Renetlisting after change of design variable

Hi Christoph,

Actually if you spectre is operating in "interactive" mode (the default), then if you only change a variable value it will communicate this to the running spectre process and spectre will change the value in memory and re-run. This allows better performance because it avoids the overhead of restarting the simulator and re-reading the netlist - it should however affect the results. Parametric analysis does the same thing.

The fact that the results don't change is the concern here rather than the lack of renetlisting.

You could try converting to batch mode by using:

envSetVal("spectre.envOpts" "controlMode" 'string "batch")

in the CIW and restarting ADE - in this case a new netlist should be produced and the simulator restarted each time. If you do that, do the results change? If they don't, maybe the circuit genuinely isn't affected by the change. If they do change in batch mode but not in interactive, please contact customer support because we need to investigate the bug.

Regards,

Andrew.

Thanks, Andrew.

I have checked that Spectre works in "interactive" mode by default. If I switch to "batch" mode, the result changes indeed.Which indicates that the circuit is affected by the variable. However, if I switch back to "interactive" there is again no change. So you think this is a bug?

Regards, Christoph

Hi Christoph,

There is a possibility that the circuit has more than one stable operating point and which one it finds can depend upon the starting point. One thing to try in interactive mode is to set rebuild_matrix to "yes" (this is on the bottom of the Algorithm tab in Simulation->Options->Analog). If it still doesn't affect the results when you change the parameter, please contact customer support (it might be worth doing that anyway, because rebuild_matrix=yes shouldn't really be necessary). The case where it might be reasonable is if you have two operating points that are completely reasonable solutions to the circuit, and there's an element of luck to which you get based on the starting condition (this can normally be shown using nodesets to show that you can get either solution depending on the node set you start with). That's best to do via customer support rather than here where it's hard to talk about without knowing the specifics of your design.

Kind Regards,

Andrew.

Hi Christoph,

There is a possibility that the circuit has more than one stable operating point and which one it finds can depend upon the starting point. One thing to try in interactive mode is to set rebuild_matrix to "yes" (this is on the bottom of the Algorithm tab in Simulation->Options->Analog). If it still doesn't affect the results when you change the parameter, please contact customer support (it might be worth doing that anyway, because rebuild_matrix=yes shouldn't really be necessary). The case where it might be reasonable is if you have two operating points that are completely reasonable solutions to the circuit, and there's an element of luck to which you get based on the starting condition (this can normally be shown using nodesets to show that you can get either solution depending on the node set you start with). That's best to do via customer support rather than here where it's hard to talk about without knowing the specifics of your design.

Kind Regards,

Andrew.