What Would Joe Do?
Peggy Aycinena is a freelance journalist and Editor of EDA Confidential at www.aycinena.com. She can be reached at peggy at aycinena dot com.
The Dictates of Fate: Andrzej Strojwas receives 2016 Kaufman Award
September 29th, 2016 by Peggy Aycinena
Interestingly, this is the first year that the Kaufman award is being presented for contributions to Electronic System Design, not EDA. Very appropriate given that Strojwas’ contributions are in manufacturing and not design. Prof. Stojwas is CTO at PDF Solutions, which per company CEO John Kibarian has never been an EDA company. And with Kibarian serving as co-chair of the ESD Alliance, the organization formerly known as EDAC has now fully embraced its role across the entirety of electronic system design.
Besides this nod to EDAC’s ongoing evolution, the larger implications in CEDA and the ESD Alliance naming Andrzej Strojwas as this year’s Kaufman recipient are profound: The problems associated with electronic systems are not so much in the design these days, but in the extraordinary difficulties associated with manufacturing those designs. It’s really tough, as you all know, when the structures being manufactured are smaller than the wavelengths of light used to etch them.
Which bring us back to Dr. Strojwas. He has been CTO at PDF for 20 years. Back in the last century/millennium, the problems of manufacturing below 193 nanometers could only have been guessed at, yet the company was already working on the intriguing issues of capturing post-manufacturing data and somehow packaging it up to make it useful: How does the semiconductor supply chain glean vital information about the vagaries of manufacturing a real chip and send it back up to the designers so they can learn from the reality when they put pen to paper to design the next hypothetical?
This engineering of the engineering demands scientific curiosity, steely eyed attitudes towards the realities of physics and material science, and a large dollop of business savvy to navigate between the needs and demands of the foundries and the needs and demands of the designers. Let’s allow Dr. Strojwas to take it from here. We spoke by phone this week after his award was announced.
WWJD: Many congratulations on your award, Dr. Strojwas. In reading the press release, I had to wonder why you did not receive this award sooner.
Andrzej Strojwas: Please call me Andrzej.
The list of former recipients of the Kaufman Award is really quite prestigious, and I definitely feel I have no absolute right to claim that I am worthy. Although [chuckling], I am old and sometimes these things are like Lifetime Achievement awards.
I would also say that maybe I work in an area that is a little bit different from traditional EDA – exploring the link between design and manufacturing, that interface. The reality is, where we are right now with the technology it is impossible not to explore the relationship between design and manufacturing.
WWJD: I admire your humility, but again looking at the press release, there is an impressive group endorsing your qualifications – CMU’s Larry Pileggi ARM’s Rob Aitken, KLA-Tencor’s Rick Wallace, Nvidia’s John Chen, CEDA’s Shishpal Rawat, and UC Berkeley’s Chenming Hu, not to mention of course, PDF’s John Kibarian. You must have done something right to have all of these folks chime in.
Andrzej Strojwas: Yes, I was extremely happy to receive these letter of recommendation, and am extremely grateful they agreed to do so.
WWJD: As we have not met before, can you tell me a little about your background, your move from Poland to the U.S., why you chose CMU? And by the way, although CMU’s Randy Bryant is also a Kaufman Award winner, most of them have been from Stanford or Cal. Are you aware of that?
Andrzej Strojwas: [Chuckling] Yes, but you also need to mention Ron Rohrer. He was also from CMU, and had been a professor there.
Actually, it’s a funny thing. Randy Bryant, Ron Rohrer and I were the only three PhD committee members for Larry Pileggi’s PhD thesis. Larry was joking [after the announcement this week], that finally his entire thesis committee has been awarded the Kaufman Award.
WWJD: So how did you end up at CMU?
Andrzej Strojwas: It all started in Warsaw, where I was an undergraduate student.
Prof. Wojciech Maly grabbed a bunch of us and started us in the area of simulation of the manufacturing process. He started that work in Warsaw – I did my master’s thesis with him – and then he went to a conference in the U.S. and met with Stephen Director.
Steve invited both us – Maly as a visiting faculty member, and me as a graduate student – to come to CMU. I was already doing my PhD in Warsaw, and working as a junior faculty member there, but I completed my PhD at CMU.
At the time, Director was doing research into the statistical design of integrated circuits, but I was more interested in design for manufacturing for integrated circuits. It was a good fit and I finished my PhD in a couple of years.
