components of design being used today. In telecomm, for instance, FPGAs are widely used for base stations, and so forth. If you're [designing chips] and your customers need high performance, but low volume, you'll probably use FPGAs for that product. In the 1980's, you could do that same design with an ASIC, but now that option's too expensive.

Wolf: Actually that's another advantage of teaching the basic design courses this way. We use Xilinx tools. They're free, pretty well integrated, and easier for students to use and see the effects of using tools. You can run random, multiple designs using the tools, and you can [execute] place-and-route changes – all processes which are within easy access of the tools.

Wolf: Students at Princeton take courses on physics and transistors. The question is, can they use that knowledge to do something that's interesting? Part of the solution is to give them interesting courses where they can understand and learn the basics, and they can also have exposure to realistic problems. Our FPGA course is a good example where students can do something practically, and also learn some circuit theory in the process. [As always], it boils down to a question of breadth versus depth.

Wolf: The sophomore students manufacture something very small, with just a few transistors, and they test it. They're able to do all of that in one semester, but they get guided through the process. If you're talking about larger chips, however, we do have some testing facilities on campus. But, there are no more functional testing facilities on campus.

In the old VLSI course that we used to teach here, I didn't have the students fabricate their designs because it just took too long to get them back. [On top of that], there was no guarantee that the students would return for the follow-on course where they would actually test their chips. Again, this is another advantage to the short implementation cycle [associated with FPGAs]. Students have an actual chip in hand much faster.

Wolf: Here at Princeton, as fair number of our EEs go into consulting work. They sometimes end up on Wall Street, where they use various sorts of rating algorithms in their work. However, some of our graduates get real engineering jobs, because there are actually a fair number of them who want to stick to engineering. They may go on to work at HP or AMD, and so forth.

Wolf: For Princeton, it's huge. In Electrical Engineering, we have the single largest graduate school enrollment – around 200 students. Those are mostly PhD candidates. You don't have to get a masters [in the process of getting a PhD], but you can get one if you want to.

Wolf: We do get Computer Science students over here, and frequently the EE students go over there. I, myself, have a courtesy appointment in Computer Science.

Wolf: Well, you talk to different schools and you get different rankings. I take those rankings [with a grain of salt]. At the graduate level, it's a lot more about the individual student and the individual researcher/faculty member [they're working under].

Wolf: I have such a nice schedule. I'm teaching one class this semester, and I'll be teaching two next semester. A lot of my time is spent on research and writing proposals for my graduate and post-doc students.

Wolf: The majority of the money at the moment comes from NSF, and some comes from the state of New Jersey as well.

Wolf: At the graduate level, we have more domestic students these days because of the economy. For a while, we were mainly getting students from the PRC, India and Taiwan, but those numbers had decreased. Now, they are starting to come back again. We also have students here from Turkey and Greece, and various other places in Europe.

Some of our foreign students are continuing to have visa issues, so we're very careful about sending them to international conferences. We always want to be sure they can return to the U.S. if they go [out of the country]. There was a while there, when we had some of our admitted students who were not able to get the visas they needed to come here to study. But, I hear rumblings that it's starting to get better.

Wolf: Yes, there has been a small decline in EE, but that [has stabilized]. A couple of years ago, financial engineering was au courant – trading algorithms and operations engineering. But, I think that has changed again.

Electrical Engineering departments face some very serious issues today. Most are going through some kind of identity crisis – whether they admit it or not – about what EE is and where it should go. I don’t have the answer yet, but I think departments should talk among themselves, and the funding agencies, about where this field is going. Electrical engineering isn't in any danger of going away, however. I think that even after Moore's law, there will be [progress that can be made]. I feel strongly that we can exploit the transistors we've got today, for quite some time to come.

To that extent, I'm a bit contrarian. And, I certainly believe we don't have to abandon the existing fields in EE. For instance, you would think there are a lot of interesting problems in power. Certainly you have to ask, would Thomas Edison recognize all of the parts of the modern power grid? He probably would, but nonetheless, there are still good schools that [are actively engaged] in research into power systems. The power grid is one of those existing technologies that people, in general, don't really care about it. They don't want to know what's behind their power [distribution system]. They just know they want it to work. However, there are still many problems that need to be addressed in that area.

