All, I think that Arpad’s three top level issues are exactly right: 1) assigning models to the system 2) how do you simulate them 3) how to interpret the simulation results I think the next step is to find out what the models need to do, and “how do you simulate with them”. One thing that we need to know for a model are what the ports are. We know that we will always need Vddq, Vssq, pad, and depending on the model also require enable, stimulus and output nodes, and there will be optional nodes for additional supplies, and possibly control logic. Obviously one task is to identify the nodes and how to hook them up. But the more important question is what is the model expected to do? One possible answer is to predict, based on the past and current values of voltages and currents at each of the nodes, the voltages and currents generated by the model. It can do this in the “.subckt” sense as being able to be hooked up to a classic spice netlist in a classic spice simulator (essential the “B” element”), or it can return intelligent information such as equation based impedance, The model could return frequency domain as well as time domain kinds of information. I am far from being authoritative on the requirements of the “model”, but I think these requirements need to be defined. Once the requirements are defined then we can evaluate parameter-based solutions, equation-based solutions, language-based solutions and procedure-based solutions. Walter