Home

MASTER`S THESIS

1. Select dataformat Create channel 2 2 Load calibration set 4 Set center frequenct span and number of points Create trace 3 as Copol2 S43 Create trace 4 as Xpol2 S23 isplay trace in window No window Yes isplay trace in Yes v Ne No window v Create window 1 Create window 4 and feed trac 1 Yes Yes and feed trac 4 v v Create window 2 Create window 3 J and feed trac 2 and feed trac 3 Window mode to Window mode to magnitude J magnitude Window mode to Window mode to magnitude magnitude v Set sweep mode to continuous Display traces i windows Yes v Set window update on End program J Appendix 5 Selectable input parameters for the measurement control software parameter explanation value title name of the measurement string measurement_point name of the measurement point string stepper 1 Rx stepper as tcpip object tcpip object stepper2 Ry stepper as tcpip object tcpip object stepper3 Tx stepper as tcpip object tcpip object stepper4 Ty stepper as tcpip object tcpip object vna VNA as tcpip object tcpip object fo measurement center freguency freguency in Hz BW meas 10 bandwidth in Hz measurement s l
2. ES OULUN YLIOPISTO UNIVERSITY of OULU DEGREE PROGRAMME IN ELECTRICAL ENGINEERING MASTER S THESIS VIRTUAL ANTENNA ARRAY BASED MIMO RADIO CHANNEL MEASUREMENT SYSTEM AT 10 GHZ Author Nuutti Tervo Supervisor Risto Vuohtoniemi Second Examiner Juha Pekka M kel Technical advisor Veikko Hovinen October 2014 Tervo N 2014 Virtual Antenna Array Based MIMO Radio Channel Measure ment System at 10 GHz University of Oulu Department of Communications Engi neering Degree Programme in Electrical Engineering Master s Thesis 58 p ABSTRACT In this thesis a 10 GHz multiple input multiple output radio channel measure ment system using four port vector network analyzer and virtual antenna arrays in both transmitter and receiver ends is presented The channel measurement sys tem measures each single antenna channel separately The radio propagation en vironment is assumed to be static during the recordings As an antenna element a dual polarized patch antenna with two feeding ports is used Linear stages and programmable stepper motors are utilized to build an X Y gantry to move the an tenna element in a plane The stepper motors and the vector network analyzer are controlled by the same measurement control software The basic principles of the control software are also presented along the measurement system Three channel measurement scenarios and their initial results are presented to verify the system operation and to dem
3. When measuring large virtual antenna arrays it is unpractical to save all of the results in the same matrix The size of the result matrices affect also to the speed of the mea surement in case of the data saving from VNA to the MATLAB workspace during the measurement Too large matrices slows down the post processing and are clumsy to handle On the other hand large number of small matrices or vectors are also unprac tical to handle In MATLAB we basically have two possible formats for large block matrices One is to use structures and the other is to use cell arrays There are some variations in the saving speeds when comparing these two to each other The selected data format is to use cell arrays with varying variable names The measurement soft ware should allow user to change the data format to larger or smaller configuration if necessary 38 The selected data format is presented in Figure 9 During the measurements we cre ate cell arrays for each combination of npx ry ntx with dynamic variable names Here npx and ngy are the number of RX antennas to x and y directions respectively Nyx and ny are the same characteristics for the TX antenna array The dimension of these cell matrices are Npoints NRy 2 Each element of the cell array is 32 bit long complex number representing one recorded S parameter value When we take into account also the cross polarization products the result is two group of matrices with variable name descr
4. 2 RADIO CHANNEL CHARACTERISTICS The radio wave undergoes many physical phenomena caused by the radio channel before reaching the RX These phenomena depends on the wave properties such as frequency as well as the properties of the propagation environment In a multipath channel the wave propagates trough several different paths between the TX and the RX causing fading and shadowing to the received signal In this chapter we will present the basic theory of radio wave propagation phenomena and the radio channels respectively 3 2 1 Electromagnetic propagation Understanding the behavior of electromagnetic waves is needed in order to understand the theory behind existing radio channel models An electromagnetic wave can be de scribed by the Maxwell s equations presented in 10 by using electrical or magnetic fields Usually the electrical field as a function of time or direction of propagation is used to describe the wave behavior The nature of the electromagnetic wave observed in very near to the source is different compared to the nature of the same wave after it has been traveled over relatively large distance The radiating field region of an an tenna can be divided into radiating near field and far field regions These regions are also called as Freshnell and Fraunhoffer regions respectively In the radiating near field the field attenuation is stronger than in the far field The distance after the field is referred to be far fie
5. Wireless Communications staff for feeling me welcome to work here I would also like to take this opportunity to thank my family and friends for all the support I have enjoyed during my University studies Especially I would like to thank my brothers Valtteri and Oskari for the everyday discussion help and support during the studies as well as during the time spend with this thesis The special thanks goes to my girlfriend Jenny for the sincere support and understanding towards my passion for science for all the years we have been together Oulu Finland October 24 2014 Nuutti Tervo LIST OF ABBREVIATIONS AND SYMBOLS AC AoA AoD DC DoA DoD GO GPS GTD GUI IDFT IF IFFT IP KED LAN LNA LOS MCode MIMO NLOS PA PDF RF RMS RT RX SCPI S parameters TCP IP TX UTD VNA 3D 5G ay a2 B Be Bp Bir B meas bi be alternating current angle of arrival angle of departure direct current direction of arrival direction of departure geometrical optics global positioning system geometrical theory of diffraction graphical user interface inverse discrete Fourier transform intermediate frequency inverse fast Fourier transform internet protocol knife edge diffraction local area network low noise amplifier line of sight machine code multiple input multiple output non line of Sight power amplifier probability density function radio frequency root mean square ray tracing receiver standard com
6. 0 36 and 1 5 in line of sight LOS environment respectively The path gain in vertically and horizontally polarized transmissions are stated to be almost the same in most of the measured environments In some special corridor rooms the path loss for horizontally polarized wave is stated to be significantly large The measurement campaign presented in the paper 9 is conducted in the university building in 3 dif ferent corridor rooms and 2 different halls and the both LOS and NLOS scenarios are considered The approach to the channel modeling and the measurements presented in the paper 9 is almost similar compared with the one that we had However using VNA with virtual antenna arrays instead of MIMO channel sounder sets its own limitations for the measurement system On the other hand it also gives a number of advantages compared to channel sounder This thesis is organized as follows In Chapter 2 some background theory related to the electromagnetic waves and radio channel models are carried out Chapter 3 introduces the used measurement setup and how it was built In Chapter 4 some channel measurement scenarios and initial results are introduced In Chapter 5 the 12 measurement system is evaluated and some improvements to the system are proposed The conclusion is drawn in the Chapter 6 Some flow charts of the measurement control software and the inputs defined by the user are found from the Appendices at the end of this thesis 13
7. affects directly to its electromagnetic prop erties For the nonorthogonal wave incidence with respect to the surface of the wall the penetration loss was also observed to vary a lot For making a theoretical model for the wall we should study better the real structure of the wall 54 6 SUMMARY The goal of this thesis was to build a MIMO channel measurement system by using four port vector network analyzer and virtual antenna arrays in both TX and RX ends The built system measures each single antenna channel separately The radio propagation environment was assumed to be static during the recordings A dual polarized patch antenna with two feeding ports was used as an antenna element The linear stages and programmable stepper motors were utilized to build an XY gantry to move the antenna element in one plane All of the measurement equipment were controlled by the same centralized measurement control software The system was able to measure antenna configurations up to 26 antenna elements in the horizontal and 25 elements in the vertical direction respectively assuming a frequency of 10 GHz and antenna spacing of half of a wavelength The measurement time for one single antenna channel was 1 9 s in the horizontal direction and 0 9 s in the vertical direction respectively The longer measurement time for the horizontal movement was caused by the external waiting to prevent the vibrations of the antenna during the measurement Three channe
8. be measured Each VNA vendor has their own calibration algorithms For the radio channel mea surements the best calibration method offered by the used device is one way trough calibration However this option was not supported by the used calibration kit and thus was not able to be used Unknown trough open short match UOSM calibration were used instead 27 3 3 Virtual antenna array When measuring the radio channel with the VNA the number of VNA ports is limiting the number of simultaneous antennas that could be used To get still a MIMO channel we move the antennas between the sweeps to get a virtual antenna array The amount of movement between the measurement points depends on the antenna spacing of the used antenna array Because of the channel correlation usually antenna spacing of half of the wavelength is used 13 There are several ways to perform the antenna movement between the measure ment point Especially in high frequencies such as 10 GHz the misplacement of the antennas may cause significant inaccuracy to the measurements When measuring large antenna configurations the best way is to move antennas automatically by using robotics Programmable devices such as robotic arms or stepper motors can be used for the movement We built two XY gantries to move the TX and the RX antennas to the vertical and the horizontal directions by using programmable stepper motors The motors where chosen such that they could be controlle
9. be used to route the device specific commands into each device separately The stands made of aluminium were built for RX and TX antenna XY gantries The antenna mast was made of plastic and with the stands it allows to use antenna heights from 1 05 to 2 0 m in the RX end and from 1 50 to 2 80 m in the TX end Naturally the antenna heights can be increased by placing the stands on some external structures The whole measurement system in operation is presented in Figure 6 3 5 Measurement control software For controlling the devices MATLAB was used as the programming language MAT LAB instrument control toolbox offers several options to control external devices 34 LMDCE572 Stepper motors mn VT16061C Mechanical switch To the main F voltage supply a ea a E Ro 1 1 Supplyivoltage Toipositive L L vpc To negative limit swicth 1 imit switch x EE To both limit In Programmable UHUDA switches multif k amp tion SAAJA J 30 VDC 12 VDC interface 7 Voltage supply Voltage 7 Ethernet b interfdce supply Figure 5 Wiring of the stepper motors and limit switches TCP IP communication protocol over the Ethernet bus is the best option for our pur poses because it is easy to use and is supported by all of the devices used for the mea surement system The communication protocol takes care of the data flow between the devices and MATLAB The Instrument control
10. can do the same The penetration loss of the wall is presented in Figure 17 As it can be seen from the figure the penetration loss of the wall varies a lot depending on the antenna location The variations on the penetration loss seem to be periodical with respect to the antenna index The refraction loss of the wall with respect to the refraction angle 0 is presented in Figure 18 Both of the RX locations are collected into same figure As it was in 47 the direct penetration loss with orthogonal wall incidence here the results varies also between the different antenna locations The chosen measurement parameters causes following problem In the measurement environment there is multipath propagation causing affecting to the recorded impulse responses The measurement bandwidth used 500 MHz gives only 60 cm resolution for to distinguish multipaths from each other In this environment there is too much multipath propagation within the resolution bandwidth giving us the a sum of multi paths in each recorded frequency point When using large bandwidth the dispersion of different frequencies might also be a problem in the measurements However same applies for the real communications trough the wall There might be several reasons why the wall penetration loss varies a lot when changing the part of the wall under test Thus plenty of sweeps in each antenna positions would be needed to measure the wall penetration loss by using this method Si
11. mechanics For example the antenna shaking after each movement af fects to the accuracy of the antenna location during the measurement In the systems we had to wait some amount of time for the antenna mast to settle This waiting time decreased the speed of the measurements The shaking were observed to be more significant when moving the antenna in the horizontal direction compared to the ver tical direction This is why the antenna movement pattern where chosen such that the horizontal movement is minimized The antenna movement patterns are presented in Figure 10 4 RX Behind TX Behind X Axis Origin Origin X Axis o x n o 4 a a gt Figure 10 The movement patterns of the antennas in X Y gantries Along the verification measurements we measured also the speed performance of the system The actual measurement speed is proportional to the selected antenna configuration and the VNA settings For the total measurement speed per measurement point is included the antenna movement the VNA sweep and the data transfer from the VNA to the laptop via Ethernet cable The measured time per measurement point and an example about the total measurement times including time for the stepper motor calibration are presented in Table 8 The used measurement settings for VNA were presented in the Table 6 Because of the external waiting time of one second added to the x direction movement to wait the antenna shaking to settle we have one second highe
