WALL COMPENSATION FOR HIGH RESOLUTION ULTRA-WIDEBAND OBSTRUCTED LOCALIZATION NURUDDEEN MOHAMMED IYA. A Thesis Presented to the

Transcription

1 ~ ~ :$ ~ ~ ~ ~ :1!i WALL COMPENSATION FOR HIGH RESOLUTION ULTRA-WIDEBAND OBSTRUCTED LOCALIZATION BY NURUDDEEN MOHAMMED IYA ~ ~ ~ -~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ 1E ~ A Thesis Pesented to the DEANSHIP OF GRADUATE STUDIES KING FAHD UNIVERSITY OF PETROLEUM & MINERALS DHAHRAN, SAUDI ARABIA In Patial Fulfillment of the Requiements fo the Degee of MASTER.OF SCIENCE In TELECOMMUNICATION ENGINEERING ~ ~ ~ Decembe 010 ~ ~ ~ff-w-w~~

2 KING FAHD UNIVERSITY OF PETROLEUM AND MINERALS DHAHRAN 3161, SAUDI ARABIA DEANSHIP OF GRADUATE STUDIES The thesis, witten by NURUDDEEN MOHAMMED IYA unde the diection of his thesis adviso and appoved by his thesis committee membes, has been pesented to and accepted by Dean of Gaduate Studies, in patial fulfillment of the equiements fo the degee of MASTER OF SCIENCE IN TELECOMMUNICATION ENGINEERING. D. ALI HUSSEIN MUQAIBEL (Adviso) c::: :so..j;,») -- ~ 3» D. SAMfR H. ABDUL-JAUW AD D. MOHAMED ADNAN LANDOLSI (Membe) f;i...-i-h LL, i V ~ (~ D. SHEIK SHARIF IQBAL (Membe) (Depatme~ D. SALAM A. ZUMMO (Dean of Gaduate Studies) Date ~\\\\,

3 DEDICATION All paises be to Allah, the Ceato and Lod of the Univese. O Allah, Have mecy and accept it as a dedication to you O Allah, Show it to me in the last day with the good deeds O Allah, Geat thanks to you fo this accomplishment. You bounties; my mothe, my fathe, a lovely wife, a geat family, and special fiends, I cannot deny, led to this accomplishment. iii

4 ACKNOWLEDGEMENT All paise is to Allah (SWT), Lod of the Heavens and the Eaths, and what they contain; Most Beneficial, Most Meciful. I seek His benediction on ou beloved Pophet, the best of mankind, Muhammad (SAW), his household and his companions, till the Day of Judgment. I thank Allah fo His infinite bounties on me, one of which is the completion of this wok; I seek His mecy and fogiveness, and indeed, I fea His wath. My gatitude, without esevation, goes to my paents fo thei unelented suppot and payes thoughout my life. I will foeve emain gateful. I would like to convey my sincee gatitude to my thesis adviso D. Ali Hussein Muqaibel fo his exceptional suppot and guidance, thesis-wise and othewise. I thank him fo being available wheneve I needed his assistance. It is my pleasue woking with him. Special thanks to distinguished membes of my thesis committee D. Adnan Al- Andalusi and D. Shaif Iqbal fist, fo accepting to be on my thesis committee, futhe, fo thei useful guidance, motivation, and pofessional contibution to my success and that of this wok. I acknowledge the contibution of M. Uma M. Joha who was instumental in caying out the expeimental pat of this wok. I also thank him fo been patient with me thoughout this peiod. I also extend my appeciation to the UWB High Resolution Positioning goup fo thei suppot and motivation. I acknowledge my wife, fo he patience; my son, fo his innocence; my family fo thei suppot; my fiends fo thei best wishes; and KFUPM, fo the oppotunity it gave me to study and accomplish this wok. To all, I emain gateful. iv

5 TABLE OF CONTENTS DEDICATION... III ACKNOWLEDGEMENT... IV LIST OF TABLES... VII LIST OF FIGURES... VIII ABSTRACT... X ABSTRACT (ARABIC)... XI 1 INTRODUCTION OVERVIEW LITERATURE SURVEY Wall Modeling Simulation of Though-the-Wall Popagation Measuements of Though-the-Wall Popagation Existing Systems and Applications MOTIVATION OBJECTIVES THESIS STRUCTURE THEORY OF UWB THROUGH-WALL PROPAGATION AND CHARACTERIZATION DIELECTRIC PROPERTIES OF MATERIALS WALL ATTENUATION AND DISPERSION TECHNIQUES FOR MEASURING ATTENUATION AND DISPERSION THROUGH WALLS Time Domain Technique Fequency Domain Technique UWB CHARACTERIZATION OF OBSTRUCTED PROPAGATION INTRODUCTION THE MEASUREMENT SETUP Component Selection Calibation The Sample Mateials MEASUREMENT PROCEDURE Tansmission Measuements Reflection Measuements ANALYSIS METHOD Single-Pass Technique Multiple-Pass Technique WALL PARAMETER CALCULATION v

6 3.5.1 Data Acquisition Un-gated Insetion Tansfe Function, Time delay and Initial guess of pemittivity Time Gating Wall (mateial) Paametes MEASUREMENT RESULTS Tansmission Measuements Results Reflection Measuements Results Between Tansmission and Reflection COMPARISON WITH LITERATURE MULTIPLE WALLS Double Laye Thee Laye ACCURACY RELATED ISSUES Souces of Eo Repeatability and Vaiability Analysis WALL COMPENSATION INTRODUCTION MEASUREMENT SETUP AND PROCEDURE WALL COMPENSATION METHODS Constant Amplitude and Delay Compensation Fequency Dependent Data Method Data Fitting Method CONCLUSION CONCLUSIONS AND RECOMMENDATIONS SUMMARY AND CONCLUSIONS RECOMMENDATIONS FOR FUTURE RESEARCH A APPENDIX A I. ANTENNA TEST RESULTS NOMENCLATURE REFERENCES VITA vi

7 LIST OF TABLES TABLE 3.1: Measuement setup equipment desciption TABLE 3.: Netwok analyze specifications TABLE 3.3: Antenna specification TABLE 3.4: Amplifie specifications TABLE 3.5: Length of cables used in the measuement TABLE 3.6: Geneal cable chaacteistics TABLE 3.7: Wall mateials showing dimensions TABLE 3.8: Coefficients of linea and quadatic fit fo the extacted paametes TABLE 3.9: Results compaison with liteatue TABLE 4.1: Showing constant amplitude and constant delay values used TABLE 4.: Taget position eos due to the walls TABLE 4.3: Pecentage similaity of wall compensation esults to No Wall esults TABLE A.1: Antenna test esults vii

8 LIST OF FIGURES Figue 1.1: Summay of eseach activities in the aea of though-wall popagation... 5 Figue.1: Time domain measuement setup... 7 Figue.: Fequency domain setup Figue 3.1: Measuement equipment Figue 3.: Losses in the cables used... 4 Figue 3.3: Tansmission measuements, (a) Pictue, (b) Schematic Figue 3.4: Reflection measuements, (a) Pictue, (b) Schematic Figue 3.5: Chat fo chaacteizing obstucted measuements Figue 3.6: Fequency domain measuements (a) measued magnitude, (b) measued Phase, (c) filte and filteed un-gated insetion tansfe function, (d) impulse esponses.. 63 Figue 3.7: The gating window Figue 3.8: (a) Un-gated time domain signal with window, (b) gated time domain signal 65 Figue 3.9: Tansmission: Insetion tansfe function vesus fequency fo diffeent walls Figue 3.10: Tansmission: dielectic constant vesus fequency fo diffeent walls Figue 3.11: Tansmission: Loss tangents vesus fequency fo vaious walls Figue 3.1: Tansmission: attenuation constant fo the diffeent walls... 7 Figue 3.13: Showing Eoneous data points fo; (a) Tansmission-insetion tansfe function, (b) Tansmission-dielectic constant, (c) Reflection-insetion tansfe function, (d) Reflection-dielectic constant Figue 3.14: Reflection: Dielectic constant vesus fequency fo the thee mateials Figue 3.15: Compaing tansmission and eflection esults fo wood; (a) insetion tansfe function, (b) dielectic constant, (c) attenuation constant, (d) loss tangent Figue 3.16: Compaing tansmission and eflection esults fo glass; (a) insetion tansfe function, (b) dielectic constant, (c) attenuation constant, (d) loss tangent Figue 3.17: Compaing tansmission and eflection esults fo a gypsum (a) insetion tansfe function, (b) dielectic constant, (c) attenuation constant, (d) loss tangent viii

9 Figue 3.18: Wall configuations Figue 3.19: Results fo double walls showing (a) magnitude of eceived signal (b) impulse esponse and (c) insetion tansfe function fo wood elative to fee-space Figue 3.0: Effect of inte-wall spacing (a) magnitude, (b) impulse esponse fo wood Figue 3.1: Block diagam of single laye and thee laye insetion tansfe functions Figue 3.: Thee laye measuements Figue 3.3: Effect of connecto mismatch Figue 3.4: Repeatability and vaiability analysis Figue 4.1: Taget measuements - (a) Taget Only (No Wall) (b) Taget + Wall Figue 4.: Reflections fom taget object with and without obstuction Figue 4.3: Illustating compensation using constant amplitude and constant delay, (a) wood, (b) glass, (c) gypsum Figue 4.4: Tansfe function of the localization scene Figue 4.5: Wall compensation using aw data fo wood sample in (a) fequency domain, (b) time domain Figue 4.6: Wall compensation using aw data fo glass sample in (a) fequency domain, (b) time domain Figue 4.7: Wall compensation using aw data fo gypsum sample in (a) fequency domain, (b) time domain Figue 4.8: Wall Compensation using fit to data fo (a) wood, (b) glass, and (c) gypsum Figue 4.9: Compensation using the thee methods fo wood wall Figue 4.10: Wall compensation fo double wall (wood-gypsum) Figue A.1: Antenna physical dimensions, gain and VSWR Figue A.: Antenna patten at vaious fequencies ix

10 ABSTRACT Name: Nuuddeen Mohammed Iya Title: Wall Compensation fo High Resolution Ulta-wideband Obstucted Localization Majo Field: Telecommunication Engineeing Date of Degee: Decembe 010 One of the inteesting popeties of an Ulta-wideband (UWB) signal is its ability to penetate walls and obstacles which comes fom the lowe fequency components of the signal. Howeve, as the signal popagates though these obstacles, it gets attenuated, slows down, and gets dispesed, which indicates the effect of the popagating medium. In this wok we demonstate wall compensation fo though-wall imaging, localization and communication eceive design puposes by fist chaacteizing wave popagation though vaious building mateials in UWB fequency ange. Fequency-domain tansmission and eflection measuements ae pefomed using a Vecto Netwok Analyze ove a fequency ange of 1 18 GHz to examine wall effects. This is done by measuing the insetion tansfe function given as the atio of two signals measued in pesence and absence of the wall. The dielectic constant and popagation loss ae extacted fom the measued insetion tansfe function using signal pocessing techniques. The wok consides typical indoo walls like glass, wood, and gypsum. Double laye walls and thee laye walls ae also investigated. Results fom tansmission and eflection measuements ae compaed with each othe and with liteatue. The esults obtained ae then futhe used to estimate and coect the position accuacy of a taget object located behind the walls using thee poposed methods namely; constant amplitude and delay, fequency dependent data, and data fitting methods. The obtained esults indicated elatively acceptable measue of wall compensation fo the thee methods. Results fom such wok povide insight on how to develop algoithms fo effective taget position estimation in imaging and localization applications. They ae also useful data fo channel modeling and link budget analysis. x

