Re: [Gnucap-devel] New gnucap development snapshot

[Dan McMahill]

Al Davis wrote:
> Thanks.  That is what I need.
>
> On Wednesday 18 January 2006 08:39, Dan McMahill wrote:
>
> > The netlist and some gnucap warnings/errors are shown below:
>
>
> > .param A4NP=0 fac_n ='0.3*2.5'
> > + toxn='8E-09+A4NP*fac_n*1.5e-10*0.5'
>
>
> Expressions for parameters are not implemented yet.  It 
> interprets that whole expression as a string.  This is on the 
> todo list.

OK.  This is fairly common.  It is often the case that a foundary will 
use expressions to provide corner libraries.  So you'll include the 
"typ" section from a library which will define a bunch of parameters and 
then it will include the section with the actual bsim models.  For slow 
corner sims, you might include the "slow" section which will define 
different values for the parameters (usually things like mobility and 
threshold voltage) and then include the models.  The expressions are 
very common.


> > .MODEL nch NMOS (
> > +LEVEL    = 49    TNOM     = 25.0    VERSION = 3.24
> > +TOX      = 'toxn'             ACM=12
> > +)
>
>
> > * Parameter and level 49 test
> > ... 3.24 TOX      = 'toxn'              ACM=12 )
> >                                         ^ ? what's this?
> > parameter 'toxn' has no value
> > parameter 'toxn' has no value
>
>
> Quote problem.  I need to fix that.  Try it without the quotes.  
> It is looking for 'toxn', complete with quotes.

No quotes (and removing the expression) made it work.

> Looking at my old HSPICE docs...
>
> ACM is "area control method".  It specifies how the area and 
> perimeter of a diode are adjusted, and how they are applied to 
> the circuit model, for example in the diode parameters and 
> parasitic resistance and capacitance. 
>
> What I will do for now is to add ACM as a dummy parameter, with 
> a default value of 10, which appears to match the Berkeley way.

ok.  Depending on what value you specify for ACM, you also end up using 
a handful of other parameters to specify the diode parameters.  For 
example, HDIF and LDIF vs explicitly specifying AS, AD, PS, PD.  The 
hspice MOS modeling manual spells all of this out fairly clearly.

> I have been hoping to put this off until I have Verilog-A 
> support.  Then this becomes a user library issue.  
> Unfortunately, without these models, users move elsewhere, and 
> the motivation for everything declines.

I have to agree.  Verilog-A support would be very cool and I don't know 
of any free Verilog-A implementations.  But, the ability to take a 
modern library file and have it Just Work is pretty important too.

I did notice that I was able to grab a level 49 model from mosis
(http://www.mosis.org/cgi-bin/cgiwrap/umosis/swp/params/tsmc-018/t58f_mm_non_epi_thk-params.txt) 
and it worked with no modification.  I didn't try to do any comparison 
of results with hspice though.  So this is a big step forward.  It does 
mean that you can probably start using gnucap more seriously for some IC 
design sims in schools.


-Dan