Chapter 11
Summary
In this chapter we summarize the topics discussed in the book and provide a framework for future work. We have subdivided the summary into four sections. First, we review our learning objectives regarding the modeling of the LTE (Long Term Evolution) transceiver system. Then we summarize our findings regarding simulation of the system model and how to accelerate it. Third, we relate what we have learned about bridging the gap between modeling and implementation and how to prototype the simulation model as C/C++ software. Finally, we review some of the topics related to the LTE PHY (Physical Layer) that we have not had the chance to study in detail. Considering the level of detail needed to do justice to these topics, we have decided that they cannot be adequately covered in this volume and have left them as subjects of a future work.
11.1 Modeling
As the first learning objective of this book, we provided an overview of the mathematical modeling of the LTE PHY. Our aim was to provide a balanced approach to the discussion in order to foster a deeper understanding. As such, we decided to incorporate three distinct yet complementary conceptual elements: (i) providing an introductory theoretical overview of LTE-enabling technologies such as Orthogonal Frequency Division Multiplexing (OFDM) multicarrier transmission and Multiple Input Multiple Output (MIMO) multi-antenna schemes; (ii) providing an introductory technical overview of LTE specifications, focusing on a more detailed coverage of downlink transmission; and (iii) providing detailed MATLAB® algorithms and testbenches for step-by-step learning and hands-on simulation of the LTE standard. This balanced multitier approach is one of the distinguishing features of this book.
In this section we will summarize what we have presented regarding each of these conceptual elements.
11.1.1 Theoretical Considerations
Throughout the book we have provided discussions regarding the theoretical background of the enabling technologies of the LTE standard. We studied the LTE multicarrier transmission schemes (i.e., OFDM in downlink and its single-carrier counterpart SC-FDM (Single-Carrier Frequency Division Multiplexing) in the uplink), as well as the multi-antenna MIMO transmission schemes.
We presented various aspects of the theoretical underpinnings for the MIMO–OFDM transmission techniques. These revealed how MIMO and OFDM are combined in the standard and helped explain the success of the technology in achieving its goals of high maximum data rates and high throughputs in mobile communications. We also discussed how incorporating the best technologies from previous standards, such as link adaptations through adaptive modulation and coding and efficient turbo coding, contribute to the overall performance of the LTE standard. Examining the theoretical background of the underlying LTE technologies can also be useful in understanding other modern communication systems. OFDM and MIMO technologies also form the fundamental basis of the WiMAX and the new wireless LAN standards.
11.1.2 Standard Specifications
Besides discussing the theoretical foundations, we provided a detailed presentation of PHY signal processing, with a special focus on downlink processing. We reviewed various channels and signals used in the standard. We also provided a more in-depth look at both the Downlink Shared Channel (DLSCH) processing and Physical Downlink Shared Channel (PDSCH) processing.
In particular, we examined in detail the composition of the time–frequency resource grid used in both OFDM and SC-FDM transmission schemes. Understanding the structure of the resource grid shed light on how the LTE standard organizes user data, control information, reference and other signals, and how it performs channel estimation and equalization operations necessary to recover the data at the receiver. It also showed how easily the standard combines the OFDM multicarrier scheme with various MIMO multi-antenna techniques. We highlighted the flexibility of the standard in maintaining a single transmission structure yet accommodating nine different transmission modes for downlink and various uplink transmission modes. We also described how different transmission modes cater to different scheduling conditions and different profiles of mobility and channel quality.
11.1.3 Algorithms in MATLAB
As the distinguishing feature of this book, we presented PHY modeling with a progressive set of algorithms and testbenches in MATLAB and Simulink. Our goal in providing MATLAB algorithms and testbenches was to introduce an initial platform for MATLAB users who are involved in communications system design. Our hope was to offer a starting point that fosters future collaborations among members of this community. Simulating an executable specification of a communications system in MATLAB and Simulink can help take the guesswork out of validating the effects of introducing innovative algorithms in system design.
Starting with MATLAB algorithms characterizing the basic scrambling, modulation, and coding in Chapter 4, we proceeded to include the OFDM multicarrier transmission in Chapter 5 and various MIMO techniques, including transmit diversity and spatial multiplexing, in Chapter 6. In Chapter 7 we presented MATLAB algorithms that model typical link-adaptation strategies and in Chapter 8 we put together an LTE transceiver covering the first four modes of downlink transmission, then provided various assessments of the quality and performance of the physical-layer simulation model. Finally, in Chapters 9 and 10 we provided MATLAB algorithms that accelerate simulations and generate C code for the prototyping of designs as standalone applications. These topics will be discussed in further detail shortly.
