Institutionen för systemteknik Department of Electrical Engineering Actuator Networks
by user
Comments
Transcript
Institutionen för systemteknik Department of Electrical Engineering Actuator Networks
Institutionen för systemteknik Department of Electrical Engineering Examensarbete Media Access Control for Wireless Sensor and Actuator Networks Examensarbete utfört i Kommunikationssystem vid Tekniska högskolan i Linköping av Muaz Un Nabi LiTH-ISY-EX--12/4603--SE Linköping 2012 Department of Electrical Engineering Linköpings universitet SE-581 83 Linköping, Sweden Linköpings tekniska högskola Linköpings universitet 581 83 Linköping Media Access Control for Wireless Sensor and Actuator Networks Examensarbete utfört i Kommunikationssystem vid Tekniska högskolan i Linköping av Muaz Un Nabi LiTH-ISY-EX--12/4603--SE Handledare: Hien Quoc Ngo isy, Linköpings universitet Mikael Gidlund ABB AB Johan Åkerberg ABB AB Examinator: Eleftherios Karipidis isy, Linköpings universitet Linköping, 11 July, 2012 Avdelning, Institution Division, Department Datum Date Division of Communication Systems Department of Electrical Engineering Linköpings universitet SE-581 83 Linköping, Sweden Språk Language Rapporttyp Report category ISBN Svenska/Swedish Licentiatavhandling ISRN Engelska/English Examensarbete C-uppsats D-uppsats Övrig rapport 2012-07-11 — LiTH-ISY-EX--12/4603--SE Serietitel och serienummer ISSN Title of series, numbering — URL för elektronisk version http://www.commsys.isy.liu.se Titel Title Media Access Control for Wireless Sensor and Actuator Networks Författare Muaz Un Nabi Author Sammanfattning Abstract In a wireless network, the medium is a shared resource. The nodes in the network negotiate access of the shared resource using the Medium Access Control (MAC) protocol. The design of a MAC protocol for a sensor node is not the same as that for a wireless transceiver. Due to the transceiver characteristics, the MAC protocol design is limited in terms of medium access methods. However, in most cases, the protocols rely on simple access methods i.e. Time Division Multiple Access (TDMA) or Carrier Sense Multiple Access / Collision Avoidance (CSMA / CA). Control and monitoring applications, running over a wireless network, are typical examples of Wireless Sensor Actuator Network (WSAN) application in industries. In an industrial network, the message deliveries must be time-bounded otherwise, they are of no use. This report aims to present the thesis work carried out at ABB AB, Västerås. The purpose of this thesis was to compare the performance of WLAN and WirelessHART when it comes to control applications. For the purpose of WLAN, the media access schemes are analyzed in terms of deadline misses. There are other metrices for the performance evaluation but our focus was on the latency, since it is very important in the field of industrial automation. NS-2 was used for the purpose of MAC layer analysis and it is also shown that PCF gives better performance as compared to DCF, in terms of deadline misses. Finally, WLAN is proven to accommodate more control loops as compared to WirelessHART for a given scenario. Nyckelord Keywords NS-2, WSN, IWLAN, DCF, PCF Abstract In a wireless network, the medium is a shared resource. The nodes in the network negotiate access of the shared resource using the Medium Access Control (MAC) protocol. The design of a MAC protocol for a sensor node is not the same as that for a wireless transceiver. Due to the transceiver characteristics, the MAC protocol design is limited in terms of medium access methods. However, in most cases, the protocols rely on simple access methods i.e. Time Division Multiple Access (TDMA) or Carrier Sense Multiple Access / Collision Avoidance (CSMA / CA). Control and monitoring applications, running over a wireless network, are typical examples of Wireless Sensor Actuator Network (WSAN) application in industries. In an industrial network, the message deliveries must be time-bounded otherwise, they are of no use. This report aims to present the thesis work carried out at ABB AB, Västerås. The purpose of this thesis was to compare the performance of WLAN and WirelessHART when it comes to control applications. For the purpose of WLAN, the media access schemes are analyzed in terms of deadline misses. There are other metrices for the performance evaluation but our focus was on the latency, since it is very important in the field of industrial automation. NS-2 was used for the purpose of MAC layer analysis and it is also shown that PCF gives better performance as compared to DCF, in terms of deadline misses. Finally, WLAN is proven to accommodate more control loops as compared to WirelessHART for a given scenario. v Acknowledgments First of all, I would like to thank ABB AB for giving me this opportunity to work with them. I am grateful to Mikael and Johan for sparing their valuable time, to supervise me for this master thesis, during my stay at ABB AB. Also, I would like to thank Hien and Eleftherios for their support at the university. Muaz Un Nabi. July, 2012 vii Contents 1 Introduction 1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Purpose and Goals . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Disposition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Introduction to Wireless Local Area Networks 2.1 Basic Architecture of IEEE 802.11 WLAN . . . . 2.2 802.11 PHY Layer . . . . . . . . . . . . . . . . . 2.3 PHY Extensions . . . . . . . . . . . . . . . . . . 2.4 IEEE 802.11n . . . . . . . . . . . . . . . . . . . . 2.4.1 Guard Interval . . . . . . . . . . . . . . . 2.4.2 High Rate FEC Codes . . . . . . . . . . . 2.4.3 MIMO Technology . . . . . . . . . . . . . 2.4.4 Modes of Operation . . . . . . . . . . . . 3 IEEE 802.11 MAC Layer 3.1 Distributed Coordination Function 3.1.1 Working Mechanism . . . . 3.1.2 Problems with the DCF . . 3.2 Point Coordination Function . . . 3.2.1 Working Mechanism . . . . 3.2.2 Problems with the PCF . . 3.2.3 Polling Mechanism . . . . . . . . . . . . . . . . . . . 4 Wireless in Industrial Automation 4.1 Industrial Automation Requirements . 4.2 Standards . . . . . . . . . . . . . . . . 4.2.1 Controller Area Network . . . . 4.2.2 PROFINET I/O . . . . . . . . 4.2.3 Bluetooth . . . . . . . . . . . . 4.2.4 Wireless HART . . . . . . . . . 4.2.5 IWLAN . . . . . . . . . . . . . 4.3 Advantages of Wireless Technologies in ix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Industrial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Automation . . . . . . . . . . . . . . . . . . . . . . . . 1 1 2 3 . . . . . . . . 5 6 7 8 9 10 10 11 12 . . . . . . . 13 13 13 15 15 15 16 16 . . . . . . . . 19 19 21 21 21 22 22 24 25 x Contents 5 Performance Analysis of IEEE 802.11 PHY 5.1 IEEE 802.11n Simulink Model . . . . . . . . . . . . . 5.1.1 Model Description . . . . . . . . . . . . . . . 5.1.2 Channels . . . . . . . . . . . . . . . . . . . . 5.1.3 Parameters . . . . . . . . . . . . . . . . . . . 5.1.4 Results . . . . . . . . . . . . . . . . . . . . . 5.2 Throughput and Delay Limits . . . . . . . . . . . . . 5.2.1 Assumptions . . . . . . . . . . . . . . . . . . 5.2.2 Derivation of MT and MD for IEEE 802.11a 5.2.3 Derivation of MT and MD for IEEE 802.11b 5.2.4 Parameters . . . . . . . . . . . . . . . . . . . 5.2.5 Results . . . . . . . . . . . . . . . . . . . . . 5.3 Packet Aggregation . . . . . . . . . . . . . . . . . . . 5.3.1 DSDSRFA . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 27 27 28 28 29 32 32 32 34 35 35 42 42 6 Performance Analysis of IEEE 802.11 MAC 6.1 NS-2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1.1 Overview . . . . . . . . . . . . . . . . . . . . 6.1.2 Network Components . . . . . . . . . . . . . 6.1.3 Tracing . . . . . . . . . . . . . . . . . . . . . 6.2 Comparison between PCF and DCF . . . . . . . . . 6.2.1 Parameters . . . . . . . . . . . . . . . . . . . 6.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . 6.3.1 Deadline misses of PCF and DCF for 1 Mbps 6.3.2 Deadline misses of PCF and DCF for 2 Mbps 6.4 Observations . . . . . . . . . . . . . . . . . . . . . . 6.4.1 Deadline Misses . . . . . . . . . . . . . . . . . 6.4.2 Bandwidth Efficiency . . . . . . . . . . . . . 6.5 Comparison between WLAN and WirelessHART . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 45 45 45 48 49 49 49 51 57 62 62 64 65 7 Conclusions and Future Work 7.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 67 68 Bibliography 69 Contents xi List of Figures 1.1 1.2 Wireless Sensor Networks (Courtesy of ABB) . . . . . . . . . . . . The Basic Block Diagram of the Wireless System (Courtesy of ABB) 2.1 2.2 2.3 2.4 2.5 Basic WLAN Architecture . . . . . . . . IEEE 802.11 Protocol Reference Model . General IEEE 802.11a Frame Format . . General IEEE 802.11n Frame Format . . The HT Control Field . . . . . . . . . . . . . . . 6 8 9 10 10 3.1 3.2 Graphical Representation of CSMA / CA . . . . . . . . . . . . . . Graphical Representation of PCF . . . . . . . . . . . . . . . . . . . 14 17 4.1 Wireless HART Mesh Network (Courtesy of ABB) . . . . . . . . . 23 5.1 5.2 5.3 5.4 5.5 5.6 30 30 31 31 36 5.15 5.16 5.17 PER vs SNR for MCS = 13, 14 and 15 (AWGN, 2x2 Direct-Map) PER vs SNR for MCS = 13, 14 and 15 (Ch D nLoS, 2x2 Direct-Map) PER vs SNR for MCS = 13, 14 and 15 (Ch D nLoS, 2x2 Beamforming) PER vs SNR for MCS = 13, 14 and 15 (Ch D nLoS, 4x2 STBC) . Ideal TUL for frame size = 1000 Bytes . . . . . . . . . . . . . . . . Maximum Throughput vs Payload size for different Data Rates (IEEE 802.11a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Minimum Delay vs Payload size for different Data Rates (IEEE 802.11a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Bandwidth Efficiency for different Data Rates (IEEE 802.11a) . . . Effect on Throughput by doubling the Rate (2x6=12Mbps)(IEEE 802.11a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Effect on Throughput by doubling the Rate (2x54=108Mbps)(IEEE 802.11a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Percentage increase in Throughput by doubling the Rate (IEEE 802.11a) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Throughput when the Channel is adding Noise (IEEE 802.11a) . . Maximum Throughput vs Payload size for different Data Rates (IEEE 802.11b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Minimum Delay vs Payload size for different Data Rates (IEEE 802.11b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Concept of DSDSRFA . . . . . . . . . . . . . . . . . . . . . . . . . Comparison of TUL . . . . . . . . . . . . . . . . . . . . . . . . . . Comparison of MT . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.1 6.2 6.3 6.4 6.5 6.6 Simplified User0 s View of NS Class Hierarchy (Partial) . . . Node (Unicast and Multicast) Link . . . . . . . . . . . . . . Insertion of Trace objects . . Trace Format example . . . . 5.7 5.8 5.9 5.10 5.11 5.12 5.13 5.14 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 3 36 37 38 38 39 39 40 41 41 42 43 43 46 46 47 47 48 48 xii Contents 6.7 6.8 6.9 6.10 6.11 6.12 6.13 6.14 6.15 6.16 6.17 6.18 6.19 6.20 6.21 6.22 6.23 6.24 6.25 6.26 Five Node Network Topology . . . . . . . . . . . . . . . . . . . . . Encountered delay vs Packets received for 5 nodes and 1000 bytes Encountered delay vs Packets received for 10 nodes and 1000 bytes Encountered delay vs Packets received for 10 nodes and 1500 bytes Encountered delay vs Packets received for 15 nodes and 512 bytes Encountered delay vs Packets received for 20 nodes and 5 bytes . . Encountered delay vs Packets received for 5 nodes and 100 bytes . Average end-to-end delay vs cbr interval (1 Mbps) . . . . . . . . . Average end-to-end delay vs cbr interval (1 Mbps) . . . . . . . . . Average end-to-end delay vs cbr interval (1 Mbps) . . . . . . . . . Encountered delay vs Packets received for 5 nodes and 1000 bytes Encountered delay vs Packets received for 10 nodes and 1000 bytes Encountered delay vs Packets received for 10 nodes and 1500 bytes Encountered delay vs Packets received for 15 nodes and 512 bytes Encountered delay vs Packets received for 20 nodes and 5 bytes . . Encountered delay vs Packets received for 5 nodes and 100 bytes . Average end-to-end delay vs cbr interval (2 Mbps) . . . . . . . . . Average end-to-end delay vs cbr interval (2 Mbps) . . . . . . . . . Average end-to-end delay vs cbr interval (2 Mbps) . . . . . . . . . The Basic Block Diagram of the Wireless System (Courtesy of ABB) 50 52 52 53 54 54 55 55 56 56 57 58 58 59 59 60 61 61 62 65 Contents xiii List of Tables 2.1 2.2 2.3 IEEE 802.11 Standards . . . . . . . . . . . . . . . . . . . . . . . . Rate Dependent Parameters for IEEE 802.11a . . . . . . . . . . . . Rate Dependent Parameters for IEEE 802.11n . . . . . . . . . . . 6 9 11 4.1 Typical Requirements for Industrial Wireless Sensor and Actuator Networks in the Process Automation Domain . . . . . . . . . . . . 20 5.1 5.2 5.3 Parameters used for Simulating IEEE 802.11n PHY Simulink Model 28 MCS Parameters for 20 Mhz, NSS = 2, NES = 1, EQM . . . . . . 29 MAC and PHY Parameters for 802.11 a/b . . . . . . . . . . . . . . 