Home

Modelling, Simulation and Implementation of an Optical Beam

1. 19 A 2 2 Pre calculation scripts API documentation 78 B OBFN Simulator Documentation 81 EI General information s sa se s sa saor ro Re Wo OR Ro 81 B2 Manual ec E UR RA Om RR OAS E ORS CR 81 B21 Touroftheinteriace lt s e464 48 0 ati 4 aS 81 B 2 2 Usage example 1 setup anew OBFN simulation 85 B3 APidocumentation Lu ea s eea nama ded eed Rb OSES EER ES 86 C Hardware Controller Documentation 89 CI General informati n a ca WX RR EERE ARR ed s 89 C2 Virtual COM port driver ee es 89 Using the Sider tool use qo Sk ee 90 CA Using the Debug tool 242 4 hd eee ewe RE X RR RR 90 C 5 Flashing the micro controller llle 90 C6 Floating point 90 D OBFN layout 91 E Ring channel and waveguide data 93 Rowley Crossfire Licenses 95 VIII CONTENTS List of Figures Ll 1 2 1 3 1 4 2d 2 2 2 3 2 4 2 5 2 6 2 7 2 8 2 9 2 10 Z1 2 12 2 18 2 14 2 15 3 1 3 2 3 3 3 4 3 5 Cobra Dane Radar se pa cd ke oe othe Rowe cx Bow System overview Schematics for a 8x1 OBEN chip 3 ring combined output 2 Data Flow Diagram DFD for the delay element simulator Compensated s
2. 3 3 4 Connectivity The simulator can be connected to the hardware implementation to verify the calcu lated ring setting using the real measurement setup Two methods are implemented each having their own advantages For both methods the commands to set a channel to the appropriate value can be send sequentially several wrch commands or in bulk a single wrchall command A switch button is provided to choose one of the two options Connect to the Java debug tool through TCP The first method to connect the board to the simulator is through a TCP internet connection For this to work the debug tool needs to be running See Chapter 4 and Appendix C for more information By default the tool is listening on TCP port 4567 for incoming connections The LabVIEW simulator can be set to connect to port 4567 on a specific IP address The address can be either localhost or a real IP address of some remote computer on the network or internet When all set the simulator tries to make a connection to the debug tool which handles all requests Replies from the controller board are sent back to the debug tool but not to the simulator In our experience this way of communicating seems much faster than the next method but is a bit more complicated to set up 3 4 SIMULATION RESULTS 41 Direct connection to COM port This is a more direct approach but also noticebly slower compared to the method just described The big advantage is the im
3. Tune ring j to 27 The phase of ring i now changes due to the change of ring j and is independent of other rings which are at 0 volt Compensate the voltage of ring i from V to V so it is back to its original position of 27 using the slider panel Convert the compensation voltage to a proper value using Equation 5 6 Store the value on the i j position in the matrix The conversion in step 9 for the effect from 7 on is done using the following equation the index of the proper ring property of Table 5 4 is shown between the parenthesis A t j SS ae ofa 5 6 AV ion J where AV V2 V Incorporating the crosstalk effect on Equation 5 2 now becomes TRUNG P Pot fset 5 7 And the output voltage Vee At Paesired Pop tect 5 8 5 4 2 Results Our measurements have been limited to a 4 x 1 subset of the full 8 x 1 OBFN Only channels ind in6 and in7 were used containing in total 3 ORRs and thus having 6 heater elements The crosstalk matrix has been determined and is shown in Table 5 5 Diagonally the a values appear which were previously calculated during the and amp calibration process A value in position 7 7 means the effect of j on i Doing a left multiplication with a column vector of uncorrected values results in having a column vector with the corrected values 66 CHAPTER 5 MEASUREMENTS 1 3 5 2 4 16 0 010596663 0 0 000618814 0 0 0 0 00
4. d 8 87 65 1 normalized group delay relative to urit delay multiply T to get real delay Power 10 7997 Bi 0 100 200 300 400 500 600 700 800 900 1000 1100 freuen Normalized delay CEN lg ter Soie dispersion 0 1453 8 0 256188 0 E o Normalized frequency m 300 400 so 600 700 1023 Ring control Load Save Values array Settings Cakulate Approximate Frequency Hem Select ring Phase Length 00 ua J do 0 25 0 1 6 3 1 E NISI Tum os fos os p us y m 7 0 8000 pole 075 7075 0 257 7175 Sy ED j c pica rim B 0 6000 En 3 b E Redraw 15 g Loss B ch T D ETE oo IE eg d e 4 02000 024 08 Normalized bandwidth ost 2s i orm 2 2 ss gous 0 3 o 1 Bic do 79113 0 100 200 300 400 soo 600 700 80 Jack van Galen Universiteit Twente Frequency Ham Ld ed Figure 2 6 The OBFN Simulator at startup 2 3 2 An approximation algorithm An analytical solution to acquire the ring parameter values and according to the mathematical model presented earlier is not possible since the number of unknowns in the model of the delay element is greater than the number of equations An estimation is necessary but requires a lot of brute force calculations to be done To overcome the burden of calculating optimal ring parameters on the fly an app
5. 4 3 5 1 5 2 9 3 5 4 5 5 5 6 5 7 5 8 5 9 5 10 5 11 5 12 A l B 1 D 1 Simulation results for an 8 x 1 OBFN width a of 0 5 equivalent to an AOA of 53 42 Simulation results for an 8 x 1 OBFN width a of 0 5 equivalent to an AOA of 53 degrees 42 Architecture of the control 46 The slider tool created by 12 is able to control the voltage for each channel of the OBFN individually 47 The three scenarios showing the responsibilities of the PC and micro controller given a AOA uuo uoce Sow Ge wR RE RS Gow wR Ew G 50 The inside of the styrofoam 54 Crosstalk measurement setup 54 Labelling of all heaters inputs and outputs of the optical chip 56 Overall stability of one ring r1 measured at a 2 minute interval for 1 OME sss et Ru E Be URS ee Se WO Vs 57 Resonance frequency drift of one ring rl measured at a 2 minute dor LAO se cossa ek xo 909 E ow eS 57 Maximum delay drift of one ring r1 measured at a 2 minute interval POP LOM oec Wh Sow wo Ble WS eRe Oe 58 Voltage differences for channel 1 0 000008 60 Voltage differences for 12 60 Kappa Voltape relation 564424444448 ko o Rx RR oA 62 Single ORR response for
6. answered and conclusions will be formulated e Suggestions for further research Indications in which further research could be directed are pointed out 1 4 Thesis organization This chapter will start with a brief introduction to OBFNs and provides some back ground information Also a short motivation why this research project is of great interest is given This chapter provided an introduction to the project its technology related work and background information a motivation and finally the organization of this research The rest of this thesis is presented as follows In Chapter 2 the design implementa tion and preliminary test results of a ring section simulator are given In Chapter 3 the development of a complete OBFN simulator is described Chapter 4 consists of implementation details concerning the porting of parts of the software to the hardware environment In Chapter 5 measurement setup and results are presented Finally in Chapter 6 conclusions and suggestions for further research are given CHAPTER 1 INTRODUCTION Chapter 2 Design and Implementation of the Delay Element Simulator The system design as described in the introduction of this thesis can be broken down into smaller pieces The ring itself must be simulated A group of rings which we will call a delay element DE has specific properties and needs careful attention while modeling Finally the OBFN a structured combination of several DEs an
7. 3 3 3 Dealing with offsets and negative AOA Due to the fact that the physical realization of the optical chip is based on an asym metrical binary tree topology negative AOAs cannot be achieved without further ad justment of the system as a whole since a negative AOA would mean a negative Ar and the ORRs can only create positive delays To solve this problem extra fixed delay lines have to be prepended to the optical chip This could be done using coaxial ca bles Basically a situation is now created where the zero AOA delays are lifted from a normalized 0 to some value x This way AOAs from 60 degrees to 60 degrees can be achieved despite of the physical limitations of the optical chip In the simulator the delays of additional coaxial delay lines can be set Given the frequency range of the system in Table 1 1 it follows that the antenna element spacing 2 should be in the order of 1 5cm 9 When combined with the AOA a ranging from 60 to 60 degrees the delay between two neighboring elements should be tunable from 7 to Using Equation 3 1 and given the speed of light in air c 3 10 and the AE spacing d 0 015m we can calculate the tuning range to be roughly 2 x 40 80ps For a 8 x 1 OBFN containing a total of 8 rings as displayed in Figure 3 5 and using a RTT of 80ps note that this value is not related to the tuning range the coaxial delays are calculated as follows Although the normalized delays of each ORR ca
8. AUR GEOR me eR AU Eos 23 2 3 5 24 290 Matlabsaiptsand APT ao co sack ee Ren 25 s Manual Dc 26 2 5 Summary and conclusions 26 VI CONTENTS 3 Design and Implementation of the OBFN simulator 3 1 Requirements As DEV simulator dosin e ees Bee Ox ERE os acd NUS ae lt e riscs ee hioi ek aba m ox Y erac A oe Datalow and siruct re sex scorse s vonie 5 oo simulator Implementation lt o s eacee Rx RE Al CECI MAIR 2 Se Rex ue C RA 3 3 2 Delay distribution within the OBEN 3 3 3 Dealing with offsets and negative Angle of Arrival AOA QR QOMT uua eR Wo Eo eed 240 Complexity und upscaling 2 seste e654 84 RR es 50 Pmt WEB c e coes semi ew ee Ex doe Ee Xu suede ed UM oo Pr 26 Summary and comclusi ns a c eraras bwa ca aa TR HES 4 Design and Implementation of the Microcontroller Software 4 1 Overview of the control system gt sorisa a2 Controller soitware PO Lus cor Sh weg ow e Yom OS Al dur SO uou voy eS daa Debug woke ec EUR SR re ee p ege A23 SOM iu zd md xy 40k RA ER ER BSP AGE Regn 4 3 Controller software microcontroller 4 3 1 Implementation ek use xor Rom RR ee 4 3 3 Ha
9. TELECOMMUNICATION UNIVERSITY ENGINEERING of TWENTE University of Twente Faculty of Electrical Engineering Mathematics and Computer Science Chair for Telecommunication Engineering Modelling Simulation and Implementation of an Optical Beam Forming Network Control Software System by Jack van Galen Master thesis Executed from July 2008 Mei 2009 Supervisor Dr Ir C G H Roeloffzen Advisors ir L Zhuang M Burla MSc Dr ir P T de Boer Summary Beam shaping and beam steering together called beam forming is needed when pro cessing the radio frequency signals received by a Phased Array Antenna PAA When correcting the arrivaltime differences between all the inputs of the PAA by adding small delays and subsequently combining them a strong signal can be obtained At the Chair for Telecommunication Engineering TE at the University of Twente re search is done on achieving these delays fully in the optical domain using an Optical Beam Forming Network OBFN With an OBFN very large bandwidths can be delayed continuously and is thus suitable for high bandwidth applications like live television reception The OBFN is based on thermo optical tuning of Optical Ring Resonators ORRs where each ORR is capable of delaying a small fraction of the bandwidth of the signal continuously The exact frequency range and the amount of delay are controlled by applying a voltage to small heater elements on top of the ORRs accordi
