Chapter 16

LTE E-MBMS Capacity and Intersite Gains

Amrico Correia, Rui Dinis, Nuno Souto, and Joo Silva Contents
16.1 Introduction ................................................................................. 587 16.2 Objectives and Requirements .......................................................... 590 16.3 Evaluation Methodology and Simulation Assumptions ....................... 592 16.3.1 Link-Level Simulator Design ................................................ 593 16.3.2 Radio Access Network System Level Simulator ....................... 595 16.4 System Level Performance Results .................................................... 597 16.4.1 BLER Results ..................................................................... 598 16.4.2 Coverage Results ................................................................ 601 16.4.3 Throughput Results ............................................................ 603 16.5 Summary and Conclusions .............................................................. 607 16.6 Open Issues ................................................................................... 608 References .............................................................................................. 608

16.1 Introduction
The Third Generation Partnership Project (3GPP) has launched the study item evolved UMTS terrestrial radio access (UTRA) and UMTS terrestrial radio access network (UTRAN), which studies the means to achieve further substantial leaps in terms of service provisioning and cost reduction. The overall target of the long-term evolution (LTE) of 3G was to arrive at an evolved radio access technology that can
587

588

Evolved Cellular Network Planning and Optimization

provide service performance on parity with current fixed-line access. As it is generally assumed that there will be a convergence toward the use of Internet protocol (IP)based protocols (i.e., all services in the future will be carried on top of IP), the focus of this evolution was on enhancements for packet-based services. 3GPP concluded the Release 8 of the evolved 3G radio access technology in 2008, with subsequent initial deployment in the 20092010 time frame. At this point, it is important to emphasize that this evolved RAN is an evolution of the current 3G networks, building on already made investments. The 3GPP community has been working on LTE, and various contributions were made to implement evolved MBMS in LTE [1]. Orthogonal frequency division multiplexing/orthogonal frequency division multiple access OFDM/OFDMA [24], used in the physical layer (downlink connection) of LTE is an attractive choice to meet requirements for high data rates, with correspondingly large transmission bandwidths and flexible spectrum allocation. OFDM also allows for a smooth migration from earlier radio access technologies and is known for achieving high performance in frequency-selective channels. Furthermore, it enables frequency domain adaptation, provides benefits in broadcast scenarios, and is well suited for multiple-input multiple-output (MIMO) processing. The possibility to operate in vastly different spectrum allocations is essential. Different bandwidths are realized by varying the number of subcarriers used for transmission, whereas the subcarrier spacing remains unchanged. In this way, operation in spectrum allocations of 1.4, 3, 5, 10, 15, and 20 MHz, respectively, can be supported. For MBMS support within a certain cell coverage area for a given coverage target, the modulation and coding scheme (MCS) of the MBMS transport channel typically has to be designed under worst-case assumptions. Except for cell-edge users that typically experience large inter-cell interference, users with better channel conditions (closer to the base station) could receive the same service with a better quality (e.g., video resolution), as their receiving signal-to-noise ratios (SNR) would allow usage of a higher-rate MCS. Hierarchical modulations [58], which have been specified for broadcast systems like digital video broadcast terrestrial (DVB-T) or MediaFLO, is one way of accounting for unequal receiving conditions. Here, a signal constellation like 16-QAM, with each symbol being represented by four bits, is interpreted in a sense that the first two bits belong to an underlying QPSK alphabet. This enables the use of two independent data streams with different sensitivity requirements. In the foregoing example, the so-called high-priority stream employs QPSK modulation and is designed to cover the whole service area. The low-priority stream requires the constellation to be demodulated as 16-QAM and provides an additional or refined service via the two additional bits. This may transport an additional MBMS channel with a different type of service, or an enhancement stream that, for example, leads to enhancing the resolution of the base stream. A design parameter that determines the constellation layout allows the control of the amount of distortion that the enhancements symbols add to the baseline constellation and can be used to control the ratio of coverage areas or service data rates. Theoretical evaluation of these types of modulations, where it is explicitly shown the dependence of the individual bit

LTE E-MBMS Capacity and Intersite Gains

589

streams performance on the constellation design parameter, has been previously presented in [9, 10]. Specifically for broadcast and multicast transmissions in a mobile cellular network, depending on the communication link conditions, some receivers will have better SNR than others, and thus the capacity of the communication link for these users is higher. Hierarchical constellations and MIMO (spatial multiplexing [11,12]) are methods able to offer multiresolution and take advantage of the different link capacities. In [1315] these two forms of multiresolution methods (considering the WCDMA technology) have been evaluated. In OFDMA-based networks, the transmission of different fractions of the total set of subcarriers (chunks) depending on the position of the mobiles is another way to offer multiresolution. All of these methods are able to provide unequal bit error protection. In any case, there are two or more classes of bits with different error protection levels to which different streams of information can be mapped. Regardless of channel conditions, a given user always attempts to demodulate both types of bitsthe most protected and the ones carrying additional resolution. Depending on its position inside the cell, more or less blocks with additional resolution will be correctly received by the mobile user. However, the basic quality will be always correctly received independently of the position of any user within the 95% coverage target. For increased distance between terminals and base station, decreased bit rates are correctly received due to the decrease of SNR. Adaptive modulation and coding (AMC) is a technique that maximizes the total throughput for unicast transmissions. The decrease of SNR with distance is common to unicast or broadcast/multicast transmissions. However, for broadcast/multicast the same video content is transmitted, and AMC is not possible without personal uplink feedback. With the introduction of multiresolution techniques, maximization of the total throughput is the goal to achieve. There will be support for MBMS right from the first version of LTE specifications. However, specifications for E-MBMS are in the early stages. Two important scenarios have been identified for E-MBMS: one is single-cell broadcast, and the second is MBMS single-frequency network (MBSFN). MBSFN is a new feature that is being introduced in the LTE specification. MBSFN is envisaged for delivering services such as mobile TV using the LTE infrastructure and is expected to be a competitor to DVB-H-based TV broadcast. In MBSFN, the transmission happens from a time-synchronized set of enhanced-nodeBs (eNBs) using the same resource block. This enables over-the-air combining, thus improving the signal-to-interference plus noise ratio (SINR) significantly compared to nonSFN operation. The cyclic prefix (CP) used for MBSFN is slightly longer, and this enables the UE to combine transmissions from different eNBs, thus somewhat negating some of the advantages of SFN operation. There will be six symbols in a slot of 0.5 ms for MBSFN operation versus seven symbols in a slot of 0.5 ms for non-SFN operation. System-level simulations for broadcast/multicast with multiresolution and different fractional frequency reuse for LTE are necessary to evaluate achievable capacity

