Discontinuity on DC Temperature Sweep for low temperature

Hi Johanny,

I'm not sure we could guess this without seeing the circuit and simulation setup. I suggest you contact customer support.

Regards,

Andrew.

Maybe the simulator found a different stable operating point for that condition. Could you try to do the sweep the other way around, not from -40 to 120?  instead 120 to -40,  previous solutions
in parameter analysis are used in the following steps unless you tell the simulator not to do that. Doing it the other way around may help the to -40 point to converge to a different point.

HI Marc.

Thanks for your suggestion. Using this approach I have a smoother curve for first derivative.  It seems like low temperature points are more unstable than high temperature points. 

I have also tried to change the homotopy algorithm without sucessful results.

One thing to be aware of is that if the circuit has more than one stable operating point (assuming that's not due to a modelling error), it means that the circuit genuinely does have such behaviour, and all you are doing by sweeping in a particular direction or re-using earlier operating points to influence the next is just increasing the likelihood of hitting one solution over the other. Both are valid though and probably something you should  investigate in case it means your circuit does not work as expected in real life...

Circuit simulations can't really detect if there are multiple operating points - they will just converge to one (which one may vary though depending on the starting point).

Andrew.

You could also try to do a "nodeset" on the critical node, that wont enforce a voltage but gives the simulator  a hint where to search first.  Or you save the initial solution of your nominal run and feed it as starting condition to your param runs.

++

Doing a transient with the sources staring from 0  could help here as well.

I think I saw kind of startup circuits in the past which helped voltage references to find the "right"point, but my circuit designer career is long over so don't ask me how that was done.

How could I do a nodeset or save the initial condition?

You wouldn't set an initial condition, but probably use a nodeset. You can save the conditions using the "write" or "writefinal" parameters of the dc analysis (see "spectre -h dc" for more details) and to read this back in use the readns parameter (these are on the options form for the dc analysis).

I think this is related to your other post, so probably best discussed there rather than here (the forum guidelines ask you not to post on the end of old threads).

Andrew