Comparison of Indoor Ray Tracing and Measurement Results for 5 GHz Band Wireless Channel. A thesis presented to. the faculty of

Transcription

1 Comparison of Indoor Ray Tracing and Measurement Results for 5 GHz Band Wireless Channel A thesis presented to the faculty of the Russ College of Engineering and Technology of Ohio University In partial fulfillment of the requirements for the degree Master of Science Mousmi Samudra November Mousmi Samudra. All Rights Reserved.

2 2 This thesis titled Comparison of Indoor Ray Tracing and Measurement Results for 5 GHz Band Wireless Channel by MOUSMI SAMUDRA has been approved for the School of Electrical Engineering and Computer Science and the Russ College of Engineering and Technology by David W. Matolak Professor of Electrical Engineering and Computer Science Dennis Irwin Dean, Russ College of Engineering and Technology

3 3 ABSTRACT SAMUDRA, MOUSMI, M.S., November 2010, Electrical Engineering Comparison of Indoor Ray Tracing and Measurement Results for 5 GHz Band Wireless Channel Director of Thesis: David W. Matolak The basic aim of this research is to compare the measured channel impulse response results with analytical results obtained with the help of Wireless Insite software (ray tracing tool) for 5 GHz wireless band indoor channel. Measurements were taken at four different locations for three positions at each location in one of the basement hallways of Stocker Centre, Ohio University. Power delay profile and root mean square delay spread were obtained from the measurements through data processing. The approximate model for the same environment was created in the software. Power delay profiles and root mean square delay spread was obtained through simulations which are referred as analytical results. The measured and analytical power delay profiles were compared. Also, a quantitative comparison of root mean square delay spread for measured and ray tracing values was carried out. In order to analyze the accuracy for ray tracing, a perturbation analysis was conducted. These included varying transmitter- receiver positions, dimensions of the objects, materials, etc. and assess the effects of this on root mean square delay spread values. The assessment showed that the measured and ray tracing results are comparable quantitatively. Also, the power delay profiles obtained from ray tracing were found comparable to the measured results in terms of the delay span. The perturbation analysis proved to be helpful in improving the inaccuracies involved during measurements and approximations of the model to some extent. Approved: Dr. David W. Matolak Professor of Electrical Engineering and Computer Science

4 4 ACKNOWLEDGEMENTS I owe my most sincere gratitude to GOD for giving me strength and courage to accomplish my thesis successfully. I would like to express my deep and sincere gratitude to my advisor, Dr. David Matolak, Professor and Graduate Chair, Electrical Engineering for his detailed and constructive comments, and for his invaluable support throughout this work without which this research would not have been possible. His wide knowledge, encouragement and personal guidance from time to time have been of great value for me. I am heartily thankful to him for his productive feedbacks and suggestions during the preparation of this thesis. I wish to express my sincere thanks to all my committee members Dr. Jeffrey Dill and Dr. Chris Bartone for their invaluable time and support for my thesis. I express my warm thanks to Indranil Sen for his help in the data processing software which was used to obtain the measured results. I am really grateful to Nidhin Davis for his untiring help and guidance in the Wireless Insite software. I would also like to thank my colleague, Prasada Reddy Kurri and other lab members for their essential assistance and help while working on this thesis. Finally, I owe my loving thanks to my mom, my family and my friends. Without their affection, encouragement and support, this thesis would not have been possible.

5 5 TABLE OF CONTENTS Abstract...3 Acknowledgments...4 List of Tables...7 List of Figures...8 List of Acronyms and Abbreviations...11 Chapter 1: Introduction Growth of Wireless Communications Importance of Channel Modeling Thesis Scope...18 Chapter 2: Background and Literature Review Multipath Propagation Types of Channel Models Tapped Delay Line Structure Channel Impulse Response and Root Mean Square Delay Spread Literature Review...33 Chapter 3: Description of Measurements Environment Description Channel Sounder Overview Measurement Procedure...45 Chapter 4: Ray Tracing Model Overview of Ray Tracing Wireless Insite Overview Environment Models...54 Chapter 5: Comparison of Results...60

6 6 5.1 Comparison of Measured and Wireless Insite PDPs Comparison of RMS-DS Multipath Thresholding Perturbation Analysis...92 Chapter 6: Summary, Conclusions and Future Work Summary Conclusions Future Work References Appendix A: Matlab Code...110

7 7 LIST OF TABLES Page Table Various units used for requested outputs Table Values for RMS-DS (ns) for three positions at location Table Values for RMS-DS (ns) for three positions at location Table Values for RMS-DS (ns) for different positions for all receiver locations Table Comparison of measured and analytical RMS-DS values (ns) Table Comparison of analytical and measured RMS-DS values (ns) after multipath thresholding Table RMS-DS values (ns) for perturbation analysis for receiver locations Table Perturbation analysis for transmitter position for location Table Perturbation analysis for transmitter-receiver position for location Table Perturbation Analysis for changing materials for location Table Perturbation analysis for change of material for additional wall Table Comparison of RMS-DS (ns) values Table Perturbation for wall, ceiling and floor material for all receiver locations Table Comparison of RMS-DS (ns) values for changing receiver positions for all locations

8 8 LIST OF FIGURES Page Figure Scattering of signal contributing to the overall channel response where each ellipse represents a constant length propagation path from transmitter to receiver Figure Tapped delay line model for linear time invariant dispersive channels Figure Tapped Delay Line Channel Model Figure Channel Impulse Response Figure Communication path with multipath reflectors Figure Diagrammatic representation of the locations for taking measurements in the indoor environment. Receiver locations denoted as Ri with i = 1, 2, 3 or Figure a) Stocker environment reading room hallway with transmitter position and first receiver location Figure b) Perpendicular hallway with second and third location of the receiver Figure c) Stocker basement hallway receiver location four Figure a) Ground Floor Plan for Stocker Figure b) Ground floor plan for Stocker with closed walls and outline for analytical environment model...42 Figure BVS channel sounder equipment Figure Transmitter and omni antenna Figure Receiver Set up Figure Project view in floor plan editor in Wireless Insite Figure Project view for Stocker ground floor plan created in Wireless Insite Figure Project view for the environment showing study area boundary Figure Top view of the project with one receiver at a time showing human absorbers... 58

9 9 Figure Top view of the project with example propagation paths Figure Chameleon s Software file conversion window Figure Comparison of measured and analytical PDPs for position 1 at location Figure Comparison of measured and analytical PDPs after noise thresholding for position 1 for location Figure Comparison of measured and analytical PDPs after noise thresholding for position 2 at location Figure Comparison of measured and analytical PDPs for position 3 at location Figure Top view of the project with example propagation paths for location Figure Comparison of measured and analytical PDPs for position 2 at location Figure Top view of the project with example propagation paths for location Figure Comparison of measured and analytical PDPs for position 1 at location Figure Comparison of measured and analytical PDPs after noise thresholding for position 1 at location 3.69 Figure Comparison of measured and analytical PDPs after noise thresholding for position 2 at location 3 70 Figure Comparison of measured and analytical PDPs after noise thresholding for position 3 at location 3.70 Figure Top view of the project with example propagation paths for location 3 71 Figure Comparison of measured and analytical PDPs after thresholding for position 2 at location Figure Top view of the project with example propagation paths for location 4.72 Figure Histogram of RMS-DS for position 1 at location Figure PDP plot for position 1 location 1 with RMS-DS = ns Figure PDP plot for position 1 location 1 with RMS-DS = ns... 75

10 10 Figure PDP plot for position 1 location 1 with RMS-DS = ns Figure Histogram of RMS-DS for position 2 at location Figure Histogram of RMS-DS for position 3 at location 1.77 Figure Histogram for RMS-DS for position 2 for location 2.79 Figure PDP plot for position 2 for location 2 for RMS-DS = ns...80 Figure PDP plot for position 2 for location 2 for RMS-DS = ns 80 Figure PDP plot for position 2 for location 2 for RMS-DS = ns..81 Figure Histogram for RMS-DS for position 1 at location Figure Histogram for RMS-DS for position 2 at location Figure Histogram for RMS-DS for position 3 at location Figure PDP plot for position 2 for location 3 for RMS-DS = ns...84 Figure PDP plot for position 2 for location 3 for RMS-DS = ns...85 Figure PDP plot for position 2 for location 3 for RMS-DS = ns...85 Figure Histogram for RMS-DS for position 2 at location Figure Comparison of analytical and measured PDPs for position 2 location 1 with multipath threshold...89 Figure Comparison of measured and analytical PDPs for position 2 location 2 with multipath threshold...90 Figure Comparison of measured and analytical PDPs for position 2 location 3 with multipath threshold...90 Figure Comparison of measured and analytical PDPs for position 2 location 4 with multipath threshold...91 Figure Top view of the project with simplest environment...93 Figure Three dimensional axis orientations for transmitter and receiver positions in ray tracing...95

