Once more, it is both a pleasure and an honor to be selected to present the second annual Phil Kaufman Award to Professor Donald Pederson of the University of California at Berkeley. Before I speak about Don, I would like to take this opportunity to personally thank EDAC for the creation of this important award—with the first two awards, EDAC has certainly set the highest of possible standards and has created an important legacy that adds a truly unique dimension to our discipline. I also thank my fellow committee members, a very talented and diverse group, and Georgia Marszalek for her hard work coordinating a group of very busy people and for the numerous other ways in which she has helped us.
I’m sure all of those present have heard of the SPICE circuit simulation program. Some of you have contributed to it technically, some have sold versions of it commercially, or tried to improve upon it, or used it for your own designs or perhaps used it in College. SPICE truly is a household name in our industry. In design technology, it is the only noun I can think of that has reached the status of a verb—we speak about "SPICEing" a circuit; "let’s SPICE that and see if it works!"; like "Xeroxing" a piece of paper. We don’t talk about "Veriloging" or "VHDLing" a design yet. But we certainly have "SPICED" some part of virtually every integrated circuit developed in the past quarter century. Virtually every one! In addition, as Dr. Ian Getreu of Analogy recently pointed out, SPICE has replaced reality in a way. Nowadays, any company or researcher who wants to introduce a new simulation technology, approximation, or model, has to compare his or her work with SPICE, no longer with any real circuit per se! There is no doubt that SPICE stands out as a unique example of the potential impact of design technology. It has no equal, not in our industry—not, that I can think of, in any industry. A truly remarkable achievement.
And the only ‘phenomenon’ in our industry that I can think of that has been more profound than SPICE itself, though certainly not as well known, is the man behind it’s development and success, Professor Donald Pederson, the winner of the 1995 Phil Kaufman Award.
However, to equate Don’s contribution to EDA to SPICE does him a gross disservice. Over the past quarter century, Don’s impact on the fields of integrated electronics and design technology, via his university teaching, his research, and his work with industry, not to mention the many other ways he has served our discipline as organizer, mentor, and inspiration, is absolutely without comparison. No other individual has provided the vision, leadership, inspiration, and models of excellence for as many key developers and leaders in our industry as Don has—whether they were his students, young faculty at Berkeley as well as other institutions world wide, or the many from industry who came to know him through his research work.
It is worthwhile to review the history of Don’s work in EDA to see how his leadership, inspiration, and enthusiasm have formed the single common thread in this odyssey, and to help understand the many other contributions to our industry he has made along the way.
Don was one of the first people to work with those new-fangled things called transistors, was even earlier in his recognition of the importance of the integrated circuit and established the first integrated circuit fabrication facility at Berkeley in 1960—it was the first fabrication facility to be established at any university and he established it at a time when his peers in industry stated publicly that such an endeavor was impossible. His first Ph.D. graduate from the IC laboratory, David Hodges, went on to develop and promote the use of ICs for memory at AT&T before returning to Berkeley as a professor. Professor David Hodges is currently the Dean of Engineering at Berkeley.
It was in the mid 1960’s when Don became interested in the application of computer programs to the analysis of integrated circuits. He and his students had used an early Bendix machine, with only paper tape input and output, to try to gain a deeper understandingof the behavior of certain linear circuit designs. In his circuit design work, he found that many of the first-order approaches to the analysis of circuits did not predict the correct behavior of real circuits. In fact, depending on the assumptions made, first-order theory could be used to predict a variety of outcomes for the same circuit! One of his recent graduates, Dr. William Howard, had been working with him at the US Harry Diamond Laboratory to better understand a particularly tricky problem involving the thermal behavior of the input characteristics of a linear JFET IC design. It was 1967 when Bill Howard first implemented a computer program at Berkeley for the analysis of the nonlinear dc operating point of an IC—he called the program BIAS—on a 16bit IBM 1130 machine. Don was now convinced that the computer was to play a central role in the design and analysis of integrated electronics.
Bill Howard left Berkeley for Motorola around 1968 where he eventually retired as a Senior Vice President. He is also a Senior Fellow at the National Academy of Engineering.
