Each of the two EPIC MOS cameras consists of seven individual 600 x 600 pixel CCDs. Because of telemetry constraints, a real time on-board recognition scheme filters out cosmic-ray tracks and exclusively transmits to the ground the information supposedly related to X-ray events. The on-board recognition scheme looks for a local enhancement of signal in flat fielded images. The signal enhancement is searched in 5 x 5 pixel matrix which is scanned over the full image. The signal is defined with respect to a threshold value set by telecommand for each observation. An event is identified if, in the 5 x 5 pixel matrix, pixels above thresholds formed a predefined pattern.
In the case of imaging mode, 32 patterns have been predefined (see figure 12, upper panel). Each one of them correspond to an isolated event i.e. to a zone above threshold completely encircled by pixels below threshold. There are however two exceptions. Pattern 30 can be connected to a pixel above threshold on the diagonal of the center pixel. Pattern 31 can have any of the border pixels above threshold. Pattern 30 and pattern 31 are designed to quantify the amount of cosmic-rays extended tracks.
On-ground calibration of the imaging mode has shown that soft X-rays mainly generate patterns 0 to 12 corresponding to compact regions of X-ray energy deposition. Pattern 0 events are single pixel events. These comprise most of the valid X-ray events with the most accurate energy resolution. For imaging mode data patterns 0 to 12 are the canonical set of valid X-ray events which are well calibrated. Selection of these patterns constitutes the best trade-off between detection efficiency and spectral resolution. However, because they deposit energy below the CCD depletion zone, higher energy X-rays also generate pattern 31 events with a probability of 20% and 50% respectively at 6 keV and 9 keV. Pattern 31 comprises largely cosmic ray events but can also include pile-up X-ray events. A large density of pattern 30, 31 events (and 2629 diagonal events) in the core of the telescope point spread function (PSF) is the signature of a piled response which needs careful analysis (see § 4.5).
In the case of timing mode data, the pattern analysis is purely 1-dimensional (i.e. insensitive to other rows, see figure 12, lower panel), because each timing "row" is actually the sum of 100 true rows, so the rows are not physically related.
High electronic noise affects occasionally some of the MOS cameras CCDs (). This effect appears most frequently, but not exlcusively, in MOS2 CCD5 and is limited to the soft (1 keV) energy range. It has been present in short phases since the beginning of the mission, but became continous on this CCD since Rev.#874. The causes are still unknown, and a correction is not available. SAS users can flag CCD affected by enhanced electronic soft X-ray noise using the task emtaglenoise, which populates keywords LENOISnn in the header of a calibrated event list (a value of '1' means that the CCD is noisy). Optional parameters filterbadccds and filteredset should be used if the user wants to remove all events in all CCDs tagged as noisy by emtaglenoise. If MOS1.evt is the event lists produced by emproc, emchain, or the pipeline processing, use the following command to flag noise CCDs:
emtaglenoise eventset=MOS1.evt filterbadccd=yes filteredset=MOS1_filtered.evt
General recommendations on filtering schemes for the generation of images and spectra are given in XMM-SOC-CAL-TN-0018 , a document describing the current status of EPIC calibration and data analysis.