The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design Languages >> Verilog-AMS >> Spectre vs Hspice for VerilogA
https://designers-guide.org/forum/YaBB.pl?num=1305939826

Message started by newic on May 20^th, 2011, 6:03pm

Title: Spectre vs Hspice for VerilogA
Post by newic on May 20^th, 2011, 6:03pm

I tried out both Spectre & Hspice to simulate a close loop system that written in VerilogA (using cadence ADE). The results are difference. Do you think spectre result is more accurate than hspice result. Why the results have such difference?

Title: Re: Spectre vs Hspice for VerilogA
Post by boe on May 23^rd, 2011, 10:11am

Newic,

please be more specific (see also forum guidelines). E.g. how big is the difference, what kind of system did you model, ...

B O E

Title: Re: Spectre vs Hspice for VerilogA
Post by newic on May 23^rd, 2011, 5:22pm

since it is a close loop system( PI CDR), the difference is actually quite big (hspice default accurate setting).
Tried out several methods on hspice accuracy setting. I only able to get matched results when i set hspice RUNLVL=6 & METHOD=Gear.

There is a mismatch even i set RUNLVL=6 & METHOD=TRAP.

Title: Re: Spectre vs Hspice for VerilogA
Post by Geoffrey_Coram on May 24^th, 2011, 6:58am

Some behavioral models simulate poorly because the variables the author chose aren't scaled properly, so the tolerances (appropriate for V's and I's) allow gross inaccuracies in the model variables.

Title: Re: Spectre vs Hspice for VerilogA
Post by newic on May 24^th, 2011, 7:24am

I got to check my ttol setting.

Besides the ttol setting, what else could cause the difference?

Title: Re: Spectre vs Hspice for VerilogA
Post by boe on May 25^th, 2011, 2:24am

Newic,

E.g., a simulation of a system that is close to the stability boundary may depend on the solver algorithm used and its implementation.

B O E

Title: Re: Spectre vs Hspice for VerilogA
Post by newic on May 25^th, 2011, 2:38am

thx for the feedback. There is no stability issue here (a lot of margin).

At least for now, spectre is more accurate for a given ttol.
I think the difference can be reduced if tightening the ttol.
No time to verify this yet.

Title: Re: Spectre vs Hspice for VerilogA
Post by boe on May 25^th, 2011, 4:52am

Newic,

A model bug can of course cause this as well.

BTW, plots would be nice...

B O E