11 ABSTRACT (ARABIC) خلاصة الرسالة الاسم الكامل: نور الدين محمد أيا عنوان الرسالة: الجدار تا ثير تعويض للتحديد عالي الدقة للمواقع المحجوزة باستخدام تقنية التردد الطيفي فاي ق الاتساع (UWB) التخصص: هندسة الاتصالات. تاريخ الشهادة: ديسمبر ٢٠١٠ من أفضل خصاي ص الا شارة ذات التردد الطيفي فاي ق الاتساع (UWB) هو قدرتها على اختراق الجدران والحواجز وهذه القدرة تا تي من الجزء ذي التردد الطيفي الا قل من الا شارة. على الرغم من ذلك إلا إن مرور هذه الا شارة ( UWB )من خلال هذه الحواجز ينتج عنه اضمحلال وبطء وتشتت للا شارة وهذا يدل على تا ثير المرور من خلال المواد. في هذا العمل سنعرض تعويض الجدار وذلك للتصوير من خلال الجدار وتحديد المواقع وتصميم جهاز الاستقبال, وذلك من خلال تمييز مرور الموجة ذات التردد المماثل ل (UWB) من خلال العديد من مواد البناء. قياسات الا رسال والانعكاس في نطاق التردد( domain (Fequency ستنفذ باستخدام محلل أنظمة متجه ) Vecto (Netwok Analyze في النطاق الترددي من ١-١٨ جيجا هرتز لدراسة تا ثير الجدار من خلال قياس دالة النقل نتيجة لادخال الحاجز function) (insetion tansfe والمعطاة آنسبة الموجتين المقاستين في وجود وعدم وجود الجدار.تم قياس معامل العزل( constant (dielectic و خسارة النشر loss) (popagation من ) insetion (tansfe function وذلك باستخدام تقنيات تحليل الا شارة. في هذا العمل تم اعتبار جدران داخلية نموذجية مصنوعة من الزجاج والخشب والجبس. تم أيضا اعتبار جدران من طبقتين ومن ثلاث طبقات. تم مقارنة النتاي ج من الا رسال والانعكاس مع تلك الموجودة في البحوث السابقة. علاوة على ذلك فقد استخدمت النتاي ج في تقدير الموقع الدقيق والصحيح لهدف ما يقع خلف الجدار وذلك باستخدام ثلاث طرق وتحديدا ) and constant amplitude (delay و data) (fequency dependent و fitting).(data وقد أشارت النتاي ج لقبول نسبي للتعويض الناتج عن تا ثير الجدار للثلاث طرق. يمكن استخدام النتاي ج المستخلصة من هذا العمل في تصميم خوارزمية فعالة لتقدير موقع الهدف او للتصوير او تطبيقات تحديد المواقع من خلال الحواجز. أيضا هذه المعلومات مفيدة لنمذجة القناة.(link budget analysis) ميزانية الرابط channel )وتحليل modeling) xi

12 CHAPTER 1 1 INTRODUCTION 1.1 Oveview Recently, ulta-wideband (UWB) systems have been a subject of inceasing inteest. UWB systems ae fomally defined by the FCC as any adio whose bandwidth is at least 0% its cente fequency o a signal whose bandwidth is 500 MHz o moe [FCC0]. UWB signals have the unique capabilities of exta-wide bandwidth, low powe and multipath immunity. In addition, they ae pomised to have the ability to penetate walls and obstacles, which addesses a vaiety of applications anging fom indoo wieless communications, though-wall imaging, detection and localization, whee thee is a desie to see into obscued aeas. The study of adio signal popagation though walls is useful in that it shows the effect of the obstacle on the popagating signal o in othe wods, shows the behavio of the signal as it popagates though the mateial. This is usually achieved by investigating the inteaction of the electomagnetic wave incident on the wall sample and extacting its electical popeties. In this wok howeve, special consideation is given to UWB signals and thei behavio when popagating though these obstacles. 1

13 Electomagnetic waves passing though a medium ae subject to amplitude and phase distotions. These distotions ae categoically attibuted to dispesive and attenuative popeties of the medium of popagation. Thee is an inceasing need to undestand and model these impaiing effects in ode to find bette ways of mitigating them. In the context of though-the-wall detection, the ultimate objective is to use the dispesion and attenuation models in developing algoithms fo detection, classification, and localization of objects behind walls. This, in tun, necessitates accuate modeling of electomagnetic effects associated with wave popagation and scatteing, in pusuit of devising cedible solutions. Ignoing the popagation effects limits the scope of ou undestanding of the sensed data, deceases esolution, and educes the effective depth fo which accuate esults can be obtained. A popagation path obstuction is defined as a man-made o natual physical object that lies close enough to a adio wave path to cause a measuable effect on the path loss exclusive of eflection effects [TIA96]. Thus, electical popeties of the mateials that make up these obstuctions ae impotant data fo indoo adio communication planning and modeling, in addition to imaging and detection [Hua96]. It is, theefoe, of paamount impotance to study the electomagnetic popeties of these mateials fo examining though-the-wall detection and imaging issues and devising the desied solutions. In addition to mateials that make the wall, the shape of the wall and its composition also influence the popagation effects. Anothe effect is caused by multiple eflections within the wall. This impact becomes moe ponounced if the wall is

14 3 heteogeneous. The dielectic constants of obstuctions and thei thicknesses intoduce vaiable delays in the popagation path. The tavel time though the thickness of an object on the signal path is citical to the delay measuement when high accuacy is desied. The poblem becomes moe sevee in typical localization applications, whee tansmit and eceive antennas ae collocated on the same side of the wall. This equies the tansmitted signal to popagate though the wall twice. Anothe impotant challenge is associated with angles of incidence of the wave on the wall o fom (to) the tansmit (eceive) antenna. Futhemoe, in pactical situations, coupling effects, adiation patten, input impedance, and polaization of tansmit and eceive antennas ae impotant factos that need to be taken into consideation. In an effot to addess these poblems, this wok investigates popagation though diffeent building walls in ode to chaacteize them ove a wide ange of fequency. This is the fist step in the wall compensation pocess. Wood, Glass and Gypsum walls wee examined ove a fequency ange of 1 18 GHz. To the best of ou knowledge, this wide fequency ange has not been studied fo wall chaacteization. Most of the studies ae at specific fequencies, lowe fequency ange o the X Band. The chaacteization method is based on measuing the insetion tansfe function, defined as the atio of two signals measued in the pesence and in the absence of the wall. The dielectic constant of the wall mateial is elated to the measued insetion tansfe function though a complex tanscendental equation that can be solved using an appoximate one-dimensional oot seach. Tansmission and eflection measuements wee caied out in fequency domain using a vecto netwok analyze and a pai of wideband antennas to extact the insetion

15 4 loss and dielectic constant fo each wall mateial. Results obtained ae in ageement with those pesented in liteatue. Multiple walls wee also consideed and the effect of spacing between walls was investigated. In imaging and localization applications, the pesence of the wall and its effect on the taget s position cannot be ignoed. The wall is mostly thee and the peceived taget position is usually shifted due to the wall chaacteistics. This poblem is investigated and we sought to use infomation about the wall obtained fom the wall chaacteization study to coect the position estimation of the taget object by de-embedding the wall and thus, emoving its effect fom the pocess. 1. Liteatue Suvey This section povides a liteatue eview of the ecent woks on electomagnetic wave popagation though obstacles. It will be seen that much has been done to chaacteize obstacles fo popagation effects in naowband fequencies using simulations, theoetical models, and expeimental techniques fo vaious mateials. Howeve, moe study is equied on wideband chaacteization of these mateials, paticulaly fo though-the-wall detection and communication puposes. Figue 1.1 classifies the wok of diffeent authos fo wall mateial chaacteization based on diffeent citeia. These ae types of building mateial (wood, glass); conditions and stuctue of mateials like wetness, and moistue content; methods used to extact wall chaacteistics, whethe though expeimentation o simulation; paametes used to chaacteize vaious walls, e.g eflection coefficient, dielectic constant, etc; and fequencies at which the paametes ae extacted. This classification as well as the list of authos, howeve, is not exhaustive.

16 5 Building Mateial Methodology Type Glass [Muq05], [Cui01], [Jat05], [Tes07a], [Lee04] Wooden doo [Muq05], [Tes07a], [Lee04] Concete [Liu07], Muq05], Bick [Cui01], [Liu07], [Pen03] Plywood [Liu07], [Muq05], Reinfoced Concete [Pen03] Styofoam [Muq05], [Liu07], [Yaz04] Cloth Patition [Liu07], [Yaz04], [Jat05] Condition Wet [Ali03] Moistue [Gul05a] Dy [Liu07], Stuctue Homogenous [Pen03], [Gul05], [Deh08], [Tes07] Inhomogeneous [Deh08] Multiple walls [Muq03b] Slab [Muq05], [Muq08], [[Tes07a], [Lee04] Rough Suface [Yoo99], [Vil08] Simulation [Liu07], [Yaz04], [Aku04], [Ali03], [Ous05], [Cui01] Theoetical [Yaz04], [Muq03a], [Pen03], [Tes07a], [Ka93] Expeimental Time domain [Muq05], [Lee04], [Sch06] Fequency domain [Cui01], [Muq05], [Tes07a] [Tes07b], [Wil0], [Sag05], [Lee04], [Jat05] Tansmission Measuements [Aku04], [Sag05], [Tes07a], [Ka93] Reflection Measuements [Aku04], [Ka93], [Vil08] Oblique Incidence [Cui01], [Pen03], [Tes07], [Ka93], [Vil08] Diect Incidence [Muq05], [Sag05], [Wil0] Chip wood [Cui01], [Tes07] Plasteboad [Cui01], [Liu07], [Tes07a] Stuctue wood [Liu07] Dywall [Liu07], [Wil0], [Lee04] PVC [Sag05] Cinde Block [Wil0], [Tes07a] Gypsum [Jat05] Fequency UWB [Yaz04], [Muq05], [Liu07], [Lee04], [Sch06], [Ka93], [Gul05], [Jat05], [Deh08] S-band [Aku04] - 16 GHz [Tes07] 1-1 GH z [Wil0] X-band [Ous05], [Sag05] Application Communications [Pen03], [Oka09] Imaging [Hua96], [Ahm07], [Wan06], [Wen08], [Aft09], [Ha08] Detection & Localization [Cha08], [Guo08], [Sac08], [Lai05], [Shi08] RF/EM Shielding [Vae88], [Ant03] Repoted Paametes Dielectic constant [Muq05], [Ous05], [Sag05], [Pen03], [Tes07a], [Sch06] Penetation loss [Has93], [Jat05], [Tes07b] Tansmission Coefficient [Aku04], [Wil0], [Pen03], [Tes07a] Reflection coefficient [Aku04], [Wil0], [Sag05], [Pen03] Loss tangent [Cui01], [Muq05], [Wil0] Delay [Sag05], [Sch06], [Jat05] Insetion loss [Muq05] Retun loss [Gul05] Conductivity [Cui01], [Liu07], [Tes07], [Pen03] Insetion tansfe function [Muq05] Figue 1.1: Summay of eseach activities in the aea of though wall popagation

17 6 Seveal studies have been conducted on the electomagnetic chaacteization of building mateials, and vaious techniques wee poposed both in the naowband and wideband anges of fequencies. Fo instance, 433 MHz, 868 MHz,.4 GHz, and 5 GHz [Ali03]; GHz [Aku04]; 5.8 GHz [Cui01]; 1 6 GHz [Lee04]; 8 1 GHz [Ous05]; and 900 MHz [Pen03] ae example fequencies o fequency anges used. Moe ecently, studies wee conducted on the UWB chaacteization of building mateials [Muq03b], [Lee04], [Muq05], [Liu07]. Electomagnetic paametes, including dielectic constant [Muq05], [Ous05], [Tes07b], [Yaz04], loss tangent [Muq05], [Wil0], conductivity [Cui01], [Tes07b], insetion loss [Muq05], etun loss, path loss o popagation loss [Jat05], [Pen03], delay spead [Jat05], [Sag04], eflection coefficients [Aku04], [Sag05], [Wil0], and tansmission coefficients [Aku04], [Tes07a], [Cui01], [Wil0], ae commonly used to chaacteize the mateials fo popagation effects. These paametes ae usually extacted though vaious theoetical and expeimental methods. It should also be noted that all these authos epoted the thickness of a wall as a citical paamete in chaacteizing the wall mateial when subjected to electomagnetic signals. Liteatue on vaious aspects of though wall popagation is eviewed in the sub-sections that follow Wall Modeling Many theoetical models, both statistical and analytical, have been poposed to extact wall paametes at vaious fequencies. Ealie investigations on such chaacteization wee summaized in a compehensive eview by Hashemi [Has93]. Moe ecently, on a moe statistical note, Liu et al. [Liu07] analyzed channel capacity fo seven diffeent

