So far as possible within the constraints of data compression and telemetry limitations, each image in the Ge:Ga and stressed Ge:Ga bands is sent to the ground for further processing, which removes cosmic ray hits and other bad data, shifts and adds the frames, and calibrates them to produce the final images of the sky. The Si:As data are fitted onboard by linear regression for each pixel and only the slopes for each image are sent to the ground (along with a difference of the first two reads). This processing is required to limit the amount of data that must be stored by the spacecraft.
32x32 Ge:Ga Array (70 µm)
The 32x32 array readouts consist of CTIA amplifiers which are ''on'' continuously, and whose outputs are sampled sequentially by an on-chip multiplexer. The MUX is cycled at a constant rate to give a read rate of 8 samples per pixel per MIPS second. Each sample is digitized and stored on the spacecraft using lossless compression. The preservation of the detailed data stream for each pixel is critical for this (and the 160 µm) array because of the large hit rate by ionizing particles, as much as 1 hit per pixel per 35 seconds. These hits can only be removed from the data stream in ground processing. The array is reset coincident with a scan mirror flyback, and then reset again half a second later to commence gathering data.
2x20 Stressed Ge:Ga Array (160 µm)
Operations are similar to the 32x32 array. One distinct difference between operations of the two arrays is that when the observer requests a 10 sec exposure, the 160 µm exposure is split into two 5 sec exposures by resetting the array half-way through the full exposure time. Because of the generally high level of background flux at 160 µm, the array saturates too quickly to be useful in many portions of the sky if a 10 sec exposure is used. An added benefit of performing an extra reset of the 160 µm array is superior rejection of cosmic-ray hits. The extra reset is entirely internal to the MIPS commanding software and instrument, and is transparent to observers.
128x128 Si:As BIB Array (24 µm)
The 128x128 array also takes data in a mode similar to that just described for the 32x32 array, but some minor differences are required because of differences in its readout, its larger format, and its differing susceptibility to hits by ionizing particles. The array is read out at a constant rate, every 0.5 MIPS sec, and the resulting data sent to the IRS/MIPS computer for processing. However, the data rate is too high to allow sending every frame to the ground. Fortunately, because it has smaller pixels than the 32x32 array (by more than 1000 in surface area), the cosmic ray hit rate is much smaller and it is permissible to process the frames from the multiple reads into a single image (the net integration between resets) before sending it to the ground. The net signal for each pixel in an image is computed by linear regression to determine the current generated by the pixel between resets. However, there is a loss in dynamic range in this process, since a source that saturates a pixel during the integration will be unrecoverable. To achieve the dynamic range requirements for MIPS, it is necessary also to retrieve the information from the initial 0.5 MIPS sec integration. To do so, the first and second reads of the array are differenced and also sent to the ground.
The readout and resetting of this array proceeds continuously over 0.5 sec, recycling to the same state for the next 0.5 sec. Therefore, the final set of reads for a given integration must start 0.5 sec before the scan mirror flyback. Following reading the array, each pixel is reset during the 0.5 sec read cycle while the scan mirror flies back. A second 0.5 sec reset cycle follows to remove residual signals accumulated during flyback before the scan mirror has settled. Integration on each pixel begins immediately after this second reset. That is, for every scan mirror flyback, there is a one second dead time for each pixel of the Si:As array, and these deadtimes are distributed over a 0.5 sec offset in the order of pixel access and read, with the first pixels beginning integration 0.5 sec after flyback is initiated and the last ones 1 sec after flyback initiation.