2. Observations and data reduction
The observations were performed on 5 June 1993 at the Swedish Vacuum Solar Telescope (SVST), La Palma, Canary Islands. This 50 cm aperture telescope yields a scale of 9.22 arcsec/mm in the primary focus (cf. Scharmer et al. 1985). A magnification system transferred the FOV onto a Kodak Megaplus Model 1.4 CCD camera of pixels with an image scale of 0.125 arcsec/pixel. Typical exposure times were 10 to 14 ms, using a bandpass of 100 Å centered at 4680 Å. The final field size of the recorded images was pixels or arcsec2. The CCD camera was connected to a DECStation via an 80 Mbyte/s DMA interface and the system was configured such as to work as a "frame selection" device: it sampled frames at a rate of 3.7 Hz, determined the rms granulation contrast in a pixel sub-area in real time, and stored the two best frames of a 15 s time interval to disk (8 bit digitization). At the end of the interval the data were written onto 8 mm tape which took about 6 s. In this way the two best frames (of a total of 55 samples) from each 21 s interval were recorded. With this set-up, an 11 hour series (more than 3700 frames) of unprecedented quality and duration was observed. For the whole series the granulation contrast, determined in a field far away from any activity, was mostly between 6% and 10.6% rms without any corrections applied. The best images of this series show details of granular structures near the diffraction limit of the telescope. Some first results on large-scale photospheric motions derived from this series were published by Simon et al. (1994).
The FOV contained the sunspot group NOAA 7519 (Fig. 1) at position N05, E15. The group was in the maximum phase of evolution on the date of observation. The leading and largest spot consisted of four umbral cores separated by three strong light bridges. The telescope drifts and low-frequency components of image motion due to seeing were minimized using a quad-cell sunspot tracker whose working principle was described by von der Lühe (1988). During the observing sequences, the tracker was locked on a small sunspot in the middle of the group, seen in the lower right corner of Fig. 1. Due to the altazimuth construction of the telescope, the observed FOV rotated around this locking point. Since the main aim of the observing run was the study of horizontal motions in the quiet photosphere, the large spot of the group was in the FOV only part of the time. Therefore we selected a subset from 9:54 to 14:20 UT consisting of 760 frames (one per 21 s selection interval) showing the whole sunspot.
After dark current and flat field corrections, we de-rotated the frames around the pivot point. The average photospheric intensity in all frames was normalized to unity to compensate for changes of transparency and/or exposure time. The frames were then registered to remove residual large-scale image motions (rigid alignment). The correction for theoretical instrumental profile of the telescope and the noise filtering were carried out simultaneously by means of a Wiener filter (as described by Sobotka et al. 1993). In order to minimize seeing distortions, the images were destretched using modifications of routines written by Molowny-Horas & Yi (1993). The destretching was applied in three consecutive steps, removing successively large- and small-scale deformations, and residual jitter. The reference frames for destretching were determined individually for each image as averages of time subseries of 5 or 7 frames surrounding the image to be destretched. This enabled us to eliminate the seeing-induced shifts and preserve the real motions. Since we wished to keep the information about waves and oscillations in the umbra, we did not apply any filtering. We find that intensity oscillations in the umbra are negligible and thus have no effect on measurements of UD lifetimes.
In order to retain only the frames of highest quality, for the final analysis we selected frames of contrast % rms and time lags s, yielding 360 frames covering almost regularly the whole time period. This choice of selection parameters represented a compromise between quality and time coverage. It should be noted that the rms contrast for this selection was determined in an area near the spot, so that the values, possibly influenced by abnormal granulation, are slightly lower than the original ones. Comparing both sets of rms values we can state that the final series contains images corresponding to quiet granulation rms contrasts %.
Due to differential seeing, the image quality changes from one location in the FOV to another. To determine the reliability of the photometric measurements, we needed an image quality criterion applicable to the location where the measurement was made, i.e. the sunspot itself. Since an rms contrast measurement in the umbra would not make much sense, we used an auxiliary measure of the image quality - the "image sharpness". This was defined as the mean value of the image after applying the Roberts gradient operator (a diagonal matrix with elements -1 and 1). In our case, the image sharpness was measured in a pixel () area containing the central umbral core, two light bridges, and the inner parts of the penumbra. To check the consistency of image sharpness with the rms granulation contrast we compared the curves of temporal variations of both quantities during the whole series and found that they have a very similar shape and display the periods of excellent, as well as of lower, image quality in a consistent manner.
For the purpose of a time series analysis, the selected frames were interpolated in time to obtain a constant time lag of 44.5 s between successive frames. The image sharpness after this interpolation was reduced on average by a factor of 0.97.
Since our data were not designated originally for the study of sunspots, no specific stray light determination was made during the observation. We believe that the core of the seeing point-spread function (PSF) is very narrow thanks to short exposure time and image selection - this is confirmed by the high values of the rms granulation contrast. However, we have no information about the far wings of the PSF which represent scattered light. Although the amount of scattered light at the SVST is usually below 1% at 4680 Å, we must consider the observed intensities and sizes of UDs as apparent and relative values. This is not a drawback in the case of time series, where temporal changes are of prime concern.
© European Southern Observatory (ESO) 1997
Online publication: March 26, 1998