Maximum likelihood estimation techniques for high rate, high throughput digital pulse processing

Many applications of radiation detection and measurement are restricted by the limited capabilities of pulse processing electronics. In particular, distortions due to pulse pile-up: reduced throughput; dead-time; and degradation of energy resolution can significantly affect performance. Analogue techniques, designed to shorten pulse duration, can reduce dead-time however the short shaping times reduce noise filtration and result in poor resolution. Digital pulse processing technologies rely on linear filtering methods, which attempt to reduce the pulse length to improve resolution. Regrettably, much of the pulse energy is at low frequencies while high-pass filters are required to reduce pulse length. Consequently, this approach significantly reduces signal energy resulting in a loss of signal to noise ratio (SNR) and ultimately degraded energy resolution. An alternate approach to resolving pulse pile-up distortion is presented in this paper. Based on the Maximum Likelihood Estimation of all events within the digitised detector output, this Pulse Pile-up Recovery (PPR) algorithm decodes information contained within piled up events. By accurately characterising individual radiation events, even in the presence of severe multi-pulse pile-up, dramatic improvements in throughput and energy resolution can be achieved. Conventional pulse processing techniques trade off energy resolution and throughput by altering pulse shaping times. The pulse pile-up recovery algorithm trades off count-rate performance (in terms of throughput, dead-time and energy resolution) against analogue-to-digital conversion (ADC) rate, ADC bit-depth and requirements for computational processing. Simulation results for a number of operating settings are presented. These are compared with the performance of a realtime version of the algorithm implemented in a Xilinx Virtex-4 field programmable gate array (FPGA). © 2008 IEEE.