The IEEE 802.22 working group is developing a standard for a cognitive Wireless Regional Area Network (WRAN) that will operate in unused television channels and provide fixed wireless access services. The 802.22 network consists of a base station and a number of client stations, referred to as customer premises equipment (CPE). However, before using these unused channels, the channel availability needs to be confirmed. To achieve this goal, the base station relies on spectrum sensing results, geolocation information, and auxiliary information provided by the network manager. This blog post focuses mainly on the ‘spectrum sensing’ option of qualifying channel availability.
Spectrum sensing involves making observations of the radio frequency (RF) spectrum and reporting on the availability of unused spectrum for use by the WRAN. Sensing is required for analog television, digital television, and wireless microphones. The required detection time for all three signal types is 2 seconds. Regarding the required sensitivity, the following table summarizes the required value for the three licensed signals types (the sensing sensitivity is the power level at which the probability of detection is 90%, while the probability of false alarms is 10%):
Signal-to-noise ratio (SNR)
IEEE 802.22 does not mandate the use of a specific sensing technique in either the base station or CPE. They did evaluate, however, a number of sensing techniques divided in two main types: blind sensing and signal-specific sensing.
The first blind sensing technique is the most intuitive one: Energy Detection. The detector estimates the signal power in the channel and compares that estimate against a threshold. This method is the one used to detect the presence of interference in the cognitive OFDM video posted on Nutaq’s website. The following test statistic was used to make a decision on channel availability:
For large value of M, T is a Normal random variable with the following distribution (from the Central Limit Theorem):
This detector has the advantage of being simple to integrate with a fast sensing time. It also performs well at a very low SNR, with a priori knowledge of the noise power. However, without that knowledge, this detector is not robust enough to handle negative SNR values.
The second proposed blind estimation technique uses the eigenvalues of the correlation matrix. First, the sensor estimates the auto-correlation function (Ryy(m)) of the received signal. Then, we fill the correlation matrix R:
The matrix is then whitened and the eigenvalues of the resulting matrix are found. By having these eigenvalues, one can use several test statistics to measure the non-whiteness of the spectrum of the received signal (in other words, to determine if the received signal is additive white Gaussian noise (AWGN) or not). Two test statistics are proposed by the IEEE 802.22 standard: the maximum-minimum eigenvalue (MME) detection and the energy with minimum eigenvalue (EME) detection (see