590

Evolved Cellular Network Planning and Optimization

and intersite gains compared to single-resolution systems, WCDMA based. Taking the 95% coverage as reference the evaluation of the achievable capacity gain (number of transmitted mobile TV channels for WCDMA and LTE) is done [16]. The intersite distance gain is also evaluated, allowing for a substantial reduction in the number of cell sites when LTE replaces WCDMA. The scenario based on the use of single-frequency network (SFN), with the multimedia broadcast over SFN (MBSFN) channel, is also evaluated for 16-QAM/64-QAM hierarchical modulations and compared with the present MBMS network based on WCDMA.

16.2 Objectives and Requirements

The multimedia broadcast and multicast service (MBMS), introduced by 3GPP in Release 6 was intended to use network/radio resources efficiently (by transmitting data over a common radio channel), both in the core network and, most importantly, in the air interface of UTRAN, where the bottleneck is placed to a large group of users. MBMS included point-to-point (PtP) and point-to-multipoint (PtM) modes. The former allowed individual retransmissions but the latter did not. MBMS is targeting high (variable) bit rate services over a common channel. One of the most important properties of MBMS is resource sharing among many user equipments (UEs), meaning that many users should be able to listen to the same MBMS channel at the same time. Thus, power should be allocated to this MBMS channel for arbitrary UEs in the cell to receive MBMS service. PtM transmission does not employ feedback and therefore needs to be statically configured to provide desired coverage in the cell. The transmitted signal is lowest at the cell border, and therefore the PtM bearer can greatly benefit from exploiting also the signals from adjacent cells transmitting the same service (i.e., from soft combining). While in the 3GPP LTE specification, two types of evolved-MBMS transmission scenarios exist: Multicell transmission (MBSFN over an SFN) on a dedicated frequency layer or on a shared frequency layer Single-cell transmission (SC-PMP: Single cell point to multipoint) on a shared frequency layer In 3GPP Release 6, the only specified transmission scenario for the MBMS transmission is SC-PMP. However, soft-diversity combining is possible as long as the delay between different base station transmissions of the same content allows macrodiversity. Multicell transmission in an SFN area is a way to improve the spectral efficiency. Because all MBMS cells transmit the same MBMS session data, signals can be combined for a UE located at a cell boundary. Furthermore, the multicell transmission may be provided over a cell group that comprises cells that transmit the same service. In contrast, single-cell transmission covers only one cell or one eNode B. In addition, the concept of a dynamic MBSFN area is introduced where the MBMS transmission

LTE E-MBMS Capacity and Intersite Gains

591

is switched off in some cells of the MBSFN area when a certain MBMS is not needed there. In some cases, the released resource can be reused for other MBMS or unicast services. The decision to turn off the MBSFN transmission in a cell is based on two factors: Local existence, which refers to the number of UEs that are both interested in the current MBMS and located in this cell Contribution for neighboring cells, which refers to the number of UEs that are both interested in the current MBMS and located in neighboring cells having the same MBSFN transmission. The cells of an MBSFN area contribute to the MBSFN transmission only if there are UEs that are interested in the particular service in this MBSFN area. The introduction of hierarchical modulation in a broadcast cellular service like E-MBMS requires a scalable video codec as shown in Figure 16.1 [13, 14], where the base layer transmission provides the minimum quality, and one or more enhancement layers offer improved quality at increasing bit/frame rates and resolutions. Besides being a potential solution for content adaptation, scalable video schemes may also allow an efficient usage of radio resources in evolved MBMS (E-MBMS).

Base layer UE2

Node B

UE1

Base layer + enhanced layer

Figure 16.1 Scalable video transmission.

592

Evolved Cellular Network Planning and Optimization

According to Release 6 of 3GPP, the single-resolution scheme corresponds to the transmission of QPSK with more than 95% coverage. The assignment of the fraction of the total transmission power reserved for MBMS has implications in the coverage and average throughput of the multiresolution based on the hierarchical 16-QAM scheme. The multicell interference distribution has also strong impact in the coverage and throughput. An interesting design parameter is the channel bit rate (and its coding rate) associated to the multiresolution scheme. An optimization of this parameter has also strong impact in achievable coverage and average throughputs. Regardless of the channel conditions and user location, a given user always attempts to demodulate both the base layer and the enhancement layer carrying additional resolution. For good multiresolution design, the basic information will be always correctly received independently of the position of any user within the 95% coverage target. However, depending on its position inside the cell, more or fewer blocks with additional resolution will be correctly received by the mobile user. The objective of this chapter is to design multiresolution schemes in the two different scenariosMBSFN and SC-PMP with inter-cell interference without and with macrodiversity supportand to measure the corresponding multiresolution gain of total throughput compared to the reference total throughput of the single resolution scheme based on the QPSK transmission.

