Re: minutes of: Verilog-AMS conf call Aug 2

In response to the query (below) about table_model, I looked
to see if it could be done with existing AMS machinery.
I found the following problems:

1) The AMS LRM does not provide any system functions for
reading from files. 1364-1995 provides $readmemb and
$readmemh (for binary and hexidecimal data); even if
this was meant to be in AMS, it would be hard to use
for a table_model. The $fscanf function from 1364-2001
would be preferable. Will this be added as part of
LRM 2.3?

2) For the 1-D case, generating the interpolation requires
solving a system of equations of size (order+1)*(numpoints-1).
(For a cubic spline, I think one really only has to solve
a system of numpoints-1 equations for the x**2 coefficients,
and the other coefficients are expressed directly in terms
of them.)

2a) The module doesn't know how big a system is involved
until it opens the data file and counts the points.

2b) It seems onerous to require the module writer to
write a matrix solver in Verilog-A, when nice math
packages are available to the simulator.

-Geoffrey

"Geoffrey.Coram" wrote:
> - Why is $table_model needed at all? David Root of
> Agilent was requesting additional features, including
> stiff splines / data filtering. Action item to
> contact him to see if table_model can't be written
> in Verilog-AMS with $fopen and existing mathematics.
Received on Fri Aug 6 07:39:22 2004