10. J van t Klooster Context design and implementation of a control system for ring resonator based optical beam forming networks MSc Thesis University of Twente October 2008 73 TA BIBLIOGRAPHY Appendix A Delay Element Simulator Documentation A 1 General information The LabVIEW delay element simulator has been built using LabVIEW 8 5 on a Mi crosoft Windows XP Professional operating system The simulator depends on sev eral Matlab routines that were developed in Matlab 7r14 so a proper installation of Matlab is also required For LabVIEW to find the proper Matlab routines Matlab should have a path reference file menu set path pointing to the directories sim ulator delayelement and coefficients lookup table Due to caching of code changes in MatLab code are not immediately effective in LabVIEW The best way to circumvent problems related to cached code is to completely restart LabVIEW A 2 Manual This section will describe the delay element simulator from a users perspective A 2 1 Tour of the interface A complete overview of the interface is given in figure A 1 Going clockwise starting at the top left we see the output window showing the normalized group delay versus frequency the power versus frequency the dispersion versus frequency and finally the phase versus frequency Note that all frequency axes are normalized The main output window on the top left shows the group delay responses for e
11. Sweeptime Figure 5 12 The response of measurements on a 4x1 subset of the 8x1 containing 3 ORRs r1 r2 and r3 The AOA has been set to 30 and 60 degrees equavalent to a Ar of 0 36 and 0 62 respectively The crosstalk correction has been included redetermined 70 CHAPTER 5 MEASUREMENTS Chapter 6 Conclusions and Futher Research This final chapter presents some conclusions and directions for further research and new that questions and ideas arose while doing this project 6 1 Conclusions The main reseach goal as stated in Section 1 3 1 was the creation of a maintainable and scalable software control system that can automatically tune all the parameters of an OBEN given only the direction of the incoming beam To achieve the goal two simulators were written in LabVIEW to see if the under lying calculations would work in theory The first of the two simulators is specifically designed to simulate the group delay response of delay elements with a variable amount of rings The settings for the rings were aquired by using an approximation algorithm with precalculated values The effects of a change of on and the loss compensation by a change of amp have been incorporated The end result is a scalable simulator capable of simulating delay elements containing a variable amount of rings The second simulator was an additional layer around the code of the first simula tor thereby creating a tool that can simulate an e
12. the height of the peak is shown in Figure 5 6 Although the fluctuations are small a clear trend emerges The maximum slowly climbs from 0 1142 to 0 1157 creating an additional delay of 0 0015ns in the course of an hour During the rest of the measurements it is assumed that this small change does not significantly influences the final results In the past there were difficulties in obtaining this level of stability A light breeze would affect the system enormously That is also the reason why it is wrapped in a styrofoam box When measuring the output response of the next channel some optical fibers have to be rewired This rewiring means opening the box which can influence the measurements To minimize this effect all three channels in5 in6 and in7 were connected at once and a selection of the channel under test was done by setting the combiners appropriately see Table 5 1 and the proper coax cable was connected to the signal generator The combiners were adjusted so that all other paths were 5 2 STABILITY AND VOLTAGE LEVELS 5T 0 14 0 1 0 08 ns Delay 0 06 0 04 002r 0 35 0 4 0 45 0 5 Sweeptime Figure 5 4 Overall stability of one ring r1 measured at a 2 minute interval for 1 hour 0 6 F o a T Sweeptime of maximum delay o o N w T 0 1 F 0 1 1 1 0 10 20 30 40 50 60 Time minutes Figure 5 5 Resonance frequency drift of one ring r
13. Matlab including documentation This section describes the steps taken to design the simulator starting with the initial design the flow of data and the software structure 3 2 1 UML model A Unified Modeling Language UML model of the core of the OBFN simulator is shown in Figure 3 1 In this layered architecture 7 we can clearly see the aggregation relationships of the separate components that model the optical chip An OBFN consist of multiple delay elements These delay elements and their functioning are described in Chapter 2 Each delay element in its turn consists of multiple rings In theory and for completeness both the OBFN as well as the delay element class can consists of zero child nodes where a child node is either a delay element in case of the OBFN or a ring in case of the delay element Each delay element is contained by exactly one OBEN and each ring is contained by exactly one delay element At the bottom of the figure the Path class is shown Since a delay element can be part of multiple paths from input to output within the OBEN there is a many to many relationship So each path can consists of zero or more delay elements and each delay element is contained by zero or more paths Structuring the OBFN like this corresponds greatly to the real world chip and gives us some advantages later on 3 2 2 Dataflow and structure Figure 3 2 shows the dataflow of the OBFN simulator Notice that the delay element simula
14. The Min Max delays matrix holding the minimum and maxi mum delays values for delay elements of a specific length e connectionmatriz The connection matrix representing the OBFN structure B 3 API DOCUMENTATION 8T e numberofinputs The number of paths in the OBFN Return values e delays An array containing the individual delays for all the delay elements in the OBFN In combination with the connection matrix this gives sufficient information to calculate the ring settings parameters 88 APPENDIX B OBFN SIMULATOR DOCUMENTATION Appendix C Hardware Controller Documentation C 1 General information This appendix provides additional information for a proper installation and setup of the hardware controller C 2 Virtual COM port driver To let Java communicate with the COM port an operating specific package needs to be installed For Microsoft Windows instructions are given below 1 Get the Java Runtime Environment from http java sun com If you would also like to edit and recompile some of the software on of the development kits is needed that includes the Java compilerjavac for example the Java SDK 2 Two files located on the CD that accompanies this thesis needs to be copied to the proper Java directories Note that if you are installing an SDK more than one Java directories will be created Check your Windows path global variable settings to see which directory is actually used Then copy the
15. a similar plot is shown but now a delay element containing 2 ORRs As we can see the combined curve is very close to the required delay of 0 5 For a delay element containing only 1 ORR the smaller than 1 delays are not achieved by shifting the curve gradually In stead is shifted by 7 and the is lowered thereby creating higher peaks to the left and right of our region of interest As a side effect the response between the peaks lowers and thus creating a delay smaller than 1 This is a direct consequence of the optimization process of the NLP solver Several attemps to adjust the boundaries of the parameter space did not change the results 2 3 5 Alternative approaches Although the methods described in this chapter work perfectly alternative approaches should be investigated This section briefly describes an alternative approach that could be further investigated in a future research project Other error functions In stead of the current MMSE method another measure of error could be provided to the NLP solver For instance one could calculate the maximum error in the function for the entire frequency range in stead of the MMSE currently used Now u becomes max Vita fo fren 2n D fo frrn 2 13 The computational complexity of this way of solving for the unknowns would likely 2 3 DELAY ELEMENT SIMULATOR IMPLEMENTATION 25 Resonance frequency Response ring 1 Response ring 2 Comb
16. a very simple tool based on the slider tool shown in Figure 4 2 The GUI implementation was replaced by a simple Command Line Interface CLI which could easily be created thanks to the Model View Controller MVC design of the slider tool In a MVC design there is a strict separation between the User Interface UI and the rest of the code The debug tool allows us to send arbitrary commands to the microcontroller similar to a standard Hyper Terminal connection but offers additional functionality For instance the pos sibility to create test cases and adding a proper delay between the commands that are 48 CHAPTER 4 DESIGN AND IMPLEMENTATION OF THE MICROCONTROLLER SOFTWARE sent Finally the debug tool can be used to receive commands from the LabVIEW simulator and send them to the connected hardware controller board Not only do we obtain higher speeds but also better logging capabilities of the output of the mi crocontroller Apparently LabVIEW only collects the beginning of large responses despite of any buffer settings Information on how to use the debug tool can be found in Appendix C 4 2 3 Configuration Both the slider tool as well as the debug tool have several configuration options Below there is a short list of available settings that can be added or changed in the settings zml located in the appropriate directory of either tool e Nr of bars Slider tool only The number of bars to display The number of bars
