Ultra-wideband Antenna Array for Crevasse Detection Ground-Penetrating Radar

Abstract

An ultra-wideband antenna array was designed, fabricated, and tested for crevasse detection using ground-penetrating radar (GPR). Conventional GPR systems for crevasse detection typically use a single pair of transmitter and receiver antennas, which may not cover the wide vehicle widths used in Antarctic region. Therefore, a multi-antenna array configuration is necessary to cover the wide vehicle widths for effective crevasse detection. To achieve sufficient penetration depth and appropriate resolution, antennas are required to operate over ultra-wideband at a low frequency. The antennas need to have a directional beam pattern to reduce sensitivity to surrounding clutter signals and be compact for array configuration. When using multiple antennas, direct coupling between antennas can become excessively strong, potentially saturating the receiving system if the crevasse signals are amplified sufficiently. Hence, an arrangement minimizing direct coupling is required. To verify the designed antenna and arrangement, an antenna array system was fabricated and tested. The antipodal Vivaldi antenna was selected for its performance at low frequencies and ultra-wideband operation, along with its directional beam pattern and compact volume. The designed antipodal Vivaldi antenna features a radiator having exponential curves and a balun. The balun smoothly transitions the 50Ω microstrip line to a 100Ω balanced stripline, acting as an impedance transformer. The designed antenna was implemented on a 0.8 mm thick FR-4 substrate, and its performance was measured. It was confirmed that the reflection coefficient was below -10 dB and the antenna gain was above 4.75 dB within the 0.5 GHz to 2 GHz range, meeting the intended specifications. By varying the horizontal and vertical separation distances of a pair of designed antennas, S21 data as a measure of direct coupling was investigated. Simulation results showed that increasing separation distances in both horizontal and vertical directions decreased direct coupling. However, it was noted that at certain separation distances, direct coupling was reduced compared to neighboring distances. This trend was confirmed using time-domain transmission and reception waveforms employing a sinc pulse within the 0.5 GHz to 2 GHz range, identifying that a horizontal separation of 425 mm and a vertical separation of 220 mm minimized direct coupling. Experimental verification using a pair of antennas confirmed the simulation results for direct coupling. Using eight designed antennas, an array with four transmit-receive channels was configured. The arrangement between the transmit-receive antennas was set to minimize direct coupling, with optimal horizontal and vertical separation distances. A vector network analyzer was used for signal transmission and reception, employing RF switches to select signal paths for the four channels, controlled by a digital I/O board. A control computer was used to control the vector network analyzer and the digital I/O board. The computer also stores and process stored the received signals. The reduction in direct coupling among arrayed antenna elements was confirmed by orienting the arrayed antennas towards the sky and measuring the transmitted and received waveforms, showing a 11.1% to 33.8% reduction in direct coupling compared to horizontal placement. To validate the effectiveness of the array radar, two sets of experiments were conducted. First, the array antenna system was moved over stationary ground targets to confirm the distance data acquisition functionality. Second, a ground scanning experiment was performed to verify the coverage performance across the vehicle width, checking data from each port individually. Frequency domain signals for each channel were converted to time domain to generate A-scan data, and sequentially acquired A-scan data was visualized in a B-scan format. During this process, through calibration was employed to remove phase shifts or time delays occurring in signal paths by directly connecting the cables without antennas. Fixed clutter signals such as antenna direct coupling, equipment, and ground reflections were removed using an average value removal technique on B-scan data. The ground scanning experiment conducted over an area with a manhole demonstrated that channels 2 and 3 detected the manhole, while channels 1 and 4 did not. This illustrated that the array radar covered scan areas that would not have been covered by a single-channel radar.