國 立 交 通 大 學
資 訊 科 學 與 工 程 研 究 所
博 士 論 文
無線區域網路第一層與第二層核心技術
之設計與實現
Design and Implementation of WLAN Layer 1 and
Layer 2 Core Techniques
研 究 生：鄭紹余
指導教授：許騰尹
無線區域網路第一層與第二層核心技術
之設計與實現
Design and Implementation of WLAN
Layer 1 and Layer 2 Core Techniques
研 究 生：鄭紹余 Student：Shau-Yu Cheng
指導教授：許騰尹 博士 Advisor：Dr. Terng-Yin Hsu
國 立 交 通 大 學
資 訊 科 學 與 工 程 研 究 所
博 士 論 文
A Dissertation
Submitted to Department of Computer Science College of Computer Sicience
National Chiao Tung University in Partial Fulfillment of the Requirements
for the Degree of Doctor of Philosophy in
Computer Science July 2011
Hsinchu, Taiwan, Republic of China
無線區域網路第一層與第二層
核心技術之設計與實現
鄭紹余
國立交通大學資訊科學與工程研究所
指導教授：許騰尹 教授
摘要
對於一個支持IEEE 802.11 的接收器來說，最近以來的一個挑戰就是讓接收 器 架 構 越 簡 潔 越 好 ， 譬 如 在 無 循 環 前 綴 單 載 波 分 組 傳 輸(non-cyclic prefixsingle-carrier block transmission, non-CP SCBT))、單輸入單輸出(single-input single-output, SISO)與多輸入多輸出(multi-input multi-output , MIMO)正交頻 分複用(orthogonal frequency division multiplexing , OFDM)間進行有效率的硬體 共享。基於頻率域類比數位轉換器(frequency-domain analog-to-digital conversion, FD-ADC)技術，本論文提出了一個多模接收器在頻率域上去處理所有的數位訊 號，為了要在頻率域回復符號時序(symbol timing)，本論文提出了一個採用了符 號速率循序並用匹配濾波器(matched filter)結果去搜尋的頻率域符號同步器(FD symbol synchronizer)，由模擬與實作結果顯示這個提出的頻率域符號同步器在低 訊雜比下仍然很強健並且在 VLSI 實作上有很低的複雜度。而為了要讓等化器 (equalizer)盡量簡潔，在無循環前綴單載波分組傳輸上，另外也提出了一個採用 了單 FFT 架構以及球面解碼(sphere decoding)演算法的單載波頻率域等化器
(SC-FDE)，因此 IEEE 802.11b 的等化可共用 MIMO-OFDM 收發機中的硬體元件。 除此之外，我們還設計了一個事前修剪的技術去更進一步降低使用空間多工多輸
入多輸出傳輸中信號偵測的複雜度，這個事前修剪的技術利用zero forcing (ZF) 的偵測結果及 Nq-QAM 星座圖上多層次結構的特性去減少傳統 K-best 演算法的 搜尋空間，因此這方法很適合同時擁有K-best 及 ZF 偵測器的接收器。 除了上述的實體層問題外，因為無線高速網際網路(Internet)的存取的增加讓 資料由存取網路(access network)轉傳到網際網路的高速無線後置網路(wireless backhaul network)的需求變的必要，而實務上更高的傳輸率要更高的基地台密 度，因此使得在高速無線後置網路的佈署中，使用基礎網路的架構變的不符成本 效益，在這情況下，IEEE 802.11s 無線網狀網路(wireless mesh network, WMN)
提供一個吸引人的方法來快速且低成本的佈署，在本論文中研發了IEEE 802.11s 無線網狀網路並實際佈署了一個 3x3 的格狀拓樸網狀網路在實驗室及一個跨三 層樓的建築物中，考量到無線網狀功能的可攜性，網狀網路的開發是在一個現成 的商用無線晶片中的純軟體延伸，其中使用模組化軟體設計及不需要高成本硬體 更動，為了要加強傳輸廣播類(broadcast-type)網狀網路控制封包的可信度，數種 廣播策略在實驗室中進行路由重建率、可接受的延遲及通道使用率等評量，對於 網狀網路的佈署上，我們的觀察指出RTS/CTS 可以增加網路吞吐量達到 87.5%， 另外比起使用 IEEE 802.11b/g，用 802.11n 傳輸可在多重資料流(multi-stream)或 多點跳躍(multi-hop)的通訊上能達到更好的公平性(fairness)，在本論文中總結的 網狀網路的實驗觀察希望能提供給要佈署小型或中型室內 IEEE 802.11s 無線網 狀網路的人一些導引。
Design and Implementation of WLAN
Layer 1 and Layer 2 Core Techniques
By Shau-Yu Cheng
Department of Computer Science National Chiao Tung University
Advisor: Terng-Yin Hsu
Abstract
Recently, one of the major challenge for a IEEE 802.11 compatible receiver is to make the receiver architecture as compact as possible, i.e., efficient hardware sharing between non-cyclic prefix single-carrier block transmission (non-CP SCBT), single-input single-output (SISO) and multi-input multi-output (MIMO) orthogonal frequency division multiplexing (OFDM) systems. Based on frequency-domain analog-to-digital conversion (FD-ADC) technology, this dissertation presents a multi-mode receiver to handle all digital signals in frequency domain. A frequency-domain (FD) symbol synchronizer adopting a symbol-rate sequential search with simple matched filter detection is presented to recover symbol timing over the frequency domain. Simulation and implementation results show that the proposed FD symbol synchronizer is robust at low single-to-noise (SNR) and low complexity for VLSI implementations. To make equalizer as compact as possible, a single-carrier frequency-domain equalization (SC-FDE) for non-CP SCBT is proposed with single-FFT architecture and sphere decoding algorithm. Thus, the equalization of IEEE 802.11b can reuse the hardware components in the MIMO-OFDM modem.
Moreover, a pre-pruning scheme is designed to further reduce the complexity of MIMO detection module for MIMO transmission using spatial multiplexing. The pre-pruning scheme reduces the search space of conventional K-best algorithm by using the zero forcing (ZF) detection result and the property of multilevel structure in
Nq-QAM constellation. Hence, it is very attractive for the receivers equipping with
both K-best and ZF detectors.
In spite the issues mentioned above in physical layer, a high rate wireless backhaul network transporting data between the access network and the wired Internet becomes essential due to the increasing of wireless high-speed Internet access. The infrastructure network becomes cost ineffective in the deployment of a high-rate wireless backhaul network due to the higher data rates requires much higher cell densities to realize in practice. Under this situation, IEEE 802.11s wireless mesh network (WMN) can provide an attractive approach for the fast and low cost deployment. This dissertation develops an IEEE 802.11s WMN and then deploys a testbed with 3-by-3 grid topology in both laboratory and field crossing three floors of the building. For the portability of mesh functions, the mesh development is a pure software extension for commercial off-the-shelf WLAN chipsets with modularized software design and without costly hardware modifications. To improve the transmission reliability of broadcast-type mesh control frames, several broadcasting strategies are evaluated based on the routing construction ratio, acceptable latency, and channel utilization in the laboratory testbed. For the WMN deployment, our observations indicate that RTS/CTS can improve throughput by up to 87.5%. Moreover, compared with the IEEE 802.11b/g, 802.11n achieves better fairness for multi-stream or multi-hop communications. The experimental observations of WMN deployment summarized in this dissertation are expected to provide guidance for the small or medium scale indoor IEEE 802.11s WMN.
