International Journal of Computer Applications (0975 – 8887) Volume 41– No.12, March 2012 42 Implementation of Transient Signal Detection Algorithms on FPGA Hamid GholamHosseini School of Engineering Auckland University of Technology Private Bag 92006 Auckland 1142, New Zealand Kang Li School of Engineering Auckland University of Technology Private Bag 92006 Auckland 1142, New Zealand ABSTRACT Radio transient signals are non-periodic and discrete obtained from high energy physical processes in space. Most challenging issues in transient signal detection are the speed and accuracy with which a signal can be detected. Cumulative Sum (CUSUM) algorithm has been employed in this paper for transient signal detection and proved to be capable of meeting the necessary requirements. However, as ordinary software- based programs are unable to handle large scale sample of signals, the current research focuses on implementing the CUSUM algorithm on Field Programmable Gate Array (FPGA) which is a specific integrated circuit within the field of semi-customized circuits that can greatly enhance the speed of detection and analysis. Therefore a FPGA-based system was devoted to implement the algorithms for an efficient transient signal detection. A detection speed of 64 ns per sample set was achieved via implementation the algorithm on an Altera Cyclone IV device with a clock speed of 50 MHZ. The analyzed result shows the power consumption of the FPGA based system can be reduced to 136.75 mW. Keywords Transient Signal Processing, CUSUM Algorithm, FPGA Implementation 1. INTRODUCTION Radio transient signals generate from high energy physical processes in space, such as solar flares, supernovae, pulsars, quasars and active galaxies. Other speculations include evaporating black holes, colliding neutron stars and a number of unknown events. The detection of radio transient is a challenge due to their short and non-periodic nature, as well as the high risk of misdetection. It requires the backend of radio telescopes to be equipped with the appropriate hardware and software. Generally, a de-dispersion procedure is used to improve detectability and test the property of the signal. However, due to the large scale of the signal from outer space, the computational demands of this method appear insufficiently robust. Therefore, a new improved algorithm and method for detecting transient signals is developed. Traditionallymost of the developed solutions have been based on software programs targeted for general purpose processors. The shortfalls of these platforms are obvious. For example, these platforms are usually constrained by a fixed number of processors, a limited operating speed and a fixed bandwidth, and are characterized by high power consumption. Most importantly, not all the resources on such platforms are used for transient signal operation, with some parts of the resources consumed by the operating system and software setup.The platform would be used to optimize only transient signal operation, thereby saving on the cost of other unnecessary components. In addition, the power consumption of the customized platform would be much lower than a general purpose computer.The most common technologies available to achieve this result are theApplication Specific Integrated Circuit (ASIC), and the Field Program Gate Array (FPGA).ASIC is an integrated circuit customized to perform a certain task.The advantages of ASIC include low power consumption, high operating frequency and high logical density. The disadvantages are that ASIC systems come with a high design cost and require specialized designer knowledge. FPGA is a specific integrated circuit within the field of semi- customized circuits that solves the lack of customized circuits and overcomes the existing limitation of gate numbers. The circuit is designed in hardware description language (Verilog or VDHL) allowing easy layout and burn to the FPGA chip. Another advantage of using FPGA is the capability of working as a co-processor for High Performance Computers (HPC). It seems easier and more reasonable to implement algorithms via FPGAs, with the design on FPGAs be able to migrate to HPCs or ASICs in most cases. 2. DETECTION ALGORITHM Transient signal detection can be considered as a complex stochastic model. Any abnormal signal can affect changes of the model. The aim of detection is to monitor the difference between input signals with the threshold. Currently, there are two ways to monitor those changes; one is from the perspective of signal processing, the other is from a statistical point of view. Signal processing methods usually transform the sampling signal to a time domain or frequency domain and observe the changes. In [1], Cornel Loana provides an adaptive time- frequency method based on the over-complete wavelet transform concepts, which lead to signal processing on interest frequency bands. This method is based on the fourth order moment, and is applied for each sub-band, in order to establish the optimal weight for each sample. The result obtained proves the capability of the proposed approach to accurately detect a transient signal, when compared with other methods (e.g. Spectrogram or Standard Wavelet Transform) [2].The author [1], discovered that the commonly used method, discrete wavelet transform (DWT) was not well suited to this kind of signal processing problem. From a mathematical point of view, the DWT is generated by the sampling in time-scale plant of a corresponding continuous wavelet transform (CWT). Despite the fact that there is an