11 11 LIST OF ACRONYMS AND ABBREVIATIONS 1G 2G 3G 4G PDP RMS-DS AMPS TACS FDMA TDMA CDMA FDD GSM PDC WAP GPRS HSCSD UMTS IMT ISI CIR MPC WSSUS First Generation Second Generation Third Generation Fourth Generation Power Delay Profile Root Mean Square Delay Spread Advanced Mobile Phone System Total Access Communication Systems Frequency Division Multiple Access Time Division Multiple Access Code Division Multiple Access Frequency Division Duplexing Global System for Mobile Pacific Digital Cellular Wireless Applications Protocol General Packet Radio Service High Speed Circuit Switched Data Universal Mobile Telecommunications Service International Mobile Telecommunications Inter Symbol Interference Channel Impulse Response Multipath Component Wide Sense Stationary Uncorrelated Scattering

12 12 LOS NLOS LTI PL FDTD UTD MED UPS Tx Rx db ns 3D NT WI Hz Line of Sight Non- Line of sight Linear Time Invariant Path Loss Finite Difference Time Domain Uniform Theory of Diffraction Mean Excess Delay Uninterruptable Power Supply Transmitter Receiver Decibels Nanoseconds Three Dimensional Noise Threshold Wireless Insite Hertz

13 13 Chapter 1 Introduction This chapter describes the evolution of wireless communications based on different generation technologies. Section 2 describes the importance of channel modeling and various parameters involved in channel modeling. Section 3 provides the scope of this research. 1.1 Growth of Wireless Communications Epoch of Cellular Concept In order to appreciate the impact of wireless technology in this modern world, it is helpful to have a prior knowledge of the history for the growth of wireless communications. Initially, wired telephones and telegraph methods were commonly used for communication. However, observing the high increase in use of these methods, it became necessary to explore new ways of communications that are more efficient and satisfy the ever-increasing demand for reliable and mobile communication. Also, a technology was certainly needed where messages could be sent over long distances between non-stationarity points, for example, air force services and naval forces, where wired telephony was not possible. In 1860, German physicist, Heinrich Rudolph Hertz drew attention to the fact that electric energy could be sent through space in the form of some kind of waves similar to light and heat. This resulted in the discovery of radio waves, a form of electromagnetic waves, which can transmit sound, pictures and even securely coded data invisibly through the air. It is said that radio waves owe their invention to wired telegraph and telephonic systems [1]. In 1866, Mahlon Loomis introduced the concept of wireless telegraphy by showing how a meter in connection with one kite was able to move another kite [1]. However, the beginning of wireless technology could be claimed in 1899 when there was a launch of first successful transatlantic radiotelegraph message transmission. Guglielmo Marconi succeeded to transmit a three dot Morse code message

14 14 for the letter S over a distance of three kilometers in 1899 [1]. This successful reception of code marked the first step towards the development of wireless communications. In 1901, radiotelegraph services were implemented between five Hawaiian Islands [2]. Afterwards, similar services were used to transfer information between armed and naval forces related to war plans and secret codes within the period of Historically, the growth of wireless communications has occurred slowly because even though the first mobile telephone was introduced in 1946, the ability to provide efficient commercial wireless services to an entire population was not achieved. In the period of , Bell Laboratories developed highly reliable solid state radio frequency hardware for mobile phones which led to the remarkable growth in cellular radio and personal communication systems all over the world [2]. One of the successful cell phone networks was ARP network in Finland in 1971, which is also referred to as a zeroth generation cellular network. Currently, wireless communications are experiencing the fastest growth in their history Different Generations of Cellular Systems The development of cellular technology can be divided into four periods. Initially, mobiles used a specific frequency band only in a particular geographic area. This resulted in severe problems of congestion and call incompletion because if one customer was making a call in a particular frequency band, another customer could not make a call on the same frequency band. Hence, the number of simultaneous calls was limited [2]. First generation (1G) cellular mobiles implemented a variety of different analogue techniques, and their use slowly became popular. Instead of allotting a band of frequencies for a particular area, the cellular idea was to break down a specific large area into smaller areas or cells. This enabled the various users to make a simultaneous call because each of the cells reused the same channels within the same area. So, this avoided call blockage and increased channel usage. In the 1980 s, 1G cellular networks were

15 15 implemented in various parts of the world including the United States with the Advanced Mobile Phone System (AMPS), United Kingdom with the Total Access Communication Systems (TACS), Germany with C-Netz, etc. These 1G cellular systems mostly relied on Frequency Division Multiple Access (FDMA) and analog frequency modulation [2]. Unlike 1G cellular standards, second generation (2G) cellular systems used digital modulation and Time Division Multiple Access (TDMA) and Code Division Multiple Access (CDMA) with frequency division duplexing FDD techniques. The most popular second generation standards include i) Global System for Mobile (GSM) Communications, deployed in Europe, Asia, Australia, South America and some parts of the USA, ii) Interim Standard 136 popularly used in North and South America and Australia, iii) Pacific Digital Cellular (PDC) used by the Japanese standards and iv) 2G CDMA or Interim Standard 95 CDMA (cdmaone) implemented by carriers in North America, Korea, Japan, China, South America and Australia [2]. With the introduction of 2G wireless networks, the spectrum efficiency increased by approximately a factor of three times in comparison to 1G technology. The 2G GSM network system also allowed users to send short messages to other subscribers in the same network. However, 2G standards supported the limited browsing of internet and the data rate used for those standards was too slow for sending quick s. So, in an attempt to increase data rates for 2G standards to provide faster internet applications, 2.5G technology was introduced. The 2.5G technologies improved upon 2G in terms of the higher data rate transmissions. The 2.5G standards also support the Wireless Applications Protocol (WAP), which enables viewing a number of internet browsers in a compressed format on portable devices like cellular phones. Some of the commonly implemented standards for 2G technology are general packet radio service (GPRS) and High Speed Circuit Switched Data (HSCSD). The most important factor in 2.5G cellular network deployment is ( backward ) compatibility with 2G technology standards, otherwise incompatibility would lead to expensive equipment changes at base station [2].

16 16 Aside from compatibility, as the growth in demand and application variety continued, 3G technology began. The 3G standards were designed with larger data rates and higher spectral efficiency than previous generations and in many ways was compatible with all the older systems. Most 3G technologies allow use of both voice and data transfer on a single user channel. As suggested in [2], 3G enabled unparalleled access to wireless networks with multi megabit access, usage of Voice Internet Protocol, voice activated calls and nearly omnipresent always-on capacity. Also, users of 3G network systems were able to receive live updates, interact on web sessions and have parallel reception of data as well as voice from various users at the same time on one single user network while moving. The global frequency band in the 2000 MHz range was selected for the operation of 3G technology which led to the evolution of CDMA 2000 based on the basics of IS-95 and IS-95B. The significant operating technology on 3G standard is Wide- Band CDMA commonly known as Universal Mobile Telecommunications Service (UMTS) which assures the backward compatibility with 2G GSM, IS-136 standard and 2.5G technologies. With the successful launch of 3G technology in the period from 1998 to 2005, further research aims were to provide an even wider range of data rates (in gigabit ranges) for faster internet access and improved security and privacy; this led to the 4G technology evolution. The most commonly cited standard for 4G is the International Mobile Telecommunications Advanced (IMT-Advanced). Different prospects are being explored for enhancing the performance of mobile communication systems. Adaptive multiple antenna systems play an important role in the 4G standards to improve data rates, reliability and coverage range of networks. Although 4G standards were introduced in 2002, they were only deployed in limited areas within the period of Recently, Europe has deployed 4G network standards from February 2010 built by TeliaSonera and Nokia Seimens which has broadened the Long Term Evolution (LTE) coverage.

