Interference imaging for aspheric surface testing.

Peak-valley accuracy of lambda/20 over a range of 2lambda is not unusual in an interferometric null test. For the larger dynamic ranges of a nonnull test, however, the fringe-imaging optics degrades the accuracy. We classify the errors introduced and analyze them in the context of both general and third-order aberration theory. We can predict the measurement error from known interferometer parameters, and we illustrate this for a single mirror. The errors are tabulated for the specific case of a fourth-order asphere with 100 mum of sag. We show that the third-order approximation is comparable with exact ray-trace results for this case.