12. option could be quite easily implemented and will be available to the next version of the measurement system One way to cancel the effect of antennas out from the measurement results could be to use fixed reference antenna in RX end The measured reference path loss could be for example used as a reference value for the data measured in different antenna loca tions However the usage of reference antenna can be replaced by external reference measurements This however can only be done if the channel is assumed to be time invariant and thus the same in real measurement and reference measurement 5 3 About the measurements The verification measurements were done mostly to verify the system operation and give an reference and test data for the DoA and DoD algorithms which will be part of the measurement system in the final product In these measurements it was clearly seen that the reflection rotates the polarization vector and thus some of the power of the reflected component is leaked to the other polarization domain This may be significant for example when considering separating users in the polarization domain It is clear that the reflection is one of the most dominating propagation phenomena at the indoor environment The anechoic chamber scenario gave a very good reference for the further measurements The results implied that the system does not cause internal spurious response for the results The test measurements performed in the cla
13. responses The channel frequency response is used to describe the channel behavior as a function of frequency When multiplying transmitted signal spectrum X f by the frequency response H f we get the received signal spectrum Y f In frequency domain this can be expressed as 20 Y f A f X f 25 In time domain the channel is described by the channel impulse response h t The impulse response is the inverse Fourier transform of the frequency response hence the received signal y t in time domain can be expressed as y t h t x t 26 where x t is the transmitted signal and x denotes the time domain convolution of the signals 20 In the radio channel modeling the power of different signal paths is often interest ing Power delay profile PDP of the channel is defined by the impulse response representing the powers received at each time instant It can be written as 3 PDP 10 27 2 3 2 Delay spread doppler spread and angular spread There are few parameters to describe the properties of the multipath channel Delay spread is a measure of the multipath richness of the propagation channel It is defined by the PDP as being the time difference between the earliest and the latest significant multipath component seen in the received signal PDP In LOS channel the earliest component is the LOS propagated component of the signal Mean delay spread and root mean sguare RMS delay spread are parameters describing
14. the TX and the RX ends In that case one pair between the RX and the TX antennas represents one single input sin gle output SISO channel in the system Thus MIMO system has several subchannels that can be combined to one MIMO channel matrix The matrix consists the subchan nel coefficients from each TX antenna to all the RX antennas If we denote h being the channel coefficient from the TX antenna i to the RX antenna j the MIMO channel matrix can be written as 3 hia hio oo ee hoi h22 STN ho mr 33 Ang hng as hnr ny In order to have an advantage on using multiple antennas the channels must be as uncorrelated as possible In case of correlated channel the rank of the channel matrix is low Furthermore this means that the number of distinguishable multipaths in the channel is low and hence MIMO cannot be successfully employed for beamforming The best advantage of using multiple antennas is exceeded when the channel is as rich as possible meaning high rank of the channel matrix H Especially in the future telecommunication systems the number of antennas are increased in order to exploit better the multipath richness of the channel If the TX and RX are equipped with very large number of antennas the system is called massive MIMO systems 1 22 3 RADIO CHANNEL MEASUREMENT SETUP When measuring the MIMO channel the measurements with a good accuracy with respect to the dynamic range of the system may take a significant amo
15. the deviation of the received signal path delays The mean delay spread is defined as 3 N Ps TPDP 7 d7 T yy 28 Jo PPP where 7 is the delay at each multipath component The RMS delay spread is defined as 3 J T F EPDP r dr 29 107 Jo PDP r dr i The coherence bandwidth of the channel can be defined as the Fourier transform of the delay spread The coherence bandwidth defines the bandwidth which the chan nel stays constant with respect to freguency Roughly it can be approximated by the inverse of the mean delay spread 3 Bo 30 3 20 If the TX the RX or the environment is in motion over time the transmitted signal experiences Doppler effect Thus in the received spectrum Doppler spectrum sev eral frequency components may be seen even if only one was transmitted The spread of the frequencies in the RX caused by the Doppler effect is called as Doppler spread The width of the Doppler spectrum is called as Doppler bandwidth Bp 3 17 Chan nel coherence time Tc is inversely proportional to the Doppler bandwidth and it can be written as 3 In a multipath channel the multipath propagated components leaves from the TX antenna and arrives to RX antenna in some specific angle with respect to some refer ence direction which is usually the direct link path between the TX and RX These angles are called as angle of departure AoD and angle of arrival AoA respectively In three dimens
16. the selected linear stages model KK6005P 600A 1 F4 nominal width 60 mm ballscrew lead 5mm rail length 600 mm resolution 5 mm rev maximum speed 340 mm s repeatability 0 003 mm accuracy 0 020 mm running parallelism 0 010 mm starting torque 15 N cm limit switches can also be used to define the origin for the steppers when the stepper is switched on Two kinds of limit switches were used in the system Hard mechanical limiters were placed to the boundaries of the linear stages to prevent the motor to force the carriage towards the stage end These switches were used in both positive and negative ends of the stages The switches were connected to the circuit in such a way that the main current was switched off in case of hard stop Inductive proximity sensors were used as soft limit switches These sensors were wired to the general purpose programmable interface of the steppers in such a way that they change register values of the stepper in case of carriage coming close to the switch inductive connection The changed register values can be used in the program to define interrupt routines that the motor stops the movement in case of reaching the limit but stays programmable which was not possible in case of the hard stop This allows the software to use these registers as inputs in programming The proximity sensors were also used to define the origo position for the stepper mo tors Furthermore the
17. 10 5 0 5 10 15 20 25 30 Refraction angle deg Figure 18 The refraction loss of the wall 49 4 5 Diffraction around a building corner The measurement system can be used also for diffraction measurement When moving virtual antennas along the x axis by distance of fractions of the wavelength we can fast perform several measurement and see the effect of diffraction for the recorded impulse response Also the changes in polarization due to the diffraction can be constructed from the measurement results The results and further analysis of this measurement are published in 35 in the first international conference on 5G for ubiquitous connectivity SGU in November 26 28 2014 50 5 DISCUSSION Accuracy and applicability for the different kind of measurements are just an examples about the characteristics describing how we have succeeded in building the system In the final measurement system there are many sources of inaccuracy Hence there are many improvements that could be made for the system to increase the speed per formance dynamic range and applicability for different kind of measurements The verification measurements showed that we should always plan well the measurements in order to be able to calculate specific measurement settings for each measurements This was taken into account in the programming phase by parameterizing the input parameters such that they could be chosen by the user In this chapter we will evaluat
18. 2 3 Thus for the reflected path the polarization vector changes and we have a stronger signal for the cross polarizations 4 and S than for co polarization 43 Figure 11 Overview of the anechoic chamber where the reference measurements were performed 50 LOS path 60 K made avaat 521 eet 43 Reflection from 941 k the floor sd 595 70 wae ce thet ia E 80 ki g Bei gab EEF ETE EVET ARS 353300 bel bad bee ETETETT TENE a vs i lt RX 0303 a TX 0303 90 PK 3 ein Wane 100 I n Iv ara ix k J i i N A k aa wy hi i i aes 110 JA 1 i y ou ao SA a ran r N Eya 0 20 40 60 80 100 Delay ns Figure 12 Sample of the impulse responses measured in the anechoic chamber 4 3 Test measurements in classroom The system is mostly made for MIMO measurements i e to measure rich multipath propagation environment A classroom with chairs tables and window blinds gives an environment where multiple reflectors and scatterers are present To test the DoA and the DoD estimation algorithms we considered to measure LOS channel with one measurement where RX was rotated to 45 angle with respect to direct link chord The classroom measurement was performed in Electrical Engineering building of the University of Oulu in the lecture room TS128 The layout of the room is presented in Figure 13a The room was chosen such that it represents the usual lecture room with whit
19. 8 Lye 6 9 20log y 0 12 1 v 0 1 21 Instead of modeling the diffraction by wedges by KED absorbing screen can also be used to model diffraction For a plane wave incidence the absorbing screen approach 18 gives us a geometrical theory of diffraction GTD diffraction coefficient with respect to 04 as Ba z as 0a 2n G On a The wavefront after diffraction is astigmatic because there is some caustic in the edge GTD and uniform theory of diffraction UTD defines also other similar coefficients for the diffraction As well as in the reflection and refraction the polarization vector of the wave may be changed due to the diffraction 16 12 22 2 2 5 Fading and shadowing Fading is defined as the deviation in radio channel causing attenuation to the transmit ted wave In a rich multipath channel the transmitted wave propagates trough many different propagation paths causing deviation to the received signal All multipaths are summed in the RX by the superposition principle Fading can occur in time space and frequency domain and it can be modeled statistically Thus fading is a random process whose quantities depends on the propagation environment and mobility in the channel 3 In Rician fading Rice distribution is used to describe the randomness of the channel Rician fading is used when one of the received multipath components are relatively strong compared to others Typical case of Rician fad
20. Pr PrGrGa 6 4rd where Pr is the transmitted power d is the direct distance between the antennas Gr and Gr are the TX and the RX amplifications respectively The path loss L experi enced by the wave can be written as 11 gt 1 aL Ga N Lis 7 2 2 2 Plane wave in the medium The dielectric and the magnetic properties of the medium can be described by param eters u permeability and permittivity The permittivity can be complex when the imaginary part e describes the dielectric losses caused by the medium Thus permit tivity can be represented as ole Jer 8 where is the permittivity of the vacuum and is the real part of the relative permit tivity 14 The dielectric properties of the propagation medium affects to the propagation loss experienced by the radio wave The loss caused by the medium can be specified by the loss tangent of the medium N o Geo and _ 9 where o is the electrical conductivity and w is the angular frequency 14 Complex propagation constant of the medium is represented as o Yp Jwy He 1 o J 10 16 where a is the propagation coefficient and is the phase coefficient 15 The attenuation of the wave is exponential with respect to complex propagation con stant yp The attenuation L d of the planar wave can be represented as Bad Ser 11 where d is the distance which the wave has been propagated in t
21. Sow tloda ables sn dare 50 5 2 Improvements proposed to the system 0 0 cee ee eee 51 5 3 About the MeASUTEMEN S ta 103 Aaa aa deel ka alaosaan oes 52 s SUMMARY ceuin naio a Mus A a E E A R tS 54 i REFERENCES 6h d pie ss I Ius Varaa ae gies Kask a eee 55 s APPENDICES sci Gs ing Tat cee A N eke AE a oe gee 58 FOREWORD This thesis has been carried out as a part of the 5G radio access solutions to 10 GHz and beyond frequency bands 5G to 10G project The project is supported by Broad com Communications Finland Oy Elektrobit Wireless Communications Oy Huawei Technologies Oy Finland Co Ltd Nokia Networks Oy and Finnish funding agency for innovation Tekes I would like to take this opportunity to thank all the project partners for their competence for this work I would also like to thank my technical advisor M Sc Tech Veikko Hovinen for the great ideas leading to this work I am graceful for the thesis examiners Lic Sc Tech Risto Vuohtoiemi and D Sc Tech Juha Pekka M kel for reading the thesis and advising me in writing I would also like take this opportunity to thank Professor Matti Latva aho for the trust to my abilities to work here D Sc Tech Marko Sonkki for helping me in the beginning of my work Anssi Rimpil inen for implementing the most of the mechanics and M Sc Tech Claudio Ferreira Dias for the great technical discussion during the work Furthermore I would like to thank all the Centre for
