Visjnoe
|
I wouldn't say Spectre is better than HSPICE. Here are some criteria:
1. Usability: HSPICE has a less steep learning curve. To my opinion, this is mostly caused by the fact that people have to learn to use SKILL (LISP variant) to extract things like GBW,PM... I agree, some functions are foreseen under ADE, but it's all a bit more clumsy than just typing simple .meas statements.
2. Simulation time: I have never made a serious comparison, but I frequently run simulations using both Spectre and HSPICE (same conditions). At first sight, I would say both lead to comparable simulation times.
3. Accuracy: a while ago, I thought Spectre was more accurate, because it checks KCL, while HSPICE does not. It is however possible under HSPICE to impose this also using a .option setting. A bit odd, I agree, but it is possible to ensure KCL.
4. Silicon correlation: I have seen bugs where Spectre simulations didn't show any strange behavior, while the errors did pop up under HSPICE. However, this could also be because of differences between the Spectre/HSPICE models delivered by the foundry.
Regards
Peter
|