User Guide

Worst-case analysis
313
Tips and other useful information
VARY BOTH, VARY DEV, and VARY LOT
When VARY BOTH is specified in the .WC statement and
a model parameter is specified with both DEV and LOT
tolerances defined, the worst-case analysis may produce
unexpected results. The sensitivity of the collating
function is only tested with respect to LOT variations of
such a parameter.
For example, during the sensitivity analysis, the
parameter is varied once affecting all devices referring to
it and its effect on the collating function is recorded. For
the worst-case analysis, the parameter is changed for all
devices by LOT + DEV in the determined direction. See
the example schematic in Figure 89 and circuit file in
Figure 90.
Figure 90 Circuit file using VARY BOTH.
In this case, V(3) is maximized if:
•
Rwc1 and Rwc2 are both increased by 10% per the LOT
tolerance specification, and
•
Rwc1 is decreased by 5% and Rwc2 is increased by 5%
per the DEV tolerance specification.
The final values for Rwc1 and Rwc2 should be 105 and 115,
respectively. However, because Rwc1 and Rwc2 are varied
together during the sensitivity analysis, it is assumed that
both must be increased to their maximum for a maximum
V(3). Therefore, both are increased by 15%.
Figure 89
Schematic using VARY BOTH.
WCASE VARY BOTH Test
Vin 1010V
Rs 121K
Rwc1 2 3 Rmod 100
Rwc2 3 0 Rmod 100
.MODEL Rmod RES(R=1 LOT 10% DEV 5%)
.DC Vin LIST 10
.WC DC V(3) MAX VARY BOTH LIST OUTPUT ALL
.ENDS
T
h
e purpose o
f
t
h
e tec
h
nique is to re
d
uce
the number of simulations. For a more
accurate worst-case analysis, you should
first perform a worst-case analysis with
VARY LOT, manually adjust the nominal
model parameter values according to the
results, then perform another analysis with
VARY DEV specified.
Pspug.book Page 313 Wednesday, November 11, 1998 1:14 PM