Re: [Ngspice-devel] Re: Ngspice-devel digest, Vol 1 #43 - 1 msg

[Paolo N. <p....@ie...>]

On Fri, 2002-07-19 at 20:49, Ala...@ao... wrote:


> Anyway, about the M parameter. If you look at my MOS3 code,
> you'll see that at that stage I did it by looking for all the equations
> and trying to figure out which ones had to be adjusted for "M".

Yes, I was displaced by the different approach taken.

 
> But somewhere I saw somebody doing it the other way, i.e. just
> multiplying all the conductances by M as you load them into the
> matrix. I can't remember where or how I saw that, but once I
> thought about it a bit more, I came to the conclusion that it's
> by far the safest way to do it. (Luckily, it's the easiest, too :-)

Mhh, you are saying that since the interface with the simulator is the
matrix and the DEVask or DEVmask routine, a change there is sufficient.
This seems right. What about internal geometry-dependent parameters ?
They will not be scaled, anyway, you may always say: "Hey dude, if you
need a wider device, you'd better use the "w" parameter!"  
 
 
 
> On the testing front, I think the technique of puting M devices
> in a circuit, and one with an M parameter in an identical circuit,
> should work fine. That's what I did. I only checked out the DC, AC
> and Transient analyses, though.

Well, I will make the necessary modifications to pole/zero analysis
which is very similar to ac load routine and noise, hope not to break
anything!

Bye,
Paolo 

-- 
--
Paolo Nenzi
GnuPG public key available on keyservers
Key FingerPrint: FCC5 FDA4 4A31 EDE6 982D  AABE 6D48 9D3A 8FC0 A981