22. VNA such as frequency region transmit power number of frequency points etc Finally the VNA is put to a mode to be ready for performing the sweep The flow chart of the VNA initialization is presented in the Appendix 4 Another program we implemented was the checking program for ensuring that the parameters for the VNA set are the same as defined by the user This is simply done by reading the settings from the VNA and comparing them to the specified ones The third program was to perform the actual sweeps to get the desired S parameters The program makes two sweeps and saves the measurement data into a vector in the raw format the VNA provides them The format for VNA data is such that the odd entries are real parts and the even ones are imaginary parts respectively Both traces 21 and S41 Or S23 and 543 are given in the same vector by one after another The mode for the sweep is to make the sweep first and then wait the sweep to be ready before making another sweep The saved trace vector is read into MATLAB in binary format and the traces are separated to two different cell matrices in the main program The flow chart of the VNA sweep program is presented in Figure 8 Start Read traces as End Clear object buffers Sweep channel 1 Sweep channel 2 program binary numbers program Figure 8 The flow chart of the MATLAB subroutine commands VNA to perform a sweep and transfer it from VNA to MATLAB vector 3 5 3 Data format
23. World Scientific Publishing Company Incorporated 23 Chu E 2012 Discrete and Continuous Fourier Transforms Analysis Applica tions and Fast Algorithms Taylor amp Francis 24 Miteq LCN 0812 amplifier datasheet accessed 13 10 2014 URL vurlihttps www miteg com viewmodel php model LCN 0812 25 Hittite HMC C026 amplifier datasheet accessed 13 10 2014 URL url http www hittite com content documents data_ sheet hmc c026 pdf 26 R amp S ZV Z5x electonic calibration kit datasheet accessed 13 10 2014 URL Nurlihttp rosenkranz elektronik de shop datenblaetter R SzZV Z25x datasheet pdf 27 Understanding the VNA calibration Anritsu accessed 13 10 2014 URL url http anlage umd edu Anritsu_ understanding vna calibration pdf 28 Lexium MDrive Ethernet TCP IP products LMDxE N57 and LMDxE N85 Product hardware manual accessed 25 8 2014 URL http motion schneider electric com lmd downloads literature LMDE pdf 29 Programming and Software Reference for Lexium MCode and Lex ium Mcode TCP accessed 25 8 2014 URL http motion schneider electric com lmd downloads literature MCode_LMD pdf 30 31 32 33 34 35 57 Hiwin KK linear stage dataheet accessed 25 8 2014 URL http hiwin com pdf 1ls Single 20Axis 20Robot_K02TE02 0701_311 pdf Panasonic GX F12A P inductive sensor datasheet accessed 25 8 2014 URL http www farn
24. a oa SHRM ae one das ee ele 26 3 2 5 Link budget and external amplifiers 0 26 3 2 6 Calibration of the VNA y420555 escawes pivws ae pene eee ees 28 Sad Vinal antenna ARa yst 2004006 UR Rs 2 2030 Meia U A ths FX s 29 3 3 1 Antenna characteristics 644 2658 Ue bees Chobe GS Ke Se 29 3 3 2 Dual polarized patch antenna lt 3 00s fie L aate da on 30 3 3 3 es VRAIN cise eek K TKI AUTAT KURSSIT WA Bea care Aiken 31 3 4 Wiring of the measurement system 2 0 0 e eee ee ee eee 33 3 5 Measurement control software 0 0 eee eee eee eee ee 33 3 5 1 Controlling the stepper motors 00 ce eee eee ee 35 3 5 2 Controlling the VNA es kasan essa Joe ASIa KIN ews 36 3 923 Data format A Sen sa daa Kumasi ln kaitaletta elk Bidets ges oe elds 37 BA A TROT COMUNE taa aan lB ct EN Ak AD aes 38 Soe oes User interface cor Aig S ESA EKKE Oh ER OOH oe Ren 39 MEASUREMENT SCENARIOS AND RESULTS 025 5 40 4 1 Used measurement settings and the system speed performance 40 4 2 Verification measurements in anechoic chamber 4 41 4 3 Test measurements in classroom s soon nnen 44 4 4 Wall penetration loss measurements 0 0 cee eee eee eee 45 4 5 Diffraction around a building corner 2 0 0 cee eee eee ee 49 gt DISCUSSION lt a aita KKT GN mena 54 STS eu eee se 50 5 1 Evaluation of the sySteni 2 aecbs dao veleadd
25. and its own RX noise figure N Fyna Thus the VNA s own noise figure affects to the noise power in the VNA As it can be seen from the Eguation 40 the thermal noise in the system is proportional to the bandwidth but not to the actual frequency The sensitivity of the VNA i e the smallest signal that the device can detect is limited by the noise floor of the VNA 5 When using VNA there are basically two ways to increase the dynamic range of the measurement system after all the input power available from VNA is used We can make multiple sweeps and use averaging or we can use narrower IF bandwidth as shown in the Section 3 2 4 Both methods increases the measurement time signif icantly It is often said that the effect to the measurement time and dynamic range is roughly the same no matter which method was used 5 However when using averag ing we consider many channel realizations and thus measure the statistical properties of the channel If the channel is assumed to be fixed during the measurements it is bet ter to use narrower IF bandwidth to increase the dynamic range By using this method we loose the statistics but we will have less measurement data to handle 5 3 2 5 Link budget and external amplifiers In case of long link distances the system requires also long cables to connect the antennas to the VNA ports The signal attenuation in cables may grow significantly 27 large and hence reduce the dynamic range of the meas
26. as the angle of the refracted wave the propagation angle can be calculated by the Snell s law for refraction ey 14 sind ea where and ez are the permittivities of first and second propagation medium 14 In case of orthogonal incidence the transmission coefficients of the wave can be expressed as 16 14 Ti 1 R andT 1 R 15 If the incidence is not orthogonal i e 0 90 the transmission coefficients can be written as 2 A cos 6 E 1201 amp 4 2 sin 0i 2 cos 0 2 6 Tj and T Vi sin 6 cos 6 16 17 2 2 4 Plane wave in rough surfaces and diffracting edges In scattering the small particles along the propagation path absorbs some energy and radiates it again to the around space while acting as small antennas by themselves When there are many of these particles along the propagation path the scattering effect can be significant and cause fading to the received signal For example clouds and bushes are just an examples about obstacles causing scattering Also rough surfaces whose roughness is close to the wavelength causes scattering For the scattering there exist several models and theorems which are not presented here Generally speaking we can note that the effect of scattering to the radio wave is random and hence must be modeled statistically 16 When some obstacle comes inside the first Fresnels zone the wave is diffracted from the edge of th
27. at the indoor measurements base station antennas are often planar arrays which supports our implementation Furthermore the used antennas are not able to receive signals from the backside beam One proportion is that we could use a programmable rotary table or a herringbone gear to rotate the XY gantry This 52 would also give an opportunity to measure to different directions even with highly directional antennas which was not possible with the current system However this is not needed if we would use omnidirectional dipole antenna in both ends For some applications dipole antennas could be a good choice but for example in diffraction or wall penetration loss measurements it is better to use directional antennas Off course it is easy to replace the current antennas by different ones if needed One possibility is to insert a small rotary table or other rounding mechanics on top of the build XY gantry and insert also a external linear stage on top of the rotary unit to move the antenna vertically The rotator and the vertical linear stage would then be used to represent virtual cylindrical antenna array and the XY gantry could be used to move the whole antenna array This would also make it possible to make virtual 3 dimensional antenna arrays During the measurements we also faced a problem in setting the origin of the an tenna array It would be useful if the origin could be freely chosen into the XY gantry for both end separately This
28. back from the wall causing attenuation to the trough propagated wave The more we increase the angle from the orthogonal incidence the more of the power of the wave is reflected back from the wall surface The theory for reflection and refraction was presented in Section 2 2 2 The penetration loss of the concrete wall were measured in the two different RX locations in 3 different antenna configurations Also LNA was included into the mea surement chain for testing the effect of LNA to the dynamic range of the system The overview of the measurement setup and environment in seen in the Figure 15a Also a reference measurement with a distance of 2 75 meters was performed to cancel the free space loss off from the results In the reference measurement the distance of the TX and RX antennas was set so that it corresponded the free space distance that the antennas had in the wall measurement case thus the wall thickness was subtracted from the results 1 29 m gt b Figure 15 a The wall penetration loss measurement overview and b TX and c RX units in operation respectively The analysis of the results can be divided into two parts At first we take all the direct orthogonal with respect to the wall surface paths between the RX and the TX and calculate the wall penetration loss by compensating the free space loss out from the results channel by channel For all the used antenna configurations in the both RX locations we
29. d the antenna configurations used nowadays This concept is often called as massive MIMO were transmitter TX and receiver RX could be equipped with hundreds or even thousands of antennas 1 Because of the limited frequency spectrum many of the future fifth generation 5G mobile communi cation applications will use higher frequencies for the communications approaching to the millimeter wave region High frequencies allows to use larger bandwidth making it possible to offer higher data rates for the users in the future Also because of higher path loss the high frequencies allows the telecommunication systems to use smaller cells and thus decrease the reuse distances In order to perform reliable link budged calculations and be able to ensure the availability the need of new reliable channel models is undisputed 2 3 There exists only few good ways to measure the MIMO channel reliably In MIMO measurements radio channel sounders are often used However there are few draw backs limiting the usage of the channel sounders such as high prize and synchroniza tion problems The drawback in most existing systems is that they does not take into account the correlation between antennas since the measurements are not performed simultaneously between all antennas which is often the case in real telecommunication systems However if the measured channel is assumed to be constant with respect to time the MIMO channel model can be constructed by m
30. d trough MATLAB along with the VNA 3 3 1 Antenna characteristics To be able to distinguish the effect of the propagation channel itself from the measured data the effect of the antennas should be compensated off from the data Since in reality antennas are not ideal components all of the power fed into the antenna is not necessarily radiated into the space The impedance of an antenna is matched into the impedance of the signal source Thus the antenna impedance is wanted to be as close as possible to 50 Q over the wanted frequency bandwidth in order to radiate well The matching of the antenna can be specified by the reflection coefficient S11 13 Reflection efficiency of the antenna takes into account the mismatch between the transmission line and the antenna The reflection efficiency can be defined as 13 er 1 yv 1519 48 30 Radiation efficiency of the antenna takes into account the conduction and dielectric losses of the antenna The radiation efficiency can be defined as R E Ru R where R and R are the loss and radiation resistances respectively 13 The total an tenna efficiency takes into account the losses at the input terminals within the structure of the antenna Thus the total antenna efficiency can be written as 13 eo cd 50 49 cd Directivity of the antenna D 0 amp is defined as the ratio of the radiation intensity in a given direction from the antenna to the radiat
31. ded dynamic range and the measurement speed is needed and the overall performance has to be optimize with respect the desired property 5 4 6 There exists many references describing virtual antenna array based channel mea surement systems Various strategies and equipments are used to move the antenna element between the antenna positions However there exits huge variations in mea 11 surement speed and accuracy between the existing systems Many of the existing sys tems uses a rotator or an XY gantry or both of them Using the rotator cylindrical arrays can be made and the array is also capable to see to the backside beam Ro tator is used for example in 4 which represents capacity measurements using large antenna arrays Planar antenna array configurations with a XY gantry are used in 6 and 7 describing channel measurements in frequencies 2 4 GHz and around 60 GHz In paper 6 optical fibre is used to degrease the cable loss in VNA based system There has been made some research and channel models about the indoor radio channels on millimeter wave region However the most of the research and channel models focuses on higher or lower frequencies than 10 GHz The measurements per formed in the higher frequencies are mostly focused to 17 GHz 28 GHz 38 GHz and 60 GHz Especially in 60 GHz there are large unallocated frequency bands which some applications of the future telecommunications systems could use Many indoor measurem
32. e the measurement system and propose a number of improvements for the system which of many will be implemented to the system in near future 5 1 Evaluation of the system When evaluating the measurement system many sources of inaccuracy must be taken into account As discussed previously in the Section 3 1 the dynamic range available and the total measurement time are inversely proportional to each other The optimiza tion of the parameters affecting to the measurement time and dynamic range should always be done with respect to the current measurement scenario This is why many of the measurement settings such as IF bandwidth of the VNA source power of the VNA ports and the maximum speeds of the stepper motors are parameterized in the measurement control software By choosing the parameters correctly the user can op timize the measurement time with respect to the dynamic range requirements needed in each measurement scenario individually Possibility to add the external amplifiers to the measurement chain were wanted to be made as an option This is because they are not necessarily needed in every case One drawback of the measurement setup is the antenna shaking especially in hori zontal movement The shaking was estimated to be even several millimeters at worst which could mean significant inaccuracy in location compared to the wavelength at 10 GHz The antenna shaking was compensated by making better acceleration and deceleration ramps fo
