The IRCAM/TUFTI system has a gain of 6 electrons per ADU.
Typically STARE mode observing results in a noise of 56 electrons and NDSTARE 47 electrons. Hence to be background limited, counts on background need to be 2-3 times greater than the square of the readout noise, or, for NDSTARE, counts need to be around 700 ADU.
Significant non-linearity sets in for IRCAM at levels below those stated in the old manual. For Standard readout bias settings, data will be 2% non-linear at 8000 counts and 5% at 12000. To retain linearity keep your counts on any target of interest below 8000 counts with Standard ND-Stare readout (or 33000 in Stare mode). The linearity correction for NDSTARE images is given by
true/measured = 1 + 3.3e-06(measured)
A postscript plot is available showing the linearity correction for NDSTARE IRCAM images (add 25,000 to the counts for the corresponding STARE level). ORACDR will correct standard readout data for non-linearity.
In Deepwell mode, used in the thermal, the array also behaves non-linearly, however in this regime counts are dominated by the thermal background and your calibrator and target will have very similar total counts in the raw frames, hence the problem is much reduced. Use of Deepwell also removes the sharp saturation effect seen at 16,000 counts in standard NDSTARE images; hard saturation instead sets in around 30,000 counts. The linearity correction for Deepwell images is given by
true/measured = 0.8298 + 8.186e-05(measured) - 1.224e-08(measured)^2 + 9.224e-13(measured)^3 - -3.207e-17(measured)^4 + 4.164e-22(measured)^5
A postscript plot is available showing the linearity correction for Deepwell images (add 25,000 to the counts for the corresponding STARE level). ORACDR does not correct Deepwell images for linearity as these frames are dominated by a high and nearly constant background level.