Overview of Wireless Channel Models For UMTS and LTE: Abbas Mohammed and Asad Mehmood

Chapter 2
Overview of Wireless Channel Models for UMTS and LTE

Abbas Mohammed and Asad Mehmood Contents
2.1 2.2 2.3 Introduction ................................................................................... 44 Multipath Propagation Channels ....................................................... 45 Modeling Methods for Multipath Channel Modeling ........................... 47 2.3.1 Deterministic Channel Computation ...................................... 47 2.3.2 Geometry-based Stochastic Channel Models ............................ 48 2.3.3 Non-geometrical Stochastic Channel Models ........................... 49 2.3.4 Correlation-based Channel Models ......................................... 49 Standard Channel Models ................................................................. 50 2.4.1 COST Channel Models ........................................................ 50 2.4.1.1 COST 259 Directional Channel Model .................... 51 2.4.1.2 COST 273 Channel Model ..................................... 52 2.4.2 ITU Channel Models ............................................................ 53 2.4.3 Extended ITU Models .......................................................... 56 2.4.4 3GPP Channel Model ........................................................... 58 MIMO Channel Models ................................................................... 58 2.5.1 3GPP Spatial Channel Model ................................................ 60 2.5.2 Extended 3GPP Spatial Channel Model .................................. 62 2.5.3 WINNER Channel Model .................................................... 65
2.4
2.5
43
44
Evolved Cellular Network Planning and Optimization
Multi-polarized MIMO Channel Models ................................ 66 2.5.4.1 3GPP Polarized Spatial Channel Model ..................... 68 2.5.4.2 Theoretical Polarized MIMO Channel Model ............ 69 2.5.5 MIMO Channel Model for LTE Evolution ............................. 70 2.5.6 Comparison of SCM, SCME, WINNER, and LTE Evolution Models ................................................................. 73 2.6 Channel Modeling Role in Cell Planning and Optimization ................. 74 2.7 Conclusion ..................................................................................... 75 References ................................................................................................ 76 2.5.4
2.1 Introduction
Designing, analyzing, and deploying communication systems requires the efficient utilization of available resources for reliable transfer of information between two parties. However, in practical systems some amount of unpredictability is tolerated in order to achieve better consumption of available resources. Thus the performance of signal processing algorithms, transceiver designs, etc., for a communication system are highly dependent on the propagation environment. A correct knowledge and modeling of the propagation channels is a central prerequisite for the analysis and design of the long-term evolution (LTE) at both the link level and the system level and also for the LTE specifications for the mobile terminal and the base station performance requirements, radio resource managements to ensure that the resources are used in an efficient way, and in RF system scenarios to derive the requirements and in system concept evolution. The use of multiple transmit/receive antenna techniques is an important feature of LTE. Multiple antenna techniques used in LTE (e.g., in spatial diversity), take the advantage of multipath dispersion to increase the capacity. However this requires that the spatial correlation between antenna elements should be low, which is difficult to obtain in practical systems. When assessing multiple antenna techniques, it is important that relevant features (e.g., spatial correlation) of the channel are modeled in an efficient way. Therefore, standard MIMO channel models also have great significance in the design and analysis of an LTE system. This chapter gives an overview of standard channel models for Universal Mobile Telecommunication Systems (UMTS) and the upcoming LTE. The emphasis is on some general channel models used during the evolution of UMTS and LTE, and specific channel models for LTE as well. The chapter is organized as follows: Section 2.2 describes the basics of multipath channel modeling. In Section 2.3, different approaches for developing generic channel models are discussed, which are used to build standard channel models for LTE. Section 2.4 describes standard channel models for UMTS and LTE and Section 2.5 gives an overview of recently developed MIMO channel models for LTE. The chapter ends with concluding remarks.
45
2.2 Multipath Propagation Channels

The difficulties in modeling a wireless channel are due to the complex and varied propagation environments. A transmitted signal arrives at the receiver through different propagation mechanisms shown in Figure 2.1: the line-of-sight (LOS) or free space propagation, scattering, or dispersion due to contact with objects with irregular surfaces or shapes, diffraction due to bending of signals around obstacles, reflection by the objects with sizes that are large compared to the wavelength of the electromagnetic wave, transmission through irregular objects, and so on. The presence of multipaths (multiple scattered paths) with different delays and attenuations gives rise to highly complex multipath propagation channels. It is significant here to note that the level of information about the environment a channel must provide is highly dependent on the category of communication system under assessment. To predict the performance of single-sensor narrowband receivers, classical channel models, which provide information about signal power level distributions and Doppler shifts of the received signals, may be satisfactory [1, 2]. The advanced technologies (e.g., LTE) built on the typical understanding of Doppler spread and fading; however, also incorporate new concepts such as time delay spread, direction of departures, direction of arrivals and adaptive array antenna geometry, and multiple antenna implementations at both sides of the wireless link, or multiple-input multiple-output (MIMO) [39].
Reec
Sca tte ri
LO
tion
ng
pa th
tion
Di
rac
Figure 2.1 Radio propagation environment. [From 3GPP Technical Specication 25.996, Spatial Channel Model for Multiple Input Multiple Output (MIMO) Simulations (Release 8),V8.0.0, 12-2008. http://www.3gpp.org.] 2008. 3GPPTM .
46
The degradation in the received signal level due to multipath effects can be classified into large-scale path loss components, medium-scale slow varying components with log-normal distribution, and small-scale fast fading components with Rayleigh or Rician distribution depending on the absence or presence of LOS components between the transmitter and receiver [2, 10]. Thus, a three-stage propagation model can be used to describe a wireless cellular environment. The three stages are: 1. Large-scale propagation model: This model is used to characterize the received signal strength by averaging the amplitude or power level of the received signal over large transmitterreceiver separation distances in the range of hundredths or thousandths of a wavelength. The large-scale models are often derived from measured data. However, semi-empirical models are employed in smaller areas to achieve higher accuracy. For this purpose, theoretical models are used, which are then fitted to measured data to obtain a desired model for a particular propagation scenario. 2. Medium-scale propagation model: This model determines the gradual variations of the local mean amplitude or the local mean power of the received signal over a time-variant multipath channel when the mobile station moves over distances larger than a few tens or hundreds of a wavelength. Some existing components will disappear while new components will appear. It is observed that variations of the local mean power of the received signal follow log-normal distribution, which is called slow fading or shadowing. The shadowing is caused by obstructions like trees and foliage. The mean and standard deviations of the received power are determined from large-scale propagation models in the environment of interest. 3. Small-scale propagation model: This model is used to characterize the rapid variations of the received signal strength due to changes in phases when a mobile terminal moves over small distances on the order of a few wavelengths or over short time durations on the order of seconds. Since the mean power remains constant over these small distances, small-scale fading can be considered as superimposed on large-scale fading for large-scale models. The most common description of small-scale fading is by means of the Rayleigh distribution. Multipath signals arrive at the receiver with different propagation path lengths, called multipath taps, and different time delays. The multipath signals with different phases sum constructively or destructively at the receiver, giving rise to time varying multipath taps. The power distribution of channel taps is described by a distribution function depending on the propagation environment. The most severe multipath channel is the Rayleigh fading channel in which there is no line-of-sight path and the channel taps are independent. In the case of the Rician fading channel, the fading dips are low due to the presence of line-of-sight components in addition to the dispersed paths.
47
The behavior of a multipath channel needs to be characterized in order to model the channel. The concepts of Doppler spread, coherence time, and delay spread and coherence bandwidth are used to describe various aspects of the multipath channel. The maximum value of delay spread gives the delay difference between the first and the last channel tap in the power delay profile. The coherence bandwidth is the inverse of the delay spread. If the coherence bandwidth is greater than the transmission bandwidth, then the frequency components of the signal will undergo frequency flat fading. A frequency selective fading results if the coherence bandwidth is less than the transmission bandwidth. The Doppler spread arises due to motion of the mobile station and gives a maximum range of Doppler shifts. If there is only one path from the mobile terminal to the base station, then the Doppler spread will be zero with a simple shift of carrier frequency. The inverse of the Doppler spread gives coherence time of the channel during which the channel statistics do not change significantly.
2.3 Modeling Methods for Multipath Channel Modeling

New channel models can be developed in two stages [5]. First, setting up a framework for a generic channel model and identifying a set of parameters that needs to be determined for the description of the channel. Second, conducting measurement campaigns and extracting numerical values of parameters and their statistical distributions. In the first stage, different methods can be employed (e.g., the deterministic methods are based on measured impulse responses and ray tracing algorithms); the stochastic approaches include tapped delay line models, frequency domain modeling, and geometry-based stochastic modeling. In this section, a brief overview of different modeling methodologies for the first stage is described.
2.3.1 Deterministic Channel Computation

