August 31, 2009
Like Father Like Son
Please note that contributed articles, blog entries, and comments posted on EDACafe.com are the views and opinion of the author and do not necessarily represent the views and opinions of the management and staff of Internet Business Systems and its subsidiary web-sites.
It is not uncommon for children to go into the family business. Nor is it uncommon for children of lawyers and doctors to follow the profession of their parents. Nevertheless I do not recall interviewing an EDA executive who mentioned their father was in the EDA business. Perhaps it is because the EDA world is so relatively young.
At DAC I interviewed Martin Vlach, the founder and CEO of Lynguent. His father Jiri Vlach was born in Czechoslovakia in 1922. During the Second World War and occupation by Germany, the universities were closed for the Czechs and the only possibility to continue some studies was in special courses. This is how Jiri got to electrical engineering in 1941. After the war he studied electrical engineering at the Technical University of Prague and worked in a communist country until 1968. Another occupation, this time by the Soviet Red Army, led to a permanent exile to Canada. In 1967 he was invited as a Visiting Professor to the University of Illinois. In 1969, he received the offer to join
the Department of Electrical Engineering of the University of Waterloo, where is he now a Professor Emeritus. He was elected IEEE Fellow in 1982.
Would you please give us a brief biography?
My original background is that I finished high school in Prague, Czechoslovakia and immigrated with my father first to the University of Illinois for a year and then to the University of Waterloo. I got interested in simulation and modeling in the early 70s. I worked with a lot of the pioneers in that field including my father who was the pioneer in modeling and simulation from the 60s. Eventually I started a company called Analogy in ’86 in Portland. I stayed with Analogy through its IPO and acquisition by Avant!. About a year later I quit. I do startups apparently. I started Lynguent in 2001. My technical claim to fame is as the author the MAST hardware description language and
creator the SABER simulator, which I consider the first analog HDL.
You mentioned your father. I do not remember anyone’s father coming up during my many interviews. Unlike, say the medical or legal professions where children often follow in their parent footsteps. Perhaps it is because EDA is such a young industry.
He was Professor of Electrical Engineering, now Professor Emeritus at University of Waterloo and a Fellow of IEEE. He has published a number of books on the subject (Computer Methods for Circuit Analysis and Design with Kishore Singhal and Basic Network Theory with Computer Applications). The first exposure I had to what I would call EDA today is the software that was written around ’64 or ’65 that my father did and ran on a Mintz computer. It had to be in a monastery in Prague, because it was the only place that was cool enough. There was no air conditioning.
I remember seeing on the Lynguent website that rather than get venture funding, you bootstrapped the operation.
Yes. We bootstrapped it primarily through two sources. One was SBIR (Small Business Innovation Research) and BAA grants from NASA/JPL and DoD. Also there was private funding. I invested a lot of my money in the company. Actually, my father did also. We were also successful because Alan (Dr. Alan Mantooth, Chief Scientist) is a professor at the University of Arkansas. Arkansas is very generous with their tax credits. We got probably close to $1 million in tax credits so far. So we have a few angeles.
Lynguent is located in Portland. Any reason for that?
Because Analogy got started in Portland and I fell in love with Portland. I am not going anywhere. That’s why we are in Portland. We are also in Prague. A large part of my investment is in the form of land that my great-grandfather bought in 1920. It was nationalized by the communists and eventually came back to my mother and me. I converted the land into EDA software.
Lynguent: Any significance to the name.
I like both natural and computer languages. As I said, I design languages professionally and I speak several. So linguistics is important. We are very good at it. We are fluent. Linguent with an “i” was taken as a website. That was the nature of things in the 90s and 2000s, so we put in a “y”. That’s where Lynguent comes from.
Tell me about Lynguent.
Lynguent is an integrated modeling environment, meaning a tool that makes it easy for people to create models. Having a tool is not enough. We have to have libraries of things, building blocks from which you can construct the models. Those are kind of Lego bricks to snap together to make models. I remember with my kids, you buy these things with which you can build a ship. There are different toolkits that you can build different style and size model. At this conference we are primarily talking about event driven mixed signal toolkit and Simulink emulation toolkit. At another conference just last week, we introduced a radiation hardened by design toolkit.
Finally, the environment is a concept into which you can plug in language packs. We plugged in the MAST language pack because we are working in the automotive and aerospace markets where MAST is present.
We work in the analog and mixed-signal modeling and simulation space. We generally exclude physical layout and digital except, of course, we do mixed-signal. As an analog mixed-signal simulation company, we provide software tools and libraries with integration to Cadence DF tools. We will have some other integrations later on. This being such a hard field to work in, analog and mixed-signal behavioral modeling, we do find ourselves providing services: modeling services and library development services.
The focus at this conference is SoC design and verification. I like to use a V diagram that shows the level of abstraction as a function of time.
The higher level of abstraction is at the top where we start from the customers’ specification, down to the architecture, to the decomposition into some hardware description models and to the implementation at a low level and then verification coming back up. There being rework, if necessary. The rework becomes more and more expensive as time goes on. What you want to do is to pull this V diagram together (shorten the time) so that you are designing from the top to the bottom and possibly bottom up, and in fact back and forth. Modeling enables that because it spans these levels of abstraction naturally. That is because the hardware description languages
VHDL AMS, Verilog and MAST support this natively. It is like a programming language. You can do things but it is hard to write programs. Just because you know C does not mean that you know how to write good C programs. This is the idea of the tool kits.
We go through Simulink equivalent models and system level down to the behavioral models, all the way to the implementation level. We help people write models. We do not write it for them and we do not automatically synthesize models. Some people claim they can take a C-Spice netlist and generate behavioral models. We do not do that. We do white box modeling. We support many different styles in our toolkits. The language is supported automatically. But, of course, it is the toolkit that makes it work. Simulink style toolkits work at the signal flow level. We do not have all the internals. I just have the information going back and forth. Analog and mixed-signal would be the level where I start introducing voltages and currents specifically. At the event driven mixed-signal what we do is use a technique that had been though of probably a decade or two ago. But it fell into disuse because nobody knew how to write those models. The ides is that you write models of analog components but it is targeted at a digital simulator. So I do not need an analog simulator to simulate these models. You lose a little bit of accuracy there but not much maybe 5% to 10%. Some of the turn-on transients are not perfect but the simulation itself is 1,000 times faster. The V diagram can close together because the system architect can now work up and down the hierarchy to explore different
architectures. The other concept of compact model is the idea of writing the models such as BSIM4, PSP, Mextram, etc. We do not use it as much as in the SoC world but it is extremely important in the rad hard world. We are able to take PDK and without changing the PDK, we can take the model that the PDK is built up from and we can add radiation effects to it with this tool. Now we can push it back out and people can assess whether their standard library and memory cells are rad hard and if not how to improve it.
You can find the full EDACafe event calendar here.
To read more news, click here.
-- Jack Horgan, EDACafe.com Contributing Editor.
Be the first to review this article