11.1.3.1 Receiver Design
As with most communications standards, the LTE standard only specifies transmitter operations. Since the receiver operations are not explicitly specified, this provides an opportunity to develop innovative receiver algorithms. The innovations, when integrated within the software and hardware implementations by the network equipment and mobile terminal manufacturers, represent the proprietary and value-added contributions of each mobile communications system provider.
MATLAB and its communications system design tools provide an easy-to-use environment for experimenting with the design of various receiver components. In this book we presented various alternatives to different receiver components of the LTE system model. For example, in Chapter 5 we discussed receiver operations related to estimation of the channel-frequency response based on received reference signals. We examined an ideal channel estimator and three different channel-estimation algorithms based on the interpolation of pilot signals. The interpolation functions expand the channel responses computed at the pilots to cover the entire resource grid. As another example, in Chapter 6 we examined various MIMO receiver operations, studying three different approaches to computing best estimates of the transmitted symbols at the receiver. These techniques were based on the Zero Forcing (ZF), Minimum Mean Square Error (MMSE), and Soft-Sphere Decoder (SSD) algorithms. In Chapter 8 we examined the effects of each of these receiver algorithms on the overall system performance. By looking at the algorithm-specific metrics, such as the memory footprint or the computational complexity, as well as the system-level metrics, such as the Bit Error Rate (BER) or the throughput, we can assess the tradeoffs associated with each.
11.1.3.2 Simulation Testbenches
Throughout the book we created and updated MATLAB testbenches (or scripts) to evaluate qualitatively and quantitatively the performance of our LTE transceivers. The testbenches included the transmitter and receiver processing chains and the channel modeling sections needed to represent a transceiver. They also included various qualitative measures, such as spectrum analyzers and constellation diagrams, and quantitative measures, such as BER and throughput computations.
11.1.3.3 Algorithmic Building Blocks
It is important to choose the right granularity for the components of the MATLAB algorithms that model a complex system such as the LTE transceiver. We used a criterion that reflects the system-modeling and simulation mandate of this book. We did not reimplement basic communications building blocks such as modulators, convolutional or turbo encoders, decoders, or space–time block-coding components. For example, in order to implement OFDM transmitter and receiver operations we used MATLAB functions for forward Fast Fourier Transform (FFT) and Inverse Fast Fourier Transform (IFFT). We also used the dsp.FFT and dsp.IFFT System objects of the DSP System Toolbox as alternative implementations. These System objects can handle fixed-point modeling and block sizes that are not a power of two. We also used modulators, turbo coders, and channel-modeling System objects from the Communications System Toolbox. Leveraging available components from the toolbox, and not spending time redeveloping basic building blocks such as turbo decoders in MATLAB, helped accelerate the pace of the creation of system models for the LTE transceiver.
11.2 Simulation
It is important to develop a full mathematical model for any communications standard. However, to validate a model's accuracy, we must perform a representative number of software simulations. Since many of the performance metrics used in communications systems, such as throughputs and BERs, are measured in a statistical sense, a large amount of data must be processed by a simulation model. Furthermore, in order to verify that a system is robust against occasional outlier degradations, the simulations should be large enough to cover these rare occurrences. These considerations prompted us to look at various ways in which a system model can be optimized for speed. We looked at different methodologies for simulation acceleration and highlighted various tools and techniques that sped up our simulation model in MATLAB and Simulink.
11.2.1 Simulation Acceleration
Acceleration of a software simulation represents a classical tradeoff between an easy-to-understand and readable description of the model on one hand and optimized performance on the other. In our step-by-step approach to developing the components of the LTE model, we took great pains to organize the MATLAB code in a way that is self-contained. To make the code easy to understand, we represented most of the more complicated algorithms as a combination of less complicated subcomponents, without using any shortcuts or code factoring.
As instructive as this approach is, in order to accelerate the simulation speed we sometimes need to take advantage of typical methodologies that factor out repeated operations and fuse processing loops and optimize for various compilers or platform-specific libraries. In Chapter 9, we highlighted many MATLAB programming techniques that rely on these typical acceleration methodologies.
One of the most important and distinguishing features of the acceleration strategies presented in Chapter 9 was their preservation of numerical accuracy. As we went through various code optimizations, we showcased the fact that as successive versions of MATLAB codes execute more rapidly, they still produce the same numerical results. On the other hand, taking a more liberal approach to acceleration can be quite useful in receiver design where no standard specification is available. Many designers use shortcuts or approximations that substantially accelerate the simulation but do not preserve the numerical results. We deliberately took a conservative approach to code optimization and limited its scope to a subset that preserves numerical accuracy in order to make the process of validation easier and more direct.
