Year
2007
Abstract
The multiple isotope material basis set (MIMBS) isotope identification algorithm has been implemented on the Analog Devices Blackfin BF561 dual digital signal processor (DSP) platform, using the Levenberg-Marquardt (LM) solver for the nonlinear part of the MIMBS problem. This software, currently running on a prototyping/development board, will be used in NaI and LaBr3 handheld isotope identifiers in development by our research group. We examined the number of iterations required by the LM solver with different convergence criteria, based on a set of 272 test spectra. An average of 7.5 LM iterations were required for convergence, although excellent identification rates were obtained using just 3 or 4 iterations with no convergence tests. Benchmark tests of execution times on the BF561 evaluation board show that accurate MIMBS identifications based on a 30-isotope library can be obtained in 0.5 to 1 s of analysis time, on average.