Direct Signal Interference Suppression and Target Detection for Low-Cost SDR-Based Passive Bistatic Radar

Abstract: Passive radar is a technology for detection of targets using echoes of existing radio transmitter, such as FM-radio. Since only receivers are needed for operation, a passive radar system has the possibility of being implemented using low-cost hardware. Using lower cost implementations to cover blind-spots of other, more sophisticated systems, could be a viable solution for full radar coverage. To achieve this, an understanding of the effects such low-cost systems have on the performance of a radar is needed.  A prominent problem for passive radar is that the interference caused by the direct signal from the transmitter used and reflections from uninteresting terrain, called clutter, can drown out the echoes from targets. This thesis compares the direct signal interference (DSI) suppression algorithms: ECA, ECA-S, ECA-B, NLMS and FBNLMS when run on data from a low-cost receiver called KerberosSDR. It is found that the low ADC resolution of 8 bits is a limiting factor for KerberosSDR. Random noise in the receiver can also limit the performance. None of the tested algorithms are any more or less affected by the ADC resolution or the noise. The first difference appears when comparing the execution times, where FBNLMS is 10–20 times faster than the other algorithms. However, the slower rate of convergence for FBNLMS and NLMS causes them to lose performance in environments where the DSI and clutter are considerably stronger than the target echoes. The algorithms FBNLMS and NLMS also lose performance due to their inability to model frequency shifted echoes as unwanted. The main disadvantages of ECA, ECA-B and ECA-S are their long execution time. It is concluded that FBNLMS would be the best candidate in most cases for low-cost hardware, as it allows execution on slower hardware and the main disadvantages not being too prominent in the use case of covering blind-spots of other systems.