Housed in the electronics tubes are the two detector assemblies. Each detector assembly consists of detector window, an S20 photocathode, three Micro-Channel Plates (MCPs), a phosphor screen, tapered fiber-optics, and a CCD (Figure 10). The CCD has pixels, of which are usable for science observations. Each pixel has a size of arcseconds on the sky thus providing a arcminute FoV. The first MCP has pore sizes of 8 micron with a distance of 10 micron between pore centers. The second and third MCPs have pore sizes of 10 micron with a distance of 12 micron between pore centers. The photocathode is optimized for the UV and blue.
Photons arriving from the Beam Steering Mirror enter the detector window and hit the photocathode. Electrons emitted from the photocathode are then amplified by the three successive MCPs which in turn illuminate the phosphor screen. The photons from the phosphor screen are then sent through the fiber-optics to the CCD. This affords an amplification of of the original signal. The detection of photons is accomplished by reading out the CCD at a high frame rate and determining the photon splash's position using a centroiding algorithm. The detector attains a large format through this centroiding algorithm by sub sampling each of the CCD pixels into virtual pixels, thus providing an array of virtual pixels with a size of arcseconds on the sky. Residuals of a pattern (referred to as mod-8) formed by creating the virtual pixels can be removed by ground processing using the UVOTMODMAP software. At present this is not done for Quick-Look data, but it is done for the ``final'' data that are sent to the Swift Data Archive (usually after about one week). Some of the data that were taken in IMAGE or IMAGE&EVENT mode were sent to the Swift Data Archive with the mod-8 noise already applied. Check the processing history in the FITS header for each exposure to determine if this correction has been applied. Unlike most UV/optical telescopes, because UVOT's CCD is read out at a high frame rate, the UVOT is operated in a photon-counting mode.
As with all photon-counting devices there is a maximum count rate limit, which can vary with the size of the frame. The frame rate of the UVOT detectors is 11.0329 ms for a full arcminute frame; therefore, for count rates above approximately 10 counts per second (for point sources) a dead time or coincidence loss correction needs to be applied during the data processing. Details of this dead time correction are provided in Poole, et al. (2008, MNRAS, 383, 627). Approximate coincidence loss corrections are given in Table 10. The UVOT photometry software corrects for coincidence lose. Care must be taken when observing bright sources as the local sensitivity of the photocathode is permanently depressed. Autonomous operations diminish the time spent on bright sources (see the Detector Safety Section). The detector's dark noise is extremely low (a mean value of counts s pixel) and can be ignored when compared to other sources of background noise.