Acknowledgements
I would like to express my sincere gratitude to my dissertation advisor, Prof. Terng-Yin Hsu, for his kind guidance, assistance, consultation, and encouragement during the course of my graduate study. Whenever I encounter difficulties or feel disappointment for my work, either in the research or in the daily life, he always tries to give me a hand and lead me to an appropriate place.
Also, I would like to thank many of present and former ISIP members: Wei-Chi Lai, Cheng-Yuan Lee, and Kai-Shu Su, whose contributions were instrumental in the development of ideas. Moreover, I will keep in mind for the friendship coming form You-Hsien Lin, Chueh-An Tsai, Shao-Ying Yeh, and Tsung-Yeh He.
The thanks also goes to the Realtek-NCTU jointly research center for its complete research environment and financial support during the several years of my Ph.D study. Particularly, I would like to thank the chief director, Prof. Ying-Dar Lin and the technical adviser, Prof. Shiao-Li Tsao, for their open management and valuable comments. In addition, I never forget the useful cooperation experience coming from Der-Zheng Liu, David Hsu, Shun-Lee Chang, and Jui-Hung Yeh.
Furthermore, I want to share my happiness with my family, especially for my dear parents for their ageless love during my whole life. In addition, I take thanks for the parents of my wife. Without their assistance and blessing, this dissertation might not be worked out and I could not reach where I am now.
Finally, and that is also the most important of all, I have to express my appreciation to my wife, Etta Fan, for her endless love, inexhaustible understanding, and dependable harbor, especially the patience and waiting for so many lonely moments.
Shau-Yu Cheng
Table of Contents
Chapter 1 Introduction...1
1.1 Introduction of WLAN Radio technologies...4
1.2 Introduction of WLAN Relay-based Backhaul Network...6
1.3 Problem Statement and Literature Survey ...9
1.4 Dissertation Overview ...14
Part I PHY Layer: Three Key modules for Multi-mode FD receiver ...17
Chapter 2 Symbol Rate Frame Synchronization with FD-ADC Architecture ...18
2.1 Frequency-Domain Analog-to-Digital Conversion...23
2.1.1 Basic Concept ...23
2.1.2 OFDM Receiver Based on FD ADC...25
2.1.3 Frequency Offset and Phase Noise ...28
2.1.4 The Advantages and Disadvantages of FD ADC ...29
2.2 System Assumptions and Problem Statement...31
2.2.1 System Assumptions ...31
2.2.2 Matched Filter Detection in FD Receiver...33
2.2.3 Problem Statement ...34
2.3 The Proposed FD Symbol Synchronization...36
2.3.1 Sequential Search...36
2.3.2 Complexity Reduction and Performance Enhancement ...39
2.3.3 Algorithm Identification Step ...43
2.5 Architecture and Implementations ...50
2.5.1 Low-Complexity Architecture ...50
2.5.2 Semi-synchronous Clock Generator ...54
2.5.3 Implementations Results...58
2.6 Summary...63
Chapter 3 FD Channel Estimation and Equalization with Single-FFT Architecture for SCBT System ...65
3.1 System Assumptions ...68
3.1.1 System Descriptions...68
3.1.2 Problem Statements ...69
3.2 The Proposed Single-FFT Processes...70
3.2.1 Frequency-Domain Channel Estimator...70
3.2.2 Decision-Feedback Aliasing Canceller ...72
3.3 Performance Evaluations ...74
3.4 Implementation and Complexity...79
3.4.1 Sphere Decoder with SCBT Decoding ...81
3.4.2 Detail VLSI Architecture ...84
3.4.3 Complexity Summary ...91
3.5 Summary...92
Chapter 4 A Cluster-based Pre-pruning Scheme for Low Complexity K-best Algorithm...93
4.1 Background...97
4.1.1 MIMO System Model...97
4.1.2 Multilevel Structure of the N-QAM Constellation ...98
4.2 The Proposed Algorithm ...102
4.2.1 Pre-pruning via Cluster-Based Detection ...103
4.2.2 Detail Matching with K-Best Algorithm ...106
4.3 Simulation Results ...109
4.4 Summary...114
Part II MAC Layer: Development/Deployment of an IEEE 802.11s System... 115
Chapter 5 Design and Implementation of IEEE 802.11s Mesh ... 116
5.1 Network Architecture...118
5.2 Mesh Functions...119
5.3 Design and Implementation Issues of An IEEE 802.11s Mesh ...124
5.3.1 Software Architecture ...124
5.3.2 Transmission Strategies for Mesh Broadcast-Type Control Frames .127 5.4 Development/Testbed Platforms...129
5.5 Experiment of Broadcasting Strategies...130
5.5.1 Experiment configuration ...130
5.5.2 Evaluation of Broadcasting Strategies ...131
5.6 Summary...138
Chapter 6 Indoor Deployment of IEEE 802.11s Mesh Networks ...140
6.1 Related Works: Effect of RTS/CTS and Rate Adaptation...143
6.2 IEEE 802.11s Testbed ...147 6.2.1 Experiment Configuration ...147 6.2.2 Experiment Methodology ...150 6.3 Experimental Results ...151 6.3.1 RTS/CTS...152 6.3.2 IEEE 802.11n vs. 802.11b/g ...157
6.3.3 Beacon Interval ...159
6.4 Lessons and Guidelines...162
6.5 Summary...167
Chapter 7 Conclusion ...168
7.1 Summary...168
7.2 Future Work ...171
List of Figures
Figure 1-1: IEEE 802.11s mesh networks architecture...8
Figure 1-2: Block diagram of the conventional 802.11 b/g/n/ac multi-mode receiver. ...12
Figure 1-3: The proposed FD receiver architecture for SCBT and OFDM systems,. .12 Figure 1-4: (a) the conventional WLAN infrastructure. (b) relay-based WLAN infrastructure. ...14
Figure 2-1: The block diagram of OFDM receivers ...21
Figure 2-2: FD-ADC block diagram...24
Figure 2-3: Block diagram of FD-ADC based OFDM receiver ...28
Figure 2-4: Definitions of data patterns. ...33
Figure 2-5: Example of the trellis diagram with path length K=3. ...43
Figure 2-6: The flowchart of the proposed sequential search...45
Figure 2-7: Error probability of symbol synchronization in IEEE SISO and MIMO frequency-selective fading with a CFO of 40 ppm...48
Figure 2-8: Error probability with different CFOs in IEEE and TGn frequency-selective fading with 50-ns and 100-ns RMS delay spreads ....49
Figure 2-9: Performance comparisons of Pf (0.5) in channel II...49
Figure 2-10: VLSI architecture and structure of the proposed sequential searcher for FD symbol synchronization in FD MIMO-OFDM modem...55
Figure 2-11: Structures and behavior of the proposed semi-synchronous clock generator (SSCG)...58 Figure 2-12: Post-layout simulation of the proposed SSCG with 20 MHz input clock
and 82.1 MHz output clock...60
Figure 3-1: The block diagram of FD-ADC based OFDM receivers ...68