17 Importance of Channel Modeling A wireless channel is considered as one of the most challenging elements for reliable communication. However, with increased research in the field of wireless technologies, it became possible to accurately characterize the unpredictable nature of the channel for enhancing the communications link. A channel can be viewed as a transmission medium through which the signal propagates from transmitter to receiver. When a wireless system is implemented, the signal travelling through the channel undergoes a generally unpredictable number of reflections, diffractions and scattering. So it becomes very difficult to predict the behavior of signal through wireless channel. Hence, channels are most often modeled statistically, and these models can be designed with a high degree of accuracy. Interference is another important aspect that should be considered in system design. Interference generally refers to undesired other signals from other transmitters. The different types of interference which may occur within the channel are co-channel interference, adjacent channel interference and inter symbol interference (ISI). Co-channel interference occurs when two radio transmitter signals using the same frequency band interfere and results in signal degradation. Channel modeling helps to establish models for received co-channel interference. Due to inadequate filtering and frequency offsets and Doppler shifts, adjacent channel interference effects are observed sometimes at the receiver. So, studying the behavior of the channel makes it easy to specify the modulation/detection type, multiplexing technique and filtering methods which need to be implemented in order to minimize this adjacent channel interference. Inter-symbol interference causes distortion of the signal and overlap of subsequent symbols; this can result in errors at the receiver. In order to minimize ISI, one must select appropriate signal bandwidth and signaling rates. Also, it may become essential to employ appropriate

18 18 equalization techniques, power control algorithms, diversity techniques and error control coding methods. Channel knowledge affects selection of these factors and plays an important role. The important aspect of multipath propagation and delay dispersion in channel modeling is discussed in the next chapter. 1.3 Thesis Scope The main objective of this research is to compare measured results with analytical results for 5.12 GHz band wireless channels. In order to obtain measured results, data was collected by conducting measurements in a selected indoor environment at a frequency of 5.12 GHz. The indoor environment selected was one of the basement hallways of Stocker Center, Ohio University, Athens, Ohio. The measurements were recorded at four different locations with three positions at each location. Both the transmitter and receiver were inside the building, with the transmitting unit stationary and the receiver moved to each location, and kept stationary during measurements. The recorded measured data was used to obtain power delay profiles and root mean square delay spreads. The analytical results were obtained from a simulation tool Wireless Insite, which employs a ray tracing method. The approximate model of the indoor environment (where measurements were taken) was created in the software and the transmission was simulated in order to obtain analytical values for power delay profiles and root mean square delay spread. The power delay profiles and root mean square delay spread were compared with those obtained by measurement. Note: in this thesis, as is commonly done in IEEE journal publications, the 2 nd person plural tense is often used; when the word we appears, this refers to (i) the author and her advisor when pertaining to overall aims of the research, (ii) the author and fellow students when pertaining to the measurements, and (iii) the author alone for all other uses.

19 19 In this research, we also aimed at improving the accuracy of ray tracing in order to obtain better agreement with the measured results. This was carried out by doing a perturbation analysis. This consists of varying, by small amounts, parameters in the ray tracing model. This includes varying building materials, transmitter-receiver positions, etc. The perturbation analysis should help in setting up some guidelines for the use of ray tracing in order to reduce computation and maintain accuracy.

20 20 Chapter 2 Background and Literature Review This chapter consists of five sections. Section 1 discusses the importance of channel modeling specifically considering multipath propagation and delay dispersion. Section 2 introduces different types of models used for channel modeling and section 3 explains the structure of tapped delay line models. Section 4 describes channel impulse response statistics and different channel parameters. Section 5 is the literature review which summarizes related channel modeling work. 2.1 Multipath Propagation In wireless communications, the transmitted signal often arrives at the receiver through different propagation paths due to reflections, diffractions and scattering. These occur due to the interactions of the signal with objects present in between the transmitter and receiver. This arrival of the signal at the receiver via multiple paths is named multipath propagation. The number of these propagation paths may be large and unpredictable. Multipath propagation can also cause fading. We may experience fading in our typical cell phones use - it may happen sometimes that if a person talking on a cell phone is moving, the signal gets weak and reception is impaired. Moving some steps forward or back, the quality of the signal often improves. This is because the multipath components combine either constructively or destructively, and this phenomenon varies on the scale of one half wavelength. This effect is referred to as small scale fading. Due to multipath propagation, it is very challenging to predict the behavior of a signal travelling through a wireless channel. Hence research interest in channel modeling has been strong for several decades.

21 Delay Dispersion While travelling through the wireless channel, the signal may undergo delay dispersion (time dispersion). Due to the presence of multipath components, the signal arrives at the receiver at several time instants, and this constitutes dispersion in time. This is called delay dispersion of the signal. The effect of delay dispersion can be quantified via several channel parameters. The delay dispersion is typically measured by power delay profiles (PDPs). The most commonly used delay dispersion parameter is multipath delay spread. 2.2 Types of channel models As noted, a transmitted signal exhibits a complex behavior while travelling through a wireless channel due to multiple reflections, diffractions and scattering. Attempts to quantify channel effects upon signals in many different types of wireless channels resulted in a number of different channel modeling techniques. These channel modeling techniques basically approximate the complexities involved and produce ideally simple and reliable models of the respective channel. Two main approaches for channel modeling are statistical and deterministic Statistical Models A statistical channel model is one in which the channel impulse response (CIR) and its components are approximated as stochastic processes. Several statistics of these random processes are used to specify the channel behavior. Since the wireless channel is for most practical cases of interest well modeled as linear, time-varying filter, an equivalent model can be constructed using the time-varying channel transfer function, Fourier transform of the timevarying CIR.

22 These models are simplified models that aim to represent the main features of the real 22 propagation channel. Designing of a statistical channel model benefits from underlying knowledge of wave propagation principles and Maxwell s equations, but these electromagnetic theory tools are typically not directly used in model development. There are certain advantages of designing statistical models for wireless channels. As stochastic processes are implied for analyzing statistical models, it becomes easy to create variability within the same environment by changing values for channel parameters, for example changing the process variance. In the case of commonly used complex Gaussian stochastic processes, model behavior is completely specified by only few parameters (mean, variance, and autocorrelation for wide-sense stationarity). In addition, although the channel parameter values may be specific for a given environment, the methodology and design of the model is general, hence it can be easily implemented for other environments Deterministic Models Deterministic models are those that aim to describe the channel using equations exactly. That is Maxwell s equations or approximations are used to provide exact values of electromagnetic field quantities at arbitrary points in space distant from a specified transmitter, in a specified environment. One popular simplification is the high-frequency approximation that is embodied in ray tracing. Current deterministic models are often designed with the help of ray tracing simulation tools to achieve highly accurate propagation characteristics for a particular environment [20]. In order to characterize channels deterministically, it is very important to know the details of the environment geometry and accurate dimensions of all objects in the particular physical environment under study. In other words, deterministic channel modeling deals with a direct electromagnetic and physical characterization of the given channel. In addition to the geometry and dimensions, a database of the various parameters for that particular environment is

23 23 also needed. The various parameters include electrical properties (conductivity, permittivity, permeability) for any materials in the environment, e.g., for walls, doors, ceiling, floor, etc. This suggests that the accuracy of these models strongly depends on the accuracy with which the object properties and dimensions are represented. These models can offer great accuracy if their inputs are accurate. A disadvantage of deterministic models is that they are computationally intensive and computation time is directly proportional to the accuracy in modeling the propagation channel. For indoor channel modeling, a very detailed database may be required for accuracy, and this may be impractical. Also, the number of multipath components may be comparatively less than present in the actual environments since ray tracing can represent only specular components, and not diffuse (scattered components). 2.3 Tapped Delay Line Structure It is well established that when the multipath components are discrete, and/or the communications receiver employs a sampling operation, then a wireless channel impulse response can be modeled by a tapped delay line structure. As discussed in a previous section, in a wireless channel, many obstacles are often present between the transmitter and receiver. Due to this, the transmitted signal reaches the receiver through different propagation paths owing to reflections, diffractions and scattering. This multipath propagation introduces delay in time required to reach the receiver, and the delay incurred by the signal to travel from transmitter to receiver depends on its path length. As suggested by [6], such a channel can be described by sets of paths of equal lengths. In the case of single reflections, this results in elliptical loci describing reflection points at a given value of delay, with the transmitter and receiver at the elliptical foci as shown in Figure