33. e the usage of the TX amplifier is limited by the maximum power that is allowed to be used accord ing to the radio permission Also the power performance of the VNA must be taken into account such that the overall transmit power does not violate the radio permission The idea is to first maximize the transmitted power and then minimize the RX noise The RF block chart of the measurement system including the external amplifiers is presented in Figure 3 28 VNA transmitter W osha RF cable Radio channel VNA receiver RF cable Figure 3 Proposed measurement system including the amplifiers We propose two types of amplifiers that can mostly be used to compensate the loss caused by long RF cables When using all four RF ports for two orthogonal polariza tions two pieces of both amplifier types are needed The attenuation of the RF cables used was approximately 0 9 dB m As discussed before an external power amplifier should be used in such a way that the radio permission is not violated The proposed TX amplifier is HMC C026 manufactured by Hittite Microwave Corporation The am plifier gives approximately 29 dB gain at 10 GHz with 1dB compression point of 25 5 dBm In order to be able to use the amplifier one should ensure that the power at the output of the amplifier does not exceed the compression point For example if 15 me ter long RF cable is used in the TX end and VNA gives 8 dBm out from the ports the total tra
34. e board chairs tables and window blinds The window blinds are made of metal so they are very good reflectors especially when they are closed Furthermore window blinds causes as well diffraction and scattering in the measurement environment The chairs and tables were organized such that there is an open area between the TX and RX The figure presenting the RX rotated to 45 angle and the TX in operation at the classroom is presented in Figure 13b The chosen antenna configurations in performed measurement cases are presented Table 10 Window blinds a b Figure 13 a Layout of the measurement environment and b the link view of the classroom measurement where RX is rotated to 45 angle 45 Table 10 Antenna configurations for each measurement case in the classroom Case 1 2 3 4 5 6 RX 3x3 1x1 20x20 20x 20 3x1 3x1 TX 13x3 20x20 1x1 1x1 1x1 1x1 note RX at 45 P2 terminated P4 terminated A sample about the recorded impulse responses are presented in the Figure 14 In these measurements we had a bug in the measurement control software The software did set the correct calibration pool only to channel 2 i e only when the VNA port 3 is the TX port This is seen in the calculated impulse response as a delay offset as well as wrong signal levels Thus we cannot use the recorded S21 and S23 to any further analysis since the results are corrupted However the other polar
35. e obstacle If the obstacle is assumed to be wedge shaped it can be ap proximated as a conducting half plane i e a knife edge By the Huygens principle every point of a radiating field can be referred to be the dot source of new electromag netic field geometrical optics GO defines the Snell s law of diffraction as n sin 6 no sin Og 17 where n and n are the refractive indices of the media 1 and 2 respectively and a is the diffraction angle of the wave The Snell s law for diffraction approximate waves as rays ray tubes and it does not take into account the attenuation that the wave undergoes because of diffraction 17 If the knife edge diffraction KED approximation is used we can also calculate theoretical diffraction coefficient The diffraction parameter v can be expressed as o 18 he where H denotes the height of the obstacle with respect to the direct link chord The KED coefficient can be expressed as 1 F u 5 1 jC 55 19 where C v and S v are Fresnel integrals defined as T C v T cos 56 d and S v T sin 36 d6 20 where amp is the auxiliary variable for the integral 12 Diffraction loss factor is the absolute value of the diffraction coefficient To avoid the calculus of complex Fresnel integrals approximations can be used to calculate the diffraction coefficient for certain v values For v gt 0 7 the diffraction loss Ly in dBs can be approximated as 1
36. easuring each single antenna channel between the antenna elements separately one by one 4 One good way to measure the radio channel is to use vector network analyzer VNA with virtual antenna arrays in both the TX and the RX ends Only few physical antenna elements are used and the antenna is moved between different positions to represent a real antenna array Robotics can be used to move the antenna making the actual mea surement smooth and automatic Advantage here compared to the channel sounders is that we do not need to perform external synchronization between the TX and the RX because the VNA takes care of that One serious drawback in VNA based measure ment systems is that the RX and the TX are in the same box meaning that we have to use long radio frequency RF cables to be able to measure trough long link distances However this problem comes less significant in higher frequencies as the reasonable link distances are decreasing meaning smaller cell size Especially at indoor propaga tion environment VNA based systems can be successfully used Another drawback in virtual antenna array based measurement systems is the increased measurement time Antenna movement between the VNA sweeps increases the measurement time sig nificantly The sweep time of the VNA is proportional to the intermediate frequeny IF bandwidth used On the other hand increasing the IF bandwidth decreases the dynamic range of the VNA The trade of between the nee
37. ed 500 MHz bandwidth The antenna layout is presented in Figure 4a The simulated radiation patterns XZ cut of the antenna polarizations in terms of total gain at 10 1 GHz are presented in Figure 4b VNA ports 1 and 3 were connected to the TX antenna and ports 2 and 4 to the RX antenna respectively The antennas were rotated in such a way that the orthogonal polarization planes were tilted at 45 angle with respect to the vertical The simulated antenna properties are presented in Table 1 31 a Figure 4 a An overview of the RX TX antennas and b simulated antenna gains of the two orthogonal polarizations at 10 1 GHz XZ cut Table 1 Specifications of the antennas used in the measurements center frequency 10 1 GHz 10 dB bandwidth 720 MHz gain gt 4 dBi boresight maximum total gain 5 8 dBi port isolation gt 24 dB cross polarization discrimination gt 18 dB 3 3 3 XY gantry Stepper motors are used in various applications to control the motion The precision of the motors is proportional to the step resolution of the motors The smaller steps the stepper can be rotated the less is the minimum distance that can be moved To convert the stepper motor rotation into the linear movement the linear stage can be used High precision linear stages are commercially available to meet the standards of the stepper motors When choosing the stepper motors programmability and controlling in
38. ed measurement setup A block diagram of the measurement system is shown in Figure 2 Idea is to use cen tralized control software such as MATLAB with instrument control toolbox to control the devices and store the measurement data The data flow between the devices is per formed via Ethernet bus by using transmit control protocol internet protocol TCP IP communication protocol The control program should be designed in such a way that it initializes the measurement performs the desired VNA sweeps and moves the an tennas to another location The used VNA was Rohde amp Schwarz ZNB20 21 VNA measurement and the antenna moving has to be synchronized such that the antennas doesn t move while the VNA is performing the sweep The S parameter measurement data measured by the VNA is transferred to the MATLAB as fast as possible and the software takes care of the data flow and indexing Radio Channel y fee antenna v RF cables To move the antehna Receiver XY Transmitter gantry RF cables XY gantry PORT 1 VNA R amp S ZNB20 Measurement USER control software i i Usercontrolsthesoftware Ethernet Mechanical gt RF cable se gt gt connection connection Figure 2 Block diagram of the measurement system 3 1 2 Scattering parameters Scattering parameters S parameters are used to describe the effects of an RF device or the radio channel to the radio wave The
39. ell com datasheets 1809919 pdf publisher Panasonic EW MATLAB Instument Control Toolbox Users Quide accessed 25 8 2014 URL http www mathworks se help instrument index html Lexium MCode Programming and Software Reference Schneider electric accessed 13 10 2014 URL url http motion schneider electric com lmd downloads literature MCode_LMD pdf Chen Z Gokeda G amp Yu Y 2010 Introduction to Direction of arrival Estima tion Artech House Tervo N Dias C Hovinen V Sonkki M Roivainen A Meinil J amp Latva aho M 2014 Diffraction Measurements around a Building Corner at 10 GHz In First International Conference on 5G for Ubiquitous Connectivity SGU IEEE 2014 Appendix 1 Appendix 2 Appendix 3 Appendix 4 Appendix 5 Appendix 6 58 8 APPENDICES The wiring of the VNA based measurement setup Flow chart of the subroutine for finding the limit switches Flow chart of the moving program for steppers The flow chart of the VNA initialization Selectable input parameters for the measurement control software The flow chart of the main program controlling the measurements Appendix 1 The wiring of the VNA based measurement setup Dual polarized TX antenna Ethernet Swicth 2 Radio Channel Dual polarized RX antenna Ethernet Swicth 1 Stepper w SP Se Se ET Instrument Control Toolbox supply Appendix 2 Flow chart of the subroutine for fi
40. ent campaigns and models are made for those frequencies because those bands are potential frequency regions for future wireless local area networks LAN Paper 8 presents channel measurements made on 10 GHz at indoor environment and compares them to the statistical channel models In the paper the authors presents measurement results of the received power envelope Measurement results are fitted to the Rayleigh Rician and Nakagami distributions The paper concludes that the proba bility that the received envelope power is below some threshold follows the Nakagami distribution with a good precision The Rayleigh and Rician distributions were con cluded to fit weakly to the 10 GHz statistical indoor channel model The paper presents only received envelope power measurement results and channel characteristics such as multipath delay spread were not calculated based on the result Paper 9 presents large scale parameters of wideband multipath channels The mod els made by the measurements are based on extensive measurement campaigns in var ious indoor environment The measurements were done by using wideband MIMO channel sounder having 400 MHz bandwidth at 11 GHz The paper characterizes the polarization characteristic of path loss shadowing factor cross polarization power ra tio delay spread and coherence bandwidth of the channel The paper states that the path loss exponent is between 2 and 3 in non line of sight NLOS environment and between
41. ent system is defined as the difference between the highest and the lowest attenuations that the system is able to measure When observing the lower limit one should avoid the RF components to drive into compression On the other hand the attenuation caused by the channel should be less than the maximum attenuation supported by the system in order to be able to distinct the actual signal from the noise 5 3 2 1 Noise in the measurement system In reality there is always some noise caused by the device in the radio system There are different kind of noise sources in RF electronics The most significant one in radio frequencies is the thermal noise caused by the resistive components The thermal noise is white meaning that it remains constant over frequency The power of the thermal noise depends on the bandwidth B and the physical temperature T and can be written as Py kgT B 40 where kp is the Bolzmanns constant It is usually assumed that T T which is the same as the nominal room temperature 11 Signal to noise ratio SN R is used to describe the signal versus noise quantity It is defined as SNR 41 where Ps and Py are the signal and noise powers respectively 14 3 2 2 Noise factor noise figure and noise temperature For the radio devices such as RX we can define parameters to describe how much noise is appended to the system by the specific device In an RF amplifier both the signal and the noise are am
42. ferent propagation phenomena presented in next Sections 13 The polarization vector p represents the polarization of the wave The polarization vector is simply the unit vector pointing to the direction of the electric field 13 2 2 Propagation in the radio channel By radio channel we mean the whole radio system including the TX the RX and the propagation channel Depending on the channel geometry and the propagation materials the radio wave may travel trough several different paths between the RX and TX Thus the radio wave undergoes several radio propagation phenomena between the TX and RX which all affects to the wave behavior in the channel In this section we will present the basic radio propagation phenomena which are valid especially for indoor propagation environment 14 2 2 1 Free space path loss The propagation medium is defined to be free space when the first Fresnel ellipsoid is clear from obstacles In case of some obstacles within the Fresnel zone the transmitted 15 signal experiences some other propagation phenomena besides free space propagation The radius of the first Fresnel ellipsoid is defined as Adi dy hp 5 di d2 where dj and d are the distances presented in the Figure 1 14 vi a ap d d gt Figure 1 The first Fresnel ellipsoid If the signal is assumed to be propagated only trough free space the received power PR can be represented by Friis equation A
