Parallel Hardware Architectures for the Life Sciences Parallel Hardware Architectures for the Life Sciences

A new type of STEM detector has been developed and tested. This detector consists of 30 rings which are split in quadrants (= 120 channels) and it allows to take further advantage of the image formation principle of STEM: the scattered electrons can be recorded separately according to their scattering angles and eight images can be acquired in parallel. Because of its electron counting capability, this detector is very well suited to analytical and quantitative applications of STEM (e. g. absolute mass determination and Z-contrast). For the implementation of the real-time data acquisition system, an Integer Processing Unit board has been developed. The unit was designed to allow implementation of parallel and pipelined multiprocessor systems. This unit is the basic building block for the fast pipelined and parallel processing system used in the computation of images from the data produced by the 128-channel silicon detector array. Systems consisting of 4, 6 and 16 such Integer Processing units were considered and the performances of these configurations were evaluated. The various channels of the detector can be combined by the real-time hardware in order to obtain images with the detector acceptance angles desired. The minimal time per pixel (t/pixel 4 μs) is approximately as short as in the case of conventional single channel detectors.