Memory from preceeding observation corrected

We found the appearance of artificial spectral features in some spectra due to memory effects from a preceeding bright observation. In order to identify affected spectra we studied the short (32 sec) dark measurements performed before each PHT-S observation. Reducing the complete sample of these measurements from the mission we determined their average signal levels and typical measurement uncertainties per pixel. If the measured short dark signal of a particular observation exceeded the average level by more than 1 sigma for a a continuous section of the spectrum (i.e. for a group of neighbouring pixels) than the observation was flagged for memory effect. Figure 1. presents an example. From tests on cases where external information on the spectral shape (theoretical models; TIMMI2 observations) was available, we concluded that subtracting the excess of the short dark signal from the astronomical observation per pixel can efficiently correct for the memory effect.

Fig.1: Example for memory effect in ISO_id=62501703. (a) the measured signal of Pix. 93 (lambda=8.45 um) decreases rather than increases with time, due to recent illumination of the pixel by a bright source. The measured signals lie above the expected transient curve (continuous line), leading to an overestimation of the derived flux. (b) the flux excess is also visible when the measured short dark signal is compared with the mission average value.