Many other "CAD" projects followed at Berkeley, based on techniques developed in Don’s laboratory as well as elsewhere (most notably at IBM San Jose laboratories and North American Rockwell), and included the work of Dr. William McCalla and Frank Jenkins—Don identified Frank’s potential and brought him into the group as a freshman—leading to programs like SLIC, Frank, and SINC. Bill McCalla, now at Cadence Design Systems, has made many significant contributions to the CAD industry over the years, including the reworking of the original BIAS program and the integration of nonlinear dc and ac analyses into a single code; Frank Jenkins went on to develop the commercial circuit simulator ASPEC in the late 1970’s as well as some early commercial logic and switch-level simulators, including LOGIS and ILOGS.
It was in Fall 1969, when the young professor Ron Rohrer returned from a leave at Fairchild and began teaching EECS 223, a course designed to apply modern system and circuit-theoretic concepts, as well as advanced numerical methods, to circuit analysis and design. He set about to build the best and most comprehensive circuit simulator he could, assigning various aspects of the task to different students in the class. At the end of the class he told them, their grade would be based on how well they had convinced Don Pederson that their contribution was the best that could be done. The outcome of that course was a program called CANCER, a program which was destined to become the starting point for SPICE 1 development. A young postdoctoral student from Belgium, Dr. Hugo De Man, was visiting Berkeley at the time and made his own contributions to the CANCER effort. Professor De Man, of KU Leaven and IMEC in Belgium, is well-known for the many significant contributions he and his group have made to EDA over the years since that time.
In the Fall of 1970 Don selected CANCER for his classroom instructional programs, rejecting the other Berkeley competitors of that time, SLIC and SINC. Unfortunately, Don’s ability to distribute the CANCER code to his friends and colleagues in industry was hampered by the fact that the program had been declared proprietary. At a time when much of the IC design and development work was going on in industry, Don felt that this was an unacceptable barrier. A young graduate student named Larry Nagel had been closely involved in the CANCER project and when his advisor Ron Rohrer left Berkeley, Don took him on as a graduate student on the condition that he could use the CANCER source as a starting point for a truly public-domain, general-purpose circuit simulator. In May, 1972, the first version of that new program, SPICE 1, was released from Berkeley.
Dr. Larry Nagel continued his work with SPICE, releasing SPICE 2A.0 before joining AT&T Bell Laboratories, where he continued to lead the AT&T in-house circuit simulation development efforts for many years.
It was another freshman that Don identified and recruited in one of his classes, Ellis Cohen, who picked up the SPICE baton from Larry and carried SPICE2 forward, making significant contributions even as an undergraduate. Ellis is, without doubt, one of the most talented computer scientists I have ever had the pleasure to work with. Much of what became the version of SPICE2 that formed the basis of the many commercial versions should be attributed to Ellis. I joined Don’s group at Berkeley in early 1975—Don recruited me from the University of Melbourne after I had already been effectively working for him there for two years, in both my undergraduate and MS programs—and was able to make some minor contributions to SPICE as well. I remember one day when Ellis Cohen had just added dynamic memory management to SPICE2. Don’t forget, SPICE2 was a Fortran program, in punched-card form, and running on a CDC 6400 computer. It’s only output device was a 132-column line printer. Adding dynamic memory management was no small feat! But, as you can imagine, it slowed the program down a bit—by almost a factor of two. Well, this was not acceptable to Don and, in his own way, he challenged Ellis to fix the problem. Within 24 hours, Ellis had designed and re-punched the cards needed to add automatic machine-code generation to the program—where SPICE itself generated the native object code needed to solve the sparse-matrix circuit equations for that particular circuit, rather than using the more general code produced by the Fortran compiler. He had debugged and installed the program and since he managed to get back the factor of two as a result, he was able to keep his dynamic memory management! Dr. Ellis Cohen is presently at Mentor Graphics Corp., Oregon.
