Introduction

To study observations, we return yet again to the definition of the cognitive radio laid out in Chapter 1 and note once more that ‘A cognitive radio is a device which has four broad inputs, namely, an understanding of the environment in which it operates, an understanding of the communication requirements of the user(s), an understanding of the regulatory policies which apply to it and an understanding of its own capabilities.’ Getting these four inputs is what we mean by the phrase ‘observing the outside world’.

We can further detail some of the observations that are needed if we go through the various action categories outlined in the last chapter. To take action from a frequency perspective the cognitive radio must observe which signals are currently being transmitted, which channels are free, the bandwidth of those channels and perhaps whether the available channels are likely to be short lived or more durable. To take action from a spatial perspective, the cognitive radio needs to make observations about the spatial distribution of systems that must be avoided, or the spatial distribution of interferers and of the target radios. The cognitive radio needs to be able to monitor its power output and the power output of other systems. To take action to make a signal more robust or to maximise the throughput of the transmitted signal, the cognitive radio needs to make observations about the signal-to-noise ratio (SNR) at the target receivers, about the bit error rates and about the propagation conditions experienced by the transmitted signal (e.g. delay spread, doppler spread).

Introduction

To discuss regulation and standardisation in the context of cognitive radio is a challenge. Currently there are almost no regulations or standards in place for cognitive radio, as cognitive radios are still very much a thing of the future. Hence this chapter is more about classifying the general types of regulations that may be needed and the standards that are emerging than discussing what is already in place. In reality there is a wealth of regulatory issues that relate directly, indirectly or just ‘kind of relate’ to cognitive radio. Chapter 1 explored the role of cognitive radios in delivering new ways of managing the spectrum and looked at applications in the military, public safety and commercial domains. The new spectrum management regimes and the various potential applications may each give rise to the need for new regulations, some of which are specifically related to cognitive radios and some of which are related to creating the kind of environment in which cognitive radio applications can thrive. The purpose of this chapter, therefore, is to give a broad sense of what those issues might be, as well as to describe the current status of the standardisation efforts.

Regulatory issues and new spectrum management regimes

Much of the discussion about ‘regulations for cognitive radio’ is about ‘regulations for new spectrum management regimes in which cognitive radios can operate’.

Introduction

The first chapter of this book focused on the application areas that will drive cognitive radio technology. This chapter acts as a bridge to the remainder of the book. It seeks to provide the reader with a broad sense of all that is involved in cognitive radio technology. In order to do this we go to the heart of the cognitive radio but not at first using technology as an example. Instead we step back and take a look at how decisions are made in a more abstract manner before returning to the radio world. The final part of the chapter provides a roadmap for the rest of the book.

Setting the scene for understanding cognitive radio

The first question to think about is: how do we make decisions? How do we reason and come to conclusions? We begin this discussion by looking at a simple example.

The lone radio

Scenario 1: I am about to go out and must decide whether I should take an umbrella with me or not. The umbrella is heavy and cumbersome and, while I don't want to get wet, I don't want to take the umbrella with me if it is not necessary.

In this example two actions are possible, namely take umbrella or don't take umbrella. I need to determine how likely it is to rain in order to decide whether to take the umbrella or not.

Introduction

We now reach the ‘decide’ part of the ‘observe, decide and act’ cycle. In very simple terms the decision-making process is about selecting the actions the cognitive radio should take. Using the vocabulary introduced in Chapter 2, it is about choosing which ‘knobs’ to change and choosing what the new settings of those ‘knobs’ should be. Decision-making goes very much to the heart of a cognitive radio.

The decision-making process: part 1

In Table 3.2 a variety of cognitive radio applications and the main highlevel actions associated with them were presented. On examining the table we noted that many of the actions, whether commercial, public safety or military based, centre on two activities:

1. The cognitive radio shapes its transmission profile and configures any other relevant radio parameters to make best use of the resources it has been given or identified for itself, while at the same time not impinging on the resources of others.

2. If and when those resources change, it reshapes its transmission profile and reconfigures any other relevant operating parameters, and in doing so it redirects resources around the network.

A re-examination of Table 3.2 will confirm that these actions are standard throughout a whole variety of applications. It therefore comes as no surprise that two kinds of decisions that regularly need to be made are decisions that map to these two activities, namely decisions about how resources are distributed and decisions about how those resources are exactly used.

