Design of stepped frequency continuous wave radars using continuous coherent calibration

Abstract

This thesis details the design of a stepped frequency continuous wave (SFCW) radar known as the Ag Radar (AGR) 120 that operates in the 120 GHz ISM band for agricultural applications. The AGR 120 hardware is centered around a transceiver that integrates transmit and receive antennas, RF mixers, RF amplifiers, and a VCO into a single IC. Due to this high level of integration, significant crosstalk is present in the received intermediate frequency (IF) signals. Methods for removing this crosstalk already exist, such as injecting a correction signal to cancel out the crosstalk prior to sampling the ADC. In this thesis, the idea of injecting a correction signal is further explored, with the correction signal being continuously derived from the received signal while the radar is pointed at a target. To do this, it is shown that the crosstalk appears as a low frequency signal in the received IF signals which can then be extracted using a low pass finite impulse response (FIR) filter. This creates a feedback loop where the received signal is sampled, a correction signal is derived using the filter, and the correction signal is applied to the received signal of the next sweep of the radar. By introducing this feedback loop, it is shown that the received signal suffers gradual degradation due to noise accrual in the correction signal. As a solution, the correction signal is periodically reset using linear regression with a polynomial model to remove the noise. By using a combination of continuously filtering and periodically resetting the correction signal, the crosstalk is successfully removed while simultaneously receiving target returns. In addition to describing the continuous coherent calibration technique, this thesis also documents the general design of the AGR 120 including a CAN communication interface that is used for reporting target data and configuring parameters of the radar such as bandwidth and a method of calculating target distance from the FFT of the received signal using noise thresholding and weighted averaging.