RobG
|
Occasionally I'm having a problem with Spectre getting stuck on a particular point in the transient run. Instead of marching forward a few us at a time it inches forward at fs for hours.
This usually happens when I'm powering up or down.
I'm simulating a switched capacitor MOS circuit including a bandgap bias that uses pnps. I've checked all the obvious nodes for oscillation and can't find anything, but the circuit is so big I could have easily missed it. In my powered down state I tied floating nodes to Vdd or Vss.
Tiny changes in the circuit or even changing the length of simulation "fix" the problem for that particular corner, but then it crops up at another corner.
My simulation tolerances are "moderate" and I loosened ibstol to 10p (from 1p).
Is there any way to check which node it is getting stalled on, or loosen the accuracy when it gets stalled?
|