Spectre bug with gmin?

[Zhiheng Cao]

I am running spectre ver. 5.0.33.120403 and I am not sure
if this is a known bug:

If I set the gmin option all the way down from 1,1e-1,1e-2,... to 1e-9
the simulator output has correct lines like
"Gmin=1nS is too large to noticeably affect result... "
(when gmin is set to 1e-9)

But as soon as I set gmin to 1e-10, then
the simulator says
"Gmin=1S is too large to noticeably affect result... 1 Ohm resistance
is put across every nonlinear elements"
where it should be 100pS.
If gmin is set 1e-12 or "1pS", the output becomes "Gmin=10mS... 100Ohm".
And, the simulation results are totally wrong.

Does anybody have had similar problem?

 

[Andrew Beckett]

On Thu, 15 Dec 2005 09:29:30 -0600, "Zhiheng Cao" <cao@ece.utexas.edu> wrote:

I am running spectre ver. 5.0.33.120403 and I am not sure
if this is a known bug:

If I set the gmin option all the way down from 1,1e-1,1e-2,... to 1e-9
the simulator output has correct lines like
"Gmin=1nS is too large to noticeably affect result... "
(when gmin is set to 1e-9)

But as soon as I set gmin to 1e-10, then
the simulator says
"Gmin=1S is too large to noticeably affect result... 1 Ohm resistance
is put across every nonlinear elements"
where it should be 100pS.
If gmin is set 1e-12 or "1pS", the output becomes "Gmin=10mS... 100Ohm".
And, the simulation results are totally wrong.

Does anybody have had similar problem?

I vaguely recall seeing something like this many years ago when somebody
was setting something using a very small engineering unit (something like
"1.0y"), which DFII understood, but not spectre - but nothing like this.

Are you using the spectre or spectreS interface?

Andrew.