18 7 building mateials and thei effects on a MIMO UWB system. Akuthota et al. [Aku04] developed an electomagnetic model to detemine the dielectic popety pofile of cement-based mateials. Thei model was based on detemining the effective popagation constant fom the tansmission and eflection popeties, and it is useful in pedicting changes in the dielectic popeties of mateials. In anothe development, Cuinas and Sanchez [Cui01] poposed an intenal multi-eflection model that takes into account the mateial thickness s effect on phase measuements. The model was futhe used to augment expeimental esults obtained fom studying electomagnetic popeties of six typical building mateials to detemine thei amplitude and phase esponses. Pena et al. [Pen03] used two ay-tacing models to estimate attenuation, pemittivity, and conductivity of a bick and concete wall at 900 MHz. Sagnad and El Zein [Sag04], [Sag05] pesented a high-esolution method based on the Matix Pencil algoithm to econstuct the impulse esponse by identifying individual multipath components within a mateial sample. This method utilizes the fact that each path is chaacteized by its own complex delay and amplitude. Othe theoetical appoaches ae available in [Muq03a], [Yaz04], and [Tes07a]. In [Muq03a], Muqaibel and Safaai-Jazi poposed a simplified model that uses a onedimensional seach algoithm fo detemining the complex dielectic constant fo wall mateials. This was obtained by analyzing the tansfe function of the wall. Yazdandoost and Kohno [Yaz04] povided a compact analytical fom of computing the complex elative pemittivity of building mateials fom fundamental pinciples of electomagnetic waves fo the UWB chaacteization of mateials. Thei esults fo the complex dielectic

19 8 constant of Styofoam with 10 cm thickness and cloth patition with 6 cm thickness closely match those of [Muq03a]. Nooi et al. [Noo08] obtained an analytical model to pedict the impulse esponses of the wall mateial fom tansmission coefficients assuming oblique wave incidence, and studied thei effect on UWB tansmission. 1.. Simulation of Though-the-Wall Popagation Simulation tools also play impotant oles in studying popagation effects. Ali-Rantala et al. [Ali03] studied how diffeent walls and wall mateials affect the attenuation of electomagnetic waves using an advanced compute simulation tool. In addition, [Ous05] developed two algoithms using theoetical fomulas fom [Muq03a] to extact mateial chaacteistics using Anosoft s High Fequency Stuctue Simulato (HFSS) softwae. These models alone, howeve, ae inadequate in poviding a compehensive desciption of the system popagation behavio. To effectively chaacteize electomagnetic wave popagation though obstacles, accuate measuements ae equied to augment the theoetical modeling. The majo difficulty facing measuement accuacy is that of adequately taking into consideation all effects, including antenna, cable, connectos, and multipath effects, and thei inteaction with the complex indoo envionment. Although antenna effects had been studied [Ada08], multipath effects wee educed though time-domain gating [Muq03b], [Muq08] o anechoic chambe measuements [Sag04], [Sag05], [Jat05], and calibation was pefomed to take cae of cable eos, measuements still have to be epeated seveal times to ensue accuacy and epeatability of the esults. Theefoe, the theoetical and expeimental appoaches ae complementay and should be used togethe.

20 Measuements of Though-the-Wall Popagation Measuements fo chaacteizing though-the-wall popagation effects can be conducted in the time domain o in the fequency domain. Time-domain measuements on samples of bick and concete wee caied out in the UWB ange of fequencies [Nem06], [Sch06], and wall paametes of thickness, pemeability, and pemittivity wee estimated at 900 MHz [Pen03]. In [Muq05], it is obseved that significant distotions occu when a bipola Gaussian pulse is passed though a bick wall as compaed to a wooden doo. Gulck et al. [Gul05a] demonstated a non-contacting detemination of moistue content in bulk mateials using sub-nanosecond UWB pulses. In anothe development, Gulck et al. [Gul05b] chaacteized mateials using UWB pulses fo the pupose of localization. Attiya et al. [Att04] used the time-domain technique to examine the potentials and limitations of though-the-wall human body detection. A easonable amount of wok has been caied out using the fequency domain technique [Muq05], [Tes07b], [Jat05], [Lee04]. In a thoough investigation, Muqaibel et al. [Muq05] pefomed a UWB chaacteization of ten commonly used building mateial samples. The esults obtained fo the dielectic constants of these mateials wee found to be in ageement with those epoted in the liteatue. Although thee ae significant diffeences and inaccuacies in the values of loss tangents fo diffeent mateials at diffeent fequencies obtained by diffeent eseaches, these discepancies ae lagely attibuted to the diffeence in mateial composition. It is also obseved that the dielectic constants of unifom mateials like wood and dywall, on an aveage, tend to decease

21 10 with fequency, while those with nonunifom stuctues like bick and concete block exhibit a moe complex behavio. Tesseault et al. [Tes07b] based thei though-the-wall measuements on tansmission powe fo doo, chipboad and wall beaing. Results indicate a slight decease of pemittivity with fequency, while conductivity, which was calculated fom the imaginay pat of the complex pemittivity, shows a moe sensitive vaiation with fequency fo the chipboad. Jatuatussani et al. [Jat05] used biconical antennas to find the dispesion and the penetation loss fo fou diffeent mateials. Lee et al. [Lee04] made obstucted LoS measuements on samples of dywall, wooden doo, and glass doo. Fequency esponses showed a maked signal loss in the case of dywall, and the impulse esponse showed signal distotion and consideable delay. It is also obseved in [Hua96] that fo walls constucted fom composite mateials, the theoetical mateial constants ae appoximations and, theefoe, wall paametes including the thickness can be estimated fom the eflected (o tansmitted) signal. In [Vil08], eflection measuements wee pefomed at diffeent angles whee an aluminum plate was used as a eflection efeence, lage enough to cove an antenna beam footpint at any incident angle. Howeve, the scatteing effects of the wall suface caused by any suface oughness wee not consideed because these effects wee vey weak accoding to the estimated standad deviation of suface oughness of 0.05 mm that is small compaed to a wavelength of 5.66 cm at 5.3 GHz.

22 Existing Systems and Applications A good undestanding of the wall effect has allowed fo bette imaging systems because getting infomation on the intenal featues of a stuctue makes it much easie fo activities in suveillance, escue and militay applications [Sin07]. It is also useful in manufactuing to detect unwanted objects in a poduction line. A typical example of a eal life application of though wall imaging is Pism 00 TM [Cam60] used by tactical opeatos which indicates the location and numbe of people behind a wall o baie and is capable of penetating concete, einfoced concete, cinde block, bick, dywall and othe common wall types, with a ange of up to 0 m. Methods on how to accuately detect a taget behind walls have been poposed by a numbe of eseaches. Chanda et al. [Cha08] applied a singula value decomposition algoithm to minimize clutte and detect a metallic taget behind plywood and bick wall in the UWB fequency ange. Guolong et al. [Guo08] illustated the impact of delay in position accuacy and used a though-the-wall compensation algoithm to coect the position of the located human to within 4 cm. Rovnakova et al [Rov09] poposed two methods to coectly tace moving tagets behind walls and compensate fo the wall effect. Imaging accuacy, howeve, is not only dependent on signal pocessing but also on the availability of detailed infomation about buildings, which includes mateial constants (pemittivity and conductivity), thicknesses of walls, as well as the stuctues of the buildings themselves [Hua96]. Recently, infomation on popagation effects is also used to econstuct buildings fo imaging and localization puposes [Ba08], [Ba06]. In [Ba08], model-based

23 1 easoning achitectues ae developed to estimate stuctual details and coectly model a building and its contents. This type of appoach obtains image infomation and, using popagation infomation, pedicts a thee-dimensional building model that best matches the obtained infomation. [Ahm07], [Wan06], and [Wen08] also demonstated the imaging poblem with pactical assumptions of unknown wall paametes. Undestanding the wall effect also helps in the development of RF/EM shielding fo impotant building locations. Vaessen et al. [Vae88] investigated RF shielding popeties of low-cost building mateials whee gids wee embedded in addition to coating, and both wee examined theoetically and expeimentally. Results illustated that unlike metal coating, wie gids exhibit significant fequency dependence. Thee is no doubt that the study of though-obstacle popagation has led to inteesting developments in wieless adio systems paticulaly in eceive design and link budget analysis. Howeve, as eseach on popagation though diffeent mateials continues in ode to cove moe gounds, moe possibilities emege poviding new diections fo study, and thus, applications like though-wall imaging and localization ceate motives fo futhe eseach. 1.3 Motivation Advances in communications and militay have led to inceased inteest in the field of though the wall popagation. One inteesting aspect of this is the detection of objects o humans behind obstacles, which has found applications in escue missions, suveillance, and econnaissance effots. Electomagnetic waves taveling though an obstacle ae subject to amplitude and phase distotion which is bought about by many

24 13 factos including the shape and composition of the wall. These factos play impotant oles in finding the exact position of a taget behind a wall. Uncetainties in wall paametes blu and defocus taget images causing them to shift away fom thei tue positions [Ahm07]. It is theefoe essential to study these advese effects with the aim of finding eliable solutions. A consideable amount of eseach has been done to study the effect of wall composition on wave popagation [Tes07a], [Vil08], [Cui01] and consequently its effect on taget detection [Ahm07]. Howeve, most of these studies ae in the naowband fequency anges o at specific fequencies. Futhemoe, little has been done to develop solutions to coect the wall effects. One of the main objectives of this wok is to investigate the inteaction of an ultawideband signal incident on the wall sample and extacting its electical chaacteistics using fequency domain adiated techniques. Ulta-wideband signals ae known to have high penetation capability and they povide high esolution. The second majo objective is to use the infomation obtained about the wall to coect the advese effect of the wall on the position accuacy of an object located behind it. Radiated methods, unlike cavity o waveguide methods, lend themselves to nondestuctive and boadband applications. Pefoming measuements in the fequency domain allows fo the application of a wide vaiety of noise eduction techniques. The infomation obtained fom such investigation is useful to the following applications: 1. Designing a wieless communication system to suit the application needed and thus, the antennas [Yan08], pulse shape [Foo04], and eceive can be tailoed to impove system pefomance.

