ADE design variable:: evaluate desVar expression/equation to scalar before simulation, in order for the netlist to contain a scalar value instead of an equation.

I deleted your original post to avoid having this posted twice in different forums (which is against the guidelines).

Anyway, three points:

1. You can force the expression to be evaluated before netlisting by ensuring it includes a SKILL function which isn't handled natively by spectre (see "spectre -h expressions" for what functions Spectre supports). I changed to (say) dV1/2+truncate(0) or even eval(dV1/2) - maybe the second is clearer - and this achieves that goal.
2. However, given that the callbacks are not invoked when you have a design variable being swept on an instance, I'm not convinced this is really your problem. The workaround above still will leave the instances set to a design variable value and will not expand the value on the instance (that's not really feasible, especially if you have hierarchy in the design and are using pPars() etc - this is covered in my article The dangers of CDF callbacks)
3. One other approach might be to use the ADE Parameters section (available in Explorer/Assembler and also ADE XL). In this case you can parameterise instances (and you could "ratio match" the two instances if you need one parameter to half of the other). With parameters the callbacks are invoked because it knows which specific instance parameters in particular cellViews are being modified.

Andrew

Thank you Andrew.

You are right  - despite the netlist issue of design variable being now scalars (the eval(dv1/2) did the trick), the callbacks are not being invoked.do

Question: Using the ADE Parameters section, do I need to set all the parameters on all the instances or only on a few - end goal being all the instances had their CBs invoked ?

I confess to never having used Parameters, so I am still fiddling around with them to try get the CBs invoked on a few/all instances.

Frank

Thanks

Frank,

frankp said:

end goal being all the instances had their CBs invoked

Why would you need the callbacks invoked on all instances? That seems a strange requirement.

Just parameterise the parameters that need to be parameterised and that should be sufficient. 

Andrew

The reason is because currently there are MANY instances that do not get their CB triggered, bevasue one of their field is populated with a desVar. Wanted to see if there was lazy way to deal with this predicament.

Just got access back to cadence - I have yet to succeed building a correct netlist with the 'paramaters' section when I sweep a desVar. Probably need to simplify my testcase first - will get back to this post with my results of simply parametrizing the needed instances then.

But is it possible to trigger all the CBs ? asking out of curiosity for now.

Thanks for your feedback Andrew. As always, much appreciated.

If the the field is populated with a design variable and the callback doesn't work with design variables, calling the callbacks is not going to help. Calling the callbacks when you have used the parameterisation is fine, because it should call it with a number in that case.

If you need to call all the callbacks (presumably you'd need to remove any design variables if the PDK doesn't support them), then you can do that using the SKILL code in my article How to call CDF callbacks procedurally from SKILL to update CDF parameters?

Regards,

Andrew

As I always, I appreciate the help Andrew.

(I had a training week last week so I my attention was elsewhere).

Returning back to this issue now (among others).

Thank you.

Parametrization did the trick!

And you were right, calling  the callbacks when they do not work with design variables does not fix the problem.

Thank you Andrew.