35 6.1 6.2 6.3 6.4 Parameters for PCF Vs. DCF Comparison Deadline Comparison of PCF and DCF . Bandwidth Efficiency (PCF Vs. DCF) . . Bandwidth Efficiency (PCF Vs. DCF) with . . . . . . . . . . . . . . . Varying . . . . . . . . . . . . . . . . . . . . . . . . . . . Number of Nodes 49 63 64 65 xv xvi Contents Abbreviations Abbreviations AWGN BCC CAN CBR CFP CP CSMA/CA CW DCF DIFS DSSS ELF EQM FCS FHSS FIFO FTP GI HART ISI ISM IWLAN MAC MCS MIMO OFDM PCF PDR PER PHY PIFS PSDU RR SIFS SM STBC UDP WLAN WSN Description Additive White Gaussian Noise Binary Convolutional Codes Controller Area Network Constant Bit Rate Contention Free Period Contention Period Carrier Sense Multiple Access/Collision Avoidance Contention Window Distributed Coordination Function DCF Inter-Frame Sequence Direct Sequence Spread Spectrum Effort Limited Fair Equal Modulation Frame Check Sequence Frequency Hopping Spread Spectrum First-In-First-Out File Transfer Protocol Guard Interval Highway Addressable Remote Transducer InterSymbol Interference Industrial, Scientific and Medical Industrial Wireless Local Area Network Media / Medium Access Control Modulation and Coding Scheme Multiple Input Multiple Output Orthogonal Frequency Division Multiplexing Point Coordination Function Packet Delivery Ratio Packet Error Rate Physical PCF Inter-Frame Sequence Physical Service Data Unit Round Robin Short Inter-Frame Sequence Spatial Multiplexing Space Time Block Coding User Datagram Protocol Wireless Local Area Network Wireless Sensor Networks Chapter 1 Introduction This Master of Science thesis is a partial fulfillment for the Master program in Communication Systems at Linköping University. The thesis work was carried out at the Corporate Research Center (CRC) of ABB AB in Västerås and this report aims to present the work carried out during the period and the corresponding results. In the modern world, industries are in the process of competing with each other in order to have an enhanced performance. In the field of industrial automation, low-cost and efficient industrial automation systems are required for the purpose. The application of wireless technologies, in industrial automation applications, are at a high pace. Due to the robust requirements of an industry, sometimes it is difficult to install a cable-based system and in most of the cases, it is impossible to do so. To have a cost-efficient system and increase the flexibility, wireless technologies are increasingly being deployed to overcome the shortcomings of a cable-based system. The most popularly used wireless-based standards, for industrial automation applications, are the Bluetooth [1], ZigBee, WirelessHART [2] and now the Wireless Local Area Network, also termed as IWLAN. 1.1 Background In a wireless network, the medium is a shared resource. The nodes in the network negotiate access of the shared resource using the Medium Access Control (MAC) protocol. The design of a MAC protocol for a sensor node is not the same as that for a wireless transceiver. Due to the transceiver characteristics, the MAC protocol design is limited in terms of medium access methods. However, in most cases, the protocols rely on simple access methods i.e. Time Division Multiple Access (TDMA) or Carrier Sense Multiple Access / Collision Avoidance (CSMA / CA). Control and monitoring applications, running over a wireless network, are typical examples of Wireless Sensor Actuator Network (WSAN) application in industries. A general architecture of a Wireless Sensor Network is shown in Figure 1.1. 1 2 Introduction Figure 1.1: Wireless Sensor Networks (Courtesy of ABB) To control an unstoppable process, message deliveries over the network must be time-bounded; the network must offer reliabilities and delays comparable to a point-to-point network. A typical example, showing the importance of timebounded message deliveries, is of a stone crushing machine in an industry. This kind of machine requires a start and a stop signal, which should be transmitted and received in a specified amount of time. For instance, due to some calculation error, a stone entering the crusher exceeds the size accommodated by the machine. To prevent any hazardous event, the stop signal should be sent and received in a pre-defined amount of time. If these message deliveries are not time-bounded, the blockade can cause a loss up to $ 50,000 to $ 100,000 per hour. More details, about the time-bounded critical application areas, are in Section 4.1. This is a major challenge and the thesis aims to investigate the problem further. 1.2 Purpose and Goals In the modern world of today, WLANs are considered to be an important part of one0 s life. The aim of this master thesis is to investigate the medium sharing techniques for industrial automation applications. The specific question this thesis wants to answer is: Can WLAN perform better than WirelessHART for industrial automation applications? Figure 1.2 shows a top-level block diagram of a Wireless Sensor and Actuator 1.3 Disposition 3 network. The Sensor Node senses the application controlled variable from the Process Plant and sends it ahead to the Gateway. The Controller is responsible for the computation of the control signals. It receives the data from the Gateway, performs the required computations and sends it back to the Gateway. It then further sends the data to the Actuator, which performs the required operation on the Process Plant. This whole process is termed as one Control loop. We are interested to answer the question: Is WLAN better than WirelessHART when it comes to control applications? Figure 1.2: The Basic Block Diagram of the Wireless System (Courtesy of ABB) When it comes to control applications, which require a fast update frequency rate, can WLAN perform better than WirelessHART? Furthermore, The MAC schemes for WLAN, are investigated. The results are analyzed under different scenarios to conclude the above stated questions. 1.3 Disposition The remaining of the report is as follows: • Chapter 2 explains the theoretical background of IEEE 802.11 WLAN and the physical layer characteristics. • Chapter 3 describes the IEEE 802.11 MAC layer and the MAC access techniques. • Chapter 4 explains the application of wireless technology in the field of industrial automation, some standards and protocols and the concept of IWLAN. • Chapter 5 shows the IEEE 802.11 physical layer analysis and theoretical limitations on throughput and delay. • Chapter 6 explains the MAC layer analysis. 4 Introduction • Chapter 7 highlights the conclusions and a guide for future work. Chapter 2 Introduction to Wireless Local Area Networks Today, wireless technologies are increasingly being used in all kinds of applications. Whether it be in home or offices to large-scale industrial applications, wireless technologies have evolved in every walk of life. A simple scenario of a Wireless Local Area Network (WLAN) is a few computers connected, wirelessly, via an access point (AP). An AP acts as a gateway between the users or stations and the wider internet. Once the station is connected to the internet via an AP, it can roam freely within the range wirelessly. The range depends on the device acting as an AP. The IEEE 802.11 defines several protocols used within the WLANs. Table 2.1 [3], [4] and [5] briefly explains different IEEE 802.11 protocols: Standard 802.11 Release 1997 802.11a 1999 802.11b 1999 802.11e 2005 802.11g 2003 802.11h 2004 Description Basic standard, uses ISM band of 2.4 GHz and maximum supported data rate is of 2 Mbps. Extension of 802.11 supporting a maximum data rate of 54 Mbps and uses the ISM band of 5 GHz. Extension of 802.11 supporting a maximum data rate of 11 Mbps and uses the ISM band of only 2.4 GHz. Enhanced QoS, resource allocation of bandwidth according to the content and the use of Wireless Multi Media (WMM). Same as 802.11a, however the ISM band is of 2.4 GHz for 802.11b backward compatibility. Standard for the dynamic frequency selection and power limitations in order to meet the requirements for the 5 GHz band in Europe. 5 6 Introduction to Wireless Local Area Networks 802.11i 2004 802.11n 2008 802.11ac 2012 Standard for the security access control and data encryption for 802.11, Wifi Protected Access 2 (WPA2) is a security certification program of this standard. Uses the ISM band of 2.4 / 5 GHz, supports a maximum data rate of 72.2 Mbps with 20 MHz bandwidth, supports a maximum data rate of 150 Mbps with 40 MHz bandwidth and backward compatible with 802.11g. Uses the ISM band of 5 GHz, supports a maximum data rate of 867 Mbps with 80 MHz bandwidth and a maximum data rate of 6.93 Gbps with 160 MHz bandwidth. Table 2.1: IEEE 802.11 Standards 2.1 Basic Architecture of IEEE 802.11 WLAN The IEEE 802.11 standard consists of two basic components: The PHY layer specifications and the MAC layer specifications. The detailed description is given in [3]. Figure 2.1 [6] illustrates how a simple WLAN looks like: Figure 2.1: Basic WLAN Architecture A distributed system (DS) connects one AP to another to form a larger network. Its main purpose is to pass data from the source AP to the destination AP. An access point (AP) gives wireless connectivity to other devices in the network. It can also serve as the center of a wireless network. The transmission of data from one station to the other is performed by a wireless medium provided. For this purpose, IEEE has defined standards [3]. These PHY access techniques are discussed in Section 2.2. A Basic Service Set (BSS) is considered to be an essential part of a WLAN 2.2 802.11 PHY Layer 7 network. A BSS is generally a collection of a number of stations that are controlled by a coordination function. These coordination functions are explained in detail in Chapter 3. Generally, two kinds of network implementations exist i.e. Independent BSS (IBSS) or the ad hoc network and Extended Service Set (ESS) or an infrastructure network. • Ad hoc Network: An ad hoc network consists of a number of networks within a single BSS. Since it has no central server, all the stations, within the BSS, can communicate freely with one another. • Infrastructure Network: On the other hand, an infrastructure network consists of an AP which serves as a server in the network. One can say that in such a network, multiple BSSs are connected together via APs. Infrastructure networks combine to form an ESS. In an ESS, multiple BSSs are connected via a common DS to enhance the coverage area. 2.2 802.11 PHY Layer In 1997, the IEEE released the 802.11 standard [3] which defines the physical (PHY) and the media access control (MAC) layer for WLANs. The IEEE 802.11 WLAN standard describes three wireless data exchange techniques for the PHY layer i.e. Frequency Hopping Spread Spectrum (FHSS), Direct Sequence Spread Spectrum (DSSS) and IR. The physical layer of 802.11 constitutes of two parts i.e. the Physical Layer Convergence Protocol (PLCP) and the Physical Medium Dependent (PMD) sub layers [3]. The PLCP uses different 802.11 media access methods to analyze the data being sent or received. Under the supervision of PLCP, PMD performs data transmission or reception and modulation or demodulation via accessing the air directly. Figure 2.2 [3] illustrates the 802.11 protocol reference model. • FHSS: The main idea behind FHSS is to transmit the data on a particular frequency for a very short duration of time and then hop to another frequency value. This frequency is a pre-set value and is known to both, the sender and the receiver. The FHSS uses the 2.4 GHz of ISM band. The center frequency of the first channel is 2.402 GHz and the rest of the channels are at a distance of 1MHz from each other. The FHSS creates three different sets of hopping sequences and each set consists of 26 hopping sequences [7]. The purpose of these sequences is to allow multiple BSSs to exist in the same location. The 1 Mbps access rate uses two-level Gaussian Frequency Shift keying (GFSK). In GFSK, 1 is coded as fc+f and 0 is coded as fc-f where fc is the center frequency. For the access rate of 2 Mbps, a four-level GFSK is used. • DSSS: In DSSS, the transmitted data is multiplied with a pseudo random binary sequence, which has a higher bit rate. It uses the same 2.4 GHz ISM 8 Introduction to Wireless Local Area Networks Figure 2.2: IEEE 802.11 Protocol Reference Model band but unlike FHSS, DSSS uses Differential Binary Phase Shift Keying (DBPSK) for the 1 Mbps access rate and Differential Quadrature Phase Shift Keying (DQPSK) for the 2 Mbps access rate.The PHY access schemes are explained in [8] in detail. 2.3 PHY Extensions In 1999, there were two extensions namely 802.11b and 802.11a. The IEEE 802.11b is based on the preceding definition of the PHY with DSSS. However, the data rate is extended to 11 Mbps. The modulation scheme used is called Complementary Code Keying (CCK) [9]. The other extension i.e. 802.11a used the 5.2 GHz band to stipulate a new radio-based PHY. It uses a modulation scheme called Orthogonal Frequency Division Multiplexing (OFDM) [10] and hence providing a data rate of 54 Mbps. In practice, IEEE 802.11b only provided up to 5 Mbps and IEEE802.11a provided almost half of the PHY data rate [11]. The major problem with these amendments was their use of different ISM bands. For the co-existence of both 11a and 11b required production cost to make the two standards compatible with each other. The IEEE later proposed 802.11g standard to resolve these issues. IEEE 802.11g has the same characteristics as IEEE 802.11a, however, based on the 2.4 MHz band. This extension resolved the issue of backward compatibility but came up with some new things to consider for instance, the simultaneous high quality video streaming for multiple customers. Table 2.2 [4] shows the rate dependent parameters for IEEE 802.11a. Figure 2.3 [4] gives a general overview of the IEEE 802.11a frame format. 