mark@mips.COM (Mark G. Johnson) (11/27/89)
Our board-design folks run SPICE rather extensively; should they be
using a more modern simulator that "always converges" and is "100X
faster than SPICE"??
They are looking at time-domain behavior on PC board traces, and
presently they trim the simulated network down so it is managebly
sized and runs reasonably fast: 30 CPU minutes on a VAX 11/780
(equivalently).
Here are their requirements:
1. Transient analysis.
2. Ability to simulate semiconductor devices, so that chip
vendor's models of I/O pin drivers and receivers can be
included in simulations. This means Gummel-Poon BJT and MOS
levels 1-3 + BSIM and PN diode.
3. Ability to simulate with "floating" (neither terminal at
ground) capacitors. This is required for crosstalk modeling.
4. Ability to simulate inductors and mutually-coupled inductors.
This is required for handling PC-trace transmission lines,
IC package noise, backplane connectors, etc.
5. Ability to use all 4 types of controlled-sources (VCVS, CCCS,
VCCS, CCVS), and also with neither the controlling element nor
the controlled source grounded. This is required to handle some
of the macromodels, e.g. those by Linear Technology and P.M.I.
6. Optional, but very handy: a good library facility for easy
inclusion of standard modules like "OutPad74F00" so the user
can build up a simulation model quickly.
Do any of the newer-algorithm simulators (waveform relaxation, dynamic
partitioning, mixed-mode, etc) contain all of these?
For example, the circuit I posted last Tuesday (<31896@hal.mips.COM>)
requires items 1-4 above; do any of the better-than-SPICE simulators
accept this input file, converge at DC, run the transient analysis,
and get an answer close to HSPICE?
Thanks,
--
-- Mark Johnson
MIPS Computer Systems, 930 E. Arques, Sunnyvale, CA 94086
(408) 991-0208 mark@mips.com {or ...!decwrl!mips!mark}