The Mobile Internet - A Success Story so far

When 3G was launched initially with WCDMA technology (Release 99), it was rather a disappointment with not many services being successful. Some years later, the mobile Internet took off when a number of factors came together:

• HSPA as a technological evolution of 3G with low latency and higher data rate

• Attractive flat-rate price plans by mobile operators

• Availability of mobile broadband hardware in terms of dongles and built-in 3G modules in notebooks

• Smart phones with attractive user interfaces, e.g., iPhone, Android

• Complete country-coverage with HSPA and HSPA+ by mobile operators.

This take-up of the mobile Internet generated substantial additional revenues for mobile operators, at a time when voice and text message revenues started to decline in saturated markets such as Europe. For example, Vodafone had a data revenue growth of 19% in financial year 2009/2010, with more than €4 Billion generated by non-SMS data. Today, only 11% of phones are smartphones, but by 2013 it is expected that more than a third of all active phones within the Vodafone network will be smartphones.

This data revenue growth comes along with a cost for mobile operators - namely data traffic growth. Fig. 2.1 shows the actual and projected traffic growth for Vodafone's European networks in Petabytes/year [Vod10]. It can be seen that data traffic has substantially surpassed voice traffic.

In this chapter, we now look into algorithm implementation aspects connected to CoMP. In Section 10.1, the issue of numerically robust and flexible multicell precoding is addressed, while Section 10.2 looks into the performance of interference rejection combining filters under practical conditions.

Robust and Flexible Base Station Precoding Implementation

As we have seen in Sections 6.3 and 6.4, spectral efficiency can be significantly increased if downlink CoMP schemes based on multi-cell joint signal processing are applied. Here, multiple base stations (BSs) perform a joint and coherent transmission towards multiple user equipments (UEs). Especially cell-edge users with symmetric links gain from the joint transmission (JT) with minimized inter-cell and inter-user interference compared to conventional cellular mobile networks.

In a cellular system, a spatial downlink resource allocation algorithm may switch between a single link transmission comprising one BS and one UE, a single BS multi-user scenario or finally a multi-cell joint transmission. The precoder must be able to cope with different system setups or matrix dimensions respectively. Such an implementation requires a higher signal processing flexibility as compared to non-cooperative setups. Physical link parameters like line-of-sight (LOS), non-LOS, or correlated transmit and receive antenna patterns influence the CoMP multi-user MIMO (MU-MIMO) eigenvalue spread and finally the spatial diversity and multiplexing gains.

Furthermore, in real-time hardware, the link parameters are related to finite precision effects.

In this chapter, we address how CoMP can be applied selectively and adaptively to well-chosen sets of terminals in a mobile communications system. While Section 11.1 focuses on scheduling approaches, where a central scheduling unit performs multi-cell resource allocation in the context of non-cooperative or joint transmission in a cellular downlink, Section 11.2 looks into radio link control and signalling aspects connected to establishing CoMP on-demand. Finally, Section 11.3 ventures into the field of ad-hoc CoMP, where cooperation is established flexibly after uplink transmission has already taken place.

Centralized Scheduling for CoMP

In this section, we discuss centralized multi-cell scheduling for a system using either non-cooperative transmission or joint transmission in the downlink. After Subsection 11.1.1 motivates the topic, Subsection 11.1.2 presents the studied scenario and its main models. Subsection 11.1.3 introduces some relevant scheduling problems, where the aim is to maximize system throughput. Subsection 11.1.4 analyzes the problems introduced earlier through system level simulations. Finally, Subsection 11.1.5 adds some final remarks on the problem of centralized scheduling and its extension to uplink scenarios.

Introduction

In previous chapters, we have typically observed transmissions between multiple base stations (BSs) and user equipments (UEs) on a single orthogonal frequency division multiplex (OFDM) sub-carrier, assuming that the assignment of system resources to the communicating entities has already taken place.

In this chapter, we focus on CoMP schemes where user data or received signals connected to multiple users are exchanged between base stations for joint signal processing. Such schemes promise larger spectral efficiency gains than pure interference coordination techniques, but typically come at the price of larger backhaul requirements and (particularly in the downlink) more severe synchronization requirements. After Sections 6.1 and 6.2 introduce centralized and decentralized uplink CoMP schemes, respectively, 6.3 and 6.4 focus on the downlink.