17. are described For all tests the 8x1 SMART optical chip was used 5 1 System overview Due to the fact that we are working with high frequency optical waves a small change of for instance temperature could have a large net effect on the measurement results For that reason a lot of measures have been taken to minimize the influence of fluctuations in the room temperature and fluctuations due to the heaters on the chip itself The optical chip is mounted on a thick copper plate that is kept to 30 degrees centigrade using a Peltier element A water cooled plate beneath guides away all the heat to a remote location where a fan cools the water The system itself is placed on a very stable table Finally to prevent airflow in the room to change the local temperature the entire system is enclosed by a styrofoam box A photograph showing the inside of this box is shown in Figure 5 1 On the left we see some modulators The optical chip is positioned on the right 5 1 1 Measurement setup All measurements were performed using the same setup To be able to compare new result with previous ones the measurement setup used to calibrate the chip and de termine the crosstalk is taken from 12 The schematics of the setup are shown in Figure 5 2 The laser is connected to the current source Curr and a Temperature Controller TEC The optical chip acting as the Device Under Test DUT is con nected to the controller board and a temperature stabilize
18. as function of loss per round trip and maximum delay Different loss compensated responses Schematic of the ORR with the Mach Zender interferometer Belanod bebwesn n uude 6 9 30 X00 Ro RR ARR Screenshot of OBFN simulator at startup Group delay curve for 3 cascaded ORRs The shaded area denote the parts that are added to the total costs Phase response plot with normalized frequency for 0 8 and 0 Curve litted polynomial hE ee diis Fitted polynomials for a delay element with 2 ORRs for a normalized uud us kb ee E bene Fitted polynomials for a delay element with 2 ORRs for a normalized Prem oi 1018 ne ka Kaeo eed bea ed RISE Response for a normalized delay of 0 5 for a delay element containing 1 609 V3 ee OE He Response for a normalized delay of 0 5 for a delay element containing 2 one we eee ee a ee ode ee we eee Alternative objective function Matlab polynomial coefficients data structure UML model for OBEN simulator 22 444 43 ok Soka ks Dataflow for the OBFN simulator A general uniform linear PAA Depth First Search DFS walk of the 8 x 1 Coaxial delay lines prepended to the OBFN ix 4 Ae 11 13 14 14 16 17 LIST OF FIGURES 3 6 3 7 4 1 4 2
19. environment to see if the approach taken could work The first measurements using the voltages ill IV SUMMARY calculated by the control system look very promising Also the system is capable with very small adjustments of tuning future chip designs or using other tuning methods than thermo optical Contents Summary iii Abbreviations xiii 1 Introduction 1 LE MoUU e a a cae oh a eee ee E 1 12 22 e scce rades ER ES Rode op CC de EKER Eun 2 121 SMART 2 53x99 ERASE ERE ede 2 L22 Overview nus ch hw eee xo 3 1 2 8 Optical Beam Forming Networks OBFN 4 LS Research organization 2 co s ue Rogo oe da wok e 6 Doo c es ok a a A a ES Oe 6 1 3 2 kn ce saias epicu KEE ERE e ES ES 7 TA me eo kn RED EOS ES RG RA 7 2 Design and Implementation of the Delay Element Simulator 9 21 TRS Re eee eS eR RES 2 2 Delay element simulator design 10 Zal Daallo sock ok hw OE ORS Se 10 PA RE i lo DLL 15 2 3 delay element Simulator Implementation 16 2 91 Graphical user interlase s esoe s saca 16 2 3 2 An approximation algorithm 17 Z5 DBonnalbisHoH 42 2 44 x omo EE SE EEE 18 04 SB dS es osx RR
20. in 1976 the radar uses a ninety five foot phased array antenna and provides 120 degree coverage of a two thousand mile corridor that spans the eastern Russian peninsula and the northern Pacific Ocean PAAs are usually flat or slightly curved and consist of a number of Antenna Elements AEs that act like individual micro antennas Normally a PAA is elec tronically steered using some form of controller The ability of an antenna to steer and focus a beam to a specific target is a huge advantage over other kinds of antennas When using electronical steering there are no moving parts and thus wear and tear is vastly reduced To use the PA As effectively the time differences due to arrival delays of the signal between the different antenna elements should be corrected by some clever control system After that the signal can be combined resulting in a signal with a high Signal to Noise Ratio SNR This signal can then be used for any suitable application This thesis consists of two parts The first presents a design and implementation of a control system simulator that is capable of compensating the arrival time differences of all the AEs The second part describes the implementation of a functional prototype which is then subjected to a set of measurements 1 1 Motivation At this moment a working prototype of an antenna system consisting of a PAA for signal reception and an OBEN for combining the signals is being developed The OBFN is con
21. matrix contains the minimum and maximum delays for delay elements containing a specific number of rings and is used to calculate the Input min maz delays when pressing the Recalculate button On the right of the tab the coaxial delay offsets can be set These are the offsets as discussed in Section 3 3 3 that compensate for the missing rings when tuning for angles smaller than 0 degrees Tab Ring settings The ring settings tabs contains all the calculated ring settings for all of the rings in the entire OBFN For now Active and Length are always 1 On the right of the ring settings two arrays of offsets are visible The phase offsets for either and coupier as discussed in Chapter 2 must be set here These values will be used to calculate the proper voltages to send to the controller board Tab Mapping The LabVIEW simulator labels each of the heater elements in a first come first serve fashion This could result in wrong commands being sent to the controller board For this reason a mapping can be applied to the heater elements used in the simulator The first column represents the number used in the simulator while the second column denotes the channel number used in the hardware setup Tab Voltages The Voltages tab contains the crosstalk matrix as discussed in Chapter 5 The matrix has to be filled in manually The simulator uses the crosstalk matrix to calculate the final output voltages displayed on the right side of the
22. middle part and choose a delay The curve will change accordingly 4 Current ring settings can be saved for later use in the Load amp Save tab Other settings can be saved using LabVIEW s own save functionality in the Edit menu scripts and API documentation The coefficients that are used to calculate the ring settings for a specific delay are pre calculated using some fully detached scripts The case of the 2 ring scripts is described here The scripts for more rings are comparable Although Matlab can more or less be used as an Object Oriented Programming OOP environment the scripts are basic procedural scripts for the sake of easily testing and changing code The precalculation program has several files that contain the following functions calcall This scripts can be run by setting the path to the appropriate directory and issue the command calcall Within the script the bandwidth range and the delay range can be specified The script will start to calculate all the settings for the whole bandwidth range and save the results to allresults2 mat Next an appropriate structure as shown in figure 2 15 is saved in a file named ringsettings_2_aboveone mat This file is required for the lookup table which we will discuss later Parameters e No parameters just run calcall Return values e File allresults2 mat file containing the entire workspace e File ringsettings 2 aboveone mat f
23. values Curve fitted polynomial of degree 3 09r 0 85 Kappa 08r 0 75 07r 1 1 1 1 1 1 5 2 2 5 3 3 5 4 Normalized delay Figure 2 9 The combined result of multiple calculations for of a 1 ring delay element The solid line represents the curve fitted polynomial of degree 3 Note only every fifth element of the calculated values is shown to avoid clutter rings the combined results seem to form a more or less smoothly decreasing line see Figure 2 9 This line is easily traced by a curve fitting function in a mathematical software tool such as Matlab As a result we are left with a curve fitted polynomial function describing the delays versus and for the number of rings we want to process Figure 2 9 To find the appropriate parameter values for and all we need to do now is fill in the blanks in the new polynomial The polynomials are solely described by their coefficients with the notion that the degree of the function is the number of coefficients minus 1 and that every term is used only once The polynomials will however be less accurate when they reach the beginning and end of the range due to the curve fitting procedure When the required delay is too large the ripple will become too large and the final delay is too much off Of course the amount of error a system can cope with is application dependent and thus a suitable suggestion cannot be given in general Therefore fo
24. values of channels 12 13 and 14 to 1200 1300 and 1400 centivolts respectively the separates the command name from the parameters The parameters are separated by a and the key value pairs are separated by an symbol In Listing 4 1 an example XML file is shown for the debug tool 4 3 CONTROLLER SOFTWARE MICROCONTROLLER 49 Listing 4 1 Example configuration file lt xml version 1 0 encoding UTF 8 standalone no gt lt DOCTYPE properties SYSTEM http java sun com dtd properties dtd gt lt properties gt lt comment gt Settings for console debug tool lt comment gt lt entry key COM gt 2 lt entry gt entry key commandparamsseparator gt lt entry gt lt entry key paramsassignmentsymbol gt lt entry gt lt entry key paramsseparator gt lt entry gt lt entry key resetdelay gt 500 lt entry gt lt entry key intercommanddelay gt 200 lt entry gt lt properties gt 4 3 Controller software microcontroller Apart from the software on the PC some software must run on the microcontroller to control the amplifiers This section describes several scenarios to gradually port the calculation process now done by the simulator to the microcontroller itself 4 3 1 Implementation scenarios There are several possibilites on where to make the separation of the functionality and responsibilities of the software on the PC and the microcontroller Three scenarios have been evaluate
25. would be 0 5 Since the production process of ORR on the chip is very reproducible in our model all values for for the directional couplers are assumed to be identical Using a heater on the upper line of 14 CHAPTER 2 DESIGN AND IMPLEMENTATION OF THE DELAY ELEMENT SIMULATOR 35r Tau max 0 dBloss 4 dBloss 3r 3dBloss 25 8 a 2 e 9 AN 15 c E 2 1 0 5 0 2 0 15 0 1 0 05 0 005 01 015 02 025 Normalized frequency Figure 2 3 Different loss compensated responses Dring Figure 2 4 Schematic of the ORR with the Mach Zender interferometer 2 2 DELAY ELEMENT SIMULATOR DESIGN 15 the MZI an additional phase shift is added to that branch which will effectively work as a tunable power coupler The now becomes 4 1 cos 2 2 6 Using equation 2 6 and the fact that the maximum value for is reached when the phase shift equals 0 degrees we can calculate the maximum value for to be 0 9951 with a minimum value of 0 for 7 2 Phi compensation Because a change of has a direct influence on the resonance frequency a compensation is needed to correct this effect Equation 2 7 is used for this compensation K naz is the value previously calculated and is set to previously determined maximum value for where m can be found by rewriting Equation 2 6 Kappa co