43. g to the stepper from which address is starts to execute commands This allows MATLAB to do other things than waiting the device to be ready while the step per programs are executed We can for example collect results or initialize the next VNA measurements while the virtual antenna is changing its position The stepper has a number of flags and registers that can be checked to know in which point of program the stepper currently is and if it is ready or not to take new commands 36 A number of different subroutines were implemented for controlling the stepper motors The first on is to set the origin of the linear stages The flow chart of the origin setting is presented in Figure 7 The subroutine uses another subroutine for finding the limit switches from the XY gantries The flow chart of this subroutine is presented in Appendix 2 The basic idea is to find the move the carriage towards the stage negative end until it reaches the negative limit switch The negative limit witch is routed to the pin 3 12 of the stepper general purpose interface As the limit switch is reached the voltage in J rises up which is programmed to stop the motor Then the carriage is moved a number of steps offset to the positive direction which is set as an origo P 0 Same is done for all of the stepper motors Tranform user Find origins for all Check that the N Walt until steppers defined offsets to steppers with epr origin was found are read
44. he medium 14 2 2 3 Plane wave in the media boundary When a planar radio wave comes to the boundary of two different propagation media some part of the wave power is reflected back form the boundary while the rest of the wave power propagates into the medium Snell s law for reflection is represented as 0 6 12 where 0 is the angle of reflected wave and 6 is the incident angle of the wave 11 The polarization of the wave affects to the wave behavior at the media boundary The amount of relative power reflecting back form the boundary can be expressed by the reflection coefficients specified for both perpendicular and parallel electric field components with respect to the boundary Hence the polarization vector of the wave may be changed due to the reflection but not the actual polarization This means that a linearly polarized wave stays linearly polarized also after the incidence The reflection coefficients for the perpendicular and parallel polarizations can be expressed as 16 cos 6 1 2 sin 6 2 sin 6 2 cos 6 R and Ri A 13 1 2 sin 6 cos 6 1 2 sin 0i 2 cos 6 As part of the wave is reflected back from the media boundary the rest of the energy is propagated trough the boundary inside the medium The propagation angle of the wave may change depending on the relation of the dielectric properties of the media Thus the wave undergoes refraction in the media boundary If we denote 0
45. high wall penetration loss On the other hand the walls gives a possibility to isolate the area better and hence avoid interfer ence coming from outside or other rooms The reflection diffraction and free space propagation are the main propagation phenomena at the indoor environment if the pen etration loss of the walls are assumed to be high Long corridors makes it possible to transmit signals trough multiple reflections from the source to the destination 3 21 From the channel modeling and measurement point of view indoor environment has two properties that makes the measurements easier First the environment can often be referred to be static Thus the mobility in the channel is low channel coherence time is large This is the case for example in the office environment In some indoor en vironments such as shopping malls there is often people moving in the environment when the channel is static anymore However the mobility is still quite low compared to the outdoor channels where we can have for example cars in the environment Sec ond the distances that the channel models needs to support is usually smaller than at the outdoor environment This means that less dynamic range is required and the channel can be assumed to be constant during the measurements 2 3 5 MIMO channel modeling To exploit the multipath channel to achieve the better system performance in a telecom munication system more than one antennas can be used in
46. hor izontal and 25 in vertical direction if the antenna spacing of half of the wavelength was used Thus large antenna arrays up to 650 650 MIMO can be measured In total it could be said that there are several properties in the current system which we have succeeded to implement Many improvements that are proposed in the next Section are just proportions and there are not necessarily needed to be implemented 5 2 Improvements proposed to the system There are a number of ways to improve the measurement system by optimizing the measurement time and adding new options to the system that can be chosen by the user To make the further analysis of the measurement results smoother in the future an external analysis software could be made to analyze the wanted results right after the measurements For the user this would give an opportunity to change parameters while measuring and thus be able to improve the results From the user point of view graphical user interface GUI would clarify better the options supported by the system Also it would give a better impression to the user that what is going on in the system MATLAB has a very good GUI tool which could be used for building the user interface Same subroutines which was made for the current system could be applied GUI would also give possibility to make for example pause option for the system such that user could pause the long measurements if needed In the channel measurements the l
47. ibing the antenna element indices in the TX and the RX ends For example cell matrix copol_1_2_3 is the data block measured in antenna positions Re 1 Ty 2 Tx 3 contains Npoinis nry 2 complex elements i e recorded S91 and S43 data in measured frequency points at each antenna positions in RX y axis Similarly xpol_1_2_3 is the same result but the measured parameters are cross polar izations S93 and S41 The initialization of the result matrices performed in MATLAB by specific subroutine Compexsample Figure 9 Data format of the saved measurement results 3 5 4 Error control For the input output IO error handling MATLAB is not the best programming en vironment to do error controlling The only error handling and interrupt control tool in MATLAB is the try catch function In the current measurement control software version we have included only the most necessary error handling In order to prevent the MATLAB to jam during the program execution the whole measurement control program is set under one big try catch structure In case of error the actual error message is printed to the display and saved to a log file In case of interrupt or error the program execution is stopped without exceptions Basically the program is terminated in case of any internal or external error This includes for exam ple that the devices are not connected properly or the device is for example switched off during the measurement The recordings
48. ing is the LOS environment The Rician distribution is a function of two parameters K and Q The probability density function PDF of the Rician distribution is defined as 2 K 1 x Qr 2 K 1 x Or 2 K 1 x f z exp K where K is said to be a Rician K factor defined as the ratio of the strongest multipath typically LOS compared to other multipaths Jo is the Bessel function and Q is the total power of all the propagation paths 19 Rayleigh fading is typically used in NLOS environment In the Rayleigh fading the magnitude of the received signal follows Rayleigh distribution described by the parameter Q The PDF of Rayleigh distribution can be written as 19 f a exp Z 24 2 3 Radio channel modeling The radio channel models are usually defined to deterministic and stochastic channel models Some of them are defined based on theory while others are based on the mea surement data The stochastic models relay on statistical distribution of the channel while deterministic models tries to model the path loss in the channel deterministically by using for example the geometry of the propagation environment In geometry based deterministic channel modeling ray tracing RT is often used By the RT we mean the 19 geometry based radio wave path estimation between the TX and RX In this section we present some key parameters and theory related to radio channel modeling 2 3 1 Frequency and impulse
49. ion intensity averaged over all direc tions Gain of the antenna is closely related to the directivity It is a measure that takes into account the efficiency of the antenna as well as its directional capabilities Thus the gain can be written as 13 G 0 9 ecaD 9 9 51 By taking into account also the impedance mismatch of the antenna the absolute gain of the antenna can be defined as 13 Gaps 9 o r caD 0 o 52 The impedance of the antenna is complex and can be written as 13 Za R X RL R 5Xa 53 where Z4 is the antenna impedance Ra is the antenna resistance and X4 is the antenna reactance 13 The 10 dB bandwidth of the antenna is defined as the frequency range where 511 is less than 10 dB Polarization of the antenna in a given direction is defined as the polarization of the wave transmitted radiated by the antenna The polarization of the radio wave was defined in Section 2 1 2 The polarization loss factor PLF describes the polarization mismatch between the antenna and the wave It can be written as PLF ppa Pow 54 where ppa and ppw are the polarization vectors of the antenna and the wave respec tively 13 3 3 2 Dual polarized patch antenna The antennas used for the measurement setup at the TX and the RX were directional dual polarized patch antennas The antennas have good impedance matching lt 13 dB and isolation gt 24 dB between the feeding ports over the measur
50. ion of the RX was done for testing the AoA algorithms The antenna config urations in each measurement case are presented in Table 9 There where no internal spurious response caused by the measurement system seen in the impulse responses Table 9 Antenna configurations for each measurement case in the anechoic chamber case 1 2 3 4 RX 3x3 20x20 1x1 20 x 20 TX 13x3 1x1 20x20 1x1 note RX rotated 18 8 A sample of the measured impulse response in direct LOS scenario is presented in Figure 12 In the figure we have marked out the LOS and reflected paths respectively As we can see from the figure the LOS path is the first arriving multipath component and it is also the strongest one The link distance was 4 92 m which for the free space loss calculated by 7 is 66 4 dB Both antennas has a gain of approximately 4 dBi in boresight so the overall loss here would be 58 4 dB The measured path loss with an tenna gains for the LOS path seen in the Figure 12 is 60 dB which corresponds almost the calculated result The slight 1 6 dB difference for the measured and calculated re sult is caused by the antennas because they are not identical in reality since they are build by hand in the workshop of the University of Oulu In the impulse response we also see the reflected path In the reflection the both of the orthogonal linear polariza tions experiences different reflection coefficient according to Section 2
51. ional 3D space AoA and AoD is often defined separately for azimuth and elevation domains Thus the azimuth and elevation angles can be extended into 3D as direction of arrival DoA and direction of departure DoD The Angular spread is a parameter to describe the spatial order of the channel 3 Te 31 2 3 3 Deterministic channel models The deterministic channel models tries to estimate the path loss and phase difference experienced by the signal as it propagates trough the channel The deterministic path loss models such as the free space path loss model presented in 7 are used to calcu late the path loss of the channel The models include a number of approximations and all of the radio propagation phenomena are not usually taken into account This means that we have to simplify the geometry of the environment in order to estimate the most significant multipaths from the impulse responses The models are parameterized to fit to the propagation environment One deterministic channel model the simplified path loss model is defined as d a 32 where L is the path loss y is the path loss exponent Lye is the path loss at the reference distance d and d is the distance in meters 3 2 3 4 Indoor channel modeling The indoor environment has many characteristics making them different from outdoor environment At indoor environment the walls are limiting the radio wave propagation especially in higher frequencies because of
52. ization can still be used The calculated impulse responses shows that in the classroom there is many multipaths between the TX and the RX There is periodical peaks in the impulse response caused by the multiple reflections AoA plots will later on show from witch angle the most significant multipaths are coming from i RX 0101 i TX 1202 100 Path Loss dB l O T 0 100 Figure 14 Example of the impulse responses in classroom S2 and S53 corrupted 4 4 Wall penetration loss measurements One application for the measurement setup is to measure the penetration of a concrete wall by placing the transmitter and RX antennas to different sides of the wall The advantage of our system here is that we can easily perform several penetration loss measurements of the wall with a small spacing between points The penetration loss 46 of the wall can vary depending on exact location and can contain different materials such as metallic structures to uphold the concrete These structures can have a sig nificant effect to the penetration loss By making several measurements in antenna positions very near to each other the fluctuations in the wall structure can be located and included into the penetration loss measurements We can also measure the pen etration loss in different RX TX path angles when the wave incidence to the wall is not orthogonal In that case the part of the transmitted power is reflected
53. l measurement scenarios and their initial results were presented to ver ify the system operation and to demonstrate the system applicability for the different cases A verification measurement was performed in the anechoic chamber to verify that the system did cause any internal spurious responses to the results The next mea surement was performed in a classroom to demonstrate the multipath propagation envi ronment The reflection diffraction and LOS propagation phenomena were concluded to be the most dominating propagation phenomena in the classroom Furthermore an indoor wall penetration loss measurement from the classroom to another was made to show that the measurement system could also be applied for the measurements requir ing an accurate antenna shifting between the measurement points The result of this measurement was that the wall penetration loss may vary drastically depending on part of the wall under test The penetration loss of the measured wall was varied between 5 and 13 dB The results measured with this setup could be used for angular domain algorithms to estimate the direction of arrival and departure respectively The mea surement system was concluded to be able to make successful measurements in the tested propagation environments The measurement system will be applied for various MIMO radio channel measurement scenarios at 10 GHz and beyond 55 7 REFERENCES 1 Massive MIMO Info Point accessed 25 8 2014 URL www ma
