The underground channel is complex, with path loss heavily dependent on frequency and soil properties (moisture, texture, density). Research suggests the 300 MHz - 400 MHz band might offer a balance between path loss and antenna size. However, simulations and some empirical studies indicate potential for communication up to 1 GHz. Studies have shown a limited range (0.5 m) at 2.4 GHz using terrestrial nodes. The soil moisture content significantly affects antenna performance, potentially shifting the operating bandwidth and increasing return loss, especially at higher frequencies. The Friis equation needs modification for underground environments, incorporating additional path loss terms related to the soil’s dielectric properties and attenuation constant. Models like Peplinski’s [9] are used to estimate soil dielectric properties based on frequency, water content, and soil composition. Path loss generally increases with both distance and frequency.

Antenna selection for WUSNs is challenging due to size constraints, variable soil conditions, and the need for directionality. Lower frequencies necessitate physically larger antennas. Directionality is also a concern; omnidirectional antennas may have nulls that limit vertical communication, while directional antennas are larger and require precise alignment. The burial itself alters antenna characteristics. Research explores the optimization of antenna design considering the impact of soil [10] .

Power consumption is critical for battery-operated buried nodes. While some systems use mobile charging, many applications require low-power operation. Amplitude-based modulation schemes (like ASK) are generally unsuitable due to signal attenuation underground. Frequency-based schemes (FSK) can be affected by noise and channel variations. Phase-based modulation (PSK), particularly QPSK, is considered more robust as phase is less affected by the medium compared to amplitude or frequency. QPSK offers better spectral efficiency than BPSK by transmitting two bits per symbol.

This work builds upon these findings by implementing and evaluating a flexible SDR-based system using QPSK in the 450/900 MHz range, with a focus on optimizing communication for pipeline monitoring while considering practical antenna constraints.

The subsurface unit is designed to be buried at depths ranging from 1 to 5 meters. Its block diagram is shown in Figure 1 .

1) SENSORS: Analog sensors relevant to water leakage detection, including temperature, humidity, and soil moisture sensors, are utilized. Sensor outputs are typically voltage signals (0 V – 5 V).

2) Analog-to-Digital Converter (ADC): An MCP3008 10-bit ADC converts the analog sensor readings into a digital format [11] . It interfaces with the processing unit via the Serial Peripheral Interface (SPI) protocol.

3) Processing Unit: A Raspberry Pi 3 Model B serves as the central processing unit. It runs a Python script to acquire data from the ADC, log timestamps, and store the data in a .csv file format.

4) SDR and Transmission: GNU Radio software, running on the Raspberry Pi, processes the .csv data. It encodes the data into packets and performs QPSK modulation. The modulated digital signal is sent via USB to a HackRF One SDR transceiver. The HackRF One converts the digital signal to RF and transmits it through an antenna.

5) Antenna: Due to size constraints, low-gain isotropic antennas are used for transmission.

6) Power: The entire subsurface unit is powered by a portable battery pack (e.g., 20,000 mAh). Power consumption is optimized by utilizing the low-power Raspberry Pi and selecting efficient modulation and antenna configurations.

The surface unit is positioned above ground, ideally directly above the buried unit. Its block diagram is shown in Figure 2 .

1) Antenna: A high-gain directional antenna (Yagi-Uda or Parabolic) is used to capture the potentially weak signal transmitted through the soil.

2) SDR Receiver: The received RF signal is fed into another HackRF One SDR, which down-converts and digitizes the signal.

3) Processing and Display: The digitized signal is sent via USB to a computer running GNU Radio. GNU Radio demodulates the QPSK signal [12] , decodes the packets, and reconstructs the original sensor data into a .csv file. A simple Python-based local web server can then display the received sensor data on an HTML page for real-time monitoring.

GNU Radio, an open-source software toolkit, provides the signal processing framework. It utilizes pre-built or custom processing blocks (coded in C++) that are linked together using Python flowgraphs. For transmission, the flowgraph includes blocks for reading the .csv file, packet encoding, QPSK modulation, and interfacing with the HackRF One (using an OSMOCOM sink). For reception, the flowgraph uses an OSMOCOM source to get data from the HackRF One, followed by QPSK demodulation, packet decoding, and writing to a file sink. The HackRF One serves as a flexible RF front-end, handling both RF transmission and reception, as well as analog-to-digital and digital-to-analog conversion.

The subsurface subsystem was placed inside a container filled with a mixture simulating typical underground conditions: 50% soil, 30% sand, 10% clay, 5% stones, and 5% garden soil, as shown in Figure 4 . For testing variable depths, the container holding the transmitter setup was buried at depths of 0.5m, 1.0m, and 1.5m below the surface.

1) Subsurface: Raspberry Pi 3, MCP3008 ADC, Temperature/Humidity/Moisture Sensors, HackRF One, Battery Pack.