26. 0187465 0 009315864 0 0 0 0 15 0 000468741 0 0 010190925 0 0 0 2 0 0 0 0 0096158 0 0 40 0 0 0 0 0088222 0 16 0 0 0 0 0 0 0084136 Table 5 5 Crosstalk matrix The numbers on the top and on the left denote the heater numbers 5 5 Delay measurements To investigate if the simulator with all its calculations is working as expected the simulated results should match the output responses of the OBFN network Using the same system setup as before several measurements have been performed 5 5 1 Determining group delay offsets When a amp is tuned at a very low value the delay starts to go to infinity at the resonance frequency The delay to the immediate left and right of this peak is almost 0 which is confirmed by both the simulations and measurements When we tune down a little more the peak completely disappears in the output windows of the network analyzer This phase shift is used as a reference value to calculate the actual group delay in nano seconds This phase shift can be determined for every output which can be used for aligning the output responses We noticed that by physically moving the optical fibers the phase offsets changed quite dramatically so any rewiring of the cables to different inputs or outputs of the optical chip has to be done with great care 5 5 2 Single ORR As a proof of concept a measurement has been done for a single ORR r1 and thus 2 heaters The results are shown in Figure 5 10
27. 1 measured at a 2 minute interval for 1 hour 58 CHAPTER 5 MEASUREMENTS 0 125 F 0 115 ER cM Maximum delay o 0 105 r 0 1 0 10 20 30 40 50 60 Time minutes Figure 5 6 Maximum delay drift of one ring r1 measured at a 2 minute interval for 1 hour Ring Heater 1 3 8 9 15 18 1 coax 4 channel 7 1065 19 68 2 coax 1 channel 6 13 55 18 55 18 39 18 23 3 coax 2 channel 5 13 55 18 39 18 23 Table 5 1 Voltage levels used for measuring output responses by switching between coax ial cables outside the styrofoam box effectively disconnected and could not interfere Using this approach the box did not have to be openened which would in theory eliminate the impact of it on the systems stability However as further investigation showed the instability was not caused by the system itself but by heaters that were still set to a high voltage level thus creating crosstalk This situation was caused by a defective reset function that should have set all heaters to 0 volt Unfortunately this didn t always happen leaving the system in an unknown state The reset function was repaired and further measurements showed that the system was indeed very stable This means that future measurements do not have to be performed with the box closed at all times Optical fibres can be rewired within the box which will give better results and eliminates any interference effects 5 3 CHIP CHARA
28. 1 The control system takes care of applying voltages to the heaters on the optical chip and consists of a Printed Circuit Board PCB containing a general purpose microcontroller and one or more amplifier boards containing 32 amplifiers each The microcontroller can be programmed to control the output voltage of each of the 32 channels It contains instructions to set a specific value to a channel of a Digital to Analog Convertor DAC on one of the amplifier boards The outputs of the DACs are then fed into amplifiers that boost the values to appropriate voltage levels to power the heater elements The microcontroller used is a Rowley CrossFire LPC2138 equipped with a LPC2138 ARM7 Reduced Instruction Set Computer RISC microprocessor from NXP The programming environment for this microcontroller is called Rowley CrossWorks Stu dio where the actual flashing of the microcontroller is done using a Universal Serial Bus USB port After the flashing process the microcontroller can be accessed via a virtual Recommended Standard 232 RS232 or COM port of a PC In our case the virtual COM port runs over a standard USB port Instructions on how to flash the 45 46 CHAPTER 4 DESIGN AND IMPLEMENTATION OF THE MICROCONTROLLER SOFTWARE DAC PCB e e e e A gt eeccee Figure 4 1 Architecture of the control system with on the left the microcontroller and on the right the amplifier boards microcontroller are provided in App
29. 146946 0 15262847 Table 5 4 Coefficients for the d voltage curve fit 5 4 where e The voltage needed to obtain a phase shift of 1 FSR or 27 e AV Reference voltage needed to obtain a pre set reference phase This is to align the resonance frequencies of the rings The values for are obtained at a Kring of 0 7 The voltage for the particular for the ring under test is determined by using the data from the from the kappa calibration in the previous section All other phase and coupling heaters should be set to 0 volt for no crosstalk This does mean however that the calibration process is slightly different than the one that is used in 12 where the focus was on calibrating at the equal voltage levels for With the two measured values a linear function as a function of V is created as follows Pring a V ni Qf fach 5 2 where 2m a 5 3 Vin 63 and b a 5 4 The equation for the output voltage V for a no crosstalk situation now becomes Vou V Qaesired b a 5 5 The method described can be used in the case there is no crosstalk We will see later on that the system does suffer from crosstalk which needs to be corrected for The crosstalk does not have an effect on the characterization The values a and b can be entered in the simulator since they are chip specific See the manual in Appendix B for more information on how to properly enter the a and 6 v
30. 27 62 CHAPTER 5 MEASUREMENTS coupler Phi Ring r1 Ring r2 Ring r3 Curve fitted ring 1 Curve fitted ring 2 Curve fitted ring 3 0 100 200 300 400 500 600 Figure 5 9 Result for determining the values of kappa in relation to the applied voltage levels Ring number 1 2 3 rad V 0 0096158 0 0088222 0 0084136 b rad 0 26061 0 17725 1 6531 Table 5 3 Coefficients for the voltage curve fit of the tunable MZI see Figure 2 5 Changing the voltage gradually gave a certain range where a delay could be achieved from the physical minimum being slightly more than one round trip time to infinite This range is subsequently used This should however not pose any problems since each ORR is characterized individually and specific calculations of the voltage levels for within the simulator is also done on a per ring basis The plots have been curve fitted with a linear function resulting in a formula of the form mg V Tb 5 1 The values of a and b for the first three rings are shown in Table 5 3 5 3 2 Phi calibration In order to calculate the voltages some fixed values depending on the optical chip need to be measured The values and for the set of ORRs are shown in Table 5 3 CHIP CHARACTERIZATION 63 Ring number 1 2 3 AV 620 24 35 25 97 24 83 AV bref 0 23 23 3 87 a rad V 0 010596663 0 009315864 0 010190925 b rad 0 5 027
31. 3 44 1796 1630 1611 1593 1500 2111 1741 1741 1722 1537 1926 1241 1167 1444 1259 1722 1241 1241 1759 1796 1537 1944 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 wrch 54 0 Figure 4 2 The slider tool created by 12 is able to control the voltage for each channel of the OBFN individually microcontroller as the simulator described in the previous chapter A small change in syntax has been made to better generalize the command structure on which we will elaborate later on Furthermore some minor adjustments were made that makes it now possible to save settings to a file at any desired location and under any name Finally when loading a file a single reset command is sent to the microcontroller to assure that the slider settings coincide with the actual voltages that are being applied One problem in the existing system is the fact that commands from the COM port are arriving too fast This causes the microcontroller the choke thereby dropping commands and leaving the entire OBFN system in an unknown state That is the sliders on the screen show different levels than the OBFN is actually set to at that moment All these minor problems were taken care of as we will see in the following sections 4 2 2 Debug tool To verify the correctness of the microcontroller code parts of it have been tested on a development PC Other parts were tested using
32. Antenna Element Angle of Arrival Application Programmers Interface Command Line Interface Digital to Analog Convertor Directional Coupler delay element Data Flow Diagram Depth First Search Dynamic link library Device Under Test Digital Video Broadcasting via Satellite Erbium Doped Fiber Amplifier Floating Point Unit Free Spectral Range Graphical User Interface Internet Protocol Minimum Mean Sqared Error Model View Controller Mach Zehnder Interferometer xiii XIV ABBREVIATIONS NLP Non Linear Programming OBFN Optical Beam Forming Network OOP Object Oriented Programming OSBF Optical Side Band Filter ORR Optical Ring Resonator PAA Phased Array Antenna PCB Printed Circuit Board RF Radio Frequency RISC Reduced Instruction Set Computer RS232 Recommended Standard 232 RSS Ring Section Simulator RTT Round Trip Time SMART SMart Antenna systems for Radio Transceivers SNR Signal to Noise Ratio SPI Serial Peripheral Interface TEC Temperature Controller TTD True Time Delay UI User Interface UML Unified Modeling Language USB Universal Serial Bus Chapter 1 Introduction Phased Array Antennas have been around for several decades now and have some unique properties and advantages over conventional dish like antennas we see every day for for example receiving satellite television signals One of the most impressive PAAs is the Cobra Dane radar located in Sheyma Alaska and shown in figure 1 1 Built
