Re: gEDA-user: possibly off topic -- minimum spice transistor size?

[sdb@cloud9.net (Stuart Brorson)]

Ngspice (& tclspice) is basically Berkeley SPICE 3f5, with some clean
up (to make it compile on Linux) by the folks in Italy who started the
project.  There is not much new simulation code in it as far as I
know.  I am unaware of any limitations w.r.t. L and W.  But I don't
use those facilities, so they might be limited & I wouldn't know about
it.  

You have two options:  1.  Ditch ngspice/tclspice.  Install LTSpice &
run it under wine.  2. Take a look at the ngspice source & see if
there is a built-in limitation, like e.g. the variables holding L and
W are only ints or small floats.  I realize this is a PITA.

You could also try posting your questions on the ngspice lists; the
folks at MultiGig Ltd who are in charge of the tclspice branch read
those lists & are pretty responsive.

Stuart




> 
> Hi all,
> 
> I started picking through various spice and design packages because of th=
> e
> flaky nature of the windows apps i use in class.  I've been a Linux user =
> for
> about five years and up until recently have only played with the various =
> *nix
> based spice packages out of boredom.  Recently though, after picking up g=
> eda,
> I have run into problems I cannot believe are what they appear.
> 
> When I set a W or L in a netlist to some value less than 1.2um, I receive=
>  an
> error that the dimension is zero.  I cannot believe that the packages are=
>  so
> outdated that they are unable to perform these simulations, are they?  1.=
> 2u
> technology has been around for quite some time - certainly this isn't a
> dead-end.  Orcad PSpice has so far dealt with every dimension I have thro=
> wn it
> without failure.  I've had this problem using spice3f4, ngspice, tclspice=
>  and
> several other small packages i had sitting around on disc.  It isn't some=
> thing
> stupid in the netlist either- i tried every notation i could think of, an=
> d
> compared results from other dimentiosn to Pspice results- it's being
> interpreted the same way in other environments, but works.  Unless of cou=
> rse one of you
> happens to know of a dumb mistake that would act this way.  Hints?
> 
> This 1.2 limitation is killing me.  Next term I wanted to put some resear=
> ch
> into short channel effects in transistors and will be comparing different
> approximations and simulations...  It doesn't look good on the Linux fron=
> t if
> I'm limited to 1.2um- unless of course i start tweaking the model in comp=
> letely
> unrealistic fashion and have an xj of 1.5u...
> 
> Anyone know what the root of this problem is, or how i get around it?  Th=
> e
> model I'm using has Ld << 1.2u, and has simulated reliably in several win=
> dows
> based apps.  A few other models I've tried, give the same error.
> 
> I really appreciate any feedback.  Thanks a lot.
> 
> Regards,
> Phil
> 
> 
> --=20
> NEU F=DCR ALLE - GMX MediaCenter - f=FCr Fotos, Musik, Dateien...
> Fotoalbum, File Sharing, MMS, Multimedia-Gru=DF, GMX FotoService
> 
> Jetzt kostenlos anmelden unter http://www.gmx.net
> 
> +++ GMX - die erste Adresse f=FCr Mail, Message, More! +++
> 
>