May 12 2008 9:10AM | Permalink | Email this | Comments (1) |
Blog This! using: Blogger.com | LiveJournal |
Digg This | Slashdot This | add to Del.icio.us
There is a nice article in former Reed Business publication ECN magazine. The title: Software-only simulation: Fact or Phony? really caught my eye. The author, Shelly Gretlein, from National Instruments makes a pretty good case to counter outfits like Cadence and even my beloved Altium when they talk about going from simulation direct to design. All those claims do is show how completely removed from the real engineering world those companies are, and it is not a good sign.
Now most of you know how down on SPICE analog guru Bob Pease is. I have participated in flame wars with him and Analog Devices fellow Barrie Gilbert. Barrie thinks SPICE is just the cat’s meow. It is interesting to note that they were arguing about orthogonal positions. Barrie was pining for SPICE in IC design, Bob was railing against its uselessness in system design. The big difference is that any decent analog IC design company will spend, oh, about 150 to 300 million dollars per year on insuring the process and the transistor models match up. And all that is being simulated in an IC house is a handful of components. SPICE for system design is a whole ‘nuther ball of wax. The models are sketchy at best, you certainly will never get your hands on the transistor-based model that the IC designer created— that would be giving away the IP for how the part is made. So you get varying levels of macro models, most that don’t properly model noise or temperature or a host of other things that will kill your design.
Very often I see engineers that love SPICE because it allows them to be ivory tower elitists that never have to go into a lab and get their hands dirty. Heck, they don’t even want a technician to make up their circuit since they would have to pick up a scope probe. As one of Bob Pease’s many pals, I can assure you he has no fear of soldering irons or scope probes. Neither does Jim Williams or Paul Grohe or Len Sherman or most other analog gurus.
I guess I just wanted to add my voice to Bob’s and this article. If you are doing real-world system level design, SPICE may be a place to test out a couple of minor design factors, but you certainly need to build a form-fit-and function prototype and test the daylights out of it before you can send it into production. Heck, even IC houses that spend 300 million to correlate process to SPICE models still have periodic disasters when the silicon does not work. The more functional companies realize that IC design is not a software project but rather, just a tiny little board layout, only the prototype costs a few hundred grand as opposed to 150 dollars from Proto Express, like us board guys can enjoy.
It is the high cost of a mask set that justifies spending so much on SPICE at IC design houses. Twenty years ago, when the space program was in vogue, some group at Ford proposed making a simulator for cars. A vice-president, I wish I could remember his name, pointed out that simulation makes sense when you have to fire a rocket off and spent a hundred million dollars just to get to your design environment. He pointed out that for a car, you could just open a garage door and drive outside. Sure there are extreme cold and heat, but that is why the automakers have winter and summer proving grounds. Like most everything else, it is analog, a little simulation in the auto business can get them started, but they sure wouldn’t put a car into production without driving it on a real road.
Related entries in: Analog |