16.3 Evaluation Methodology and Simulation Assumptions

Typically, radio network simulations can be classified as either link level (radio link between the base station and the user terminal) or system level (several base stations with large number of mobile users). A single approach would be preferable, but the complexity of such a simulator (including everything from transmitted waveforms to multicell network) is far too high for the required simulation resolutions and simulation time. Therefore, separate but interconnected link and system level approaches are needed. The link level simulator is needed for the system simulator to build a receiver model that can predict the receiver block error rate/bit error rate (BLER/BER) performance, taking into account channel estimation, interleaving, modulation, receiver structure, and decoding. The system level simulator is needed to model a system with a large number of mobiles and base stations and also to evaluate algorithms operating in such a system. As the simulation is divided in two parts, an approach for linking the two simulators must be defined. Conventionally, the information obtained from the link level simulator is inserted into the system level simulator through the utilization of a specific performance parameter (BLER) corresponding to a specific SNR

LTE E-MBMS Capacity and Intersite Gains

593

estimated in the terminal or base station. Figure 16.2 shows the interaction between the simulators.

16.3.1 Link-Level Simulator Design

The link level simulator (LLS) was developed in Matlab and took into account the specifications of 3GPP MBMS Release 7 [17] regarding the signal processing of transport and physical channels. It satisfied two essential requirements: Serve as reference for all the link level simulations with multiresolution and parameters estimation Serve as a platform to the different multiresolution improvements tested and quantified A typical time interval of each link level simulation is 0.5 s (as shown in Table 16.1). The entire OFDMA signal processing at the transmitter was included in the LLS as well as in several different receiver structures. To achieve reliable channel estimation and data detection, a receiver capable of jointly performing these tasks through iterative processing is used. The structure of the iterative receiver is shown in Figure 16.3 (see also [18]). Clearly, the receiver structure for additive white Gaussian noise (AWGN) channel is less complex (only a few turbo-decoder iterations, and no channel estimation or channel equalization is required). Multipath Rayleigh fading channels were considered in the simulator, as it comprises a more realistic scenario for evaluating hierarchical high-order QAM modulations due to their sensitivity to the channel parameters estimation. As indicated, the receiver structure is nonlinear, iterative, and includes channel parameters estimation for the analyzed multipath Rayleigh fading channels [19].
Simulation parameters System level simulator SNR SNR BLER

Results

Link level simulator BLER

BLER

SNR

Figure 16.2 Interaction between link level simulator and system level simulator.

594

Evolved Cellular Network Planning and Optimization

Table 16.1 Link and System Level Simulation Parameters for an Urban Macrocellular Scenario
Transmission bandwidth Cyclic prefix size FFT Size Carriers space (kHz) Available bandwidth Sample time (ns) Max Tx power (dBm)/sector Number of used subcarriers/sector Number of used subcarriers/cell Freq. reuse Subframe duration (ms) Interfering cells transmit with % of max power Cellular layout Sectors Number of cell sites Antenna gain of the base station Width of beam of the antenna at 3 dB Front/back ratio of the antenna Antenna pattern radiation of the base station Propagation model Downlink thermal noise Cable loss Fadeout standard deviation due to shadowing 10 MHz 72 1024 15 9 MHz 130 46 200 600 1/3 0.5 90 Hexagonal three sectors/cell 19 17.5 dBi 70 degrees 20 dB Gaussian Okumura-Hata 100 dBm 3 dB 10 dB

LTE E-MBMS Capacity and Intersite Gains

595

Deinterleaver Rk,l DFT Channel equalization Demodulator

Channel decoder

log2 M parallel chains 2 Deinterleaver Channel decoder

^ (q) Hk,l

^ Sk,l Channel estimator Transmitted signal rebuilder

Decision device

Decision device

Figure 16.3 Iterative receiver structure.

16.3.2 Radio Access Network System Level Simulator

For the purpose of validating the work presented in this section, a system level simulator was developed in Java, using a discrete event-based philosophy, which captures the dynamic behavior of the RAN system. This dynamic behavior includes the user (e.g., mobility and variable traffic demands), radio interface, and RAN with some level of abstraction. The system level simulator (SLS) works at a subframe duration rate; a typical time interval of each simulation is 600 s. Table 16.1 shows the simulation parameters. It presents the parameters used in the link and system level simulations based on 3GPP documents [2023]. The channel model used in the system level simulator considers three types of losses: Distance loss Shadowing loss Multipath fading loss (one value per 0.5 ms) Model parameters depend on the environment. For distance loss, the OkumuraHata model from the COST 231 project was used (see [24]). Shadowing is due to the existence of large obstacles like buildings and the movement of UEs in and out of the shadows. This is modeled through a process with a log-normal distribution and a correlation distance. The multipath fading employed in the system level simulator corresponds to the 3GPP channel models, where the ITU vehicular A (see [21] Annex B) and the MBSFN environments were chosen as references. The latter models

596

Evolved Cellular Network Planning and Optimization

were also used in the link level simulator but at a much higher rate. Vehicular A (with velocity v = 30 km/h) channel model was chosen because it is an important test channel in 3GPP specifications, furthermore, it allows direct comparison against previous system level simulations [16]. In OFDM systems, an important parameter is the maximum delay of the multipath profile and its relation with the duration of the time guard between OFDM symbols to avoid intersymbol interference. 3GPP has specified a short time guard with about 4.75 s and a long time guard with 16.67 s. The latter was considered in the model for achieving the results next presented, making the performance less sensitive to the chosen propagation channel. However, there is a reduction of the transmitted bit rates. A uniform distribution of mobile users is generated at the beginning of each simulation. A typical number of users chosen for each simulation run was 20 per sector. Each mobile has random mobility with the specified speed of 30 km/h. Dynamic system level simulators like the one presented in this chapter are very accurate; the main limitation is the hypothetical urban macrocellular test scenario that is different from any real one. Figure 16.4 illustrates the cellular layout (trisectorial antenna pattern) indicating the fractional frequency reuse of 1/3 considered in the system level simulations.

