How to get a fixed time step with Hspice

[skrainbow]

I'm trying to find the relationship between time step used and the simulation accuracy.But as Hspice resolve to some adaptive time step calculation step that the final time step in the '.tr' file is much smaller than the prescribed time step I have written in the netlist.
Is there some options for fixing the time step exactly as I have set in the netlist?

Thank you in advance

 

[erikl]

Is there some options for fixing the time step exactly as I have set in the netlist?

I think you can't fix the internally used timestep, but you can fix its timestep output to the .tr file, see the PDF: View attachment HSPICE.OPTION_INTERP.pdf

 

[skrainbow]

That's too bad.It's not exactly what I'm expecting.
Another question:
Have you ever used Powerspice provided by IBM?Is it the same that I'm not capable for deciding the internal time step as in Hspice?Is it because all the internal time steps are decided by the internal algorithm and solving integral equation needs?

Expecting your reply.
Thank you in advance.

 

[erikl]

... Have you ever used Powerspice provided by IBM?

Sorry, no.

Is it the same that I'm not capable for deciding the internal time step as in Hspice? Is it because all the internal time steps are decided by the internal algorithm and solving integral equation needs?

It's the same method like all SPICE programs work, AFAIK.

 

[steve_guo]

Try spectre, an option named strobeperiod could be a choice.

 

[kgl_13gr]

Strobeperiod only affects at what time the output is saved, has nothing to do with simulation.
If you do turn to spectre, look also at this thread: https://www.edaboard.com/threads/31621/.

Extrapolating from there though, does HSPICE have a minstep and/or a maxstep option?

 

[erikl]

Strobeperiod only affects at what time the output is saved, has nothing to do with simulation.

Right. Spectre's STROBEPERIOD .option does the same as Hspice's INTERP .option.

... does HSPICE have a minstep and/or a maxstep option?

No such options. HSPICE (like many other SPICE programs) uses quite a lot of control options (e.g. ABSVAR, DELMAX, DVDT, FS, FT, IMAX, IMIN, ITL3, ITL4, ITL5, TIMERES, to name just a few) which affect or control the internal timestep. There is, however, no option which fixes this timestep: wouldn't be beneficial for convergence.

For plotting and/or further usage of equidistant time data results, the INTERP resp. STROBEPERIOD options are quite useful.