24 24 Figure Scattering of signal contributing to the overall channel response where each ellipse represents a constant length propagation path from transmitter to receiver [6]. If the bandwidth of this channel (signal) is W, then the signals arriving at time (delay) instants separated by a minimum of Δτ = 1/W seconds, can be differentiated. The channel impulse response can be seen as the summation of all components from each scattering or reflection. As the size of ellipse in Figure increases, the reflected energy takes longer to reach the receiver. However, the above description is simplified; the energy of components reflected two or more times is not considered. This simplification is common in most ray tracing models. The single bounce scattering principle assumes that the energy carried by components that undergo multiple reflections is comparatively weaker than that carried by the singly reflected components, and can hence be neglected. The tapped delay line model is shown in Figure

25 25 Figure Tapped delay line model for linear time invariant dispersive channels [6] The model in Figure is a discrete tapped delay line model in which the multipath components (MPC s) are represented at discrete intervals separated in delay by the sampling (often, symbol) time T s. The boxes denote delays and the MPCs at a given value of delay are actually the aggregate of those received within a delay bin (of width 1/W), i.e., the MPCs h[i] are the summation of the MPCs over the specified period of the bin. If the channel input is x(i) and the output is y(i) then the output of the time invariant channel is given by y i = D 1 d=0 h d x i d + n i (2.1) where h[d] is the channel response (MPC) at delay d and n[i] is a sample of complex-gaussian noise. Since h[d] is as voltage (or E-field) quantity, [h(d)]² is proportional to power. Plots of [h(d)]² vs delay are termed power delay profiles PDPs. Currently, many standard models for tapped delay lines are available for particular environments and bandwidths. Reference [7] provides an example of the tapped delay line model for Wide Sense Stationary Uncorrelated Scattering (WSSUS) channels as well as Wideband Channels. The CIR can be defined as follows:

26 26 N h t, τ = i=1 c i t δ τ τ i (2.2) where N is the number of taps (MPC s), the c i (t) are time dependent complex coefficients for the taps and τ i is the delay of the i th tap. In wideband models, the most widely employed model is the N-tap Rayleigh fading model where the amplitudes of all the taps are assumed to have Rayleigh statistics. Generalizing from this to allow, in addition, a dominant, or non-fading (often LOS) component, the model becomes the Ricean model, in which the impulse response is then h t, τ = a 0 δ τ τ 0 + N i=1 c i t δ τ τ i (2.3) where a 0 is amplitude of the LOS component, and the c i (t) are zero-mean complex Gaussian random processes. The author of [8] has suggested a correlated tapped delay line model for band limited multipath channels for analyzing direct sequence code division multiple access (CDMA) transmission technique. The author assumed a two-path Rayleigh fading propagation channel and the CDMA receiver implements Rake combining. The author of [17] employs a tapped delay line for a statistical channel model where time variant MPCs are considered. The model is identical to the time-invariant case, except the MPCs are random processes, as in Figure 2.3.3

27 27 Figure Tapped Delay Line Channel Model [17] There are in general several ways to compute the number of channel taps based upon knowledge of the measured CIRs. As in common, the author of [17] calculates the number of taps (L) based upon the mean value of RMS Delay Spread (RMS-DS) as shown in Equation 2.4. where x denotes the greatest integer less than or equal to x and T c is the signal symbol duration. Note that, the number of taps depends upon the signal bandwidth, and L expresses the duration of the CIR in units of T c. Generally, it takes more taps to represent non LOS (NLOS) channels than LOS channels. (2.4) 2.4 Channel Impulse Response (CIR) & Root Mean Square Delay Spread (RMS-DS) Introduction As noted, due to the seeming random nature of channel, statistical channel modeling is often preferred. When a signal reaches the receiver through multiple paths owing to reflections, diffractions and scattering, this may yield distortion in a signal at the receiver. This introduces

28 28 small scale variations in the transmitted signal. The CIR captures these small scale variations in the MPCs. Since the channel is linear, the CIR provides all information essential for analyzing any type of signal transmission through the channel Channel Impulse Response (CIR) The channel impulse response (CIR) is the output of the channel when the input is an impulse applied at time instant t. For a real channel, the impulse response must be causal and hence, it is zero for t < 0. If one has a good model for the CIR, one can use this information to counteract any channel induced distortions. For example, as suggested in reference [9], knowledge of the CIR can be used to design filter to eliminate the multipath distortions. These filters are called equalizers, and they can be adaptive for use in linear time varying channels. For any linear channel with impulse response h(t), the receiver output is calculated by convolving the impulse response with the channel input. Figure Channel Impulse Response

29 29 Figure shows the LTI channel with impulse response h(t). As shown, if the input signal is Dirac Delta function δ(t) then the receiver output r(t) will be the input signal convolved with h(t), which is identically h(t). r t = δ t h t (2.5) Channel Propagation Parameters Increasing interest in research on wireless channels led to the conduction of a number of radio propagation experiments in different urban-suburban areas and indoor environments. The aim of these experiments was to quantify channel behavior, in order to be able to design more reliable transmission and reception techniques. Although these studies provide results that can be used for determination of suitable data rates, signal bandwidths etc., research is still going on for characterizing channels accurately in multiple environments. Experiments have been conducted to obtain information on multiple characteristics of wireless channels. Two of these characteristics are described next. 1. Path loss: Path loss (PL) is a measure of average RF attenuation undergone by the transmitted signal in reaching the receiver [6]. Analytically, path loss in db is expressed as PL db = 10 log P t P r (2.6) where P t is the transmitted power and P r is the received power in linear units (e.g., watts). Path loss is useful in estimation of signal coverage area for wireless systems. There are two main types of path loss models, empirical path loss models and site-specific path loss models [10]. The path loss or signal attenuation is described mainly as a function of distance.

30 30 However, in some models, path loss depends not only on distance, but also on other parameters like measurement environment, building height, antenna heights, terrain profile, etc [8]. One of the most commonly used empirical models is the Okumura Hata model. This model considers the carrier frequency and antenna height for calculating path loss. The path loss equation for this model is given by (2.8). PL db = A + B log d + C (2.8) where A, B and C are dependent on frequency and antenna height. Factor A is directly proportional to carrier frequency and inversely proportional to the antenna height. Factor B is dependent on the path loss exponent, which decreases with the increasing height of the base station. This model is simple to apply and is useful for large cells (base station placed at height above surrounding rooftops) [8]. Ray tracing technique and finite difference time domain (FDTD) method are site-specific path loss models which are commonly used to estimate path loss analytically. Ray tracing uses geometrical optics and uniform geometrical theory of diffraction (UTD) for estimation. For more complex structures, FDTD method is also used which provides numerical solution with the use of Maxwell equations. The main advantage of these models is their accuracy. However, a hybrid technique employing ray tracing and FDTD method together is also suggested for achieving more accurate path loss estimation in radio-wave propagation [10]. 2. Delay Spread: The spread of the CIR in delay is measured by the delay spread. Often, the RMS Delay Spread (RMS-DS) is calculated as a single representative measure of this spread. The RMS-DS is defined as the square root of second central moment of power delay profile (PDP) [10]. The PDP is the distribution of the received power versus delay when an impulse

31 is input to the channel. Thus, the PDP can be considered as the CIR, with the MPC amplitudes squared RMS Delay Spread Figure Communication path with multipath reflectors To begin with, consider the example of the several simple communication paths shown in Figure The transmitter (Tx) and receiver Rx are connected by four paths, and there are two reflectors denoted A and B in the area. The shortest path is path 1, the direct path. This path typically contains the maximum energy. The signal also travels through the additional paths 2, 3 and 4. Obviously, these paths will take more time to reach the receiver as they travel a longer distance than the direct path. Thus the delays of the transmitted signal are spread about a range of values. This spread is called the delay spread of the signal.