33. CTERIZATION 59 5 2 2 Voltage levels The voltage levels that are calculated in the simulator are sent to the microprocessor which subsequently drives the DAC after which the voltages are amplified Because the system is very sensitive to small voltage changes in the order of tenths of volts the voltages must be stable in any case Measurements for two channels 127 and in6 of the 4 x 1 subset of the full 8 x 1 OBFN show that the voltages are hardly only 0 001 volt affected by voltages ranging from 0 to 30 volt applied to the other channels We can in this particular case safely conclude that a voltage applied on one channel is not affected by a voltage applied on another As we will see later on the relation between and voltage is captured by a linear equation in V measured at two points It is therefore important to know how the actual voltage relates to the calculated voltage For the same two channels as above in7 and in6 this relation is determined Figures 5 7 and 5 8 show the responses of the amplifier boards measured by a voltage meter The voltage source has been set such that the output voltage of channel 4 when set to 30 00 volts was exactly 30 00 volt Both the absolute difference and relative difference in voltage level have been plotted We see that for low voltages the output differs relatively much from the calculated voltage Low voltages should thus be avoided as much as possible Furthermore we see that the mea
34. For a set of AOAs between 10 and 60 degrees measurements have been done To compare the results of the simulation and the measured values all values are first denormalized to ns The differences between AOA i and i 1 for both the simulation as well as the measurements are shown in Table 5 6 T a b means the delay difference in ns between an AOA of a degrees and b degrees With respect to the resonance frequency we see a small shift of the resonance frequency to the right for larger delays presumably caused by the lack off crosstalk compensation for the amp heaters as mentioned earlier in Section 5 4 5 5 DELAY MEASUREMENTS 67 T 60 50 T 50 40 T 40 30 30 20 Simulation ns 0 0050 0 0062 0 0071 0 0079 Measurement ns 0 0059 0 0061 0 0068 0 0077 Absolute difference ns 0 0009 0 0001 0 0003 0 0002 Table 5 6 0 18 10 degree 0 16 20 degres 30 degree 40 degree 0 14 50 degree 60 degree 1 1 1 0 3 0 32 0 34 0 36 0 38 0 4 0 42 0 44 Sweeptime Figure 5 10 Single ORR response for several AOAs 68 CHAPTER 5 MEASUREMENTS Delay normalized 0 200 400 600 800 1000 1200 Frequency normalized Figure 5 11 The response of a simulation of 4x1 subset of the 8x1 OBFN containing 3 ORRs r1 r2 and r3 The AOA has been set to 30 and 60 degrees equavalent to a A7 of 0 36 and 0 62 respectively 5 5 3 4x1 OBFN As
35. KUIO 94HOZ S0G5Z YIK5X MAQYT RGYZP CM3FP 6CZMU 020042 OF8AA U8B43 KRK JK 46CYO0 T9OUAY ZANEB G1MQJ IAGOI BLHKN PZGAN XO3TV O6T2F IT4XE 3B14P PX49P ZQHKP LS6NK DOBPC P8TQA OWBBE 020007 08VXK NGCBS BAF14 W78ZV CW95V E2EQI TAZDS 9WQQ5 KKQ9X AUAZM KG7NG ZWB5J HJDHD A4BXTK BEYML F9089 FGIS3P OKA4GR Q8864 HAA4HW 020011 0CASZ CKSBR 9TO04P 8HAAU KCS3CX CNCA8 B8065 P1HHG I6XMS FYQTT R1HB5 HNUYQ 46TVV WXM2U K3GO0T 418CW 38EL2 0XAJ0 6328M QJ10Y 02003B OHUON IEP3B 20 JOA T97PE OBZGS AUGM5 P5H7P R2H6K SM61 J EMI6H PKYVA VSQ9E KWPST N0O170 E4EAF 6L1G7 3BZU5 5AG5K 28CSG 81AP1 02002B OY5CP 9F5ZB WGMF8 PK9B7 6Z7V2 M7WCT I1100Y X8UIO 9ZNYO GA171 DHGKM 1LJDL 1CDM1 53JZK POIA5 8XXE2 9LOFM BLUSZ JR436 MDSG8 02002A OVI2C FUBA5 9845R C1I265 38MKS A7GNB A30DI BIQMC V19GF RAFUV JVO3X FS3C2B HIWQQ YTDOX TOKCL ENGT7F DAVNG THIA4I MLK8P N2218 95
36. N AND IMPLEMENTATION OF THE DELAY ELEMENT SIMULATOR Phase 0 0 2 0 4 0 6 0 8 1 1 14 1 6 Normalized frequency Figure 2 8 Phase response plot with normalized frequency for 0 8 and 0 NLP solver implementation In stead of optimizing the parameters for the delay function directly Equation 1 2 the phase function is used see Equation 2 11 Although the use of the delay function to determine the minimum error values for specific parameters would be intuitively appealing and easily comparable with prior research result the delay error does not play a direct role in the output power of the optical detector The absolute values of the addition of the complex phase vectors determine the output power As a third method for determining the error power functions could be used Although theoret ically optimal it has some practical disadvantages Besides that results show only small differences compared to phase tuning Therefore phase tuning is used in the approximation algorithm 6 The proper equation is shown below and is plotted in Figure 2 8 W f arctan sin 2n fT 1 cos 2rfT v1 amp sin 27 fT arctan Vi amp eos 2n fT 5 2 11 Having an error function u X Weotat fo fren 25 D fo fren 2 12 When we repeat the process for a wide range of delays and for a fixed number of 2 3 DELAY ELEMENT SIMULATOR IMPLEMENTATION 21 O Calculated
37. a variable should automatically force recalculation of the values of other variables The graphical programming language LabVIEW is widely based on the idea of dataflow and is thus very well suited for this approach In our case a change of the required delay forces the parameters of each ring to be automatically recalculated The alteration is achieved by a somewhat lengthy process and is therefore broken down into several smaller steps These steps are displayed in a DFD as shown in Figure 2 1 Only the most important activity is displayed being a change in the required delay by the user In the DFD some of the blocks deal with the physical capabilities and limitations of the optical chip We will discuss them one by one 2 2 DELAY ELEMENT SIMULATOR DESIGN 11 dii i A Delay E d E _ p 1 Ring section simulator General Dela ring settings ID Datastore General settings Bandwidth 1 1 Get ring settings 1 1 1 Get coefficients Calculate the kappas and phis using 1 1 2 Calculate kappas the approximation algorithm Calculate value Kappa Apply loss Kappa loss ID File ringsettings mat Determine ibe Model proper value using ring loss Limit kappa the oefficients SPOT limited range of valid values 1 1 3 Calculate phis on M z 3 Calculate value Get Phi compensation Determine the The kappa has a direct proper v
38. a final measurement the results of the simulator were tested on a 4 x 1 OBFN The response of the OBFN for input in5 in6 and in7 have been measured The results are shown in Figure 5 11 simulated and 5 12 measured Although the results seem quite good again there is an offset to the left that is increasing with decreasing angles This effect is most likely caused by the lack of correction of the on other rings These measurements have not been performed yet The effect of heaters 2 4 and 16 on heaters 1 3 and 15 are filled in with zeros at this moment 5 6 Summary and conclusions In this chapter measurement setup execution and results have been discussed When the OBEN is properly characterized the simulator seems to work very well for the tested 4 x 1 OBFN The method used could prove to be usable for larger systems Although the software system is ready to compensate all linear crosstalk effects only half of them have been entered in the crosstalk matrix Some measurements still have to be done hopefully resulting in a perfect alignment of the resonance frequencies in the measured output responses For our measurements the stability of the system itself was sufficient The rewiring of the optical cables caused the offset phase shift of the input signal to be altered Whenever cables are rewired phase offsets should be 5 6 SUMMARY AND CONCLUSIONS 69 Delay ns 0 1 j 0 35 0 4 0 45 0 5 0 55
39. ach ring of the delay element individually and also a combined group delay 75 76 APPENDIX A DELAY ELEMENT SIMULATOR DOCUMENTATION Edt View Project Operate Tools Window Help d m 13pt Application Font 2 131 8 65 normalized group delay relative to urit delay multiply by T to get real delay 10 7997 10 0000 Normalized delay 8 i 0 1453 0 256188 0 EI o Normalized frequency Hm 1003 Ring control Load Save Values array Settings Calculate Approximate A Length Total normalized delay 3 075 1 1 25 5 5 mis E UNE ley 08 os X us AP a 5 0257 Jus TA E it TL 23 1043153 EH i e Redraw 9 E Loss DB chi m 0 4 06 15 2 ost 25 02 4 a Normalized bandwidth Q S gis Seeds Jo Afo 791136 0 100 200 300 400 50 600 700 80 Jack van Galen Universiteit Twente Frequency E B Figure A 1 Screenshot of the delay element simulator at startup Tab Ring Control The bottom left contains two control panels On the left a panel having 4 tabs is visible and has a combined input output function The first tab block is used to apply settings for each ring By selecting a ring the knobs are turned to the current settings Turning a knob has a
40. alue using influence on the phi We the approximation need to compensate algorithm this effect Phis 1 1 4 Calculate chis Calculate Chis Chi is the Phi value for setting the kappa M phis and chis for all the rings in the ring section 1 2 Simulate rings 1 3 Convert for display Mimick the behaviour of cascaded Normalized cascaded ring outputse The ring outputs must be made physical rings using a direct visible Some scale conversion is implementation of the done here mathematical model Plottable data Graphical output in Y N GUI y Figure 2 1 DFD for the delay element simulator 12 CHAPTER 2 DESIGN AND IMPLEMENTATION OF THE DELAY ELEMENT SIMULATOR Loss This method as shown in the returns the power coupling coefficient when loss is applied The resulting value equals the amp to be set When loss would not be taken into account in the simulator the resulting output response differs too much from the real measurement data The loss could be modelled and solved for in the Non Linear Programming NLP solver which we will describe later However every change in the production process of the chip that would change the loss properties of the optical chip would mean that the entire precalculation process has to be restarted Therefore it is desirable that the loss can be corrected for afterwards so the precalculated coefficients which are calculated f
41. alues 64 CHAPTER 5 MEASUREMENTS 5 4 Crosstalk The tuning of the optical chip is done by applying heat to the specific parts of the chip to cause some change in the behavior of the ORRs This tuning process of the optical chip suffers from crosstalk Crosstalk is the unwanted effect that tuning of one ring has on another For example when tuning ring 77 for the heat that is created has some influence on the rings surrounding it The smaller the distance the greater the effect This effect can be both positive and negative meaning that other rings experience a phase addition or phase subtraction The positive effect is caused by some extra heat provided to one ring by another The other effect is explained by 12 as being an electrical crosstalk effect In both cases compensation is necessary The system we will use to compensate for crosstalk will be able to deal with basically any form of crosstalk provided that the effects are linear In that case we can use a simple matrix multiplication to compensate the crosstalk effects The matrix would be a square matrix of size n x n where n is the number of heater elements of the chip that need compensation The diagonal of the matrix is filled with the self values the effect of the heater that belongs to the ring without any crosstalk All the other n x n n values are the crosstalk factors They can be either negative due to thermal crosstalk or both positive and negative due to elec