At this point, I would like to mention the important role played by the Army Research Office (ARO) in the development of SPICE. In these days of tight budgets and congressional pressures, many agencies which fund university research are being asked to show increasingly shorter term payoffs, force technology transfer to industry, and reduce their long-term commitments to research programs. In discussing the SPICE project with Don Pederson, it is clear that the long term research funding commitment ARO made to Pederson’s effort was instrumental in giving him the flexibility to continue his work to completion over many years. There were a number of occasions where the ultimate pay-off of the work was questioned, where the path was not perhaps as clear to those outside of the SPICE group, but the ARO support continued.
The SPICE2 work continued until the early 1980’s when we converted the program to the C language, added new models and generalized some analysis techniques, resulting in SPICE3.
Don and his students have made many other contributions to EDA along the way as well, in areas from device modeling, mixed-mode simulation, rule-based circuit diagnosis, to macromodels. In fact, his analog macromodel for the operational amplifier, developed in conjunction with Jim Solomon and Dr. Graeme Boyle in the 1970’s, is still a standard today. Jim Solomon, formerly of National Semiconductor and founder of SDA Systems (now Cadence Design Systems) received his MS degree under Don’s supervision at Berkeley while Dr. Graeme Boyle received his Ph.D. with Don before joining Tektronix.
From the early 1960’s to today, we all worked for Don, whether we realized it or not. We all still do, in many ways. I cannot think of any former student who does not turn to Don, at times, for advice and counsel—who does not consider knowing him an invaluable part of their life. And in every case, Don has been there. Before coming to Berkeley in 1975, I met Dr. Ian Getreu, at that time a recent graduate of Don’s and now a Vice President at Analogy Corp., and I asked him what he thought about working for Don. He gave me two pieces of advice: "First, just stand outside Don’s office door and the entire world of microelectronics will flow by you. Second, Don very rarely gives anyone direct advice—he’d rather leave the decision up to you. But if he ever does give you advice, think carefully before you choose not to take it!" I know of individuals who have actually flown half way around the world, with a good bottle of wine in hand, to seek Don’s counsel regarding their professional careers and the choices they were about to make.
As well as those listed above, there are many other leaders in out field who worked as graduate students with Don Pederson. They include Dr. Gary Baldwin, Laboratory Director at HP Labs; Professor Mohammed Gaussi, Dean of Engineering UC Davis; Professor Gary Hachtel, University of Colorado; Dr. Bob Pepper, President and CEO of Level 1; and Professor Bruce Wooley, Stanford, to name a few. I cannot think of a single major design technology company of today which has not been influenced significantly by at least one of Don Pederson’s former graduate students.
The industrial impact of Pederson’s early work in EDA, then circuit simulation, is best measured by the use of the technology he and his colleagues developed over a quarter century ago. As mentioned in the introduction, virtually every semiconductor company and the vast majority of electronic system design companies throughout the world use a version of SPICE, or a program derived directly from it. In addition to companies who have obtained SPICE from Berkeley and who have adapted it for their own in-house use, over the past twenty years more than a dozen companies have been formed based around SPICE versions. In almost every case, the basic architecture of the program, as originally designed by Don Pederson and his students is still used and, in most cases, virtually all of the original code is still present. These companies sell both enhancements to the original program as well as user support . For example, the Microsim Corporation has distributed over 25,000 copies of their PSPICE program and Intusoft Corp. has sold over 8,000 copies of their IS-SPICE version, both for use on personal computers. Berkeley distributed over 1,500 copies of SPICE, free of charge, in 1993 alone. Commercial versions of SPICE are sold ranging in price from a few hundred dollars on a personal computer, to almost $100,000 in some cases. Not only has Don’s pioneering work been of vital importance to the semiconductor industry, it also contributed significantly to the formation of a significant, new industrial base—integrated circuit CAD companies.