Deterministic propagation modeling is intended to reproduce the actual radio propagation process for a given environment. These methods are suitable for environments where the radio waves interact with fairly simple geometric obstacles such as streets and buildings. The geometric and electromagnetic characteristics of the corresponding environment and of the radio links can be stored in files (environment data) and the corresponding propagation process can be modeled through analytical formulas or computer programs. Deterministic channel models are potentially accurate and meaningful. Due to the high accuracy and adherence to the real propagation process, deterministic models can be advantageous to use in situations when time is not sufficient to conduct measurements or some specific cases that are difficult to measure in the real world. Although electromagnetic models such as finite difference in time domain (FDTD) and methods of moments (MoM) can be helpful in the study of near field problems, these models are not used for field prediction at radio frequencies
48
because of small wavelengths with respect to the dimensional scale of the environment. The most appropriate deterministic method for radio propagation, at least in urban area, is the ray-tracing model [11]. In a wireless channel, multipath propagation at higher frequencies can be modeled by applying geometrical optics (GO) theory. This theory is based on ray approximation, which is suitable when the wavelength is significantly small compared to the dimensions of the obstacles. Under this assumption, the electromagnetic waves following multiple paths can be expressed as a set of rays or beams where each ray represents a piecewise linear path connecting two terminals. In the ray-tracing method, the position of the transmitter and the receiver is specified initially and then all the possible rays (paths) between the transmitter and the receiver can be determined by applying geometrical optics rules and geometric considerations. The rays can be characterized from the propagation environment by their amplitude, Doppler shifts, delays, angle of departure, angle of arrival, and polarization. Once the complete information (database) about the environment is known, including the positions of the transmitter and the receiver, then by applying the fundamental laws of electromagnetic waves propagation, channel properties can be derived from the positions of the scatterers. If instead of rays, beams of finite transverse dimension are used, then the resulting model is called beam launching or ray splitting. The beam launching models are suitable for large areas and permit faster field strength prediction. On the other hand, the ray-tracing method is more suitable for point-to-point field prediction and gives accurate results as compared to the beam launching method. High computational burdens and a difficulty to maintain suitable and detailed databases are the main drawbacks of ray-based models.
2.3.2 Geometry-based Stochastic Channel Models

Geometry-based stochastic channel modeling approach also uses the ray-based modeling concept, which is a commonly used approach in directional channel modeling for performance assessment of systems including adaptive antenna arrays. In deterministic geometrical modeling approaches (e.g., ray tracing in previous subsection), the locations of scatterers are prescribed deterministically from a database. On the other hand, in geometric-based stochastic channel models (GSCM) the locations of the scatterers are chosen in a stochastic manner with a certain probability distribution of multipath delays and where the direction of departures and arrivals is determined by the ray-based approach. The distribution of scatterers depends on the environments. In indoor environments, the scatterers are located around both ends of the wireless link. In the case of highly mounted antennas, a scatterers presence only around the mobile terminal is considered. Each scatterer can be characterized by its own direction of arrival, direction of departure and propagation delay using the raybased approach. When the channel parameters of each ray (i.e., complex amplitudes, Doppler frequency, delays, direction of departures and arrivals) are determined, the channel behavior can be characterized in multidimensions.
49
The GSCM approach has some advantages [12, 13]. The approach used by GSCM is more practical, and channel parameters can be obtained through simple geometric considerations. Many effects like delay drifts, direction of arrivals, and small-scale fading by superposition of individual scatterers, are implicitly reproduced. All the information lies in the distribution of the scatterers, which do not make the model complex. The transmitter, receiver and scatterer locations, appearance/disappearance of propagation paths, and shadowing can be determined in a straightforward way. Different versions of GSCM are described in [11].
2.3.3 Non-geometrical Stochastic Channel Models

In non-geometrical stochastic channel models, all parameters such as the location of the scatterers, direction of departures, direction of arrivals and so on, describing the paths between the transmitter and the receiver are determined statistically, without referencing the geometry of the physical environment. There are two nongeometrical stochastic channel models in the literature [11]: the extended SalehValenzuela model and the Zwick model. The Zwick model considers the multipath components (MPCs) individually, while the extended Saleh-Valenzuela model uses the cluster of the MPCs. The details of these models are explained in [10].
2.3.4 Correlation-based Channel Models

Correlation-based channel models are simple and have a low computational complexity compared to geometric-based channel models. These models are used to describe the correlation properties between all transmit/receive antenna pairs and are useful in the study of correlation impacts on any performance parameter of the system. Spatial correlation can be defined explicitly by the spatial correlation matrix. In ray-based modeling, correlation is present in the channel matrix implicitly. For zero mean complex circularly symmetric Gaussian channels, the channel correlation matrix for the description of the MIMO channel behavior is defined in [14]. R full = E [Vec ( H )Vec ( H ) H ] (2.1)
In the preceding equation, Vec () stacks all elements of the matrix H into a large vector, H is the channel matrix for single tap delay, and () H is the Hermitian transpose. The channel correlation matrix R is different for each channel tap. MIMO channel spatial correlation properties are captured by the matrix R at both ends of a wireless link. The correlation matrix R based on the Rayleigh fading channel is defined in [15] as R = Vec 1 ( R 1/2 g ) (2.2)
50
where, g is a circularly symmetric Gaussian vector having zero mean and unit variance, and Vec 1 is the inverse vectorization operation. To simulate the Ricean fading K-factor, the LOS signal is included in the signal. The spatial correlation matrices can be derived from ray-based models, channel matrices based on measurements, or from analytical calculations. The most popular correlation-based model is the Kronecker model, which is computationally simpler than the full correlation matrix R full . This model requires that the correlation matrix at the receiver be independent of the direction of transmission. In this case, channel matrices are obtained [15] using, H = R TX GR RX
1/2 1/2
(2.3)
where, G is the i.i.d. (independent identically distributed) complex Gaussian matrix, R TX and R RX are correlation matrices of the transmitter and the receiver, respectively. The main advantage of correlation-based channel models is that these models are simple and have low computational complexity. On the other hand, these models cannot be generalized simply to other configurations since spatial correlations depend on antenna configurations, and so new correlation coefficients are required for each configuration.
2.4 Standard Channel Models

When designing an LTE system, different requirements are considered (e.g., UE and BS performance requirements, radio resource management requirements, RF system scenarios) to derive the requirements. The standard channel models play a vital role in the assessment of these requirements. In the following section, some standard channel models are discussed that are used in the design and evolution of the UMTS-LTE system.
2.4.1 COST Channel Models

COST stands for the European Co-operation in the Field of Scientific and Technical Research. Several COST efforts were dedicated to the field of wireless communications, especially radio propagation modeling; COST 207 for the development of Second Generation of Mobile Communications (GSM), COST 231 for GSM extension and Third Generation (UMTS) systems, COST 259 Flexible personalized wireless communications (19962000) and COST 273 Toward mobile broadband multimedia networks (20012005). These projects developed channel models based on extensive measurement campaigns, including directional characteristics of radio propagation (COST 259 and COST 273) in macro, micro, and picocells, and are appropriate for simulations with smart antennas and MIMO systems (MIMO
51
models are discussed in Section 2.5). These channel models form the basis of ITU standards for channel models of beyond 3G systems (e.g., LTE).
2.4.1.1 COST 259 Directional Channel Model

COST 259 was a European Research initiative in the field of Flexible Personalized Wireless Communications, which encompassed representatives of the key manufacturers, many network operators, and universities. One of the contributions of COST 259 was to propose set standards to overcome the limitations of channel models developed in the past [16, 17]. The COST 259 directional channel model was originally developed for simulations of systems with multiple antennas either at the base-station or mobile terminal (i.e., MISO systems). This channel model is developed for 13 different types of environments, covering macro-, micro-, and picocells, which are given in Table 2.1 [16]. Each environment is illustrated by a set of external parameters (e.g., radio frequency, MS and BS heights, BS position) and global parameters, which are probability density functions describing a specific environment (e.g., scatterers are characterized by Poisson distribution), defining the propagation characteristics as a whole. A layered approach, which distinguishes between the external (fixed) small-scale effects and the large-scale effects, allows well-organized parameterization. It is categorized into the following subsequent layers: The upper layer describes different propagation scenarios that represent a group of environments with similar propagation characteristics. The middle layer deals with non stationary large-scale effects. These effects include angular spreads, delay spreads, shadowing, and the appearance/ disappearance of far-scattering clusters. These effects are described by their
Table 2.1 Different Types of Propagation Environments for Macro-, Micro-, and Picocells
Macrocell General typical urban (GTU) General rural area (GRA) General bad urban (GBA) General hilly terrain (GHT) Microcell General urban microcell (GUM) General urban bad microcell (GUBM) General open place (GOP) General open place NLOS (GPN) Picocell General office LOS (GOL) General office NLOS (GON) General corridor LOS (GCL) General corridor NLOS (GCN) General factory hall (GFH)
52
corresponding probability density functions with different parameters for different propagation environments. These parameters may be modeled as correlated log-normally distributed random variables. The lower layer deals with small-scale fading effects caused by interference due to rapid fluctuations of amplitudes and phases in multipath components. The statistics for small-scale fading are obtained from large-scale fading effects. In a specific scenario, clusters of scatters are distributed at random fixed places in the coverage area according to a specific probability density function. The clusters of scatterers are characterized by the angular spreads and the RMS delay spreads, which are correlated random variables and are obtained deterministically from the positions of the MS and the BS. The intra-cluster variations are modeled stochastically. Each scatterer is described by a random complex coefficient that follows Gaussian distribution. In spite of its general applicability, COST 259 has some limitations which restrict its applicability. First, the scatterers are assumed to be stationary, so the channel variations originate only due to the MS movement. Secondly, a rich scattering environment is required to describe the envelope of delay attenuation as complex Gaussian, which is the case in this model. However, this assumption is not supported by some environments of the channel model, which is a common assumption for all other channel models.
2.4.1.2 COST 273 Channel Model