42. and thereby channels is unlimited e COM The COM port to connect to Note that in general a COM port can be opened only once Combined usage of for instance the simulator and the slider tool is therefore not possible e commandparamsseparator The symbol that separates a command from the pa rameters These values must be changed when using older versions of the micro controller software All the current microcontrollers have been loaded with the newest version of the software so the default settings do not have to be changed e paramsassignmentsymbol The symbol that is used for the assignment of a param eter key to a parameter value e paramsseparator The symbol used to separate parameters from each other e intercommanddelay Debug tool only The time the controller software running on the PC will wait before sending a new command to the micro controller Since at this moment there is only one way communication the micro controller must have enough processing time to handle all the incoming requests When set too low an overload of commands can cause errors or unexpected results The value is the time to wait in milliseconds e resetdelay Debug tool only One special command that requires some time to finish is the reset command setting 0 volt on all available channels The amount of delay to wait can be entered here The value is the time to wait in milliseconds As an example the command uwrchall 12 1200 13 1300 14 1400 sets the
43. ange the Min and Maz delays matrix 4 Coax delay offsets adjust the coaxial delay offsets to proper values 5 Mapping if the simulator is used to control the controller board the mapping has to be adjusted fit the heater numbers to the real channels 6 Crosstalk matriz finally adjust the size of the crosstalk matrix to 2n x 2n where n is the number of rings used in the entire OBFN Run the simulation When all the steps above are done the simulator is ready to be used Press the Run Continuously button in the LabVIEW tool bar Try to change the angle by turning the big knob en see the results adjust almost instantly When an error occurs there is probably some miscommunication between LabVIEW and Matlab Try to restart both programs and rerun the simulation Saving the new settings All settings can be saved by stopping the simulation by pressing the red Abort button followed by Edit menu Make Current Values Default The next time when the simulator is loaded the settings will be restored B 3 APl documentation This section will give a description of all the important public Matlab functions that can be used Most of the functions that are used by the simulator have already been discussed in Appendix A calculateDelays This functions calculates the proper delays for all the delay elements given the total path delays of all paths Parameters e totalpathdelays The total path delays for all paths e minmazdelays
44. ation process for the chip By sending only a AOA the microcontroller calculates the voltage levels thereby tuning the chip This last option does however requires a lot of processing power and a lot of memory to store the coefficients for the approximation algorithm as described in Section 2 3 Also additional information such as the AE spacing must be provided In our situation only the first option has been used extensively However a start has been made for the implementation of scenarios 2 and 3 Although not ready to use yet it should provide a good starting point for expanding the computational capabilities of the chip and aiming for a more stand alone version of the complete system A full API documentation for the functions that have been implemented are listed in Appendix C 4 3 2 Command parser A very simple mechanism has been build into the microcontroller code that accepts messages from the COM port splits them to command and parameters and then splits parameters to parameter parameterz Using separate functions for handling all the command makes it easy to create combined commands and to add functions for processing new types of commands later on 4 3 CONTROLLER SOFTWARE MICROCONTROLLER 51 4 3 3 Hardware software communication For now the microcontroller support communication over a COM port In 12 a suggestion is done to use D2XX for communication D2XX drivers allow direct access to the USB devi
45. ays for all the ring settings are out of range using the Min and Maz delays matrix on the OBFN structure settings tab When one of the lights turns on the systems operates outside the safe zone and results can be unexpected Tab Serial port settings To properly communicate with the controller board directly the COM settings must be right In Windows XP the COM port of the controller board can be determined by right clicking My Computer Properties Hardware Device Manager Ports The other settings are shown in Table B 1 When communicating with the controller board directly using COM the results are displayed in the feedback panel The command that has been sent is shown in the Concatenated string panel B 2 MANUAL 85 Tab TCP settings The second way to communicate with the hardware board is via a TCP connection to the console debug tool Only two parameters have to be known provided that the settings of the console debug tool are correct and that it is running The first parameter is the Host name This can be either a host name as the name suggests but can also be an Internet Protocol IP ad Tab Inputs When all settings are done and the Run continuously button has been pressed the lower section of the simulator displays the results when changing the Angle On the left the array Required delays per input shows the total delay of each path within the OBFN Below that the distributed delays for all the
46. ce through a Dynamic link library DLL Application software can access the USB device through a series of DLL function calls The benefits of using D2XX would be faster data transmission At the moment for controlling 16 channels we need commands like wrchall 1 1234 2 1234 3 1234 4 1234 5 1234 6 1234 7 1234 8 1234 9 1234 10 1234 11 1234 12 1234 13 1234 14 1234 15 1234 16 1234 with a total length of 126 bytes Using a COM speed of 115200 bits 14400 bytes per second means being able to sent roughly 110 commands of this type per second Keeping in mind that there is also some processing time involved for processing the commands on the chip the communication speed is really not the issue here Also when focusing on gradually moving the responsibilities from a PC to the microcontroller the level of communication would further decrease thereby reducing the need for speed even more A big improvement would be to create a simple response parser like the one used in the microcontroller code that handles messages coming from the microcontroller According to the type of return message appropriate action can be taken or new commands can be sent Going one step further is to create a reliable two way commu nication channel preferably over TCP using a new version of the microcontroller that is a little faster has more memory and can implement a TCP stack A fully reliable two way communication channel can be setup bet
47. ctive time delay to the modulated RF signal and a Free Spectral Range FSR of 1 T 13GHz The group delay for a single lossless ORR as a function of frequency f is expressed by 2 KT Hon CN RED Pring 1 1 Of course no chip could be fabricated that behaves like the mathematical equation above Although declining due to new production techniques we have to take optical loss into account When we consider the optical loss the formula becomes gt TAI per ry amp eos 2nfT bring n T r 1 1 2 2 p oT cos 2m fT ring The equations shown depend on the RTT the power coupling coefficient and additional round trip phase shift of the ring The equation involving the loss uses r 10 20 with the loss of the ring in dB When the loss is 0 dB the second equation is of course equal to the first Using heater elements it is effectively possible to control the phase shift o and the power coupling coefficient Both parameters can be used to change the shape of one of the dotted curves shown in Figure 1 4 When changing the height of the curve will be altered When changing the position on the x axis frequency will be changed The total area under each dotted line is constant so there is a trade off between peak delay and bandwidth As a solution to the demand of higher delays for fixed bandwidths ORRs can be cascaded resulting in a curve that is simp
48. d All have both advantages and disadvantages over the others A schematic of the separation of responsibilities is shown in Figure 4 3 1 Keep the microcontroller as simple as possible and only write values to channels 2 Store the crosstalk matrix and other chip characteristics on the chip 3 Implement the whole system on the chip only AOAs have to be provided The first option is the simplest but also the slowest due to the communication speed However since we are in an experimental phase this is not an issue at this moment All the calculations of the voltage values for the individual channels are calculated by the simulator and subsequently sent to the microcontroller The micro controller then simply activates the channel The second scenario can be interesting when using multiple chips of the same type The hardware board has knowledge of the optical chip s characteristics and only needs 50 CHAPTER 4 DESIGN AND IMPLEMENTATION OF THE MICROCONTROLLER SOFTWARE Calculate the Calculate Apply the Calculate required delays ring settings chip specific the voltage corrections levels Figure 4 3 The three scenarios showing the responsibilities of the PC and microcontroller given a AOA to know the s and for each channel to tune the chip correctly Basically the conversion of ring settings to voltages is moved from the PC to the microcontroller The last option is to create a full implementation of the entire calcul
49. d combiners following a specific schematic must arise This chapter will deal with the first two steps that will result in a working DE simulator The third step will be dealt with in the next chapter 2 1 Requirements As every integrated hardware software system the one we are building has several general non functional requirements These are 1 Maintainability Although the simulator can be seen as a stand alone application it would be nice if new features could be added in the near future by others For that reason a programming environment should be chosen of which knowledge is widely available 2 Scalability The simulator must be designed to cope with a wide variety of config urations now and future versions This means having the possibility of changing the number of inputs changing the physical layout of the chip and changing the different ORR parameters like loss and length 3 Resource usage The simulator must work fluently even on an every day computer A proper design of the simulator makes the most out of the available CPU cycles thereby maximizing the speed and responsiveness of the system 9 10 CHAPTER 2 DESIGN AND IMPLEMENTATION OF THE DELAY ELEMENT SIMULATOR 4 Ease of use The simulator s Graphical User Interface GUI must be readily us able for anyone who has some knowledge about OBFNs Clutter and unimportant input elements must either be hidden or not be created at all Also the ability to sa
