Direction finding on spread-spectrum signals using the time-domain filtered cross spectral density

In an earlier paper, it was shown that spread-spectrum signals of arbitrary unknown waveform could be detected at received powers well below the receiver noise level, by crosscorrelating the signals from two spatially separated antennas, windowing the crosscorrelation function (CCF) over a narrow range about zero time shift and taking its Fourier transform, thus forming the time-domain filtered cross spectral density (TDFCSD). The authors show that, having detected the signal and established its bandwidth, it is possible to estimate its angle of arrival (AOA) by measuring the phase slope in the TDFCSD across the signal bandwidth. An analysis of the technique is presented showing how the direction finding (DF) accuracy varies with signal and receiver parameters. Computer simulation results show that, with an antenna spacing as low as 1 m, a DF error of less than 1° can be obtained on a signal with a bandwidth of 100 MHz at a received power level lower than -100 dBm, using an integration time of a few milliseconds. A practical system based upon an acousto-optic correlator capable of computing the CCF in real time is described. Preliminary results from the system show DF errors that are in good agreement with theory