Re: Cooment on Version 7 of document

Jeroen -
I had thought that the limiting was done on "vcrit" at nominal
or reference temperature, but now I see that you are correct.
This is a little tricky, because we don't have a temperature
limiting algorithm in Spice to point to, and if you don't
limit the temperature, then your "vcrit" could be negative.

What I had in mind was Spice's fetlim, which uses a threshold
voltage in the algorithm; someone mentioned that BSIM3 uses
the bias-dependent value from the previous iteration. This
requires the simulator to keep track of yet another state
variable.

Since temperature appears in the denominator of exp(v/$vt),
it seems that limiting based on $vt($temperature) would be
at worst conservative relative to $vt($temperature + V(nt));
you would not have the case that the exponential would blow
up because the argument was allowed to get too large.

-Geoffrey

jeroen.paasschens@philips.com wrote:
>
> Dear Geoffrey,
>
> unfortunately, I will not be able to phone in tonight.
>
> I do have a comment, however.
> Just above section 5.5., it is mentioned that the expressions in $limit may not depend on voltage, but they may depend on temperature. Please realise that in aal self-heating models the temperature is actually modelled by a voltage. This could lead to problems in limiting.
> With kind regards
> Jeroen Paasschens
Received on Tue Apr 6 04:52:37 2004