Tilted Wave Interferometry has been invented and developed over the last 12 years as a exible and very fastmethod to test precision aspheres and freeform surfaces effciently. It measures surface deviations full field,with high lateral resolution, without any null compensator like CGH and without moving the tested part whilemeasuring. The test of non-spherical optical components is a topic of high relevance for optics industry, ascurrent optic designs rely heavily on those elements, since the small form factors and high performance of actualdesigns would be impossible with traditional spherical functional surfaces. As all precision components, aspheresand freeform surfaces need accurate inline quality control. Because of the high exibility and high measurementspeed of typically less than 30 sec., the TWI is well suited for close integration into the fabrication chain. Due tothe special illumination scheme, the first implementations of this new interferometer have been of Mach-Zehndertype. In this contribution we demonstrate, how the tilted wave interferometer principle can be implemented ina Fizeau configuration. The benefit of this configuration compared to the Mach-Zehnder configuration is thecommon path feature. Here, the reference beam and the measurement beam follow the same optical track insidethe interferometer, making the interferometer much more robust against temporal environmental inuences suchas vibrations and air turbulences. At the same time, the form tolerances of the interferometer components inthe common path area can be relaxed. These advantages of Fizeau configuration are well known. However,the multiple source illumination of the tilted wave interferometer leads to the generation of multiple referencewavefronts that can be disturbing. We therefore present a new TWI setup that avoids these problems. It relieson a new illumination design with four sets of illumination patterns that each generate their own referencewave. The new approach has been implemented in a lab setup and shows in first measurements the expectedimprovements in stability. We tested the system performance too in extensive Monte Carlo simulations. Thecommon path approach showed a reconstruction error of the test specimen of up to an order of magnitude lowercompared to the Mach-Zehnder configuration.
展开▼
