Signal processing techniques for high-rate x-ray detectors

Many experiments carried out at modern synchrotrons, and in future free electron lasers, require that detectors observe events at higher and higher rates. These high count rate demands are motivated by the large amounts of radiation available at new sources. High flux and high brightness sources are capable of creating high event rates that not only permit existing types of experiments to be conducted faster, but also make possible some experiments that probe phenomena with small cross sections. Here we present signal processing techniques that increase the maximum event rates for two types of photon counting detectors of fundamental importance at synchrotrons, energy resolving detectors and position resolving detectors.;In the case of energy resolving detectors, we exploit the time structure of the synchrotron generated radiation to permit the processing of shaper pulses that would otherwise be rejected or corrupted due to pile-up. In cases such as XAFS type experiments, the energy of both the fluorescence and scattered photons are also known a priori. When pile-up occurs an estimation of an event sequence is performed based on periodically sampled observations of the shaper amplifier output. Isolated pulses are processed with a simple threshold decision much like an SCA while non-isolated pulses that produce sequence segments of finite length are processed as vectors. Since these piled-up pulses are not rejected, the over all count rate of the experiment is increased.;In the case of position resolving gas detectors the pile-up problem can be greatly amplified by the traditional method of sampling all cathode signals when a trigger is generated by the anode signal. The total count rate is therefore limited by pile-up effects seen on the anode signal. Here, each cathode signal is continuously sampled and processed in parallel so that no common dead time is introduced to event detection. Each detected event is then listed in a FIFO. Since all the data held in the FIFO hold event data, the DSP is only concerned with centroid calculations and not burdened with searching for event data.