2.4 IEEE 802.11n Mode 1 2 3 4 5 6 7 8 Modulation BPSK BPSK QPSK QPSK 16-QAM 16-QAM 16-QAM 64-QAM 9 Code Rate 1 /2 3 /4 1 /2 3 /4 1 /2 3 /4 2 /3 3 /4 Data Rate 6 Mbps 9 Mbps 12 Mbps 18 Mbps 24 Mbps 36 Mbps 48 Mbps 54 Mbps NDBP S 24 36 48 72 96 144 192 216 Table 2.2: Rate Dependent Parameters for IEEE 802.11a Figure 2.3: General IEEE 802.11a Frame Format 2.4 IEEE 802.11n Due to the shortcomings of the previous IEEE standards and the growing demands of high throughputs, IEEE came up with a new standard i.e. 802.11n in 2003. The initial target of this standard was to provide a data rate of a minimum of 100 Mbps, which is almost twice as the rates supported by previous standards. A special Task Group (TGn) was formed and given the job to do so. The upper limit for the PHY layer data rate was assigned to be 600 Mbps. The current version of the IEEE 802.11n standard is defined in [5]. In order to provide this huge amount of data rate, improvements have to be made in the existing MAC and PHY layers. Following are the PHY layer characteristics for the IEEE 802.11n standard. 10 Introduction to Wireless Local Area Networks Figure 2.4 [5] depicts the general frame format for IEEE 802.11n. Compar- Figure 2.4: General IEEE 802.11n Frame Format ing it with the original 802.11 frame format [3], one can observe two additions namely, the QoS Control and High-Throughput (HT) control. The HT control field is shown in Figure 2.5 [5]: Figure 2.5: The HT Control Field Note: Following are simply the highlights for the IEEE802.11n amendments. For detailed description, refer to [11], [5] and [12]. 2.4.1 Guard Interval IEEE 802.11n uses OFDM [10] as a modulation technique. Each OFDM symbol constitutes of a guard interval. The purpose of guard interval is to reduce the Inter Symbol Interference (ISI). The duration of the guard interval depends on the environment conditions, which is normally a multipath environment. For IEEE 802.11 a/g, the length of the guard interval is 0.8 µs. for IEEE 802.11n it is the same, however, it can be lowered as low as 0.4 µs depending on the channel conditions. This reduction in guard interval increases the data rate by a value of 10 % [12]. 2.4.2 High Rate FEC Codes Mode BPSK QPSK QPSK 16-QAM 16-QAM Modulation 1 /2 1 /2 3 /4 1 /2 3 /4 Code Rate 1 2 3 4 5 Data Rate 6.5 Mbps 13 Mbps 19.5 Mbps 26 Mbps 39 Mbps 2.4 IEEE 802.11n 64-QAM 64-QAM 64-QAM 11 2 /3 /4 5 /6 3 6 7 8 52 Mbps 58.5 Mbps 65 Mbps Table 2.3: Rate Dependent Parameters for IEEE 802.11n As in IEEE 802.11a, the BPSK modulation, having code rate of 3 /4 , is removed from the IEEE 802.11n standard. It is not at all suitable for the performance improvement at any SNR value [13]. Table 2.3 [14] shows an addition in the form of 5 /6 puncturing for a 64 QAM. This scheme has been tested for a 10 % increase in the data rate [12] . This Modulation and Coding Scheme (MCS) [5] works well if the channel is in good condition. 2.4.3 MIMO Technology The major advancement in the IEEE 802.11n was the inclusion of the Multiple Input Multiple Output (MIMO) transmission technology. In simple terms, MIMO implies the usage of more than one antenna at both, the transmitter and the receiver side. MIMO technology has been applied on large scales, since it is capable of providing enhanced data rates at a minimal cost of extra transmitted power. In IEEE 802.11n, the MIMO techniques are usable along with three other techniques namely, Transmit Beamforming, Space Time Block Coding (STBC) and Spatial Multiplexing (SM) [14]. • Transmit Beamforming: Beamforming is a technique, applied at the transmitting antennas (MIMO) to improve SNR. In MIMO scheme, multiple antennas for the transmission and reception of radio signals. These radio signals, due to multipath environment, follow different paths before reaching the receiver side. They arrive at different times thus affecting the phase of the transmitted signals. Beamforming wisely sets the phase at the transmitter in order to increase the SNR at the receiver end. However, if the channel conditions are known to the transmitter and the receiver, this process can further be enhanced but it involves an extra overhead. Therefore this technique is recommended to use with one receiver. • Space Time Block Coding: STBC corresponds to transmitting the same signal on multiple transmitter antennas for the purpose of redundancy. Since they follow the multipath scheme, the same signal travels via different conditions before reaching the receiver. For the signal to reach the receiver safe and sound, it is advised to send more copies of the signal. [15] can be referred for more details about beamforming Vs. STBC performance. • Spatial Multiplexing: SM is a transmission technique which parses transmitted data streams into sub-streams and sends them with the help of multiple antennas with one sub-stream per antenna. Each individual sub-stream 12 Introduction to Wireless Local Area Networks practices the effects of the multipath environment. At the transmitter side, special mechanisms are applied to re-group the sub-streams. SM is more effective in terms of providing high data rates at high SNR values. SM can be employed along with the STBC but in this case, only one sub-stream will be transmitted via STBC and the rest will be sent via SM [5]. 2.4.4 Modes of Operation In IEEE 802.11n, the PLCP preamble [3] is also modified, depending on the mode of operation. There are three modes of operation: Green field mode, Non-HT mode and Mixed mode. In Green field mode, only IEEE 802.11n stations are involved in the process of communication. In Non-HT mode, legacy 802.11 station and Mixed mode corresponds to a combination of both. The first one is optional but the rest are compulsory. Chapter 3 IEEE 802.11 MAC Layer The MAC layer of IEEE 802.11 specifies the schemes to access the channel in the process of communication. The two techniques are namely Distributed Coordination Function (DCF) and Point Coordination Function (PCF). 3.1 Distributed Coordination Function The DCF mechanism is the conventional scheme for the channel access. As the name says, DCF does not require a master station to control the sharing mechanism. It uses the concept of Distributed Coordination. DCF uses Carrier Sense Multiple Access with Collision Avoidance (CSMA/CA) and a random exponential backoff [16] to decide the control of channel. The backoff interval is randomly defined between 0 and the length of the Contention Window (CW). In other words, one can say that DCF is used for the contention dependent services. DCF is particularly for asynchronous data transmission e.g. e-mail or web browsing. These data exchanges are not time dependent. The technique CSMA/CA is the reason why its time independent, since when it senses the medium to be free, it allows a particular station to transmit without a time bound [16]. 3.1.1 Working Mechanism The working mechanism of CSMA/CA, for the DCF scheme is shown in Figure 3.1. In CSMA/CA, the station senses the channel to be idle, following Wait and Listen scheme. It keeps on observing the required channel. As long as there are transmissions going on the channel, it waits. When it senses the channel to be idle, the station starts sending data packets [17]. In DCF, when the channel is sensed idle, the station will wait for another DCF Interframe Space (DIFS) before starting the transmission. This DIFS interval varies depending upon the type of the physical layer [18]. After the transmission, the transmitter waits for an ACK. If the ACK is not received, it assumes that the packet maybe lost. As a result, it tries for a re-transmission. When the channel is sensed idle, the backoff timer, set to some value, is decreased. The backoff timer is selected, randomly, between 13 14 IEEE 802.11 MAC Layer zero and the maximum size of contention window. In Figure 3.1, the minimum size of contention window is 31 and the maximum size of contention window is 1023 (these values are for IEEE 802.11b). The station is allowed to transmit only when the backoff timer reaches 0 [19]. Every station has to come up with a CW. This CW determines the waiting time for a station (in terms of time slots) before it can initiate a transmission. The size of the contention window is 2K - 1. Each station has a Short Interframe Space (SIFS) to transmit the ACK. Since SIFS shorter then DIFS, no medium sensing is required by the station before sending the ACK. The SIFS also depend on the type of PHY and are listed in [3]. Figure 3.1: Graphical Representation of CSMA / CA 3.2 Point Coordination Function 3.1.2 15 Problems with the DCF The Hidden node problem is a major hurdle in the implementation of DCF. This problem is resolved by using Request to Send (RTS) and Clear to Send (CTS) flags. However, RTS/CTS should be efficiently implemented, since they deliberately increase the overhead and as a result, affect the total throughput of the system. A wise method, for instance, is to implement RTS/CTS algorithm at the user wireless NICs [20]. The drawbacks of using DCF for real-time communication applications are listed below: [20] and [21] • Cannot support QoS. • The guarantee of the packet transmission, during a specified time, is not confirmed. • DCF cannot differentiate between different types of data and hence has no concept of prioritization. • It is not suitable for highly loaded traffic since it uses wait and listen concept. It will eventually affect the total throughput of the system. • Theory of contention. • Even if RTS/CTS are used, in the presence of RF interference, it will sense the channel to be busy and will not allow initiating the process of transmission. • Denials of Service (DoS) attacks also affect the overall performance of the system using DCF as a channel access scheme. The above mentioned drawbacks of DCF are tackled by PCF, which is a polling based medium access mechanism devised by IEEE 802.11. 3.2 Point Coordination Function PCF is a polling based algorithm for MAC layer. The main aspect of this technique is that there is single Point Coordinator (PC), usually an Access Point (AP), which keeps on polling the stations within the Basic Service Set (BSS) to initialize the data transmission process. PCF is basically designed for synchronous data transmission i.e. real time data such as video etc. synchronous means that the data transmission is time-bounded i.e. a process has to occur in a specific interval of time. The PCF process initiates with a Contention-free Period (CFP). However it also has a room for Contention Period (CP) [16]. The time access for the channel is divided into these CFP and CP. A combination of alternating CFP and CP is called a SuperFrame [19]. 3.2.1 Working Mechanism Before PCF initializes and when the AP is in CP, the stations requiring transmission of some data request the PC in AP to add them in the polling list. Reason being 16 IEEE 802.11 MAC Layer once the technique begins, the PC will only poll those stations for the channel access which are in its polling list. Once the AP is activated with PCF mode, the PC polls the stations, in its list, to access the channel by sending a BEACON signal. A BEACON frame consists of three things: the maximum duration of the CFP, the BEACON interval and the BSS identifier. Before sending the BEACON signal, the PC will wait for a PCF Interframe Space (PIFS). Once there is no activity, the PC starts with the polling process by following the polling list. Another important thing is that PIFS is less than DIFS. As a result, PCF always gets priority over DCF and no interruption is encountered [16]. PIFS for different PHY types are listed in [3]. Each station maintains its own Network Allocation Vector (NAV) [16]. This NAV stores the maximum duration of the CFP so that the stations cannot occupy the channel (since the stations to get access to the channel are polled by the PC and they already exist in the polling list). After the exchange of some signals, the station is ready to transmit. If the station has nothing to transfer, it has to send a NULL frame. In the same way, the PC continues the polling process following the list maintained in the CP. Since PCF also has the provision of CP, when the AP reaches that state, the channel access scheme is switched to DCF. Figure 3.2 [16], [19] and [21] shows the general working of the PCF scheme. 3.2.2 Problems with the PCF Although PCF is a lot better in terms of QoS as compared to DCF, it still has some shortcomings, some of which are discussed as follows [19] and [22]. • Polling algorithms are important aspect for effective implementation of PCF. In most of the cases, RR scheme is used. Its drawback is that it cannot differentiate between the types of data traffic and which type should be given preference. • The time span of the SuperFrame (SF) is important, since it directly affects the total delay encountered by system. • As described above, a SF alternates between CP and CFP. This increases the overhead and affects the throughput of the system. • In PCF, the PC lacks the tendency to determine the queuing order. • PC does not have the inner conditions of the channel to be accessed. Maybe, it will keep on polling the station when the channel is in bad condition and it will result in a drop of throughput. 