Two questions about HSPICE

[yaly]

1. Does the HSPICE simulation accuracy depend on the transistor number in the circuit?
2. If I simualte a circuit with different converge method, is there any difference between each run?

These two questions jump into my mind while i'm running a simualtion now . Can any HSPICer tell me the answer befor i can verify myself?

 

[yaxazaa]

Hspice accuracy is not dependent of nomuber of devices. When you set, it's done. Hspice is a tool to help you verify the results in your mind instead of let it tell you. So before use the tool, in your mind you should know what's results.

 

[leo_o2]

HSPICE accuracy depends on model.
Convergence is not related to run.

 

[cetc1525]

I think maybe it is.I simulate a big circuit in windows&solaris.they are some differents.

 

[chinito]

I have simulated medium sized ckts in PSPICE (windows) and cadence (Solaris) and never noticed any difference. What difference are you talking about.

 

[markty]

The accuracy depends on the spice model.

In my opinion, the options you set may slightly change the reulsts. Anyway, i think the designer can not totally count on the simulation results!

 

[bear7679]

I think the accuracy is not affected by the number of transistors but affected by different converge method.

 

[layes2]

HSPICE accuracy depends on model level
not on size of chip
diff mode has impact on hspice of course

 

[nickooeda]

HSPICE accuracy is very difficult for us to understand because the accuracy is controlled by many parameters and options,the most important is the model accuracy.

 

[Hughes]

Ideally, the simulation accuracy doesn't depend on the number of transistor in the circuit. But since the simulator find the solution by numerical method, the process can not be so ideal. The accuracy will more or less depend on the size of the circuit.

 

[limingchou]

HSPICE accuracy lies in its algorithm and the model of foundary. the model is the important.

 

[playdc2002]

HSPICE accuracy is dependent on models.

 

[wwang147]

HSPICE internally tries different methods to find a converged one. I am afraid it won't try another way when it finds a convergence. It's very hard for the user to judge the accuracy among various approaches.