During the production phase of this book, the FCC released two reports that are of relevance to this book. At that stage it was too late to include details of the reports in the main body of the text. This short appendix addresses the issues briefly.

On 15 October 2008 the FCC released their report (FCC/OET 08-TR-1005) on the Evaluation of the Performance of Prototype TV-Band White Space Devices Phase II. The opening paragraph of the report summarises what the report shows:

The Federal Communications Commission's Laboratory Division has completed a second phase of its measurement studies of the spectrum sensing and transmitting capabilities of prototype TV white space devices. These devices have been developed to demonstrate capabilities that might be used in unlicensed low power radio transmitting devices that would operate on frequencies in the broadcast television bands that are unused in each local area. At this juncture, we believe that the burden of ‘proof of concept’ has been met. We are satisfied that spectrum sensing in combination with geo-location and database access techniques can be used to authorize equipment today under appropriate technical standards and that issues regarding future development and approval of any additional devices, including devices relying on sensing alone, can be addressed.

The report goes on to state that

All of the devices were able to reliably detect the presence a clean DTV signal on a single channel at low levels in the range of – 116 dBm to – 126 dBm; the detection ability of each device varied little relative to the channel on which the clean signal was applied.

Introduction

In Chapter 1 the working definition for cognitive radio used throughout this book was presented. That definition ended with the statement ‘A cognitive radio is made from software and hardware components that can facilitate the wide variety of different configurations it needs to communicate.’ In this chapter we look at the hardware involved. There is no one right way to build a cognitive radio so the chapter merely aims to give a sense of what kind of hardware can be used and some of the related performance issues.

A complete cognitive radio system

In a cognitive radio receiver, the antenna captures the incoming signal. The signal is fed to the RF circuitry and is filtered and amplified and possibly downconverted to a lower frequency. The signal is converted to digital format and further manipulation occurs in the digital domain. On the transmit side the opposite occurs. The signal is prepared and processed and at some stage is converted from digital to analogue format for transmission, upconverted to the correct frequencies and launched on to the airwaves via the antenna.

Throughout this book we have been using the terms ‘cognitive radio’ and ‘cognitive node’ interchangeably. The reason for this is that a cognitive radio will almost all of the time function as a node in a network. Therefore it is useful to think of the complete cognitive radio system in terms of a communication stack.

Having covered the fundamentals of meshes, we now arrive at the point where we may begin to consider the big and often asked questions about mesh, four of which we consider together, via our list of hypotheses. As a reminder, these are that

1. meshes self-generate capacity,

2. meshes improve spectral efficiency,

3. directional antennas help a mesh, and

4. meshes improve the overall utilisation of spectrum.

We will examine them formally, via analysis of existing peer reviewed publications, followed by some more recent analysis and insight of our own [1, 2]. A key problem in assessing the published literature is that different assumptions are made in different published papers; a direct comparison is thus at risk of being inconsistent. We spend some time at the outset to ensure we avoid this issue.

We will bear in mind that we are predominantly interested in our six application examples of Chapter 2. This will set helpful bounds to our scope for testing the hypotheses.

When we look at Hypothesis 1 which is concerned with capacity, we form our initial viewpoint via a simple thought experiment, which looks at how we expect the capacity of a mesh might behave versus demand, relative to the known case of cellular. This is followed by a summary of four important peer reviewed research papers in the field, which concern system capacity. We contend that the important conclusions presented in these papers were never intended to be used by readers as evidence that a real-world mesh can self-generate capacity.

The aim of using a mobility model is to reflect as accurately as practicable the real conditions themselves. One way to do this is to use motion traces, which are logs of real-life node movements over a representative period of time. There are not many such logs available for use even with established cellular schemes, and none are known to this author which cover mesh environments. The focus then must move to synthetic models. Such a model will deal with a number of nodes and may include parameters such as speed and direction of movement, the ability to pause at some locations and a bound to the model area. The models available are mostly fairly simple to implement, since they are intended for use in simulators where a tractable run time is expected. It is probably the case that present models err on the side of simplicity at the expense of realism. On the other hand, moving too close to the actual environment requires a very specific model – which may then not be adequately representative of all environments. The choice of model is thus a subject which needs to be understood, in order to interpret specific protocol and other simulation results for wider contexts.

Camp et al. [1] review 12 different mobility models which have been applied to mesh simulations at various points in the published literature. Their work is an often quoted indication that the choice of model alone can strongly affect the results when testing the exact same routing protocol. For the purposes of this book three models are noted as being appropriate.

