Chapter 5 Robust Performance Tailoring with Tuning - SSL - MIT

RMS performance, [µm]
400
350
300
250
200
150
100
50
A B C
Nominal PT
Worst−case PT
Requirement
Nominal RPT AO
Worst−case RPT AO
0
0 5 10 15 20 25
Uncertainty, [%]
Figure 3-11: RMS OPD for PT (–) RPT MM (- -) and RPT AO (.-) designs vs.
% Uncertainty: nominal performance (o), worst-case performance (�), requirement
(black –).
squares. Since the PT formulation does not include uncertainty, the nominal PT
performance is independent of uncertainty level and remains at a value of 100.53µm
across the entire range. However, as the uncertainty increases, the worst case perfor-
mance of the PT design increases dramatically. In fact, the worst case for values for
uncertainty over 3% are too large to show on this plot. The RPT AO formulation
does include uncertainty in the cost function and therefore produces different designs
at each value of ∆. The worst-case RPT performance predictions at all uncertainty
levels are much lower than those of the corresponding PT design, while the nominal
RPT performance is larger than the PT nominal and increases with the uncertainty
level. This behavior is the classic trade between nominal performance and robustness
explored in the previous section. As the uncertainty increases the nominal perfor-
mance of the RPT design gets slowly worse, but the design is relatively insensitive to
uncertainty so the worst-case performance stays close to the nominal prediction.
103