Despite efforts made in the previous COST projects, difficulties arose in designing channel models for MIMO systems (MIMO models are discussed in Section 2.5). The 3GPP made efforts to develop a channel model that is used for third-generation cellular systems (UMTS). The description of this channel model is limited to a particular set of environments with specific parameters. This model is specific from a simulation point of view but restricts its general applicability. Thus, COST 273 decided to develop a new channel model for MIMO systems. The channel model can be considered as an extension of the COST 259 directional channel model [18, 19]. The main difference between this model and the COST 259 directional channel is that it uses the same generic channel model for all environments while the COST 259 model uses different models for macro-, micro-, and picocells. In addition, some new environments are defined to incorporate new MIMO applications (e.g., fixed wireless access scenarios and peer-to-peer). Two types of parameters are used to describe each environment. 1. External parameters: These parameters describe the environment and keep fixed values during a simulation run (e.g., carrier frequency, base station and mobile station antenna heights, building heights, antenna orientations, antenna scenarios, path loss models).
53
2. Global parameters: Global or stochastic parameters are a set of probability density functions and a set of statistical moments describing a specific environment (e.g., the number of scatterers is characterized by Poisson distribution). The COST 273 channel model includes three types of scatterer clusters, local clusters around the BS or the MS, single interaction clusters, and twin clusters to model the concept of multiple interactions. A cluster is divided into two representations, one as seen by the BS and the other corresponding to the MS side. The advantage of splitting up a cluster in two is that the angular distributions of energy at the BS and the MS can be modeled independently based on the marginal densities of the angular spectra of corresponding clusters. Each ray radiated from the transmitter is scattered by a scatterer in a cluster and it reaches the receiver after bouncing at the corresponding scatterer of the twin cluster. The twin clusters are linked through the stochastic cluster link delay concept. The link delay guarantees realistic path delays (e.g., obtained from measurements), while the position of the cluster is determined by the angular statistics of the cluster. All scatterers inside a cluster have the same link delay [6]. The mean angles and delays of the clusters are modeled by geometric considerations, and the small-scale fading and intra-cluster spreads can be modeled by either the tapped delay line approach or by the geometrical representation. The total impulse response can be written as the sum of the clusters double directional impulse responses, which is as follows [6]:
BS MT (BS ) PMT (MT ) P (MT ) P (, BS , BS , MT , MT ) = P () PBS (BS ) P (2.4)
In the preceding equation, is the delay, BS , BS , MT , MT are the respective azimuth and elevation spreads at the BS and the MS, respectively. The model assumes that azimuth spreads, elevation spreads (i.e., angular spreads), and delay spreads in a cluster are independent on a per-cluster basis. However, as a whole there can be significant coupling between DoDs and DoAs.
2.4.2 ITU Channel Models

The International Mobile Telecommunications (IMT-2000) was an initiative of the International Telecommunication Union (ITU) for the evolution of European Telecommunications Standardization Institute (ETSI) standards for second-generation mobile systems (GSM) to third-generation Universal Mobile Telecommunications Systems (UMTS). ITU standard channel models were used for the development of the 3G IMT-2000 group of radio access systems [20, 21]. The aim of these channel models is to develop standards that help system designers and network planners with system designs and performance verification. Instead of defining propagation models for all possible environments, ITU proposes a set of
54
test environments in [21] that adequately spans all possible operating environments and types of user mobility. The proposed ITU test environments may not resemble the actual mobiles user operating environments, but they give a very good overview of how a mobile user performs in different operating environments. The complete description of all possible scenarios can be found in [22]. 1. Indoor office test environment: The indoor office test environments, where both the base station and users are located indoors, are characterized by small cells and low transmit powers. Path losses and shadowing effects are due to scattering and the attenuation by floors, walls, and metallic structures such as partitions and filing cabinets. Fading can follow Rayleigh or Rician distribution depending upon the location of the user. Indoor channel models based on ITU recommendations are used for modeling indoor scenarios. The average powers and the relative delays of taps for ITU channel models in indoor scenarios are given in Table 2.2 [21]. 2. Outdoor to indoor and pedestrian test environment: For outdoor to indoor and pedestrian environments, base stations with low antenna heights are situated outdoors, while pedestrian users are to be found inside buildings and residences. Path loss rules of R 2 to R 6 can be applied for different ranges (e.g., LOS on a canyon-like street rule where there is Fresnel zone clearance to the region where there is no longer Fresnel zone clearance [21]). Shadowing, caused by hindrance from trees and foliage, follows log-normal distribution and results in the received signal power variations with standard deviations of 12 dB for indoor and 10 dB for outdoor environments, respectively. The building penetration loss average is 12 dB, with a standard-deviation of 8 dB. ITU recommends that in modeling microcells, the outdoor to indoor and
Table 2.2 Average Powers and Relative Delays for ITU Indoor Ofce Test Environment
Channel A Relative Delay (ns) 0 50 110 170 290 310 Average Power (dB) 0 3 10 18 26 32 Channel B Relative Delay (ns) 0 100 200 300 500 700 Average Power (dB) 0 3.6 7.2 10.8 18 25.2 Doppler Spectrum Classical Classical Classical Classical Classical Classical
Tap No. 1 2 3 4 5 6
55
Table 2.3 Average Powers and Relative Delays for ITU Indoor to Outdoor and Pedestrian Test Environment
Channel A Relative Delay (ns) 0 110 190 410 NA NA Average Power (dB) 0 9.7 19.2 22.8 NA NA Channel B Relative Delay (ns) 0 200 800 1200 2300 3700 Average Power (dB) 0 0.9 4.9 8 7.8 23.9 Doppler Spectrum Classical Classical Classical Classical Classical Classical
Tap No. 1 2 3 4 5 6
pedestrian models are to be used to represent multipath conditions. The average powers and the relative delays for the taps of multipath channels based on ITU recommendations are given in Table 2.3 [21]. 3. Vehicular test environment: This type environment is categorized by large macrocells with higher capacity, limited spectrum, and a large transmit power. A path loss exponent of 4 and log-normal shadow fading with a 10 dB standard deviation are suitable in urban and suburban areas. In rural areas, path loss may be lower than previous, while in mountainous areas, if the BS location is suitably selected to avoid path blockages, a path loss attenuation exponent closer to two may be appropriate. The vehicular models (Table 2.4 [21]) are used to model multipath propagations in macrocells regardless of whether the user is inside the car or not. 4. Mixed test environment: This type of environment takes account of environments [e.g., a vehicular environment (macrocells) and outdoor-to-indoor test environment (microcells) in the same geographical area]. In this test environment, fast-moving terminals are connected to macrocells and slow-moving terminals (pedestrians) are associated with microcells to achieve higher capacity. For example, a dense urban environment may be modeled as consisting of 30% of the pedestrian channel model at a speed of 50 km/h and 70% of pedestrian channel model at a speed of 3 km/h. Likewise, other environments (e.g., suburban or rural environments) may be modeled as percentage mixtures of ITU channel models at various speeds. To assess these propagation environments, reference models for each operating environment have been given both on system level calculations and link level software simulations. The key parameters to describe each propagation model are time delay
56
Table 2.4 Average Powers and Relative Delays for ITU Vehicular Test Environment
Channel A Relative Delay (ns) 0 310 710 1090 1730 2510 Average Power (dB) 0 1 9 10 15 20 Channel B Relative Delay (ns) 0 300 8900 12900 17100 20000 Average Power (dB) 2.5 0 12.8 10 25.2 16 Doppler Spectrum Classical Classical Classical Classical Classical Classical
Tap No. 1 2 3 4 5 6
spread, its structure and statistical variability, overall path loss prediction, including path loss, excess path loss, shadowing, maximum Doppler shifts, and operating radio frequency [22].
2.4.3 Extended ITU Models

The analysis done by ITU-R showed that evolution of 3G systems to future generation networks will require technology changes on large scale while new quality of service (QoS) requirements will require increased transmission bandwidth. Thus, LTE channel models require more bandwidth as compared to UMTS channel models to account for the fact that channel impulses are associated to the delay resolution of the receiver [23]. The LTE channel models developed by 3GPP are based on the existing 3GPP channel models and ITU channel models. The extended ITU models for LTE were given the names Extended Pedestrian A (EPA), Extended Vehicular A (EVA), and Extended Typical Urban (ETU), which do not actually represent urban environments because of small cell sizes with large delay spreads of up to 5 ps. Another proposal to categorize these models in terms of delay spreads are representing these with low delay spread (LD), medium delay spread (MD), and high delay spread (HD), respectively. The low delay spread models are used to represent indoor environments and small cell sizes, while medium delay spread and high delay spread are used to model urban environments with large cell sizes. The high delay spread models are according to the typical urban GSM model [24]. The resulting model parameters, number of taps, RMS delay spread, and maximum excess tap delay are shown in Table 2.5, and RMS delay spreads values for tap delay line models are given in Table 2.6 [24].
57
Table 2.5 Power Delay Proles for Extended ITU Models

EPA Model EVA Model ETU Model
Tap Excess Average Excess Average Excess Average No. Delay (ns) Power (dB) Delay (ns) Power (dB) Delay (ns) Power (dB) 1 2 3 4 5 6 7 8 9 0 30 70 80 110 190 410 0 1 2 3 8 17.2 20.8 0 30 150 310 370 710 1090 1730 2510 0 1.5 1.4 3.6 0.6 9.1 7 12 16.9 0 50 120 200 230 500 1600 2300 5000 1 1 1 0 0 0 3 5 7
The Doppler frequencies for these LTE channel models are defined on a basis similar to what was used for UTERA. Just as the three channel models are classified on the basis of low, medium, and large delay spreads, a similar approach is adopted to define Doppler frequencies (low, medium, and high) for the Doppler environments. The Doppler frequencies for LTE channel models with low, medium, and high Doppler conditions are 5 Hz, 70 Hz, and 900 Hz, respectively [24]. The delay spreads and the Doppler frequencies provide a framework from which possible scenarios for the operating environment can be selected. The following combinations of delay spread and Doppler spread are proposed in [25]: extended pedestrian A 5 Hz, extended vehicular A 5 Hz, extended vehicular A 70 Hz, and extended typical urban 70 Hz.
Table 2.6 Summary of Delay Proles for LTE Channel Models