11.2.2 Acceleration Methods
In Chapter 9, we showcased various techniques used to accelerate simulations of our LTE system model in MATLAB and Simulink. We presented a series of six types of optimization applied to control-channel processing. The techniques either provide ways of optimizing MATLAB programs or gain performance improvements through the use of additional computing power or by retargeting the design to compiled C code. We started with a baseline algorithm and through successive profiling and code updates introduced the following optimizations:
- Better MATLAB serial programming techniques (vectorization, preallocation)
- Use of System objects
- MATLAB-to-C code generation (MATLAB Executable, MEX)
- Parallel computing (parfor, spmd)
- GPU (Graphics Processing Unit)-optimized System objects
- Rapid accelerator mode for simulation in Simulink.
We also showed how to further accelerate simulations by combining two or more of these techniques. To take advantage of some of their benefits, specialized product capabilities beyond what is offered in application-specific toolboxes must be used. For example, MATLAB parallel-computing products provide computing techniques that take advantage of multicore processors, computer clusters, and GPUs. MATLAB Coder provides the ability to automatically convert a MATLAB code to C code, which can be compiled to provide faster simulations.
11.2.3 Implementation
Besides discussions regarding modeling and simulation, in Chapter 10 we went through the first steps involved in implementing the LTE-standard model. In order to bridge the gap between modeling and implementation, we used the MATLAB Coder to generate a prototype of the model as C code. We showed how the ANSI/ISO C source code generated by MATLAB Coder can be integrated with existing C/C++ testbenches and applications.
11.3 Directions for Future Work
There is a lot more to be done before we can adequately specify every detail of the PHY model of the LTE standard in MATLAB. In this book, our approach has mostly been pedagogic and educational. We focused on the LTE-enabling technologies, aiming to shed light on user-plane signal processing. We also covered as much detail as needed regarding various physical signals and channels, the organization of data in the OFDM resource grid, and the handling of multi-antenna techniques. These discussions clarified the underlying approach to transmission and explained the feasibility of achieving high data rates and improved system throughputs, as mandated by the standard specifications.
The next level of modeling is to provide a software solution that can be used as a reference to verify conformity to the LTE-standard requirements. If our objective is to ensure standard compliance, we must incorporate much more detail in our simulation model. The resulting LTE simulation model in MATLAB needs to incorporate all standard tests and cover all transmission modes and scenarios.
Next we will go through a list of modeling components that need to be added in order to evolve our baseline simulation model to the next level. With these upgrades, we can ultimately turn the LTE system model into a simulation platform for LTE-standard compliance testing. We will present these details in three sections: user-plane modeling, control-plane modeling, and system-access modules.
11.3.1 User-Plane Details
In order to update the LTE simulation model developed in this book, we need first to cover all aspects of user-plane modeling. These include the inclusion of both FDD (Frequency Division Duplex) and TDD (Time Division Duplex) duplexing for time framing, a complete treatment of both downlink and uplink shared-channel processing, and the inclusion of the LTE-Advanced features. These items are discussed in this section.
11.3.1.1 FDD and TDD Duplexing
As we saw in this book, two types of frame structure are specified in the LTE standard. Type 1 frames are used in FDD mode and type 2 frames in TDD mode. We have provided details relating to the FDD and type 1 frames. With minor modifications, we can present MATLAB functions that represent the time framing applicable to the TDD duplexing modes. Similarly, throughout the book we used normal cyclic prefix lengths, and again with minor modifications of the MATLAB code we can also accommodate extended cyclic prefixes in OFDM and SC-FDM transmissions.
11.3.1.2 Uplink Processing (PUSCH)
We have focused entirely on downlink transmission details in this book. The future work should contain the signal processing chain of the Physical Uplink Shared Channel (PUSCH). Many of the MATLAB components developed for downlink transmission can be used for uplink modeling almost without modification. However, there are some differences specifically related to the reference signals that are based on Zadoff–Chu sequences in the uplink specifications.
11.3.1.3 Complete Downlink Transmission Modes
We examined in detail the first four downlink-transmission modes. A complete model should include all of the modes, including the Downlink Enhanced MIMO modes (modes 7, 8, and 9), UE (User Equipment)-specific beamforming modes, and single-layer spatial-multiplexing modes. The modeling should include the generation and placement of various types of reference signal, including the Channel State Information Reference Signal (CSI-RS) and the Demodulation Reference Signal (DM-RS).