32 32 The time dispersion of the signal is an important factor in channel modeling and characterization. There are several time dispersion parameters that quantify the amount of delay spread, including excess delay spread (X db), root mean square delay spread (RMS-DS), delay window, and delay interval. Of these, RMS-DS is the most widely used parameter Calculation of RMS-DS The PDP gives the distribution of the transmitted power versus delay over the various paths in a multipath model of the channel. If h(t) represents the impulse response of the channel, then the corresponding PDP, P(τ), can be expressed as P τ h(τ) 2 (2.9) Typically, the PDP is normalized so that the received power P = h(τ) 2 dτ = 1. The mean excess delay (MED) is the first moment of the power delay profile given as L 1 2 k k k 0 L 1 2 k k 0 (2.10) where α k is the amplitude of the k th CIR impulse at delay τ k. This presumes a discrete PDP: a corresponding integral form for (2.10) (and (2.11)) would pertain for a continuous PDP. Then, the RMS-DS is the square root of the second central moment of the power delay profile: L k k k 0 2 L 1 2 k k 0 (2.11)

33 33 The RMS-DS and MED can be obtained from a single PDP, but are usually computed using a spatially averaged profile, where the average is taken over an area for which large scale fading variation is negligible [2]. In mathematical statistics, since these measures pertain to a specific set of data, MED and RMS-DS are equivalent to the sample mean and sample standard deviation for grouped data. The size of the averaging area is not absolute, but it is often taken as wavelengths in cluttered NLOS areas. Values of RMS-DS typically range in the order of a few tons of nanoseconds for indoor channels and can reach several tens of microseconds for outdoor channels [2]. 2.5 Literature Review In order to begin research in channel modeling for wireless communications, it is very important to review previous research carried out in the field, and determine how the results and conclusions drawn from those published articles can be helpful in our research. The prime focus of this review is on research carried out on implementing ray tracing techniques and comparison of simulated results with measured results, in indoor as well as outdoor propagation environments. Reference [10] provides a review of various propagation models for path loss, delay spread and fading, in mobile communication. Various path loss models are discussed for both outdoor and indoor environments. Statistical path loss models for indoor cases were provided and compared for different buildings including grocery store, retail stores, factories, offices etc. The authors also described some deterministic models for path loss using ray tracing techniques, the image method and the Finite Difference Time Domain (FDTD) method. All the propagation models discussed in [10] serve as useful guidelines for channel modeling. The article also discusses research on impulse response measurements in both outdoor and indoor environments, with a tabular comparison of the RMS-DS values obtained for each

34 34 environment for different channel sounding techniques. The RMS-DS values ranged from 7-16 nanoseconds for measurements within a room, nanoseconds for office buildings and nanoseconds for an indoor sports arena. These values served as a guideline when we took measurements in an indoor environment for our research. The author of [11] primarily discussed modeling vehicle to vehicle (V2V) wireless channels and the unfavorable channel fading effects observed in such environments. The article centers its focus on describing existing V2V models (statistical and site-specific), citing values for some channel parameters. It also provides tapped delay line models based upon measurements. These measurements were taken for different conditions including urban areas with antennas inside & outside vehicles, and in open areas and small cities. Measurements were also taken with the varying traffic densities. The models provided pertain to bandwidths of 5 MHz, 10 MHz, and 20 MHz. Reference [12] describes the ray tracing method used for indoor channel modeling for a channel with center frequency in the 60 GHz band. The paper compares analytical and measured results for the selected indoor environment. The experiment was conducted in an empty room with a network analyzer. The transmitter antenna was at height 2.24 m in one fixed position and the measurements were taken at 18 different receiver locations (at height 1.12 m). The root mean square delay spread (RMS-DS) values for measured and analytical methods ranged from 10 to 60 nanoseconds and were found to be in good agreement. Reference [11] also shows two ray tracing models (room and hallway) in order to study the effect of wall configuration on the channel parameters. It was observed that delay spread ranges from to nanoseconds for the room and from ns to nanoseconds for the hallway. The authors of [13] carried out channel characterization for 5 GHz band wireless channels for domestic and office environments. However, reference [13] implements only ray tracing. The simulation was carried out for different floor plans of several residential areas with varying

35 35 number of households. The office area selected was comparatively wider than the residential area and consisted of 56 rooms connected with a long corridor. Simulations were carried out for the transmitter located at different places in the corridor and also in different rooms inside the office area. All the simulation results for different rooms/floors for both the buildings were compared. RMS-DS values were in the range of 4 to 10 nanoseconds for small rooms on both ground and top floor of residential areas. However, for the office scenario, most of the RMS-DS values ranged from 0 to 25 nanoseconds except for the conditions where the receiver was at a distance of more than 20 m from the transmitter, RMS-DS values for such conditions were found in the range of 50 to 100 nanoseconds and some above 100 nanoseconds. Reference [14] showed results for spatial variation of RMS-DS for measurement and analytical methods for indoor Line Of Sight (LOS) conditions at 5.2 GHz. Measurements were made in six business and residential areas. Complete room characteristics were also described in the paper for the analytical method (ray tracing technique). Measurement and modeling results showed that RMS-DS increases with an increase in the transmitter-receiver distance in LOS indoor environments, but does reaches a maximum value. The authors suggested that this maximum value of RMS-DS is dependent on the room dimensions and parameters of the walls. Reference [15] shows a comparison of measurement results and analytical results for 5 GHz band wireless channels for an outdoor environment. The author conducted the experiment in a parking lot of the Ohio University campus, Athens. The measurements were taken with only one car parked in the lot. Four sets of measurements were taken at four different locations with varied positions of transmitter and receiver. A ray tracing model for the experiment was created using the software Wireless Insite TM. Several approximations were employed, e.g. the car was modeled as a metal box, objects like sign posts were modeled as simple metal poles, etc. Measured and analytical PDPs were comparable qualitatively, but the measured profiles contained a larger number of multipath components than did the measured ones. RMS-DS values for measured and

36 36 analytical methods were found in reasonable agreement, e.g., for one position, the measured RMS-DS value was near 6 nanoseconds and the analytical value was 11 nanoseconds. The author also carried out a perturbation analysis in ray tracing to see if better agreement between measured and analytical results could be obtained. The perturbations included increasing complexity of the environment, varying the transmitter and receiver positions by small amounts, etc. It was seen that by doing this, the analytical RMS-DS value became closer to the measured RMS-DS value, for example, for the same position mentioned above, decreasing transmitter distance from the building decreased the analytical RMS-DS value. In reference [16], the authors attempted to assess the performance of an analytical approach by comparing its results with measured ones for the indoor radio channel over a 1-GHz bandwidth centered on 3.5 GHz. Three channel characteristics, path loss, Ricean K-factor and RMS-DS were used in the comparison of site-specific ray tracing analytical and measured results. The measurements were taken in the ORBIT laboratory of Rutgers University for 18 different transmitter-receiver paths. The channel response for the environment was measured with a vector network analyzer (VNA). Very good agreement was seen between the simulated and measured results. RMS-DS values obtained for 18 different paths ranged from 40 to 70 nanoseconds for the measured results and from 30 to 60 nanoseconds for the analytical results. The authors note that the complexity of ray tracing increases if diffraction is a primary propagation mechanism. Reference [17] provides a characterization of an outdoor wireless channel for the 5 GHz band. Measurements were carried out on the Ohio University campus with the transmitter on top of three different buildings and the receiver was moved slowly on a cart around the transmitter building area. The measured results included PDPs in both line of sight (LOS) and non line of sight (NLOS) regions. The author also provided statistical tapped delay line models for this environment, and found delay spread values in the range of 10 to 800 nanoseconds. This paper provides a review for characterization of the wireless channel in 5 GHz band via channel

37 37 sounding measurements, useful for our research since we used an identical channel sounding technique in our indoor measurements. In reality, an exact environment cannot be perfectly emulated in any deterministic channel modeling approach. For indoor channels, complexity is large because the propagating signal will be affected by a large number of objects, making it difficult to predict the behavior of the channel. This leads to certain approximations in ray tracing models in order to reduce their complexity while still obtaining good agreement with actual measured results. The authors of [18] focused on these problems in ray tracing for indoor channel modeling and have described various methods to improve the accuracy of ray tracing techniques. In order to improve accuracy for effects of scatterers and dielectric obstacles whose size approaches the wavelength of operation, the authors suggested a ray tracing approach with a relative phase method, and compared this with other methods like finite difference time domain (FDTD). A Hybrid of the FDTD and Kirchhoff methods was shown to be useful as a standard for the performance evaluation of the ray tracing technique. The authors also developed a new set of diffraction coefficients to minimize the inaccuracies involved in ray tracing. If there is large number of scatterers and dielectric obstacles present in the indoor environment selected, then these sets of coefficients helps to improve accuracy. Reference [19] is a useful reference for beginners in ray tracing; it is the user s manual for Wireless Insite software. This reference provides indoor and outdoor propagation tutorials with stepwise explanations for designing models and conducting ray tracing simulations. It also provides a list of materials and descriptions of various features available in the software. In addition to this manual, various textbooks proved to be very useful for conceptual understanding of different features and parameters involved in channel modeling. References [2] and [4] describe propagation modeling, cellular standards and modulation techniques. References [3] and [8] describe different channel models and wideband channel characterizations.