Figure 3-2: Power delay profiles for IEEE and JTC channel models...76
Figure 3-3: Simulation of the linear (8, 4) code in IEEE and JTC fading ...78
Figure 3-4: Simulation of CCK in JTC fading － office A, residential C and office B with residual CFOs. ...78
Figure 3-5: SDR platform for SCBT measurements...79
Figure 3-6: Block diagram of the proposed FD-CE. ...80
Figure 3-7: Block diagram of the single-FFT SC-FDE with DF-AC...80
Figure 3-8: Tree structures of SD search in MIMO-OFDM and DSSS-CCK modes..84
Figure 3-9: Detail architecture and complexity of a 4x4 MIMO-OFDM modem with the proposed SC-FDE ...88
Figure 3-10: Complex multiplier with an additional conjugation output ...89
Figure 3-11: The rule of signal multiplications of G. ...89
Figure 3-12: CCK mapping on four sub-sets for parallel SD searching...90
Figure 4-1: The Multilevel structure of 64-QAM constellation. ...100
Figure 4-2: The cluster-based search tree corresponding to the multilevel structure of 64-QAM constellation ...103
Figure 4-3: The behavior of cluster-based detection where the black points are the candidate points at that stage. ...109
Figure 4-4: CDF of error distance between ZF estimation result and its sliced one .112 Figure 4-5: PER performance with 4x4 64-QAM in IEEE TGn E channel with 100-ns. ...112
Figure 4-6: Comparison of complexity in terms of the number of search points for 4x4 64-QAM K-best search tree in IEEE TGn E channel with 100-ns ...113
Figure 5-1: A logical view of a wireless mesh network and its mapping to IEEE
802.11s ...119
Figure 5-2: The proposed software architecture and the control/data plane flows....127
Figure 5-3: Illustration of the testbeds. ...131
Figure 5-4: The relation between PER and data rate in our experimental deployment ...134
Figure 5-5: Experiments for routing establishment. ...138
Figure 6-1: Topologies of the testbed in laboratory and field...149
Figure 6-2. Pictures of the testbed in laboratory and field...150
Figure 6-3: Effects of enabling RTS/CTS (single-stream, chain topology)...155
Figure 6-4: Comparison of IEEE 802.11b/g/n rates (single-stream, chain topology). ...155
Figure 6-5: Effects of setting MPP’s location at corner in the field (downlink multistream,grid topology). ...156
Figure 6-6: Effects of setting MPP’s location at center in the field (downlink multistream, grid topology). ...156
Figure 6-7: Channel capture effect of IEEE 802.11b/g rates in the field (uplink multistream, chain topology). ...159
Figure 6-8: Comparison of the total throughput when setting MPP at center in field (multi-stream, grid topology)...162
List of Tables
Table 1-1 Summary of IEEE 802.11 PHY related standards ...5
Table 2-1 Comparisons of Symbol Synchronization Methods in IEEE 802.11a Systems ...50
Table 2-2 Summary of the Proposed Implementation ...60
Table 2-3 Comparisons of Symbol Synchronizers...63
Table 3-1 The VLSI complexity of a 4X4 MIMO OFDM modem...91
Table 4-1: Performance and Complexity Comparisons ...113
Table 6-1: Summary and Comparison on the Previous Work...146
List of Acronyms
The following acronyms are used through this dissertation.
ACK acknowledgment
ACI adjacent channel interference
ADC analog-to-digital conversion
AFC automatic frequency control
AGC auto gain control
AODV Ad Hoc On-Demand Vector
AP access point
ARF automatic rate fallback
AWGN additive white Gaussian noise
B&B branch and bound
CCK complementary code keying
CFO carrier frequency offset
CFR channel frequency response
CMOS complementary metal-oxide-semiconductor
CoMP coordinated multipoint transmission
CP cyclic prefix
CPU central processing unit
CTS clear to send
DF-AC decision-feedback aliasing canceller
DFT discrete Fourier transform
DHCP dynamic host configuration protocol
DSP digital signal processor
DSSS direct sequence spread spectrum
DST destination node
FD frequency-domain
FD-ADC frequency-domain analog-to-digital conversion
FD-CE FD channel estimator
FFT fast Fourier transform
FPGA field programmable gate array
HWMP hybrid wireless mesh protocol
IBI interblock interference
IFS inter-frame space
IEEE Institute of Electrical and Electronics Engineers
JAER just-acceptable error rate
JTC Joint Standards Committee
LUT look up table
LTE-A Long Term Evolution-Advance
MAC media access control
MANETs mobile ad hoc networks
MAP mesh access point
MCN multihop cellular network
MIMO multi-input multi-output
ML maximum-likelihood
MLD maximum likelihood detection
MP mesh point
MPP mesh portal
MU-MIMO multi-user MIMO
OFDM orthogonal frequency-division multiplexing
OLPC One Laptop per Child
PED partial Euclidean distance
PAPR peak-to-average power ratio
PER packet error rate
PHY physical
PREP path response
PREQ path request
QPSK Quadrature phase-shift keying
RMS root mean square
ROM read-only memory
RTS request to send
SC single-carrier
SCBT single-carrier block transmission
SC-FDE single-carrier frequency-domain equalization
SD sphere decoder
SDR software-defined radio
SISO single-input single-output
SM spatial multiplexing
SNR signal-to-noise ratio
SRAM static random-access memory
SRC source node
SSC semi-synchronous clocker
SSCG semi-synchronous clock generator
STA station
TCP Transmission Control Protocol
TD time-domain
TDC time-to-digital converter
TSMC Taiwan Semiconductor Manufacturing Company
UCG unified channel graph
VHT very high throughput
VLSI very large scale integration
WDS wireless distribution system
WLAN wireless local area network
WMN wireless mesh network
Chapter 1
Introduction
In recent years, broadband wireless communications have been considered as the key enabling technology for innovative future consumer products. Moreover, applications like multimedia sharing and cloud computing grow the demands of higher data rate transmission in wireless communication system. Being the most popular and successful wireless broadband network standard, IEEE 802.11 [1] wireless local area network (WLAN) has be adopted in client devices over several hundreds million worldwide. WLAN radios are appearing not just in PCs and laptops, but in equipment as diverse as mobile phones/pads, security cameras and home entertainment equipment, etc. With large, diverse and rapidly growing WLAN installed bases, the research and development investment in WLAN technology becomes very important.