54. ld is defined as 2D2 a 1 where Dy is the largest dimension of the antenna measured in perpendicular to the antenna radiation direction and A is the wavelength of the wave 11 Tf 2 1 1 Plane waves and spherical waves When observing the whole wavefront that the antenna is transmitting the wavefront is seen as spherical Electric field of the spherical wavefront can be written as E jkr E r Eo 0 9 J 2 where r is the radial distance from an the antenna Eo 0 amp is the electrical field vector at distance r 0 as a function of direction of propagation 0 amp k is the wave number and j is the unit imaginary number Wave number k can also be expressed as 12 27 k W 3 In the antenna far field region the spherical wave can be locally approximated as a plane wave This is because every source looks like a dot source when observing it from far enough The electric field of the plane wave can be expressed as e jkr E r Eo 9 ji 4 14 where k is a complex wave vector and r is a position vector defining a point in 3D Three Dimensional space 12 2 1 2 Polarization The polarization of the electromagnetic wave describes the time varying direction and relative magnitude of the electric field vector In the 3D vector space the polarization describes the function of how the electric field vector varies among the direction of propagation or wt axis We can classify different kind of polarizations t
55. ll s Equations Cambridge Uni versity Press 11 Pozar D M 2005 Microwave Engineering John Wiley amp Sons third ed 12 Bertoni H L 2000 Radio Propagation for Modern Wireless Systems Prentice Hall PTR 13 Balanis C A 2005 Antenna Theory Analysis and Design John Wiley amp Sons third ed 14 R is nen A amp Lehto A 2003 Radio Engineering for Wireless Communication and Sensor Applications Artech House 15 Miquel A C 2009 UWB antenna design for underwater communications Mas ter s thesis Delft University of Technology 56 16 Lindell I 1996 Radioaaltojen Eteneminen Otatieto fourth ed 17 Saunders S amp Arag n Zavala A 2007 Antennas and Propagation for Wireless Communication Systems John Wiley amp Sons second ed 18 ITU R Recommendation P 526 7 Propagation by diffraction accessed 13 10 2014 URL url https www itu int dms_pubrec itu r rec p R REC P 526 7 200102 S PDF E pdf 19 Parsons J 2000 The Mobile Radio Propagation Channel John Wiley amp Sons 20 Phillips C Parr J amp Riskin E 2013 Signals Systems amp Transforms Pearson Education 21 Rohde amp Schwarz ZNB ZNBT Vector Network Analyzer User manual accessed 25 8 2014 URL http www rohde schwarz com en manual r s znb znbt user manual manuals gbl 78701 29151 html 22 Sundararajan D 2001 The Discrete Fourier Transform Theory Algorithms and Applications
56. locks Even though the high precision rubidium clocks were used there is still some imprecision in timing The third problem is the large antenna arrays used for the measurements For every freguency range measured specified antenna array has to be designed individually Other good possibility is to use VNA with virtual antenna array Now only few physical antenna elements are used and the antenna is moved between positions to represent an antenna array One drawback is that the produced measurement result does not take into account the correlation between antennas in the one end On the other hand simple measurement setup can be used with the device that can be later on applied to the various applications Furthermore the system can be scaled to other freguencies simply by changing the VNA parameters and using different antennas If the VNA s own freguency range is not enough mixers can be used to increase VNA freguency range However using external mixers may increase the system noise as well One drawback in virtual antenna array based systems is the increased measurement time Antenna movement between the VNA sweeps increases the measurement time significantly Also if narrow IF bandwidth was used one sweep would take hundreds of milliseconds of time The trade of between the needed dynamic range and the measurement speed is needed and the overall performance has to be optimized with respect the desired property 5 23 3 1 1 The propos
57. m can be written as N points 1 1 Siml fp e er Nevins 35 hi ta N points k 0 where t is the n th time instant and Npoints is the number of frequency points 22 The measured bandwidth in freguency domain determines the time resolution in time domain Inverse fast fourier transform IFFT algorithm is used to calculate the IDFT When using direct IFFT for the freguency domain samples measured by VNA we obtain a time resolution of i gt Jasa where Breas is the measured bandwidth In time domain the number of points Npoints is the same as in freguency domain if zero padding is not performed while executing IFFT Zero padding increases the resolution but not accuracy because of interpolation and thus it should not be used in case of measured data Thus the length of the impulse response is written as 23 t 36 At Npoins 1 6t 37 The distance resolution of the recorded impulse responses can roughly be calculated as d dtcg 38 where co denotes the light speed in the free space The length of the impulse response in distance domain i e the maximum detectable path can be written as Ad Atco Npoins 1 6d 39 25 3 2 Dynamic range of the measurement system For each measurement one should ensure that the dynamic range offered by the mea surement system is reasonable for performing successful radio channel measurements The dynamic range of the channel measurem
58. made before the error is saved into a temporary file from where they can be reloaded again afterwards 39 3 5 5 User interface The measurement software is developed in such a way that the user can parameterize the configuration by modifying a single input file called inputs m Before the mea surements the user should perform the calibration manually and save the calibration pool into the memory of the VNA User can make his own calibrations for the both of the measurement channels to speed up the measurement The name of the applied calibration file has to be specified in the measurement input file and the software sets automatically the defined calibration pools into the measurement channels The input parameters that the user can set are given in the Appendix 5 The main script of the control software uses subroutines defined in Sections 3 5 1 and 3 5 2 to control the devices The inputs given by the user as defined in Section 3 5 5 are used to parameterize the program A waiting period is added in the vertical direction movement because of vibrations of the antenna mast The flow chart of the measurement control software is presented in Appendix 6 40 4 MEASUREMENT SCENARIOS AND RESULTS The measurement system can be applied for various measurement scenarios The recorded S parameters can be used to calculate information about the propagation paths dominating propagation phenomena channel correlation and many others The rec
59. mands for programmable instruments scattering parameters transmit control protocol internet protocol transmitter uniform theory of diffraction vector network analyzer three dimensional fifth generation signal entering to the 2 port input signal leaving from the 2 port output bandwidth coherence bandwidth of the channel doppler bandwidth intermediate frequency bandwidth measurement bandwidth signal reflecting from the 2 port input signal reflecting from the 2 port output Fras F v him h t Ln d Lret NF N Fvna Fresnel cosine integral light speed in vacuum largest dimension of an antenna absorbing screen diffraction coefficient antenna directivity distance distance that the wave has propagated in the medium reference link distance distance from the TX distance from the RX electrical field vector as a function of r electrical field vector at distance r 0 Neper number antenna radiation efficiency antenna reflection efficiency total antenna efficiency total noise figure of cascaded RF blocks noise factor knife edge diffraction coefficient freguency n th recorded freguency sample center freguency gain RX antenna Gain TX antenna Gain antenna gain absolute antenna gain height of an obstacle MIMO channel matrix channel freguency response radius of first Fresnel ellipsoid channel coefficient from TX antenna i to RX antenna j impulse response between ports 1 and m channel impulse res
60. nce the variations are so huge we could also decrease the antenna spacing and hence get more precise results Here the result is that a wall penetration loss can vary a lot depending on the antenna location even in orthogonal incidence with respect to the wall under test The variations in the horizontal direction is observed to be stronger than the variations in the vertical direction This may be caused by the structure of the wall as well as the reflections from the roof and floor Hence an office room might not be separated from the other rooms when considering a communication system where each room has its own BS i e one cell consists of one room MN Wall under terst M49 Wall to corridor 1 5f o TX positions o RX positions Y direction m 4 3 2 1 0 1 X direction 1 Figure 16 Antenna positions when measuring wall penetration loss 48 4 z gt j x A x 6 3 s A E E A sum S vertical movement 14 sum S43 vertical movement sum S horizontal movement i sum S43 horizontal movement x 16 I T I 0 2 4 6 8 10 12 14 16 18 20 Antenna position in X Y direction RX TX Figure 17 The penetration loss of the wall in with respect to X direction antenna position 50 55 L 60 5 E x ic 65 gt o sum S x sum S 3 70F mean sum S n mean sum S 75 1 I I 1 i J
61. nding the limit switches Start S Flush the object input H Define limit stop mode tobe LM 4 Define input 1 as Define input 2 as positive limit negative limit Move stepper 20000 microsteps to j 4 positive diretion Move to negative direction No Yes v Stop movement Move the amount of offset to positive direction Clear error register SetP 0 End program g Appendix 3 Flow chart of the moving program for steppers Flush object input I Go to address where the program will be written I Clear programs from that address H Select stopping mode as AS 2 H Define limit swicthes I Set Initial speed and maximum speed I Set acceleration and deceleration l Set hold current and run current Move relatively the amount of steps defined by user End program Appendix 4 The flow chart of the VNA initialization Creat channel 1 4 Load calibration set 1 H Set center freguenct span and number of points Create trace 1 as Copolt S21 Create trace 2 as Xpol1 S41 isplay trace in JA Se Vv A Start PUI Delete all current channels v
62. neaarisille polarisaatioil le Antennielementti siirret n tasossa k ytt m ll ohjelmoitavia askelmootto reita ja lineaariyksik it Kaikkia mittausj rjestelm n laitteita ohjataan samalla ohjausohjelmistolla jonka toimintaperiaate on my s esitetty t ss ty ss Mittausj rjestelm n toiminta varmistetaan ja sit demonstroidaan suoritta malla kanavamittauksia erilaisissa etenemisymp rist iss Varmennusmittaukset suoritetaan kaiuttomassa huoneessa jotta voidaan varmistua siit ettei j rjes telm tuota sis isi harhatoistoja jotka vaikuttaisivat mittaustuloksiin Monitie etenemisymp rist demonstroidaan kanavamittauksilla luokkahuoneessa My s luokkahuoneiden v list etenemisvaimennusta mitataan Mittaustuloksia voidaan k ytt muodostettaessa erilaisia radiokanavamalleja ja niit voidaan soveltaa my s aallon tulo ja l ht kulman estimointiin k ytt m ll siihen tarkoitettuja al goritmeja Avainsanat millimetriaallot staattisen kanavan mallinnus massiivi MIMO ka navamittaukset TABLE OF CONTENTS ABSTRACT TIIVISTELM TABLE OF CONTENTS FOREWORD LIST OF ABBREVIATIONS AND SYMBOLS 1 INTRODUCTION siisi SA KASSAAN eee TALK De A SS 10 2 RADIO CHANNEL CHARACTERISTICS o sosss nnen 13 2 1 Electromagnetic propagation oos soks eee cece ee eee ee 13 2 1 1 Plane waves and spherical waves oossoo eee eee 13 21 2 Polarization 30 hs tissa pracie nsien en a aa ieee 14 2 2 Pr
63. nsmitted power would be 23 5 dBm plus the antenna gain The specifications of the proposed TX amplifier are presented in 24 As discussed before the external LNA can even increase the noise in the RX end if short cables are used Thus it is not beneficial to use LNA if the VNA is placed near to the RX end However if long RF cables are used in RX end we propose to use LNC 0812 LNA manufactured by Miteq It gives us maximum noise figure of 1 8 dB at 10 GHz with gain of 25 dB at minimum The overall effect for the RX can be calculated by Equations 47 if the noise figure of the VNA is known The specifications of the proposed LNA are found from 25 3 2 6 Calibration of the VNA The VNA has to be calibrated before the measurements The calibrations should in clude all of the RF components used in the RF chain If the external amplifiers where used they should be included into the calibrations or canceled out from the results afterwards There are many possible calibration methods supported by the used VNA However only few of them are supported by the used calibration kit which was R amp S 29 ZV Z5x 26 VNA usually measures all of the S parameters even though only few of them are selected to be saved into defined traces The ones not saved into traces are dummy measurements which only increases the measurement time This means that we can speed up the measurement by calibrating the VNA only for the parameters which are supposed to
64. nt decrement ry decrement rx Decite moving direction Move TY stepper one step Move TX stepper Increment one step decrement ty eranient Break measurement N Move saved results y loop to the result folders Close and clear Drive stepper TCPIP objects motors home gt End program
65. nts afterwards If VNA would support the external amplifier selection by adding it to the rear panel of the VNA we could also use that option to compensate the amplifier off from the results However our VNA did not support this option When using amplifiers we also need to ensure that the received power does not reach the level that drives the RX into compression Using external amplifiers does not necessarily increase the system dynamic range but shifts it down in the power region The measurement system should be able to be modified in such a way that the dy namic range must be able to be scalded if needed in order to achieve good speed with respect of accuracy The external amplifiers were added to the measurement chain only if the VNA s own sensitivity was not enough This is the case when the long RF cables are used causing external attenuation to the signal hence decreasing the dynamic range left for the channel measurements External LNA could be placed right after the RX antenna to increase the RX sensitivity Power amplifier could be placed right before the TX antenna to increase the transmit power If the measurement envi ronment requires to use long cables long links are wanted to be measured the best option is to use long cables in both ends However this would require that the VNA should most probably be placed between the antennas inside the measurement environ ment affecting to the radio channel that is to be measured Furthermor
