OPTICAL TESTING OF LARGE TELESCOPES USING MULTIPLE SUBAPERTURES

Abstract

The construction of telescope systems with large apertures (≃10 meters) is currently being planned. These entire telescope systems should be optically tested in a double-pass configuration. The high cost of manufacturing optical flats large enough to test a large telescope has stimulated research on a new type of testing in which several smaller flats, or subapertures, are distributed over the telescope aperture. The problem is to combine the partial data obtained only over the subapertures in order to obtain the wavefront over the entire aperture. It was the purpose of this dissertation to prove experimentally that subaperture testing is feasible. The question of the necessity of phasing the subapertures relative to each other was specifically addressed in the experiment. A brief review is given of two algorithms utilizing Zernike polynomials. A third subaperture testing analysis algorithm, the Stuhlinger method, is developed in this work; this provides raw phase data over the entire aperture of the system under test. A statistical analysis of this algorithm is given. A 6 in. diameter array of seven subapertures was used in this small-scale test. Data were obtained with the array, a monolithic flat, and a mask simulating the array placed over the monolithic flat. The results of the experiment are in good agreement with control data measured with a Zygo interferometer. Data and analysis for the Stuhlinger method are also presented. Error analysis shows that Zernike coefficients derived using subaperture testing are 5 times less accurate than those derived using monolithic testing for the subaperture configuration used here. It is shown that knowledge of the subaperture tilts can produce accurate wavefront information with as few as 30 data points per subaperture, as compared with 750 data points per subaperture if tilts are unknown. In conclusion, subaperture testing indeed functions in the absence of subaperture phasing. Tilt information influences mostly the lower order Zernike coefficients; lack of such information may be compensated by the use of more data points. Algorithms yielding either Zernike coefficients or raw phase data were shown to function. (Abstract shortened with permission of author.)