"The development of curved detectors allows to enhance the performances of the optical system used (telescope or astronomical instrument), while keeping the system more compact. We describe here a set of five curved CMOS detectors developed within a collaboration between CEA-LETI and CNRS-LAM. These fully-functional detectors 20 Mpix (CMOSIS CMV20000) have been curved to different radii of curvature and spherical shapes (both convex and concave) over a size of 24x32 mm^2. Before being able to use them for astronomical observations, we assess the impact of the curving process on their performances. We perform a full electro-optical characterization of the curved detectors, by measuring the gain, the full well capacity, the dynamic-range and the noise properties, such as dark current, readout noise, pixel-relative-non-uniformity. We repeat the same process for the flat version of the same CMOS sensor, as a reference for comparison. We find no significant difference among most of the characterization values of the curved and flat samples. We obtain values of readout noise of 10e− for the curved samples compared to the 11e− of the flat sample, which provides slightly larger dynamic ranges for the curved detectors. Additionally we measure consistently smaller values of dark current compared to the flat CMOS sensor. The curving process for the prototypes shown in this paper does not significantly impact the performances of the detectors. These results represent the first step towards their astronomical implementation."
Very good dissertation by Georgios Dogiamis on strain in image sensors might go into a bit more depth: https://duepublico2.uni-due.de/servlets/MCRFileNodeServlet/duepublico_derivate_00036522/Dogiamis_Diss.pdf
ReplyDeleteThe difference in dark current between D and E seems only to depend on the temperature (taking into account that the two devices have the same dark current) !? An increase of 5 deg.C increases the dark current by a factor of 3. This does not sound realistic to me. So I do expect that OR I do miss something, or something else is wrong .... ???
ReplyDeleteInteresting that this runs contrary to the only example I know of curved sensors on a large telescope, the DARPA SST. I'm pretty sure the curved MIT/LL CCDs increased both in read noise and dark rate -- at f/1 though that doesn't matter so much.
ReplyDeleteUnfortunately,
ReplyDeletethese numbers are absolutely meaningless, since:
1. the series deviation of each single characteristic listed, is much higher than the deviations observed here and
2. the sample size for each category A-F is just one, so too low to suppress series deviation