2 1 3 2 1 3 2 1 3 2 1 1 3 2 3 2 1 1 3

Figure 16.4 Cellular layout including the frequency reuse of 1/3 (gray shadows/ numbers of the cells).

LTE E-MBMS Capacity and Intersite Gains

597

One third of the available bandwidth was used in each sector to reduce multicell interference. As indicated in Figure 16.4, the identification of the sources of multicell interferencethe use of the same adjacent subcarriers (named physical resource blocks)is given by the sectors with the same gray shadow/number: dark gray/one, average gray/two, or light gray/three. For 16-QAM hierarchical constellations, two classes of bits with different error protection are used. The darkest gray around the antennas only indicates the approximate coverage of the weak bits blocks, whereas the other gray shadows indicate the coverage of the strong bits blocks. In the analysis of the single-cell point-to-multipoint scenario (SC-PMP), there is one radio link between the mobile and the closest base station. It does not assume any time synchronism between the transmissions from different base stations with the same gray shadow resulting in interference from all cells without the same gray shadow. However, in the SC-PMP scenario with macrodiversity combining the two best radio links, it is assumed that there is time synchronization between the two closest base station sites with the same gray shadow. In this case, multicell interference is reduced because only the other base station sites with the same gray shadow remain asynchronous and capable of interference. In the MBSFN scenario, there are at least three radio links between the mobile and the three closest base stations. Time synchronisation is assumed between the transmissions from the closest base stations with the same gray shadow, resulting in much less interference from the cellular environment. This results in macrodiversity combining of the three best radio links. In addition, the interfering base stations must be at least 5 km away from the reference base station considering a cyclic prefix (CP) of 16.67 s and a frequency of 2 GHz. Only distant base station sites are capable of introducing interference.

16.4 System Level Performance Results

To study the behavior of the proposed OFDM multiresolution schemes, several simulations were performed for 16-QAM hierarchical modulations. The 16-QAM hierarchical constellations are constructed using a main QPSK constellation where each symbol is, in fact, another QPSK constellation, shown in Figure 16.5. The main parameter for defining one of these constellations is the ratio between d1 and d2 as shown in Figure 16.5: d1 =k d2 (16.1)

where 0 < k 0.5. For 16-QAM, two classes of bits with different error protection were used (for 64QAM, three classes are used). Each information stream was encoded with a different

598

Evolved Cellular Network Planning and Optimization

I 01 00 01

0101 00

0100

d1 0001 0000

0111

0110

0011

0010

Q 11 10

+
11 10

=
1101 1100 1001 1111 1110 1011

1000 Q

1010

Basic

Enhancement

d2

Figure 16.5 Signal constellation for 16-QAM hierarchical modulation.

block size, depending on the coding rate. Two different coding rates were considered: 1/2 and 3/4. This leads to block sizes of each information stream of 2400 and 3600 bits, respectively. This occupies a subframe 0.5 ms long. The total corresponding transmitted information bit rates per cell sector are 4800 kbps and 7200 kbps, respectively. A reference for measuring the capacity is the number of transmitted TV channels, each with a bit rate of 256 kbps. For the chosen fractional frequency reuse of 1/3, we get 18.75 and 28.125 TV channels, respectively. We want to compare the OFDM/OFDMA results directly with those obtained previously with the WCDMA technology [17]. All the parameters used for OFDM during these simulations were based on 3GPP documents [2023]. About one-third of the total physical resource blocks (PRB) are transmitted in each sector. This corresponds to an instantly occupied bandwidth of 3.0 MHz, where we have considered an average of 16.67 PRBs per sector, each with 180 kHz of adjacent bandwidth (corresponding to 12 subcarriers with a frequency spacing of 15 kHz). The number of adjacent subcarriers in each PRB was chosen according to 3GPP specifications. With the dynamic allocation of the resources per sector, sectors 2 and 3 have 17 PRBs, and sector 1 has 16 PRBs in the first subframe duration. Sectors 1 and 3 have 17 PRBs, and sector 2 has 16 PRBs in the second subframe duration. Finally, sectors 1 and 2 have 17 PRBs, and sector 3 has 16 PRBs in the third subframe duration. On average, there are 16.67 PRBs per sector. We can conclude that the transmission of each TV channel with LTE technology requires less than one PRB for any analyzed coding rate.

16.4.1 BLER Results

In link level simulations, we have evaluated the hierarchical 16-QAM and 64-QAM with two different coding rates. In Figures 16.6 and 16.7 we consider the vehicular A propagation channel to be used in the SC-PMP scenario, and we present BLER versus E s / No for the hierarchical 16-QAM and 64-QAM, respectively. In the legend,

LTE E-MBMS Capacity and Intersite Gains

599

100 QPSK, rate = 1/2 H1, rate = 1/2 H2, rate = 1/2 H1, rate = 3/4 H2, rate = 3/4 101 BLER 102 103 0

10

15

20 25 Es/N0 (dB)

30

35

40

Figure 16.6 BLER vs. E s / N0 for hierarchical 16-QAM, VehA 30 km/h.

100

101 BLER

102

103

H1, rate = 1/2 H2, rate = 1/2 H3, rate = 1/2 H1, rate = 3/4 H2, rate = 3/4 H3, rate = 3/4 0 5 10 15 20 25 Es/N0 (dB) 30 35 40

Figure 16.7 BLER vs. E s / N0 for hierarchical 64-QAM, VehA 30 km/h.