25 14. Develop though-wall imaging and localization systems. These systems ae paticulaly impotant in suveillance, escue [Sin07], and militay applications. These systems ae based on shot-pulse wavefoms which can penetate mateials and povide pecise anging infomation [Mah05]. Ulta wideband has emeged as an excellent candidate fo such applications [Aft09], and [Ha08]. 3. Designing building stuctues to withstand RF and EM wave penetation [Ant03], [Vae88] commonly known as EM/RF Shielding. Indoo wieless netwoks tansmit RF signals that often popagate outside the physically contolled aea of a building posing a secuity isk to sensitive data. Telecommunication [Lem0], medical, eseach, and militay installations located nea AM, FM and TV stations ae subjected to RF ingess, equipment malfunction, test pocedue difficulties, woke health poblems, and compomised secuity situations paticulaly at defense facilities. Since shot-pulse wavefom systems like UWB ada can be used in imaging applications to locate humans o objects behind obstacles [Sac08], [Shi08], and [Lai05], anti-imaging walls can also be designed to hinde ada penetation paticulaly in secuity situations, and in poviding pivacy. 4. Modeling the obstacle fo communication puposes as in office (industial and hospital) buildings [Pen03]. Infomation on building mateial popeties like attenuation and insetion loss will be useful in choice of mateial fo constucting say a adio eseach facility fo example. Additionally, modeling the obstacle can be used in healthcae, whee the human body is envisioned as a fequency-dependent, iegula, lossy, and inhomogeneous

26 15 dielectic mateial [Fo07], and theefoe EM wave popagation though it is geatly influenced. An example of this can be seen in wieless body aea netwoks (WBAN) [Fo07], [Dib06]. While many studies had been caied out on chaacteization of obstacles fo popagation effects, it is wothy of note that most of them elate only to cetain applications, like ada, o at specific fequencies, o fo indoo popagation, etc, and theefoe cove only a sub-goup of the numeous eseach aeas. Thus, many issues ae still subject to futhe eseach. 1.4 Objectives In the couse of pefoming this wok, the objectives ae as follows: 1. Develop a test-bed fo obstucted UWB localization. The system is a fequency domain setup based on a Vecto Netwok Analyze. This task includes choosing and acquiing mateial samples, wideband antennas, low noise amplifies, wall mount fo easy movement of wall samples, and appopiate cables.. Conduct tansmission and eflection measuements to chaacteize some typical indoo walls (wood, glass, gypsum). The chaacteized walls should be pesented in tems of the insetion loss function (magnitude and phase) poviding infomation about the impact of the specific thickness and how the etieved data can be futhe genealized. Paametes like dielectic constant and loss tangent ae also obtained. Diffeent scenaios (diffeent distance fom the wall, multiple walls, space between wall, etc) ae consideed and study the impact of on the wavefom.

27 16 3. Demonstate wall compensation by using the pio knowledge of the wall to coect the position estimation and pefom eo analysis. This is achieved using thee methods namely; Constant Amplitude and Delay, Fequency Dependent Raw Data Method, and Data Fitting Method. 1.5 Thesis Stuctue Afte having an intoduction which includes a eview of liteatue and an insight on the objectives and motivation behind this thesis wok, Chapte gives a theoetical backgound on though-wall popagation. This includes an oveview of EM wave popagation in dielectic media, vaious methods used in wall chaacteization, including the fequency domain technique we used. Chapte 3 details the desciption of ou measuement setup fo this wok including the equipment, accessoies and mateials used. Expeiments fo studying UWB signal popagation though typical walls ae also pesented thee. Conducted tansmission and eflection measuements fo wood, glass and gypsum ae descibed and the esults ae pesented. Repeatability and vaiability analysis ae also coveed. In chapte 4, wall compensation is demonstated by using esults obtained fom chapte 3. The expeiment done to detemine an object behind the wall is descibed and the methods used to coect its position ae discussed. Chapte 5 concludes the wok with a summay and gives suggestions fo futue eseach. This thesis also includes an appendix and a nomenclatue list. The appendix povides the manufactues test esults fo the antenna used in the measuements. The Nomenclatue povides a list of all symbols used in the text with thei meanings.

28 CHAPTER THEORY OF UWB THROUGH-WALL PROPAGATION AND CHARACTERIZATION.1 Dielectic Popeties of Mateials The popagation chaacteistics of electomagnetic waves taveling though walls ae lagely detemined by the type of mateial composing these walls. Wall compositions ae, in geneal, dielectic and nonmagnetic in natue. Thus, they exhibit no esponse to magnetic fields. Howeve, when such mateials ae subjected to electic fields, numeous electic dipoles ae ceated within thei molecula stuctues. These dipoles tend to align along the diection of the extenal electic field, E. The cumulative effect of the localized shift between bound positive and negative chages is called polaization, P. It coesponds to a state of stess within the mateial, which gives ise to potential enegy stoage. This enegy is eleased when the extenal electic field is emoved. The ability of a mateial to be polaized o stessed by extenal fields is a popety detemined by its molecula stuctue. Within the wall mateial, the density of electic field lines (i.e., the electic flux density, D) is enhanced due to polaization, such that 17

29 18 D D P (.1) o whee D is the fee-space electic flux density. The atio of the numbe of field lines o inside the mateial to that in fee space (absence of mateial) is called dielectic constant o elative pemittivity of the mateial: P 1 (.) E o The dielectic constant is, thus, a measue of the enegy stoage capability of the mateial. The pemittivity of a mateial,, is that of fee space, o, multiplied by the dielectic constant. The time-vaying (o ac) natue of the extenal field has pofound effects on the polaization of the mateial and its pemittivity. These effects tanslate into fequencydependent behavios and incemental changes in the mateial conductivity. The mateials fequency dependence gives ise to the phenomenon of dispesion, a subject that is addessed late in this section. To study the effects of time vaiations on electic popeties of mateials, it is customay to epesent the electic dipole by the hamonic oscillato model o the classical mass-sping system. Moe sophisticated models exist but ae beyond the scope of this wok. The following diffeential equation govens the displacement l of an electic dipole

30 with chage q and mass m in esponse to a time-hamonic applied electic field with angula fequency ω [Bal89]: 19 l m t l k t sl qe e o jt (.3) whee k is the damping coefficient (fiction), s is the tension paamete (sping), E is the o amplitude of the applied electic field. The steady-state solution fo the displacement is eadily obtained by using / t = jω in (.3) and solving fo l. The esult is [Bal89] jt qeoe l( t) m (.4) s k j m m The polaization due to N simila electic dipoles pe unit volume within the mateial is given by P Nql(t) (.5) Accodingly, the elative pemittivity is obtained as

31 0 1 o j m k j m s m Nq (.6) which is, in geneal, a complex quantity. The eal and imaginay pats of ae given by 1 m k m s m m s Nq o (.7) m k m s m m k Nq o (.8) The eal pat of the elative pemittivity accounts fo the enegy stoing capability of the mateial, while its imaginay pat gives ise to an incemental change in the mateial conductivity. Thus, the effective conductivity of the mateial becomes o s a s e (.9) whee s is the static conductivity (sometimes called the dc conductivity) esponsible fo the ohmic losses inside the dielectic mateial; it is vey small fo good dielectics. The ac

32 field conductivity, a, gives ise to heating of the dielectic mateial due to its dipole oscillations. The atio of the two enegies, namely, the stoed (displacement) enegy and the enegy lost due to conductivity, defines the loss tangent of the mateial. Including ac vaiations, the loss tangent, tan δ, is given by 1 s tan (.10) o a The detemination of the electical popeties of the wall mateials is essential, as these popeties affect a vaiety of applications including though-the-wall imaging. The ability of walls to alte the popagation of ulta-wideband (UWB) signals is lagely attibuted to the eal pat of the complex pemittivity and the effective conductivity of the wall mateial. These quantities can, in fact, be design paametes fo bette localization. In the light of the above discussion, the total electic cuent density within the mateial can be expessed as J E j E j (1 j tan E (.11) e o o ) The time-hamonic wave equation becomes E j E E E (.1) e o

33 whee o is the pemeability of the mateial and is unity fo non-magnetic mateials. Also, γ is the complex popagation constant defined as j j( j ) (.13) e o The paametes α and β ae best known as the attenuation and phase constants, espectively. They can be expessed in tems of the mateial popeties and fequency as o e o Np/m, (.14) and o e o ad/m (.15) The attenuation suffeed by a signal impinging on a wall can be, in geneal, attibuted to conductivity loss, eflection loss, and multiple eflections within the wall. The conductivity loss may be a significant attenuation facto especially at highe fequencies and in the pesence of liquids. Howeve, most dy wall mateials do not

34 3 exhibit significant ohmic losses. The losses due to eflections depend on the degee of contast in the dielectic constant between the wall mateial and fee space and the angle of incidence. Effects of multiple eflections inside the wall stuctue become significant when inhomogeneities ae pesent and when the wall thickness is much lage than the signal wavelength. The ealization of an object s position behind the wall equies the coect building up of the signal phase. Walls can contibute significantly to the alteation of the signal phase. While the signal attenuation due to though-the-wall popagation can be easily compensated, say, by inceasing the level of tansmit powe, the teatment of the signal phase equies moe elaboate schemes.. Wall Attenuation and Dispesion As mentioned ealie, UWB systems pomise excellent wall penetation abilities. The popagation depth of an electomagnetic wave is popotional to its wavelength [Lee04]. The fine time esolution (faction of a nanosecond) achieved by UWB systems makes it a candidate fo pecise anging and localization of pesons o objects behind walls and obstacles. This is unlike the naowband systems that achieve high esolution using shotwave adio which in effect does not povide good penetation capabilities. Consequently, the popagation of UWB signals is significantly affected by the fequency-dependent popeties of mateials compising the popagation medium. Ove such a wide ange of fequencies, mateials exhibit divese behavios when inteacting with electomagnetic waves. As the fequency of the inteacting field inceases, the molecula dipoles of the mateial subjected to the field cannot espond instantaneously.

35 The esult of this sluggish esponse of the mateial to electomagnetic waves is dispesion. This phenomenon causes diffeent spectal components of a UWB signal to tavel at diffeent speeds. The attenuation caused by eactive losses (dipole oscillation) is also fequency dependent, implying that diffeent components of the signal ae attenuated by diffeent amounts. The most impotant manifestation of these effects is in pulse boadening, loss of amplitude, and, geneally, signal distotions. The diect consequences amount to the loss of all o some of the following: bandwidth, accuacy, and identification capability. In impulse adio UWB systems, naow time-domain pulses ae used. The pulse popagation chaacteistics ae best explained by the goup velocity concept. The goup velocity, denoted as g, is the velocity with which the electomagnetic enegy tavels, and is detemined fom 4 d g d (.16) In nondispesive media, the goup velocity is the same as the phase velocity, p and is constant at all fequencies. In dispesive media, the goup velocity is a function of fequency; thus, a delay diffeence develops between the tavel times of diffeent spectal components of a pulse. This delay diffeence causes pulse boadening, that is, the time width of the pulse is inceased and hence assumes a naowe bandwidth (loss of bandwidth). The pulse boadening phenomenon is moe ponounced fo vey shot pulses.

36 5 Dispesion theoies pedict that, away fom the esonance egions, the dielectic constant of a mateial inceases with fequency [Hay06]. In fact, it has been shown that the dispesion cuves of most mateials follow a few well-known classical models, such as Debye and Loentz models. In the Debye model, a elaxation time is defined to account fo how electic dipoles follow the behavio of the applied electic field. At low fequencies, the dipoles can follow the field closely, esulting in stong polaization. Howeve, as the fequency of oscillation of the applied field inceases, the dipoles stat to lose tack of field vaiations and weake polaization is expected. Thus, mateials with long elaxation times exhibit weak polaizability, while those with shot elaxation times show high degees of polaization. Fo dielectic mateials, the dispesion elation in (.6) is ewitten as s 1 j (.17) whee is the optical dielectic constant, is the static dielectic constant and τ is the s mateial elaxation time. The Debye model in (.17) is a mathematical expession that can be fitted to expeimental data. Once the fit is attained, the model can be used in analytical o numeical simulations. The Loentz model takes into account multiple esonances of the mateial as well as dipole coupling effects. It is diectly deived fom the motion Equation (.3). Intoducing Nq m p o and s o m, (.6) is expessed as

37 6 o s p (.18) j whee ω o is the mateial esonance fequency and Γ is the damping facto. Fo fequencies well below the esonance fequency of the mateial, the dielectic constant is eal and independent of fequency, and hence thee is no dispesion. As the fequency neas esonance, the dielectic constant inceases and attenuation also becomes significant..3 Techniques fo Measuing Attenuation and Dispesion though Walls Unlike naowband wieless systems, in which signal distotion is essentially caused by multipath components, in UWB wieless systems, signals may suffe significant distotions due to the dispesive popeties of wall mateials in the popagation path, multipath components, and also bandwidth limitations of tansmit and eceive antennas. In this section, techniques fo the wideband measuement of attenuation and dispesion of signals popagating though walls ae discussed. In paticula, time-domain as well as fequency-domain fee-space adiated measuement techniques, which lend themselves to in situ and nondestuctive applications, ae discussed..3.1 Time Domain Technique The time-domain technique consists of a pai of tansmit and eceive UWB antennas such as TEM hons, a pulse geneato, a digital sampling oscilloscope, and a tiggeing signal geneato as shown in Figue.1. Vey shot-duation Gaussian-like pulses ae adiated by

