A software-based receiver sampling frequency calibration technique and its application in GPS signal quality monitoring

Abstract

Due to manufacturing and environmental reasons, a GPS receiver's actual and manufacturer specified sampling frequency may differ significantly. Accurate knowledge of the sampling frequency is fundamental for high sensitivity and high accuracy receiver signal processing.Additionally, applications such as pre-correlation signal quality monitoring employing periodic averaging and dithered sampling techniques to enhance the signal-tonoise ratio (SNR) and sampling resolution depend on the accuracy of the sampling frequency. A refined mathematical model of software-correlator based receiver processing in the presence of clock error is established in the paper, and a novel inline method is presented to estimate the accurate sampling frequency. The method is a solely software-based technique requiring no additional hardware other than the GPS receiver RF front end output samples. Neither a priori knowledge of the specific frequency plan of the RF front end circuit nor complex receiver output data fitting are needed. The impact of the sampling frequency error on the performance of receiver signal processing and pre-correlation periodic averaging, and the performance of the frequency calibration method are evaluated using simulated signals as well as live GPS signals collected by GPS data acquisition equipments manufactured by different vendors. Several experimental measurements are presented to support the evaluation including the receiver observables and navigation solutions, as well as the pre-correlation time domain waveforms and eye patterns, Power Spectrum Density (PSD) envelopes, amplitude probability density histograms, correlation function and S-curve bias after periodic averaging. Our simulation results show that the method can calibrate the sampling frequency with an accuracy resolution down to 10 -9 of the true sampling frequency online, and the pre-correlation SNR can be potentially improved by 39dB using periodic averaging.