600

Evolved Cellular Network Planning and Optimization

for 16-QAM, H1 means strong bit blocks and H2 weak bit blocks. For 64-QAM, H1 means strong bit blocks, H2 medium bit blocks, and H3 weak bit blocks. We have also included for comparison the QPSK performance for coding rate 1/2. We conclude that QPSK has the lowest E s / N0 , consequently it will have the highest coverage. As expected, coding rate 1/2 provides a smaller E s / N0 compared to rate = 3/4, resulting in higher coverage. However, coding rate = 3/4 provides bit rates that are 1.5 higher than rate 1/2. There is a trade-off between bit rate (or throughput) and coverage. We will consider later on macrodiversity combining to increase the coverage (and throughput) at the cell borders. When we compare Figures 16.6 and 16.7, we observe that due to the higher bit rates offered by 64-QAM, exactly 1.5 times the bit rates of 16-QAM, the corresponding E s / N0 are higher than 16-QAM, resulting in less coverage for 64-QAM. It seems that 64-QAM with coding rate 1/2 has a small E s / N0 advantage compared to 16QAM rate 3/4 (both provide the same maximum bit rates). However, the sensitivity of 64-QAM to channel estimation errors is a feature that should not be forgotten, especially for hierarchical 64-QAM. The introduction of macrodiversity combining will increase the coverage of 16-QAM H2 blocks and H3 blocks (weak bits). Figure 16.8 shows the BLER versus E s / N0 for the hierarchical 16-QAM in the MBSFN scenario with the MBSFN propagation channel. Comparison between Figures 16.6 and 16.8 indicate that the MBSFN channel due to the longer multipath power delay profile provides higher multipath diversity. This can be confirmed by a clear increase of the BLER performance of both coding rates compared to VehA. There is also a higher inherent intersymbol interference in the MBSFN channel, which is evident for rate 3/4 and weak bit blocks (H2). In spite of this, there is no significant loss in the BLER performance for rate 3/4 due to the redundancy of the channel coding. Recall that as the MBSFN scenario has lower inter-cell interference,
100
H1, 1/2 H2, 1/2 H1, 3/4 H2, 3/4

101
BLER

102

103

10

15 20 Es/N0 (dB)

25

30

35

Figure 16.8 BLER vs. E s / N0 for hierarchical 16-QAM, MBSFN 30 km/h.

LTE E-MBMS Capacity and Intersite Gains

601

it is thus expected to compensate the lower bit rates resulting from the use of a longer guard time to avoid the effects of intersymbol interference.

16.4.2 Coverage Results

In system level simulations, mobile users receive strong and weak bits blocks transmitted from base stations. Each block undergoes small and large scale fading and multicell interference. In terms of coverage or throughput, the SNR of each block is computed taking into account all the above impairments. Based on the comparison between the reference SNR at a BLER of 1% and the evaluated SNR, it is decided whether the block is or not correctly received. This is done for all the transmitted blocks for all users in all sectors of the 19 cells, during typically 10 min. Figure 16.9 presents the coverage versus the fraction of the total transmitted power (denoted as E c / Ior ), for SC-PMP scenario where there is interference only from one-third of the sectors due to the frequency reuse of 1/3 (see Figure 16.4). All interfering sites transmit with a maximum power of 90% according to the parameters indicated in Table 16.1. The cell radius is 750 m or 1500 m, and strong blocks (H1) are separated from weak blocks (H2) without including macrodiversity combining, denoted as 1RL, and also with macrodiversity combining the two best radio links (2RL). Recall that the basic scenario SC-PMP does not include macrodiversity. Multicell interference is 90% of the maximum transmitted power in each site. Previous coding rates were consideredrates 1/2 and 3/4, respectively. Recall that it is necessary to ensure the coverage of 95% for strong bit blocks (H1). The only case that never reaches the required coverage is when the cell radius has 1500 m
100 90 80 Average coverage (%) 70 60 50 40 30 20 10 0 0 10 20 2RL, H1, 1/2 (R = 1500 m) 2RL, H2, 1/2 (R = 1500 m) 1RL, H1, 1/2 (R = 750 m) 1RL, H2, 1/2 (R = 750 m) 2RL, H1, 3/4 (R = 1500 m) 2RL, H2, 3/4 (R = 1500 m) 30 40 50 60 70 Multicast channel Ec/Ior (%) 80 90 100

Figure 16.9 Average coverage (%) vs. E c/ Ior , of SC-PMP.

602

Evolved Cellular Network Planning and Optimization

and the coding rate is 3/4, even with a macrodiversity of 2RL. For coding rate 1/2, it is equivalent to have 1RL and cell radius of 750 m or 2RL and cell radius of 1500 m. For E c / Ior = 50% and rate 1/2, the coverage of H1 is 98%, and H2 coverage is around 89%. For the same E c / Ior , but with rate 3/4, the coverage values of H1 and H2 are 77% and 57%, respectively. According to the coverage results of this figure, we can conclude that with a coding rate of 1/2, we can increase the cell radius from 750 m to 1500 m as long as we include macrodiversity combining of the two best radio links. However, for coding rate 3/4, we cannot double the cell radius and ensure the designed coverage, even if we add a macrodiversity of 2RL. We should choose between increasing the number of TV channels (or the TV channel bit rates) or increasing the cell radius. In Figure 16.10 the coverage performance curves for MBSFN scenario, versus E c / Ior , are presented for both cell radii of 750 m and 1500 m and should be compared to the corresponding results of Figure 16.9 for the SC-PMP scenario. As expected, there is a difference in the coverage between the two scenarios where MBSFN takes advantage of its lower inter-cell interference. The coverage values are above 95% even for small values of E c / N0 , such as 25%; the only exception are the weak bit blocks (H2) of coding rate 3/4 and cell radius R = 1500 m. There is a coverage similarity between rate 1/2 with R = 1500 m and rate 3/4 with R = 750 m. This means that we can opt between increasing the coding rate (the average throughput) or increasing the coverage. When we increase both coverage

100 90 80
Average coverage (%)

70 60 50 40 30 20 10 0 0 10 20
H1, 3/4 (R = 750 m) H2, 3/4 (R = 750 m) H1, 3/4 (R =1500 m) H2, 3/4 (R = 1500 m) H1, 1/2 (R = 1500 m) H2, 1/2 (R = 1500 m)

30 40 50 60 70 Multicast channel Ec/Ior (%)

80

90

100

Figure 16.10 Average coverage (%) vs. E c/ Ior , of MBSFN.