50. delay elements are displayed The matrix has a similar structure as the connection matrix The two graphs display the results of the actual simulation using the distributed delays The left graph shows the output response of all the paths The right graphs also shows all the input responses but now the coaxial offsets are taken into account This graph gives good indication of the correctness of all the algorithms used in the calculation process When results seem wrong the look up tables containing the coefficients for the approximation algorithms must be verified Also sometimes a restart of both LabVIEW and Matlab works wonders B 2 2 Usage example 1 setup a new OBFN simulation Since the OBFN simulator is a bit more complex than the delay element simulator special care has to be taken to properly set up the system We will show you how to change the settings of the simulator from a 2 x 1 OBEN to a 4 x 1 OBFN Setting up the environment First the environment has to be set up properly For more information about this process read Section B 1 of this appendix Change the OBFN structure settings The following settings have to be adjusted in order to simulate a new and different OBFN 1 Number of inputs change the value to 4 2 Connection matriz adjust the connection matrix to accommodate 4 inputs For details see Section 3 3 1 86 APPENDIX B OBFN SIMULATOR DOCUMENTATION 3 Optionally Min and Maz delays if needed ch
51. e When losses are applied the output amplitude drops e Output intensity The magnitude of the output amplitude e Equivalent Coupling Bends Adds extra coupling because of bends within the ORR e Actual Ring RTT The round trip time of each ring in nanoseconds e Center Frequency The frequency in THz on which the main output window is centralized Tab Calculate Using the MMSE method described in chapter 2 a real time calculation is done to get the best ring settings possible for an optimal combined output group delay response for a given bandwidth and normalized delay When a delay element is simulated that has a large number of rings the calculation time could become large on older computers Tab Approximate For the reason of large calculation times an approximation algorithm was used also described in chapter 2 Again a normalized delay and normalized bandwidth serve as an input for getting an approximation of the ring settings Changes to the delay are processed near real time after which the main window is updated to show the effect A 2 2 Usage examples Example we want to see the response of a delay element containing 2 rings and save the ring settings to a file 78 APPENDIX A DELAY ELEMENT SIMULATOR DOCUMENTATION 1 Click the LabVIEW play button to run the simulator top left corner 2 Click the settings tab lower left corner and change the Number of rings to 2 Click the Approximate tab lower
52. e compensated 5 10 15 20 25 30 Voltage Figure 5 7 Voltage differences for channel 1 Voltage percentage Absolute voltage difference uncompensated Relative voltage difference uncompensated Absolute voltage difference compensated Relative voltage difference compensated 5 10 15 20 25 30 Voltage Figure 5 8 Voltage differences for channel 2 5 3 CHIP CHARACTERIZATION 61 K Tnorm 8 A offset V 098 13 1 04 3 744 16 856 6 61 0 28 096 15 12 4 32 16 28 8 06 0 40 094 16 1 28 4 608 15 992 8 71 0 9 0 92 18 144 5 184 15 416 94 0 57 09 19 1 52 5 472 15 128 9 75 0 64 08 2 6 2 08 7 488 13 112 11 29 0 93 Oy 34 2p 9 792 10 808 12 26 1 16 06 45 3 6 12 96 7 64 13 1 137 05 59 4 72 16 99 3 608 138 1 57 04 8 6 4 23 04 2 44 14 5 1 77 03 117 9 36 33 696 13 096 15 16 1 98 Table 5 2 Example measurements for determining the voltage relationship of ring r1 channel 2 For the first 3 ORRs r1 r2 and r8 the values have been measured using the setup described earlier The results of one set of measurements are shown in Table 5 2 First using Equation 2 1 is converted to the maximum normalized delay The delay is converted to real delay in nano seconds by multiplying it by 0 08 107 Using a signal frequency of 100 109 MHz with T 1 100 108 s the expected signal phase shift is calculated by 360 T Using
53. e of the delay elements A value within the matrix indicates the number of rings contained that delay element The values within one column must thus be the same since this is in reality 1 delay element When summing a row the total 36 CHAPTER DESIGN AND IMPLEMENTATION OF THE OBFN SIMULATOR Listing 3 1 Pseudo code Depth First Search preorder node v 1 1 for each child w of v 1 preorder w dfs vertex v visit v for each neighbor w of v if w is unvisited dfs w add edge vw to tree T Taensa tanken oss o Tu 79 Em Ed Hae Re n pp PE ED get ee i m t X KR bap dar fd v 1 N gt s ae 1 m mum wm mw Wm M e o errr a ae sass heh be 2 Sa m 2 s 1 1 2 ctm Soest a a eS ee E 2 es d t Figure 3 4 DFS walk of the 8 x 1 OBFN 3 3 OBFN SIMULATOR IMPLEMENTATION 37 number of rings in a path is known Counting the non empty positions in a column results in having the number of times a delay element is shared between paths As a concrete example take a look at Table 3 2 Table 3 2 Connection matrix 8x1 OBFN DE DE DE DE DE DE Input 1 Input 2 1 Input 3 N Input 4 Input 5 Input 6 Input 7 Se A A a Input 8 3 3 2 Delay distribution
54. en there are no events at hand Although a bit subjective we think the GUI is a good example of a simple interface combined with only the bare necessities for controlling the simulator Finally the software has been successfully tested on a the Microsoft Windows operating system as well as on Apple OS X Both operating systems are very well capable of running the simulator 28 CHAPTER 2 DESIGN AND IMPLEMENTATION OF THE DELAY ELEMENT SIMULATOR Chapter 3 Design and Implementation of the OBFN simulator Using the previously described delay element simulator as a building block a complete OBFN can be modelled This chapter describes the design and implementation of such an OBFN simulator Not only does the software presented here acts as a simulator but can also be used as a control tool for actively controlling the hardware amplifier board Both aspects will be discussed in this chapter 3 1 Requirements As for the delay element simulator the OBFN simulator also has several general non functional requirements They are 1 Maintainable Although the simulator can be seen as a stand alone application it would be nice if new features could be added in the near future by others For that reason a programming environment should be chosen of which knowledge is widely available 2 Scalability The simulator must be designed to cope with a wide variety of optical chips now and future versions This means having the possibilit
55. endix C 4 2 Controller software PC To be able to use the microcontroller commands have to be sent to it by a PC This section describes the tools that have been used to interact with the microprocessor for controlling and debug purposes 4 2 1 Current software The software we have used is not built from scratch but is built upon an already existing framework This framework allows for Serial Peripheral Interface SPI com munication between the microcontroller and the DAC Also methods for receiving commands from the COM port from the simulator were already available Although working quite well the system would sometimes halt This issue could be traced back to a memory leak in the code which eventually caused a memory overflow Because all communication stops when the code running on the microcontroller crashes no further info could be given to the user To control the chip a special tool has been developed in 12 that allows for sep arate tuning of all the available channels on the chip A screenshot of the system in action is shown in Figure 4 2 Basically the tool sends the same commands to the 4 2 CONTROLLER SOFTWARE PC 47 ller values_O xml Edit Help Heater driver interface 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 1704 1704 1537 1556 1759 1407 1519 1389 1593 1185 1148 981 944 870 870 833 1500 1481 1481 1463 1426 1796 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 4l 42 4
56. er order polynomial of degree 10 Research shows that the error is minimal and better fits are accomplished for delay elements containing more rings Also the approach is more generic and better suits further extension of the number of rings The range of bandwidths that needs to be simulated can be adjusted in the precal culation Matlab scripts For the remainder of this thesis we will work with a value of 0 09 which coincides with a suitable bandwidth for the system that is being developed Having a FSR of 14Ghz the actual bandwidth for which the optimization is done is1 26 GHz 2 3 DELAY ELEMENT SIMULATOR IMPLEMENTATION 23 Calculated kappa Fitted kappa Calculated phi Fitted phi Error total cost 2 4 6 8 10 12 Normalized delay Figure 2 11 Fitted polynomials for a delay element with 2 ORRs for a normalized band width of B 0 16 2 3 4 Small delays As discussed previously in the part about limitation in Section 2 2 1 the value for is limited between 0 and 0 995 In theory when equals 1 all the light is coupled into the ring and after exactly 1 round trip fully decoupled back into the optical guide The normalized delay would thus be 1 When amp is tuned to a very small value only a part of the light is coupled into the ring where the intensity starts to build up at the resonance frequency As a result a infinitesimal small frequency band is delayed for infini
57. esired delay the software will calculate the result and only then updates the screen The event name of this event in the current LabVIEW model is called User Event 2 3 delay element Simulator Implementation Using the design described previously in this chapter we can start implementing the delay element simulator in LabVIEW For additional functionality and specific small algorithms we will revert to MatLab LabVIEW and Matlab work seamlessly together and is thus a good combination of a visual programming environment combined with a solid text based programming language This section describes implementation details about the GUI the approximation algorithm and the dataflow 2 3 1 Graphical user interface The first thing one sees when working with a simulator is the graphical user interface GUI Although GUF s are an interesting topic of research by themselves we have aimed at developing an easy to use interface just by using some common sense This means building an interface with a minimum of clutter a logical work flow and one that required no unnecessary scrolling in order to maximize a clear overview A screenshot of the interface can be seen in Figure 2 6 2 3 DELAY ELEMENT SIMULATOR IMPLEMENTATION 17 variable rings nieuwe structuur vi Front Panel 181 Eie Edit View Project Operate Tools Window mm s s e m Appication Font
