The MIPS-24 pipeline is discussed in further detail in the Gordon et al. paper, as well as Masci et al. (2005) and Engelbrachet et al. (2007), all of which are available through this website. A summary of the most significant steps follows here. Also see Figure 5.1 for a flowchart of the MIPS-24 pipeline.
The Si:As array is read out at a constant rate, but this data rate is too high to allow all of the data to be sent to the ground. Multiple reads of the array are processed on board into a single image consisting of the net integration between resets; the data are fitted onboard by linear regression for each pixel and only the slopes (DN/sample time) for each image are sent to the ground. Another component of the raw data sent to the ground is the 'difference image,' which represents the difference of the first 2 reads. On board the spacecraft, the difference image (a 0.5 MIPS-second exposure) is set to zero everywhere except for those pixels whose count rate is large enough to lead to an A-to-D (analog to digital) saturation in the data ramp over the total image integration. This is performed on-board using a nominal (and conservative) difference threshold value of 600 DN/sample time (where one sample time = 0.5 MIPS second = 0.524288 real seconds) over a 30 MIPS second integration, or equivalently, ~50.3 MJy/sr. Later in pipeline processing, a larger difference image saturation threshold is used, ~63 MJy/sr. Above this level, the slope pixels are replaced with difference pixels. In the end, this yields more reliable fluxes for bright sources.
More specifically, the pipeline re-scales this nominal threshold according to the image integration time since a shorter integration can tolerate a larger count rate before the ramp starts to saturate, and hence bias the slope value. The following scaling is used:
Difference_Sat_Threshold = 1907.34 x (30/EXPTIME) x 0.0447 [MJy/sr]
(5.1)
where the factor of 1907.34 is (1000/0.524288]) DN/real seconds, i.e., the nominal threshold for a 30 MIPS second integration, EXPTIME is the image integration time in MIPS seconds, and the factor of 0.0447 is the conversion factor from DN/real seconds to MJy/sr.
Of course, some sources may saturate the ramp even in the first read, leading to what's called 'hard saturation' and then causing the difference value to be zero. The latter cannot be corrected, and the sources will appear to have 'holes' where sources saturate (usually in the centers of the PSFs) after processing. Also, as the difference pixels approach complete saturation, the signal will drop before becoming zero, so the central holes of hard saturation will be surrounded by pixels of decreasing or even negative slope.
After conversion to floating point and DN/sec, the first step is the 'read-2 correction.' An offset in the second read of the 24-micron array causes the SUR (sample-up-the-ramp) slope to be too large. This effect, referred to in shorthand as 'the read-2 effect,' results in a gradient across the array that depends on the array position and background level. The correction, which is very small (<<1%), is applied in the regular MIPS pipeline as an analytic function subtracted from SUR slope values.
The Si array suffers from the 'droop effect' whereby the output for a pixel is proportional to the photon signal that fell on that pixel plus a signal proportional to the average signal falling over the entire array. (Also see IRS Instrument Handbook for more discussion of 'droop.') Because saturated pixels will cause the droop to be underestimated, pixels are first 'desaturated' using the nonlinearity model and the difference image before the droop correction is calculated. Droop correction is then applied, followed by dark subtraction, nonlinearity correction, flat field division (using mirror position-dependent flat fields), and conversion to units of MJy/sr. Finally, suspected saturated pixels are replaced by those from the first difference image to create reliable slope data wherever possible, resulting in a larger dynamic range. Latent images and potential radhits (cosmic rays) are flagged but not removed at the BCD level. The post-bcd mosaic software is effective in removing radhits from the mosaics (see section 5.2.2).
Starting in S18.12 (the final reprocessing data version for the cryogenic Spitzer mission), the Enhanced BCD (EBCD) pipeline was introduced for MIPS-24 photometry mode. Scan mode remains the same as in previous software versions. The EBCD products generally have improved flat fielding, especially with regard to removing spot residuals (see section 6.3.4, Figure 6.4, and Figure 6.5). Occasionally, the EBCD pipeline will not find a good spot match, but in general, there is a great improvement over previous data products. Parallel data (taken when MIPS-70 or MIPS-160 were the primary arrays) also have improved flat fielding, especially for the 160 Enhanced mode, which had no spot correction previous to S18.12 (see Figure 6.5). For photometry data, both EBCD (new flat fielding method) and BCD (old flat fielding method) products are archived. However, in the final reprocessing, only one version of the post-BCD mosaic is produced, and that mosaic is made from EBCDs. The calibration flat field files for the the EBCDs have a new format of a data cube shifted to match the spot positions of each photometry AOR, with the planes of the cube in order of ascending scan mirror position (CSM_PRED). The flat fields applied are the product of the appropriate spotflat and the overall gainflat. The spotflat used in each EBCD is reflected in the EBCD header keyword CSMLAYER. For more detail on calibration data products, see Appendix D.
Figure 5.1: MIPS 24 micron pipeline.
Figure 5.2 and Figure 5.3 show example MIPS-24 raw and BCD products; Figure 5.4 and Figure 5.5 show example MIPS-24 final mosaics for photometry and scan maps, respectively.
Figure 5.2: Automated MIPS-24 pipeline products : raw data (DCE). Two large islands of bad pixels that are circled are included in the MIPS-24 p-mask (see section 6.5); note cosmic rays, which are flagged - but not removed - by the pipeline. Compare the background here, which varies by 30% before flat fielding, to the much flatter background in the next figure.
Figure 5.3: Automated MIPS-24 pipeline products: single BCD. This is on a different scale than the previous figure; note cirrus structure that has appeared (compared to raw image) in the background.
Figure 5.4: Automatically-produced mosaic combining multiple BCDs from a photometry observation.
Figure 5.5: Automatically-produced mosaic combining multiple BCDs from a scan map observation.