3.2.3 Polling Mechanism In PCF, the polling schemes play an important role in the efficient application of the technique. Some of the polling schemes, for the efficient utilization of PCF in terms of multimedia services, are proposed in [21]. The schemes are Round-Robin 3.2 Point Coordination Function 17 Figure 3.2: Graphical Representation of PCF (RR), First-In-First-Out (FIFO), Priority and Priority-ELF (Effort-Limited Fair). RR, FIFO and priority schemes have one thing in common. When the PC is transmitting frames to a particular frame, it also combines the polling frames with it. If a station requires an ACK from the PC, instead of sending it separately, PC combines the ACK with the transmitted data packet. This whole process is for the efficient utilization of the wireless medium. • Round Robin: In this polling method, the CFP is set to a specific value, say Max_CFP. Once the PC has access over the channel, it searches for the station with the minimum address in the BSS. Once the PC finds that it has to perform data transmission with the station, it starts with the PCF procedure by sending the respected frame. There is a particular list of stations that are being served by a BSS. In RR, the ACK from the station is not that important. Even if the PC does not receive the ACK, it will switch to the next station according to the maintained polling list. Once the PC has entertained all of the stations, it re-starts the process with the lowest address station. This continues until we have reached Max_CFP. • First-In-First-Out: This scheme is quite clear from its name. The station to be polled depends on the queue maintained in the PC0 s polling list. The difference between FIFO and RR is that when the station has not received 18 IEEE 802.11 MAC Layer the ACK from the polled station, it assumes that there has been some transmission error and it prepares for the re-transmission. It also maintains a counter for the number of retries and continues to transmit until it has reaches the maximum value of the number of retries. When the PCs queue is empty, it acts as RR but as soon as any frame enters the queue, it switches to the FIFO scheme. • Priority: In this scheme, priority is given to the station having the highest Type of Service (TOS). Once the PC recognizes the station with the highest TOS, the process of frame transmission begins. In the same way, the PC switches to the station with the second highest TOS after the data transmission of the first one. The PC will serve all the stations with the TOS greater than 1 and when it comes to Best Effort Traffic (e.g. data traffic), the PC adapts the RR scheme for polling. In this scheme, every station has a flag segment in the MAC header. The purpose of this flag is to tell the PC that a particular station has data to send or receive. This flag comes into play once the station is polled by the PC. In the response, if the flag is set to TRUE, the PC will have to send a poll frame to that station. • Priority ELF: In ELF, there is a threshold i.e. the per-flow threshold. The ELF scheduling confirms that the data flows, experiencing an error rate below this per-flow threshold, will get their required services. In this scheme, the priority scheme utilizes the concept of ELF scheduling to perform the polling mechanism. The mechanism of TOS is the same as in the Priority scheme. However, in priority ELF, the PC also keeps track of the counter exhibiting the amount of data transmitted by a station in a particular period of time. It should be greater than 1. If the counter value is less than 1, the PC will not poll the corresponding station for the data transmission. The management of the polling list is very important. An efficient way to do this is shown in [22]. There are three main aspects when considering the polling mechanism namely, throughput, performance and the fairness of the channel accession. The polling list is divided into 4 categories: good channel with data (GD), good channel with no data (GN), bad channel (BC) and no response (NR) [22]. The PC should give the GD with the highest priority, followed by GN, BC and finally NR. In case of GD, the sorting of the stations is necessary. This sorting depends if the station has uplink, downlink or both kinds of data to transmit. There is a new variable i which assigns the priority of stations within the GD. The formula for channel efficiency calculation is given in [22]. Once getting these values, sorted in descending order and then a polling scheme is applied to the new polling list (the scheme is Non-Preemptive: the PC continues to poll the station as long as it has data to transmit or receive). This will have a positive impact on the total throughput of the system. One of the dis-advantages of using NP polling scheme is that it does not caters for the fairness. For fairness, RR polling scheme is more suitable. An efficient polling algorithm, which shows better results as compared to the RR or NP polling schemes, is also proposed in [22]. Chapter 4 Wireless in Industrial Automation The deployment of WLANs in industries is getting quite common in the world of today such as the control and automation industry. An industrial environment may include hostile surroundings such as extreme heat and movement of heavy machinery. In these conditions, sometimes, it becomes almost impossible to deploy wired communication systems. These conditions and scenarios gave rise to the extensive use of wireless LANs in the industrial industries. One of the major vendors for the IWLAN devices, deployed in various industries, is Siemens [23]. Wireless Sensor Networks (WSNs) are also getting popular among the industrial applications. WirelessHART [2] is the industrial standard which provides a seamless wireless communication to be used in industries. WirelessHART provides a solution for the replacement of wired fieldbuses in the automation industry in order to reduce the cost and the mechanical wear and tear. Fieldbuses are basically computer network protocols and they support real-time distributed controls among communications. These fieldbuses lie between the Programmable Logic Controllers (PLC) and the devices which are actually performing the work such as sensors and actuators. With the help of IWLANs, the distributed I/O devices can communicate wirelessly. Some of the common applications are the driverless transport machines, conveyor belts, and remote control applications [24]. 4.1 Industrial Automation Requirements The use of WLAN for Industrial Automation is increasing day by day. However, the conventional IWLAN scheme is not suitable for all the Industrial applications, deterministic communication, being the major one. For effective utilization of WLAN in industrial automation, there are some requirements which have to be met. From [25] and [26], the major industrial automation requirements are listed below: 19 20 Wireless in Industrial Automation • Safety and Security requirements: Safety, in terms of equipments and people working in the industry, should be the first priority before deploying any kind of system. Network security is also an important factor. It should be taken into account that the transmitted data is safe from unauthorized users and avoiding access to hackers. • Delay requirements: In industrial automation, data must be transmitted with the minimum possible delay. In order to have a precise control communication, latency and jitter should be minimized. • Deterministic requirements: The data transmission should be time-bounded. Deterministic communication is very important in industrial automation. Incase of mobility, this requirement is necessary hence giving rise to real-time handover. • Network Redundancy: In a given network, there should not be a single point of faliure. For instance if something goes down at one point, it should not affect the whole system and it should continue with the production. • Support for large number of devices: The network topology should be flexible i.e. any addition to the existing network should be possible. In a given industrial environment, the network might have nodes of the order of hundreds. For conventional IEEE 802.11, the efficiency is reduced with the increase in the network size. In an industrial application, latency is very important. Usually, the transmitted signals are of very short length and if the transmission is not within the specified deadline, it is of no use. Table 4.1 [26] highlights some of the requirements for the successful deployment of a large scale Wireless Sensor Network (WSN). Sensor Network Application Monitoring and supervision Vibration sensor Pressure sensor Temprature sensor Gas detection sensor Closed loop control Control valve Pressure sensor Temprature sensor Flow sensor Torque sensor Variable speed drive Interlocking and Control Proximity sensor Motor Valve Protection relays Delay Range Battery Life Update Freq. Security s ms s ms 100 100 100 100 m m m m 3 3 3 3 years years years years sec - days 1 sec 5 sec 1 sec low low low low ms ms ms ms ms ms 100 100 100 100 100 100 m m m m m m > > > > > > 5 5 5 5 5 5 years years years years years years 10 - 500 10 - 500 500 ms 10 - 500 10 - 500 10 - 500 ms ms ms ms ms medium medium medium medium medium medium ms ms ms ms 100 100 100 100 m m m m > > > > 5 5 5 5 years years years years 10 10 10 10 ms ms ms ms medium medium medium medium - 250 250 250 250 Table 4.1: Typical Requirements for Industrial Wireless Sensor and Actuator Networks in the Process Automation Domain 4.2 Standards 21 • Monitoring and Supervision: Sensors providing diagnostics and monitoring functions lie in this category. The update time for these applications typically varies from 1 sec to several days. These type of applications are insensitive to packet losses and / or jitter, since they are reserved for monitoring and supervision purposes. • Closed Loop Control: The main purpose of closed loop is to stabilize the process by controlling the actuators using preset readings. Typically, these kind of applications are affected by delays and jitter. These applications can bear the loss of packets and the wait time for retransmissions to some extent. • Interlocking and Control: Interlocking and control are very important in terms of a large control application. They correspond to discrete signaling. For example, to enable a start command for a particular control application, there might be several conditions to be fulfilled. These start, stop etc commands are termed as interlocks and are highly sensitive to delays. Generally, these applications are required to start or stop a particular machine and hence can cause a lot of damage if we have deadline misses. 4.2 Standards Depending on the requirements, there are different standards and protocols for data transfer in industrial applications. In some cases, data is tranmitted via cable networks e.g. CAN, EtherCAT and PROFINET I/O. Bluetooth, Wireless HART and IWLAN are the wireless networks used for the purpose of data transfer. 4.2.1 Controller Area Network CAN bus is the most extensively used industrial cable network. CAN bus was originally developed to be used with the automobile industry. However, due to some of its properties, such as speed, cost-effectiveness and utilization in real time applications, CAN bus has made its way into other industrial applications. For a given scenario, where a large number of short messages are to be transmitted for multiple users, CAN bus is ideal to use. CANopen and DeviceNet are some of the protocols used in industrial applications. However, users can also design their own protocols according to the application requirements. 4.2.2 PROFINET I/O PROFINET IO is an Ethernet based fieldbus protocol used in the automation industry. It is designed for the transmission of both real-time and non-real time data. For that, it uses a switched 100 Mbps network [27]. It is one of the IEC 61784 standards. Since real time data transmissions are of great importance industries, 22 Wireless in Industrial Automation PROFINET IO has a dedicated layer for that which lies above the Ethernet [27]. PROFINET IO is termed as a unique Ethernet based protocol and it has some benefits over the traditional physical layers [28]: • Provides high speed i.e. it skips the TCP/IP stack and the time required for the message delivery in case of real-time data traffic. • Unified integration with PROFIBUS [29]. • Supports time-dependent applications. • Easy deployment. PROFINET IO is divided into three types of devices: IO-Controllers, IODevices IO-Supervisors. IO-Controllers initiate the automation process whereas the IO-Supervisors are the end point devices e.g. monitoring or the diagnostic tools. IO-Devices serve as moderators between these end point devices e.g. sensors, actuators etc [28]. 1412 bytes is the maximum payload size for PROFINET IO [27]. This is to avoid fragmentation in case of real-time transmissions. 4.2.3 Bluetooth The wireless technology, Bluetooth, serves the purpose of data exchange over short ranges. Generally, bluetooth has a range of approximately 10 m but this range can be extended upto several hundred meters by using a long-range module. It has a maximum data throughput of 780 Kbps. However, Bluetooth v2.1+EDR (Enhanced Data Rate) gives an approximate data throughput of 2.1 Mbps [30]. In terms of delay, the bluetooth technology provides a latency of 5-10 ms. Considering the above mentioned properties, the Bluetooth technology (IEEE 802.15.1), in terms of industrial applications, serves the purpose of wirelessly integrating various automation devices. Devices which are cost-effective and have low power consumption are practically used within the Bluetooth technology. Using the Bluetooth PHY and a proprietary MAC, ABB has developed the Wireless Interface to Sensors and Actuators (WISA) system [1]. The communication between the sensors, equipped with the WISA technology, takes place in accordance with the IEEE standard radios which are same as that of bluetooth. The protocol for communication is designed by ABB, which works on the principle of TDMA and frequency hopping to provide enhanced and collision-free transmission. The design, implementation and working of the WISA system is explained in [31]. 