While designing a thin film coating, the designer should be concerned about
whether the coating can be manufactured. TFCalc has a feature that helps the
designer determine approximately how sensitive a coating design is to random
errors in the thickness and index of layers. TFCalc uses a "Monte Carlo" method to
generate many random designs. Then it displays the results. The plot below
shows how sensitive a 17layer antireflection coating is to a 1% manufacturing
error. 1000 random coatings were computed.
The heavy curve is the performance of the nominal design; the two thin curves
show the range of variation at each wavelength. TFCalc also has the capability
of optimizing sensitivity; that is,
designing a coating that is less sensitive to manufacturing errors.
In version 3.3, extensive changes were made to the sensitivity analysis:
 Worstpossible sensitivity analysis. On occasion, the designer may want to
know what the worst possible performance is for a given manufacturing error. In
the "Set Sensitivity Parameters" dialog, there is a new analysis type: Worstpossible.
When this option is selected, then subsequent use
of the Run menu's "Compute Sensitivity" command will compute the upper and lower
bounds of the design's performance. Unlike the other two analysis types, which
use random numbers to simulate manufacturing errors, the "worstpossible" calculation
is deterministic; at each wavelength, optimization is employed to find the two designs (with thicknesses
within the manufacturing tolerance) having the highest and lowest performance.
 Sensitivity analysis: vary QWOT. The designer may
choose to vary the quarterwave optical thickness of layers. This type of analysis,
which can be used only on dielectric (k < 0.01) layers, is important for studying the
effects of thickness errors when coatings are optically monitored during their production.
 Index variation in sensitivity analysis. In the past, sensitivity analysis
varied only the physical thickness of the layers. Now the refractive index can
also be varied. When index is varied, the layers' optical thickness remains unchanged.
It is possible to vary thickness and index simultaneously.
 Increased the number of trials allowed in the sensitivity computation. The maximum
number of trials was 32,767. It has been increased to 2,147,483,647.