IEEE 802.11 is a set of standards consisting of the base standard [1] defined in 1997 and a series of amendments started from 1999 [2]-[5]. The evolution of IEEE 802.11 is still ongoing. The base standard is originally designed for 1 and 2 Mbps throughput and is now being upgraded to support 600 Mbps in 802.11n and is being considered as a high-throughput (up to several Gbps) wireless interface for the multimedia data service in the scope of the next-generation of wireless systems. To support higher data rate with backward compatibility during the evolutions of amendments, IEEE 802.11 adopts two radio modulation schemes: non-cyclic prefix single-carrier block transmission (non-CP SCBT) and orthogonal frequency division multiplexing (OFDM). In addition, the radio system is evaluated from single-input single-output (SISO) to multi-input multi-output (MIMO) for the transmission rate over hundreds Mbps. These advances of radio technologies make a high demand for multi-mode receiver design to support non-CP SCBT, SISO OFDM and MIMO OFDM transmission in modern WLAN applications.
It is also recognizes that the later amendments usually provide higher data rate with shorter transmission distance. For example, typical access points can provide 54 Mbps data rates only up to tens of feet whereas they can extend 11 Mbps data rates up to hundreds of feet. This is because these higher rates require higher levels of signal-to-noise ratio (SNR) at the receiver. In other words, these higher data rates will
require much higher cell densities to realize in practice. To construct a high-data rate network with conventional infrastructure architecture, the deployment cost could be unacceptable. Therefore, the relay-based network architecture is essential for current and next generation WLAN applications. Currently, the relay-based network can be achieved by deployment with Ad hoc or wireless distribution system (WDS) modes. However, Ad hoc can not let client devices connect to Internet while WDS loses the flexibly of deployment with manually configurations and static network topology. To overcome those drawbacks, The 802.11s Task Group is now working on an infrastructure mesh amendment to allow 802.11 access points or cells from multiple manufacturers to self-configure into multi-hop wireless topologies. Example usage scenarios for mesh networks include interconnectivity for devices in the digital home, unwired campuses, and community area networks or hotzones.
To build a low cost, high rate, high flexibility WLAN network, there are still many challenges issues need more research investment. In this dissertation, we focus on the design of IEEE 802.11 compatible multi-mode receiver and development of IEEE 802.11s wireless mesh works.
1.1 Introduction of WLAN Radio technologies
The radio technologies of IEEE 802.11 family include two modulation schemes: non-CP SCBT and OFDM modulations. The non-CP SCBT is a type of single-carrier (SC) modulation, which is adopted in the original (legacy) standard 802.11-1997 and the first amendments IEEE 802.11b [2]. The non-CP SCBT modulation owns the good characteristics of friendly front-end implementation and insensitive impact on CFO [6]. However, non-CP SCBT modulation provides lower frequency spectrum efficiency than OFDM modulation scheme and complex design of equalizer in time domain. Therefore, IEEE 802.11a [3] provides an alternative approach with OFDM modulation to provide high data rate up to 54Mbps in the 5GHz unlicensed bands. The 2.4GHz OFDM system is first started from 802.11g [4] in 2003, which is a backwards-compatible extension to the 802.11b standard allows data rates up to 54 Mbps through use of OFDM or complementary code keying (CCK) modulation in the 2.4GHz band. The advantages of OFDM modulation is well known to have high spectrum efficiency and robust to multipath fading (i.e., frequency-selective) channels. However, the requirement of the front-end for OFDM is stricter than non-CP SCBT modulation because its nature of high peak-to-average power ratio (PAPR) and sensitive to carrier frequency offset (CFO) [7]. In order to support the data rate up to
few hundreds megabits per second, OFDM-based MIMO spatial multiplexing (SM) technique is introduced in the latest amendment IEEE 802.11n [5]. MIMO system can significantly improve the spectrum efficiency and achieve a much higher transmission rate than a SISO system. In addition, combining OFDM and MIMO can simplify the implementation of MIMO detection in receiver architecture. Table 1-1 summarizes the radio technologies in IEEE 802.11 legacy and its amendments.
Table 1-1 Summary of IEEE 802.11 PHY related standards
802.11 std Release Freq. (GHz) BW (MHz) Max. Data rate (MBit/s) Max # of stream Modulation Legac y [1] Jun 1997 2.4 20 2 1 non-CP SCBT (DSSS)
a[3] Sep 1999 5 20 54 1 OFDM
b[2] Sep 1999 2.4 20 11 1 non-CP SCBT (DSSS/CCK) g[4] Jun 2003 2.4 20 54 1 OFDM, non-CP SCBT (DSSS/CCK) 20 288.8 n[5] Oct 2009 2.4/ 5 40 600 4 OFDM
1.2 Introduction of WLAN Relay-based Backhaul
Network
Infrastructure-based wireless networks provide convenient access to the Internet, and are becoming increasingly popular in spite of their costly wired deployment. On the other hand, mobile ad hoc networks (MANETs) eliminate the need for infrastructure, decreasing deployment time and alleviating network construction costs. However, having routing functions on all nodes in a MANET complicates the design of networking devices [8]. The fact that MANET usage is typically limited to military and specialized civilian applications also hinders its growth [9]. By combining an infrastructure-based wireless network and a MANET, a wireless mesh network (WMN) presents a low-cost and fast-deployment solution compared to an infrastructure-based wireless network, and a reliable and less complicated solution compared to a MANET. A WMN is similar to a multihop cellular network (MCN) [10], which has been proved to improve aggregated throughput linearly due to spatial division.
In [8], the authors classify the WMN architecture into three types: infrastructure, client, and hybrid. An infrastructure WMN is organized as a hierarchical network, functionally consisting of mesh gateways, relay points, access points, and terminals. A
mesh gateway is a device capable of bridging the wireless mesh and wired infrastructure. A relay point implements a routing algorithm to relay packets in a mesh. To support non-mesh terminals, the mesh uses access points to bridge the WMN and non-mesh terminals. In a client mesh there is no gateway and non-mesh terminal because this kind of mesh emphasizes flat peer-to-peer communications. A hybrid mesh includes both infrastructure and mesh terminals that provide interfaces for end users and mesh routing capability.
Diverse mesh architectures result in various usage scenarios [8]-[9], and a considerable number of challenges for designing and realizing a WMN [8], [11]–[13]. Industrial organizations have also prepared standards and recommended practices for existing wireless technologies, such as IEEE 802.15.5 for low-rate wireless personal area networks (WPANs). Among these efforts, IEEE 802.11s [14], which defines a WLAN mesh using IEEE 802.11 MAC and PHY layers, is one of the most active standards and has increasing commercial opportunities. Figure 1-1 shows the architecture of mesh IEEE 802.11s network. The IEEE 802.11s standard uses the IEEE 802.11 MAC and PHY, and has drawn numerous research and commercial interests in recent years. Unlike ad hoc networks and sensor networks that are motivated by military or crisis applications, WMNs introduce the commercial applications such as the last-mile wireless access or home wireless networking.