And it was then I felt like the story of my lifetime became clear: I have made very few decisions of my own, it is fate that has dictated my moves.
As I was finishing my PhD in December 1981, marshal law was declared in Warsaw and the rest is history. That [development in Poland] made the decision for me to settle in the U.S.
Then in 1982, the SRC created two centers of excellent for IC design. One was at UC Berkeley and one was at Carnegie Mellon. In fact, it was actually a joint center [housed on the two campuses], and my research was part of it. And that is how it all started.
WWJD: Do you still have family in Poland?
Andrzej Strojwas: Oh, yes. My wife and I met when we were 14 entering high school, where English was an elective. We got married in our third year of undergrad studies – a long, long time ago – and had a couple of children.
I came to Pittsburgh in August of 1979, but at that time it was not possible to leave Poland with an entire family. It took more than a year for my wife and children to join me, and then we were re-united.
I finished my PhD, and then she did her PhD in mathematics. She was teaching for a while, and doing research at Carnegie Mellon, but then my crazy life started with a lot of travel. She sacrificed her career to take care of the children. Something which she regrets from time to time, as she was a very promising mathematician.
Of course, we were very lucky with our children. They did extremely well. Our daughter is a lawyer, and our son has a PhD from Harvard Business School.
WWJD: Like your children, my husband is also a first-generation American. He hardly knew his relatives in the old country.
Andrzej Strojwas: Yes, the hardest part of [immigrating] was being away from our parents as they were aging. Eventually both of my parents passed away – that was the hardest part.
Another thing that was hard is that in Poland there are big families where the children can listen to the stories of uncles, aunts, grandparents. Yes, it is always a tradeoff to [embrace the opportunities here] while missing the chance to be with your family there.
WWJD: Do your children speak Polish?
Andrzej Strojwas: Yes. My daughter was 6 and my son was 2 when they came to the U.S. My son was basically learning Polish and English at the same time. Interestingly, he is the one who has insisted on speaking with us in Polish, but their spoken Polish is excellent for both of them.
Eventually we were able to take them to see family in Poland, and now we are taking our grand kids.
WWJD: Returning to your Kaufman Award, how have you managed to continue with your academic duties and also hold a position in industry?
Andrzej Strojwas: First of all, I should make a more general comment. An award like this is an individual award, but it’s really a team award.
Over the years, I have graduated 48 PhDs, countless masters students, and certainly this award is a testimonial to my students – plus luck. Coming from Poland and having the opportunity [to work in this technology].
Returning to your question about PDF, I was lucky to have John Kibarian as a student. Not only is he absolutely brilliant, but he always had an inclination for business. [Chuckling] He was the only PhD student that I knew who would read the Wall Street Journal every day.
It was really very fortunate for me, that the founders of PDF decided that what we had been working on in our research group at CMU – heavily supported by the SEMATECH Rapid Yield Learning project – had a chance to be transplanted into industry.
The other part of the equation is Carnegie Mellon. I must say that throughout this time period – even through different administrations at the university – they have been extremely good and patient with me, allowing me to do both activities.
You know, there is a huge benefit to PDF because of the opportunity to do advanced research with us, and then move some of the ideas [into industry].
Most of the early employees of PDF were my former PhD students and that continues, although now they are only a small fraction of the employees.
But I think, from the university’s viewpoint, they have an understanding that in my research area, this work cannot be done in isolation. There is no semiconductor industry in Pittsburgh – the administration understood that the action was somewhere else – so after 5 years of PDF being in Pittsburgh, the company moved to Silicon Valley. The CMU administration and the department heads had to come up with for my time sharing.
Of course, there is an understanding that you have to fulfill your duties as far as your teaching, research, and committee work are concerned, but if those things are well attended to, you are welcome to explore.
WWJD: I think it requires a great deal of energy, this multi-tasking between industry and academia.
Andrzej Strojwas: I am not sure about the energy required, but once you get involved …
We are exploring the bleeding edge of technology at PDF Solutions and this requires very advanced research, so there has continued to be very good synergy between my research group at CMU and what is needed for PDF.
Also, because a lot of design for manufacturing was really started a long time ago – although perhaps we were doing it ahead of that time – if you look at some of the early papers, the early research into these problems, there were important foundations being made in the technology, but the industry was not ready for it.
As it became a total necessity, however, the natural way to [implement the technology] was to do it in a company.
Moreover, in this area where I have had access to the newest process technologies – it is a very different [opportunity] to do this in industry, applying our work based at the university.