66. o be linear circular and elliptical When the electric field is oscillated only in one line the wave can be said to be linearly polarized Linear polarization can always be reduced to two polarization components vertical and horizontal In vertical polarization the electrical field is oscillating vertically among the y axis with respect to the direction of propaga tion time axis In horizontal polarization the same happens in the horizontal plane 1 e the electrical field is oscillating among the x axis In circular polarization the elec tric field goes around the circular orbit over the time axis with a constant amplitude In elliptical polarization the electrical field vector goes around the elliptical orbit over the time axis and the amplitude is also varying In case of elliptically polarized wave we can define the axial ratio of an ellipse that the electrical field vector is tracing The axial ratio is the ratio of the magnitudes of the major axis and minor axis In case of circular polarization the axial ratio is one 13 The polarization of an antenna is said to be same as the polarization of the radio wave the antenna is radiated Therefore vertically polarized antenna receives poorly horizontal polarized waves Same goes the other way around However in the radio channel the polarization is not always the same in the TX and the RX Thus the polarization can change while the wave is traveling trough the radio channel due to the dif
67. ocation of the antennas should always be able to be measured with a good precision When going higher in the frequency spectrum the position accuracy may became even more important To make the location measure ment automatic programmable laser distance meters could be used by attaching them to the antenna mast structure It could be possible to read the distance data automati cally and store them within the measurement data This would make further analysis of the results easier and data fitting could even be made during the measurements Also for example global position system GPS coordinate could be recorded if for example VNA has an internal GPS receiver which is the case for many devices However the precision of the location measured by GPS is not very precise but it could still be used to rough estimation of the measurement location and thus help the future analysis On the other hand all external measurements such as distance data read from the distance meters increases the measurement time especially if the data is read between each an tenna positions The usage of them should be able to be selected by user to be optional Making improvements by adding the new options would also increase the prize of the system which is not the aim if the system is wanted to pay itself back in the future by giving good measurement results One problem in the current system is that it is only capable to measure planar antenna configurations However
68. onstrate the system in different cases A verification measurement is performed in an anechoic chamber to verify that the system does not cause internal spurious responses to the results The next mea surement is performed in a classroom to demonstrate the multipath propagation environment Furthermore an indoor wall penetration loss measurement from the classroom to another is made to show that the measurement system can also be applied for the measurements requiring an accurate antenna shifting between the measurement points The results measured with this setup can be applied angular domain algorithms to estimate the direction of arrival and departure respectively Keywords millimeter wave static channel modeling massive MIMO channel measurements Tervo N 2014 Virtuaalista antenniryhm k ytt v MIMO radiokanavan mit tausj rjestelm 10 GHz n taajuudelle Oulun yliopisto tietoliikennetekniikan osas to s hkotekniikan koulutusohjelma Diplomity 58 s TIIVISTELM T m diplomity esittelee moniantenniradiokanavan mittausj rjestelm n jossa mittauslaitteena k ytet n vektoripiirianalysaattoria ja kahta virtuaalista ta soantenniryhm J rjestelm mittaa jokaisen antennielementin v lisen kanavan erikseen Etenemisymp rist on oletettu staattiseksi radiokanavan tallennuksen aikana Antennielementtin k ytet n kaksoispolaroitua mikroliuska antennia jossa on omat sy tt portit molemmille ortogonaalisille li
69. opagation in the radio channel 0 0 cee eee eee eee 14 Dio Nt Free spacepath 10880 AL Ae ARs 2238 Ah Oe MA ee 3284 14 2 2 2 Plane wave in the Medium lt 26 22 2 64 4 kaisa tees A ek Ss 15 2 2 3 Plane wave in the media boundary 0005 16 2 2 4 Plane wave in rough surfaces and diffracting edges 17 2 2 5 Fading and shadowing 1 Bho tg ope Ne NAS 8 ge X 18 2 3 Radio channel modeling 5 4 044 widow sa RS Ga sew Talossa PSE KR ae eles 18 2 3 1 Frequency and impulse responses 0 00 000s 19 2 3 2 Delay spread doppler spread and angular spread 19 2 3 3 Deterministic channel models lt veers eek eee Rk 20 2 3 4 Indoor channel modeling 2 0 0 cee eee eee eee 20 2 3 5 MIMO channel modeling 0 0 cee eee eee eee ee 21 3 RADIO CHANNEL MEASUREMENT SETUP 00 00 22 3 1 VNA based measurement system 0 0 e eee ee eee 22 3 1 1 The proposed measurement setup 0 0002 eee 23 3 1 2 Scattering parameters thus aoc eka Ge Ga Bete eee hee 23 3 1 3 VNA time domain analysis 0 0 cee eee NNN 24 3 2 Dynamic range of the measurement system 0005 25 3 2 1 Noise in the measurement system 00 eee 25 3 2 2 Noise factor noise figure and noise temperature 25 3 2 3 Noise in cascaded radio blocks 0 0 e ee ee ee eee 26 3 2 4 Noise in the VNA us
70. orded data can also be used to estimate the DoA and DoD The used DoA algo rithms to analyze the verification measurements are presented in 34 In this chapter we represent some examples about the measurement scenarios which for the system can be used The measurements were performed in the Oulu University campus area 4 1 Used measurement settings and the system speed performance Same measurement settings were used for all of the verification measurements The used VNA settings are presented in Table 6 The center frequency 10 1 GHz and bandwidth 500 MHz were chosen to satisfy the radio permission The number of frequency points used was 201 the IF bandwidth 10 kHz and the transmit power 8 dBm which was the maximum power available from the used VNA According to the Section 3 1 3 the used VNA settings gives us time domain impulse responses whose characteristics are presented in Table 7 Table 6 VNA settings in the verification measurements 0 B meas B IF P T N points 10 1 GHz 500 MHz 10 kHz 8 dBm 201 Table 7 Properties of the calculated impulse responses number of samples 201 delay resolution 2 ns unaliased length time 400 ns path resolution 60 cm maximum detectable path length 120 m In the final measurement system there are many sources of inaccuracy that must be considered The accuracy of the XY gantry is limited by the used moving equipment and self build
71. plified to the output The noise factor can be defined as SN Rin SN Rout Fy 42 where SN Rin and SN Row are the signal to noise ratios at the input and output The noise figure is the noise factor expressed in decibels and it can be written as 11 NF 10 logy Fy 43 Noise equivalent temperature 7 describes the thermal noise existed in a radio block For a radio component it can be defied as GkB 44 26 where No is the noise power delivered to the output and G is the gain of the component The relation of the noise temperature to the noise factor is 11 T F 1 To 45 3 2 3 Noise in cascaded radio blocks The radio device consists many different kind of blocks in a chain which all affects to the noise of the whole system In a chain of RF blocks connected in cascade the overall noise temperature can be calculated as 11 TaT 5 Tai 46 cas el T FT i 2 ja G where Niocks is the number of blocks connected in cascade Similarly the overall noise factor of the cascaded chain can be defined as Nplocks Feas Fi f X SN 47 i 2 j 1 G 3 2 4 Noise in the VNA Noise power in the VNA is proportional to its RX bandwidth as it was shown in Section 3 2 1 The bandwidth B is limited by the IF bandwidth of the RX From the Equation 40 we see that doubling the bandwidth doubles also the noise power 5 Because the VNA has its own RX it also has its own low noise amplifier LNA
72. ponse Bessel function imaginary unit Rician K factor wave number complex wave number Bolzmann constant path loss free space path loss knife edge diffraction loss loss in the dielectric medium path loss at the reference distance noise figure noise figure of VNA N blocks N Fr No MR NRx NRy NT NTx NTy ne PDP N R Ps IT Ra R Ry Ry Tf Sim fn SNR SN Rin SN Rou S v PLF X f a t Y f y t number of RF blocks in the RX chain number of frequency points noise power delivered to the output number of RX antennas number of RX antennas into x direction number of RX antennas into y direction number of TX antennas number of TX antennas into x direction number of TX antennas into y direction refraction coefficient for the medium 1 refraction coefficient for the medium 2 power power delay profile noise power received power signal power transmitted power antenna resistance antenna loss resistance antenna radiation resistance reflection coefficient for perpendicular polarization reflection coefficient for parallel polarization distance from the antenna far field distance S parameters measured between ports and m signal to noise ratio signal to noise ratio at the input signal to noise ratio at the output Fresnel sine integral reflection coefficient of the 2 port input transmission coefficient of 2 port physical temperature in Kelvins channel coherence time total noi
73. r measurement time in the x direction compared to the y direction Table 8 The measurement system speed measurements case time one point in x direction 1 9 s one point in y direction 0 9 s test case with antenna configuration RX 3 x 3 2 min 27 s TX 3 x 3 4 2 Verification measurements in anechoic chamber To verify that the measurement system works properly we decided to perform few verification measurements in as robust and stable LOS propagation environment as possible This measurement can also be used as a reference measurement for the fur ther measurements and for the angle estimation algorithms The anechoic chamber placed in the Oulu University campus area was a good place to do this The measure ment system was assembled to the anechoic chamber and the room was isolated from the most of the reflecting and diffracting surfaces by using absorber pillows By doing this we wanted to ensure only LOS component is seen in the measured impulse re 42 sponse However the floor of the chamber contained some metal which caused some reflection in the measurement environment The overview of the anechoic chamber is presented in Figure 11 The TX was set to the right hand corner and the RX was set to left hand corner near the door The mea surements were performed in direct LOS cases with different antenna configurations and LOS with the RX rotated 18 8 counterclockwise with respect to the direct path The rotat
74. r the stepper motors However that slows down the measure ment Also we added an optional offset delay for the both directions in the X Y gantry as an option in the measurement control software The user can choose by himself if he uses the external delay or lower acceleration and deceleration ramps to prevent the antenna shaking Best way to avoid the shaking would be to update the antenna mast to a stronger structure than it is now For example we could add additional structures to support the mast and reduce shaking One could also add acceleration sensors near to the antenna to detect if the antenna is steady enough for the next measurement Another serious drawback in the system is the poor error and interrupt control The error control in very long measurements where tens of thousands of VNA sweeps are performed is a very good way to prevent to corrupt the measurement data and cause problems even during the measurements The used limit switches ensures that the carriage stays inside the allowed region of the XY gantry However if the stepper is for some reason switched off during the measurement the limit switch interrupt 51 routine may vanish from the stepper memory The problem was overcame by defining the limit switches in the measurement control software in the moving program The maximum moving range in the XY gantry is 39 cm in horizontal direction and 38 cm in vertical direction In 10 GHz this corresponds 26 antenna elements in
75. se equivalent temperature of cascaded RF blocks noise equivalent temperature transmission coefficient for perpendicular polarization transmission coefficient for parallel polarization room temperature in Kelvins time antenna reactance Polarization loss factor transmitted signal frequency response transmitted signal in time domain received signal frequency response received signal in time domain antenna impedance path loss exponent propagation constant maximum detectable path distance maximum detectable path delay path distance resolution path time resolution loss tangent permittivity real part of the permittivity imaginary part of the permittivity vacuum permittivity permittivity of the medium 1 permittivity of the medium 2 elevation angle diffraction angle incident angle reflection angle transmitted angle wavelength permeability fresnel diffraction parameter auxiliary integral variable polarization vector polarization vector of the antenna polarization vector of the wave electrical conductivity delay mean delay spread RMS delay spread azimuth angle Rayleigh distribution parameter angular frequency 1 INTRODUCTION In the modern telecommunication systems the presence of multiple input multiple output MIMO has taken a huge role when trying to increase the capacity of the wireless systems In a multipath radio channel one way to increase the capacity is to increase the number of antennas beyon
