SPADnet Network Modeling, Simulation and Emulation Chockalingam Veerappan, Esteban Venialgo, Claudio Bruschini and Edoardo Charbon Abstract– A complete simulation environment, designed to aid the SPADnet network design and to perform scalability studies, is presented in this work. Two simulators, namely a coincidence network simulator and a data transfer simulator, designed to simulate two network channels, were optimized for speed and memory. Furthermore, these simulators were coupled with GATE in order to emulate a real PET system network. Finally, they were modified so as to incorporate measurements from the network prototype model being implemented within the SPADnet project to ensure realistic scenarios. I. INTRODUCTION The SPADnet project [1] aims to realize a completely digital pre-clinical PET system. In [2] we proposed a sensor network based PET concept, to address the many data acquisition challenges introduced by the use of digital SiPMs [3] [4]. In the proposed scheme, sensor nodes are communicating with each other in a high-speed data network, whereas each node is implemented in hardware as a so-called photonic module. Photonic modules comprise a scintillator, a sensor tile, and a data processing and communication unit (DPCU). See Fig. 1. Fig. 1. (a) Completely digital SPADnet network with photonic modules represented as circles, (b) SPADnet photonic module construction The DPCU estimates in real time the gamma event’s energy, timing, and scintillation coordinates, thus discarding inappropriate events. Event filtering within the DPCU leads to This work is supported in part by the European Community within the FP7 SPADnet project. Chockalingam Veerappan is with Delft University of Technology, Delft, 2628 CN, The Netherlands (e-mail: C.Veerappan@tudelft.nl). Esteban Venialgo is with Delft University of Technology, Delft, 2628 CN, The Netherlands (e-mail: E.VenialgoAraujo@tudelft.nl). Claudio Bruschini is with École Polytechnique Fédérale de Lausanne (EPFL), 1015 Lausanne, Switzerland (e-mail: claudio.bruschini@epfl.ch). Edoardo Charbon is with Delft University of Technology, Delft, 2628 CN, The Netherlands (e-mail: E.Charbon@tudelft.nl). a scalable DAQ system, appropriate for multi-ring, fully reconfigurable systems. In the proposed DAQ scheme, coincidence detection [5] [6] [7] and data transfer to an external PC were carried out in a distributed way using inter-node networked communication. The network prototype based on FPGAs successfully demonstrated its functionality [8]. However, to facilitate the study of network scalability, a complete simulation environment was built to test the proposed concept for various PET types (pre-clinical, brain and human configurations). In this work we present the corresponding network models, simulations, and emulations. In [2], considering the data type versatility and the amount of data that needs to be transferred to the external PC, the network was designed to operate using two channels, where one of them was used to perform distributed coincidence detection and the other to transfer coincidence detected packets or singles to the computer. To perform a detailed study of these two channels, two simulators were built, namely the data transfer and the coincidence detection simulator. Furthermore, these simulators were coupled with GATE [9] – which generates realistic gamma photon distributions – in order to emulate a real PET system network. Finally, the simulators were modified so as to incorporate measurement results from the network hardware prototype being built using FPGAs [10], to ensure realistic scenarios in terms of the packet latency and data bandwidth variations introduced by the communication protocol. II. NETWORK SIMULATORS A. Data transfer network simulator Fig. 2. Data transfer network simulator concept Scintillator Sensor tile PCB DPCU (a) (b) P1 P2 Buffer Occupancy Clock cycle Simulation clock Gamma events Next node buffer changes to full status Previous node buffer changes from empty status