Uplink Centralized Joint Detection

In this section, uplink centralized joint detection with infinite or limited backhaul capacity is studied, as already introduced in Section 4.3.1. A cluster of base stations (BSs) sends either raw or preprocessed receive signals from the user equipments (UEs) to a CoMP central unit (CCU), where joint processing is performed to deal with inter-cell interference. The CCU can be a separate entity, or any of the BSs involved in the cooperation. This is practical in systems such as LTE-A employing a flat network architecture. The detection algorithms are applicable to various static or dynamic clustering schemes as described in Chapter 7.

Introduction

When multiple BSs are connected via perfect backhaul links with infinite capacity, uplink centralized joint detection resembles a multiple access channel (MAC) problem, where the CCU is a super-receiver, and the BSs form a distributed antenna system (DAS) [Mol01].

As mentioned in previous chapters, CoMP has the capability to significantly enhance spectral efficiency and cell-edge throughput. However, CoMP may require additional signaling overhead on the air interface and the backhaul, in particular joint signal processing CoMP as introduced in Chapter 6. Therefore, in practice only a limited number of base stations can cooperate in order to keep the overhead manageable. This raises the question which base stations should form cooperation clusters in order to exploit the advantages of CoMP efficiently at limited complexity.

In general, one can distinguish between static and dynamic clustering algorithms. Static clusters are kept constant over time and designed based on geographical criteria as the positions of base stations and the morphology of the surroundings. In the case of dynamic clustering, the system continuously adapts the clustering strategy to changing parameters such as user equipment (UE) locations and radio frequency (RF) conditions. Here, the central question is on which information the adaptation of clusters shall be based, and where in the system clustering decisions are made.

To illustrate concrete clustering results and their corresponding performance, we use two different setups in this chapter. On one hand, we consider an idealistic setup, i.e. a hexagonal layout of up to M = 111 cells, grouped into sites of 3 cells each, with an inter-site distance (ISD) of 500 m.

In this chapter, we address a last, but absolutely not least important challenge connected to CoMP, namely the fact that most base station cooperation schemes require information exchange over a backhaul infrastructure. Depending on the existing infrastructure of a mobile operator, both backhaul capacity and latency requirements of some CoMP schemes may be the main cost drivers or potential show stoppers on the roadmap towards CoMP. The chapter starts with addressing fundamental aspects of backhaul-constrained cooperation in Section 12.1, after which concrete backhaul capacity and latency requirements of various uplink and downlink CoMP schemes and their scaling behavior are derived in Section 12.2. Finally, Section 12.3 gives an overview on existing and upcoming backhaul technology options, and hence gives the reader a feeling of whether particular CoMP schemes can be expected to be technically and commercially feasible in the near future or not.

Fundamental Limits of Interference Mitigation with Limited Backhaul Cooperation

As we have seen in previous parts of this book, cooperation among base stations (BSs) via infrastructure backhaul networks can help mitigate interference by forming distributed multiple-input multiple-output (MIMO) systems, while the rate at which BSs cooperate is limited in wide-band cellular systems. How much interference can one bit of backhaul cooperation mitigate? In this section, we study the two-user Gaussian interference channel with limited backhaul cooperation to answer this question in a simple setting.

In this chapter, we introduce CoMP schemes where no or little information is exchanged between cooperating base stations. In Section 5.1, we observe an interference-aware downlink transmission scheme where each base station performs individual intra-cell beamforming, while the terminals are able to mitigate inter-cell interference to a certain extent through a particular interference estimation and rejection concept. The level of base station cooperation is then increased in Sections 5.2 and 5.3, where joint multi-cell scheduling and link adaptation, and multi-cell coordinated beamforming are investigated, respectively.

Downlink Multi-User Beamforming with Interference Rejection Combining

In this section, we evaluate a non-cooperative downlink transmission scheme, i.e. where no explicit cooperation takes place between base stations (BSs), but where interference-aware transmission and reception is performed within cells. The BSs perform intra-cell precoding based on limited feedback from the user equipments (UEs), in conjunction with interference-aware scheduling and interference rejection combining (IRC) at the terminal side. This section is based on “Interference-aware scheduling in the synchronous cellular multi-antenna downlink”, by L. Thiele, M. Schellmann, T. Wirth and V. Jungnickel, which appeared in [TSWJ09]. © 2009 IEEE.

