I had a chance to interview at length Keith Mueller, currently the VP of Worldwide Sales and Marketing at Apache Design Solutions. Keith has had a very successful track record as an entrepreneur in the industry. His partner is Andrew Yang, the company CEO. Keith will share below his opinions based on his past successes about how to create a successful company as well as their current approach at Apache.
We are ~4.5 years old at this point. My last startup by the way was SPC, Silicon Perspective Corporation. This thing is on a very similar track. It has been kind of interesting. We both had the same amount of money, around $5 million. This is pretty small compared to Magma and Monterey, each of whom took around $115 million to get started. Sequence is probably in the range of $80 to $90 million. I don't know the exact numbers. We are kind of on the extreme low end. The other company that Andrew and I were involved with and where we first met each other was Anagram. Anagram was a circuit simulator. It was kind of the first generation HSPICE. In fact we were competing with EPIC at
the time with their TimeMill. The product was called ADM. We had a really tech sounding terms for what the letters stood for. I don't remember it now. Andrew was actually the founder of that company. He was a professor at the University of Washington at the time. He got a couple of his graduate students, locked them in an apartment and fed them food and water under the door until they got the program done. Then we started the company, Soliday. That one took $100K to get started. We sold it to Avant! for $74 million after about a year and a half. You can see the central philosophy of the company: Let's minimize the expenses. Let's not get a lot of marketing hype. Let's not get a
beautiful lobby. Let's not build volcanoes like Magma did. Let's focus on customers and customers' success. Then there's the other extreme that kind of hypes themselves with a lot of money and marketing dollars and goes that route. There seems to be two diametrically opposed approaches to doing a startup. A software company shouldn't take that much cash. There are no NREs, there are no hardware expenses, there's no nothing.
Did Andrew do the same thing with this company, i.e. lockup a bunch of graduate students? Where did the technology come from?
This one he did a little bit more formally. He actually grew up. When he got acquired, he left the University and joined Avant!. He ran the simulation and the StarRC team for a couple of years under Gerry Hsu. He actually retired at that point for about 4 years. When he was “retired”, he did a lot of investing in EDA companies. He was the lead investor in CADMOS. He invested in Sapphire which became part of Sequence. Until the beginning of this year when Sequence had to reorganize and refinance the entire company, he was a major stockholder in Sequence. Mohave was one of his recent investments. He was the lead investor. He sold that one to Magma about a year and a
half ago. He has been doing a lot of that.
The technical founders of Apache are Shen Lin and Norman Chang. These guys came from HP Labs and Shen worked previously at IBM. They weren't EDA veterans. The technical guys were published. They have written a book on “Interconnect Analysis and Synthesis”, so they are well renowned in their field. The common link was that they were inside the labs at HP and at IBM. This is their first venture company. The good news is that we have these very sharp technical guys and we have Andrew and me, who are kind of serial entrepreneur types. This is our fifth EDA startup. It goes all the way back to Silicon Compiler which was acquired by Mentor Graphics, onto QuickTurn which
went public, onto Anagram which got acquired by Avant!, on to SPC which got acquired by Cadence. Now this one! We have some experience as well as some new highly energetic fresh blood working on this new area. I didn't know Andrew before Anagram. I knew the president, Edmund Chang from Silicon Compliers. That's how I got into Anagram. I met Andrew there. In my experience sometimes a CEO can be very good technically and not so good business wise. On the other extreme they can be very good businesslike but not so good technically. Andrew is one of the few CEOs that I have met that is actually very strong on both sides. This is evidenced by his record both being a professor and
obviously knowing the technical side and also being able to fund and envision what's coming next. When you are doing a startup, you want to do something unique. You want to be ahead of the market. You don't want to compete just on capacity and speed because that is very fleeting and in difficult economies it doesn't even get you orders. You have to view something that is coming and maybe isn't quite there but you don't want to be too far ahead, where you can't survive long enough for the market to realize you are right. I think that that one of his real strengths is being able to envision what's coming, what the next issues are and be ready for them when the issue comes to the
marketplace. I think that's what had happened with Apache.
Basically we are focusing on noise and the noise comes from different areas. Our first real endeavor was the power supply but it also encompassed the package. We take package models either RLC or S models and we include the effects of those. One of our products looks at the chip including the power grid, checks the power grid and IOs out through the package onto the board. We can simulate that whole path. As these busses get wider and the currents get dramatically faster and the edge rate go up, on chip inductance as well as packaging inductance becomes significant. We have been extracting on-chip inductance for several years. Only people doing high frequency, 1 GigaHz plus type
designs, need to worry about that at this point. But there are more and more designs getting into that space. Looking from the border of the package including IOs out is an area where they have been different DDR interfaces and some graphics interfaces that are extremely fast. These can cause issues when those IO switch at the same time.
As you go down to 65 nm and below there are things like on-chip variation in this space. We are not playing in all of them at the current time. We are taking it one by one. We now also have tools in place to handle the leakage issue and some of the advanced techniques that TI and Qualcomm and some of these wireless guys are using to reduce leakage currents by headers or switches to completely cut off different parts of the circuit and disconnect even the leakage current. We have the ability to analyze that effect when you turn it back on, the transient effects turning on different parts of the circuit at different times. We are focused on the leakage area as well. Both from an accuracy
standpoint you have to accurately predict the leakage current and then be able to analyze the advanced techniques that customers are using to really control leakages switches become more and more critical.
Signal integrity is an area we just got into. At DAC this year we announced a product called PSI Sidewinder where P stands for power and SI is for Signal Integrity. This is the first product that concurrently analyzes the effects of signal integrity and power integrity. From the chip's standpoint it really doesn't know if this noise is coming from power or from signal. It doesn't really care. It just knows that the timing and some of the other aspects of the chip have been affected by the noise. This is a tool that helps analyze them together.
You can find the full EDACafe event calendar here.
To read more news, click here.
-- Jack Horgan, EDACafe.com Contributing Editor.