Geoffrey_Coram wrote on Jan 4th, 2016, 5:34am:I think BSIM4 can model subthreshold behavior, but the question is whether the foundry has extracted the model parameters accurately enough for that region.
Thank you, Geoffrey! I think, you're right, the parameters precision outside the working range is very questionable. From the other hand, the near threshold voltage (NTV) is not something unusual: NOR4 cell, for example, includes 4 pmos between supply and output, so some of that pmos is likely operates at NTV.
Geoffrey_Coram wrote on Jan 4th, 2016, 5:34am:You may also find that you are more susceptible to interconnect parasitics when you operate with a low supply voltage: if there's a lot of parasitic resistance, for example, then the IR drop may push the supply too low.
I'm not sure i clearly understand you. But, it seems not possible to design synchronous circuit for NTV supply because of all things you describe.
I'm planning to use asynchronous (delay-insensitive) digital design which is robust to PVT variations (and to the IR-drop in particular). So asynchronous logic must operate at NTV. The only question is how close the simulation to the real silicon life.