WMNs can largely reduce the cost and complexity of network deployment by multi-hop relaying. The 802.11s WMN is composed of mesh portals (MPPs), mesh points (MPs), mesh access points (MAPs), and wireless stations (STAs). An MPP connects the WMN and the Internet. MPPs, MAPs, and MPs communicate with one another via the wireless medium and form a wireless backbone network. Then, STAs that associate with an MAP can communicate with the other STAs or access the Internet.
1.3 Problem Statement and Literature Survey
From the Table 1-1, it is clearly that the IEEE 802.11 receiver needs to be multi-mode design to support both non-CP SCBT (802.11b) and OFDM (802.11 a/n) modes. In practice, most of the implementations of multi-mode receivers use dedicated hardware to support both non-CP SCBT and OFDM systems as shown in Fig 1-2. However, those dedicated modules are difficult to be shared and merged for different modes, which leads high cost and large die size in implementations. For example, the different packet format and coding scheme between non-CP SCBT and OFDM may lead difficult hardware sharing in synchronization, equalization and data decoding modules. Moreover, the less hardware sharing introduces complex control path and cost inefficiently in receiver design. Although the software-defined radio (SDR) solution [15]-[16] implemented with digital signal processors (DSPs), central processing unit (CPU) and field programmable gate array (FPGA) provides the flexible solutions for multi-mode applications, the designers still need to consider the complex data path introduced by different signal formats, i.e. time-domain signals in the pre-fast Fourier transform (pre-FFT) phase and frequency-domain signals in the post-FFT phase. The complex data path might cause bad influences in the pipeline design, e.g., data hazards, hard to identify pipeline stages and complex instruction sets,
which results in higher clock rate and more complex control circuit in hardware implementation [17]. To avoid the switch of data path between time and frequency domains, handling all signals in frequency domain after conventional ADC followed by FFT is investigated in the research [18].
Recently, the concept of FD receiver with FD-ADC technology [19]-[23] is proposed as a potential solution for a multi-standard receiver. Unlike the typical OFDM receiver architecture based on conventional analog-to-digital conversion (ADC) and time-domain (TD) synchronization followed by FFTs and data decoding in frequency domain, a FD receiver directly handles all signal processes in frequency domain. To reduce signal formats transformation between domains and increase the hardware sharing between two modes, this dissertation adopts the FD receiver architecture to support for WLAN multi-mode applications.
Based on the FD-ADC technology, Fig. 1-3 shows the proposed multi-mode FD receiver architecture. To realize this FD receiver, two problems need to be overcome: symbol synchronization and non-CP SCBT equalization over frequency domain. Although a number of methods for symbol synchronizations [24]-[32] have been proposed, they are all developed based on TD-ADC based receiver which decides the symbol boundary by a sliding window before FFT. However, FD ADC transforms a segment of continuous time-domain signal to a set of frequency coefficients. Since
those synchronization algorithms performs sample-by-sample search for correlation peak within the sliding window, they can not function with FD-ADC technology properly. Therefore, the synchronization problem of the FD receiver architecture is to detect symbol boundary with frequency coefficients of FD-ADC outputs. For OFDM-based packets, the channel frequency response (CFR) can be obtained from the frequency response of received OFDM symbol divided by the frequency response of the frequency domain training symbol (long training symbol) due to the existence of cyclic-prefix (CP) [33]. For equalization of non-CP SCBT over frequency domain, a single-carrier frequency-domain equalizer (SC-FDE) is developed to eliminate FFT aliasing without a circular property in some approaches [34]-[35], e.g., overlap-and-save and overlap-and-add methods. Yet, additional DFT units were included—hardware complexity of the multi-mode FD receiver may increase significantly. Thus, one of the major challenge for multi-mode integrations is to make equalizers as compact as possible, i.e., consolidation of non-CP SCBT, SISO OFDM and MIMO OFDM. Finally, it is also important to reduce the complexity of MIMO detection module for a low cost receiver design. Numbers of methods have proposed in the literature [36]-[43]. However, those methods include some of the following drawbacks, i.e., poor performance, unacceptable complexity, not favored for VLSI implementation. Therefore, the design of a low complexity MIMO detection method
is investigated in this dissertation.
Figure 1-2: Block diagram of the conventional 802.11 b/g/n/ac multi-mode receiver.
Modern output Eq ua liz er w ith MIMO d e te cti o n
Figure 1-3: The proposed FD receiver architecture for SCBT and OFDM systems, where dash lines are the control signals, thin solid lines are the synchronization and channel estimation paths, and bold solid lines represent the data paths.
Besides the multi-mode integration issue, another important issue for modern WLAN applications is to provide cost effective and high rate networks over a wider coverage area. From radio technique perspective, MIMO scheme can provide the high date rate transmission yet it also requires good link quality which means the coverage of high rate transmission is very small. As shown in Fig. 1-4, the coverage can be extended by deploying an IEEE 802.11s WMN. Therefore, the knowledge related to the development and deployment of an IEEE 802.11s system becomes emergency topic for WLAN research filed recently. Although Prior studies, such as the mesh on XO-laptop for One Laptop per Child (OLPC) [44] and the open80211s project for Linux, has evaluated the network performance of the IEEE 802.11s mesh, few studies examine system architectures and mesh stability. Moreover, conventional mesh deployment [8], [45]-[50] focuses on the outdoor environment, which regards the WMNs as backbone networks. For WLAN application, the backhaul network is usually considered as an indoor deployment. However, indoor and outdoor WMNs possess distinguishable attributes and limitations. To the best of our knowledge, only a little previous work focuses on indoor WMNs [45]-[46].
(a)
(b)
Figure 1-4: (a) the conventional WLAN infrastructure. (b) relay-based WLAN infrastructure.
1.4 Dissertation Overview
The designs of FD receiver architecture and its key components of synchronizer and equalizer are discussed in Chapters 2, 3 and 4 while the design, implementation and deployment of IEEE 802.11s mesh network are discussed in Chapters 5 and 6. Details of the research contributions in each chapter are as follows.
robust symbol synchronization in a 4x4 FD MIMO-OFDM modem. Based on the FD ADC, the searcher finds symbol boundary at symbol-rate and then the hardware sharing can successfully accomplish low-complexity implementations Moreover, this chapter also includes the design and implement of a very low complexity clock generator, called semi-synchronous clock generator (SSCG), to provide a high-speed clocking.
