User Guide

Chapter 12 Monte Carlo and sensitivity/worst-case analyses
314
Gaussian distributions
Parameters using Gaussian distributions are changed by
3σ (three times sigma) for the worst-case analysis.
YMAX collating function
The purpose of the YMAX collating function is often
misunderstood. This function does not try to maximize
the deviation of the output variable value from nominal.
Depending on whether HI or LO is specified, it tries to
maximize or minimize the output variable value itself at
the point where maximum deviation occurred during
sensitivity analysis.
RELTOL
During the sensitivity analysis, each parameter is varied
(multiplied) by 1+RELTOL where RELTOL is specified in
a .OPTIONS statement, or defaults to 0.001.
Sensitivity analysis
The sensitivity analysis results are printed in the output
file (.OUT). For each varied parameter, the percent change
in the collating function and the sweep variable value at
which the collating function was measured are given. The
parameters are listed in worst output order; for example,
the collating function was its worst when the first
parameter printed in the list was varied.
When you use the YMAX collating function, the output
file also lists mean deviation and sigma values. These are
based on the changes in the output variable from nominal
at every sweep point in every sensitivity run.
Manual optimization
You can use worst-case analysis to perform manual
optimization with PSpice. The monotonicity condition is
usually met if the parameters have a very limited range.
T
h
is may resu
l
t in maximizing or
minimizing the output variable value over
the entire range of the sweep. This collating
function is useful when you know the
direction in which the maximum deviation
occurs.
Pspug.book Page 314 Wednesday, November 11, 1998 1:14 PM