Early on, as well, I spent a sabbatical at Texas Instruments working on a revolutionary new manufacturing process which was single-wafer manufacturing. Towards the end of that sabbatical, they offered to transplant the work to Pittsburgh, to move the equipment from the TI pilot line to CMU because we already had a clean room.
Unfortunately, I had to politely decline. The problem was, it cost so much to maintain the equipment, and upgrades were impossible because the costs were prohibitive. It just needed to be done in industry, and that is what [motivated] me to continue working with PDF.
Also, all of my PhD students would spend semesters and/or summers at leading-edge semiconductor companies to be sure that our research was still relevant. It turned out eventually, that they could sanitize the data enough from their [experiences] to base their PhD theses on their work.
Of course, the industry has now changed with [the advent] of foundries and fabless companies. And even though PDF Solutions started by working with vertically integrated [enterprises], the company began to have many clients among the bleeding-edge foundries and fabless companies.
WWJD: Does CMU own the patents that PDF bases its work on?
Andrzej Strojwas: Yes, originally PDF commercialized the platform licensing the technology from CMU. But then PDF became much broader and it wasn’t just about manufacturing simulation, but much more about doing test structures and consulting for the semiconductor industry.
So although initially the technology was licensed, eventually the company’s work went beyond that. PDF made a lump payment to the university and moved on from there.
WWJD: A long history of cordial relations between the university and PDF?
Andrzej Strojwas: Absolutely, definitely cordial. That was the only way it could work.
WWJD: Over the years, have you seen an evolution in the attitude and openness of the foundries with respect to their customers, a willingness to release manufacturing data into the wild?
Andrzej Strojwas: First of all, the information is not released into the wild.
In fact, one of the characteristics of the way we work with the foundries is that we are very tight-lipped. Their IP is the process, and although we are helping them to ramp up those processes, we never transfer any information that’s proprietary for a given foundry.
That is an absolute necessity for us. We cannot afford to slip ever. That would be the end of PDF Solutions.
Point number two: We have been investing a lot at PDF in the research and building of the entire infrastructure for characterizing the processes, very important in developing the technology.
There are a number of ingredients in this – test structures, massively parallel testers for fast testing, and a customized set of analysis tool just to provide the unique set of capabilities that the foundries need. This is at the core of what we do at PDF – even the largest foundries in the world are relying on us for that.
The other part of the equation is the fabless companies. We work closely with them, providing a similar infrastructure, but customized to what they need, the types of products they do, their design styles. We allow the fabless companies to characterize the process capabilities and optimize them to their design processes.
Again, this is very propriety, which is why we do this work one fabless company at a time.
In the end, it’s a win-win situation for everyone. The foundry is incentivized to provide fast ramp-up for the processes, and the fabless company wins by having more information to optimize their designs to the newest technology node.
WWJD: I have often wondered why the foundries don’t do this type of work in-house?
Andrzej Strojwas: It’s a matter of resources versus what you can accomplish.
At PDF, we have been spending $30 million on R&D and we have an outstanding team of people doing this work. At some point, it is just more economical for the foundries to hire PDF rather than do it in-house.
And the same is true for the fabless. All of the fabless companies have foundry interface groups, but we are not competing with the foundries OR the fabless companies. We are enabling them all to do their jobs better, more efficiently, and that is reflected in the ROI.
If the foundry can bring up their process faster, it has a very precise, quantitative benefit. If the fabless company can bring their product to market by the Holiday shopping season – the penalty for missing that market is humongous – they also benefit.
And that is how we function. But we have to be ahead of the curve and constantly come out with new offerings – like design for inspection. A brand new capability to provide valuable information in-line, which is important for foundries who need to see where the problems are.
At the center of this, of course, and key to the PDF team is the genius of John Kibarian. He has constantly pushed things ahead of the curve to come up with unique solutions that make us relevant to the most bleeding-edge nodes. In fact, now we have a number of engagements at 7 nanometers.
WWJD: Speaking of shrinking nodes, what do you think about the future of CMOS? Is Professor Hu’s FinFET going to save the day for Moore’s Law?
Andrzej Strojwas: Of course, it depends on who you ask. And it depends on your favorite definition of Moore’s Law.
There are a number of articles that say Moore’s Law ended at 28 nanometers, where as Intel would claim it still continues. But actually, it’s all a question of economics: What are the needs of the industry?