In Chapter 3, we design and implement a single-FFT approach SC frequency-domain equalization (FDE) approach for non-CP SCBT system. This approach makes IEEE 802.11b data packet can be equalized in frequency domain for FD receiver architecture. Moreover, the approach can be implemented by sphere decoder (SD) algorithm [51] which is widely adopted in MIMO-OFDM modems. Thus, the efficient hardware sharing of equalization module in FD receiver can be achieved.
In Chapter 4, we propose a pre-pruning scheme to reduce the search space of
K-best algorithm. Based on the property of multilevel structure in Nq-QAM
constellation, the scheme adopts a cluster-based search to find the reliable constellation points according to the ZF detection results. Compared to the conventional K-best algorithm [40] with the same K value, the proposed work achieves the same performance with fewer search nodes (67.6%~77.21% of the
conventional K-best algorithm). Hence, the proposed pre-pruning scheme is attractive for those receivers equipping with both K-best and ZF detectors.
In Chapter 5, we develop a prototype of WLAN mesh network based on the draft D2.03 of IEEE 802.11s, and port the mesh functions to the commercial off-the-shelf WLAN chipsets such as Realtek RTL8186 and Realtek RTL8192SE+RTL8196B platforms. Then, we evaluate the design strategies for broadcast-type mesh control frames.
In Chapter 6, we present a two-phase deployment plan with 3-by-3 grid topology to establish the benchmark of the implemented mesh network in last chapter. In the first phase, the testbed was deployed in a laboratory to evaluate its basic capacity and performance for a dense deployment. In the second phase, the grid WMN was deployed in the sixth to eighth floors of the Microelectronics and Information Systems Research Center (MIRC) at National Chiao Tung University. The experiments mainly focus on the effects of different IEEE 802.11 settings, including RTS/CTS, links with IEEE 802.11n or IEEE 802.11b/g, and beacon interval. Finally, the experimental observations are summarized to provide guidance for small or medium scale indoor 802.11 WMNs.
In Chapter 7, we draw some conclusions of this work and outline possible topics for further research.
Part I PHY Layer:
Three Key
modules for Multi-mode FD
Chapter 2
Symbol Rate Frame
Synchronization with FD-ADC
Architecture
Based on the FD ADC, this chapter builds a low-complexity sequential searcher for robust symbol synchronization in a 4x4 frequency-domain (FD) multiple-input multiple-output orthogonal frequency-division multiplexing (MIMO-OFDM) modem. The proposed scheme adopts a symbol-rate sequential search with simple matched filter detection to recover symbol timing over the frequency domain. Simulation results show that the detection error is less than 2% at SNR ≦ 5 dB. And performance loss is not significant when carrier frequency offset (CFO) ≦ 100 ppm.
Using an in-house 65-nm CMOS technology, the proposed solution occupies 84.881K gates and consumes 5.2mW at 1.0V supply voltage. This work makes the FD ADC more attractive to be adopted in high throughput OFDM systems.
Symbol synchronization is one of the essential processes to correctly detect symbol boundary in wireless packet-based receiver. The incorrect boundary disarranges the amplitude as well as the phase of the received signal. Moreover, it causes the error position of FFT windows within orthogonal frequency-division multiplexing (OFDM) symbols, introducing intersymbol and interference (ISI) and significant performance degradation during data demodulation. To guarantee the system performance, the reliability of symbol synchronization is therefore critical to OFDM-based systems.
Based on the receiver architecture using conventional ADC (Fig. 2-1(a)), numbers of symbol synchronizers have been revealed in the literature. By exploiting the known structure of the preamble or the cyclic prefix, a number of methods for OFDM symbol synchronization have been proposed in the literature. The autocorrelation-based method [24] is presented to find the peak of correlation between the previous and current received symbols. The implementation is rather simple but the correlation peak of the timing metric exhibits a plateau which causes large variance for the timing estimator [25]. The cross-correlation method [26] uses a
“clean” copy of the preamble to form a matched filter which is applied to the received symbol. In spite of the advantage of more accurate timing detection, more sensitive to the carrier frequency offsets [27] and higher implementation complexity of the matched filer [28] are the main drawbacks. The double correlation metric [29] is presented to develop a reliable frequency and time synchronization scheme. However, its complexity is also doubled. Although those correlation-based methods mentioned above are suitable for hardware implementations, the performances are easily affected by multipath fading [30]. To overcome this limitation, the generalized Akaike information criterion (GAIC) [31] and [32] are presented to jointly estimate the channel order and establish the symbol timing. Both of the two method provide a pretty good performance when SNR ≥ 10 dB. However, the two methods are more complex than the correlation-based methods.
Recently, frequency-domain analog-to-digital conversion (ADC) for OFDM applications [19]–[23] has been reported to gain several advantages, e.g., the relaxation of ADC requirements, robustness in narrow band interference (NBI) suppression. Such structure is also named FD receiver [20] since all baseband processes are operated in the frequency domain as shown in Fig. 2-1(b). To make an FD receiver function well in practices, a reliable FD symbol synchronization is necessary. Yet, this technique is difficultly ported to most broadband standards, e.g.,
IEEE 802.11 and 15 series, because those algorithms perform sample-by-sample search for correlation peak within the sliding window. In contrast, FD ADC transforms a segment of continuous time-domain signal to a set of frequency coefficients. Therefore, a symbol rate symbol synchronizer that processes signals in the frequency domain is needed for the FD receiver architecture.
Figure 2-1: The block diagram of OFDM receivers: (a) using the conventional ADC, and (b) using the FD ADC.
To the best of our knowledge, so far no FD symbol synchronizer has been built for FD-OFDM receivers. This work develops a low-complexity sequential searcher in
a 4x4 MIMO-OFDM modem with IEEE 802.11a/n [3], [5] compatibility for the robust symbol synchronization using FD-ADC techniques. Only simple matched filter detection with TD preambles is required to detect the correct symbol timing over the frequency domain. Moreover, an all-digital semi-synchronous clock generator (SSCG) is proposed without PLLs to make high-speed inner clocking much easier in VLSI implementations, so that a shared architecture can work efficiently. By using an in-house 65-nm CMOS library, our all-digital SSCG only requires 3.03 gates. To build in a 4x4 FD MIMO-OFDM modem, the total gate count of the proposed sequential searcher is 84.881K and the power consumption at 1.0 V supply voltage is 5.2 mW. In conclusion, the proposed sequential searcher for robust symbol synchronization is attractive to very high throughput (VHT) wireless LAN (IEEE 802.11 ac/ad) [52] and Multi-Gbps WPAN (IEEE 802.15.3c) [53] specifications implemented using FD-ADC techniques.
The organization of this chapter is described as follows. Section 2.1 introduces the technology of FD ADC. Section 2.2 depicts the system model and the problem statement. Section 2.3 addresses the algorithm of the proposed sequential searcher for FD symbol timing synchronizations. Section 2.4 shows the performance evaluations of SISO- and 4x4 MIMO-OFDM systems and also gives a comparison with other methods. Section 2.5 then presents the low-complexity VLSI architecture and
implementation results. Finally, conclusions are given in Section 2.6.
