The current implementation of WFI read-out modes doesn't allow to read only one chip nor to read a specific region of the whole array (windowing). The read-out time for binning 1 images is 45 seconds; using a binning of 3, the read-out time falls to around 10 seconds, value which is still too high for fast photometry pourposes.

One option which allows to do fast photometry with WFI is to use the WFI autoguider CCD. This CCD has the same characteristics of the CCDs used in the scientific array and is located on one side of the array; the advantage of using this CCD is that it supports windowing allowing to dowload images with a rate of approximately 2 frames per second.

On the other side, the disadvantage in adopting this solution is that the filters mounted in the tracker CCD window are of poorer optical quality than the ones used for the science array. However, optical quality strongly depends on each filter. As examples, good results were obtained using the Rc filter during the Pluto applulse on 2006-04-09 (whose data was used for the plots below), while on the same occasion the I203 filter was rejected due to the very bad image quality it gave. Also, since off-axis distorsions are of low entity, in such cases it may be optionally chosen to not to use any filter.

It is thus of critical importance verify in advance the image quality for the desired configuration.

Hereafter some plots based on data taken on 09 April 2006 are shown, when Pluto passed very close to a star. For this event, the requirement was to sample with a frequency of one image per second. The adopted exposure time was of 0.6 seconds and binning 1; the windowing box was 170x100 pixels and it was centered at (1410,1950), i.e. at the side of the chip closer to the WFI array. This because of a double reason:

• despite the many cosmetical defects (bad columns, hot pixels, etc.) spread over the chip, that region is one of the cleanest;
• being closer to the optical center, distorsions are much less and the gradient in illumination is almost negligible.

The plot below shows the distribution of the effective exposure time (as recorded in the header) for all the images (around 4000) taken during the event. In the fast photometry configuration, the WFI shutter always remains open, i.e. integration is controlled via only the powering cycle of the CCD.

As it can be noted, the accuracy in exposure time is better than 0.01 seconds, while the precision is of the order of milliseconds. The following plot, instead, shows the global overheads between two successive images for all the frames. These values were derived from the exposure times and the MJD dates of the beginning of the exposures, recorded in the header of the frames.
In this case data is almost equally distributed among two well defined levels separated by 0.09 seconds and in only very few cases the overhead was greater than 0.5 seconds or smaller than 0.30.

The combination of the two previous plots is given in the image below, showing the distribution of the time separation between two successive images, i.e. exposure time + overhead. The bi-partition of data is of course still present A particularity: in each line the spread of data seems non-existent. This is due to the number of digits used to record in the header the values of MJD which allows to reach the precision of 0.004 seconds while for the exposure time the precision is of 1 millisecond.

SciOps: instructions for fast photometry can be found here.