On applications of linear prediction filtering to small wavelength Doppler weather radar signal processing

This paper examines the problem of reduced signal-to-noise ratios in small wavelength (2 cm) Doppler weather radars. Using adaptive linear prediction filters, two solutions to the noise problem are presented. Through simulation techniques, an adaptive mean frequency estimator, based on Griffiths´ method of instantaneous frequency measurement, is tested and compared to the pulse pair estimator. It is shown that when both the adaptive and pulse pair estimators are used on narrowband signals (sampled at the pulse repetition frequency) the pulse pair method is superior. However, when the adaptive estimator is tested on signals sampled at the radar´s intermediate frequency, the adaptive method has a smaller variance. Finally, the use of an adaptive noise canceller is shown to significantly increase signal-to-noise ratios and improve the pulse pair estimator´s quality when the adaptive noise canceller is used as a preprocessor before sampling at PRE rates. In both of the adaptive filtering cases, it is noted that the non-stationarities associated with airborne weather radar implementations do not affect filter performance since adaptive filtering techniques have the ability to track changes in the input signal´s statistics.