strange simulation curves

[lhlbluesky]

i use spectre (mmsim 7.1 version) for my circuit simulation. but when i see my simulation results, i find that the curve is sawtooth shape in the whole half clock period (Tclock/2, and my circuit has many switches), it looks like that it will never settle at all. i'm confused.

for ex: one part of my circuit can be seen as the folloiwng path:

switch S1, cap C1, Buffer1, switch S2. first, S1 closed, signal is sampled to C1, and then, after a period of time, S2 closed, the sampled signal in C1 transfers to the following circuit, such as a buffer, too.

in this example, the signal in C1 and after S2 are all sawtooth shape during the whole period, so i don't how to find the exact signal value at the end of every clock phase.

maybe my question is not very clear, but pls help me. thanks all.

Added after 4 minutes:

pls tell me the reasons, which reason can cause the problem, from circuit itself or simualtion method or others? and how to solve the problem.

 

[Fei]

Have you tried to turn on the "skipdc" option?

 

[lhlbluesky]

i don't, what is the role of turning on the "skipdc" option? turn on or turn off the "skipdc" option, is there any difference?

 

[lhlbluesky]

i tried to turn on the "skipdc" option, but it doesn't work. any other advice?

 

[dick_freebird]

Change the transient method to something besides "TRAP"
or "TRAPGEAR2"; trapezoidal integration is not damped
enough to squash the numerical residues and sawtoothing
is common. Try GEAR or EULER (the latter being almost
too well damped, great for getting a quiet result but may
be unrealistically so, just as TRAP is unrealistically noisy).

You can also select "conservative" tolerances and drive
down (if not out) the behavior.

 

[AdvaRes]

I think It is a matter of simulation step. Choose a smaller step in the tran option.

 

[lhlbluesky]

thanks all, i will try it again.