38 7 Tigge Pulse Geneato Step Geneato Pe-Tigge Data Acquisition Sampling Oscilloscope Low-noise Pe- Amplifie ai ai Low-noise Amplifie Wideband tansmit antenna Wideband eceive antenna Wall Figue.1: Time domain measuement setup

39 8 an antenna whose bandwidth is sufficiently lage that it causes negligible signal distotions. The signal is eceived by anothe antenna and detected by means of a wideband detecto such as a digital sampling oscilloscope. An ulta-wideband powe amplifie may be used at the feed point of the tansmit antenna if highe adiated powe is needed. The synchonization of both tansmit and eceive sides of the popagation channel is an impotant equiement in time-domain measuements. To maintain synchonization, a low jitte tiggeing signal is established between the pulse geneato and the digital sampling oscilloscope. The sampling oscilloscope equies a pe-tigge. This is achieved by using a step geneato dive that can supply the equied tigge and pe-tigge signals. The time delay intoduced by the tiggeing cables and the popagation path of the pulse is compensated by adjusting the time delay between the pe-tigge and the delayed tigge signals. Calibation and noise ae two othe impotant issues that need to be addessed. The pupose of calibation is to eliminate the effects of non-ideal chaacteistics of the measuement instuments fom the measued data. Also, eceived signals may suffe degadation due to the intefeence and noise fom vaious souces. The naowband noise is usually due to electomagnetic intefeence fom neaby naowband systems, and often takes the fom of a sinusoidal wavefom added to the eceived signal. This type of noise can be eliminated though bandpass filteing. The wideband noise, on the othe hand, efes to the themal noise in the eceive. The wideband noise typically appeas in the

40 fom of andom shot pulses and can be significantly educed though multiple signal aveaging, a featue geneally available in sampling oscilloscopes Fequency Domain Technique The fequency domain technique uses netwok analyzes to pefom swept fequency measuements within the intended bandwidth. Netwok analyzes can povide a wealth of knowledge about a device unde test, including its magnitude, phase, and goup-delay esponse. To accomplish this, a netwok analyze must povide a souce fo stimulus, signal-sepaation devices, eceives fo signal detection, and display/pocessing cicuity fo eviewing esults. The souce is usually a built-in phase-locked (synthesized) voltagecontolled oscillato. Signal-sepaation hadwae allows measuements of a potion of the incident signal to povide a efeence fo atio measuements, and it sepaates the incident (fowad) and eflected (evese) signals pesent at the input of the device unde test [Pac97]. The fequency domain setup also consist of a two pot S-paamete test set that povides both fowad and evese measuements. The RF powe is available fom eithe pot 1 o pot, and eithe test pot can be connected to the vecto netwok analyze s eceive inputs. Such test sets allow the use of full two-pot eo coection techniques fo the highest measuement accuacy. A pai of hon antennas is used with one of them as the tansmitte connected to pot 1 and the othe as the eceive connected to pot of the S-paamete test set as indicated in Figue.. The netwok analyze pefoms a swept fequency measuement within the equied fequency band. Each data point obtained fom the measuement is in

41 30 ai ai Wall Pot 1 Vecto Netwok Analyze S Paamete Test Set Pot Data Pocessing Invese DFT Pocessing Figue.: Fequency domain setup

42 31 a complex fom epesented by a magnitude and a phase tem. Thus, the fequency band and equied numbe of points has to be specified. Fo wideband chaacteization, whee a wide fequency ange needs to be swept, one has to make a tade-off between the fequency esolution and the equied numbe of measuements that is diectly popotional to the time it takes to pefom the expeiment as well as data stoage equiements. Diect phase measuements in wideband though-the-wall popagation ove long distances should be dealt with vey caefully, as available VNAs ae geneally designed fo the measuement of complex S-paametes of small two-pot netwoks athe than long popagation distances. Eos in diect phase measuements occu due to the diffeence between the measued fequency and the fequency of the eceived signal caused by the popagation delay time though the wall. In the sweep mode opeation, the VNA fequency changes linealy with the sweep time. With long popagation paths, the fequency of the signal at the end of the channel would be diffeent fom that at the beginning, esulting in eos in both magnitude and phase measuements. To avoid sweep mode eos, fequency stepping instead of fequency sweeping may be used. In the fequency stepping mode of opeation, the duation of the fequency step should be lage than the time delay between tansmit and eceive antennas. It is emphasized that the measued signal is the one ecoded at the end of the step athe than the aveage of measued signals eceived duing the time step. In the next chapte, a detailed desciption of the expeimental setup used in this wok fo chaacteizing UWB popagation though walls is given.

43 CHAPTER 3 3 UWB CHARACTERIZATION OF OBSTRUCTED PROPAGATION 3.1 Intoduction In communication and ada systems, the effect of signal distotion needs to be equalized. Obstuctions with linea popeties as well as those with non-linea popeties intoduce distotion in both magnitude and phase to signals passing though them. Linea effects impose changes in the time wavefom of signals by alteing the amplitude o phase elationships of the spectal components that make up the signal. No new signals ae ceated. On the othe hand non-linea effects can shift an incident signal in fequency o add othe fequency components esulting in a modified pulse shape o fequency content. Measuing both magnitude and phase components ae impotant in that they ae needed to fully chaacteize the obstuction and undestand its effect on the signal. In addition, impulse esponse chaacteization equies magnitude and phase infomation of the tansfe function in ode to pefom IFFT. Calibation, which uses vecto eo 3

44 33 coection, impoves measuement accuacy and needs magnitude and phase data to build an effective eo model. Measuements ae impotant in obstucted popagation studies because they povide a diect way of extacting obstacle popeties and pesent a clea undestanding of its effects on signal popagation. Measuements can be pefomed using both time domain and fequency domain techniques. In the time-domain method, a peiodic tain of pulses is applied to a mateial sample of known thickness using an antenna connected to a pulse geneato and is eceived by an identical antenna connected to a digital sampling oscilloscope. While in the fequency domain, sinusoidal signals ae used instead of pulses and a Vecto Netwok Analyze (VNA) with an antenna connected to each of its two pots is used to pefom sweep fequency measuements. Additionally, when extacting wall paametes like dielectic constant and loss tangent, a vaiety of methods ae employed togethe with the time domain o the fequency domain techniques. Such methods can be sampled methods like coaxial o tansmission line methods o they can be in-situ methods (e.g fee-space adiated measuements). In this wok, a fee-space adiation fequency domain technique is employed. The VNA povides highe dynamic ange than time domain equipment and also thee is no need fo synchonization between tansmitte and eceive as is the case with the time domain technique. Also, the fequency domain technique allows fo ease of application of a wide vaiety of noise eduction schemes. We pefomed the fee-space adiated measuements fo both tansmission and eflection. Thee ae seveal easons fo this: 1. Measuements using fee-space pocedues ae contactless and nondestuctive.. Cavity

45 34 and wave guide methods equie that sample mateials be machined popely to fit the cavity and/o waveguide coss-sections with negligible ai-gaps. This equiement affects the accuacy of measuements fo mateials that cannot be machined pecisely. 3. Some mateials ae inhomogeneous in stuctue due to vaiation in manufactuing pocesses. Because of this, unwanted highe ode modes can be excited at the ai-dielectic inteface in the waveguide and cavities. 4. Chaacteization using fee-space adiated measuements matches the configuation of the final application fo which the expeiments ae caied out, which is communications and ada. 3. The Measuement Setup This section details the desciption, the expeimental setup (including accessoies) fo both tansmission and eflection measuements, and the contibution of each to the measuement esults Component Selection Table 3.1 summaizes the hadwae used to build the though-wall UWB localization setup and Figue 3.1 shows thei pictues. Each of these components is then futhe descibed in the following headings. The Vecto Netwok Analyze System Vecto netwok analyze systems measue the magnitude and phase chaacteistics of netwoks and of components. The system consists of the souce, the s-paamete test set, the vecto signal pocesso display. Togethe these compise a complete esponse system

46 35 TABLE 3.1: Measuement setup equipment desciption Equipment Netwok Analyze System Desciption HP 8510C Netwok Analyze with 8361B RF souce and 8514B S- paamete test set Antenna Pai of boadband hon antennas (1 18 GHz) JXTXLB Amplifie Cables Tipod Connectos Low-noise amplifie (1-18 GHz) with SMA female LA1018N309 Times Micowave StipFlex low loss high pefomance coaxial cable (1 18 GHz) Tipod fo boadband hon antennas Two N-Type female to SMA male fo connecting to amplifie pots

47 36 (b) N-to-SMA connnectos (a) antenna (c) low w noise ampliifie (dd) vecto netw wok analyze (e) widebaand cables Figue 3.11: Measuemeent equipmennt

48 37 that povides stimulus to the device unde test and measues the signal tansmitted though the device o eflected fom its input. The Netwok Analyze System we used opeates between 45 GHz and 0 GHz. The dynamic ange is given in Table 3. below TABLE 3.: Netwok analyze specifications [Agi00] (GHz) 8 (GHz) 8 0 (GHz) Maximum powe (at pot ) +0dBm +11 dbm +10 dbm Refeence powe (at pot 1) + dbm - dbm -6 dbm Minimum powe (at pot ) -66 dbm -95 dbm -95 dbm Antenna The antennas used wee a pai of boadband TEM hon antennas opeating between 1 18 GHz. The specifications ae shown in the Table 3.3 below. A TEM hon uses TEM wave popagation having a velocity that is constant fo all fequencies. At the hon thoat, all fequencies geneated aive at the apetue togethe, and a wideband pulse is tansmitted fom the apetue. An antenna apetue is a measue of how big a piece of an incoming wave font an antenna can intecept. Hons ae egaded as constant apetue antennas because the apetue emains fixed with fequency, thus the tansfe function in tansmit will be linea with fequency, while the tansfe function on eceive is constant. As fequency f inceases, the size of this apetue in units of wavelength inceases as f. This naows the patten and inceases the antenna gain as f.this is paticulaly desied on the eceive side as it adds diectly to link pefomance [Ge03].

49 38 Wideband antennas povide popagation infomation ove a wide bandwidth, making it easy to identify and eliminate effects of unwanted eflections in time domain by way of time gating. Hon antennas ae diectional and ae theefoe useful in clutte eduction and enhancing oveall system pefomance. This also, makes them useful in time domain gating. TABLE 3.3: Antenna specification Manufactue Chengdu Ainfo Inc. China Pat Numbe JXTXLB Fequency ange Gain (Typical) Polaization 1 18 GHz 11 dbi Linea VSWR (Typical).0:1 Size Net Weight 41mm x 160mm x 40mm 1.38 kg The manufactue test esults fo the antenna showing the gain and VSWR ae shown in Appendix A-I. Amplifie The maximum output at the pot of the HP8510C netwok analyze is +17 dbm. Using path loss fomula fee space measuements with the wideband hon antenna at 18 GHz and antenna sepaation of 1.8 m, we have

50 39 pathloss 4 R (3.1) 0 log(0.0167) 0 log(4 ) 0 log(1.8) = db Thee is a maximum cable insetion loss of. db pe mete [Dyn05] at 18 GHz giving 16.5 db fo total length of ou cables (7.5 m). The loss was computed to be aound 79.1 db. Theefoe, an amplifie on the tansmitte was equied to have sufficient gain and input powe levels to achieve a easonably high tansmit powe. Based on these equiements and limitations, a low noise amplifie with specifications given in Table 3.4 was used. The amplifie is equied to exhibit constant good gain and phase esponses ove the band of inteest. Cables Long cables connecting the S-paamete test set pots to the tansmitte and/o eceive geneally ceate a majo limitation, because at high fequencies cable losses incease exponentially, causing a significant eduction in the dynamic ange. The cables used fo the measuement ae labelled and the lengths ae indicated in Table 3.5 below. The cable losses ae also measued and the plots ae shown in Figue 3.. Table 3.6 also summaizes the geneal chaacteistics of the cables.

