Auto-regressive estimation of spectral width of chirp sounder signals

Abstract

A major limitation of HF chirp sounders is the use of fast Fourier transform (FFT) analysis to resolve the different multipath components. To reduce the effects of spectral leakage associated with the FFT, the signal is usually weighted by a suitable window prior to spectral analysis. The choice of window function and its effects on the measured width of the dispersed chirp pulse has been previously studied. It was found that the window with the narrowest main lobe gave the best resolution in the least dispersive part of the ionograms whereas the window with the highest attenuation of the edges and hence the widest main lobe had the best resolution in the most dispersive parts. Since the mainlobe of the window is the narrowest width that can be measured non-dispersed pulses will be masked by it. Hence, a spectral estimation technique which is capable of avoiding the limitations of FFT analysis is desirable. Modern spectral estimation techniques such as autoregressive (AR) modelling extend the signal outside the observation interval rather than windowing it to zero as in the FFT. The implied extrapolation results in a higher resolution than is obtained from the FFT which is limited to the inverse of the observation interval. The enhancement of resolution with AR analysis depends on the signal to noise ratio (SNR). The HF spectrum is highly congested. Hence, techniques such as filtering and interference reduction to improve the SNR prior to AR analysis would be required. This paper discusses the analysis stages for both FFT analysis and AR analysis. The results of applying both techniques to ionospheric data obtained over a short skywave link in the UK are presented.