Model Extended pedestrian A model (EPA) Extended vehicular A model (EPA) Extended typical urban model (ETU) Channel Delay Max. Access Taps Spread (rs) Tap Delay (span) 7 9 9 45 357 991 410 ns 2510 ns 5000 ns
58
The propagation scenarios for LTE with speeds from 120 km/h to 350 km/h are also defined in [25, 26] (e.g., the high-speed train scenario at speed 300 km/h and 350 km/h). The maximum carrier frequency over all frequency bands is f c = 2690 MHz and the Doppler shift at speed v = 350 km/h is 900 Hz.
2.4.4 3GPP Channel Model

Since 1998, ETSIs standardization of third-generation mobile systems has been carried out in the 3rd Generation Partnership Project (3GPP). The 3GPP standards for radio propagation in UMTS are mainly based on ITU models. These standards stipulate certain performance tests for both mobile stations and base stations under different propagating conditions covering additive white Gaussian noise and multipath fading environments. As an example, the 3GPP propagation models used for performance evaluation of different multipath environments are shown in Table 2.7 [27]. All taps have the classical Doppler spectrum [28]. These models are used as reference for the specification and testing of both uplink and downlink data channels with defined parameters and data rates of 64 kbps, 144 kbps, 384 kbps, and 2048 kbps. Case 1 in Table 2.7 is a single tap heavily faded model and is almost identical to the ITU pedestrian S model at speed 3 km/h. Case 2 and case 4 are similar (mobile speed of 3 km/h), but the former has less fading and a higher multipath diversity. Case 3 has four channel taps and a mobile terminal speed of 120 km/h, which is similar to the vehicular A channel. Case 5 is similar to case 1, but the mobile terminal speed is 50 km/h. Case 6 and case 3 have the same number of taps, but the UE speed is 250 km/h in case 6. The signals arriving at the receiver are considered independent (i.e., no correlation among the received signals), a case appropriate for these channel models.
2.5 MIMO Channel Models

The spatial characteristics of a radio channel have a significant effect on the performance of multiple-input multiple-output (MIMO) systems. The MIMO techniques take the advantage of multipath effects in the form of spatial diversity to significantly improve SNR by combining the outputs of de-correlated antenna arrays with low mutual fading correlation. The other technique to improve the gain of a system using multi-antenna arrays is spatial multiplexing, which creates multiple parallel channels between the transmitter and receiver sides. Multiple antennas at the transmitter and/or the receiver side can be used to shape the overall beam in the direction of a specific user to maximize the gain. This technique is called beamforming. The large MIMO gains can be achieved by low spatial correlation. The antenna separation, in terms of the wavelength of the operating frequency, has a significant impact on the spatial correlation. To achieve a low fading correlation, the antenna separation
Table 2.7 Average Path Powers and Relative delays for 3GPP Multipath Channel Models for specic cases
Case 1 Speed 3 km/h Relative Delay (ns) 0 976 Avg Power (dB) 0 10 Case 2 Speed 3 km/h Relative Delay (ns) 0 976 20000 Avg Power (dB) 0 0 0 Case 3 Speed 120 km/h Relative Delay (ns) 0 260 521 781 Avg Power (dB) 0 3 6 9 Case 4 Speed 3 km/h Relative Delay (ns) 0 976 Avg Power (dB) 0 0 Case 5 Speed 50 km/h Relative Delay (ns) 0 976 Avg Power (dB) 0 10 Case 6 Speed 250 km/h Relative Delay (ns) 0 260 521 781 Avg Power (dB) 0 3 6 9
59
60
should be large. The small sizes of wireless devices restrict large antenna separation, depending upon the wavelength of the operating frequency. An alternative solution to achieve low correlation is to use antenna arrays with cross-polarizations (i.e., antenna arrays with polarizations in orthogonal or near orthogonal orientations (discussed in Section 2.5.4)). Different channel modeling approaches (see Section 2.3) are used to develop MIMO channel models for LTE. The 3GPP/3GPP2 spatial channel and its extension SCME model, described in Sections 2.5.1 and 2.5.2, respectively, are rayor geometric-based stochastic channel models, while ITU models, described in Sections 2.4.2 and 2.4.4, are correlation-based channel models. Section 2.5.3 details the IST-WINNER channel model which is also the geometric-based stochastic channel model. Section 2.5.4 describes the concepts of polarized antenna arrays and in subsections 2.5.4.1 and 2.5.4.2, an overview of the 3GPP-polarized SCM model and the theoretical MIMO channel model is presented using polarized antenna arrays. Section 2.5.5 deals with the LTE evolution channel, and in Section 2.5.6, a comparison of standard MIMO channels is made.
2.5.1 3GPP Spatial Channel Model

The standardization bodies for third-generation cellular systems (3GPP and 3GPP2) jointly established a double directional geometry-based spatial channel model (SCM) for modeling, analysis and evaluation of MIMO concepts in outdoor environments for a system with bandwidth 5 MHz and 2 GHz frequency band (e.g., HSDPA, LTE) [29]. The modeling formation is similar to the COST 259 directional channel model but with some differences. First, it is not a continuous model but advises a precise discrete implementation (i.e., DOAs and DoDs are supposed to be fixed), and second, the movement of the mobile terminal is not continuous on a large scale within a cell; the model illustrates diverse positions of the mobile terminal within a cell. The SCM model describes two types of models: a calibration-level model based on tapped delay line approach taps is described in both the delay and angular domains, and a geometric-based (ray-tracing) stochastic model for system level simulations. The spatial description of the channel is illustrated in terms of directional distribution of multipath components and the angular energy distribution at both the mobile station and the base station. The calibration model is used to give only one snapshot of the channel behavior [29]. This model is not intended for performance assessment of the behavior of systems or algorithms; its purpose is to only verify the accuracy of simulation implementations. The calibration model, described by 3GPP/3GPP2, can be seen as a spatial extension of ITU-R channel models [21], where the channel characteristics are described as tap delay lines. The taps are independently fading with different delays and each tap is illustrated by its own mean angular spread, power azimuth spectrum (Laplacian or uniform), and mean direction at both the BS and the MS. These parameters are fixed so the channel characterizes stationary channel
61
conditions. The Doppler spectrum is characterized by introducing the direction of travel and speed of the MS. The SCM model used for the comparison and performance evolution of MIMO systems is called the system level simulation model. The system level simulations typically consist of multiple UE, base stations, and multiple cells/sectors. Performance metrics such as power delay profiles, angle spreads, and throughput are generated over a large number of drops Da drop is specified as a simulation run for given cells/sectors. These channel observation periods (i.e., drops) are significantly separated in space or time and channel large-scale parameters remain constant within a drop but the channel suffers from fast fading according to the movement of the mobile terminals, which varies randomly. The base station can schedule its transmission according to the channel state information provided by the mobile terminals. The spatial parameters are described according to the geometrical framework shown in Figure 2.2 [29]. The SCM model distinguishes between three propagation environments: urban macrocell, suburban macrocell, and urban microcell. The overall methodology is similar for these environments but there are some optional features in the basic propagation scenarios; far scatterer clusters in case of bad urban environment, modified angular distribution at the MS in the urban canyon model representing dense urban areas for both the urban macro and urban micro scenarios, and a LOS component in the urban microcellular case based on the Ricean K-factor. In addition, the spatial parameters like delay spread, angular spread, and so forth, are different for each of these environments.
Cluster n
N Sub-path m nmAoA nmAoD nmAoA MS broadside array nAoA MS V
BS nAoD
nmAoD
MS MS array
BS BS broadside array BS array
Figure 2.2 Geometry of BS and MS angular parameters.
62
The simulation model is a geometry-based stochastic model in which the movement of the mobile terminal within a given cell and the orientations of antenna arrays chosen at random are modeled geometrically. This channel model is based on ITU-R models described in Section 2.4, so the number of propagation paths with different delays is six for each environment. The paths are described by their mean angles and delays, which are correlated random variables with normal or log-normal probability density function [30]. Each path arrives at the BS or at the MS with angular dispersion. This dispersion is modeled by representing each path by a number of sub-paths with the same delays but different DoDs and DoAs dispersed around the mean angles with different fixed offsets i , where i represents the number of sub-paths. In all environments, angular dispersion for each path is composed of 20 sub-paths. The per-path angular dispersion at both the UE and the base station is described by Laplacian distribution, which is obtained by giving 20 sub-paths the same power, and fixed azimuth directions with respect to the nominal direction of the corresponding path. Addition of the different sub-paths gives Rayleigh or Rice fading. The angle spread, delay spread, and shadow fading are correlated random variables. Path losses for the environments are determined by the COST 231-Hata model for urban and suburban macrocells and the COST 231-Walfish Ikegami model for microcells. The SCM model was designed for different antenna radiation patterns, antenna orientations, and geometry to be applied. For example, antenna patterns and antenna spacing at the base station can be varied using antenna patterns for three-sector, sixsector cells, or omni-directional pattern and inter-element spacing of 0.5, 4, and 10 wavelengths. The composite angle spreads, delay spreads, and shadow fading which can be correlated random variables depending on the employed scenario, are applied to all sectors or antennas of the given base. When all parameters and antenna effects are specified, we can extract analytical formulation from the physical model. During each drop, a different correlation matrix is obtained for the analytical model. Table 2.8 shows the main parameters for this channel model.
2.5.2 Extended 3GPP Spatial Channel Model