51 40 TABLE 3.4: Amplifie specifications [A-INF] Manufactue Pat Numbe Fequency ange Gain Noise Figue P 1 db Flatness IP3 Chengdu Ainfo Inc. China LA1018N GHz Min: 8 db, Max: 36 db Max: 4.5 db Min: +9 dbm Max: +/-. db 18 db (typical) VSWR. Cuent (+1V) Connecto Net Weight 50 ma (typical) SMA female kg TABLE 3.5: Length of cables used in the measuement Cable Label Length (metes) A 1.5 B 1.5 C 4.5

52 41 TABLE 3.6: Geneal cable chaacteistics Manufactue Categoy Model Desciption Impedance Max. Opeating Fequency Cut-Off Fequency Nominal Impedance Times Micowave Stipflex SF-14B Flexible 50 Ohms 18 GHz 34.1 GHz 50 Ohms Nominal Velocity Of Popagation 70.7 % Nominal Delay Maximum Opeating Voltage Maximum CW Powe Rating Maximum Retun Loss Maximum Insetion Loss Nominal Insetion Loss 1.44 Ns / feet 1,879 VRMS 85 Watts at 18.0 GHz -0 db at 18.0 GHz 75.1 db / 100 feet at 18.0 GHz 68.3 db / 100 feet at 18.0 GHz

53 4 0 - Cable A Cable B Cable C -4 Magnitude, db Fequency, GHz Figue 3.: Losses in the cables used

54 43 Connectos The fequency ange of any connecto is limited by the excitation of the fist cicula waveguide popagation mode in the coaxial stuctue. Deceasing the diamete of the oute conducto inceases the highest usable fequency; filling the ai space with dielectic lowes the highest usable fequency and inceases system loss. Pefomance of all connectos is affected by the quality of the inteface fo the mated pai. If the diametes of the inne and oute conductos vay fom the nominal design, if plating quality is poo, o if contact sepaation at the junction is excessive, then the eflection coefficient and esistive loss at the inteface will be degaded. As the inteface between instuments, cable, DUT, test fixtue, the connecto is a citical element in achieving good measuement esults. Connectos may look mundane, but they ae fagile, pecisionmachined components and ae highly sensitive to cae in handling [Hie08]. An adapte with N Type-to-SMA connecto was used to connect the cables to the amplifie. N Type is one of the most common RF connectos in use aound today. It is a high pefomance connecto designed by Bell Labs in the 1940 s. The N connecto is ugged, elatively inexpensive and the standad vesion is capable of mode-fee opeation to 11 GHz with pecision vesions suppoting up to 18 GHz. The SMA was oiginally intended fo use on semi-igid coaxial cable and late fo flexible cable. SMA opeates up to 18 GHz with the pecision vesions extending the uppe fequency limit to 6.5 GHz [Han09].

55 Calibation A measuement calibation pocedue tansfes the accuacy of the calibation standads (Open, Shot, and Load) to the measuement of the device. Since the esponse of the standads is known to a high degee of accuacy, the system can measue one o moe standads, then use the esults of these measuements to povide data to algoithms which pocess the measued data fo display. This pocess is called measuement calibation, accuacy enhancement, o eo coection [Agi01]. A Full -pot measuement calibation was pefomed in ou case because it povides the best magnitude and phase measuement accuacy. The fou calibation standads used ae shielded open cicuit, shot cicuit, a load and a thu. This model povides full eo coection of diectivity, souce match, eflection and tansmission signal path fequency esponse, load match and isolation fo S 11, S 1, S 1, and S. Ou measuement also has an inheent calibation in that the pocedue involves pefoming a fee-space efeence measuement and a though wall measuement using the exact setup. Any eos in the efeence measuement will be cancelled by the second measuement The Sample Mateials Thee typical building mateials wee used as wall sample fo chaacteization. These ae wood, gypsum and glass. Table 3.7 shows the mateials and dimensions of each.

56 45 TABLE 3.7: Wall mateials showing dimensions Mateial Dimension (cm) Wood Glass Gypsum Boad Measuement Pocedue As mentioned ealie, the fequency domain technique was employed. This technique is descibed in section.3. Fo the pupose of wall chaacteization, a HP8510C Vecto netwok analyze with two-pot S-paamete test set was used and each data point obtained fom the measuement is in a complex fom epesented by a magnitude and a phase tem. The maximum numbe of points obtained by the HP8510C VNA is 801. The output of the netwok analyze pot 1 is fed though a 1.5 m cable to a wideband powe amplifie; anothe 1.5 m cable is connected to the amplifie output to a wideband antenna mounted on a tipod. An identical antenna mounted on a simila tipod stand is used as a eceiving antenna. The eceive antenna output is fed though a 4.5 m cable to the netwok analyze pot. The device unde test (DUT) is the sample wall mateial and it was chosen to be quite lage enough to cove beam footpint at any incident angle in ode to avoid edge diffaction and scatteing. Also its thickness was lage enough to povide consideable delay, which should be geate than the incident pulse width, to enable distinguishing

57 46 between multiple eflections inside the wall. Time gating can futhe be used to extact the main equied pulse, and eliminate all delayed pulses due to multiple eflections in the wall. On the othe hand, if the wall is too thick, it can seveely attenuate the signal. To chaacteize the walls, tansmission and eflection measuements ae conducted to obtain the insetion loss function in tems of magnitude and phase. This infomation is futhe used to coect the position estimation of a taget behind the wall as we will see late Tansmission Measuements This method uses a wave that is sent by a tansmitting antenna, popagates though a mateial of some thickness and is captued by a eceiving antenna. The mateial unde test is assumed to be the two-pot device (DUT) with an oveall tansfe function which elates the output pulse to the input excitation pulse [Au96]. The tansmission measuement also allows fo measuing both the attenuation and dispesion though the wall samples. Two measuements ae pefomed in this case. A fee-space measuement pefomed with empty space between tansmitte and eceive, and a though-wall measuement conducted with a wall mateial inseted between the two antennas as depicted in Figue 3.3. The tansmitte and eceive antennas ae aligned fo maximum signal eception. The wall mateial is placed at exactly midway between the antennas so that the wall is in the fa-field egion of each antenna, i.e. the sepaation distance between the antenna and the wall should be lage enough to ensue fa-field o Faunhofe egion

58 appoximation whee the antenna adiation patten is independent of the sepaation distance, and the electomagnetic field incident on the wall is essentially a plane wave. To achieve this, the Faunhofe distance, d f in this case, given as 47 d f D a must be much geate than both the lagest physical linea antenna dimension D a, and the wavelength. Fo the pai of antenna used, D a is 0.41 m, and the coesponding fa field distance at 18 GHz is 6.9 m. Howeve, a few pactical issues limited applying the fa field appoximation. Fist, at the time of pefoming these measuements, an anechoic chambe was not available, which would have allowed a consideable eduction of multipath effect. Theefoe, the measuement was pefomed in a lab which contains fixed tables and othe souces of multipath. Secondly, the available cables ae a bit long and signal loss inceases with distance and fequency, thus, amplifies ae needed on both the tansmit and eceive sides of the setup. Howeve, the available amplifie has 30 db gain and the maximum input powe the pots of the HP8510C VNA system is +17 db. So as not to damage the VNA, only one amplifie was used, which is on the tansmitte side as mentioned ealie. In ode to compensate fo a weak eceived signal and stay clea of multipath, while ensuing the accuacy of the esults, two actions wee caied out. Fist, the distance to wall fom the antenna was educed to less than the fa field appoximation such that eceived signal stength is inceased and edge diffaction fom the wall is educed. Next, time gating is pefomed to emove the effect of eflections fom the floo

59 48 (a) Fee Space E i E t fs Region II Region I (Ai) Region III (Ai) E i E t E d z (b) Figue 3.3: Tansmissionn measuements, (a) Pictue, (b) Schematic

60 49 and tables aound. The wall effect is based on extacting the complex dielectic constant fom the measued insetion tansfe function obtained as the atio of two tansmit signals given as [Muq03a] Et ( j) Ei ( j) Et ( j) X t ( j) H ( j) (3.) fs fs fs Et ( j) Et ( j) X t ( j) E ( j) i whee E i ( j) is the incident wave, ( j) is the tansmitted wave though the wall, and E t E fs t ( j) is the fee space tansmitted wave. The scatteing paamete elated to the insetion tansfe function in this case is given as S 1 ( j ) j 0 H ( j ) e (3.3) The tems X fs t ( j) and X t ( j ) ae fee-space efeence and though-wall fequency domain signals obtained in the absence and pesence of the wall espectively. The feespace measuement is used as efeence to take cae of the effects of components othe than the wall. Assuming a fictitious laye of fee-space of the same thickness as the wall, then the popagation delay though this laye is d c whee d is the laye thickness and c is the speed of light in fee space. Exta caution should be taken to make sue that exact setup is used to pefom both expeiments in ode to avoid measuement inconsistencies. 0

61 Reflection Measuements In eflection expeiments, a metallic eflecto, aluminum whose eflection coefficient is close to unity was used fo efeence measuements. Two measuements ae conducted hee also as with the tansmission measuements, but in this case, both antennas ae positioned on the same side of the wall as shown in Figue 3.4. The efeence signal is obtained with only the aluminum plate in place. While the second is pefomed with the wall pessed against the aluminum sheet. To enable us use the same model with that of the tansmission measuements in computing the dielectic constant, the wall thickness is assumed to be twice the oiginal size. This is because in the eflection expeiment, the signal popagates though the wall twice befoe aiving at the eceive, i.e., when it goes fom tansmitte though the wall to the aluminum sheet and when it is eflected by the aluminum sheet goes though the wall to the eceive. In addition, fo the model to wok fo eflection, we consideed only nomal signal incidence on the wall. The eflected wave only changes its polaity undegoing a 180 phase change. Cae was taken to ensue that thee ae minimum ai gaps between the aluminum sheet and the wall samples. The antennas wee aanged one above the othe with eceive being above at a distance of 1 m fom the floo, and ae conveniently spaced apat to educe coupling. A sepaation distance of 0.6 m was found to be convenient and the eceived signal was good. 3.4 Analysis Method The fee-space and though-the-wall measuements would be most accuate if pefomed inside an anechoic chambe, which absobs all multipath components and eflections fom

62 51 (a) Wall Mateial Metallic Reflecto (b) Figue 3.4: Reflection measuements, (a) Pictue, (b) Schematic

63 5 the floo and the ceiling. Ideally, to avoid scatteing fom the edges, the sample wall to be measued should be infinitely wide. Also, samples unde test have to be in the fa-field egions of tansmit and eceive antennas, typically seveal metes fo the fequency ange of inteest and dimensions of the antennas used. Maintaining these equiements is not a convenient task, keeping in mind that absobes and chambe envionment do not allow easy movement of lage samples. Fotunately, time gating can be used to significantly educe the undesied effects such as scatteing fom edges and eflections fom suounding walls. Fo time gating to be efficiently implemented, thee conditions have to be met. Fist, the tansmit and the eceive antennas should be positioned away fom the eflecting sufaces. Second, samples should have elatively lage suface dimensions in ode to minimize the edge effects. Finally, thee should be flexibility in adjusting the distance between the antennas and the sample. Time gating can also be used to isolate a desied potion of the eceived signal; namely, the fist single-pass of the signal tansmitted though the wall. In this application, the sample thickness should be lage enough to yield sufficient delay. Thus, the fist pulse can easily extacted, and all delayed pulses due to multiple eflections inside the wall can be emoved Single-Pass Technique The single-pass technique can be used if the potions of the signal due to multiple eflections inside the wall ae eithe negligible o can be eliminated by means of time gating. We assume that the wave is nomally incident on the mateial suface and the duation of the pulse is smalle than the pulse tavel time though the mateial. The