2) Surface: Laptop/Computer, HackRF One, High-Gain Antenna.

Based on frequency and role, different antennas were used, balancing performance and size constraints.

1) 450 MHz:

Transmitter (Subsurface): Dipole antenna (2.15 dBi gain, 20.5 cm length).

Receiver (Surface): Yagi-Uda directional antenna (11.2 dB gain, eight elements, 120 cm length).

2) 900 MHz:

Transmitter (Subsurface): Ceiling-mount omnidirectional antenna (2.5 dBi gain, approx. 17 cm diameter).

Receiver (Surface): Grid parabolic directional antenna (15 dB gain, 60 cm × 90 cm size).

Key specifications are summarized in Table 1 . Directional antennas were used at the surface receiver to maximize signal capture from the known transmitter location below. Low-gain, smaller antennas were used for the buried transmitter.

Frequencies: 450 MHz and 900 MHz.

Modulation: Primarily QPSK, with BPSK and 8-PSK tested for comparison. QPSK was chosen for its balance of robustness and spectral efficiency [13] .

SDR Settings: GNU Radio and HackRF One were used. The sample rate was set to 2 Msps. Bandwidth was fixed at 800 kHz for transmission tests. Receiver gains (RF, IF, BB) were adjusted for optimal reception. The transmitter RF gain was kept low (0 dBi - 2.15 dBi) to simulate power constraints.

Data Protocol: Sensor data was formatted into .csv files, then packetized, modulated, transmitted, received, demodulated, and decoded back into .csv files.

The system’s performance was evaluated based on:

1) Signal characteristics: FFT plots, constellation diagrams, waterfall diagrams.

2) Signal-to-Noise Ratio (SNR): Measured using GNU Radio probes.

3) Data Rate: Calculated by transmitting a known file for a fixed duration (10 seconds) and measuring the amount of data successfully received.

Experiments were conducted with varying burial depths (0.5 m, 1 m, 1.5 m), operating frequencies (450 MHz, 900 MHz), and modulation schemes (BPSK, QPSK, 8-PSK). Each test was repeated multiple times, and average values were reported.

1) Spectrum: FFT plots confirmed transmission centered around the target frequencies (450 MHz/900 MHz) with the expected bandwidth (8pprox.. 800 kHz). The received signal power was significantly lower than the transmitted power (e.g., −45 dB received vs. −25 dB transmitted at 450 MHz), indicating substantial path loss through the medium. Waterfall diagrams visually confirmed the presence of the signal and its bandwidth over time at both the transmitter and receiver ends, as shown in Figure 5 and Figure 6 [14] .

2) Constellation: Constellation plots were used to assess the quality of the QPSK modulation. At the transmitter, clear constellation points were observed, as shown in Figure 7 . The received constellation showed more dispersion due to channel noise and distortion, but remained distinct enough for successful demodulation at the depths tested, as shown in Figure 8 .

3) Time Domain: Signal scope plots of the modulated waveform at the transmitter and the attenuated/noisy version at the receiver are shown in Figure 9 and Figure 10 .

Table 2 summarizes the key performance results for different configurations.

1) Effect of Depth: As expected, increasing the burial depth consistently decreased the average SNR and, consequently, the achievable data rate for all configurations. This aligns with channel models showing increased path loss with distance.

2) Effect of Frequency: Comparing 450 MHz and 900 MHz at the same depth and modulation (QPSK, 1.5 m depth), 900 MHz surprisingly yielded a higher data rate (1.953 Mbps) than 450 MHz (1.253 Mbps) despite similar SNR values (around 24.2 dB). While higher frequencies generally experience more path loss, the specific antenna gains (15 dBi parabolic at 900 MHz vs 11.2 dBi Yagi at 450 MHz for the receiver) and potentially different multipath/channel characteristics at these frequencies in the test environment might explain this outcome. At shallower depths (0.5 m), 900 MHz also provided higher data rates.

3) Effect of Modulation: Increasing modulation complexity (BPSK → QPSK → 8-PSK) generally increased the raw data rate, as more bits are encoded per symbol. For instance, at 900 MHz and a depth of 1.5 m, data rates were 1.245 Mbps (BPSK), 1.953 Mbps (QPSK), and 2.895 Mbps (8-PSK). However, higher-order modulations, such as 8-PSK, are typically more susceptible to noise (lower SNR tolerance), potentially leading to higher bit error rates, although error rates were not explicitly measured in this case. QPSK offered a good compromise, providing double the data rate of BPSK with reasonable robustness, as indicated by the stable SNR values of around 24 dB - 28 dB across depths.

The system successfully logged sensor data, transmitted it using the SDR pipeline, received and demodulated the signal, and displayed the results. The end-to-end data flow demonstrated the feasibility of the WUSN DAQ system for the target application. The achieved data rates (e.g., greater than 1 Mbps with QPSK) are sufficient for transmitting typical sensor readings in pipeline monitoring.