Introduction

Transmission with multiple antennas both at the transmitting and receiving ends of a wireless link has become increasingly mature in recent years. From theory, the fundamental capacity gain of the multiple-input multiple-output (MIMO) radio link, being proportional to the minimum of the number of transmit and receive antennas, is well understood for an isolated point-to-point link.

In this chapter, we finally provide field trial results for different CoMP schemes discussed in this book. While Section 13.1 observes the performance of successive interference cancelation (SIC) algorithms and uplink macro diversity in laboratory and outdoor drive tests, Section 13.2 provides results on a large-scale field trial in Dresden, where two terminals were moved within a setup of 16 base stations, and the gain of uplink joint detection is assessed. Sections 13.3 and 13.4 then present two slightly different implementations of downlink joint transmission CoMP, where the former puts a primary focus on various challenges encountered, while the latter section discusses an evaluation methodology that enables to predict downlink CoMP performance over larger areas. The chapter is concluded with a review on the lessons learnt through field trial implementation and test in Section 13.5.

Real-time Implementation and Trials of Advanced Receiver and Uplink CoMP Schemes

In this section, field trial activities connected to uplink CoMP schemes and advanced receiver algorithms are described. These new features are promising methods to increase performance for LTE-A networks and have been implemented in a Bell Labs eNodeB prototype. In a first step, an uplink (UL) successive interference cancelation (SIC) algorithm is considered in a single-cell scenario. Simulations and lab tests are used to select the best algorithm approach and to optimize involved parameters. To show the feasibility and to assess the advantages of such a SIC receiver algorithm in a real deployment scenario, field trials in the EASY-C research test bed in Dresden [I+09] have been conducted.

This chapter deals with another major challenge connected to CoMP, namely the synchronization of cooperating and cooperatively served devices in time and frequency. On one hand, there are different local oscillators in each base station and mobile terminal that lead to deviations in the carrier frequency according to its nominal value. On the other hand, there are variations in the symbol timing between each transmitter and receiver station. Both effects need to be compensated by synchronization techniques.

In cellular networks, we can distinguish between a network synchronization among all involved base stations and the alignment of the user equipments to that time and frequency reference. The basic definitions of the synchronization terms as well as procedures for the reference network synchronization are described in Section 8.1. The impact of symbol timing mismatches on CoMP is then treated in Section 8.2, before Section 8.3 concludes this chapter with the analysis of the impact of residual carrier frequency offsets on CoMP performance.

Synchronization Concepts

Synchronization is the process of establishing a common notion of time among two or more entities. In the context of wired and wireless communication networks, synchronization enables coordination among the nodes in the network and can facilitate applications such as distributed sensing. Precise synchronization can also facilitate scheduling of communication resources as well as interference avoidance in multi-access networks.

Summary of this Book

In this book, coordinated multi-point (CoMP) has been investigated from the point of view of a multitude of authors, some of which have been working in this field since a decade. Clearly, the different opinions stated throughout the book emphasize that CoMP is still a controversially discussed topic, and it is still too early to draw final conclusions. Nevertheless, we now want to briefly review the contributions in this book and try to extract a more or less conclusive big picture.

Most Promising CoMP Schemes and Potential Gains

First, let us summarize the key CoMP schemes discussed in this book, which were categorized according to the extent of base station (BS) cooperation involved:

Non-cooperative, but interference aware transceiver techniques have been investigated in Section 5.1. Here, LTE Release 8 style precoding is used in the downlink, in conjunction with multi-cell interference estimation at the receiver side, which was also covered in Section 10.2. The results predict large gains from using smart multi-cell pilot and channel estimation schemes, even though no cooperation between BSs is required.

Interference coordination schemes, in comparison, are based on some limited extent of information exchange between BSs. In Section 5.2, multi-cell interference prediction through the exchange of scheduling tables and coordinated scheduling by a central unit were introduced, yielding average spectral efficiency gains of more than 20%, according to system level simulation results in Section 14.3.