The channel models play a vital role for the performance evolution and comparison of communication systems. With the advancement of communication technologies, these models need to be refined to incorporate challenging advanced communication algorithms. The SCM model, described in Section 2.5.1, operates in the 2-GHz frequency band and supports bandwidth up to 5 MHz. An extension to the 3GPP/3GPP2 model was made and used within the European IST-WINNER project. This is known as the SCME (SCM-Extension) model [31, 32]. The main contribution of this spatial channel model is broadening the channel bandwidth from 5 MHz to 100 MHz in the 2- and 5-GHz frequency bands. The focus was to extend the model in such a way that it remains backward-compatible with the conceptual
63
Table 2.8 The Main Parameters of 3GPP Spatial Channel Model Realizations
Parameter No. of paths (N) No. of sub-paths per path (M) Mean angle spread at BS Per-path angle spread at BS (Fixed) Mean angle spread at MS Per-path angle spread at MS (Fixed) Mean total delay spread (r) Std. deviation for log-normal shadowing Path loss model (dB) Suburban Macro 6 20 5 2 68 35 0.17 s 8 dB Urban Macro 6 20 8 , 15 2 68 35 0.65 s 8 dB Urban Micro 6 20 19 5 (LOS and NLOS) 68 35 0.25 s NLOS:10 dB, LOS:4 dB
31.5 + 35 log10 (d) 34.5 + 35 log10 (d) NLOS:34.53+38 log10 (d) LOS:30.18 + 26 log10 (d)
approach of the 3GPP/3GPP2 SCM model. The extension is based on the shortcomings of the existing SCM model (i.e., large bandwidth support, no LOS component in case of macrocells, and short-term time variations in system level model). To extend the model, the bandwidth extension is done in such a way that it remains compatible with the original 5-MHz bandwidth, by introducing the concept of the intra-cluster delay spread. The idea was initially proposed by Saleh and Valenzuela for indoor channel modeling. The idea of the intra-cluster delay has also been employed for outdoor scenarios in COST 259. The 20 sub-paths of a path are divided into subsets, called mid-paths. These mid-paths define the intra-cluster delay spread and have different delays and power offsets relative to the original path. Each mid-path consisting of a number of sub-paths acts as a single tap (delay resolvable component). Grouping together a number of sub-paths makes the fading distribution of that tap approximately Rayleigh distributed. The angle spreads (AS) assigned to the mid-paths are optimized in such a way that the angular spread of all mid-paths combined is minimized. The resulting SCME impulse response has a good approximation to the respective SCM impulse response. Due to bandwidth extension, the number of tap delays increases from 6 to 18 or 20 depending upon the propagation scenario. Table 2.9 [31] shows mid-paths powers and delays for SCME.
64
Table 2.9 The SCME Mid-Path Power-Delay Parameters

Scenario No. of Mid-paths per Path Power 1 2 3 4 10/20 6/20 4/20 Suburban Macro Urban Micro 3 Relative Delay (ns) 0 ns 7 ns 26.5 ns Power 6/20 6/20 4/20 4/20 Urban Micro 4 Relative Delay (ns) 0 ns 5.8 ns 13.5 ns 27.6 ns
One other contribution of SCME is the evolution of spatio-temporal parameters for fixed tap delay line (TDL) models called cluster delay-line models. The model parameters: power, delays, and angles of departures and arrivals are assigned fixed values, illustrating all MIMO propagation parameters. The tap-delay line model is similar to the SCM link level model; however, it can be closely approximated to the SCM system level model, which is optimized for small frequency autocorrelations. The path loss for the SCME model in the 5 GHz band is proposed on the basis of path loss models used in the SCM model with an offset of 8 dB to the 2-GHz path loss model. The COST 231 Walfish Ikegami model is selected as the standard path loss model for all scenarios. The SCME model offers a number of optional features which can be employed depending upon the specific simulation purpose. In the SCM model, the LOS option is for urban micro only. The SCME model also incorporates the K-factor option (i.e., the LOS option for urban and suburban macro scenarios by assigning the same parameters to both scenarios). The SCME also features the time evolution of system level parameters (i.e., it introduces the optional drifting of the path delays and angles of arrivals and departures). In the SCM model, all the propagation parameters stay fixed and independent during the observation periods, which are significantly separated from each other in space or time. This approach is also followed in the SCME model; the length of these intervals is extended by adding the short-term time unpredictability of some channel parameters within the drops. The channel parameters stay independent between the drops. Because of fixed geometry assumption, the sub-path delays and scatterer angles do not change at the BS, but due to movement of the MS, these parameters vary during a drop as seen from the MS. The drifting of these parameters is intended to the testing of beamforming algorithms. Another optional feature of the SCME model is drifting of shadow fading, which is modeled by the exponentially shaped spatial autocorrelation function, which shows that correlation of shadow fading decreases exponentially with distance.
65
2.5.3 WINNER Channel Model

The European WINNER (Wireless World Initiative New Radio) project is responsible for developing new radio concepts for Beyond-3G systems (e.g., LTE) using a frequency bandwidth of 100 MHz and a radio frequency lying between 2 and 6 GHz in spectrum. The latest developments in MIMO channel modeling are made within WINNER WPs (Work Packages). The WINNER models used the GSCM principle and generic approach for all scenarios with the same generic structure. Generic multilink double-directional models are developed for system level simulations while cluster delay line (CDL) models with reduced statistical variability of small-scale parameters are used for calibration and comparison purposes. Extensive measurement campaigns conducted by five partners in different European countries provide the background for the parameterization of various scenarios. Initially there was no broadly accepted channel model suitable for WINNER system parameters. In the beginning, the 3GPP/3GPP2 SCM model was selected for outdoor simulations. Due to the limited frequency applicability range and narrow bandwidth, some modifications were made to cope with more advanced simulations. However these initial models were not adequate for advanced level simulations. The main requirements were the proper categorization of spatial properties for MIMO support, consistency in space, time, and frequency (e.g., inherent association between Doppler and angle spreads); a set of possible channels and some limited randomized channels; statistical variability of bulk parameters, and extended polarization support. Consequently, new WINNER models had to be developed. The WINNER channel models were developed in two phases of the IST-WINNER project [33]. In the first stage, the so-called WINNER generic channel model for immediately required propagation scenarios with a limited number of parameters was created based on channel measurements at 2 and 5 GHz. The stochastic channel modeling approach provides unlimited double-directional channel realizations. This generic channel model is a ray-based multilink double-directional model, which is scalable, antenna independent, and capable of channel modeling for MIMO systems. Channel characterization parameters (e.g., delay spreads, angle spreads, and power delay profiles), cross-polarization, shadow fading, and path loss extracted from measurements for the scenarios of interest and the respective statistical distributions can be integrated into the generic model. The following scenarios are of interest in phase 1: indoor typical urban microcell, suburban macrocell, typical urban macrocell, rural macrocell, and stationary feeder link. In the second stage [34], the WINNER-l channels models were upgraded and new multidimensional channels were developed based on the measurement campaigns. More parameters were included and the frequency range was increased to cover the 2 to 6 GHz spectrum. The numbers of scenarios are increased to 13 based on the feedback from other work packages. The propagation scenarios of concern are indoor office, indoor to outdoor, outdoor to indoor, large indoor hall, urban micro cell, bad urban microcell, stationary feeder, suburban macrocell, urban macrocell,
66
rural macrocell, and rural moving networks. Table 2.10 shows the specific scenarios according to the environments [i.e., for wide area (WA), metropolitan area (MA), and local area (LA) environments] [34]. Measurement campaigns showed that the differences between indoor-to-outdoor and outdoor-to-indoor scenarios are negligible; therefore, these scenarios are merged together. These geometric-based stochastic channel models use a generic channel modeling approach, which means that the number of antennas, antenna configurations and geometry, and antenna beam patterns can be changed without varying the basic propagation model. The new features of the second stage include representation of the elevation of rays, LOS components taken as random variables, and moving scatterers in fixed links. This method facilitates the same channel data in different system level and link level simulations [34]. Path loss models for various propagation scenarios are also developed on the basis of measurement campaigns conducted within the WINNER and from open literature. The general structure of the path loss model is of the form [15]: Pl = A log10 (d [m ]) + B + C log10 The free space path loss is of the form: Pl = 20 log10 (d ) + 46.4 + 20 log10 fc 5.0 (2.6) fc GH z 5.0 +X (2.5)
where f c is the carrier frequency and d is the separation between the transmitter and the receiver. The parameters A , B , and C are respectively the path loss exponent, intercept, and path loss frequency dependence. The parameter X is optional for specific cases. Details about these parameters are given in [15].
2.5.4 Multi-polarized MIMO Channel Models

The multiple-input multiple-output systems using arrays of spatially separated antennas at both ends show a dramatic increase in capacity by exploiting the multipath effects. However, in a LOS scenario, the MIMO systems show reduced performance since the LOS components overpower the multipath components in the received signal. The exploitation of polarization dimension results in the improved performance of MIMO systems. Indeed the orthogonal polarization ideally offers a complete channel separation, with a full de-correlation between the transmitter and receiver sides [35]. With spatially separated and cross-polarized antenna arrays, both the polarization diversity and polarization multiplexing can be achieved (e.g., two dualpolarized spatially separated arrays form four antenna arrays. This concept can be extended for nr nr MIMO systems with the assumption that antenna arrays consist of n t /2 and nr /2 dual-polarized sub arrays.
Table 2.10
Scenario
Selected Propagation Environments of the IST-WINNER Channel Model