4.2.4 Wireless HART HART stands for Highway Addressable Remote Transducer. WirelessHART is termed as a wireless sensor networking technology. It uses mesh-network topology. It is considered to be a robust wireless protocol for various industrial applications e.g. process management, control and asset management applications. It 4.2 Standards 23 utilizes the 2.4 GHz ISM band with IEEE 802.15.4 compatible Direct Sequence Spread Spectrum, channel hopping and Time Division Multiple Access [27]. The key characteristics of WirelessHART are Simple (ease of installation), Reliable (robust, compatible with other wireless technologies) and Security (uses Advanced Encryption Standard (AES) for security) [2]. Figure 4.1: Wireless HART Mesh Network (Courtesy of ABB) Figure 4.1 depicts the general network of wirelessHART. A WirelessHART networks constitutes of 5 parts:Gateway: provides connection between the wireless and the control network, Access Point: resides inside the gateway, provides the radio interface, Network Manager: responsible for the wireless network management, configures and sustains the meshed network, Security Manager: sharing and management of security keys, also holds the authorization list, Field Devices: end point devices, performing the real operation. [2] and [27] In a WirelessHART wireless network, each field device in the mesh network acts as a router. Even if the data packet is not assigned to the corresponding device, it will redirect the package to the respective device using its routing capabilities. It is the responsibility of the Network Manager to determine and establish the most efficient routes for the data transfer. It also adds the concept of redundancy. E.g. if a particular path is congested, the network manager will assign another path for the data to travel, which is going to be reliable and efficient also. WirelessHART works on the principle of Time Division Multiplexing (TDMA) and uses the concept of Frequency Hopping Spread Spectrum (FHSS). With TDMA, the participants are allocated a slot of 10 ms each, for transmission of data, whether it is from the sensor node to the gateway or from the gateway to the actuator. 24 Wireless in Industrial Automation This synchronization allows to have a reliable and collision-free transmission. Due to these properties, WirelessHART is ideal for industrial automation applications. As it is said, every cloud has a silver lining. WirelessHART also comes with some drawbacks. When it comes to applications requiring a fast update frequency rate, WirelessHART fails due to the predefined 10 ms slot. Considering the control loop defined in Figure 1.2, an application with an update frequency rate of 10 ms will not be supported by WirelessHART since it will take 10 ms time from sensor to the gateway and another 10 ms delay from the gateway to the actuator. This means that the total delay time will be 20 ms plus the time taken by the controller and the process plant. Another drawback is the limitation on the data payload size (0-127 bytes). 4.2.5 IWLAN The use of WLAN for industrial applications is termed as IWLAN. In an industrial environment, sometimes it is not feasible to use wired connectivity for data transfer. Also, on many occasions, it is impossible to deploy wired connections between the devices. IWLAN application arises in these situations. Some of the applications of IWLAN are in Automation Control, Manufacturing Industry, Process Control and Motion Control. The application of wireless communication technology in the control industry implies to the extension of a particular network to physical locations, some of which are not easily accessible using a wired connection. In an industrial environment, due to some physical parameters, robustness is required and there are some areas, where it is not possible for cable network to arrive. As a solution, the concept of IWLAN is introduced. As discussed in Chapter 3, there are two multiple access schemes for IEEE 802.11 namely, DCF and PCF. DCF being the non-deterministic scheme and PCF being the deterministic scheme. Generally, for the IWLAN implementation, CSMA/CA is used for the purpose of medium access which corresponds to the DCF scheme. This method is also termed as the Contention Scheme where each node, in the network, has to content with each other in order to win the medium access. The details are explained in Section 3.1. It can be seen that CSMA/CA is based on the principle of deterministic scheduling. Since the medium is shared by multiple nodes, it also increases the transmission delay, encountered by the transmitted data packets. In industrial applications i.e. control applications etc, latency plays an important role and there are strict thresholds for this purpose. In case of CSMA/CA, the system will encounter longer delay, which will eventually affect the overall system performance. PCF is considered to be a deterministic approach and is thought to give better results in terms of throughput and more importantly, delay. Section 3.2 explains the access scheme in detail. 4.3 Advantages of Wireless Technologies in Industrial Automation 25 4.3 Advantages of Wireless Technologies in Industrial Automation The wireless sensor network technology is increasingly being implemented in modern world Industrial Automation applications. Following are some of the highlighted benefits of using WSN in industrial Automation. The wireless technologies are increasingly being implemented in modern world Industrial Automation applications. Following are some of the highlighted benefits of using wireless networks in industrial Automation. • Cost: Cost effective techniques play an important role in Industrial automation. Using wireless networks in industrial applications highly reduces the cost and are easy to deploy. As an example, green field installation could cost upto $ 200 per meter to deploy a wired network in an industrial field and around $ 1000 in an offshore installation. In addition, maintainance, troubleshooting etc are required on regular basis which increases the cost and requires more time. • Flexibility: With the use of wireless technologies, it is possible to extract useful data from equipments, which used to be non-instrumented previously. The flexibility factor come with the benefit that the deployed wireless networks in automation industries can be utilized to temporarily measure certain process values without even having a wired network installed. • Emerging Applications: Wireless applications such as wireless control applications, empoweing mobile workers, integration with the non-traditional signals etc arise with the use of wireless technologies in Industrial automation. These emerging wireless applications are useful in terms of Industrial applications. • Availability: In order to prevent any catastrophic event, communication in an Industrial application should be deterministic. For instance, any precision error in a stone crushing industry can cause a loss upto $ 50,000 to $ 100,000 per hour. Wireless technologies, with the properties of being deterministic, are useful for industrial automation applications. Chapter 5 Performance Analysis of IEEE 802.11 PHY The following Sections contain the simulational results, along with the arguments, for the IEEE 802.11 PHY. For the purpose, MATLAB is used. 5.1 IEEE 802.11n Simulink Model The purpose of analyzing the IEEE 802.11n PHY simulink model is to study the IEEE 802.11 physical layer (Section 2.2) and observe the behavior under different circumstances. The simulation reflects an indoor environment. The effects of MIMO (Section 2.4.3), along with STBC and transmit beamforming are also analyzed. 5.1.1 Model Description The generic Simulink model for IEEE 802.11n is available at [32]. Following are the characteristics of the model: • 802.11n PHY model. • All the optional and compulsory data rates are supported i.e. 6, 9, 12, 18, 24, 36, 48 and 54 Mbps. • BPSK, QPSK, 16- and 64-QAM modulation schemes supported. • Convolutional coding with supporting rate: 1 /2 , 2 /3 , 3 /4 and 5 /6 . • OFDM with 20 MHz supported: 52 data carriers. • MIMO Detection. • TGn channel supported. 27 28 5.1.2 Performance Analysis of IEEE 802.11 PHY Channels This Section is about the MIMO channel modeling for indoor wireless communications. However, having a single transmit antenna and a single receive antenna is considered to be a part of MIMO. Large-scale propagation and Small-scale propagation, are the two types of propagation models in a wireless channel model. The former deals with the signal losses caused over distance and the latter corresponds to the signal losses caused by small changes in the distance. The details about the propagation models are in [11]. For indoor wireless communications, a number of MIMO channel models, based on the TGn specifications, are designed and are detailed in [33]. The High Throughput Task Group (TGn) has designed six channel models, namely A - F. These channel models are designated for different indoor configurations. Channels A - C are specifiied for smaller environments i.e. residential areas and channels D - F are specified for larger indoor areas such as office-type or industrial environments. In an industry, multi path effect is an important factor to consider, since there is no LoS communication in industrial applications. Also, the rms delay factor, of the channel selected, reflects an indoor industrial environment. Here, channel D, with nLoS, is used for the simulation purposes, which is a Rayleigh fading MIMO channel, corresponds to larger areas and has an rms delay spread of 50 ns. Normally, this category of channel models (D - F) show an rms delay spread of 50 to 150 ns [11]. On the other hand, channels A - C encounter an rms delay spread varying from 0 to 30 ns [11]. In wireless communications, rms delay spread corresponds to the multipath properties of a communication channel. In simple words, it is the time duration between the arrival of earliest and the latest multipath component. 5.1.3 Parameters Table 5.1 [5] shows the parameters used for the simulation purposes. Attributes Channel MCS Payload Size MIMO Configurations Value AWGN and TGn 13, 14 and 15 1000 Bytes 2x2 2 x 2 with beamforming 4 x 2 with STBC Table 5.1: Parameters used for Simulating IEEE 802.11n PHY Simulink Model As stated above, the Simulink model for IEEE 802.11n supports OFDM with 20 MHz. Corresponding to that, three values for Modulation and Coding Scheme (MCS) i.e. 13, 14 and 15 are used in the simulation. Table 5.2 shows the rate dependent parameters when the number of spatial streams (NSS ) is equal to 2. 5.1 IEEE 802.11n Simulink Model MCS Modulation R 13 14 15 64-QAM 64-QAM 64-QAM 2 NBP SCS (iss ) NSD /3 6 /4 6 5 /6 6 3 29 52 52 52 NSP NCBP S NDBP S 4 4 4 624 624 624 416 468 520 Data Rate(Mbps) 104.0 117.0 130.0 Table 5.2: MCS Parameters for 20 Mhz, NSS = 2, NES = 1, EQM NOTE: The data rates in the last column correspond to a Guard Interval (GI) of 800 ns. The parameters in Table 5.2 are defined as follows: NSS = Number of spatial streams MCS = Modulation and Coding scheme index R = Code rate NBP SCS (iss ) = Number of coded bits per single carrier for each spatial stream, iss = 1,...,NSS NSD = Number of complex data numbers per spatial stream per OFDM symbol NSP = Number of pilot values per OFDM symbol NCBP S = Number of coded bits per OFDM symbol NDBP S = Number of data bits per OFDM symbol NES = Number of BCC encoders for the DATA field The transmit beamforming is a method in which the spatial streams, for the signal to be transmitted, are determined using the channel estimates. The purpose of beamforming is to improve the signal strength at the receiving antenna. Beamforming can either be explicit or implicit. In the former technique, the required matrices, for the purpose of beamforming, are send by the receiver side. In the latter method, the transmitter side uses the transpose of its own channel estimation and performs beamforming. Explicit beamforming comes with the assumption that the channel response, from the transmitter to the receiver antenna, and vice versa is same. 5.1.4 Results Following are the Packet Error Rate (PER) vs Signal to Noise Ratio (SNR) curves, obtained after running 15 number of simulations. Figure 5.1 depicts PER Vs. SNR curve for MCS values of 13, 14 and 15 in the presence of an AWGN channel. MIMO technique, with 2 transmitter and 2 receiver antennas is used. The performance of IEEE 802.11n, in the presence of channel D nLoS, is shown in Figure 5.2. The channel is a typical delay spread channel [11]. It can be observed that at lower values of SNR, PER is on the higher side. 30 Performance Analysis of IEEE 802.11 PHY Figure 5.1: PER vs SNR for MCS = 13, 14 and 15 (AWGN, 2x2 Direct-Map) Figure 5.2: PER vs SNR for MCS = 13, 14 and 15 (Ch D nLoS, 2x2 Direct-Map) 5.1 IEEE 802.11n Simulink Model 31 Figure 5.3: PER vs SNR for MCS = 13, 14 and 15 (Ch D nLoS, 2x2 Beamforming) Figure 5.4: PER vs SNR for MCS = 13, 14 and 15 (Ch D nLoS, 4x2 STBC) 32 Performance Analysis of IEEE 802.11 PHY Therefore, in order to attain the same performance as AWGN channel, a higher averaged-SNR is required. The MIMO configuration is same as before i.e. 2 transmitter and 2 receiver antennas. Figures 5.3 show the effect of transmit beamforming (Section 2.4.3). The purpose of applying beamforming is already explained in Section 5.1.3. One can see a visible improvement in SNR (although minute, 2-3 dB). The reason for this small improvement in SNR is due to the use of equal number of spatial streams and transmitter antennas i.e. 2. This reduces the diversity order. Finally, the effect of STBC (Section 2.4.3) is shown in Figure 5.4. The performance is close to AWGN and an almost 5-8 dB improvement in SNR as compared to 2 x 2 channel D nLoS and beamforming. For this scenario, 4 transmitter and 2 receiver antennas are used. This performance improvement is due to the increase in the transmit diversity, which is due to the use of 4 x 2 transmission. Data rate increases with the increase in MCS value (Table 5.2). High data rate corresponds to high signal transmission power which directly affects the SNR at a given value of PER. This can be seen in Figures 5.1, 5.2, 5.3 and 5.4. 