76. ss room was in reality the first measure ment scenario performed with this system Thus there were many unsure things in the measurement control software and the system itself which had never been tested before The bug in the code discussed in the Section 4 3 did cause corrupted data be cause of wrong calibration pool in the VNA channel 1 The calibration problem was later on corrected Anyway the measurements gave an valuable information about the wave behavior in the class room environment Also the recorded data has been used to test the angle algorithms for a multipath environment Yet the results could be analyzed further to calculate channel parameters such as delay spreads and angular spreads However several RX positions should be considered in order to calculate for example path loss exponent for the environment 53 The wall penetration loss in 10 GHz is important especially for calibrating RT tools to produce a 3D channel model about the room As it was seen in the measurements the penetration loss of the wall may vary even 8 dB depending on the part of the wall under test To be able to calculate trustable mean penetration loss of such wall we should perform several measurements for similar walls and compare them Indoor walls might have many different kind of structures which causes huge variations for the penetration loss If the wall is made of concrete the way how the wall was masoned e g water concentration used for concrete
77. ssivemimo eu 2 International Telecommunication Union official website accessed 25 8 2014 URL http www itu int en 3 Goldsmith A 2005 Wireless Communications Cambridge University Press first ed 46 48 p 4 Hoydis J Hoek C Wild T amp ten Brink S 2012 Channel Measurements for Large Antenna Arrays In International Symposium on Wireless Communication Systems ISWCS 2012 pp 811 815 5 The Essentials of Vector Network Analysis From a to Zo Anritsu Company 2007 6 Payami S amp Tufvesson F 2012 Channel Measurements and Analysis for Very Large Array Systems at 2 6 GHz In Sixth European Conference on Antennas and Propagation EUCAP 2012 pp 433 437 7 Ranvier S Kivinen J amp Vainikainen P 2007 Millimeter Wave MIMO Radio Channel Sounder IEEE Transactions on Instrumentation and Measurement 56 pp 1018 1024 8 Abouraddy A amp Elnoubi S 2000 Statistical modeling of the indoor radio chan nel at 10 GHz through propagation measurements I Narrow band measurements and modeling IEEE Transactions on Vehicular Technology 49 pp 1491 1507 9 Kim M Konishi Y Chang Y amp Takada J I 2014 Large Scale Parameters and Double Directional Characterization of Indoor Wideband Radio Multipath Channels at 11 GHz IEEE Transactions on Antennas and Propagation 62 pp 430 441 10 Fleisch D 2008 A Student s Guide to Maxwe
78. t number of points SP ene Chay IFBW measurement intermediate bandwidth in Hz bandwidth TT sui Cal Set name of the calibration pool suis used for channel 2 sour pow pl source power used in channel 1 power in dBm s sour_pow_p2 source power used in channel 2 power in dBm s number of receiving antennas to nRx ee integer x direction number of receiving antennas to 5 nRy KUUN integer y direction number of transmitter antennas nT x ee integer to x direction number of transmitter antennas nTy N S integer to y direction antenna spacing in antenna spacing meters _lambdas Appendix 6 The flow chart of the main program controlling the measurements Start program v Create Result User defined Inputs Initialize result Folders matrices y Write moving Initialize VNA I amp Open TCPIP objects gt programs to steppers U Find stepper origins and drive them home teppers are read for the VNA sweep Make one VNA Save sweep Measure S gt results parameters v Decite movin Decite movin ee a orcs nee direction direction v N Move RY stepper Move RX stepper one step one step Increment Increme
79. terface were the most important properties Other properties such as step resolution and torque power were also parameters to be considered The chosen stepper motor was Lexium MDrive LMDCE572 with Ethernet interface manufactured by Schneider The chosen linear stage was HIWIN KK6005P 600A1 F4 To be able to move both RX and TX antennas to the horizontal and the vertical directions 4 stepper motors and linear stages were needed The specifications of the motor and the linear stage are represented in Tables 2 and 3 The hardware manual for the stepper motor is found from 28 and the programming reference is found from 29 The manual for the used linear stages is found from 30 The function of the limit switches is to limit the motion range of the motors such that it knows when linear stage is at the maximum or the minimum positions The limit switches prevents the motors to push the carriage towards the boundary of the stage and hence prevents the motors to miss steps as well as broke themselves The 32 Table 2 Specifications of the selected stepper motors model Schneider LMDCE572 NEMA 23 microstep resolution 51200 microsteps rev general purpose interface and Ethernet interface TCP IP Ethernet IP ModbusTCP programmable interfaces memory RAM programming language MCode maximum voltage 48 VDC maximum holding torque 0 86 Nm maximum required power supply current 3 5A Table 3 Specifications of
80. the initial parameters for the VNA was chosen to be programmed to the VNA by the developed measurement control software instead of entering them by hand before the recordings By doing this we can ensure that all of the VNA settings are correct for each measurement We implemented three different controlling subroutines for VNA that can be utilized in the main measurement program The first one is the initialization In the initializa tion the VNA is set to a right measuring mode with right initial parameters Before using this we have to make calibration files available into the device memory First we delete all channels from the VNA memory to ensure that no uncorrect parame ters would be measured Then we choose the data transferring mode to be 32 bit real 37 number in binary format The amplitude and the phase measured will be transferred in binary format so that even numbers are the amplitudes and the odd ones are the phases respectively Then we create two measurement channels with two traces and select the sweep mode to be single for all the traces The analyzer can only sweep one channel at time so making two channels with their own calibration pools can decrease the mea surement time Third we choose the correct calibration pools for both channels to be used for the measurements The calibrations pools are automatically set to all of the traces in the channels After the calibration pool selection we set the correct settings for the
81. toolbox can be used to debug device specific languages such as machine code MCode and standard commands for pro grammable instruments SCPT to the devices MATLAB can also be used as a post processing tool for the measurement results and is therefore an versatile programming environment for our purposes 32 35 Figure 6 The measurement setup in operation 3 5 1 Controlling the stepper motors The stepper motors were connected to the laptop via Ethernet cable as presented in the Appendix 1 The Lexium MDrive LMDCE572 stepper motor understands MCode programming language that can be debugged to the device via different programming environment When the right TCP IP object is created to the MATLAB using right IP address with a correct port number the used stepper motor can directly be used under a TCP IP protocol via MATLAB The port number for MCode TCP option for LMDCES72 is 503 The command reference for the stepper motors can be found in 33 When communicating with the stepper motors trough MATLAB we have to make sure that the debugged commands are queued correctly so that the whole program is ex ecuted properly without exceptions Instead of debugging the commands to the device one by one we chose to use stepper internal random access memory RAM to store the programs to steppers beforehand The programs are stored to the stepper motor memory spaces to a certain memory addresses and executed from the device memory by only tellin
82. unt of time In massive MIMO systems where both ends contains hundreds of antennas the increase of amount of measurement time is multiplicative with respect the number of anten nas Furthermore increasing the number of antennas increases the amount of recorded measurement data One key design principle of the designed measurement system was to do the system as versatile as possible so that it could be applied to several channel measurements in the future We wanted to parameterize the software so that the system supports various measurement options chosen by the user In this chapter we introduce the measurement system for measuring the MIMO radio channel at 10 GHz VNA based measurement setup with virtual planar antenna arrays in the RX and the TX ends is presented The programming of the devices is introduced but the actual MATLAB implementations are not included to this thesis However some flow charts of the measurement control software are found from the Appendix 3 1 VNA based measurement system There are only few good ways to measure the radio propagation channel When mea suring a MIMO channel channel sounders are usually a useful devices to be used However there are few drawbacks limiting the usage of the channel sounders First drawback is the prize of the apparatus The commercial channel sounders are expen sive and are not always easily available for high freguencies Second drawback is the synchronization of the RX and TX c
83. urement system As mentioned before increasing the dynamic range of the system by averaging several sweeps or using narrower IF bandwidth in the VNA increases also the measurement time Thus using external amplifiers may be necessary to compensate the cable loss and keep the received signal above the RX sensitivity 5 Using external amplifiers causes two problems As mentioned in the Section 3 2 3 the additional components in the RX chain increases the noise of the measurement system Especially using external LNA in the RX front end increases the RX noise by its noise figure N Fina The overall noise factor noise figure can be calculated by using Equation 47 Also the received signal should always be above the LNA s own noise i e the sensitivity of the LNA should be adequate for the smallest received signal level If the external LNA is used it has to have better noise figure than the VNA s own RX in order to increase the sensitivity of the whole system Thus the sensitivity is not improved directly by the amplification of the LNA If the LNA is placed directly after the RX antenna to compensate the long RF cable it is useful The second problem is how to include the amplifiers into the measurement calibra tions in such a way that they will not damage the devices during the calibrations The calibration problem can however be overcame by adding the amplifiers to the RF chain after the calibrations and canceling them out from the measureme
84. y succesfully microsteps defined origin offset Figure 7 Flow chart of the stepper origin setting The subroutine writing the stepper movement programs is presented in Appendix 3 We command the stepper to start writing a program from specific address The stepper has its specific programming mode which can be used by command PG Before the movement we set the initial speed V maximum speed Vm and acceleration A to be suitable Too large values for speed and acceleration would overload the stepper causing missed steps and heats up the stepper motor The load and run currents Hc and Rc are set to 50 and 100 respectively Then the relative movement is performed into wanted direction and the program is saved into the device RAM memory The program can be executed from stepper memory by command EX lt address gt 3 5 2 Controlling the VNA The VNA was connected to the laptop via Ethernet cable similarly to the stepper mo tors The connections are presented in the Appendix 1 For controlling the VNA standard SPCI language was used with some external commands offered by Rohde amp Schwarz The commands can be debugged trough MATLAB in the same way as it was done in the case of stepper motors However here waiting commands WAT after the VNA sweep was used instead of writing the program into the VNA memory This was because there were no need to do other tasks with MATLAB while the VNA is per forming the measurement Most of
85. y are also used to define the maximum distance that the carriage can be moved in stage The specifications of the used limit switches are presented in Table 4 The wirings of the stepper motors and limit switches are presented in Figure 33 5 The manual for the used proximity sensors is found from 31 The settings used in the voltage sources of the XY gantries are presented in Table 5 Table 4 Specifications of the used limit switches model GX F12A P VT16061C switch type inductive mechanical stable sensing range 0 3 3 mm 0 K needed output operation normally open normally open PNP open collector axis 0 04 mm output E assive P trans stor P along sensing axis repeatability perpendicular to sensing not needed supply voltage 12 24 V can be chosen up to maximum AC 250 V maximum source current 100 mA 16 A Table 5 Settings of the voltage supplies of the build antenna spacer main voltage 30 VDC main current limit 2A limit switch voltage 12 V limit switch current limit 150 mA 3 4 Wiring of the measurement system The wiring of the overall measurement system including Ethernet routing RF cables and stepper voltage source wiring is presented in Appendix 1 Two Ethernet switches were used to route the connections between the laptop and devices Each stepper motor and the VNA has its own specified Internet Protocol IP address which can
86. y can be measured by VNA by connecting the antennas to the VNA ports S parameters can be defined theoretically for 2 port meaning a box with an input and output Let us denote the incoming wave to the input port as a and wave seen at the output port as a2 Furthermore let us denote the wave 24 reflecting back towards the input and output ports as b and be respectively For the S parameters we can write by a2 S11 and S91 34 ai ai Similar coefficients than 34 can be defined for wave coming to the output port S21 can be referred to be the transmission coefficient of the 2 port and S11 as the reflection coefficient of the input port S parameters can be expanded for n port as they were defined for 2 port 11 3 1 3 VNA time domain analysis S parameters are usually presented in the frequency domain The transition from fre quency domain to time domain can be done via inverse Fourier transform There are two possibilities to get impulse responses out from the VNA Many analyzers allow to measure the impulse responses directly in time domain which are also called as time domain S parameters by VNA vendors However we decided to measure the parameters in frequency domain and transform them into time domain via inverse dis crete Fourier transform IDFT Let us denote Sim fn as the frequency domain S parameters where and m are the port indices and f is the n th recorded frequency sample By the IDFT impulse responses A

MASTER`S THESIS

Contents

Download Pdf Manuals

Related Search

Related Contents