Solido, analog EDA tool automation, and Bob Pease on DFM: “Take that Carver Mead”

Hi folks, we just posted my write up of startup Solido Technology’s first tool release, the SolidoStat and we also posted a write up of Synopsys’ new SystemC DesignWare model portfolio.

I think the SolidoStat announcement in particular is very interesting because as I wrote in my coverage of the tool, new tools for the analog space typically fall into two categories: they either are a faster, larger capacity and cheaper version of an existing tool or they require designers make a radical change or addition to their tool flows. However, the SolidoStat tool seems a bit unique in that it seemingly allows users to dig deeper into a task that they typically do anyway.

Years ago when I was at EE Times, the analog tools beat was covered by Steve Ohr, and so it was one of the tool categories that I wasn’t so familiar with. Then circa 2001, Steve left EE Times to do other projects for the publisher and eventually took an industry analyst gig at Gartner Dataquest. When he left, it meant Goering and I had to start covering analog design tools, as well as digital design tools. So to get up to speed, I did a series of briefings with the analog tools players and with a few analog designers and learned a few surprising things (which I admit are somewhat generalities that may not apply to all of you folks doing analog black magic voodoo). First, is just that: analog design is black magic voodoo. It isn’t just ones and zeroes, it isn’t black and white like digital, it’s all shades of gray and every other color in spectrum and every other shade of all those colors—you have to account for every variable in analog design, RF is worse, and microwave is worse yet. Two is that the tools are arcane. I remember attending a panel Ohr was moderating and the discussion was something about the state of the art in analog design tools. What was the state of the art? there wasn’t one—folks are using spice and essentially the same tools they were 15 years ago? I thought: There is no such thing as synthesis in this space, no HDLs, what??? Three, (I don’t yet fully buy this statement), is that analog designers by nature are super conservative and are reluctant to change and thus if anyone were to introduce a phenomenally great automated tool, it probably wouldn’t get adopted in the first place because analog designers just don’t like change because “they are artists” as one EDA vendor dominant in the analog space once portrayed to me.

Then last year, or was it the year before, I attended a panel on DFM and was surprised to see analog guru Bob Pease was on the panel (DFM’s usually a topic associated with digital standard cell design). At any rate, Bob went off on a rant on how spice simulation wasn’t even accurate. He walked over to an old school overhead projector, waited for it to warm up, and started throwing real foils (old school plastic foils) on the screen showing a circuit design that spice said couldn’t be designed. He then proceeded to lift up a breadboard the size of one my early surf boards above his head saying something to the effect: “here is the design that the spice simulator said I couldn’t design—and it’s working and is a chip in production today…take that Carver Mead.” (Mead is one of the father’s of EDA, though not the inventor of spice).

Wow! I’ll never forget that. I didn’t get his point immediately but in retrospect I think what Bob was saying was that even the first fundamental building block of tool automation, spice, couldn’t be trusted—therefore how can you trust all these other tools, including synthesis and DFM, that are built on top of spice? It’s essentially the junk in junk out argument demonstrated with a giant breadboard…thanks Bob.—BTW, I don’t remember what the other panelists said in their presentations or if there was even a theme but Bob’s presentation has stuck with me. Spice isn’t even reliable? Holy smokes, analog design is hard, black magic voodoo.

So with all this background, I’ve thought, wow, here is an area (analog design tools), that really has been underserved from an EDA standpoint (and also from a trade journal standpoint) and could use more attention and hopefully as a result progress and receive some greater degree of automation. It’s one of the reasons, especially now that I’m at EDN, that I’ve been paying so much attention to the Cadence Pcell/Skill/open access controversy. Also, I can’t help but have great respect (and or pity) for the folks (or fools) who are brave enough to try to introduce tools for the analog EDA market (I think they must be very idealistic or insane, maybe both). Creating an analog EDA tools startup has to be a tough business when the tools you are trying to sell are going up against a juggernaught EDA company that has dominated the space forever and is very territorial and protective of that space (Cadence gains 1/3 of its revenue each quarter from its analog tools lineup) so much so even it’s main competitors haven’t put up great efforts to compete. On top of that, you seemingly have to sell new tools to a user base that is reluctant to adopt new technologies and some users don’t even trust tools that are considered mature in the first place.

So I’m wondering what you folks in the analog design world think of the EDA industry? Is it serving you as well as it could? What do you think of EDA startups trying to introduce new tools and new degrees of automation and the notion that analog designers are not willing to try new things? Any insights will help my ongoing education on this very tough subject.