As mentioned above and shown in Fig. 99, photons entering the OM detector hit a photo-cathode, which is located at the backside of the detector entrance window. Electrons emanating from the photo-cathode are amplified by the MCPs, creating photon splashes at the location of the OM CCD. The detection of a photon entering the detector is performed by reading out the CCD and determining the photon splash's centre position using a centroiding algorithm, which is part of the onboard software. In the process of centroiding a grid of 88 ``in-memory'' pixels is defined, leading to an array of 20482048 in-memory pixels with an approximate size of 0''.476 on the sky. In the resulting images at some level there is always a pattern repeating on an 88 grid, resulting from a limitation in the algorithm. This spurious pattern can be removed by subsequent processing on ground.
As with all photon-counting detectors, there is a limit to the maximum count rate achievable before saturation sets in. The frame time6of the OM detector is about 11 ms at slowest, so a linearity correction must be applied in the offline data processing for count rates above ca. 10 counts/s for point sources. Both deadtime and coincidence losses contribute to the non-linearity of the OM detector. These effects are corrected by using SAS in the data reduction. Deadtime losses are due to the lack of instrumental response during a frame transfer. Coincidence loss is observed whenever the count rate is such that more than one photon arrive in the same pixel within a given readout frame. Deadtime becomes important for short frametime. On the other hand, longer frametimes are more likely affected by coincidence losses. These effects are quantitatively described in § 3.5.5.
In addition, sources which are too bright can depress the local sensitivity of the photocathode: this is a cumulative effect, so that fainter sources observed for long times have the same effect as brighter sources observed for shorter periods. This places some operational constraints on the instrument. Pointing of bright sources may also yield ghost images in subsequent observations due to fluorescence.
Cosmetically, the OM detectors are good, with few hot or dead pixels, and little global variation in quantum efficiency. Pixel to pixel sensitivity variations on the OM CCD are in some way smoothed by the centroiding mechanism producing the final pixels, as described above. Local sensitivity gradients are negligible on scales up to two minutes of arc.