64 53 deivations petaining to the single-path popagation analysis ae available in the appendix of [Muq03b]. Assuming low loss, the esulting dielectic constant becomes eal-valued. The esults ae summaized below: ( f ) 1 d sp ( f ) ( f ) 1 1 df 0 0 (3.4) 1 1 ( f ) tan ln H sp ( f ) f0 ( f) 4 ( f) (3.5) whee H ( f ) H ( f ) exp[ j ( f )] is the single-pass insetion tansfe function. It is sp sp sp the atio of the Fouie tansfom of the single-pass eceived signal when the slab is in place to the Fouie tansfom of the eceived signal in the absence of the wall. It should be noted that the deivative tem d ( f )/ df in (3.4) is based on the assumption that the sp phase vaies linealy with fequency. The advantage of using the deivative of the phase is to avoid tacking the unwapped phase function. As a function of the unwapped phase, the dielectic constant is given by ( f ) sp ( f ) ( f ) 1 1 f 0 0 (3.6) 3.4. Multiple-Pass Technique If the single-pass signal cannot be gated out satisfactoily, multiple eflections fom the wall inteio that constitute pat of the eceived signal must be consideed. This situation

65 54 paticulaly aises when the tansit time though the thickness of the wall is small compaed to the pulse duation. In this case, an insetion tansfe function that accounts fo multiple eflections of a homogenous wall is needed. The model to be pesented can be used to detemine the complex dielectic constant fom the measued insetion tansfe function. To obtain an expession fo the insetion tansfe function H ( j), we assume a plane wave nomally incident on a dielectic wall of thickness d and pemittivity j, as depicted in Figue 3.3(b), establishes a eflected wave in egion I (ai), a set of fowad and backwad-taveling waves in egion II (wall), and a tansmitted wave in egion III (ai). Using electic and magnetic field expessions and bounday conditions fo the electic and the magnetic fields at the wall ai intefaces, the tansmission coefficient which is equivalent to S 1 in scatteing paamete teminology is eadily obtained as [Muq03a]. T e d 4 o e o d o o (3.7) Based on the definition of insetion tansfe function given in (3.), E E E i fs t E t i T j0d Te H ( j) j0d e (3.8)

66 55 Thus, o o d o o d d j e e e j H 4 ) ( 0 (3.9) whee o o o, ) ( o o j j j o o o o j A multi-pass tansfe function is obtained that accounts fo all the tansmitted waves including the ones esulting fom multiple eflections within the wall. Equation (3.9) can be solved fo the dielectic constant when the insetion tansfe function H(jω) is obtained by measuements as descibed in Section 3.3. It should be noted that equation (3.9) is a complex equation and its numeical solution equies two-dimensional oot seach algoithms. Assuming the wall occupying egion II is low loss, the elation 1 is satisfied, allowing consideable simplifications to be made, and and to be obtained fom sepaate eal expessions. The benefit of the simplified solution is that, instead of

67 56 time-consuming two-dimensional oot seach techniques, only a one-dimensional oot seach needs to be implemented. The following appoximations can be made [Muq03a] j ) ( 0 0 j j j j j j and ) ( o j Then, 1 1 o o and (3.9) becomes ) ( ) ( ) ( 0 d j d j d j e e e j H (3.10) Rewiting the tansfe function in tems of magnitude and phase, we get

68 ) cos( ) ( d d d e e j H (3.11) and d j H 0 ) ( (3.1) Whee ) tan( tan 1 d e e e e d d d d (3.13) In a compact fom, (3.13) can be put as ) tan( 1 1 tan 1 d Q e Q e d d (3.14) whee 1 1 1) ( 1) ( Q (3.15)

69 58 when X e d, then ) cos( ) ( d X X j H, (3.16) o 0 1 ) ( 8 1 ) cos( X j H d X which is a quadatic equation in X. Solving the equation fo X, we have ) ( 8 1 ) cos( ) ( 8 1 ) cos( d j H d j H d e X (3.17) Only the solution with the negative sign is valid. Substituting fo X fom (3.17) into (3.14) we get an equation which is only in tems of. 0 ) tan( 1 1 ) ( tan 0 d QX QX j H d (3.18) Solving this equation numeically, is eadily detemined. X and ae then found fom (3.17). Finally, is calculated using c (3.19)

70 Wall Paamete Calculation This section details the signal pocessing pocedues used to extact the wall paametes fom the measuements. Recall that two measuements ae caied out in each expeiment; a efeence measuement and a though measuement. Figue 3.5 summaizes these pocedues Data Acquisition Using the fequency domain technique, 801 complex data points epesenting magnitude and phase infomation ae obtained ove a fequency ange of 1 18 GHz. This is the maximum numbe of data points that can be obtained ove a given fequency ange on the HP8510C netwok analyze. This data is acquied and conveted to files eadable by MATLAB Un-gated Insetion Tansfe Function, Time delay and Initial guess of pemittivity The un-gated insetion tansfe function is obtained by dividing the though wall data by the efeence data. A finite impulse esponse filte of ode 100 is then used to emove the noise at the low fequencies and those beyond the antenna bandwidth. The fequency domain signals ae then conveted to time domain using invese fast fouie tansfom to get the impulse esponses. Zeos ae padded fo highe time domain esolution. Figue 3.6 shows the magnitude, phase, un-gated insetion tansfe function, and the impulse esponses fo the efeence and though signals.

71 The impulse esponses obtained fom fequency-domain measuements ae coelated using a sliding coelato to obtain the fist guess on the delay and effective dielectic constant. An estimate of the aveage dielectic constant could also be obtained though peak-to-peak impulse time delay,. This aveage dielectic constant, which does not eflect the fequency dependence, is given by 60 1 d c (3.0) whee d is the wall thickness and c is the speed of light in fee-space Time Gating In the absence of an anechoic chambe, multipath components, multiple eflections in the wall, and eflections fom the floo, ceiling and othe stuctues become a theat to the measuement accuacy. In ode to educe this effect of multipath, time gating is used to selectively emove o include the undesied esponses in time. The emaining time domain esponses can then be tansfomed back to the fequency domain with the effect of the gated-out esponses being emoved. The delayed signals can be filteed out by imposing a window function ove the dominating signal that is identified to be the desied one. In ou case, this window is based on the modified Kaise window [Muq03b] whee the paamete contols the amount of oll-off of the window function. Note howeve that, wall paametes ae independent of the type of window used. We used a of 0 and adjusted the window paametes (ise time, fall time and width) in ode to obtain accuate values fo the wall paametes.

72 61 S 1 fequency Domain measuements Un-gated Insetion tansfe function and FIR filte Ifft on S 1 data, with zeo padding to obtain impulse esponse Detemine peak-to-peak time domain diffeential delay using a sliding coelato Pefom time gating to emove unwanted eflections Fist estimate of delay and loss fft with zeo padding Insetion tansfe function Use equations to 3.17, 3.18, & 3.19 to extact unknowns, Dielectic constant, Loss tangent, Attenuation constant Figue 3.5: Chat fo chaacteizing obstucted measuements

73 6 The associated efeence and though wall fequency domain signals ae then ecoveed using fast fouie tansfom and equation (3.) is used to obtain the insetion tansfe function. Figue 3.7 shows the window, and in Figue 3.8 an un-gated time domain wavefom with a window, and time domain gated wavefom afte applying the window ae shown.

74 Feespace Though Wall Feespace Though Wall Magnitude, db Phase, ad (a) Fequency, GHz (b) Fequency, GHz x Feespace Impulse Response Though wall Impulse Response -0 4 Magnitude, db Amplitude in (V) Insetion Tansfe Function Filte Filteed Insetion tansfe function -4-6 (c) Fequency, GHz (d) time in secs x 10-8 Figue 3.6: Fequency domain measuements (a) measued magnitude, (b) measued Phase, (c) filte and filteed un-gated insetion tansfe function, (d) impulse esponses. 63

75 Amplitude time, ns Figue 3.7: The gating window

76 65 x Feespace Impulse Response Though wall Impulse Response *Window(Fee space) *Window(Though wall) 4 Amplitude in (V) time, ns (a) 10 x Fee-Space Impulse Resopnse Though Impulse Response 6 4 Amplitude (V) time, ns (b) Figue 3.8: (a) Un-gated time domain signal with window, (b) gated time domain signal

77 Wall (mateial) Paametes Both magnitude and phase infomation ae needed fo accuate chaacteization of walls. The dielectic constant, loss tangent and attenuation constant fo the sample mateials ae calculated. The multiple-pass technique fom [Muq03a] using the one-dimensional seach is employed. The dielectic constant and loss tangent ae obtained fom the insetion tansfe function (See equations 3.17, 3.18, & 3.19). 3.6 Measuement Results This pat of the wok pesents and discusses the esults fo the wall chaacteization expeiments. As mentioned ealie, tansmission and eflection measuements wee conducted on thee diffeent wall mateials including wood, glass and gypsum. Wall paametes namely, dielectic constant, insetion loss, attenuation constant and loss tangent wee extacted fo the given mateials. Results fo both tansmission and eflection measuements indicate a close ageement with each othe, and also with esults found in the liteatue. We futhe consideed effect of multiple walls, including double walls, when we have ai gap between the walls, and the effect of vaying the thickness of this ai gap. Tiple wall scenaios wee also investigated. Repeatability and vaiability analysis wee conducted to evaluate the measuement pecision Tansmission Measuements Results Results obtained fom tansmission measuements fo the magnitude of the insetion tansfe function calculated fom equation (3.) ae shown in Figue 3.9 fo the thee

78 67 mateials. Both the measued aw data of the tansfe function and thei fits ae shown in Figue 3.9 (a) and (b) espectively. Wood shows a highe loss as expected with up to 6 db loss at 18 GHz attibuted to its composition and thickness. We also can obseve the wiggling in the glass cuve which is elated to its small thickness; theefoe eos ae bound to occu. In Figue 3.10, the extacted dielectic constants ae given fo each mateial. Eoneous data have been emoved fom the esults as seen at aound 16 GHz (see Figue 3.13 fo eoneous data). We elate this issue to the spectal limits of the antenna as indicated by the null at aound 15.5 GHz 16 GHz in Figue 3.6(a) and is moe ponounced fo eflection-type measuements as will be seen late in Figue 4.5(a).This means that, at that fequency, vey little signal is tansmitted. This poblem is also eflected in the antenna test esults (see Figue A) by the numeous lage side lobes in the antenna patten stating fom aound 15 GHz. The esults fo the antenna gain also showed a loss of 4 db fom 14 db at 13 GHz to 10 db at 16 GHz. This has manifested on most of the esults including those fom eflection measuements if we obseve. Because of the pesence of noise in the data, the one-dimensional algoithm uses a seach that is bounded (minimum constained seach) to enable it convege. An initial guess is taken fo the dielectic constant fom equation (3.0) and the solution obtained is used as an initial guess fo the next point till all fequency points ae solved. Figue 3.10(b) also shows quadatic fits fo the dielectic constants. The coefficients fo these fits can be found in Table 3.8. All thee mateials have unifom stuctue, theefoe, examining Figue 3.10, we see that the dielectic constants, in the aveage sense; exhibit a deceasing tend with

79 68 Insetion Tansfe Function (Magnitude), db Wood Gypsum Glass Tansmission Fequency, GHz (a) Insetion Tansfe Function (Magnitude), db Wood - fit Gypsum - fit Glass - fit Tansmission Fequency, GHz (b) Figue 3.9: Tansmission insetion tansfe function vesus fequency fo diffeent walls, (a) Measued, (b) Fits to measued data