38 38 Chapter 3 Description of Measurements This chapter consists of three sections. Section 1 describes the indoor environment in which measurements were carried out. Section 2 provides an overview description of the measurement equipment ( channel sounder ) used in this research and section 3 describes the measurement procedure. 3.1 Environment Description In this research, we compared measured and analytical results for indoor channel characteristics, specifically delay spread. Hence the first step was to select an indoor environment (within the university) for taking measurements with the channel sounder. Accordingly, one of the basement hallways of Stocker Center, Ohio University, Athens was selected as a fairly simple environment. Both line of sight (LOS) and non-line of sight (NLOS) conditions for the indoor channel are considered for this research. The measurements were taken, with the transmitter placed near one of the entrances to the basement hallway (the reading room hallway in Stocker). In order to minimize time variation, we wanted no pedestrian motion between the transmitter and receiver when the channel was measured; hence the measurements were taken on a weekend during the night time when few people were in the hallway. The measurements were taken at four different receiver locations with three actual receiver antenna positions at each location. The transmitter position and different receiver locations are illustrated in Figure 3.1.

39 39 Figure Diagrammatic representation of the locations for taking measurements in the indoor environment. Receiver locations denoted as Ri with i = 1, 2, 3 or 4 In Figure 3.1.1, Tx represents the transmitter position and the Ri points represent different locations of the receiver. The first set of measurements was carried out with the receiver at a distance of 10 m from the transmitter. At this location, the signal power was measured at three different local positions which are separated by a distance of 2.4 ʎ where ʎ is the wavelength of the signal equal to 5.85 cm. The wavelength of the signal is calculated with the following equation: ʎ= c f (3.1) where c is the speed of light (3 x 10 8 m/s) and f is the frequency of the signal, equal to 5120 MHz. For the three local positions, the receiver cart was moved 14 cm from the central point at distance 10 m in two directions: one in direction away from the transmitter and one toward the transmitter.

40 40 The second set of measurements was carried out with the receiver at a distance of 25 m from the transmitter, using the same approach of measuring at three positions. Locations 3 and 4 represent completely NLOS conditions. The channel impulse response (CIR) was measured at three different positions for the same location in order to analyze the effect on the CIR when the receiver is moved by small distances. For measurements, the Raptor Software on the laptop was invoked. With the help of this software, the delay span was set to 1 microsecond. Our channel sounder functions as a stepped correlator. Once recording is started, the receiver correlates with the received signal, with shifted lag times, for the specified delay span. For our research, every set of measurements was recorded for 20 seconds at each position for all the locations. This recorded data is the correlated output at the receiver corresponding to the power collected over the specified delay span. In this research, the transmitter position was fixed for measurements at all the locations. Even though the receiver was moved to different locations, the receiver was kept stationary when actual measurements were recorded at each position. In order to keep the propagation environment as simple as possible, laboratory and classroom doors within the area were kept closed. Figures and show photographs of the different measurement locations and the ground floor plan for Stocker. Figure (b) shows the floor plan with closed doors for classrooms and laboratories and the highlighted boundary for approximated model created in Wireless Insite.

41 41 Figure (a) Stocker environment reading room hallway with transmitter position and first receiver location, and (b) perpendicular hallway with second and third location of the receiver Figure (c) Stocker basement hallway receiver location four

42 42 Figure a) Ground Floor Plan for Stocker Figure b) Ground floor plan showing closed doors and outline of analytical environment model

43 Channel Sounder Overview In this research, a channel sounder was used for taking the measurements. It was manufactured by Berkeley Varitronics Systems (BVS) and is a modified version of their Raptor channel sounder. Modifications include frequency band of operation (5 GHz instead of 2.4 GHz [17]), higher output power, selectable values of chip rate (50 Mcps or 25 Mcps), selectable center frequency (5 values), and higher transmitter output power. The channel sounder consists of a transmitter and a receiver. Each unit also requires a power supply. However, the transmitter and receiver are required to move in selected environments according to the desired measurement procedure. So, an uninterruptable power supply (UPS) was also used with the sounder, and batteries can power both units. Figure shows the channel sounder with transmitter (Tx) on the right and receiver (Rx) on the left. This picture was taken inside the Mobile Communications laboratory, School of EECS, Ohio University, Athens. The equipment was mounted on a cart that was used to move the equipment to the desired locations. As mentioned in section 3.1, for this research, once set up at its location, the transmitter was stationary and the receiver was moved to different locations with the help of cart and UPS. Figure BVS channel sounder equipment, receiver (left) and transmitter (right).

44 44 Figure Transmitter and omni antenna. The transmitter with the omnidirectional antenna is shown in Figure The center frequency for the measurements was set at 5120 MHz. The power level was set to 33 dbm and the chip rate of 50 Mcps was used. The receiver set up, with laptop computer and omni antenna, is shown in Figure

45 45 Figure Receiver set up 3.3 Measurement Procedure This section discusses the training of the channel sounder as well as the measurement procedure we followed in the indoor environment. In order to have accuracy in measurements, the channel sounder needs to be calibrated, referred to as training the channel sounder. The transmitter and receiver have rubidium oscillators, which offer short term stability for observation intervals less than about one day. Training of the channel sounder is done in order to synchronize the frequencies of these rubidium oscillators. The training time is directly related to the time span for reliable measurements, i.e. the longer the sounder is trained, the longer the duration is available for reliable measurements. The required time for our measurements was approximately 2 hours. Hence, the channel sounder training was done overnight and the measurements were taken on the next day in the evening.

46 46 For training the channel sounder, the transmitter was connected to its power supply and turned on for warming up; RF transmission was kept off at this time. A laptop, with the Raptor software installed, was connected to the receiver with its AC power supply connected. Once the laptop and Rx were turned on, an RF cable was connected from Tx to Rx through an attenuator with 40 db minimum attenuation. This ensures that the received power is less than 35 dbm, for protecting the Rx front end. The transmitter radio frequency was set to 5120 MHz and the output power set to 5 dbm for training. The transmitter chip rate was set to 50 Mcps. After this, the transmitter RF output power was turned on and the state of the sounder was observed on the Raptor software on the laptop. The software showed Raptor Stable and Raptor locked which ensures that it is ready to start training. The Rx was also set to a frequency of 5120 MHz and 50 Mcps chip rate (identical to that of the transmitter) through the Raptor software, then the training was initiated. Once the training was completed, the maximum of two sounder parameters, current count, and least count was noted. This maximum is the nominal measurement time, in seconds. Training was stopped via the software, and the Tx RF transmit power was turned off. For taking measurements, the RF cable from transmitter to receiver was disconnected, and omni-directional antennas were connected to both transmitter and receiver. The AC power supplies were unplugged, but both the transmitter and receiver were connected to the UPS so that their power supply was uninterrupted. The entire set up was then taken to the ground floor of Stocker to the reading room hallway. Once the selected indoor location was reached, the transmitter was again connected to the AC power supply. The receiver was then connected ( make before break ) to a battery supply. The receiver was moved to the four different locations (and three positions at each location) using a cart and the measurements were taken as described in section 1 of this chapter.

