How to change a transistor model parameter with ADE variables?

Hi,

that variable should still work.  Did you put it into the .cdsinit before the virtuoso launch? It's not working on the fly.

Kind regards,

Marc

We tried:
- putting it in the.cdsinit file
- putting it in the .cdsenv file
- directly typing it in the Virtuoso log window

Nothing worked. The tool shows no error or warning during start-up when reading the .cdsinit file. Nevertheless, it does not show up in the Cdsenv Editor. Here a screenshot:

 In the meantime, we also checked if we can implement a workaround where we put those variables in a .scs file that is read in along with the other model files. No matter if we load it first or last, it does not seem to work.

Hi,

that setting will make sure that the all the model include files are above the  parameters section in the Spectre netlist.  If you say it wont work for you,  you have "parameters"  before the "includes"?  

Kind regards,

Marc

The bahaviour is actually correct: In my netlist, the includes of all the .scs files are above the parameter section, i.e. the netlist looks something like this:

simulator lang=spectre
include "path/to/model1.scs"
include "path/to/model2.scs"
include "path/to/model3.scs"
parameter dvth=30m

Spectre even issues a warning that looks something like this:

Warning from spectre in `<device_name>', during circuit read-in.
WARNING (SFE-2946): "/path/to/model.scs": `<device_name>': Parameter `dvth' redefines parameter of same name defined at higher level. Local parameter value will be used.

All of that looks correct to me: The order in the netlist is right and the warning is issued, as it is supposed to be. The reason I believe that the environment variable is not working correctly is because the log message says that the warning ocured during the read in of the model file, which makes me believe that the parameter (the variable I set in ADE) was read first. 

Anyway, despite the correct looking netlist, I can see no changes in the transistor's threshold voltage.

The reason it doesn't show up in the cdsenv editor is that (for some reason) the netlistModelFileFirst cdsenv variable for spectre is private (and hence undocumented). I have no idea why this is, because it's documented for hspiceD - I suspect it's an oversight.

It does (as you say) work.

The problem you've got is possibly indicative of the fact that the parameter dvth is being defined as a subckt parameter (that's typically what that message means - it's not telling you that the global parameter has been redefined). If it's a subckt parameter, then setting it globally won't help. Without knowing which models you're using, it's hard to be sure... 

Perhaps contacting customer support would be a good idea?

Andrew

Thanks for the clarification.

The MOSFETs are indeed defined as inline subckt. Is there a way to overwrite those parameters with ADE variables (or any other way)?

It's not the fact that they are an inline subckt (or any kind of subckt) that's the issue, I believe. It's if there is a parameter statement within the subckt that's the issue - that means that the parameter is defined locally to the subckt (although it could then be override on the instance, provided that the CDF in Virtuoso is set up to allow it to be specified on the instance). 

Without seeing your models though, I can't confirm this is the situation you're facing...

Andrew

Dear Andrew

Thank you for your help. What information (files) do I need to share for you to be able to tell?

That's why I suggested customer support - I doubt you can share the model files here (which is what I would need to see). Knowing which PDK you're using might be enough, but given that I'm going to be on vacation after tomorrow I'm not going to be able to realistically check the details of a specific PDK.

Regards,

Andrew

That's exactly why I hesitated to directly share model files / log files. Fortunately, we were able to figure it out; so I can answer my own question here.

The hint that there could be parameter statement within the subckt was the right clue: In our PDK(s), there were two variables to add a offset to vth. One was called dvth, and the other one dvth_t. The former one is defined locally in the context of the inline subckt and cannot be modified in the way I intended. The latter one is actually the correct one. If I change the variables to that name, everything works as intended.

Thanks to Andrew and Marc for their help.