Wolf: There are people here who specialize in signal processing and in computer engineering. The real cutting edge today is biological machines, and there is some very interesting work going on here as well in that area. Also, there is important work being done here in nanotechnology. For the students, however, we offer a fairly traditional degree in EE.

Wolf: I'm interested in the physical nature of computing – low-power, real-time performance, cost, etc. My PhD was on layout compaction, but I'm not doing physical design anymore. Generally, I work in embedded systems, which includes a lot of work related in some way to a 'smart' camera. The work requires looking at the entire system, from the application down to the chips.

A 'smart' camera is basically a camera with a processor and algorithms that analyze the image or video. One possible application might be that you would have a bunch of cameras to cover the different parts of a room. If you stand in front of one of the cameras, it will tell you where you are in the room. The camera can follow you, and the computation of your location will move with you.

Our original motivation was to create a 'smart' room, a meeting room that's covered with cameras that have full knowledge of everyone in the room. That's clearly one application, but it's not the only application for our systems.

The [larger question is] is – if you want to understand a space, for whatever reason, one camera would never be enough. You would need multiple cameras, which require a way to handle all of that data. You don’t want to be dragging that data back to the server, so we put the processing near to the camera. That way, you can actually install and operate these systems in the field. We're trying to make a decision with this system. It's a form of distributed processing, and I think the most interesting question then is what do you do with the information you gather [using the system].

Maybe in certain systems, for instance, you don't actually want certain people in the room, however people might get into that room by 'tailgating' somebody else as they swipe their [legitimate] security badge and enter the room. Eventually, I would like to have some kind of system that says who are the bad guys, who are the terrorists – but as far as I can tell, it's not yet possible to relate general activity or body gestures to a psychological state. So, if you want to just watch somebody, and say they're suspicious – I don't know of any way today to do that as yet. Certainly we know that watching for certain things, certain behaviors, can be useful – but not from a psychological basis.

Smart cameras, in general, have a lot of different uses – medicine, emergency response, etc. Any amount of information on the upper floors of the World Trade Center would have made it safer for the responders on September 11th, for instance. Cameras are getting cheaper thanks to advances in VLSI, so [we're beginning to be able to] afford putting multiple cameras on a subject. [Going forward], when you build buildings, you could install this stuff during construction [and the capability would be there when you needed it].

In general, however, I always try to be agnostic about the applications of the research we do here.

Wolf: As far as I understand the law, the students here at Princeton own their course work, but any sponsored work – funded research – is a different matter. John Ritter is the head of our licensing program and I believe he does a good job. There are a fair number of departments here at Princeton who don't generate any IP at all, so it's really a department by department thing.

Wolf: I was actually in industry for 5 years (if you call Bell Labs the real world) before I decided to teach. It was a big decision to [transition to academia], but I think it has allowed me to have more of an audience for my ideas. If you want to show your impact inside of a company, there are actually only a limited number of people you can talk to about your technology.

I've always been interested in designing big systems. We used to worry about 10,000 rectangles, whereas now it's 100 million transistors. From that perspective, I'd probably be [looking at the same problems] if I was still in industry. Either way, I would have been interested in climbing the abstraction level. At a university, you have the opportunity to bang your head on the wall and learn some new stuff.

Wolf: The main difference is that, if you lose in politics in a company, you get fired. Of course, there are politics in academia, but university politics are kind of like the battle for Stalingrad, whereas industry politics are more like guerilla warfare – the sort of thing where you're solving the problem of the day.

Wolf: I spend time with my 2-year-old, who's really interested in understanding things. I also spend time cooking – I do several different types of cuisines. Besides that, I'm working on the 2nd edition of my Embedded Systems Book. My textbook on FPGAs just came out this June. That one took me about one and a half years to write, but it wasn't entirely from scratch because it was derived somewhat from my VLSI book.

Wolf: Telling jokes requires delivery, and I'm not sure I've got what it takes. Although, other people always seem to keep laughing when I'm speaking, so maybe I'm funnier than I think.

********************************

Editor's Note: To learn more about Wayne Wolf, and to determine if he has a sense of humor, check out his website: http://www.princeton.edu/~wolf/bio.html