RE: [SI-LIST] : SCALE factor in HSPICE simulations

About this list Date view Thread view Subject view Author view

From: Ingraham, Andrew (Andrew.Ingraham@compaq.com)
Date: Tue Sep 26 2000 - 10:30:50 PDT


The bottom line is that there is no good way to fix this, except by
insisting that all model vendors do not use the SCALE and SCALM parameters.

In my opinion, SCALE/SCALM are evil and will only cause endless headaches.
They are fine as long as you are creating models STRICTLY FOR YOUR OWN
CONSUMPTION which will never be used by anyone else; but anytime a model is
created by a vendor to be used by his/her customers, or by co-workers, there
must not be a SCALE or SCALM parameter in it! (i.e., both must be left at
the SPICE default of 1.0.)

You've hit one of my HSPICE hot buttons. Grrr.

HSPICE does have the PARHIER option which affects whether parameters are
local or global; and HSPICE support once tried to tell me that it would fix
the problem. However, it does not affect the SCALE/SCALM options which are
always global.

The same can also be said about any .GLOBAL nodes, even GND nodes,
especially (but not only) when they are hidden within encrypted models.
They should never exist on models intended for use by other than the model's
creator. Let me choose how to connect ground, and let me insert package
parasitics without having to reverse-engineer the SPICE model (which I have
had to do many times).

If your model vendor can't or won't fix the model, you could embarrass the
heck out of them until they change their ways.

If you have unencrypted models, you might be able to apply suitable scaling
factors to all device dimensional parameters and fix it yourself. Not
recommended for any but the simplest of models.

Many SPICE model vendors live in a vacuum, and seem to think we work in one.
I never use a vendor's model all by itself. Just about the only reason I
need SPICE models is so that I can simulate the interconnect between chips,
and there is rarely any interconnect that goes from a chip to itself without
touching some other semiconductor device. Connecting together models from
different vendors is the norm, NOT the exception. Get with it, guys!

Regards,
Andy

**** To unsubscribe from si-list or si-list-digest: send e-mail to
majordomo@silab.eng.sun.com. In the BODY of message put: UNSUBSCRIBE
si-list or UNSUBSCRIBE si-list-digest, for more help, put HELP.
si-list archives are accessible at http://www.qsl.net/wb6tpu
****


About this list Date view Thread view Subject view Author view

This archive was generated by hypermail 2b29 : Tue May 08 2001 - 14:29:35 PDT