Subband channelized radar detection and bandwidth estimation for FPGA implementation

Abstract

The theory of optimum radar detection is well known and is generally implemented in expensive ASICs or supercomputers. However, today's state-of-the-art FPGAs are capable of performing relatively complex algorithms and provide the added benefit of being reconfigurable with new algorithms or methods on-site. Los Alamos National Laboratory has undertaken the goal of developing a receiver that is capable of performing detection and bandwidth estimation of pulsed radar systems. It is designed to function in electronic intelligence (ELINT) applications, where the goal is to determine the capabilities of threatening systems, such as radars which guide aircraft or missiles to targets. This thesis addresses methods of pulse detection and bandwidth estimation that are able to be implemented on an FPGA. The framework is that which is commonly used in this application: a polyphase filter bank subband frequency decomposition of the RF signal, followed by statistical detection methods. The optimal fixed-sample-size (FSS) estimator for this subband decomposition is shown to be the F-test, based on the output statistics of the filter bank, which are found to be chi-squared. An alternative to fixed-sample-size methods, the sequential probability ratio test (SPRT) is, however, more suited to ELINT due to its ability to adapt the test length to the received data. The SPRT is shown to achieve a higher probability of detection with approximately 1/5 the required sample size of the FSS method. The complexity of the SPRT is equivalent to that of the FSS method, and the statistic that results from the optimal SPRT implementation also lends itself to easy calculation of the bandwidth of the signal.