5.2 Throughput and Delay Limits This Section describes the theoretical limitations, in terms of throughput and delay, for IEEE 802.11a and IEEE 802.11b. It also shows the impact of using aggregation on the throughput for different data rates. The equations for the throughput and the delay analysis are adapted from [34] and [35]. 5.2.1 Assumptions • The channel is functioning under ideal conditions i.e. there is no InterSymbol Interference (ISI) and there is no noise added by the channel. In non-ideal case, the achieved throughput is always less than Maximum Throughput (MT) and the experienced delay is always greater than Minimum Delay (MD). • For a particular transmission cycle, only one station is able to send data packets and the rest of the nodes in the network can only receive and acknowledge packet reception. 5.2.2 Derivation of MT and MD for IEEE 802.11a The following derivation is based on [34] and [35]. A DCF transmission cycle constitutes of a DIFS interval, Backoff time, an SIFS interval, Data and Acknowledgement transmission time. TDCF = TDIF S + TB + TDAT A + TSIF S + TACK (5.1) 5.2 Throughput and Delay Limits 33 TB is the average backoff time and is given by: TB = CWmin σ 2 (5.2) CWmin = Minimum contention window size σ = Slot time The data transmission time (TDAT A ) is given by: TDAT A = TP + TP HY hdr + TM AChdr + TP AY LOAD (5.3) TP = Physical preamble transmission time TP HY hdr = PHY header transmission time TM AChdr = MAC header transmission time TP AY LOAD = Data payload transmission time Similarly, the acknowledgement transmission time (TACK ) is given by: TACK = TP + TP HY hdr + TACK_f rame (5.4) TACK_f rame = Acknowledgement frame transmission time From Figure 2.3, the DATA field corresponds to the number of coded bits in an OFDM symbol [4]. For the purpose, the message length in increased to be the multiple of data bits per OFDM symbol (NDBP S ). Therefore, to cater the pad bits, a ceiling function is used. We have: 16 + 6 + 8LM AChdr + 8LP AY LOAD TDAT A = TP + TP HY hdr + τ (5.5) NDBP S LM AChdr = MAC header length in bytes LP AY LOAD = Data payload length in bytes τ = OFDM symbol delay NDBP S = Data bits per OFDM symbol In (5.5), let us consider that the fraction, within the ceil function, be denoted by z. For IEEE 802.11a, the value of LM AChdr is 28 bytes, corresponding to the data rates i.e. 6, 9, 12, 18, 24, 35, 48 and 54 Mbps, NDBP S is 24, 36, 48, 72, 96, 144, 192 and 216 respectively. In Figure 2.3, the PSDU consists of 28 bytes of MAC header, 0 - 2312 bytes of data payload and 4 bytes of FCS. So, for a given data rate, say 6 Mbps, dze ≥ 3. The same follows for the acknowledgement transmission time: 16 + 6 + 8LACK_f rame TACK = TP + TP HY hdr + τ NDBP S LACK_f rame = Acknowledgement frame length in bytes (14 bytes) (5.6) 34 Performance Analysis of IEEE 802.11 PHY From (5.1), CWmin σ TDCF = TDIF S + + 2TP + 2TP HY hdr + TSIF S 2 16 + 6 + 8LM AChdr + 8LP AY LOAD 16 + 6 + 8LACK_f rame +τ +τ NDBP S NDBP S (5.7) The maximum throughput (MT) is given by: MT = 8LP AY LOAD TDCF + 2δ (5.8) Table 5.3 shows the parameters for IEEE 802.11a/b. Note that NDBP S is rate dependent as shown in Table 2.2. Applying the limit theorem (NDBP S →∞), gives us the Throughput Upper Limit (TUL). TUL = 8LP AY LOAD 2TP + 2TP HY hdr + 2δ + TDIF S + TB (5.9) The time between the starting of a transmission cycle and the packet0 s successful reception is termed as Packet Delay. The minimum encountered delay consists of a DIFS interval, backoff time and data transmission time. M D = TDIF S + TB + TDAT A (5.10) The values of TDIF S , TB and TDAT A have already been derived in the first part this Section. Substitution gives us: CWmin σ 16 + 6 + 8LM AChdr + 8LP AY LOAD M D = TDIF S + +TP +TP HY hdr +δ+τ 2 NDBP S (5.11) Note that NDBP S is rate dependent as shown in Table 2.2. Applying the limit theorem (NDBP S →∞), gives us the Delay Lower Limit (DLL). DLL = TP + TP HY hdr + δ + TDIF S + TB (5.12) δ in (5.9) and (5.12) is the propagation delay. 5.2.3 Derivation of MT and MD for IEEE 802.11b The equations for MT and MD, for IEEE 802.11b are the same as those for IEEE 802.11a i.e. (5.8) and (5.11), respectively. The only difference comes in the data transmission time (TDAT A ) and the acknowledgement transmission time (TACK ). In IEEE 802.11b, ratio of the packet size to the rate, at which the transmission occurs, equals the transmission time. Therefore, for IEEE 802.11b, (5.3) takes the form as follows: TDAT A = TP + TP HY hdr + 8LM AChdr + 8LP AY LOAD 1000000RDAT A (5.13) 5.2 Throughput and Delay Limits 35 RDAT A = Transmission rate Similarly, for TACK , (5.4) takes the form as follows: TACK = TP + TP HY hdr + 8LACK_f rame 1000000RACK (5.14) RACK = Control rate The equations for the Throughput Upper Limit (TUL) and the Delay Lower Limit (DLL) are the same as (5.9) and (5.12), respectively. The supported rates for IEEE 802.11b are 1, 2, 5.5 and 11 Mbps. Therefore, the (data rate, control rate) pairs are (1, 1), (2, 2), (5.5, 2) and (11, 2). 5.2.4 Parameters Table 5.3 [5] defines the attributes and their corresponding values used for the simulations to follow in this Section. Attributes TSIF S (µs) Slot time - σ (µs) TDIF S (µs) TP HY hdr (µs) TP (µs) CWmin (slots) MAChdr (bytes) CRC (bytes) Propagation delay - δ (µs) OFDM symbol delay - τ (µs) PHY layer peak rate (Mbps) 802.11a 16 9 34 4 16 15 28 4 1 4 54 802.11b 10 20 50 48 144 31 28 4 1 11 Table 5.3: MAC and PHY Parameters for 802.11 a/b 5.2.5 Results In real-time applications, we come across some limitations for the available throughput and encountered delay. Figure 5.5 shows the throughput upper limit (TUL) and the ideal throughput attained when the frame size is fixed to 1000 bytes [35]. It is observed that one cannot exceed 50 Mbps throughout with a payload size of 1000 bytes. Figure 5.6 shows the maximum achievable throughput for different data rates with a varying payload size. For a payload size of 1000 bytes at data rate of 54 Mbps, the throughput is almost 25 Mbps. However, for higher rates the performance is not good. 36 Performance Analysis of IEEE 802.11 PHY Figure 5.5: Ideal TUL for frame size = 1000 Bytes Figure 5.6: Maximum Throughput vs Payload size for different Data Rates (IEEE 802.11a) 5.2 Throughput and Delay Limits 37 This shows that in practice, the achievable maximum throughput is always less than the actual rate. Figure 5.7 illustrates the Delay lower limit (DLL) and the delay experienced for different data rates. The DLL is the same for all the rates i.e. 121.5 µs. Figure 5.7: Minimum Delay vs Payload size for different Data Rates (IEEE 802.11a) For a payload size of 1000 bytes and rate of 54 Mbps, the maximum delay is almost 280 µs. The Bandwidth efficiency is defined as the ratio of throughput and the data rate. Figure 5.8 depicts the bandwidth efficiency curves for different rates. For a payload size of 1000 bytes, the bandwidth efficiency for 54 Mbps is almost 46 %. The bandwidth efficiency for other rates can be seen from Figure 5.8. However, by doubling the rate, one can have an enhanced throughput. Figures 5.9 and 5.10 show the effect of doubling the rate for 6 Mbps and 54 Mbps, respectively. For 2 x 6 = 12 Mbps, we have an increase of around 6 dB. For 2 x 54 = 108 Mbps, we experience an increase of almost 10 dB. Figure 5.11 shows the percentage increase in throughput by doubling the rate for different rates. 38 Performance Analysis of IEEE 802.11 PHY Figure 5.8: Bandwidth Efficiency for different Data Rates (IEEE 802.11a) Figure 5.9: Effect on Throughput by doubling the Rate (2x6=12Mbps)(IEEE 802.11a) 5.2 Throughput and Delay Limits 39 Figure 5.10: Effect on Throughput by doubling the Rate (2x54=108Mbps)(IEEE 802.11a) Figure 5.11: Percentage increase in Throughput by doubling the Rate (IEEE 802.11a) 40 Performance Analysis of IEEE 802.11 PHY Figure 5.12: Throughput when the Channel is adding Noise (IEEE 802.11a) In Figures 5.5 to 5.11, an ideal channel was considered (Section 5.2.1). Figure 5.12 depicts the actual effective throughput when the frame size is varying and the PHY layer data rate is 216 Mbps. The sharp curve for BER = 10−3 is showing the concept of CSMA. The lower the bit error rate is, the more the effective throughput we have. 5.2 Throughput and Delay Limits 41 Figure 5.13: Maximum Throughput vs Payload size for different Data Rates (IEEE 802.11b) Figure 5.14: Minimum Delay vs Payload size for different Data Rates (IEEE 802.11b) 42 Performance Analysis of IEEE 802.11 PHY Figure 5.13 shows the TUL and the MT for IEEE 802.11b for different data rates. It can be observed that, at a payload size of 1000 bytes, the TUL is almost 11.5 Mbps. This shows that, even if the rate maximized, there exists a limitation to maximum achievable throughput. Figure 5.14 shows the DLL and the MD for IEEE 802.11b for different data rates. The lowest achievable delay, in the case of IEEE 802.11b, is almost 523 µs. 5.3 Packet Aggregation The IEEE 802.11 MAC has overhead which affects the effective throughput. [36] suggests some IEEE 802.11 enhancements for the MAC overhead improvement. One method, along with the simulation results is explained in this Section. The reader can refer to [36] for the details for other techniques. 5.3.1 DSDSRFA Figure 5.15 [36] depicts the concept of Distributed, Single-Destination and SingleRate Frame Aggregation (DSDSRFA). It shows k PHY frames being sent without any space involved. Figure 5.15: Concept of DSDSRFA The start of the frame aggregation is initialized by PHY frame 1 and PHY frame k indicates that the frame aggregation has ended. There are k PHY frames in frame aggregation. 5.3 Packet Aggregation Figure 5.16: Comparison of TUL Figure 5.17: Comparison of MT 43 44 Performance Analysis of IEEE 802.11 PHY Figures 5.16 and 5.17 show the comparison of the original and the aggregated versions in terms of TUL and MT, respectively. For both simulations the value of k is taken as 2. The data rate used is 54 Mbps. It is clearly seen that there is an improvement in both, TUL and MT, when frame aggregation is applied. For a data payload size of 1500 bytes, the improvement in TUL is approximately 58 Mbps. For a same payload size, the improvement is MT is about 6 Mbps. Chapter 6 Performance Analysis of IEEE 802.11 MAC The following Sections contain the simulational results, along with the arguments, for the IEEE 802.11 MAC layer. For the purpose, NS-2 [37] is used. 6.1 6.1.1 NS-2 Overview NS-2 is an object-oriented, event driven network simulator which is developed at UC Berkley. Essentially, NS-2 is written in C++ and OTcl. OTcl is an objectoriented subdivision of Tcl [38]. At present, NS-2 is being widely used for the network simulation of local and wide area networks. NS-2 is open source and [39] states the contributions made to the NS-2. NS-2 is implemented using C++ and OTcl. There are always two corresponding hierarchies for every protocol or network objects implemented in NS-2, the compiled C++ hierarchy and the interpreted OTcl hierarchy [40]. Figure 6.1 [40] shows a simplified user0 s view to NS. 6.1.2 Network Components An NS network comprises of Compound network components and Basic network components. A partial class hierarchy is shown in Figure 6.2 [40]. TclObject acts as the root i.e. it is the superclass for all the OTcl library objects. As we move down the hierarchy, there is NsObject, which is the superclass for all the basic networking components. These basic network components are responsible for tasks such as packet handling. The network may also contain some compound network objects e.g. nodes and links. Possible number of output data paths decide whether the basic network components branch into Connector or Classifier. A single output data path corresponds to the Connector class and multiple output data paths correspond to the Classifier class. 