80 fequency. Glass shows a slightly moe negative slope and this is because dielectic constants tend to be highe at low fequencies. As mentioned ealie, the loss tangent and dielectic constant wee deived fom the insetion tansfe function using the low-loss analysis method (one-dimensional oot seach). In Figue 3.11, the loss tangents ae shown. It is wothy of note that fo most mateials, measuing the loss is moe difficult than measuing the eal pat of the complex pemittivity [Gey90]. The attenuation constant in db/m, is calculated fom equation (.14) whee the unit is in Nepes/m (Np/m) using the following, ( db / m) 0 log( e) ( Np / m) ( Np / m) (3.1) Figue 3.1 shows the attenuation constants fo wood, glass and gypsum boad. This indicates the attenuation of the field stength as it passes though the media. The attenuation constant fo wood inceases oughly linealy with fequency, fom aound 60 db/m at 4 GHz, to 50 db/m at 16 GHz. TABLE 3.8: Coefficients of linea and quadatic fit fo the extacted paametes Insetion tansfe function Dielectic constant (tansmission) Dielectic constant (eflection) Coefficients of linea (Figues 3.8(b), 3.10(b), 3.11(b)) o quadatic fits (Figues 3.9(b), 3.13(b)) to extacted the paametes af + b o af + bf + c, f(ghz) Wood Glass Gypsum a b c a b c a b c Loss tangent i Attenuation constant i i i

81 70 10 Tansmission Diecletic Constant Wood Gypsum Glass Fequency, GHz (a) 10 Tansmission Diecletic Constant Wood - fit Gypsum - fit Glass - fit Fequency, GHz (b) Figue 3.10: Tansmission: dielectic constant vesus fequency fo diffeent walls

82 Tansmission Wood Gypsum Glass Loss tangent Fequency, GHz (a) 0.15 Tansmission Loss tangent Wood - fit Gypsum - fit Glass - fit Fequency, GHz (b) Figue 3.11: Tansmission: Loss tangents vesus fequency fo vaious walls

83 7 Attenuation constant, db/m Tansmission Wood Gypsum Glass Fequency, GHz (a) Attenuation constant, db/m Tansmission Wood - fit Gypsum - fit Glass - fit Fequency, GHz (b) Figue 3.1: Tansmission: attenuation constant fo the diffeent walls

84 73 Insetion tansfe function, db Fit Wood Sample Fequency, GHz Fequency (GHz) (a) (b) Dielctic Constant Fit Wood Sample Insetion tansfe function, db Fit Wood Sample Dielctic Constant Fit Wood Sample Fequency (GHz) Fequency (GHz) (c) (d) Figue 3.13: Showing Eoneous data points fo; (a) Tansmission-insetion tansfe function, (b) Tansmission-dielectic constant, (c) Reflectioninsetion tansfe function, (d) Reflection-dielectic constant

85 Reflection Measuements Results The same model was used to compute the wall paametes fo the eflection measuements. In section 3.3, we indicated that the pocedue fo caying out the eflection expeiments was slightly diffeent fom that of tansmission because both antennas ae collocated on the same side of the wall, and to enable the use of the same model fo chaacteization, the wall thickness is assumed to be twice the oiginal size. This is because the signal has to popagate twice though the wall befoe aiving at the eceive. The dielectic constant is shown fo wood, glass and gypsum, in Figue Calculating the dielectic constant and hence, the loss tangent fo glass fom the eflection measuements was difficult because of the multiple eflections aiving at the eceive at the same time fom the glass suface, the aluminum sheet behind it, and fom intenal slab eflections making time gating difficult. In paticula, a lage pat of the signal is eflected fom the glass suface even befoe passing though the wall. This has also made getting the fist estimate of the dielectic constant fom equation (3.0) using peak-to-peak delay difficult. Thus, the oot seach algoithm will not convege popely leading to false solutions as indicated by the peaks in the dielectic constant fo glass given in Figue 3.14(a). Consequently, the attenuation constant and loss tangent ae also affected as shown in Figue 3.16(c) and (d) espectively.

86 75 Diecletic Constant Reflection false solutions Wood Gypsum Glass Fequency, GHz (a) Diecletic Constant Reflection Wood - fit Gypsum - fit Glass - fit Fequency, GHz (b) Figue 3.14: Reflection: Dielectic constant vesus fequency fo the thee mateials

87 Between Tansmission and Reflection Having mentioned that the same model was used in calculating the dielectic constant fo both tansmission and eflection measuements, it should be noted that in both cases only nomal incidence was consideed. In eflection howeve, the phase of the signal is shifted 180. In Figue 3.15, esults fo tansmission and eflection obtained fo wood sample ae shown. We can obseve a close ageement between the esults fo the two methods. Some of the inconsistencies ae as a esult of souces of eos like the fa-field (see section 3.9) o emaining ai-gap between aluminum and wall sample. Figues 3.16 and 3.17 shows esult compaison fo glass and gypsum espectively. Plots fo insetion tansfe function ae showing its magnitude in db.

88 77 Insetion Tansfe Function, db Wood Tansmission Reflection Dielectic Constant Wood Tansmission Reflection Fequency, GHz (a) (b) Fequency, GHz Attenuation Constant, db/m Wood Tansmission Reflection Fequency, GHz (c) (d) Fequency, GHz Figue 3.15: Compaing tansmission and eflection esults fo wood; (a) insetion tansfe function, (b) dielectic constant, (c) attenuation constant, (d) loss tangent Loss Tangent Wood Tansmission Reflection

89 78 Insetion Tansfe Function, db glass Tansmission Reflection Dielectic Constant glass Tansmission Reflection Fequency, GHz Fequency, GHz (a) (b) Attenuation Constant, db/m glass Tansmission Reflection Loss Tangent glass Tansmission Reflection Fequency, GHz Fequency, GHz (c) (d) Figue 3.16: Compaing tansmission and eflection esults fo glass; (a) insetion tansfe function, (b) dielectic constant, (c) attenuation constant, (d) loss tangent

90 Insetion Tansfe Function, db Tansmission Reflection Gypsum Fequency, GHz Fequency, GHz (a) (b) Dielectic Constant Gypsum Tansmission Reflection 79 Attenuation Constant, db/m Gypsum Tansmission Reflection Loss Tangent Tansmission Reflection Gypsum Fequency, GHz Fequency, GHz (c) (d) Figue 3.17: Compaing tansmission and eflection esults fo a gypsum (a) insetion tansfe function, (b) dielectic constant, (c) attenuation constant, (d) loss tangent

91 Compaison with Liteatue To measue the accuacy of the esults, we compae them to what is obtained in the liteatue. Table 3.9 summaizes compaison between ou data and published esults. These compaisons clealy indicate that the esults fo the dielectic constant of mateials tested ae elatively accuate. Howeve, the accuacy of loss tangent, conductivity o is less cetain. This uncetainty is also widely eflected in epoted data. Fo instance [Sag04] epoted a mean dielectic constant of 4.45 fo gypsum which is entiely diffeent fom.4,.4, and.5 epoted by [Cui01], [Muq05], and [Tes07a] espectively. Also, [Cui01] gave a conductively of 0.35 S/m fo glass, which is significantly diffeent fom 10-1 ohm -1 m -1 given in [Vae88]. (Note that conductivity and loss tangent ae elated by the expession tan whee tan o ). We should howeve, undestand that the same mateials used by diffeent eseaches may vay in composition and sizes, and theefoe diffeences in the esults ae expected. In geneal, we believe ou esults ae elatively accuate even with pesence of cetain limitations (see souces of eo in section 3.9) to ou measuements. It should also be noted that in many cases the insetion loss of a wall is lagely due to eflection and much less due to absoption of the signal, so inaccuacies in the loss tangent does not have a majo impact on path loss evaluation [Muq05].

92 81 TABLE 3.9: Results compaison with liteatue Wall Results Wood Liteatue [Sag04], [Sag05] epoted a mean value of = 3. fo dielectic constant at 8 1 GHz and = 0.31 fo imaginay pat of complex pemittivity coesponding to loss tangent of [Vae88] calculated dielectic constant fo X- Band fequency to be between 3 and 7. Also epoted a = 10-1 ohm -1 /m. Ou Findings Fom Figue 3.10, we ead an aveage value of = 3.0 at 8 1 GHz fo tansmission measuements and an aveage of 3. fom Figue 3.14 fo eflection measuements. Figue 3.11 shows a loss tangent of 0.11 at that fequency ange. In Figue 3.14, ou esult fo dielectic constant using eflection measuements lie within 3 7 at a mean of 3.5. Glass Liteatue [Cui01] measued and epoted a mean value of = 6.06 at 5.8 GHz with a max value of 6.31 fo the dielectic constant. They also measued conductivity to be 0.35 S/m. [Muq05] esults showed value of 6.7 at 5 GHz fo dielectic constant. [Be00] povided a ange fo fom at 3 GHz. [Tes07a] epoted an aveage of 6.6 fo dielectic constant and of coesponding to a loss tangent of in the X Band ange. [Vae88] epoted a ange of 5-10 fo dielectic constant of glass and a = 10-1 ohm -1 /m. [Wil0] epoted of 6.38 fo fequency ange of 7 GHz and loss tangent of 0.06 Ou Findings At 5.8 GHz, Figue 3.16(b) eads a value of = 6.5 and 6.3 fo tansmission and eflection measuements espectively and we have a loss tangent of about 0.01 fom Figue 3.11(a). Figue 3.10(a) eads a value of 6. fo dielectic constant at 5 GHz. Ou esult fo of glass at 3 GHz in Figue 3.14 is 7, which lies in the ange 3 8. We have an aveage of 6.8 fo dielectic constant in the X-Band fom Figue 3.10(a). And in Figue 3.14(a) an aveage = 6.6 in the ange of 9 11 GHz. Figue 3.16(d) shows an aveage loss tangent of 0.09 fo tansmission. Tansmission measuements esults in Figue 3.10(a) indicate a value between in the ange 7 GHz. Gypsum Liteatue In [Cui01], esults show mean =.0 with a max value of.4 fo the dielectic constant. [Ali03] epoted a value of =.3 fo dielectic constant and = 0.03 S/m. [Muq05] epoted =.4 fo dielectic constant at 5 GHz and a loss tangent of [Tes07a] also epoted =.41 and a coesponding loss tangent of Ou Findings In Figue 3.10(a) a value of.7 can be ead fo dielectic constant using tansmission measuement at 5.8 GHz. We also have between 4 GHz and 6 GHz an aveage value of fo the loss tangent fom Figue Again, Figue 3.10 shows =.6 at 5 GHz and Figue 3.11 shows an aveage loss tangent of 0.05 in the X-Band fequency ange.

93 8 3.8 Multiple Walls The main idea behind consideing multiple wall chaacteization was to investigate how effects such as the penetation loss compae with that of a single wall. Since we ae analyzing the wall tansfe function, ou expeiments cascading two(thee) diffeent(same) wall mateials togethe (to make one wall) gives us an insight on whethe having infomation on a single wall will enable us pedict and make decisions about the multiple walls. In this pat of the wok, we deal only with the insetion tansfe function, so we ae able to say how much loss, ove the wide fequency band, is thee in a single wood wall compaed to that of double its size. Futhemoe, in the multiple laye configuations, we compae the poduct of the single laye tansfe functions to the tansfe function of the cascaded multiple layes (oveall tansfe function). This aangement is illustated in Figue 3.1. Anothe aangement investigated is when the double walls ae spaced by an ai-gap of vaying sizes. The insetion loss though double walls spaced 5 cm and 10 cm apat wee investigated Double Laye A section of single wood and double glass walls ae shown in Figue 3.18, Figue 3.18(c) shows a spaced double wall, while Figue 3.19 shows magnitude of the eceived signal, insetion tansfe function and impulse esponse fo single and double wood walls obtained fom tansmission measuements. The loss is highe at highe fequencies with up to 4.5 db diffeence between single and double walls. Fo the effect of inte-wall spacing, no significant diffeence was obseved fo all the wall mateials. Figue 3.0 pesents the esults fo wood.

94 83 (b) Single wood wall (a) Double glass wall ai gap (c) spaced wall: wood and gypsum (d) Thee laye: glass-woodglass Figue 3.18: Wall configuations