Sub-nyquist sampling techniques for cognitive radio applications

Cognitive Radio (CR) has emerged as the promising solution to overcome the limited spectral resources available to support the incessant demand for higher data throughput in today’s wireless communications. CR operation exploits the underutilized spectral resources characteristics of typical radio channels by transmitting the data when a chan- nel is found to be idle. In order to increase the probability of finding unused spectrum and therefore increase its transmission throughput, a CR will try to monitor as many channels as possible by performing spectrum sensing (SS) over a wide frequency range. However, due to the Nyquist sampling requirement, monitoring a wideband spectrum, which is termed as wideband spectrum sensing, would require very high sampling rate and is limited by existing analog-to-digital converters (ADCs) technologies. This thesis presents the study of using a sub-Nyquist sampling technique to extend the capability of CR system, primarily through the adoption of Compressive Sampling (CS) technique in CR implementation. Compressive sampling is a signal-processing framework which enables a system to reconstruct a sparse signal that is sampled at a sub-Nyquist rate. In practice, the CS reduces the required sampling rate with the tradeoff in higher data reconstruction time cost. It hence typically limits the CS technique to off-line data processing applications, which is not possible for CR systems that require the SS process to be performed in real time. This thesis hence presents several novel approaches to overcome the existing CS limitations with the aim to minimize the time required for CS based SS operations and further enhance the performance of the CR system. The first contribution presented in this thesis is a new CS based SS technique pro- posed for hybrid CR that uses the combination of underlay and interweave transmission modes. Unlike existing CS based signal processing operation, the proposed CS based technique does not require the reconstruction process. As such, it is able to achieve much lower sampling rate and greatly reduce the detection processing time compared to other known SS techniques. In addition, the proposed approach also incorporates the learned feature which can further improve the accuracy of the SS process. As a result, the proposed technique is able to achieve higher transmission throughput compared to other well known SS techniques, while operating at sub-Nyquist sampling rate without the need to use complicate ADC hardware architecture. The second contribution presented in the thesis is a novel matrix optimization algorithm that can be incorporated into CS based CR receiver to enhance the detection and reconstruction accuracy for OFDM-based signal transmission. This is important as it is usually not feasible to implement optimal sensing matrix for CS based CR since its frontend receiver circuit is typically hardwired, and the need to remain compatible with standard digital OFDM receiver’s operation. Simulation results show that the proposed approach can consistently produce smaller CS reconstruction error in term of BER under various operating conditions when comparing to existing published systems. The third contribution of this thesis is to further extend the use of the matrix optimization algorithm to MIMO-OFDM based system, which is the dominant air interface for the latest 4G and 5G broadband wireless communications. The proposed technique enables the enhancement of CS related data transmission performance with reduced number of ADCs required at the MIMO-based receiver. Extensive simulation results have strongly confirmed the promising performance of this proposed approach. In summary, this thesis proposed several new ideas on how the sub-Nyquist CS based technique can be adopted for CR systems that require real-time operations, without compromising the performance while at the same time reduces the complexity of the hardware circuitry required in the CR implementation.Doctor of Philosophy (SCE