Definition
LOS/NLOS
Environment
Frequency (GHz)
Mobile Velocity (km/h)
Notes
A1
Indoor/residential
LOS/NLOS
LA
2--6
0--5
A2
Indoor to outdoor
NLOS
LA
2--6
0--5
AP inside UT outside. Outdoor, environment urban
B1
Typical urban microcell
LOS/NLOS
LA, MA
2--6
0--70
B2
Bad urban microcell
NLOS
MA
2--6
0--70
Same as B1 with long delays
B3
Large indoor hall
LOS/NLOS
LA
2--6
0--5
B4
Outdoor to indoor microcell
NLOS
MA
2--6
0--5
Outdoor typical urban B1. indoor A1
B5
LOS stationary feeder
LOS
MA
2--6
C1
Suburban
LOS/NLOS
WA
2--6
0--120
C2
Typical urban macrocell
LOS/NLOS
WA/MA
2--6
0--120
C3
Macrocell bad urban macrocell
NLOS
WA/MA
2--6
0--70
Same as C2 with long delays
C4
Outdoor to indoor macrocell
NLOS
MA
2--6
0--5
Outdoor typical urban C2-Indoor A1
D1
Rural macrocell
LOS/NLOS
WA
2--6
0--200
D2
1)Moving networksBS-MRS rural 2)Moving networksMRS-MS rural
67
LOS
WA
2--6
0--350
Large doppler variability
LOS/OLOS/NLOS
LA
2--6
0--5
Same as A1 NLOS
68
The cross-polar transmissions (e.g., from a horizontally polarized transmit antenna to a vertically polarized receive antenna) should be zero. But in real propagation scenarios there is always some depolarization due to the following reasons: linearly polarized antennas have nonzero patterns for cross-polarized fields. Therefore, signals arriving at the, say, vertically polarized antenna from a horizontally polarized antenna will not be zero. Also, due to the multipath scattering effects (i.e., diffuse scattering, diffraction, reflection, and so on), the polarization of the incident electromagnetic wave at the receiver may change [35]. In the following subsections, the two MIMO channel models employing the concept of polarization arrays are described briefly.
2.5.4.1 3GPP Polarized Spatial Channel Model

The antenna separation has a significant impact on the spatial correlation (i.e., the larger the separation between the antenna arrays, the lower the spatial correlation, and vice versa). The large MIMO gains can be achieved with low spatial correlation. Since multiple antennas on the handheld devices require spacing much less than half of the wavelength of the carrier frequency, polarized arrays are likely to be the primary choice to implement multiple antennas [36]. The 3GPP/3GPP2 spatial channel model described in Section 2.5.1 has several optional features. One of the optional features is the use of multi-polarized antenna arrays at both the transmitter and receiver sides. All other features, like the angular spread, delay spread, DOAs, AOAs, power delay profiles, path loss modeling, movement of the mobile terminal to model Doppler spreads, the drop concept, the number of propagation environments and the procedures to characterize them, are the same as for the SCM model. The sub-paths in the case of multi-polarized arrays are to be determined as follows. As described in Section 2.5.1, the multipath signals arrive at the BS or at the MS with angular dispersion; this dispersion is modeled by expressing each path (total paths are six) by a number of sub-paths with the same delays but different DoDs and DoAs distributed around the mean angles. In all environments, angular dispersion for each path is composed of 20 sub-paths. In case of polarized arrays, to consider the effects of signal leakage into the cross-polarized antenna orientations due to scattering, additional M sub-paths at the BS and M sub-paths at the MS are created. The angle of departures and angle of arrivals for these sub-paths are calculated in the same way as in the case of the co-polarized SCM model, which follows Laplacian or uniform distribution. The phase offsets for the cross-polarized elements are also (x , y ) determined. The phase (n , m ) is the phase offset of the mth sub-path of the nth path between x-component (vertical-polarized or horizontal-polarized) of the MS antenna array and y-component (vertical-polarized or horizontal-polarized) of the (y , x ) (y , y ) x,x) BS antenna array. The phase offsets ( (n , m ) , (n , m ) , and (n , m ) uniformly distributed in (0, 360 ), are also determinedthe fading is seen independent between orthogonal polarizations, therefore, sub-path phases are modeled randomly. The propagation
69
characteristics of horizontal-to-horizontal paths are equivalent to those of vertical-tovertical paths. The co-polarized and cross-polarized sub-paths are decomposed into vertical and horizontal components based on the co-polarized and cross-polarized orientations. The leakage of power P2 of each sub-path in the horizontal direction is set relative to the power P1 of each sub-path in the vertical direction according to the cross-polarization discrimination relation (XPD) ratio (i.e., XPD = P1/P2). For all scenarios, the XPD is calculated from the following distribution [29]: P2 = P1 A B (0, 1) (2.7)
The term (0, 1) is a Gaussian random number with zero mean and unit variance. P1 and P2 are the respective powers of sub-paths in the horizontal or vertical directions; A and B are the relative mean path powers and standard deviation of crosspolarization discrimination variation, respectively. The coupled powers of the V-H (vertical-horizontal) and H-V (horizontal-vertical) XPD are the same by symmetry. At the receiving antennas, the horizontal and vertical components are decomposed into components that are co-polarized with the receiving antennas and are added. The bulk parameters, path losses, and log-normal shadow fading are calculated in the same way as for the SCM model. Based on these calculations, a number of independent channel realizations are found using the drop concept.
2.5.4.2 Theoretical Polarized MIMO Channel Model

Widely used channel models such as the 3GPP/3GPP2 spatial channel model described in the earlier subsection, have limited support for cross-polarization MIMO channels in LOS scenarios. A modified SCM for cross-polarized MIMO channels for both line-of-sight and non-LOS (NLOS) scenarios are found in [37]. This generic theoretical cross-polarized MIMO channel model is derived using the co-polarized MIMO channel model for both LOS and NLOS scenarios. This model shows that cross-polarization discrimination (XPD) depends on the polarization mismatch between the transmit and the receive antenna pairs and antenna patterns. Another feature of this model is that it can be fitted to the SCM model to predict the dualpolarized MIMO channels for LOS and NLOS scenarios. For LOS case, the channel model for co-polarized antenna elements is derived first and is then extended by element-wise multiplication with a matrix having the polarization mismatch loss between the antenna pairs at both ends and the effect of the azimuth direction of the mobile terminal as well. The channel matrix for the line-of-sight case can be written as [37]: HLOS =
V HN M A V N M HV HN M A N M H HN M A V N M HH HN M A N M
(2.8)
70
Equation 2.8 can be written in matrix form as: HLOS = H2 N 2 M A 2 N 2 M (2.9)
where the dot between the two matrices denotes an element-wise multiplication. H2 N 2 M is the LOS channel matrix for the 2M-transmitter and the 2N-receiver copolarized MIMO system. The matrix A 2 N 2 M illustrates the polarization mismatch between the transmit and receive antenna elements. When antennas at both the ends are strictly aligned, cross-polar transmissions are considered zero, but in reality there is always some depolarization. To represent this, the polarization rotation angle is introduced in the model. Since the normal vectors of the transmitting and receiving antennas do not lie along the LOS path, in this model azimuthal displacements of the antenna pairs are also taken into account. To accomplish this, a displacement H HH angle factor is multiplied to the A V N M and A N M of the polarization mismatch matrix [37]. This concept can be extended to the M N MIMO channel to obtain the arbitrarily polarized MIMO channel. In the case of NLOS, the signals transmitted with horizontal or vertical polarization are not zero at the cross-polarized receiver. The model describes the polarization rotation is distributed within (0, 2). The model for NLOS scenarios models the amplitude and phase changes during multipath effects and also the polarization rotation angle for each path, respectively. The elements of the channel matrix are the sum of all multipath fields for the corresponding antenna pairs. It is shown that in rich scattering environments, elements of the channel matrix can be modeled as Gaussian random variables. Thus, in a rich scattering environment, there is not much impact by polarization on the channel statistics. This theoretical model can be integrated into the SCM model. The details are found in [37].
2.5.5 MIMO Channel Model for LTE Evolution

Initially, the SCME channel model, discussed in Section 2.5.2, was taken as the standard model for the design of LTE, BS and UE performance requirements, radio resource management requirements, and RF designing to derive requirements without applying the full complexity of the channel model in [38]. The full implementation of the SCME channel requires a significant amount of simulation time, and its backward compatibility with the SCM puts limitations on its performance. Therefore, some simplifications were made in the SCME channel model to obtain a MIMO tap delay line model for design purposes of LTE. The modifications were performed using the following three steps. First, the statistical variability of the SCME model is removed by defining the fixed tap delay line models. The fixed tap delay line models are defined for four environments given in Table 2.10. The optional feature, polarization, of the SCM and SCME models, which gives independence between the antennas, are also included
71
in the simplified models. These models are also intended to be used in the system level simulations. In the second step, the antenna configurations at the MS and BS are defined. At the BS, two spatially separated dual polarized (+45/45) antenna elements are taken with three-sector or six-sector antenna patterns according to the calibration model in the SCM model. The radiation pattern for the three-sector or six-sector antenna is as follows: A () = min[12(/3dB )2 , A m ] where 180 180
For a three-sector antenna pattern: 3dB = 70 , A m = 20 dB and maximum gain: 14 dBi For a six-sector antenna pattern: 3dB = 35 , A m = 23 dB, maximum gain: 17 dBi The separation between antenna elements is chosen to be 0.5 or 4 with polarization assumed to be unchanged over all AoDs. The azimuthal directions of the BS antennas are set so that the angle of departure of the first tap occurs at +20 in all scenarios. Two types of MS scenarios are considered: a laptop with two dual-polarized spatially separated antennas (verticalhorizontal) and a handset with two orthogonally polarized (verticalhorizontal) antennas as shown in Figure 2.3 [38]. The antenna pattern shapes are the same as in the case of the BS but with a wider beamwidth
Laptop 2 Top view Side view + radiation pattern Back view 45 First tap Handheld unit 45 60
Back view
Side view + radiation pattern Data Talk position
Figure 2.3 Antenna positions for two scenarios, using handset and laptop.
72
and different side lobe levels. The polarizations are assumed to be pure horizontal and vertical in all directions with an antenna nominal position. The handset can be in the talk or web browsing position. In talk position, the lobe is in the horizontal direction and the handset is turned 60 (polarizations are also rotated). In the data position, the MS is at 45 such that the lobe has its maximum partially downwards. The azimuthal directions of the MS antennas are adjusted such that the angle of arrival of the first tap occurs at +45 in all scenarios. The parameter values for antenna patterns are: Handheld, talk position: 3dB = 120 , A m = 15 dB, maximum gain: vertical: 3 dBi, horizontal: 0 dBi Handheld, data position: 3dB = 120 , A m = 5 dB, maximum gain: vertical: 3 dBi, horizontal: 0 dBi Laptop: 3dB = 90 , A m = 10 dB, maximum gain: 7 dBi, spatial separation: 2 In the third step, using the angular and polarization conditions with the antenna configurations described earlier, correlation matrices per channel tap are calculated for the LTE evolution model. The polarization covariance matrices are determined instead of the correlation of the polarization combinations to account for the power imbalances between different combinations of antenna polarizations and between the MS antennas. The Kronecker product of the BS and the MS spatial correlation matrices and the polarization covariance matrix is used to obtain the total per tap covariance matrix, R tap = p tap g BS , tap g MS , tap A B (2.10)
where p tap represents the relative power of the tap, g BS , tap is the BS antenna gain, and g MS , tap is the gain at the MS antenna. A and B are the correlation matrices of the BS and the MS, respectively, and represents the polarization covariance matrix. The channel models, the propagation environments and the BS and MS arrangements are given in Table 2.11. [38].
Table 2.11 The Channel Models and Propagation Environments for the BS and the MS Arrangement
Model SCM-A SCM-B SCM-C SCM-D Propagation Environment Suburban macro Urban macro (Low spread) Urban macro (High spread) Urban micro BS Arrangement 3-sector, 0.5 spacing 6-sector, 0.5 spacing 3-sector, 4 spacing 6-sector, 4 spacing MS Arrangement Handset, talk position Handset, data position Laptop Laptop
73
The two models SCM-C and SCM-D are used for evaluating laptops with two receiving antennas. In these models, channel realizations are calculated with one of the two dual-polarized antennas.
2.5.6 Comparison of SCM, SCME, WINNER, and LTE Evolution Models