The design of an efficient multiple access and multiplexing scheme is more challenging on the uplink than on the downlink due to the many-to-one nature of the uplink transmissions. Another important requirement for uplink transmissions is low signal peakiness due to the limited transmission power at the user equipment (UE). The current 3G systems use the wideband code division multiple access (WCDMA) scheme both in the uplink and in the downlink. In a WCDMA downlink (Node-B to UE link) the transmissions on different Walsh codes are orthogonal when they are received at the UE. This is due to the fact that the signal is transmitted from a fixed location (base station) on the downlink and all the Walsh codes received are synchronized. Therefore, in the absence of multi-paths, transmissions on different codes do not interfere with each other. However, in the presence of multi-path propagation, which is typically the case in cellular environments, the Walsh codes are no longer orthogonal and interfere with each other resulting in inter-user and/or inter-symbol interference (ISI).

The problem is even more severe on the uplink because the received Walsh codes from multiple users are not orthogonal even in the absence of multi-paths. In the uplink (UE to Node-B link), the propagation times from UEs at different locations in the cell to the Node-B are different. The received codes are not synchronized when they arrive at the Node-B and therefore orthogonality cannot be guaranteed.

In the previous chapter, we discussed how multiple transmission antennas can be used to achieve the diversity gain. The transmission diversity allows us to improve the link performance when the channel quality cannot be tracked at the transmitter which is the case for high mobility UEs. The transmission diversity is also useful for delay-sensitive services that cannot afford the delays introduced by channel-sensitive scheduling. The transmission diversity, however, does not help in improving the peak data rates as a single data stream is always transmitted. The multiple transmission antennas at the eNB in combination with multiple receiver antennas at the UE can be used to achieve higher peak data rates by enabling multiple data stream transmissions between the eNB and the UE by using MIMO (multiple input multiple output) spatial multiplexing. Therefore, in addition to larger bandwidths and high-order modulations, MIMO spatial multiplexing is used in the LTE system to achieve the peak data rate targets. The MIMO spatial multiplexing also provides improvement in cell capacity and throughput as UEs with good channel conditions can benefit from multiple streams transmissions. Similarly, the weak UEs in the system benefit from beam-forming gains provided by precoding signals transmitted from multiple transmission antennas.

MIMO capacity

A MIMO channel consists of channel gains and phase information for links from each of the transmission antennas to each of the receive antennas as shown in Figure 7.1.

The LTE system requirements mandate significant improvement in performance relative to the Release 6 HSPA system. In particular, the spectrum efficiency improvement targets for the downlink are three to four times that of the Release 6 HSPA system. The spectral efficiency improvement targets for the uplink are relatively modest with two to three times improvement over Release 6 HSPA. One of the reasons for lower improvement targets for the uplink is that the same antenna configuration is assumed for the LTE system and Release 6 HSPA system. On the other hand for downlink, LTE assumes two transmit antennas while Release 6 HSPA baseline system assumes only one transmit antenna at the Node-B. Similar targets are set for the peak data rates and also cell-edge performance improvements. The spectral efficiency target for the MBSFN, which is a downlink only service, is set at an absolute number of 1 bps/Hz.

An evaluation methodology specifying the traffic models and simulation parameters was developed for assessing the performance of the LTE and Release 6 HSPA systems. The goal of the evaluation methodology is to provide a fair comparison as all the parties participating in the simulations campaign can evaluate performance under the same set of assumptions. In this chapter, we describe LTE simulations methodology and provide relative performance of the LTE system and Release 6 HSPA system.

Traffic models

In this section, we discuss various traffic models considered in the performance verification. The traffic mix scenarios are given in Table 19.1.

With the exception of a scheduling request, all uplink control consists of feedback information to support downlink transmissions. The channel quality feedback is provided to support downlink channel-sensitive scheduling and link adaptation. The rank and precoding matrix indication is used for selecting a downlink MIMO transmission format. The ACK/NACK signaling provides feedback on downlink hybrid ARQ transmissions. In contrast to uplink control, the only feedback information on the downlink is ACK/NACK signaling to support uplink hybrid ARQ operation and transmission power control (TPC) commands to support uplink power control. The reason for this asymmetry is simply the fact that both the uplink and the downlink schedulers resides in the eNB. Therefore, the bulk of downlink signaling involves uplink and downlink scheduling grants that convey information on the transmission format and resource allocation for both the uplink and downlink transmissions. In order to support the uplink channel-sensitive scheduling, the uplink channel quality is estimated from the uplink sounding reference signal (SRS).

