The critical challenge facing the NSO adaptive optics (AO) program at is developing a solar wavefront sensor. Solar wavefront sensing is difficult because point source sensing targets are never available everywhere on the solar disk. Most nighttime AO systems use Shack-Hartmann wavefront sensors (SH-WFS), in which a point source target (natural or laser beacon) is imaged through an array of sub-apertures. The displacement of each sub-aperture image with respect to a reference is proportional to the wavefront slope within that sub-aperture. Displacements can be measured by simple center-of-gravity algorithms or quad-cell devices. Small sunspots or pores can and have been used as SH-WFS targets: for example, the Lockheed AO system tested at NSO/SP in the early 90's was based on a quad-cell SH-WFS. Such tractable sensing targets, however, are scarce on the Sun, and laser beacons are not bright enough to provide practical substitutes. In general, solar wavefront sensing has to be performed using the granulation, an extended, low contrast and evolving structure, as its sensing target.
Two wavefront sensing approaches are currently under consideration at NSO/SP: 1) a modified Shack-Hartmann wavefront sensor that employs cross-correlation techniques to measure the displacements of granulation images formed by a sub-aperture array, and 2) a spatial filtering approach that uses a mask placed in an image plane to modify the light in a way that makes wavefront gradients visible as intensity fluctuations in a pupil plane. The latter technique, first proposed by O. von der Luehe (1988), can be understood as a generalization of the classic Foucault knife edge test, in which the "mask" is a straight edge placed over the image of a point source. Two orthogonal edges are required to measure the two components of the wavefront gradient. A complex target structure like the granulation requires a more complex set of masks. Typically the partial derivatives of the target structure are mapped in binary form on a transmissive liquid crystal screen, which is placed in an image plane. An image of the pupil is then formed on a small, fast CCD, which records the wavefront information.
There are advantages and disadvantages to both approaches. Shack-Hartmann wavefront sensors are widely used in nighttime AO, and are comparatively well understood. The low (order 2%) contrast of the granulation image formed by a small sub-aperture is not a problem. The Mark II correlation tracker recently completed at Sac Peak has successfully demonstrated its ability to stabilize the granulation image formed by a 7-cm sub-aperture - i.e., a single channel of a correlating solar SH-WFS has been implemented in hardware. To gain more detailed understanding of the SH-WFS approach, we have also recorded simultaneous SH-WFS data and granulation images. The measured instantaneous wavefronts are being used to deconvolve a time sequence of granulation images (see the figure). Matthias Roeser, (PhD student, Kiepenheuer Institut, Freiburg, Germany) visiting NSO/SP, is working on these data. Preliminary results show that the SH-WFS provides phase information with the required accuracy. However, a SH-WFS for a solar AO system operating on granulation requires the implementation of high-speed cross-correlation algorithms. Although the hardware needed to do this is now commercially available, the bandwidth requirements point to a massively parallel system, which would be very costly.
In principal, the "LCD Knife-Edge" approach offers an elegant and inexpensive solution, since much of processing is performed optically rather than digitally. Wavefront slopes are directly coded as pupil plane intensity variations by the appropriate mask. No sub-apertures are involved. Theoretical studies and simulations performed several years ago at NSO/SP by J-M. Conan and R. Radick did suggest potential problems with sensitivity and resolution when granulation is used as target structure. Engineering issues such as noise due to the refresh cycle of the LCD and the imperfect contrast of the LCD mask pose further practical problems. As a first step towards an experimental concept validation, we performed an experiment about a year ago that simultaneously measured wavefront slopes with both a correlating SH-WFS and a Foucault knife-edge test, using sunspots and the planet Venus as high contrast targets. The wavefronts measured were virtually identical (Rimmele and Radick 1996). In December 1996, observations were made aimed at proofing the actual "LCD Knife-Edge" approach, using an active matrix LCD computer display as the focal plane mask. The optical experiment at the VTT was designed and set up by R.B. Dunn.
Various target structures (sunspots, pores, and granulation) were observed, and various approaches for generating a mask on the LCD were explored. A small-format DALSA camera was used to record the pupil images at a rate of 800Hz. The data are currently being analyzed. The principal result emerging from this analysis is that the "LCD Knife-Edge" approach encounters increasingly severe S/N problems as the contrast of the target structure decreases. With a sunspot as the sensing target, we typically achieved a S/N of 3.5. A small pore delivered a S/N of about 2. Granulation, however, produced a S/N of unity or less, even under good seeing conditions. We are currently exploring ways to improve the S/N: unless this can be done, the correlating SH-WFS may be the only practicable approach to solar wavefront sensing.
T. Rimmele, R.B. Dunn, R. Radick, M. Roeser