Spectrum Awareness Testbed

Abstract

This thesis presents a flexible and scalable spectrum awareness testbed targeting a wideband frequency range. Due to the static frequency allocation scheme, spectrum scarcity has become a problem in communications. Opportunistic spectrum access will allow secondary users to take advantage of empty portions of the spectrum to increase the efficiency of spectrum use. To allow that to happen, opportunistic users must be able to identify, characterize, and geolocate nearby transmitters.

The spectrum awareness testbed is capable of recovering the approximate carrier frequencies of an input transmission. It operates in the 5G Frequency Range 1 (5G FR1) and is currently configured to sense one transmission. The testbed uses the Modulated Wideband Converter (MWC) as a sub-Nyquist sampling scheme to acquire the input signal in hardware. The input is split into multiple channels. Each channel is then mixed with a periodic waveform, lowpass filtered, and sampled at a low rate for digital processing. The periodic waveform defines a relationship between the low-rate samples in each channel and the support of the input signal, which is recovered through a compressed sensing (CS) technique.

To verify operation of the testbed, the MWC system was simulated in Matlab. For the parameters selected for the hardware implementation, the simulation achieved a successful support recovery rate greater than 90% for SNR values larger than 5 dB. The MWC system was constructed in hardware and tested using a Hardware-in-the-Loop (HWIL) setup. Multiple carrier frequencies and signal bandwidths of 10 MHz and 80 MHz were evaluated. The largest successful percentage of support recovery for a signal with an 80 MHz was 43.28% for the carrier frequency 2.0 GHz. For a signal with 10 MHz bandwidth, the largest successful percentage of support recovery was 36.86% for the carrier frequency 2.0 GHz.

The performance of the testbed MWC hardware implementation did not meet the performance seen in simulation. This is likely attributed to low input signal power levels, the frequency range of the chosen mixer, and analog component inaccuracies. Possible solutions are suggested to improve performance. Overall, the hardware implementation functions as a proof of concept for a wideband spectrum awareness testbed.