If you look at consumer types of products, it’s all about the cost – as long as you can actually decrease the cost and put in more functionally. So 7 nanometers is happening, 5 nanometers is going to happen, while 3 or 3-1/2 are still a little bit of an open question. But those are all with traditional architectures.
Clearly the nature of computing is changing. Perhaps traditional Von Nauman architecture is not the right way to go. So you will see changes in the architecture, particularly in what people are calling neuromorphic computing.
That will affect the way the future process nodes will develop. Whether it is Moore’s Law, or More than Moore, the question is almost secondary.
You will clearly need to change the architecture in such a way that you improve performance and power. And what has to be done to do that? Humungous amounts of memory that’s all integrated.
Right now you waste most of the delay and power in taking data in and out, and that is where the 3D structures will have an affect. There is room in all of this for more novel memories, and that is how things will develop. With traditional scaling in 2D there is a limit of how much you can accomplish, so 3D integration will help solve this.
WWJD: Just for the record, what was the title of your PhD thesis?
Andrzej Strojwas: My title was “The Diagnosis of Manufacturing Problems”, and that was the subject of John Kibarian’s thesis as well. Although mine was more simulation-oriented and his was more test-structure oriented, they were both squarely based in this area.
WWJD: And if you could do your PhD again, what would you work on?
Andrzej Strojwas: If I were to do it today, I would probably go into exploring the new architectures. But I would want to do it in such a way that it is relevant to the technological capabilities.
Looking out 5 years, that is where I would go right now in my own PhD, if I were do it again, so I did originally, I would still be trying to be ambitious and cover the full spectrum from the technology to the actual manufacturing.
For example, right now the graduate courses that I teach explore both the technical roadmap and the design of mobile platforms and servers – what is necessary to do these things?
My students do a couple of projects in these classes – one using today’s current scenario, and one that looks 5 years ahead. These designs are not being done in all of their full gory details, but the student has to decide which process node, the number of cores, how many memories.
WWJD: There has been so much work done in Design for Manufacturing, what grade would you give DFM today?
Andrzej Strojwas: The grade has definitely been improving over the years because it just had to be done, but the grade right now would still only be a B or B+. There is so much room for improvement.
But I would also say that if I were to grade it over the last 20 years, it has taken us way too long to get to this grade. I don’t think the industry as a whole was doing enough, for example, in exploring regular paths to layout. We should have jumped on that much, much earlier. Now, of course, it’s happening by necessity.
Dealing with an efficient improvement in the process control, in spatial distribution on the wafer and so on, this also took us way too long. But we are getting there now, and in some sense are finally doing it much better.
But again, most of it is because we have no choice. Without doing it, it would be a colossal disaster.
WWJD: In talking with you, it seems as if you have almost had a perfect life, your work a success and so closely aligned with your interests.
Andrzej Strojwas: There is a certain multi-dimensionality to a perfect life. From the point of view of scientific curiosity you are right, but it has not necessarily been an easy life.
In all of it, however, John Kibarian has been really fantastic. I like to think that for a good father, one who is super proud, you don’t mind if your kid is always getting much better than you are.
I have that similar feeling for John, and some of my other PhD students at PDF Solutions. I can only stand and admire what they have been able to accomplish. And I am very happy that they still want me to work with them.
That is why I have to conclude that something like the Kaufman Award is really a team award. It involves fate and a lot of luck in who you have the honor to work with over your lifetime.
Andrzej J. Strojwas, Keithley professor of Electrical and Computer Engineering at Carnegie Mellon University, has been selected as the recipient of the 2016 Phil Kaufman Award for Distinguished Contributions to Electronic System Design (EDA). The award is presented yearly by the Electronic System Design Alliance (ESD Alliance) and the IEEE Council on Electronic Design Automation (CEDA).
The award ceremony and dinner will be held at the Fourth Street Summit Center in San Jose, Calif., Thursday, January 26, beginning at 6:30 p.m. Dr. Strojwas is being recognized for his pioneering research in the area of design for manufacturing in the semiconductor industry.
According to Dr. Larry Pileggi, the Tanoto professor of Electrical and Computer Engineering at Carnegie Mellon University, virtually every fab in the world, as well as a vast majority of chips manufactured today, use methodologies Dr. Strojwas developed.
“Since the 1970s, he has done as much as anyone to co-optimize IC design and manufacturing, driving the more sophisticated use of design data in fabs.”