47 47 Chapter 4 Ray Tracing Model This chapter consists of three sections. The first section gives an overview of ray tracing discussing its uses, limits etc. Section 2 provides an overview of the Wireless Insite software, describing the features available in the software, and section 3 describes the environment models created for our research. 4.1 Overview of Ray Tracing Rapidly growing interests for analyzing radio signal coverage with the help of computational methods has resulted in designing of various site-specific propagation models (deterministic channel models). Most of the deterministic models use the ray tracing method for studying radio signals in indoor and outdoor environments. The ray tracing technique uses the principle of Geometric Optics (GO) for estimating the propagation of electromagnetic rays through a channel. According to the GO principle, the energy of a signal is assumed to be radiated via rays and these rays are assumed to travel in straight lines, normal to the surface of equal signal power [10]. Ray tracing is thus known as a high frequency approximation. However, in many real environments, there are numerous obstacles present in the communication channel, so most of the rays travel from source to destination via multiple paths due to reflections, diffractions and scattering. The ray tracing technique approximates this interaction of electromagnetic rays and calculates the progress of radio signals through the wireless channel. It can account for both reflection and diffraction, but cannot model scattering (scattering violates the high frequency assumption). Both two dimensional and three dimensional ray tracing models are being widely used in wireless propagation studies. The original ray tracing method using the GO technique only captures the effect of direct or reflected rays and does not consider diffracted rays. However, for non line-of-sight (NLOS)

48 48 paths, the signal often reaches the receiver through diffractions (bending around the edge of an obstacle to reach the receiver). So, the concept of diffraction was applied to ray tracing with the use of the geometrical theory of diffraction (GTD). Although this improved the accuracy of the ray tracing model, adding diffraction increases model complexity. Increasing interests in ray tracing led to simplifications for the GTD approach. One attempt was applying wedge diffraction in GTD theory. In this approach, instead of using a general shape, the diffracting object was considered as a wedge (an object triangular in cross section) which simplified the numerical computations of GTD to some extent. However, the most widely used method is the Fresnel knife edge diffraction model which simplifies the diffraction model [3]. Diffraction parameters like roughness of the surface, conductivity etc. are not considered in this model. In real environments, the transmitted signal may reach the receiver through only diffracted rays or through both reflections and diffraction. These models designed for multiple reflections and diffractions may have a large number of multipath components, which will slow the simulation and potentially reduce accuracy. Hence, while dealing with ray tracing, it is always attempted to approximate the actual environment in a simpler way by using the smallest number of objects or obstacles, while still preserving the main propagation features. Ray tracing approximates the effect of reflection using simple equations. Hence, ray tracing requires less computation than other methods based on Maxwell s equations. The accuracy of approximation in ray tracing depends on the ratio of wavelength to dimensions of the scatterers and the spatial volume of interest [21]. If the receiver is positioned in a way such that it is at a distance of many wavelengths from the closest scatterer and scatterers are of dimensions larger than the wavelength, ray tracing will give fairly accurate results. The accuracy can be increased further by comparing the results of ray tracing with experimental data obtained from measurements taken in the specific environments. Ray tracing accuracy also depends on the

49 49 availability of databases for environment geometry and building dimensions, exact transmitterreceiver positions, etc. Although ray tracing techniques are being widely used these days, there are certain limitations for implementing this method in radio propagation prediction. A complex environment refers to large number of reflectors present, which may result in a large number of multipath components. If the number of multipath components is very large, then the complexity of the ray tracing program is large [3]. As the complexities in the environment increases, the accuracy for ray tracing typically degrade. Apart from root mean square delay spread (RMS-DS), other propagation parameters are not always well predicted by ray tracing [3]. It is impossible to develop models using ray tracing which will exactly match the actual environment. There will be certain inaccuracies involved in describing environment features such as materials properties, sizes and shapes of the objects, relative locations of transmitter and receiver, etc. Also, designing a model with diffracted rays is time consuming because a single ray bending around the object generates a new family of rays [10]. So, we have to limit the number of diffractions to be considered for practical reasons. In this research, we have selected the number of diffractions as one in order to comparatively reduce the time required for simulation. There are various computer software packages based on ray tracing, including Wireless Insite, Wireless Systems Engineering Software, and Wireless Valleys Site Planner etc. In our research, we have used Wireless Insite to design a model for the indoor environment using the ray tracing method. 4.2 Wireless Insite Overview Wireless Insite is a software package used for site-specific analysis and design of radio propagation models in wireless communication systems. It is basically an electromagnetic modeling tool used to predict the effects of buildings and terrain on the propagation of

50 50 electromagnetic waves [19]. It helps in predicting propagation and communication channel characteristics in indoor as well as outdoor environments, efficiently and accurately. We can model the physical characteristics of the indoor building features or outdoor terrain or city plan in Wireless Insite. The software routine then performs electromagnetic calculations for the designed model and estimates various signal propagation characteristics. Wireless Insite allows the user to either construct the approximate environment model with the help of an available database or import the available database files into the project window. The files to be imported should be in any one of several formats such as Drawing interchange format/drawing exchange format (DXF), shapefile, digital terrain elevation data (DTED) or USGS (US geological survey). Apart from databases for the model, transmitter and receiver locations can also be imported from a data file if available. If such data files are not available, the transmitter and receiver positions can be specified with the help of the software s tools. There are many advanced features available in the software for the ease of user manipulation. We will discuss some of them in this section. One of the main features for Wireless Insite is the study area. If the environment for the study is an urban area or an entire city, Wireless Insite enables the user to select a specific section for studying signal propagation within the model by defining the study area. Once the study area is specified, software calculates the signal strength (and other parameters) via ray tracing within the specified area, without considering objects outside the study area for calculations. For example, if a particular experiment is conducted with two or more transmitters and the study area is created in such a way that only one transmitter is within the study area, the effects of other transmitters on the signal are not considered. There are two ways in which study area can be created; one way is the manual approach in which the study area boundary can be drawn manually and the other way is the auto fit approach in which the study area boundary is automatically fit around some project feature. Wireless insite considers reflections from the ground/surfaces, diffractions and even transmission through a wall [19].

51 51 Different study area with varying properties can be created; this can be useful for comparing results. Editing the study area properties enables the user to manually set the number of reflections and diffractions for simulations. The larger the number of diffractions specified, the longer the time required for simulation. The resulting signal characteristics are evaluated by calculating the effect of each interaction of the rays along the path from transmitter to receiver. This evaluation can predict several channel characteristics, namely path loss, power delay profile, delay spread and impulse response. Wireless Insite also provides a visual representation of transmitter coverage areas, power distributions, etc [19]. The floor plan editor tool helps in building the environment by creating interior floor plans, windows, doors, ceilings for indoor channels. Similarly, for outdoor channels, this editor tool enables the user to build city plans, urban areas, parking lots etc. depending on the environment under study. Once the floor plan is drawn or edited, the software allows the user to have a complete project view two dimensionally or three dimensionally. There are many features available for this project view, including solid or wired body view, adding legend for distinguishing materials, the ability to rotate the project view in all directions, etc. Rotating the project view enables the user to see propagation paths for rays from transmitter to receiver, easily and efficiently. This software also allows the user to create objects of various desired shapes and sizes in the features tool. There is a database of materials provided in the software, the properties of which contribute to determining the electromagnetic properties. The reflection and transmission coefficients are determined through material properties and the diffraction coefficients from these determined coefficients. The properties of each material such as thickness, permittivity, conductivity, reflection coefficient and transmission coefficient are built into the Insite software and can be changed according to the application [19]. Various types of materials used for creating buildings and objects are provided, for example, concrete, metal, brick, perfect absorber, wood, glass, layered drywall etc.

