OK. I figured it out. It is happening b/c I applied the “option scale” in a dumb way. the scale factor gets applied inside of a function during the BSIM4 setup. Unfortunately, that function gets called any time a “dependent parameter” is updated. At the moment, Xyce considers params that contain AGAUSS to be “dependent params” meaning that they may need updating after initialization. At the moment, they are tagged in this way, even if the simulation is NOT doing any sampling. Anyway, the result is that the setup function that applies “scale” gets called more than once, and thus “scale” gets applied more than once, resulting in bad values for L, W.