Spectre results differ in ADE from SKILL script

[hamidrezah]

I have a simple circuit. When simulated in ADE, I see one set of
results. When simulated using a SKILL script, I get completely
different results.

The only difference between the netlists in two cases was a saveOption
statement missing from the SKILL script. I added that to the script
and the results match now!

But wait a minute, how in the hell saving or not saving some currents
can have any impact on the simulation results?

 

[Andrew Beckett]

On 03/23/12 08:51, hamidrezah wrote:

I have a simple circuit. When simulated in ADE, I see one set of
results. When simulated using a SKILL script, I get completely
different results.

The only difference between the netlists in two cases was a saveOption
statement missing from the SKILL script. I added that to the script
and the results match now!

But wait a minute, how in the hell saving or not saving some currents
can have any impact on the simulation results?

Saving currents will sometimes cause iprobes to be inserted
automatically in the simulation, which alters the matrix formulation.
I'm talking about spectre here (you didn't say which simulator you're
using).

The symptoms you describe above are usually (pretty much always in my
experience) due to the circuit having more than one stable operating
point (you didn't say what kind of simulation you were doing). A small
difference can cause the circuit to converge on one valid solution or
another valid solution - and hence give different results.

Regards,

Andrew.