A comparison of the SCM, its extension SCME, and the WINNER model is made in terms of small-scale and large-scale effects in the MIMO radio channel in [39]. All three models use the same generic ray- or geometric-based stochastic approach. The delay spreads, angular spreads, departure and arrival angles are used to characterize these channel models, which are chosen randomly from the appropriate probability distributions. The values of these parameters are different in different simulation runs for a single user or in multi-simulation runs for different users. The channel models are compared in [39] in terms of system bandwidth, the correlation between large-scale parameters at system level, antenna arrays and polarization, drop concept, cluster delay spreads, cluster angular spreads, and complexity issues. The channel model for LTE evolution, discussed in Section 2.5.5, is a simplified version of SCME analogous to link level calibration models. The procedure of finding spatial correlation matrices in the SCM/SCME and the WINNER is different from that of the LTE evolution model. The spatial correlation for the SCM/SCME and the WINNER is defined by the angular geometry and nominal direction of the sub-paths of each delay tap while the LTE evolution model describes deterministic correlation values for different clusters. Table 2.12 describes the main features and Table 2.13 shows key parameter values of these models [39].
Table 2.12 Different Attributes of SCM, SCME, WINNER, and LTE Evolution Models
Attribute Indoor scenarios BW >20 MHz Indoor-to-outdoor and outdoor-to-indoor scenarios Elevation angle AoD/AoA Intra-cluster delay spreads Cross-correlation between LSPs
SCM SCME WINNER LTE Evolution No No No No No No No No Yes No No Yes No Yes Yes Yes Yes Yes Yes Yes Yes No Yes No No Yes No No
Time evolution of model parameters
Large-scale parameters
74
Table 2.13 Key Parameters of SCM, SCME, WINNER, and LTE Evolution Models
Parameter No. of scenarios No. of taps Maximum bandwidth (MHz) Carrier frequency (GHz) No. of clusters No. of mid-paths per cluster No. of sub-paths per cluster

SCM 3 6 5 2 6 1 20
SCME 3 18--24 100* 2--6 6 3--4 20
WINNER 12 4--24 100** 2--6 4--20 1--3 20
LTE Evolution 4 18 20 6 3 -
Artificial bandwidth from 5 MHz bandwidth Based on 100 MHz measurements
2.6 Channel Modeling Role in Cell Planning and Optimization

