The full resolution spectral information is generated in 8 science channels (table 2-6). These employ two types of detectors. For the UV-VIS-NIR range covered by channels 1-5, standard Silicon photodiodes (RL 1024 SR, EG&G RETICON, California) with 1024 pixels are used which are sequentially read out. Additionally, UV channels 1 and 2 are electronically divided into two virtual bands 1a/1b and 2a/2b, which can be configured separately. The SWIR channels 6-8 employ Indium Gallium Arsenide detectors (InGaAs by EPITAXX, New Jersey) specifically developed and qualified for SCIAMACHY (Hoogeveen et al. 2001). In the SWIR channels all pixels are read out in parallel. In order to be sensitive to wavelengths beyond 1700 nm, the detector material in the upper part of channel 6 above pixel number 794 (named channel 6+) and channels 7 & 8 were doted with higher amounts of Indium. All channels have to be cooled to achieve the specified signal-to-noise performance. The operational temperature range is channel dependent and lowest for the SWIR channels 7 & 8 (fig. 2-9).
Trace gas features are distributed non-uniformly over the spectrum. The limited total data rate would therefore prohibit the detailed sampling of those ranges of interest if the full spectrum had to be downlinked as one block. SCIAMACHY avoids this situation by using spectral clusters and co-adding. The 1024 pixels per channel can be sub-divided into a number of Clusters identifying regions where trace gas retrieval will take place. Each cluster can be sampled by on-board data processing applying Co-adding factors to the readout of the pixels of this cluster. This results in an integration time (= pixel exposure time × co-adding factor) which defines how many subsequent readouts of each pixel of a cluster are added to generate one measurement data readout. By appropriately setting the integration time, high or low temporal resolution – equivalent to high or low spatial resolution – can be selected. Thus, depending on the executed measurement states, variable ground pixel sizes as a function of spectral region, i.e. trace gas features, are achieved. Efficient setting of co-adding factors is also required ensuring that the volume of the generated data does not violate the assigned nominal data rate limit of 400 kbit/sec. The measurement data stream finally consists of cluster sequences representing different wavelength regions read out at different rates.