LTE E-MBMS Capacity and Intersite Gains

603

and throughput at the same time, we observe a decrease of coverage, in particular, H2 blocks.

16.4.3 Throughput Results

Figure 16.11 presents the average throughput distribution as a function of the E c / Ior for the SC-PMP scenario network with and without macrodiversity for both cell radii of 750 m and 1500 m. We observe a considerable gain in throughput when macrodiversity (2RL) is considered compared to the single radio link case. This is particularly true for the high coding rate of 3/4. For E c / Ior above 50%, the average throughput for coding rate 3/4 is above 256 kbps, which is the maximum throughput for rate 1/2. However, not all UEs are able to achieve such high throughput, as users located at the cell borders never reach such high values of throughput. Figure 16.12 considers the throughput distribution as function of the distance between UEs and BS for the E c / Ior = 90%, with and without macrodiversity for the same cell radius of 1500 m and different coding rates. For the chosen E c / Ior , both 2RL and 1RL ensure the maximum throughput for users located near the base station. As the distance between UEs and BS increases the throughput of 1RL decreases significantly. However, the decrease in throughput is more obvious for rate 3/4 and when mobile users are at the cell bofrders. It is observed that with 2RL, only for rate 1/2, is the throughput almost independent of the distance. For the high coding rate of 3/4, a single radio link offers high throughput only for
350 300 Average UE throughput (kbps) 250 200 150 100 50 0 2RL, 1/2 (R = 1500 m) 1RL, 1/2 (R = 1500 m) 2RL, 1/2 (R = 750 m) 1RL, 1/2 (R = 750 m) 2RL, 3/4 (R = 1500 m) 1RL, 3/4 (R = 1500 m) 0 10 20 30 40 50 60 70 Multicast channel Ec/Ior (%) 80 90 100

Figure 16.11 Average UEs throughput vs. E c/ Ior , SC-PMP.

604

Evolved Cellular Network Planning and Optimization

400 350 Average UE throughput (kbps) 300 250 200 150 100 50 0 2RL (rate = 1/2) 1RL (rate = 1/2) 2RL (rate = 3/4) 1RL (rate = 3/4) 0 500 Distance (m) 1000 1500

Figure 16.12 Throughput vs. distance for SC-PMP.

users close to the base station. This figure complements the previous one where the average throughput for the cell area was presented. When we consider that approximately three-fourths of users are located in the area between the cell border and the half radius of the cell, we realize the importance for the average throughput within in this area. For users located beyond 1200 m when there is 2RL, coding rate 1/2 provides higher throughput than rate 3/4. Again, we must choose between increasing the intersite distance (coverage) or increasing the number of TV channels (capacity). Figure 16.13 presents the average throughput distribution as function of the E c / Ior for the MBSFN scenario for both cell radii of 750 m and 1500 m. As expected, we observe that for high coding rate 3/4 the throughput performance is higher for R = 750 m compared to R = 1500 m due to the higher coverage of H2 blocks. However, for the E c / Ior value of 90 (dedicated carrier to MBSFN) there is almost no difference between the two performance curves. The average throughput of 256 kbps, which is the maximum throughput for rate 1/2 is achievable for E c / Ior = 45%. This means that we can have two MBSFN carriers each transmitting at least 18 TV channels. The operator must always choose between increasing the capacity, 2 18 = 36 TV channels, keeping R = 750 m or increase the coverage to R = 1500 m with only 28 TV channels (see Table 16.2). In Figure 16.14, the throughput distribution as function of the distance between UEs and BS is presented for the E c / Ior = 90% and considering both scenarios. For the chosen E c / Ior , both scenarios assure the maximum throughput for users

LTE E-MBMS Capacity and Intersite Gains

605

400 350 Average UE throughput (kbps) 300 250 200 150 100 50 0 3/4 (R = 750 m) 3/4 (R = 1500 m) 1/2 (R = 1500 m) 0 10 20 30 40 50 60 70 Multicast channel Ec/Ior (%) 80 90 100

Figure 16.13 Average UEs throughput vs. E c/ Ior for MBSFN.

located near the base station. As the distance between UEs and BS increases, the throughput of SC (single cell with a macrodiversity of 2RL) decreases significantly. However, it is more obvious that there is more of a decrease in throughput for coding rate 3/4 than for rate 1/2. The SFN assures not only higher coverage than SC (single cell with 2RL) due to lower inter-cell interference but also higher throughput. For coding rate 3/4, the throughput gain of MBSFN compared SC-PMP, where we measure the gain, as the aggregate throughput in all cell areas (weighted by the distribution of users) under the two-throughput performance curves indicates a value close to 1.5 considering the use of the 16-QAM multiresolution scheme. Notice that for broadcasting mobile TV channels, it is also important to increase the intersite

