represent the AOC of the primary user. One can see how the secondary user AOC depicted by the circle on the left‐hand side causes unintended interference with the primary user's receiver since the transmitting signal from the primary user could not be detected. One can see how distributed cooperative sensing or the use of a centralized arbitrator can better address the hidden primary user problems. In order to reduce the chance of encountering a hidden node problem, one can rely on:
1 geographically dispersed spectrum sensors
2 a centralized DSA decision process, which can have a bird's eye view of the area of operation
3 the centralized DSA decision‐making process having an algorithm that can estimate the AOC of the primary user based on multiple sensor input.
Figure 2.16 The hidden node problem.
2.5.2 Angle of the RF Beam Detection
When primary users use directional antennas such as microwave links, the locations of the primary users and the directions of their RF beam – which include the azimuth and elevation angle – can be detected by the geographically dispersed spectrum sensors and leveraged by the secondary user's DSA decision‐making process to create opportunistic spectrum use. The spectrum sensing techniques will estimate the geolocation or position of the primary users and the direction of their beam. With a directional antenna, if a primary user is transmitting in a specific direction, the secondary user can transmit in the opposite direction without creating interference, as shown in Figure 2.17.
Figure 2.17 Directional secondary user's leveraging of the primary user's beam angle to create spectral opportunity.
Obviously, the case illustrated in Figure 2.17 points out to the importance of having a centralized arbitrator or the secondary user having knowledge of peer nodes location and the primary user nodes locations and signal characteristics.
2.6 Other Sensing Techniques
There are many other sensing techniques that have been proposed in the literature and are not detailed here, including the following:
Multitaper spectral estimation. This technique is widely used in neuroscience and other biomedical engineering applications. It estimates the power spectrum of the signal that is a stationary ergodic random process with finite variance (wideband signals can carry these characteristics). It detects contiguous realization of the signal and uses a maximum likelihood estimator to calculate the signal's power spectral density.
Random Hough transform. While most non time domain spectrum sensors use Fourier transform (FT), this approach proposes the use of a different transform domain for signal detection. The Hough transform domain suits signals with periodic patterns where it exploits the statistical covariance of noise and signal. This spectrum sensing technique can be effective at detecting digital television (DTV) signals.
Wavelet transform based estimation. This approach uses another transform domain. Wavelet18 domain is used for detecting edges in the power spectral density. These edges can result from the transition from the occupied band to the empty band and vice versa. Analog implementation of wavelet transform based sensing have the advantage of needing low power consumption and it can be implemented in real time.
2.7 Concluding Remarks
Regardless if one is building spectrum sensing capabilities from the bottom‐up or utilizing an existing spectrum sensing technology, one must understand the different spectrum techniques that can be used as reviewed in this chapter. The decision‐making process using spectrum sensing information is covered in the next chapter.
When designing a system that uses DSA, one may build the best spectrum sensing capabilities and choose the best spectrum sensing hardware and configure it for the appropriate bands to sense and use the sensing parameters appropriately; however, if some critical factors are ignored, the design can be way suboptimal. One of the critical factors worth mentioning at the conclusion of this chapter is making sure a concept called blanking signal is applied. On a platform that has both a communications waveform and a spectrum sensor, it will be important for the communications waveform to inform the spectrum sensor of transmission time intervals. During these intervals, the spectrum sensor should refrain from collecting spectrum sensing information since spectrum sensing will be dominated by the emitted communications signal. Even if sensing is at a different frequency band from the transmission band, frequency domain harmonics can impact the sensing accuracy, as explained in Chapter 8.
Another critical factor worth mentioning is the relationship between dwell time and the signal characteristics. When performing same‐channel in‐band sensing, many aspects of the signal characteristics are known. For example, a frame transmitted over the air can have a preamble. The presence of the preamble can inform the energy detection process that the sensed energy level includes the presence of the communications signal. The absence of the preamble can inform the energy detection process that the sensed energy level is for noise or noise plus interfering signal. One can map dwell time to the frame time, taking multiple samples from the frame, or one can sample multiple frames in tandem with a larger dwell time. This can yield a good estimation of the noise floor when sensing the in‐band frequency. With augmented sensors, performing energy detection is separate from the demodulation process. Cyclostationary characteristics can also help define a dwell time that informs the energy detection process whether or not the sensed energy level includes the presence of the sensed signal. Having a dwell time that can allow the energy detection process to overlay noise, in‐band signal, and interfering signal can lead to a lower probability of misdetection and a lower probability of false alarm.
There are other critical factors to consider as discussed in the following chapters and introduced in Chapter 1, such as abstraction, the value of same‐channel in‐band sensing, making the best out of local, distributed, centralized, and hybrid decisions, the reduction of spectrum sensing control traffic, the powerful features of augmented sensing, and the role of policies and security with DSA systems.
Exercises
1 Create a table of 11 entries that map dBm values to milliwatt values in the range of 1–50 mW. Use this table to create six approximate RSSI thresholds to express:Undetectable signalVery weak signalWeak signalGood signalVery good signalExcellent signal
2 State some of the dimensions you would want a spectrum sensor to consider in sensing a 5G commercial cellular signal.
3 Consider the case of M‐AM (M‐level amplitude modulation, where M is an even number) where the number of bits transmitted per symbol is (log2M). The transmitted signal is related to a data symbol xi ∈ {0, 1, …, M − 1} by si(t) = (2xi − M + 1) φ (t), where φ (t) is the common signal shape to all signals. This is essentially a one dimensional signal in space. Assume the net amplitude modulation is symmetric at each AM constellation point. Arrange the M constellation points on a straight line