2.1 Frequency-Domain Analog-to-Digital Conversion
2.1.1 Basic Concept
Fig. 2-2 displays the block diagram of FD ADC. FD ADC projects the received signal ( )
r t over the complex exponential functions to a set of Nb basis functions such that
the frequency coefficients R
n can be obtained via sampling of thecontinuous-time signal spectrum at the frequencies
Nb01n n F .
2 0 c n T j F t n r t e dt
R (2.1)where n0Nb 1. Each R
n outputted by the sampler is then quantized by an individual quantizer Qn with associated bit resolutions to generate a digital word
n Qn
n
R R . To avoid discrete-time aliasing, the frequency sample spacing
1 n n
F F F
needs to follow F
1 /Tc
, and the optimal number of frequencycoefficients necessary to provide fully sampling over the signal spectrum of
bandwidth W should be larger than the time-bandwidth product WTc . Thus, the
condition of the optimal number of frequency coefficients Nopt is given by
/
opt c
where denotes ceil function. Clearly, the FD ADC provides N frequency samples
every Tc seconds. Compared with conventional ADC operated at 1 /T sampling s
speed (T is the sample duration), FD ADC samples with the relaxed rate of a factor s
/
c s
T T at the cost of Nb parallel devices (i.e. Nb multipliers, integrators and ADCs in
Fig. 2-2). 0 2 j F t e 1 2 j F t e 2 2 j F t e 1 2 Nb j F t e
0 c T dt
0 c T dt
0 c T dt
0 c T dt
0 Q
1 Q
2 Q
1 N Q
r t
0
R
1
R
2
R
Nb 1
R
0
R
1
R
2
R
Nb-1
R2.1.2 OFDM Receiver Based on FD ADC
For an OFDM system, Tc can be set to the OFDM symbol period Tsymbol, which
makes Nb equal to the number of subcarrier NS. However, it might be infeasible to let
the conversion-time Tc equal to the duration of OFDM signal Tsymbol if the OFDM
system employs large number of subcarrier. The system becomes impractical due to the complex implementation of too many parallel devices. To reduce the number of coefficients N used in FD-ADC based OFDM systems, the method of segmentation of the OFDM signal is investigated in research [20], which discusses the case of
c symbol s
T T M (Ms is the segmentation number). Let OFDM signal x(t) define as
1 2 0 , 0 s s N j f t s s x t a e t T
(2.3)where a is the data modulated at s-th subcarrier tone. In order to reflect the effect of s
segmenting the signal x(t) into Ms time-slots, the following window signals is defined
as
1,
1
0, elsewhere c c m mT t m T w t (2.4)
m m x t w t x t (2.5)where m= 0,…, Ms-1. Then, the Fourier transform of xm
t can be derived as [20]
1 2 1 0 sin = s c s m m N c s j T m F f s s s X F x t T F f a e F f
F (2.6)where F
means the operation of Fourier transform. To detect the transmittedsymbols 1
0
|Ns
s s
a in the receiver side, the matched filter scheme can be used in time
domain as
0 * | = s s t T T s a r t p t r p T d
(2.7)Let the channel frequency response is H f
and assume the channel is flat withineach sub-band. Then, the matched filter impulse response is given by
* * 2 * = s s s s j f t s p t H f x T t H f e (2.8) where
j2 f ts sx t e is the carrier at frequency f . Let s p Ts
t
is
* sg t . The
segmented signals can be expressed as
m m r t r t w t (2.9)
, s m s m g t g t w t (2.10)Then, the a can be rewritten as s
1 1 * 0 1 * , 0 = s c c s M m T s mT s m M m s m m a r g d r g d
(2.11)To express the matched filter operations in the frequency-domain, the above equation can be applied Parseval’s theorem leading to [20]
1 * , 0 1 * , 0 1 1 * , 0 0 = s s s b M s m s m m M m s m m M N c m n s m n m n a r g d R F G F dF F R F G F
(2.12) where
1 0 |Nb m n R F and ,
|Nb01 s m n nG F are the samples from the spectrum of rm
tand gs m,
t , respectively. A good approximation of the above equation can beobtained if the conversion frequency spacing ΔFc avoids discrete time aliasing and samples are taken only in the frequency band of interest where most of the signal energy is concentrated. If the number of spectrum samples Nb satisfies Equaltion (2.2)
which avoids discrete-time aliasing and makes the error in (20) negligible, then the probability of error associated with this receiver will be the same probability of error
of a conventionally implemented OFDM system [20]. In addition, *
, s m n
G F needs to
be estimated from channel frequency response H f
, which is can be estimated by[20]
1 * , , 0 1 2 , 0 ˆ I s i s i i s I s i i a a H f a
(2.13)where I means constant channel blocks number. Then Gs m,
Fn is derived as [20]
, , 2 1 sin e c s s m s m c s j T m F f s s G F g t T F f H f F f F (2.14)BandPass Filter Matched Filter 0 Matched Filter 1 Matched Filter Ns-1 0 m R F 1 m R F
b1
m N R F 0 ˆa 0 ˆa ˆ s N a
r tFigure 2-3: Block diagram of FD-ADC based OFDM receiver [20]
2.1.3 Frequency Offset and Phase Noise
Let
t 2 ft and n
t 2f tn n where f and f are thefrequency offsets. In order to investigate the effect of frequency offset and phase noise on the performance of the FD receiver, the received signal is rewritten as
1
2
0 s s N j t j f t s s s r t a H f e e z t
(2.15)and the local oscillators
10 |Ns n n O t are defined as
j2 F tn jn t n O t e e (2.16)Then, the frequency samples
10
|Nb
m n n
1 2 0 2 + s n n s n n m n m n N j t t j F f t s s m s j t j F t m R F r t O t dt H f a e w t e dt z t e e dt
(2.17)when only frequency offset is taken into account, the above expression reduces to
1
, 0 s n N j m n s s s m n n m n s R F H f a e G F f f dt Z F
(2.18) where
j2 F tn jn t m n m Z F z t e e dt
Thus, frequency offset introduces a constant
complex phase equal to
n
and an offset in the frequency sample equal to
fn f
. The frequency estimation in FD ADC is detailed in the literature [21].2.1.4 The Advantages and Disadvantages of FD ADC
As the bandwidth of the OFDM signal is increased to achieve higher data rates or to accommodate more users, the time-domain sampling approach of the conventional OFDM receiver poses serious challenges to the sampling speed requirement of the ADC. The FD ADC approach has the advantage of sub-Nyquist sampling rate, enabling parallel digital signal processing, flexibility, and scalability in receiver design, making it an attractive option for applications of high speed multi-mode wireless communication systems.