Table 16.2 Capacity Values for MBSFN and SC-PMP Scenarios Using 16-QAM Hierarchical Multiresolution OFDMA (BW = 10 MHz)
QoS 256 kbps 256 kbps 256 kbps 384 kbps #TV Channels 18.75 18.75 28.125 18.75 Spectral Efficiency 0.48 bps/Hz/cell 0.48 bps/Hz/cell 0.72 bps/Hz/cell 0.72 bps/Hz/cell ISD 1500 m 3000 m 3000 m 3000 m Scenario SC-PMP 1RL SC-PMP 2RL MBSFN MBSFN

606

Evolved Cellular Network Planning and Optimization

400 350 Average UE throughput (kbps) 300 250 200 150 100 50 0 SFN, 1/2 (R = 1500 m) SC, 1/2 (R = 1500 m) SFN, 3/4 (R = 1500 m) SC, 3/4 (R = 1500 m) 0 500 1000 Multicast channel Ec/Ior (%) 1500

Figure 16.14 Throughput vs. distance.

distance (ISD) to 3000 m to reduce the number of sites and this is only easily ensured with MBSFN. Recall that as the intersite distance ISD = 2 R when we double R , we double the ISD allowing for a 50% reduction in the number of sites. To obtain the 16-QAM multiresolution gain over the single resolution with QPSK used by MBMS with the WCDMA technology specified in Release 6, the aggregate throughput in all cell areas with multiresolution should be computed and divided by the single resolution aggregate throughput in the cell area. As the coverage of QPSK blocks becomes the same as strong bits blocks of hierarchical 16-QAM due to macrodiversity combining, comparing aggregate throughputs is based on different coverage of the weak bits blocks. It is clear that the smallest throughput gain is achieved for coding rate = 1/2 (256 kbps). For this case, taking into account that the single resolution throughput of QPSK, which is 128 kbps, the throughput gain is 2. The highest throughput gain is achieved for coding rate = 3/4 (384 kbps) and the MBSFN scenario. For this case, the throughput gain is almost 3. However, for the SC-PMP scenario with macrodiversity the throughput gain stays close to 3/1.5 = 2 (see Figure 16.14). In the MBSFN scenario due to the smaller inter-cell interference and substantial macrodiversity combining, in order to achieve higher multiresolution gains, it is suggested that both the channel bit rates from 256 kbps (i.e., the channel coding rate of 1/2) to 384 kbps corresponding to coding rate of 3/4 be increased; in addition, the ISD can also be extended from 1500 m to 3000 m. For the high channel bit rate of 384 kbps, the spectral efficiency achieved per cell sector considering the transmission

LTE E-MBMS Capacity and Intersite Gains

607

Table 16.3 Capacity Values for a QPSK Single-Resolution WCDMA Scheme for Bandwidth BW = 10 MHz
QoS 256 kbps #TV Channels Spectral Efficiency 14 0.358 bps/Hz/cell ISD Scenario

1000 m SC-PMP 2RL (25MHz)

of 18.75 TV channels using all the 10 MHz bandwidth is 0.72 bps/Hz/cell. This value of spectral efficiency is valid for users at the cell border. The ISD associated to this spectral efficiency is 3000 m. Alternatively, 28.125 TV channels with 256 kbps could be transmitted at the same time as indicated in Table 16.2. Table 16.3 shows the capacity of MBMS single resolution taking into account results for the standard MBMS normalized in Release 6 presented in [17] for the same SC-PMP scenario with the macrodiversity of two radio links. The comparison between Tables 16.2 and 16.3 is not straightforward due to the different ISD used. However, it is possible to draw a capacity gain of at least 2 between hierarchical 16-QAM and QPSK (notice that the higher ISD is an advantage for broadcasting). The ISD gain is related to the decrease in the number of sites due to a longer ISD. It has been said in this chapter that for the MBSFN scenario the cell radius can be increased (the ISD is the double the cell radius) from 750 m up to 1500 m without any decrease of average throughput as long as the fraction of the total transmitted power dedicated to E-MBMS reaches 90%. This corresponds to a reduction in the number of sites equal to 50%. An alternative is to use the same carrier for E-MBMS traffic and other unicast types of traffic without increasing the ISD.

16.5 Summary and Conclusions

We have evaluated the use of multiresolutions with hierarchical modulations for the E-MBMS to be standardized in the next release for the LTE-advanced. In this chapter, link and system level simulations of LTE multicellular networks considering broadcast/multicast transmissions using OFDM/OFDMAbased LTE technology were presented. The evaluation of the capacity, in terms of number of TV channels with given bit rates or total spectral efficiency and coverage, was presented. Taking the 95% coverage as reference the evaluation of the achievable capacity gain (number of transmitted mobile TV channels for WCDMA and LTE) is next presented. The intersite distance gain is also presented allowing for a substantial reduction in the number of cell sites when LTE will replace WCDMA. Scenarios based on the use of SFNs with the MBSFN channel and the SC-PMP network with the vehicular A channel were both evaluated for 16-QAM/64-QAM hierarchical modulations and compared with the present WCDMA-based MBMS network. In general, it can be stated that multiresolution is suitable for any of the

608

Evolved Cellular Network Planning and Optimization

analyzed scenarios MBSFN and SC-PMP. Indeed, it works fine in any single-cell scenario without macrodiversity combining or in multicells with macrodiversity. In the SC-PMP scenario without macrodiversity (1RL), due to multiresolution, the channel bit rate of each TV channel (compared to single resolution provided by QPSK) for users can be increased close to the base station for the ISD equal to 1500 m. When macrodiversity (2RL) of the two best radio links is added, the multiresolution schemes become less sensitive to the used channel bit rates and it is possible to increase the channel coding rate, keeping the same ISD or to increase the ISD to 3000 m, keeping the channel bit rate. The operator must choose between the trade-off of increasing capacity or coverage (see Table 16.2); it is not possible to increase both at the same time.