45 46 Performance Analysis of IEEE 802.11 MAC Figure 6.1: Simplified User0 s View of NS Figure 6.2: Class Hierarchy (Partial) One of the essential components of NS-2 is a Node. A node entry object and a classifier make up a node. Figure 6.3 [40] shows the structure of nodes in NS-2. By default, NS has Unicast nodes. A unicast node consists of an address classifier and a port classifier. The former is responsible for unicast routing. In case of multicast operations, it is possible to creat Multicast nodes. Multicast nodes have the same structure of unicast nodes. However, in addition to that, they have a classifier to distinguish between unicast and multicast packets and a multicast classifier to undertake multicast routing. 6.1 NS-2 47 Figure 6.3: Node (Unicast and Multicast) Another important component of NS is a Link. Generally, the link object corresponds to a uni-directional simplex link. A bi-directional duplex link can be created by using the duplex-link member function of the Simulator object. It generates two simplex links in both the directions as shown in Figure 6.4 [40]. Figure 6.4: Link During network generations, an Agent can either be in an active or a reactive state. On the other hand, nodes and links act as reactive elements. Routing agents is such an example. Packets are actively generated by the routing agents and are sent to other agents, in the same line, in order to define possible routes within the network. Traffic generators are responsible to generate data within the network. They 48 Performance Analysis of IEEE 802.11 MAC can either be a data source (e.g. CBR) or an application simulating source (e.g. FTP). 6.1.3 Tracing NS-2 has a unique way of analyzing the results obtained after running the simulations. For the purpose, trace objects i.e. EnqT, DeqT, DrpT and RecvT are inserted within the corresponding link objects. Figure 6.5 [40] depicts the trace object insertion structure. As a result, useful information, about the packets passing through the trace objects, is stored in a predefined trace file. Figure 6.6 [40] shows a general format of a trace file. Figure 6.5: Insertion of Trace objects Figure 6.6: Trace Format example The segments of the trace file in Figure 6.6 are defined as follows: • event: Each trace file starts of with an event descriptor. +,−, d and s are some event descriptors which specify enqueue, dequeue, packet drop and packet receive respectively. • time: For a particular event, this part gives the simulation time in sec. • from node and to node: Source and destination nodes. • pkt type and pkt size: Type of the packet e.g. cbr, tcp, ack etc and the packet size in Bytes. • fid: IPv60 s flow id. The user can define this fid in the input OTcl script. • src addr and dst addr: Addresses for the source and destination in the form of node.port. 6.3 Results 49 • seq num: This part specifies the network layer protocol packet sequence number. For the case of UDP, this field is not used but NS keeps track for the sequence number for future analysis. • pkt id: Since this packet id is unique, it serves as an important part of the trace file. 6.2 Comparison between PCF and DCF In Industrial Automation, latency is an important factor. Usually, the transmitted signals are of very short length and if they are not received in time, they are of no use. The purpose of this Section is to compare PCF and DCF, in terms of deadline misses, under different scenarios. NS2 [37] is used for the simulation purposes. 6.2.1 Parameters Table 6.1 shows the attributes, and their values, used in the simulations. Attributes Slot time (µs) SIFS(µs) DIFS (µs) PIFS (µs) Transmit power (W ) Transmission range (m) Simulation time (sec) CWmin (slots) CWmax (slots) PLCP preamble (bits) PLCP header (bits) Beacon period (TU ) CFPMaxDuration (TU ) Values 20 10 16 12 0.2818 250 600 31 1023 144 48 200 180 Table 6.1: Parameters for PCF Vs. DCF Comparison Note: TU is Time Unit. 1 TU = 1024 µs. 6.3 Results This Section explains the generated scenario, along with the results and observations. The PCF patch for ns2 is available at [41]. The network size and data payload is varied from 5 to 20 nodes and 5 to 1500 bytes respectively. For simulating PCF, the number of priority nodes is also varied. All these scenarios are simulated with 1 and 2 Mbps of data rates. Figure 6.7 shows a simple five node star topology. PC is the server which is the point coordinator for PCF and base station incase of 50 Performance Analysis of IEEE 802.11 MAC DCF. The PC is a wired station and is connected to the wired backbone. The PC has a wireless range and the nodes (N) are randomly distributed within the range. Exchange of data packets is done between the PC and the nodes. Wired Backbone PC N N N N N Figure 6.7: Five Node Network Topology In the following, different environments are listed, under which PCF and DCF are compared. • 5 nodes, 1000 bytes. • 10 nodes, 1000 bytes. • 10 nodes, 1500 bytes. • 15 nodes, 512 bytes. • 20 nodes, 5 bytes. • 20 nodes, 100 bytes. 6.3 Results 6.3.1 51 Deadline misses of PCF and DCF for 1 Mbps Figures 6.8 to 6.13 depict the encountered delay vs number of received packets, in seconds, for the above mentioned scenarios for 1 Mbps. The deadline point is highlighted by a blue horizontal line in the Figures. This is also termed as the cbr interval i.e. the time between packet transmissions. As mentioned earlier, time is very crutial for Industrial applications. So, it is important for the packet to be received within the specified deadline. The x-axis shows the number of packets received and the experienced end-to-end delay is shown along the y-axis. By closely observing the Figures, one can see that the DCF access scheme is encountering deadline at an higher cbr interval than the PCF access scheme. The DCF and PCF schemes are explained in Sections 3.1 and 3.2 respectively. In case of DCF, the concept of CSMA / CA is used for the channel access. According to CSMA / CA, the station senses the channel to be idle. Once it has sensed the channel to be idle, the station waits another DIFS before starting the data transmission process. It also has a backoff timer which is reduced whenever the channel is sensed idle. This backoff timer is set randomly varying from 0 to the maximum size of contention window (CW). For the PCF access scheme, the concept of polling is used. The PC already maintains a polling list before the data transmission starts. There is a superframe or the beacon period which has the provision of both, the CFP and the CP. Once the process starts, the PC waits a PIFS before it sends the beacon signal to the stations, that are already in the PC’s polling list. Since, PCF has the provision of both, the CFP and the CP, and PIFS is less than DIFS, PCF scheme always gets priority over the DCF channel access scheme. Let us discuss Figure 6.8 in detail. The rate is 1Mbps, we have a setup with 5 active nodes and a payload size of 1000 bytes. One can see that when the cbr interval is set to 0.0499 sec, the DCF is encountering deadline misses at an higher cbr interval than the PCF scheme. The reason for the DCF scheme to encounter deadline misses at an higher cbr interval than the PCF scheme is due to the use of CSMA / CA. In DCF, since the nodes have to content for the channel access and it depends on the CW size, there is an increase in the delay factor, causing the deadline misses. On the other hand, for the PCF scheme, the concept of polling is used for the maximum size of the predefined contention free period, which overcomes the channel access fight between the stations in the polling list of the PC. However, as already stated above, PCF has both CFP and CP. When the CFP end is received, the channel access scheme automatically switches back to the conventional DCF channel access scheme. 52 Performance Analysis of IEEE 802.11 MAC Figure 6.8: Encountered delay vs Packets received for 5 nodes and 1000 bytes Figure 6.9: Encountered delay vs Packets received for 10 nodes and 1000 bytes 6.3 Results 53 Sudden spikes in case of PCF are due to that reason. The rest of the Figures (Figures 6.9 to 6.13) show the same behavior under different scenarios. Figure 6.10: Encountered delay vs Packets received for 10 nodes and 1500 bytes Figures 6.14 to 6.16 show the average end-to-end delay vs cbr interval for a data rate of 1 Mbps under the scenarios mentioned in Section 6.3. The deadline is marked with "deadline" in the Figures. In Figure 6.14, PCF and DCF are compared for two cases i.e. 20 nodes, 5 bytes and 5 nodes, 1000 bytes. The increase in the difference between the average end-to-end delay, at the point of deadline (0.042 sec for 20 nodes, 5 bytes and 0.0499 sec for 5 nodes, 1000 bytes), shows that DCF encounters deadline misses. Same explaination is valid for the rest of the cases (Figures 6.15 and 6.16). 54 Performance Analysis of IEEE 802.11 MAC Figure 6.11: Encountered delay vs Packets received for 15 nodes and 512 bytes Figure 6.12: Encountered delay vs Packets received for 20 nodes and 5 bytes 6.3 Results 55 Figure 6.13: Encountered delay vs Packets received for 5 nodes and 100 bytes Figure 6.14: Average end-to-end delay vs cbr interval (1 Mbps) 56 Performance Analysis of IEEE 802.11 MAC Figure 6.15: Average end-to-end delay vs cbr interval (1 Mbps) Figure 6.16: Average end-to-end delay vs cbr interval (1 Mbps) 6.3 Results 6.3.2 57 Deadline misses of PCF and DCF for 2 Mbps Figures 6.17 to 6.22 depict the encountered delay vs number of received packets, in seconds, for the above mentioned scenarios (Section 6.3) for 2 Mbps. Figure 6.17: Encountered delay vs Packets received for 5 nodes and 1000 bytes The blue horizontal line marks the deadline. In Figure 6.17, the cbr interval or the deadline is 0.0285 sec. From the Figure, it is quite evident that DCF encounters deadline misses at an higher cbr interval than PCF. The same explanation goes for the rest of the Figures. The reason for DCF, to encounter deadline misses at an higher cbr interval than PCF, is the same as described in Section 6.3.1. The only difference is that we can observe an enhanced performance, in all cases, since the data rate is 2 Mbps. High data rate means that the transmission power is high, which directly affects the encountered delay time. 58 Performance Analysis of IEEE 802.11 MAC Figure 6.18: Encountered delay vs Packets received for 10 nodes and 1000 bytes Figure 6.19: Encountered delay vs Packets received for 10 nodes and 1500 bytes Figures 6.23 to 6.25 show the average end-to-end delay vs cbr interval for a data rate of 2 Mbps under the scenarios mentioned in Section 6.3. The deadline is 6.3 Results 59 Figure 6.20: Encountered delay vs Packets received for 15 nodes and 512 bytes Figure 6.21: Encountered delay vs Packets received for 20 nodes and 5 bytes 60 Performance Analysis of IEEE 802.11 MAC Figure 6.22: Encountered delay vs Packets received for 5 nodes and 100 bytes marked with "deadline" in the Figures. In Figure 6.23, PCF and DCF are compared for two cases i.e. 20 nodes, 5 bytes and 5 nodes, 1000 bytes. The increase in the difference between the average end-to-end delay, at the point of deadline (0.035 sec for 20 nodes, 5 bytes and 0.0285 sec for 5 nodes, 1000 bytes), shows that DCF encounters deadline misses. Same explaination is valid for rest of the cases. 6.3 Results Figure 6.23: Average end-to-end delay vs cbr interval (2 Mbps) Figure 6.24: Average end-to-end delay vs cbr interval (2 Mbps) 61 62 Performance Analysis of IEEE 802.11 MAC Figure 6.25: Average end-to-end delay vs cbr interval (2 Mbps) 6.4 6.4.1 Observations Deadline Misses Table 6.2 shows the points, at which PCF and DCF are encountering deadline misses, when different parameters are varied. An important observation is that when the number of stations is increased, there is a rapid improvement in the PCF channel access scheme as compared to the DCF scheme in terms of deadline misses. When the network size increases, in DCF, there are more nodes in the process of contention and the probability of collision increases. When collision occurs, the CW increases about the power of 2 and eventually it encounters deadline misses earlier as compared to PCF. For the PCF, when we have more nodes and increase in the priority nodes, maximum data transfer will take place in the CFP. Hence, difference between the deadline misses (PCF Vs. DCF) is more obvious in a bigger network. Another important observation is that when the packet size increases, the improvement level of PCF falls down. Here, we can say that there is a trade-off between the packet size and the number of nodes. 