As well as its clear commercial impact of the SPICE effort, one cannot overlook the educational impact this contribution has had. In fact, the first versions of SPICE, in use at Berkeley in May 1972, were intended to augment instructional laboratories. As Brian Preas mentioned in his nomination, he had counted twelve text books and monographs at the Stanford University student bookstore which contained the word "SPICE" in the title, such as "Integrated Circuit Design Using SPICE." I agree with him when he says that he cannot think of a significant undergraduate electrical engineering instructional program anywhere in the world today that does not use the SPICE program as an integral part of its curriculum. No other EDA tool or technology has had such a broad educational impact.
While there were numerous important scientific contributions in the original SPICE1 and SPICE2 programs, the success of this technology cannot be attributed to science alone. In fact, even the original developers of SPICE admit that there exist "better" algorithms than those used in SPICE, when considered separately. It is the engineering contribution—the way in which the algorithms and ideas were combined—and the unique and open relationship Don Pederson and his students established with industry and maintained for the continued evolution of the programs that have resulted in a family of programs that have lasted almost intact and without significant competition for over a quarter of a century. With the possible exception of the IBM operating system OS/360, I cannot think of any program that has lasted as long, is still in widespread use, and is without significant changes, let alone a particular application program or design tool. Of course, many researchers have tried to improve upon both the basic architecture and the specific algorithms used in SPICE. Any improvement would be very worthwhile—in one company, their extensive use of SPICE is such that even a 2% improvement in run time would save the company enough in computer charges to pay the salaries of the entire eight-person SPICE support group! While many programs have been developed for specific technologies, or which produce approximate results (use simpler models and algorithms) and so can run faster than SPICE on larger circuits, for general purpose, robust circuit simulation, the user community seems always to fall back to the trusted and well-tested SPICE-based solution.
At the time the SPICE program was being developed, there were a number of similar programs under development in major semiconductor and system houses throughout the world, for example ECAP, CIRCUIT, TRSN, Sceptre, and the TRAC. As mentioned earlier, these programs used precursors to many of the techniques present in SPICE and a number of the early innovations in circuit simulation were developed as company-proprietary. However, the original vision of Pederson’s SPICE program development team was to put together the best combination of algorithms, to code them in as flexible and portable a style as possible, and to make them freely available. The only restrictions placed on users was that they should never charge any third party for the SPICE program itself—Pederson considered it a public-domain resource. As a result, new ideas and contributions to SPICE flowed from many sources, both in universities and in industry. Companies as well as universities ported SPICE to many different operating systems and made these versions available in the public domain as well. For example, the first versions of SPICE for IBM computers—TI-SPICE—were created at Texas Instruments and TI itself distributed these versions and updates for many years free of charge to any interested party in the United States. Similar arrangements existed at other sites for CDC Cyber versions, Digital, Prime, Honeywell, Cray, and Burroughs machines. In addition, Don saw the circuit simulator as a means to an end, not as an end in itself. His attention was always directed to the solution of circuit design problems, both through heavy classroom use and in research; the simulator was developed as the best way to achieve that goal. It was the process used to develop the technology, as much as the technology itself, that represents the great insight Don Pederson contributed to the steps needed to develop and transfer the knowledge embodied in engineering research. In fact, Berkeley has continued to use this model, pioneered by Pederson, for its ongoing research in other CAD areas. The Berkeley CAD faculty believe that much of the general EDA success at Berkeley over the past fifteen years can be attributed to the model used for EDA research, as much as the specific details of the projects themselves. Other successful universities, for example Carnegie-Mellon University, have adopted a similar model for their interaction with industry, based on the early Berkeley success.
I know I have painted what might see to some like a larger-than-life image of the impact SPICE has had on the world, but I am also sure that those who have traveled the SPICE path with me over the past quarter century know that it truly is larger than life.
I also know that the image I have painted of Professor Don Pederson seems equally larger than life. But to those who know, Don deserves every bit of praise I have given him and more. He too, truly is larger than life.
So, in summary, on behalf of Mrs. Kaufman, EDAC, and all present here tonight, I am both pleased and honored to be able to present the Phil Kaufman Award to Dr. Donald Pederson, a man who has played a central role in creating the EDA industry, through his teaching and his research, but perhaps most importantly through the leadership and inspiration he has provided to many generations of professionals in our industry today.