Implementation of a parallel virtual machine in the data acquisition system of a high resolution animal PET

Positron emission tomography is based on the coincident detection of two colinear emitted gamma rays with subsequent registration by a data acquisition (DAQ) system. To preserve the systems sensitivity and provide accurate tomographic imaging, a large fraction of the detected events has to be registered. Therefore, a high and constant readout capability of the DAQ architecture is desired. We have realized a low cost DAQ system for our small animal sector tomograph MADPET (2.5 mm resolution, coincidence event registration, <10 kHz readout rate). Two standard computers (AMD-K6, Linux, 10 MBit ethernet) form a parallel virtual machine (PVM-master and PVM-slave). This allows for the distribution of tasks like graphical interface, online data processing and display to the PVM-master while preserving the resources of the PVM-slave for data acquisition. A full ring dual layer 3D tomograph is currently under construction (≲1.5 mm resolution, single event registration, <5 MHz=20 Mbyte/s singles readout rate). The DAQ system of this device will be based on SUN workstations. Reflective memory modules in each (connected with fibre optical cable) facilitate automatic memory update between both machines within 700 ns. According to our experiences with MADPET, PVM will be implemented in this system