6.4 Observations Scenario 1 Mbps Payload size = 100 bytes 5 nodes 10 nodes 15 nodes 20 nodes Payload size = 512 bytes 5 nodes 10 nodes 15 nodes 20 nodes Payload size = 1000 bytes 5 nodes 10 nodes 15 nodes 20 nodes 2 Mbps Payload size = 100 bytes 5 nodes 10 nodes 15 nodes 20 nodes Payload size = 512 bytes 5 nodes 10 nodes 15 nodes 20 nodes Payload size = 1000 bytes 5 nodes 10 nodes 15 nodes 20 nodes 63 PCF DCF 12 22 35 51 14 28 43 59 ms ms ms ms ms ms ms ms 41 ms 65 ms 86 ms 119 ms 42 ms 69 ms 90 ms 126 ms 47 ms 97 ms 148 ms 197 ms 50 ms 99 ms 151 ms 201 ms 11 17 28 38 ms ms ms ms 12 22 33 50 ms ms ms ms 22 32 50 68 ms ms ms ms 23 38 56 76 ms ms ms ms 26 ms 53 ms 86 ms 109 ms 28 ms 57 ms 89 ms 114 ms Table 6.2: Deadline Comparison of PCF and DCF 64 6.4.2 Performance Analysis of IEEE 802.11 MAC Bandwidth Efficiency Bandwidth efficiency or the bandwidth utilization is another important factor, which helps us in the process of investigating the behaviour of PCF and DCF, under different scenarios. Table 6.3 shows the utilization of bandwidth for PCF and DCF, for a given scenario. Payload Size (bytes) 100 200 300 400 500 Bandwidth Efficiency (%) PCF DCF 47 32 69 48 77 60 82 65 85 75 Table 6.3: Bandwidth Efficiency (PCF Vs. DCF) In terms of bandwidth efficiency, PCF has a better performance than DCF. The scenario shown in Table 6.3 is for a simple 2-node network. The data transmission rate is of 1 Mbps, 10 ms is the cbr interval or the update frequency and the data payload size is varied from 100 bytes to 500 bytes. When we increase the number of stations in the network, the bandwidth utilization for PCF remains the same. The reason is simply that PCF is deterministic. Since the delay time is same for all of the transmissions, increasing the number of nodes will also increase the overall system delay in a deterministic way. For example, a network has 5 nodes. If each nodes has a delay time of 1 ms, then the overall system delay will be 5 ms. However, for the case of DCF, the encountered delay consists of a random backoff. With the increase in number of nodes, the probability of collision also increases which consequently, increases the random backoff. In CSMA/CA, the nodes sense the channel to be idle before starting to send data. If a node never senses the channel to be idle, there will be an increase in the random backoff, which increases the overall system delay. The working mechanism of CSMA / CA is explained in Section 3.1.1. This behaviour is shown in Table 6.4. 6.5 Comparison between WLAN and WirelessHART Payload Size (bytes) 100 5 nodes 10 nodes 200 5 nodes 10 nodes 300 5 nodes 10 nodes 400 5 nodes 10 nodes 500 5 nodes 10 nodes 65 Bandwidth Efficiency (%) PCF DCF 47 47 28 19 69 69 35 27 77 77 51 40 82 82 59 40 85 85 65 58 Table 6.4: Bandwidth Efficiency (PCF Vs. DCF) with Varying Number of Nodes 6.5 Comparison between WLAN and WirelessHART This Section explains, for a given scenario, how WLAN is better then WirelessHART in terms of control loops. We have an update frequency of 20 ms and a data payload size of 100 bytes. Figure 6.26: The Basic Block Diagram of the Wireless System (Courtesy of ABB) For WirelessHART, we have a 10 ms fixed time slot, since it follows the concept of TDMA for channel access. This means that there is a fixed window for the data 66 Performance Analysis of IEEE 802.11 MAC transmission and the reception of acknowledgement. Ideally, with a data rate of 250 Kbps, maximum supported by WirelessHART, a packet takes around 4 ms to travel from a source to a destination. Figure 6.26 shows a top-level block diagram for a wireless sensor and actuator network. Considering a packet travelling from the Sensor Node to the Actuator, via the Gateway, for the case of wirelessHART, it will require two time slots, 10 ms each, to complete the one way process, since the slot is reserved for this purpose. The time required by the Controller and the Process Plant will be the same for both, WirelessHART and the WLAN. Therefore, the difference in the number of control loops is affected by the time from the Sensor Node to the Actuator, via the Gateway. On the other hand, WLAN encounters a delay of only 1.72 ms. This means that with an update frequency of 20 ms, WirelessHART can only entertain 1 control loop (ideally), whereas WLAN can squeeze in 5 control loops. Chapter 7 Conclusions and Future Work 7.1 Conclusions The use of Wireless Local Area Networks, in various applications, is at a high pace. Recently, their use in industrial applications has been started. In the field of industrial automation, there is a strict requirement for a signal to be received with in the specified deadline otherwise it is of no use. Encountering a deadline miss can cause a lot of damage in terms of money and personal health. The target of this thesis was to investigate the IEEE 802.11 PHY and MAC layer for its use in industrial automation. Section 5.1 shows the PHY layer analysis using the IEEE 802.11n simulink model. This simulation was carried out under the influence of a larger indoor area using different conditions. The results show that when the channel is AWGN, the effects on the PER Vs. SNR curves are not that severe. This is because in an AWGN channel, the only impairment is the white noise with a constant spectral density and it does not include the effects of fading. The affect of fading is observed when a n-LoS channel D is used. In this case, PER is higher at lower values of SNR. This behavior was catered by using the concept of beamforming which shows some improvements. However, the use of STBC shows good improvements since, there is an increase in the transmit diversity. Section 5.2 shows the theoretical upper and lower bounds on the throughput and the delay, respectively. These limitations show that by simply increasing the available data rate, without reducing the overhead, these values are bounded. Therefore, overhead reduction is necessary in order to have an enhanced performance in terms of throughput and delay. This problem ca be solved, to some extent, by using packet aggregation (Section 5.3). Section 6.2 investigates the main target of this thesis, i.e. the comparison of 67 68 Conclusions and Future Work MAC access schemes (PCF Vs. DCF) in terms of deadline misses. PCF and DCF were investigated under various scenarios. Observing the MAC schemes in different environments leads us to the conclusion that PCF has a far better performance than DCF in terms of deadline misses. From Table 6.2, it can be concluded that there exists a trade-off between the packet size and the number of stations involved in the network. However, increasing the available data rate and reducing the overhead, can enhance the performance of IEEE 802.11 (WLAN) in industrial applications. PCF is also efficient, in terms of bandwidth utilization, as compared to DCF, with varying number of nodes. Applications requiring a fast update of frequency cannot be handled easily by the wirelessHART. For example, with an update frequency of 10 ms, WirelessHART fails. Here, WLAN comes into play. However, it is not wise to use WLAN for control applications with slow update frequency rate, since WLAN requires more transmission power as compared to WirelessHART. WLAN is also suitable for monitoring applications where a large amount of data payload size is required. Finally, for a given scenario, more control loops can be squeezed in, in case of WLAN than WirelessHART. 7.2 Future Work In any field of study, there is always room for further research. As explained in Chapter 3, the PCF channel access scheme works on the principle of polling. However, the current PCF implementation has some deficiencies. The PC maintains a polling list, which contains the stations who need to use the channel for the purpose of data transmission. Currently, the polling mechanism is not adaptive i.e. there is no proper method to detect the level of emergency among the stations. This is very important in terms of industrial automation applications. This issue can be solved by putting the high-risk sensors in the priority list but is open for future research. A solution might be by keeping the count of recent polls and put the stations, with maximum number of recent polls, in the priority list. In the current implementation of PCF access scheme, if a polled station has no data to send, the PC waits until the poll times out. However, if the polled stations have a NODATA frame and the stations have no data to transmit, they can send this NODATA frame to the PC, notifying that the PC should move on to the next station in its polling list. This can save the waiting time of the PC and consequently, reduce the encountered delay time. Bibliography [1] R. Steigmann and J. Endresen, “Introduction to WISA: WISA - Wireless Interface for Sensors and Actuators,” White paper, ABB, July 2006. [2] H. C. Foundation, “Wireless HART Technology,” 2011. [Online]. Available: http://www.hartcomm.org/protocol/wihart/wireless_technology.html [3] IEEE 802.11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications, doi:10.1109/ieeestd.2007.373646 ed., IEEE, June 2007. [4] Part 11: Wireless LAN Medium Access Control (MAC) and Physical (PHY) layer Specifications: High Speed Physical Layer in the 5 GHz band, IEEE 802.11a WG, September 1999. [5] IEEE Standard for Information technology–Telecommunications and information exchange between systems–Local and metropolitan area networks–Specific requirements Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Specifications Amendment 5: Enhancements for Higher Throughput, IEEE, doi: 10.1109/IEEESTD.2009.5307322. [6] M. S. Gast, 802.11 Wireless Networks: The Definitive Guide, Second Edition. O’Reilly Media, Inc., 2005. [7] B. Crow, I. Widjaja, L. Kim, and P. Sakai, “IEEE 802.11 Wireless Local Area Networks,” in IEEE Communications Magazine, vol. 35, no. 9, pp. 116 –126, September 1997. [8] B. O’Hara and A. Petrick, The IEEE 802.11 Handbook: A Designer’s Companion. Standards Information Network IEEE Press, 1999. [9] B. Pearson, “Complementary Code Keying Made Simple,” Intersil Corporation, Tech. Rep., May 2000. [10] http://www.s3.kth.se/signal/grad/OFDM/URSIOFDM9808.htm. [11] T. Paul and T. Ogunfunmi, “Wireless LAN Comes of Age: Understanding the IEEE 802.11n Amendment,” in IEEE Circuits and Systems Magazine, vol. 8, no. 1, pp. 28 –54, First Quarter 2008. 69 70 Bibliography [12] J. Petranovich, S. Mujtaba, A. Kasher, C. Hansen, and J. Trachewsky, “Joint Proposal PHY Overview,” doc.: IEEE 802.11-05/1161r2. [13] D. Qiao and S. Choi, “Goodput enhancement of IEEE 802.11a wireless LAN via link adaptation,” in Proc. IEEE International Conference on Communications, vol. 7, June 2001, pp. 1995 –2000. [14] O. P. Ostos, “Performance Evaluation of IEEE 802.11 DYN: Extension of OFDM-based 802.11 a/g by dynamic OFDM schemes,” Diploma Thesis, Technische Universität Berlin, 2007. [15] M. Kobayashi, G. Caire, and D. Gesbert, “Opportunistic beamforming vs. space time coding in a queued downlink,” in Proc. 14th IST Mobile & Wireless Communication Summit, Dresden, GERMANY, June 19-23, 2005. [Online]. Available: http://www.eurecom.fr/publication/1674 [16] S. Shaaban, D. H. M. E. Badawy, and P. A. Hashad, “Performance Evaluation of the IEEE 802.11 Wireless LAN Standards,” in Proc. of the World Congress on Engineering, July 2008, pp. 781 – 786. [17] http://www.zytrax.com/tech/wireless/802_MAC. [18] IEEE 802.11 Standards, 2007. [19] J. Villalon, P. Cuenca, and L. Orozco-Barbosa, “Limitations and Capabilities of QoS Support in IEEE 802.11 WLANs,” in Proc. IEEE Pacific Rim Conference on Communications, Computers and Signal Processing, August 2005, pp. 633 – 636. [20] J. Geier, “802.11 Medium Access Methods,” 2002. [Online]. Available: http://www.wi-fiplanet.com/tutorials/article.php/1548381 [21] J.-Y. Yeh and C. Chen, “Support of Multimedia Services with IEEE 802.11 MAC protocol,” in Proc. IEEE International Conference on Communications, vol. 1, 2002, pp. 600 –604. [22] J. Kim, Y. Moon, T. Kim, and H. Lee, “Efficient Polling MAC Scheduler for IEEE 802.11 WLAN,” in Proc. International Conference on Information Networking, January 2008, pp. 1 –5. [23] http://www.automation.siemens.com/mcms/industrialcommunication/en/industrialwirelesscommunication/network_components/ Pages/networkcomponentsiwlan.aspx. [24] M. Blaho, L. Mrafko, M. Urban, and J. Murgas̆, “Networked Control Systems with PROFINET and IWLAN,” ISBN: 978-1-61804-031-2. [25] Y. Zheng, A. Xu, Y. Song, W. Zhao, and M. Liu, “Research on Deterministic Communication of Industrial Wireless Based on WLAN,” in Proc. International Conference on Electronic Measurement Instruments, August 2009, pp. 3–260 –3–264. Bibliography 71 [26] J. Åkerberg, “On Safe and Secure Communication in Process Automation,” Ph.D. dissertation, Mälardalen University, Sweden, 2011. [27] J. Åkerberg, M. Gidlund, T. Lennvall, J. Neander, and M. Björkman, “Efficient Integration of Secure and Safety Critical Industrial Wireless Sensor Networks,” in EURASIP Journal on Wireless Communications and Networking, September 2011. [28] http://www.rtaautomation.com/profinetio/. [29] http://www.rtaautomation.com/profibus/. [30] M. Andersson, “Wireless Technologies for Industrial Applications,” March 2012, version 2.1. [Online]. Available: http://www.connectblue.com/ [31] D. Dzung, C. Apneseth, J. Endresen, and J.-E. Frey, “Design and Implementation of a Real-Time Wireless Sensor/Actuator Communication System,” in Proc. 10th IEEE Conference on Emerging Technologies and Factory Automation, vol. 2, September 2005, pp. 433 – 442. [32] http://www.mathworks.com/matlabcentral/fileexchange/22137-simulinkmodel-of-the-ieee-802-11n-phy-layer-model. [33] “TGn Channel Models,” http://www.ieee802.org/11/DocFiles/03/11-03-094004-000n-tgnchannel-models.doc, May 2004, IEEE Std. 802.11 - 03/940r4. [34] Y. Xiao and J. Rosdahl, “Throughput and Delay Limits of IEEE 802.11,” in IEEE Communications Letters, vol. 6, no. 8, pp. 355 – 357, August 2002. [35] Q. Ni, T. L. Li, T. Turletti, and Y. Xiao, “Aggregation with FragmentRetransmission partial MAC proposal for IEEE 802.11n,” in doc.: IEEE 802.11-04/949r0, August 2004. [36] Y. Xiao, “IEEE 802.11 n: Enhancement for Higher Throughput in Wireless LANs,” in IEEE Wireless Communications Magazine, vol. 12, no. 6, pp. 82 – 91, December 2005. [37] http://www.isi.edu/nsnam/ns/. [38] http://www.tcl.tk/. [39] http://nsnam.isi.edu/nsnam/index.php/Contributed_Code. [40] http://nile.wpi.edu/NS/. [41] http://web.archive.org/web/20051227033205/http://www.sm.luth.se/ ∼dugdale/index/software.shtml.