11.3.1.4 LTE-Advanced Features
LTE-Advanced features should also be included in the LTE MATLAB receiver model. These include in particular an uplink MIMO transmission and carrier aggregation. A multi-user uplink MIMO example populates and transmits PUSCH subframes in such a way that multiple UEs can share resources in a transmission. This technique is quite effective in boosting the uplink throughput. Carrier aggregation is another LTE-Advanced feature that enables downlink transmission to cover multiple carriers. By leveraging up to five contiguous carriers, carrier aggregation is the main technique responsible for achieving the maximum data rate of 1 Gbps provisioned within the LTE-Advanced standard. Functions that handle these two features must be part of a standard compliant MATLAB model for LTE PHY. As each of the processing chains in each of the carrier-aggregation bandwidths is independent, parallel processing can provide an obvious boost to the processing time needed for implementation. As such, the techniques we learned in Chapter 9 are directly applicable here.
11.3.2 Control-Plane Processing
As one of the features of this book, we focused on user-plane shared-channel processing. We did not study in any depth the control information needed to make the user-plane transmission possible. The collection of Downlink Control Information (DCI) and Uplink Control Information (UCI) must be part of a comprehensive LTE system model in MATLAB.
11.3.3 Hybrid Automatic Repeat Request
In the LTE standard, a Hybrid Automatic Repeat Request (HARQ) protocol is specified to ensure the reliability of data packet transmission and to manage occasional retransmissions. With a positive acknowledgment of a received packet, new data is transmitted. However, a negative acknowledgment initiates the retransmission of a previously sent packet. In order to provide a continuous supply of data packets at the receiver and minimize the waiting time for new data, we can send different data packets on different HARQ process numbers. In the LTE downlink specification, the DCI format contains explicit signaling related to the HARQ process. This includes an incremental-redundancy version and a new data indicator. In this book we have not presented the MATLAB functions necessary to implement the HARQ process. As an area of future work, inclusion of these routines will help contain the system delay resulting from excessive retransmissions and will update the way DLSCH handles channel coding with the inclusion of HARQ information.
11.3.4 System-Access Modules
In this book we focused on developing routines and functions that enable communications between UE and eNodeB (enhanced Node Base station) once initial access has been established. The LTE standard provides many components, signals, and capabilities for the initial phase of system access, cell search, and handoff procedures. A comprehensive system model in MATLAB should include these types of functionality. Two particular examples are described in further detail in this section.
11.3.4.1 Cell Search and Frame Timing
Encoded within the resource grid in the downlink transmitted signals are blocks of information that are essential to system access, cell search, and frame timing by a mobile unit. As we saw earlier, some of the initial system information is conveyed in the Master Information Block (MIB) and encoded and represented in the grid with a fixed modulation and coding scheme. The MIB contains information regarding system bandwidth, System Frame Number (SFN), and Physical Hybrid ARQ Indicator Channel (PHICH) configuration. We studied the Primary Synchronization Signal (PSS) and Secondary Synchronization Signal (SSS) and the Physical Broadcast Channel (PBCH) (containing the MIB) in Chapter 5. However, we did not present the MATLAB algorithms and functions that encode and transmit this information or the receiver operations that use it to obtain the initial system bandwidth and other critical information.
11.3.4.2 Random Access
In order to initiate access to the network, the UE uses the Physical Random Access Channel (PRACH) to transmit a preamble. Since this corresponds to the first communication from the UE to the eNodeB, the system does not know the type or specifications of the UE device. Various transmission modes, such as Cyclic Delay Diversity (CDD) and Precoding Vector Switching (PVS), provide a transparent way of decoding the preamble information. As we have not presented the uplink transmission details, we have not presented the MATLAB algorithms and functions needed for initial system access.
11.4 Concluding Remarks
In this chapter we summarized the learning objectives of this book and provided directions for further study. We subdivided the topics covered into two main categories: modeling and simulation. Within the modeling context, we elaborated on our stated goal of providing a balanced approach in presenting three distinct aspects related to understanding the LTE standard. We covered theoretical and mathematical descriptions of various enabling technologies, presented standard specifications as needed, and provided MATLAB programs and testbenches that enable hands-on experimentation with concepts through simulation. In the section on simulation, we highlighted the necessity of an adequate simulation speed for effective use of software that models a complex system like LTE. We reviewed various simulation acceleration techniques and prototyping mechanisms presented in this book. Finally, we presented a list of additional topics that need to be covered in a future work in order to provide a complete treatment of LTE-standard PHY modeling.
Depending on the interest in the LTE and MATLAB communities, the completion of our work in producing a fully standard-compliant LTE model in MATLAB may require another book. Having laid the foundation here by focusing on the enabling technologies and principles, the next volume would focus on standard compliance and full coverage of standard specifications.