You can use the

$rdist_uniform(seed, start_range, end_range); for uniform

or

$rdist_normal(seed, mean, sigma); for gaussian

if you make the arguments variables in ADE i guess you should be able to do what you want...

ACWWong,

Thanks for the suggestion. I had thought of sweeping a random core random number generator seed but unfortunately not all of the parameters are in VerilogA blocks. My subcircuit is a mix of VerilogA blocks, resistors, and capacitors.

I think the problem is associated with pPar parameters. I'm using pPar parameters to specify the VerilogA and Spectre primitive parameters inside the subcircuit. Sourcelink gives an example of how to setup in-line subcircuits for Monte Carlo analysis but that does not seem to work with out-of-line(?) subcircuits, circuits defined through a Composer schematic. I would also like to use the Monte Carlo historgram display, which is not available in the parametric plotter. There's a sweep button on the Monte Carlo interface but I see no way to set the limits of the sweep.

-Jess

Hmmm... Perhaps Ocean is indeed one way to proceed. However, that gives me another idea. If I must use a scripting language, I might as well use a potent one. I'm not very familiar with Ocean but I believe its data analysis capabilities are far inferior to some external tools. I am more familiar with Mathematica and Matlab. Both external tools have much better data manipulation capabilities and these capabilities are well documented and easy to use. I could use Ocean to run the basic analysis but then embed the Ocean run command in an external Mathematica or Matlab loop. Part of the external loop could create a file with the random parameters that the netlist imports through an include or model file. I could also record the results of interest with a VerilogA block. In general, another part of the external loop could read the recorded results and alter the design parameters to semi-automate the design and/or data reduction process.

It's too bad there is no easy way to do such a basic thing completely within the ADE environment. Thanks for the suggestion.

-Jess

Cadence helped me find the problem with my Monte Carlo analysis. I knew I had to describe the tolerances and distributions for the variables in a file and then add that file to the model list. I did not know that that file MUST have a ".scs" suffix or the tolerances are ignored. My file did not have any suffix. Adding the suffix solved the problem. Cadence also explained that I need to use inherited parameters for the toleranced parameters inside my VerilogA blocks. There's a thorough description of the inherited parameter trick in Sourcelink.

-Jess