New Focus Tool for Estimating Hubble‘s PSF

M. Lallo,, C. Cox,, S.-M. Niemi,, G. Hartig,, and R. van der Marel,

Knowing the point-spread function (PSF) in Hubble’s cameras is useful for several common steps in data analysis, such as centroiding stellar images in order to perform astrometry, estimating the total flux in stellar images to perform photometry with small field apertures, subtracting the wings of the stellar image for coronography of faint companions, and removing the instrumental signature of the PSF, by deconvolution, in order to discover the finest possible detail in an astronomical image. The Institute recently released an on-line tool to determine the Hubble focus—the main uncertain parameter in the PSF—which facilitates these analysis operations when the PSF cannot be adequately estimated from the data, and an independent handle on the time-dependent variations of focus is required.

The focus tool is built on our best thermal model of Hubble’s focus position. Knowing the focus, the latest version of Tiny Tim can be used to estimate the PSF. The focus tool and Tiny Tim capabilities are available to the community through a web page devoted to Hubble’s optical variations, and to the monitoring, modeling, and maintenance of its image performance.


Explanation of the new focus tool

Hubble is a Cassegrain telescope, with its primary and secondary mirrors supported by a graphite-epoxy truss. The design of this truss—and its system of passive thermal controls—ensure extreme stability of the optical alignment, especially the avoidance of non-axial movements of the telescope mirrors—so-called tip, tilt, and decenter. Hubble does, however, experience noticeable changes in defocus—the axial position of its focus—due to small, time-dependent variations in the separation of the primary and secondary mirrors. These axial movements are minute—usually between 1 and 5 microns. Nevertheless, they can have observable effects on the PSF in images taken by Hubble’s cameras. Their magnitude and temporal variation pose problems for science analyses striving to achieve the greatest possible dynamic range, astrometric or photometric accuracy, or reveal the smallest image details.

Since the launch of Hubble in 1990, the telescope truss has shrunk by about 0.15 mm—about 3 × 10-5 of its length—due to the slow outgassing of water in the graphite epoxy. To compensate for this drift and to preserve the quality of the Hubble PSF, commands from the ground have adjusted the axial position of the secondary mirror some 20 times. There remain, however, thermal fluctuations in the length of the truss due to the changes in the heating experienced by the spacecraft at different orientations and positions with respect to the Sun and Earth. As a result, the secondary mirror typically oscillates axially with an amplitude of 3–5 microns over a Hubble orbit. One micron of such movement at the secondary mirror (referred to as “despace”) produces at the instrument apertures, a resulting root-mean-square (RMS) wave-front error of ~6 nanometers (of defocus). The thermal changes in focus vary on a timescale too short to allow active compensation. Nevertheless, our new services enable users to estimate the variations in focus at the times of their observations, and to perform limited correction of images long after the fact.

As explained at the website, astronomers use our new online service by first entering the date and time of the exposure in the focus-modeling tool. The tool looks up the truss temperatures at that time in the engineering record, and returns an estimate of the “despace” in microns. The focus-modeling tool can also plot the estimated despace together with measured values, if they are available (Figure 1).

Figure 1Figure 2















After obtaining the value of despace, the user can use Tiny Tim to convert the input despace value into focus error and generate the corresponding PSFs for any combination of camera, filter, chip, and pixel position. The output consists of links to FITS images of the modeled PSFs, together with documentary text files and graphics (Figure 2).

Historically, empirical methods for characterizing the Hubble PSF have been the most successful and effective, and we urge the observer to “self-calibrate” whenever practical. In many cases, however, model-based characterizations are the most convenient—and perhaps the only—method available. In such cases, our new service provides a standard solution.

Currently, we have only limited experience using the new procedure for science characterizations. Therefore, we are anxious to get feedback from users in the community, especially any performance comparisons between the results of our focus model and results obtained by earlier models or empirical approaches.