Re: [Ngspice-devel] Fix for Numparam under cygwin

[Steven B. <ste...@di...>]

On 22 Apr 2005, at 16:37, Hitoshi Tanaka wrote:

> Dear all,
>
> I have tested the function of 'NUMPARAM' and I have got some errors.
> Following examples have problems.
>
> (1) The X line is before '.subckt' line.
> In this case, 'X1 1 2 3 4 mos' has been transformed to 'X1'.
> Ngspice outputs "error: unknown subckt: x1".
>
> numparam test
> .param wx=10u
> .param wy=3
> **
> X1 1 2 3 4 mos
> **
> vdd 1 3 1
> vss 3 0 0
> vgg 2 0 1
> vbb 4 0 0
> **
> .subckt mos 1 2 3 4
> m1 1 2 3 4 nch w={wx*wy} l=0.5u
> + ad={wx*wy*1u} as={wx*wy*1u} + pd={(wx*wy+1u)*2} ps={(wx*wy+1u)*2} + 
> nrd={0.01u/wx/wy} nrs={0.01u/wx/wy}
> .ends mos
> .dc vdd 0 3 0.01 vgg 0 3 0.5
> .model nch nmos level=14 vtho=0.5 .end

I have made some small changes to the numparam library that I've not 
yet submitted, but still get the same problem (after fixing the line 
breaks in the above).  I'll have a look at this shortly to see if there 
is a easy fix.

For now it seems that .subckt must be before Xxxx lines that call them.

>
> (2) '.subckt' line has '.params:'. This causes the program freeze.
>
> numparam test
> .param wx=10u
> vdd 1 3 1
> vss 3 0 0
> vgg 2 0 1
> vbb 4 0 0
> **
> .subckt mos 1 2 3 4 params:wy=3
> m1 1 2 3 4 nch w={wx*wy} l=0.5u
> + ad={wx*wy*1u} as={wx*wy*1u} + pd={(wx*wy+1u)*2} ps={(wx*wy+1u)*2} + 
> nrd={0.01u/wx/wy} nrs={0.01u/wx/wy}
> .ends mos
> **
> X1 1 2 3 4 mos
> **
> .dc vdd 0 3 0.01 vgg 0 3 0.5
> .model nch nmos level=14 vtho=0.5 .end

This is a limitation of numparam's implementation I think.

There is this comment in the src/frontend/numparam/readme.txt...

   The <value> parts are supposed to be default values of the parameters.
    However, in the current version of Numparam, they are not used and 
each
    invocation of the subcircuit must supply the _exact_ number of actual
    parameters.

On my version the program dose not freeze.  I get...

******
** ngspice-15 : Circuit level simulation program
** The U. C. Berkeley CAD Group
** Copyright 1985-1994, Regents of the University of California.
** Please submit bug-reports to: ngs...@li...
** Creation Date: Thu Apr 21 19:47:35 BST 2005
******

Circuit: numparam test

15:  Mismatch: 1  formal but 0 actual params.
15: WY=$;
15: Syntax error: letter [$]
15:  Expression err: $
15:  Formula() error.
  Copies=15 Evals=14 Placeholders=7 Symbols=3 Errors=5
Numparam expansion errors: Run Spice anyway? y/n ?

This slight improvement might, or might not, be due to the fixes that I 
yet to submit.

>
> The following example is able to analyze normally.
> numparam test
> .param wx=10u
> .param wy=3
> vdd 1 3 1
> vss 3 0 0
> vgg 2 0 1
> vbb 4 0 0
> **
> .subckt mos 1 2 3 4
> m1 1 2 3 4 nch w={wx*wy} l=0.5u
> + ad={wx*wy*1u} as={wx*wy*1u} + pd={(wx*wy+1u)*2} ps={(wx*wy+1u)*2} + 
> nrd={0.01u/wx/wy} nrs={0.01u/wx/wy}
> .ends mos
> **
> X1 1 2 3 4 mos
> **
> .dc vdd 0 3 0.01 vgg 0 3 0.5
> .model nch nmos level=14 vtho=0.5 .end

Yes, this works fine.

>
> Regards,
> Hitoshi Tanaka
>

I think we need to treat the numparam library is rather crude and 
'experimental' at the moment.  One of the changes I'm about to submit 
implement a switch so it can be turned on/off before a circuit is 
loaded. I'll arrange it so it defaults to being off.

For the record I've attached the three circuits as files - they'll be 
good as test circuit if we can fix the numparam library

Regards,
Steven