In general data obtained from arrays that have been saturated by exposing them to bright sources is undesirable. When they become saturated, the detector operating equilibrium is disturbed and the calibration of the following exposures may be affected. These issues are particularly important with the 70 and 160 µm arrays because the readout amplifiers can no longer maintain the detector bias with a saturated signal. Consequently, there is increased cross-talk to neighboring pixels and the saturated pixel exhibits long time constant drifts that reduce its sensitivity and shift its calibration.
Nonetheless, in some circumstances, saturation cannot be avoided. The MIPS electronics provide a short-exposure look at each source along with the requested long exposure, so that information can be recovered that would otherwise be lost due to saturated signals. These short exposures provide a measure of source brightness within the first second after the array reset that begins a DCE. Measured fluxes for any source that causes saturation in <1 sec will therefore be compromised. Saturation by extended sources is a more severe problem. If many pixels attached to a single readout saturate in a given DCE, the response of the amplifier can be seriously impacted. The 1 sec saturation limits for point sources, and the 10 sec limits for extended sources, are given in Table 2.7. These limits include the effects of pixel-to-pixel responsivity variations for all three arrays, and expected responsivity changes between thermal anneals for the 70 and 160 µm arrays. Observations of targets or regions that approach these limits must be carefully considered and planned.
Table 2.7: MIPS saturation levels.
Point Source Saturation in 1 sec
Extended Source Saturation in 10 sec
4.1 Jy (*)
70 µm default scale
23 Jy (**)
70 µm fine scale
SED (@ 60, 75, 90 µm)
3 Jy (**)
(*) see also next table (**) see also additional paragraphs below
The general case for observations involves a combination of extended and point-like emission. The implications of saturation for the 24 μm array can be computed by assuming a combination of the effects in Table 2.7, as seen in Table 2.8. For example, if the estimated background at 24 µm is 100 MJy/ster, it uses up 40% of the dynamic range in a 10 second integration. Without the background, a source of 0.5 Jy would not saturate in a 10 second integration. However, only 60% of the dynamic range is left on top of the background, so the brightest measurable source in a 10 second integration is 0.4 Jy (from the 10-sec formula in Table 2.8).
Many users attempt to derive Table 2.7 from Table 2.8. We would like to point out that this is not possible. The values for truly ideal cases are in Table 2.7. The formulae in Table 2.8 actually describe the general case, which includes a combination of point and background flux, so they break down in the limiting cases. Further complicating matters, Table 2.8 also has incorporated into it subframes that are part of the observing sequence AND the impact of some substitution of values for point sources - there is a half-second data frame taken at the beginning of the observing sequence, and we can use the first difference to substitute in values for point sources that are saturated later in the exposure. The values in Table 2.7 are meant to be single-point values, with nothing hidden. The values in Table 2.8 are a better approximation to reality: a mixture of point and background sources, subframes that are part of the observing sequence, and the use of some of these subframes to substitute saturated pixels.
For extended source 24 µm saturation limits, the 10-second limit from Table 2.7 (260 MJy/sr) can be scaled by a factor of 9.5 /(exposure time - 0.5).
At 70 µm, some targets slightly brighter than these limits can be usefully observed. However, the consequences for the immediately following reads are significant, with latent images and a degradation of linearity. If known about beforehand, an attempt was made to schedule such observations near a thermal anneal, but this was not always the case. Note that, since our calibrators are not this bright, observations of objects this bright may not be as well-calibrated as fainter objects.
At 160 µm, the saturation limit for a 3 second integration is about 1 Jy. For sources brighter than this level, up to about 4 Jy, useful data will be obtained on the first few reads, but the brightest pixels will saturate before the end of the integration. As a result, there will be some degradation of the results in the readouts immediately following, but the recovery will be relatively fast. Note that there is no equivalent to Table 2.8 for 70 and 160 µm; the equivalent effects (ramp fitting to just the points before saturation) are already included in the values given in Table 2.7 above.
Several programs had observations with severely compromised signal-to-noise ratios because the observers chose long exposure times (30 sec) in regions where the background due to zodiacal light was very high. In these cases, the slope image becomes saturated during the exposure and a large fraction of the array had the ''soft'' saturated 30 sec slope image replaced with pixels from the unsaturated, but much noisier, difference image (see section 5.1.2). The difference image represents only about 0.5 sec of exposure time. Thus, the use of the 30 sec exposure time has instead resulted in BCD images dominated by only 0.5 sec of exposure time! Better designed programs to detect faint sources superposed on high backgrounds made use of coadding many shorter exposures (for photometry mode, 3 or 10 sec exposures).
A simple calculation using the numbers from Table 2.8 indicates that the extended 24 µm saturation limit for a 30 sec exposure should be 84 MJy/sr. We note that this is the hard saturation level at which even the difference image will be saturated. The approximate soft saturation limits are given in