In this analysis (actual WCS.SCP file), a reference simulation is first run on the design and a plot of the data is stored to save the simulation results, notably at nodes or devices specified by the user (i.e., Vout, iR2, etc.). Then a new simulation is run for each device containing tolerance parameters, perturbing the parameter by a small fraction. The difference between these two sets of measurements are saved. The absolute value of operations performed on the difference measurements are summed (for the WCS_HI analysis) or subtracted (for the WCS_LO analysis), and saved in the IsSpice4 output file for viewing, and in a format that can be read back into the "Results" dialog accessed from SpiceNet's "Simulation Control" dialog (i.e., for viewing design "measurements"). This is not as rigorous as an EVA analysis or a worst case by optimization, however, it is the most computationally efficient method. In summary, differences in electrical measurements of interest between the reference simulation and sensitivity simulations, are continuously summed in a positive (WCS-HI), or negative (WCS-LO) direction to give a final measurement reading at any design node device specified by the user. You can set the min-max measurement test limits for prescribed nodes or devices across the design by expanding the measurements to "pass with symmetry" (amongst other choices) in Simulation Control's "Results" dialog; as shown below.
The WCS analysis is based on the assumption
that each measurement is a linear function of all of the parameters.
For most moderately complex circuits this assumption is invalid.
Generally you will get tolerances larger than the 3 sigma limits
of a statistical analysis, however, you should run a Monte Carlo
analysis for at least 6 cases and set tolerances based on the
Monte Carlo analysis, usually to 5 sigma, before expanding the
WCS data in order to be sure that non-linear relationships are
taken into account.
