Sims are very useful in determining how well your model of a device is
in describing its performance. Especially for non linear circuits where
analytical non-linear math ability limited in engineering community training.
Like the standard equations we use for biasing, they come from a model
that is a simplification of device behavior. Works well enough that jillions
of circuits have been designed using those inaccurate bias equations.
Its great for showing AC performance as well. Especially in complex networks,
of many poles and zeroes. And if it deviates from measured then modal was
insufficient/inaccurate.
The real problem in industry is models, as supplied by vendors purposely, in many
instances, not complete to hide proprietary design techniques and methods in their
IC.
And there is the propensity of users to not read the simulator setup properly.
like time step, initial conditions, BLAH BLAH BLAH.
So I blame the models (vendors) and users for most sim limitations, not the
simulator.
Of course this does not discuss measurement error and imperfect test systems used to
confirm bench results, a whole other topic of errors and limited capability.
Regards, Dana.