Spectrum monitoring in cognitive radio networks using error vector magnitude.
Type of content
The proliferation of new operators, innovative services and wireless technologies has caused radio spectrum resources to become scarce. This scarcity has led to the concept of cognitive radio (CR) communication. Spectrum sensing is a critical component of CR networks for spectrum utilization. It is the process in which the CR users or the secondary users (SUs) exploit unused licensed or unlicensed frequency bands providing minimal interference to the primary users (PUs). Spectrum sensing can be categorized into two types : out-of-band sensing and in-band sensing. Out-of-band sensing searches for an idle channel by sensing multiple channels sequentially until an available channel is found. In-band sensing monitors a channel periodically while using it, in order to detect the return of PUs, so that SUs can vacate the channel immediately upon detection of returning PUs. This periodic spectrum sensing causes SUs to halt its communication frequently. This causes a tradeoff between maintaining high communications efficiency in the secondary network and avoiding disruption to the primary network.
In this thesis, we use error vector magnitude (EVM) to develop a novel in-band sensing technique that allows SUs to perform spectrum monitoring during transmission. EVM is being increasingly employed in the wireless industry and has already become part of several wireless standards. However, this metric has been overlooked in the context of CR despite having significant cant advantages. One advantage is that the SU transmitted signal gets canceled out during EVM calculation, therefore we do not require a sophisticated self-interference cancellation (SIC) technique as required by full duplex (FD) techniques. The other advantage is that the EVM technique utilizes fewer symbols, does not require any subcarriers to be reserved, provides results much before demodulation and decoding thus giving us real time results. This monitoring method differs from a spectrum sensing method in that the monitoring is applied during reception of packets. It involves detecting the emergence of PUs during periods in which the SUs are communicating. This method doesn't halt the SU communication, hence a signi cant gain in SU throughput is obtained. This spectrum monitoring technique supplements the traditional spectrum sensing and provides enhanced communications efficiency.
In the fi rst part of this thesis, we study the performance of the EVM based PU monitoring technique in orthogonal frequency division multiplexing (OFDM) based CR networks. We utilize the pilot tones that are inherent to many OFDM based standards to measure the EVM as the difference between the received and transmitted pilot tones. We show that a step change in the EVM curve is sufficient to detect a PU during ongoing SU transmission. The technique also allows us to locate the bits corrupted by the PU's arrival by looking at the EVM values of the pilot tones. In the second part of this thesis we analytically characterize the performance of the proposed method. We derive the probability density function (PDF) of the EVM based statistic. We then analyze the detection performance with the complementary receiver operating characteristics (CROC) curve in terms of type I and type II error probabilities. In order to simplify the analytical expression, a Laplacian approximation method is also provided. We also present the joint PDF of the test statistic towards the detection of the reappearing PU. A throughput performance of the proposed detector is also analyzed. Simulation results illustrate that the EVM based detection performs better than the energy detection (ED) method.
In a later part of the thesis, we show the application of the sequential change detection also known as the quickest detection method to the EVM based detector. Motivated by the results from CROC analysis and considering the importance of detection delay in spectrum sensing, we develop an exact quickest detection scheme by using a traditional cumulative sum (CUSUM) test. Simulation results show that the quickest EVM detection significantly outperforms the quickest ED.