This week, I spent a few more days trying to understand the errors related with strong radar targets in the analysis of weak targets in passive radar. As a consequence of this, I wrote a linear least squares estimator to remove this contribution more effectively (and consistently). I also optimized my code a bit, and got a massive speed-up. Now the signal processing and plotting all runs in real time.
This video shows a lot of aeroplanes around the New England area detected using a simple passive radar setup, consisting of: one USRP and two yagi antennas, a quad core Linux PC. Every now and then an occasional specular meteor echo is observed too. Because this is FM radio, the waveform is not always optimal for inverting the echoes, which results in a blow up of the solutions when the waveform is narrow band (eg., somebody talking, or silence). This is not a problem with digital modulation schemes, which will be explored next. I updated my code. I came up with a way to efficiently estimate ground clutter and transmitter self-noise using a linear least squares matrix equation, which works better than the previous attempt. I also made some optimizations that allow me to run this in real-time using a normal PC. Next up: imaging using the midas-mini cube.