Tracking Spectre bias progress?

I'm a relatively new Cadence user, having come to it from custom in-
house tools. One thing that I'm currently missing is the ability to
track the progress of a bias simulation.

In our own version of Spice, there was a convergence tracking number
that we could monitor to determine whether our simulation was
converging or just getting hung.

In Spectre it appears that once I start a simulation and it's actually
crunching the bias, there appears to be NO indicator from Spectre on
the progress of the simulation.

Does anyone else have this problem?
Is there some convergence monitor that I haven't seen?

The only one that I've seen is 'spectre.out' and it isn't helpful in
this regard.

Mark

 

Hi Mark,

Using Spectre, you could setup an option called 'diagnose'. This is a
snippet from the doc ($MMSIMHOME/doc/spectreref/spectreref.pdf):
diagnose (default=no) . Print additional information that might help
diagnose accuracy and convergence problems. Possible values are no or
yes.

Setting up this option is very likely to help you out with your
convergence problems. Besides, I would advice you to go through the
'Correcting Convergence Problems' section in the Spectre Circuit
Simulator User Guide ($MMSIMHOME/doc/spectreuser/spectreuser.pdf).
This section treats both DC and transient convergence issues.

Regards,
Riad.

 

Unfortunately, this didn't give me the information that I want to
see. It did tell me about the 'maximum value achieved' for I,V,U but
it didn't give me any indication of the progress of the bias.

I'm looking for some numerical indicator which show how close I am to
achieving bias during the bias run.

 

markbe wrote, on 04/15/09 16:36:

This doesn't exist, as far as I know. Normally spectre's convergence is good and
fast, so I guess this has never been asked for.

I suggest that you contact Cadence customer support and request this as an
enhancement - some kind of DC Convergence progress indicator. That said,
convergence is not always that straightforward - you can sometimes appear very
close, and then diverge!

We do have similar feedback when doing PSS and QPSS analyses - these give
feedback as to how these analyses are converging. But there's nothing for a DC
operating point (as it's normally referred to, rather than "bias").

Regards,

Andrew.