16.6 Open Issues

Evolved MBMS is not a closed issue in the standardization carried out by 3GPP. It is not finalized if the hierarchical constellations already used by the DVB and MediaFLO standards will be chosen for LTE-advanced in the next release of E-MBMS. The combined use of 64-QAM hierarchical constellations and MIMO (spatial multiplexing) in the LTE-advanced as an additional flexible multiresolution scheme for the E-MBMS network is a topic that remains to be evaluated. The scenario evaluated here was that MBSFN was based on a regular cellular grid. Real-life scenarios are not so uniform that will result in the reduction of the gain figures presented in this chapter. This is an issue that should be considered and evaluated in the near future when LTE deployment is carried out.

References
[1] 3GPP TR 25.905, version 7.2.0, Release 7 Feasibility study on improvement of the multimedia broadcast multicast service (MBMS). http://www.3gpp.org, January 2008. [2] H. Sari, Y. Levy, and G. Karam, An analysis of orthogonal frequency-division multiple access, IEEE GLOBECOM97, November 1997. [3] I. Koffman and V. Roman, Broadband wireless access solutions based on OFDM access in IEEE 802.16, IEEE Communications Magazine, vol. 40, no. 4, pp. 96103, April 2002. [4] J. A. C. Bingham, Multicarrier modulation for data transmission: An idea whose time has come, IEEE Communications Magazine, vol. 28, no. 5, pp. 514, May 1990. [5] T. Cover, Broadcast channels, IEEE Transactions on Informational Theory, vol. IT-18, pp. 214, January 1972. [6] K. Ramchandran, A. Ortega, K. M. Uz, and M. Vetterli, Multi-resolution broadcast for digital HDTV using joint source/channel coding, IEEE Journal on Selected Areas in Communication, vol. 11, January 1993. [7] H. Jiang and P. A. Wilford, A hierarchical modulation for upgrading digital, IEEE Transactions on Broadcasting, vol. 51, no. 2, pp. 223229, June 2005.

LTE E-MBMS Capacity and Intersite Gains

609

[8] S. Wang, S. Kwon, and B. K. Yi, On enhancing hierarchical modulation, Proceedings of the IEEE International Symposium on Broadband Multimedia Systems and BroadcastingBTS, Las Vegas, NV, March 31April 2, 2008. [9] P. K. Vitthaladevuni and M.-S. Alouini, A closed-form expression for the exact BER of generalized PAM and QAM constellations, IEEE Transactions on Communications, vol. 52, pp. 698700, May 2004. [10] N. Souto, F. Cercas, R. Dinis, and J. C. Silva, On the BER performance of hierarchical MQAM constellations with diversity and imperfect channel estimation, IEEE Transactions on Communications, vol. 55, no. 10, pp. 18521856, October 2007. [11] G. Foschini, Layered-space-time architecture for wireless communication in a fading environment when using multi-element antennas, Bell Labs Technical Journal, pp. 4159, Autumn 1996. [12] G. Foschini and M. Gans, On limits of wireless communications in fading environments when using multiple antennas, Wireless Personal Communications Journal, vol. 6, pp. 315 335, March 1998. [13] A. Soares, N. Souto, J. Silva, P. Eusbio, and A. Correia, Effective radio resource management for MBMS in UMTS networks, Wireless Personal Communications Journal, vol. 42, no. 2, pp. 185211, July 2007. [14] A. Soares, J. Silva, F. Leito, A. Correia, and N. Souto, MIMO based radio resource management for UMTS multicast broadcast multimedia services, Wireless Personal Communications Journal, vol. 42, Issue 2, pp. 225246, July 2007. [15] A. Correia, N. Souto, J. Silva, and A. Soares, Chapter 17, Air interface enhancements for MBMS, Handbook on Mobile Broadcasting, Borko Furht and Syed Ahson, eds., CRC Press, Taylor and Francis, New York, 2008. [16] A. Correia, J. Silva, N. Souto, L. Silva, A. Boal, and A. Soares, Multi-resolution broadcast/multicast systems for MBMS, IEEE Transactions on Broadcasting, vol. 53, no. 1, pp. 224234, March 2007. [17] 3GPPP TR 25.814, version 7.1.0, Release 7, Technical specification group radio Access Network; Physical Layers Aspects for Evolved (UTRA), September 2006. http://www.3gpp.org [18] N. Souto, A. Correia, R. Dinis, J. Silva, and L. Abreu, Multiresolution MBMS Transmissions for MIMO UTRA LTE systems, Proceedings of the IEEE International Symposium on Broadband Multimedia Systems and BroadcastingBTS, Las Vegas, NV, March 31April 2, 2008. [19] 3GPP, 25.101, version 6.2.0, Release 6, User equipment radio transmission and reception (FDD), June 2006, http://www.3gpp.org [20] 3GPP TR 25.912, version 7.1.0, Release 7, Feasibility study for evolved universal terrestrial radio access (UTRA) and universal terrestrial radio access network (UTRAN), http://www.3gpp.org [21] 3GPP TR 25.892, version 6.0.0, Release 6, Feasibility study for orthogonal frequency division multiplexing (OFDM) for UTRAN enhancement, September 2006, http://www.3gpp.org [22] 3GPP TR 36.942, Release 8, Evolved universal terrestrial radio access (E-UTRA); radio frequency (RF) system scenarios, December 2008, http://www.3gpp.org [23] 3GPP, R1-070674, LTE physical layer framework for performance verification, 3GPP TSG-RAN1#48, February 2007, http://www.3gpp.org [24] E. Damosso, European Commission COST 231, Digital Mobile Radio towards Future Generation Systems, European Commission, 1999, Brussels.