On Monday 02 April 2007 12:44, Patrick Doyle wrote: > I'll take a look at that. I read on Stuart's web page that > ngspice "experienced a burst of development during 2004, and > incorporates a number of new patches which have increased its > stability and augmented its feature set, and is therefore the > preferred open-source SPICE (for now)." So I started by > looking there. > > I was probably a little scared away from gnucap since it > didn't have "spice" in its name :-)
Spectre and Nanosim don't have spice in the name either, yet they are the preferred commercial simulators among those in the know. Some people equate "spice" with circuit simulation. To me that is equivalent to equating "FORTRAN" with mathematical programming, despite that fact that most of it is done in C++. It's long past time to move away from Spice. It has served us well, but we need more. Gnucap is our project to make a more advanced simulator, closer to the commercial simulators like spectre and nanosim. Spice has too much baggage for that. It has pretty much remained stagnant since about 1990. There have been minor enhancements and new models added, but the inside is the same. NG-spice has collected the impovements that were scattered and made some bug fixes, so it is the preferred open-source SPICE. But, it's SPICE. Gnucap was stalled around 2004, for reasons I cannot reveal now but will probably eventually leak out. Development has been very active over the last year. The idea is not to duplicate spice. Instead, learn from all of past work including spice, but not limited to it. Back off, decide how to make something better and do it. It is more interactive, has better scripting, better probe capability. The development version supports plugins, which dramatically adds to its capability in the long run. One capability it has is to accept the models designed for other simulators, such as spice. I'm referring to the "C" code here. The current snapshot has plugins with the complete set of BSIM models, from the oldest up to the new BSIM460 that hasn't made it to most commercial simulators yet. There is also a "spice3f5" set, and a "ngspice17" set. You attach the ones you want, and the rest don't get in the way. I expect soon to be able to use other formats also as plugins. The input language is changing to use Verilog-AMS by default, with Spice through a plugin. I think (and so do several other experts) that the only thing holding back Verilog-AMS at replacing the Spice format is that no free simulators have it. When I have to go back to Spice for something, it is always frustrating. Spice probes are very limited. There is no help when you run into problems. The fourier analysis is not accurate enough to be useful. Gnucap has its drawbacks, but I am working on it. The biggest one is the lack of beginner documentation. _______________________________________________ geda-user mailing list [email protected] http://www.seul.org/cgi-bin/mailman/listinfo/geda-user