52 52 There are also different sets of transmitters and receivers provided in Wireless Insite. The simplest set is with points where each position of transmitter and receiver can be specified independently. When transmitter or receiver points are to be placed along streets in an urban environment, a route set can be used which defines the path of the transmitter/receiver with evenly spaced points along the connecting lines of the route specified. However, if the receiver or transmitter is in motion, trajectory sets can be used. When a large area is to be considered with evenly spaced points, an X-Y grid enables the user to select the complete area to be filled with equidistant points. Wireless Insite allows the user to specify distance in meters between the two evenly spaced points while defining the location for receiver. Various shape types are also present allowing the user to define the multiple points of transmitter/receiver locations in an arc (horizontal and vertical), cylinder, sphere and polygon. Other types of sets available are vertical surface and points-on-face. We have used a set of points for specifying the transmitter- receiver locations in this research. For all the types of sets, the software enables the user to define height, xyz directions, and antenna patterns for transmitter as well as receiver. Wireless Insite offers the user different antenna patterns which may be specified for transmitter-receiver sets [19]. Various antenna patterns available in the database are horn, half wave dipole, omnidirectional, short monopole, linear dipole, linear monopole, quarter-wave monopole, square loop, circular loop, rectangular aperture, etc. This software also has a provision to calculate the antenna gain, and transmission line loss, and analyze the waveform and plot the specified antenna pattern. For user convenience, Wireless Insite computes various types of outputs. The required output may vary for each user s needs and thus, before starting the simulation for the created project, the user can select the desired outputs from a list. The list has a number of output types: receiver power, received power in free space with and without antenna patterns, path loss, path gain, free space path loss, excess path loss, propagation paths, time of arrival, mean time of arrival, delay

53 53 spread, electric field magnitude, electric field phase, direction of arrival, direction of departure, total receiver power, impulse response, power delay profile, electric field vs. frequency, electric field vs. time, Doppler shift and radiated power. Default units for some outputs are listed in Table [19]. Table Various units used for requested outputs ([19]) Output Received Power Time Frequency Length Units dbm Seconds Hertz (Hz) Meters Phase Degrees (-180 to 180) Direction Degrees (0 to 360) Electric Field Path Loss V/m db However, for some of the requested outputs, Wireless Insite allows the user to change units. Once the simulation is over, the project hierarchy feature in the software enables the user to view the requested outputs. All the propagation paths are listed in a window and each path reaching the receiver can be observed in the project view by just clicking on the list, one path at a time. All the paths can also be viewed together. All the requested outputs are automatically loaded into the project hierarchy window. For our research, we had propagation paths, power delay profile and delay spread as requested outputs after simulation.

54 54 Wireless Insite offers several different ray based propagation models, namely Urban Canyon, Fast 3D Urban, Full 3D, Free Space and Vertical Plane Model. These models use the ray tracing algorithm combined with UTD [19]. The Urban Canyon model is used for urban environments with tall buildings/sky scrapers where the height of transmitter and receiver are comparatively small. In order to consider interactions of rays with buildings, heights for buildings are assumed infinite and diffractions over the rooftops are assumed negligible compared to diffracted rays between different buildings. The Fast 3D urban model is for conditions where three dimensional (3D) studies are required for urban environments with low height buildings or hilly areas [19]. The Full 3D model is the only ray-based propagation model available in software for indoor environments and this model has no restrictions on object shape. A Free space model is desirable when the objective is to study the effects of transmitting antenna patterns. Features like blockages due to buildings, floors, objects, etc. are not considered. The Vertical plane model is similar to the Urban Canyon model with 2D geometry instead of 3D. This model is preferred over very high frequency or ultra high frequency regions with irregular terrains. Apart from ray based propagation models, Wireless Insite also provides urban canyon FDTD model, moving window FDTD model, HATA model and Wireless Insite real time module. For a beginner, the software also provides project tutorials. These contain complete designed projects for outdoor as well as indoor propagation predictions. The software also has some simple irregular terrain database files and propagation models in order to help the user understand importing of the files in the software. 4.3 Environment Models In our research, we are comparing analytical results with measured results, with the analytical results being those obtained via ray tracing method Wireless Insite. We have attempted to create

55 55 an approximate model of the indoor environment (where the actual measurements were carried out) using the floor plan editor in Wireless Insite. Figure shows the floor plan of the model viewed in floor plan editor. However, the transmitter T and receiver Rx (x=1, 2, 3, 4) are marked externally because the floor plan editor does not show the transmitter and receiver positions. Figure Project view in floor plan editor in Wireless Insite We have already shown the complete ground floor plan for Stocker in Figure in Chapter 3. However, practically, a signal travelling through the hallway beyond location 4 either reaches the receiver with a large number of multipath components or never reaches the receiver. So, as a part of our approximation, we closed the hallway after location 4 by creating a wall instead of doorway leading towards the elevator (symbolized as E1 in Figure 3.3). Figure shows the

56 approximated Stocker ground floor location where the measurements were taken. Tx1 represents the transmitter position and Rx (1, 2, 3 and 4) represents the receiver locations (position 1). 56 Figure Project view for Stocker ground floor plan created in Wireless Insite We have shown all four locations of the receiver in the project view but in actual simulation we have keep only one receiver active because there was only one receiver at the time of measurements. The wall, ceiling and floor material was approximated as concrete and the door material as wood. The grid spacing was kept at 0.5 m so the dimensions of the hallway were approximated to the nearest integer multiple of 0.5 m for creating a wall or door, for example, if the length of a wall was 2.6 meters, then it was approximated as 2.5 meters. The water fountain was approximated as a metal box. The dimensions of the hallway and objects like doors, fountains, etc., were measured manually via measuring tape. Ceilings are present but set to invisible for viewing the model clearly and analyzing propagation paths easily.

57 57 In order to limit the analytical simulation time to a reasonable value (e.g., under one hour), the model study area must be limited in complexity. Thus the ray tracing model is of necessity a simplified version of the actual environment. Throughout this work, the model complexity was increased in several ways in order to improve agreement between analysis and measurements (see Chapter 5), while still keeping the simulation time moderate. For the most accurate simulation results, all objects in the environment that are of size comparable to a half-wavelength ʎ/2 (ʎ 6 cm) or larger should be included. Out of nine propagation models available in the software as described in section 4.2, the Full 3D propagation model was used for simulation as it was the only propagation model that can be used for indoor environments. Transmitter and receiver locations were specified with points and antenna patterns were set to omnidirectional. The number of diffractions was specified as 1 in order to obtain simulated results without extensive simulation times. The study area was created manually along the boundary of the project because the auto fit feature would assume the complete model to be fit in a cube which would not be accurate for our case. The white line surrounding the entire environment in Figure is the study area created manually. Figure Project view for the environment showing study area boundary

58 58 One more step in approximation was creating two humans standing near the receiver (me the author and a colleague) while taking measurements. Approximate human bodies were inserted as block shaped objects composed of perfect absorber material. Figure shows the top view of the model showing only one receiver location with the human absorbers. Figure Top view of the project with one receiver at a time showing human absorbers As mentioned in section 4.2, the requested outputs for our research were power delay profile, delay spread and propagation paths. After simulation, Wireless Insite showed twenty five propagation paths by default. Figure shows some of the example propagation paths visible from the top view of the project.

59 59 Figure Top view of the project with example propagation paths The different colored lines in Figure represent varying propagation path power. The color coding for rays is set as starting from dbm (dark blue color) to dbm (dark red color). Hence every propagation path color is shown according to its power. The power delay profile results obtained from Wireless Insite were plotted by importing those files into MATLAB. Delay spread values for the profiles were directly obtained from the Wireless Insite for comparison.

60 60 Chapter 5 Comparison of Results This chapter discusses the results obtained from the comparison of measurements and ray tracing analysis. Section 1 deals with the analysis of measured PDPs and provides a discussion of comparison of measured PDPs with analytical ones. Section 2 discusses the statistics of measured RMS-DS. Section 3 discusses multipath thresholding, and section 4 describes the perturbation analysis carried out to evaluate the ray tracing accuracy. 5.1 Comparison of Measured and Wireless Insite PDPs The data files recorded from the channel sounder are saved on the laptop computer in.rap format. This file format cannot be directly read in MATLAB. In order to use the data in this file in MATLAB, the manufacturer-provided Chameleon software is used to convert this log file into an ASCII file that is readable by MATLAB. After conversion, the ASCII files are saved in a.out format. The file conversion window for the Chameleon software is shown in Figure

61 61 Figure Chameleon s Software file conversion window The converted.out file is a matrix containing the values of power (dbm), phase (radians) and channel power (dbm) for various samples. The number of profiles obtained for the power matrix is different for different locations. A comma is used as a delimiter in order to separate the field values in the file. This output file is then imported into MATLAB for processing of the measured data. After loading the processed data file into MATLAB, each measurement set is separated into individual files for power record, phase record and received signal strength indicator (RSSI). The power records are the PDPs, which are the estimates of power contained in each multipath component. In our research, we obtained from 375 to 392 PDPs for the different locations. In order to circumvent this discrepancy and smooth out minor variations caused by equipment imperfections, when comparing measured PDPs with the analytical ones, we have averaged the