Designing a cellular network is a challenging task that requires the use of available resources in an efficient way to allow networks to accommodate as many users as possible. One of the main purposes in the design of a cellular network is to meet the traffic demands of all potential users with an adequate quality of service (QoS) and with an acceptable percentage of blocked calls in the network [40]. Standard channel models assist network planners and system designers to meet these objectives. Therefore, realistic spatial and temporal radio propagation models are a critical part of any radio system design or network planning and optimization process. To design wireless networks with optimal user and frequency allocation and cellular planning, important statistical characteristics (e.g., path loss, log-normal shadowing, smallscale, or fast fading) must be predicted [41]. Detailed cell planning also includes coverage, spectral efficiency, capacity and parameter planning, parameters related to power control, neighbor cells, signaling, radio resource management, and so on. On the basis of these characteristics, cellular network designers are able to predict strict link budgets to generate cellular maps of areas of service with desired QoS and to optimize the data transfer within each radio communication channel [41, 42]. There are two ways to design a radio network [42]: the network planners can either produce their own propagation models for cell planning of different environments in the given area, or existing channel models are used that are generic in nature. Although the use of standard channel models is economical from a cost and time perspective, these models are usually not employed directly. The reason is that standard models
75
are developed by taking particular cities into account. So changes must be made through drive tests called correction factors. The use of different channel models in cellular planning and optimization have trade-offs between them. The use of deterministic channel models based on the uniform theory of diffraction (UTD) and the geometrical theory of diffraction (GTD) has been a widely adapted deterministic technique for radio propagation modeling in indoor, outdoor, and rural environments [40, 43]. The combined use of deterministic channel models, ray-tracing acceleration techniques, and topographical/ morphological databases give more accurate results for microcellular networks planning and design. However, these techniques are computationally complex, requiring large amounts of data and computer memory for simulations. When designing the interfaces between picocells and macrocells or between distant picocells, a better approach to work out propagation losses is to apply empirical or semi empirical methods. This is because in modeling these environments, deterministic methods require large amounts of computational resources and the improvement in precision is insignificant. These channel models can provide enough information necessary in the network design process in the case of macrocells and microcells. The deterministic channel models are adequate for microcells and picocells independently, but their implementation in macrocell designs makes them less favorable due to large computation times.
2.7 Conclusion
This chapter presents an overview of the important features of wireless channel modeling and standard channel models for UMTS and LTE communication systems. Standard channel models play a vital role in the design and performance assessment of advanced transceivers techniques and smart antennas employed to establish reliable communication links in mobile communication systems such as UMTS and LTE. From different channel modeling approaches, it is evident that there is no standalone master method to obtain radio channels with desired characteristics. There is always a trade-off between complexity and accuracy in modeling a radio channel. The channel models SCM and SCME, based on 3GPP specifications, are geometry-based stochastic models, suitable for LTE and give more accurate results for real propagation scenarios. However, these models are complex to some extent and require long simulation times. On the other hand, extended ITU models are correlation-based and show low computational complexity in the performance assessment of systems involving advanced antenna array concepts. The state-of-art channel models such as WINNER models provide additional features for state-of-the-art communication systems like UMTS and LTE. An efficient modeling of the spatial characteristics in MIMO systems is necessary to determine the improved transmission techniques for LTE. The spatial characteristics in multi-antenna channel modeling, including polarization effects, are expected
76
to be crucial in the coming LTE standards and future communication systems. Thus, new and improved channel models are necessary to evaluate the parameters and performance of these future systems.
References
[1] P. A. Bello, Characterization of Randomly Time-Variant Linear Channels, IEEE Trans. Comm. Systems, CS-11, 360, 1963. [2] M. Ibnkahla, (ed.), Signal Processing for Mobile Communications, CRC Press: New York, 2005. [3] H. Holma and A. Toskala, (eds.), LTE for UMTS: OFDMA and SC-FDMA Base Band Radio Access, John Wiley & Sons Ltd, 2009. [4] S. R. Saunders and A. A. Zavala, Antennas and Propagation for Wireless Communication Systems, John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester, West Sussex, PO19 8SQ, England, 2007. [5] A. F. Molisch, A Generic Model for MIMO Wireless Propagation Channels in Macroand Micro Cells, IEEE Trans. Signal Processing, vol. 52, no. 1, 2004. [6] M. Steinbauer, The Radio Propagation Channel-A Non-directional, Directional, and Double-Directional Point-of-View, Ph.D. dissertation, Vienna University of Technology, Vienna, Austria, 2001. [7] M. Steinbauer, A. F. Molisch, and E. Bonek, The Double Directional Radio Channel, IEEE Antennas and Propagation Magazine, vol. 43, no. 4, pp. 5163, 2001. [8] J. K. Cavers, Mobile Channel Characteristics, Kluwer Academic Publishers, New York, 2002. [9] J. Maciej, Nawrocki, M. Dohler, and A. H. Aghvami, (ed.), Understanding UMTS Radio Network Modeling, Planning, and Automated Optimization, John Wiley & Sons, The Atrium, Southern Gate, Chichester, West Sussex PO19 8SQ, England, 2006. [10] T. Rappaport, Wireless Communications, Principles and Practice, Prentice-Hall: Englewood Cliffs, NJ, 1996. [11] P. Almers, E. Bonek, A. Burr, N. Czink, M. Debbah, V. Degli-Esposti, H. Hofstetter, P. Kyosti, D. Laurenson, G. Matz, A. F. Molisch, C. Oestges, and H. Ozcelik, Survey of Channel and Radio Propagation Models for Wireless MIMO Systems, EURASIP Journal on Wireless Communications and Networking (special issue on space-time channel modeling for wireless communications), 2007. [12] G. D. Durgin, Theory of Stochastic Local Area CHANNEL Modelling for Wireless Communications, Doctor of Philosophy in Electrical Engineering, Blacksburg, Virginia, December, 2006. [13] A. F. Molisch, A. Kuchar, J. Laurila, K. Hugl, and R. Schmalenberger, Geometry-based Directional Model for Mobile Radio Channels: Principles and Implementation, European Transactions on Telecommunications, vol. 14, no. 4, pp. 351359, 2003. [14] J. D. Parsons, The Mobile Radio Propagation Channel, 2nd Edition, John Wiley & Sons Ltd, 2000. [15] J. Meinil a, T. J ams a, P. Ky osti, D. Laselva, H. El-Sallabi, J. Salo, C. Schneider, and D. Baum, IST-2003-507581 WINNER: Determination of Propagation Scenarios, D5.2 v1.0, 2004. [16] A. F. Molisch, H. Asplund, R. Heddergott, M. Steinbauer, and T. Zwick, The COST 259 Directional-Channel Model A-I: Overview and Methodology, IEEE Transactions on Wireless Communications, 12, pp. 34213433, 2006.
77
[17] L. M. Correia, (ed.), Wireless Flexible Personalised Communications (COST 259 Final Report), John Wiley & Sons, Chichester, UK, 2001. [18] L. Correia, (ed.), Mobile Broadband Multimedia Networks, John Wiley & Sons, New York, 2006. [19] I. D. Sirkova, Overview of COST 273 Part I: Propagation Modelling and Channel Characterization, Sofia, Bulgaria 29 June 1 July 2006. http://www.lx.it.pt/cost273. [20] S. Sesia, I. Toufik, and M. Baker, LTE The UMTS Long Term Evolution, John Wiley & Sons Ltd, 2009. [21] ITU-R M.1225 International Telecommunication Union, Guidelines for Evaluation of Radio Transmission Technologies for IMT-2000, 1997. [22] ITU-R M.1034 International Telecommunication Union, Requirements for the Radio Interface(s) for International Mobile Telecommunications-2000 (IMT-2000), 1997. [23] T. B. Srensen, P. E. Mogensen, and F. Frederiksen, Extension of the ITU Channel Models for Wideband (OFDM) Systems, in Proc. IEEE Vehicular Technology Conf., Dallas, Sept. 2005. [24] 3GPP Technical Specification 36.803, User Equipment (EU) Transmission and Reception, Release 8, V0.3.0, May 2007. [25] Ericsson, Nokia, Motorola, and Rohde and Schwarz, R4-070572: Proposal for LTE Channel Models, www.3gpp.org, 3GPP TSG RAN WG4, meeting 43, Kobe, Japan, May 2007. [26] Ericsson, R4-070994LTE Channel Models: High-speed Scenario, TSG-RAN Working Group 4 (Radio) meeting #43 bis Orlando, FL, June 2529 2007. [27] 3GPP Technical Specification 25.101, UE Radio Transmission and Reception (FDD), Release 5, V5.2.0, 2002-3. [28] 3GPP Technical Specification, 25.943, v4.2.0, Deployment Aspects, (Release 4), www.3gpp.org. [29] 3GPP Technical Specification 25.996, Spatial Channel Model for Multiple Input Multiple Output (MIMO) Simulations (Release 8), V8.0.0, 12-2008. http://www.3gpp.org. [30] A. Algans, K. I. Pedersen, Associate Member, IEEE, and Preben Elgaard Mogensen, Member, IEEE, Experimental Analysis of the Joint Statistical Properties of Azimuth Spread, Delay Spread, and Shadow Fading, IEEE Journal on Selected Areas in Communications, vol. 20, no. 3, April 2002. [31] D. S. Baum, J. Salo, G. Del Galdo, M. Milojevic, P. Ky osti, and J. Hansen, An Interim Channel Model for Beyond-3G Systems, in Proc. IEEE Vehicular Technology Conference, Stockholm, May 2005. [32] Elektrobit, Nokia, Siemens, Philips, Alcatel, Telefonica, Lucent, and Ericsson, R4-050854: Spatial Radio Channel Models for Systems Beyond 3G, 3GPP TSG RAN WG4 Meeting 36, London, UK, August 29September 2, 2005. [33] Commission of the European Communities, IST-WINNER Project, http://www.istwinner.org. [34] Commission of the European Communities, IST-WINNER II Project Deliverable D1.1.2, V1.2 WINNER II Channel Models, http://www.ist-winner.org. [35] C. Oestages and B. Clerckx, MIMO Wireless Communications: From Real World Propagation to Space-code Design, AP. [36] T. Hult, A. Mohammed, Z. Yang, and D. Grace, Performance of a Multiple HAP System Employing Multiple Polarization, Invited Paper, Special Issue of Springer Wireless Personal Communications Journal, vol. 52, Issue 1, January 2010. [37] L. Jiang, L. Thiele, and V. Jungnickel, On the Modeling of Polarized MIMO Channel, Fraunhofer Institute for Telecommunications, Heinrich-Hertz-Institute Einsteinufer 37, D-10587, Berlin, Germany.
78
[38] Ericsson, R4-060101: LTE Channel Models for Concept Evaluation in RAN1, TSGRAN Working Group 4 (Radio) meeting 38 Denver, CO, February 1317, 2006. [39] M. Narandzic, C. Schneider, R. Thom a, T. J ams a, P. Ky osti, and X. Zhao, Comparison of SCM, SCME, and WINNER Channel Models, in Proc. IEEE Vehicular Technology Conference, Dublin, April 2007. [40] M. F. C aatedra and J. P.-Arriaga, Cell Planning for Wireless Communications Artech House Mobile Communications Library, Artech House, Inc., 1999. [41] N. Blaunstein and C. Christodoulou, Radio Propagation and Adaptive Antennas for Wireless Communication Links, John Wiley & Sons, Inc., 2007. [42] A. R. Mishra, Fundamentals of Cellular Network Planning and Optimization 2G/2.5G/ 3G . . . Evolution to 4G, John Wiley & Sons Ltd, The Atrium, Southern Gate, Chichester West Sussex PO19 8SQ, England, 2004. [43] C. Smith and C. Gervelis, Cellular System Design and Optimization, McGraw-Hill Professional, 1996.

Overview of Wireless Channel Models For UMTS and LTE: Abbas Mohammed and Asad Mehmood

Uploaded by

Copyright:

Available Formats

Overview of Wireless Channel Models For UMTS and LTE: Abbas Mohammed and Asad Mehmood

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Overview of Wireless Channel Models For UMTS and LTE: Abbas Mohammed and Asad Mehmood

Uploaded by

Copyright:

Available Formats

Chapter 2

Overview of Wireless Channel Models for UMTS and LTE

Evolved Cellular Network Planning and Optimization

Overview of Wireless Channel Models for UMTS and LTE

2.2 Multipath Propagation Channels

Evolved Cellular Network Planning and Optimization

Overview of Wireless Channel Models for UMTS and LTE

2.3 Modeling Methods for Multipath Channel Modeling

2.3.1 Deterministic Channel Computation

Evolved Cellular Network Planning and Optimization

2.3.2 Geometry-based Stochastic Channel Models

Overview of Wireless Channel Models for UMTS and LTE

2.3.3 Non-geometrical Stochastic Channel Models

2.3.4 Correlation-based Channel Models

Evolved Cellular Network Planning and Optimization

2.4 Standard Channel Models

2.4.1 COST Channel Models

Overview of Wireless Channel Models for UMTS and LTE

2.4.1.1 COST 259 Directional Channel Model

Evolved Cellular Network Planning and Optimization

2.4.1.2 COST 273 Channel Model

Overview of Wireless Channel Models for UMTS and LTE

2.4.2 ITU Channel Models

Evolved Cellular Network Planning and Optimization

Overview of Wireless Channel Models for UMTS and LTE

Evolved Cellular Network Planning and Optimization

2.4.3 Extended ITU Models

Overview of Wireless Channel Models for UMTS and LTE

Table 2.5 Power Delay Proles for Extended ITU Models

Table 2.6 Summary of Delay Proles for LTE Channel Models

Evolved Cellular Network Planning and Optimization

2.4.4 3GPP Channel Model

2.5 MIMO Channel Models

Overview of Wireless Channel Models for UMTS and LTE

Evolved Cellular Network Planning and Optimization

2.5.1 3GPP Spatial Channel Model

Overview of Wireless Channel Models for UMTS and LTE

N Sub-path m nmAoA nmAoD nmAoA MS broadside array nAoA MS V

BS BS broadside array BS array

Figure 2.2 Geometry of BS and MS angular parameters.

Evolved Cellular Network Planning and Optimization

2.5.2 Extended 3GPP Spatial Channel Model

Overview of Wireless Channel Models for UMTS and LTE

Evolved Cellular Network Planning and Optimization

Table 2.9 The SCME Mid-Path Power-Delay Parameters

Overview of Wireless Channel Models for UMTS and LTE

2.5.3 WINNER Channel Model

Evolved Cellular Network Planning and Optimization

2.5.4 Multi-polarized MIMO Channel Models

Selected Propagation Environments of the IST-WINNER Channel Model

Mobile Velocity (km/h)

AP inside UT outside. Outdoor, environment urban

Overview of Wireless Channel Models for UMTS and LTE

Typical urban microcell

Bad urban microcell

Same as B1 with long delays

Large indoor hall

Outdoor to indoor microcell

Outdoor typical urban B1. indoor A1