The three downlink control channels transmitted every subframe are physical control format indicator channel (PCFICH), physical downlink control channel (PDCCH) and physical hybrid ARQ indicator channel (PHICH). The PCFICH carries information on the number of OFDM symbols used for PDCCH. The PDCCH is used to inform the UEs about the resource allocation as well as modulation, coding and hybridARQ control information. Since multiple UEs can be scheduled simultaneously within a subframe in a frequency or space division multiplexed fashion multiple PDCCHs each carrying information for a single UE are transmitted.

The goal of power control is to transmit at the right amount of power needed to support a certain data rate. Too much power generates unnecessary interference, while too little power results in an increased error rate requiring retransmissions and hence resulting in larger transmission delays and lower throughputs. In a WCDMA system, power control is important particularly in the uplink to avoid the near–far problem. This is because the uplink transmissions are nonorthogonal and very high signal levels from cell-center UEs can overwhelm the weak signals received from cell-edge UEs. Therefore, a very elaborate power control mechanism based on the fast closed-loop principle is used in the WCDMA system. Similarly, power control is used for the downlink of WCDMA systems to support the fixed rate delay-sensitive voice service. However, for high-speed data transmission in WCDMA/HSPA systems, transmissions are generally performed at full power and link adaptation is preferably used to match the data rate to the channel conditions.

The LTE uplink uses orthogonal SC-FDMA access and hence the near–far problem of WCDMA does not exist. However, high levels of interference from neighboring cells can still limit the uplink coverage if UEs in the neighboring cells are not power controlled. The cellular systems are generally coverage limited in the uplink due to limited UE transmit power. The increased levels of interference from neighboring cells increase Interference over Thermal (IoT) limiting coverage at the desired cell. Therefore, uplink power control is beneficial in an orthogonal uplink access as well.

The LTE system supports fast dynamic scheduling on a per subframe basis to exploit gains from channel-sensitive scheduling. Moreover, advanced techniques such as link adaptation, hybrid ARQ and MIMO are employed to meet the performance goals. A set of physical control channels are defined in both the uplink and the downlink to enable the operation of these techniques. In order to support channel sensitive scheduling and link adaptation in the downlink, the UEs measure and report their channel quality information back to the eNB. Similarly, for downlink hybrid ARQ operation, the hybrid ARQ ACK/NACK feedback from the UE is provided in the uplink.

Two types of feedback information are required for MIMO operation, the first is MIMO rank information and the second is preferred precoding information. It is well known that even when a system supports N × N MIMO, rank-N or N MIMO layers transmission is not always beneficial. The MIMO channel experienced by a UE generally limits the maximum rank that can be used for transmission. In general, for weak users in the system, a lower rank transmission is preferred over a higher rank transmission. This is because at low SINR, the capacity is power limited and not degree-of-freedom limited and therefore multiple layers transmission is not helpful. Moreover, when the antennas are correlated, the channel matrix is rank deficient leading to a single layer or rank-1 transmission. Therefore, the system should support a variable number of MIMO layers transmission to maximize gains from MIMO.

An important requirement for the LTE system is improved cell-edge performance and throughput. This is to provide some level of service consistency in terms of geographical coverage as well as in terms of available data throughput within the coverage area. In a cellular system, however, the SINR disparity between cell-center and cell-edge users can be of the order of 20 dB. The disparity can be even higher in a coverage-limited cellular system. This leads to vastly lower data throughputs for the cell-edge users relative to cell-center users creating a large QoS discrepancy.

The cell-edge performance may be either noise-limited or interference-limited. In a noise-limited situation that typically occurs in large cells in rural areas, the performance can generally be improved by providing a power gain. The power gain can be achieved by using high-gain directional transmit antennas, increased transmit power, transmit beam-forming and receive beam-forming or receive diversity, etc. The total transmit power is generally dictated by regulatory requirements and hence limits the coverage gains possible due to increased transmit power.

The situation is different in small cells interference-limited cases, where, in addition to noise, inter-cell interference also contributes to degraded cell-edge SINR. In this case, providing a transmit power gain may not help because as the signal power goes up, the interference power also increases. This is assuming that with a transmit power gain all cells in the system will operate at a higher transmit power.