“At CMU, Dr. Strojwas and his colleagues Dr. Wojciech Maly and Dr. Stephen Director realized that semiconductor yield involves more than controlling contamination in a fab,” says Dr. Rob Aitken, R&D fellow at ARM, Inc.
“It is also a consequence of design decisions. This insight led to the observation that yield influencers could be modeled, and that yield could be predicted by building a set of systematic test structures and analyzing the results. This systematic approach in many ways has enabled the fabless semiconductor ecosystem.”
“Dr. Strojwas’ contributions to improving and streamlining fabrication in the 1990s cannot be overstated,” adds Rick Wallace, CEO of KLA-Tencor.
“Deciding which wafers to inspect and where on the wafer to look were significant decisions. Dr. Strojwas developed a novel methodology using product layout design information to determine critical area-based sampling, reducing the amount of inspection time required and making in-line inspection affordable and effective. All major semiconductor fabs today employ these methods.”
Semiconductor manufacturers have benefited from doubling of yield learning rates, resulting in billions of dollars in increased revenue since the 0.35 micron node.
“Dr. Strojwas was instrumental in enabling this by using statistically accurate test chips and simulation for rapid yield learning,” states Dr. John Kibarian, CEO of PDF Solutions, Inc., and co-chair of the ESD Alliance. “His contributions have become industry standards in yield improvement.”
“Dr. Strojwas has published his research work extensively, including three books, more than 80 journal papers, and nearly 250 conference papers,” comments Dr. Shishpal S. Rawat, president of CEDA.
“He has successfully transformed his research into developing ULSIC designs with superior manufacturability as well as diagnosing manufacturing process issues that continuously improve line yield for these chips.”
“Andrzej has been more effective in transferring university EDA research to industry than nearly all other academics,” remarks Dr. Chenming Hu, TSMC distinguished professor emeritus at the University of California at Berkeley, and the 2013 Phil Kaufman Award recipient.
“The result is an exemplary four-decade long career that has significantly advanced the art and science of design for manufacturability.” John Chen, vice president at NVIDIA Corporation, observes: “Dr. Strojwas’ ‘Design for Inspection’ methodology allows us to ‘detect the undetectables,’ allowing the industry to achieve the perfection required to cost effectively manufacture today’s chips containing billions of transistors and tens of billions of contacts and vias.”
“The relationship between design and manufacturing has never been more important, and we must give Dr. Strojwas great credit for recognizing early on that design for manufacturing needed to be a key element of chip design automation,” concludes Bob Smith, executive director of the ESD Alliance.
“His work was instrumental in bringing design and manufacturing together in a way that has benefited the semiconductor design and manufacturing communities as well as the broader electronics products markets.”
Dr. Andrzej J. Strojwas, the 2016 Phil Kaufman Award Recipient Andrzej J. Strojwas is the Joseph F. and Nancy Keithley Professor of Electrical and Computer Engineering at Carnegie Mellon University.
Since 1997, he has served as Chief Technologist at PDF Solutions.
He has held positions at Harris Semiconductor Co., AT&T Bell Laboratories, Texas Instruments, NEC, HITACHI, SEMATECH and KLA-Tencor.
He received multiple awards for the best papers published in the IEEE Transactions on Computer-Aided Design of Integrated Circuits and Systems, IEEE Transactions on Semiconductor Manufacturing and IEEE-ACM Design Automation Conference.
Dr. Strojwas is a recipient of the SRC Inventor Recognition Award.
He was the Editor of the IEEE Transactions on CAD of ICAS from 1987 to 1989 and served as Technical Program Chairman of the 1988 ICCAD and Conference Chairman of the 1989 ICCAD.
In 1990, he was elected IEEE Fellow.
Dr. Strojwas received a Master of Science degree in Electrical Engineering from the Technical University of Warsaw, Poland, and his Ph.D. from Carnegie Mellon University in Pittsburgh.
Tags: 2016 Phil Kaufman Award, Andrzej Strojwas, Carnegie Mellon University, CEDA, Chenming Hu, Design for Manufacturing, EDA Consortium, ESD Alliance, IEEE Council on Electronic Design Automation, John Chen, John Kibarian, Larry Pileggi, Moore's Law, PDF Solutions, Randy Bryant, Rick Wallace, Rob Aitken, Ron Rohrer, Semitech, Shishpal Rawat, SRC, Stephen Director, Texas Instruments, Wojciech Maly