Improving Cross-Correlation Spectrum Sensing Using Two Antennas

Spectrum sensing is used in Cognitive Radio (CR) to be able to detect which frequency can be used for communication. Spectrum sensing is done with Spectrum Analyzers (SAs). An existing one-antenna cross-correlation SA (XCSA) uses cross-correlation to achieve a very low noise figure (NF) to be able to sense the spectrum more accurately. Uncorrelated noise added in two signal paths is averaged out using spectral averaging. The NF resulting from the correlated noise added by the XCSA is called the correlated NF. This thesis analyzes the improvements which can be made by sensing the spectrum with a two-antenna XCSA. To achieve that goal, first the correlated NF of the existing one-antenna XCSA is analyzed. This one-antenna XCSA uses a resistive splitter and a resistive matching network to split the signal into the two signal paths and match the impedance of two mixers, which form the input-stage of each signal path, to the impedance of the antenna. It is found that the resistive splitter and matching network account for 4.3dB of the correlated NF of the one-antenna XCSA, independent of the attenuation of the matching network. Furthermore, noise folding accounts for 1.6dB of correlated NF. The total amount of correlated NF, 5.9dB, is higher than the measured correlated NF of 2-4dB in previous work. From literature, a theory is derived to be able to calculate the correlation of the noise in two antennas. The correlation factor of the noise is dependent on the timedelay of the noise between the two antennas, which results in a frequency and a distance dependency. With this theory the correlated NF of a two-antenna XCSA is calculated. The results show that an improvement can be made of 3-4dB, depending on the correlation factor of the noise in both antennas. A low correlation factor of the signal, due to fading and other multipath effects, has a negative effect on the correlated NF of the two-antenna XCSA. Other effects, like moving sources and propagation delays between both paths, are analyzed. They are shown to have negligible effect in mobile devices, where antenna distance is small, the signal bandwidth is smaller then 6MHz and the measurement time longer then 20ms. Measurement have been performed to verify the two-antenna theory. For these measurements a low correlated NF XCSA is implemented with off-the-shelf components and PC-based spectral averaging software, similar to the existing XCSA. The effect of timedelay on correlation of two noise sources in multiple receive paths is verified using an idealized setup. In this case the measurement results match the theory. Furthermore, the noise floor in a lab-environment is measured with a two-antenna XCSA and a oneantenna XCSA. The measurement results show that the correlation of the noise seems very random and no frequency dependency or antenna distance dependency is found, unlike the theoretical results, this is probably due to fading and multipath effects in the lab-environment, which can resemble a real CR environment. Using two antennas for cross-correlation spectrum sensing with the XCSA will improve the correlated NF of the XCSA with at least 3dB, because the resistive splitter is lost. The effect of signal fading in case of two antennas needs to be investigated in the future, because it may decorrelate the signal, negatively affecting the sensing capabilities.