[Previous] | [Session 16] | [Next]
A. Zoglauer (MPE/UCB), G. Weidenspointner (CESR), C. Wunderer, S. Boggs (UCB), R.M. Kippen (LANL), ACT Collaboration
Advanced Compton Telescopes (ACT) intend to measure more parameters of Compton events (e.g. several Compton interactions, the recoil electron direction, etc.) with a higher accuracy than their predecessor COMPTEL. But all ACT designs have the same fundamental problems in common: Due to their compact design the time between the interactions and their sequence can not easily be measured, and space background will dominate the source signal. Thus, one crucial step in the data-analysis is to order the interactions and to identify most of the background. We compare two approaches to this fundamental problem of event classification: The classical approach is based on test statistics and figures-of-merit, which describe the interaction kinematics and probabilities. It has moderate computational demands and can be easily transfered to different ACT designs but lacks the ability of determining an absolute value for the quality of the event. The second option considered is a Bayesian approach, which has the ability to determine an absolute event quality - at the expense of very high computational demands for simulating the underlying multi-dimensional probability density functions, which in addition are only valid for one distinct ACT design. We apply both the classical and the Bayesian event reconstruction techniques to different event classes (source photons, beta decays in the detector, etc.) in a Si/Ge ACT design and discuss the algorithms' capability for event reconstruction and background rejection.
[Previous] | [Session 16] | [Next]
Bulletin of the American Astronomical Society, 36 #3
© 2004. The American Astronomical Soceity.