Sampling at frequency domain provides the several main advantages: 1) relaxation of ADC sampling rate; 2) possibility of optimally allocating the available
number of bits for ADC; 3) easily suppression of narrowband interference. Due to the lower sampling rate, FD ADC is more tolerant to the front-end distortions, e.g., clock jitter [22], carrier frequency offset and phase noise [20]. For example, the research [22]
shows that FD-ADC receiver can tolerate up to 5-psrms clock jitter while the
conventional ADC requires 0.5-ps clock jitter at the same specifications.
Both of the analog front-end and digital complexities of FD-ADC approach are marginally higher than that of the conventional approach of FFT with time-interleaved ADC [21], [22]. However, the complexity overhead is not a critical drawback, and it is justified by the critical savings in the clock generation, routing, and driving circuits [22]. There are two main drawbacks for FD ADC architecture. The first one is the noise added by the Nb filters could degrade the overall receiver performance [22]. The
second drawback is that too little literature can be referred for the issues of synchronization and distortion compensations when people considers to build a practical OFDM system based on FD-ADC technology. It is because many synchronizations and distortion compensations processes, including sample and symbol timing recovery, and compensation of I/Q imbalance, are developed in time-domain and may fail to work in frequency domain. Currently, the designs of multicarrier and direct spread systems based on FD-ADC architecture can be referred to the research [19].[21] and [23], respectively.
2.2 System Assumptions and Problem Statement
2.2.1 System Assumptions
The packet structure, depicted in Fig. 2-4, is assumed to contain several preambles without any guard interval (GI) in the packet header. The preamble and its cyclic shift versions are also assumed to have good autocorrelation and low cross-correlation with other shift versions. Let the preamble be denoted as c
c
0 , 1 , ,c c L
1
withL samples length. Then, the baseband-equivalent model of the preamble is given by
1 0 ( ) ( ) ( ) L s T i c t c i t iT f t
(2.19)where is the denotes continuous time convolution, ( )t is a function of BPSK
modulation, and Ts is the sample period of preambles. To be consistent with the
spectrum mask, c t( ) is passed through the transmission filter fT(t) (shaping filter). The duration of one preamble Tp, is derived from Ts by Tp=LTs. The signal c t( ) is
then transmitted through a multipath frequency-selective fading channel h t( ), e.g., IEEE random phase and Rayleigh fading SISO channel [55] and IEEE TGn MIMO channel [56].
At the receiver, the signal is first passed through the RF filter and then down converted into baseband signal. After the automatic gain control (AGC), the baseband signal of received preambles is given by
2( ) ( ) ( ) ( ) j fCFOt ( )
R
r t t c t h t f t e z t (2.20)
where
t is an AGC compensated error, fCFO is the carrier frequency offset,( )
w t represents additive white Gaussian noise (AWGN), fR(t) is the lowpass
equivalent receiver filter response, and h t( ) is the channel response of multipath
frequency-selective fading. FD ADC transforms a segment of continuous time-domain received signal into a vector of digital frequency coefficients by [19]
2 0 c n T j F t q n r t qT ec dt
R (2.21) where n0Nb1, [ , , ,0 1 b 1] T q R R RNR is the frequency coefficient vector
with the frequencies of
0Nb 1n
F , T is the segment duration, and q is the segment c
index. T can be equal to the whole preamble period Tc p or a segment of preamble, i.e. c p s
T T M . Due to the preamble period is usually shorter then an OFDM symbol, Tc
is assumed to be equal to the preamble period Tp to simplify the following derivations.
Hence, Nb is set to be equal to the the number of preamble samples L to avoid timing
Figure 2-4: Definitions of data patterns: (a) TX preamble patterns; and (b) FD signal vectors and an estimated example of L 2 in RX.
2.2.2 Matched Filter Detection in FD Receiver
In practice, the matched filter detection is important to timing synchronization and data detection for many communication systems. It can be efficiently realized in the
FD receiver structure, too. For example, the matched filter metric between the
time segment of received signal r t (duration of T
p) and the preamble c t( ) canexpress in continuous time domain as the conventional form of
* ini p ini t T t r t c t dt
(2.22)filter. With the help of the Parsevals’ theorem [54], Equation (2.22) can be derived from the frequency-domain signals as
* * ini p ini t T t r t c t dt R F C F dF
(2.23)where R F
and C*
F represent continuous frequency domain signals generatedby Fourier transform of r t and
c t , respectively. However, the output of FD *
ADC are Nb discrete frequency samples
1 0
R N
n
n from the spectrum of r t
.Therefore, Equation (2.23) is approximated as
* 1 * 0 b N c n F C F dF F n n
R C R (2.24) where
*
1 0 b N n n C is the set of frequency samples from the spectrum of c t . Note *
that the truncate error in Equation (2.24) can be negligible [20] and the discrete-time
aliasing can be avoided when the number of frequency samples Nb is optimal,
e.g.,N WTc in Equation (2.2). Therefore, the matched filter function can be
adopted in the FD receiver as well as the conventional receiver.
2.2.3 Problem Statement
With the conventional TD ADC, most algorithms find the symbol boundary by serial search with a sliding window over the consecutive received signal samples. However,
FD ADC outputs the frequency-domain digital signals from the received signal segments of continuous time-domain signal in parallel. Without sampling index in FD receivers, the serial search is a failure because it cannot get any timing information in time domain. Meanwhile, directly utilizing the matched filter function with the frequency samples of one received signal segment might be a solution for the synchronization with FD ADC. However, it still could obtain a poor synchronization performance in the frequency selective channel [32]. To the best of our knowledge, so far no existing FD symbol synchronizer has been built for FD-ADC approaches.
Considering a low complexity design and implementation for FD synchronizer with FD ADC architecture, one way is to use a simple metric computation and reduce the detection space based on algorithm level. The other way is to decrease the gate counts of VLSI implementations via shared architectures, i.e. shared hardware by high-speed inner clocking without the help of PLLs. Finally, the proposed solution must be workable, reliable and robust at low SNR (i.e. SNR ≤ 10 dB) due to many broadband standards provide low-rate, medium-rate and high-rate transmissions for different environments (SNR).
2.3 The Proposed FD Symbol Synchronization
2.3.1 Sequential Search
Since the transmitted preamble repeats periodically as shown in Fig. 2-4, any time segment of duration Tp within the transmitted preambles can be defined as
1 ( ) ( ) ( ) 1 , 0 1 L s T L L i s s c t c i t i T f t T t T L
(2.25)where 0, , L1 is the preamble pattern index, and
L
means “modulo L”. The
segment is one of the L possible preamble patterns where c t0( ) is the preamble with
zero shift, and the other c t( ) are the preambles with cyclic shift of samples.
On the receiver side, the continuous time-domain received signals can be
segmented into consecutive signal segments. Let q be the segment index and the
corresponding segment is defined as
q q r t r t w t (2.26) where
1,
1
0, elsewhere p p q qT t q T w t (2.27)Because preambles have good correlation property, the similarities between each received signal segment r tq