58. files from the CD JDK1168 javaxcomm directory as follows Copy win32com dll to the bin directory e g c jdk1 6 bin b Copy com jar to the lib directory e g c jdk1 6 lib c Copy avazx comm properties to the lib directory e g c jdk1 6 lib 3 You should be able to use the COM port now If the steps above did not work make sure you have copied the files the proper Java directory and reboot your system 89 90 APPENDIX C HARDWARE CONTROLLER DOCUMENTATION C 3 Using the Slider tool The slider tool can be started by navigating to the Slider tool directory on the CD and double clicking click the slider jar file The jar file can be copied to any location and is fully stand alone C 4 Using the Debug tool The debug tool can be started by navigating to Debug tool and double clicking the debugtool jar file The jar file can be copied to any location and is fully stand alone C 5 Flashing the micro controller To flash the ARM7 micro controller the CrossWorks studio application is used A manual can be found in the Manuals directory on the CD labeled Starting the micro controller environment doc When using a specific micro controller for the first time on a computer a license needs to be installed For all micro controllers that were in possession during the writing of this thesis licenses were requested and can be found in Appendix F C 6 Floating point operations The micro controller does not have ha
59. g tool code Bibliography 1 2 10 11 12 Online Available http www lionixbv nl L Zhuang Time delay properties of optical ring resonators Master s thesis Universtiy of Twente May 2005 C Roeloffzen Passband flattened binary tree structured add drop multiplexers using sion waveguide technology Ph D dissertation University of Twente 2002 M Ruiter Design of a system for driving heaters on optical ring resonators Master s thesis University of Twente 2006 W D Shoaff How to write a master s thesis in computer science Department of Computer Sciences Florida Institute of Technology 2001 R Blokpoel Staggered delay tuning algorithms for ring resonators in optical beam forming networks Master s thesis University of Twente 2007 T C Lethbridge and R Laganiere Object Oriented Software Engineering Prac tical Software Development using UML and Java 2nd ed McGraw Hill 2001 S Baase and A V Gelder Computer Algorithms Unknown Ed Addison Wesley 2000 A Meijkerink Functional design of the demonstrator chipset deliverable for the FlySmart Project M Tijmes Simulation of a ring resonator based optical beamformer system for phased array receive antennas Master s thesis University of Twente 2009 T Vrijmoeth Implementation of a heater driving system Master s thesis Uni versity of Twente 2006
60. g with phase shifters instead of T TDs the position of the beam changes with frequency The building blocks are combined to form an OBFN A 8x1 OBFN for a receiving phased array is shown in Figure 1 3 The OBEN is designed using a binary tree topology Using this layout only a small amount of ORRs have to be used to achieve a large range of delays per path while the dimensions of the chip can be kept to a minimum Although the freedom of tuning for every path is more restricted than in for example a parallel topology the tuning complexity is reduced The OBFN shown has 2 rings in one of the branches of each stage where n is the stage number see Figure 1 3 Using this approach every 1 2 BACKGROUND 5 path has a unique linearly increasing number of rings Although this method seems attractive and uses a lot less rings than a parallel topology the number of rings grows exponentially Fortunately the achievable delay with a ring section of a certain size is not linearly dependent on the number of rings so far less rings have to be used The physical layout of the chip actually produced is shown in Figure D 1 in Appendix D and shows only 8 rings An ORR consists of a straight waveguide with a circular waveguide coupled to it Using a ORR with a circumference L of 1 5cm and a waveguide group index n of 1 55 we can calculate the Round Trip Time RTT to be T L n 3 10 An ORR has a periodic group delay response representing the effe
61. have some irregular shape The simulator could be expanded to cope with grid PAAs The problem of handling curved surfaces could likely be solved in the simulator by adding an extra SubVI between the input of the AOA and the calculations of the required delay per path e Some of the crosstalk effects have not yet been measured These effect do however contribute to the shift of the resonance frequency and should thus be included in the crosstalk matrix e Although the crosstalk correction matrix works it would be better to have some form of thermal feedback from each heater directly by the use of integrated ther momethers Perhaps the resistence of the heaters on the current chip could be used for that purpose By characterizing the optical effects of each tuning element for all temperatures the heater and feedback combination would be responsible for achieving the desired optical effects The crosstalk matrix and the relation from voltage to optical effect can then be eliminated leaving an easier to calibrate and tune system especially when the OBFN grows in size e The communication with the microprocessor is currently one way Possible prob lems due to timing are now solved by adjustable delays between the commands that are sent Of course this is only a temporary solution Better is to have a two way reliable communication channel with the chip giving feedback This could be implemented using a simple response parser in the Java debu
62. he coaxial delay offsets and the the total amount of delay required per input of the OBFN can be calculated This is done using the algorithm that will be discussed in more detail in Section 3 3 2 34 CHAPTER 3 DESIGN AND IMPLEMENTATION OF THE OBFN SIMULATOR Determine overflow Using the delays per delay element and the information that is made available by the user concerning the minimum and maximum delays for a delay element of a specific length a warning will be issued When a warning is given calculations of the simulator are out of bounds and cannot be used reliably The software will however continue to Work Get ring settings When the delays for all the delay elements are known the individual ring settings for each ring within the delay element can be calculated Note that the number of the step in the flow diagram corresponds to Figure 2 1 in the previous chapter This step is indeed a reuse of the model and code used for the delay element simulator A SubVI is created to abstract the inner workings of the previous simulator For details about this step see Section 2 2 1 Simulate the delay elements Again the SubVI of the delay element simulator is used to simulate the output response when the signal travels all the concatenated ORRs The resulting responses for all the paths within the OBFN are summed Create plot data The responses coming from the delay elements are being converted for plotting two graph
63. ile containing the processed coefficients along some other useful info See the documentation of the calculatedOptimizedCoeffi cients 2rings function for details calculatedOptimizedCoefficients 2rings A function to calculate coefficients according to the phase optimization function Given a bandwidth parameters for the whole delay range will be optimized A 3 PRE CALCULATION SCRIPTS AND API DOCUMENTATION 79 Parameters e bandwidth The bandwidth to optimized for e fromdelay The starting point of the delay range normalized e todelay The end point of the delay range normalized Return values e resi The coefficients for the first curve fitted function in this case the e res2 The coefficients for the second curve fitted function in this case the e Z1 The raw optimization results for the first parameter e Z2 The raw optimization results for the second parameter 6 xas The x axis used This servers basically as an index for the delay range B The bandwidth used e E Error values from the NLP solver for the whole delay range phase 2opti Sets some options and then calls Matlab s fmincon function to start the NLP solver As an objective function to be minimized phase fun2 x is used Parameters e points Number of points to check within the frequency range e bandwidth Normalized bandwidth e height The target delay e varargin Values used as a starting point for the NLP solver Ret
64. ined response 7 Normalized group delay 0 2 0 15 0 1 0 05 0 0 05 0 1 0 15 0 2 0 25 Normalized frequency Figure 2 13 Response for a normalized delay of 0 5 for a delay element containing 2 ORRs be slightly lower since we do not have to square the function and sum it Besides one can argue about the meaning of the error and if it suits the problem better than the previous solution In our system a high error in some part of the bandwidth is unacceptable and thus this direct approach would deal with that by punishing high errors immediately for any given frequency To test the new error function the coefficients have been precalculated for a DE containing 5 rings Recall that because of symmetry only 5 not 10 parameters have to be optimized In this case and 2 The results of the calculated parameters are shown in Figure 2 14 In spite of what one may have guessed the calculation took about twice the time it took for the regular objective function and the results are not smooth The optimal solutions for different sets of parameters are further apart than when using the previous error function The NLP solver evidently has a harder time trying to find an optimal solution The function cannot be reliably curve fitted with a relatively low order polynomial and thus the road previously chosen will be used 2 3 6 Matlab scripts and API For the delay element simulator delay elements contain
65. ing connection settings The middle part takes care of all the settings related to the OBFN itself Lastly the bottom part serves purely to display the results Please be aware of the caching of Matlab functions that LabVIEW performs For substantial changes in settings the simulator has to be restarted If you experience any errors press the Make current values default option in the Edit menu and restart LabVIEW Things should be working again now Lets discuss all the separate parts one by one starting with the middle part going down to the bottom part and finally the top part 8l 82 APPENDIX B OBFN SIMULATOR DOCUMENTATION jo lo Figure B 1 Screenshot of the OBFN simulator at startup B 2 MANUAL 83 Tab OBFN Structure settings Multiple settings can be adjusted to fit the desired OBFN simulation On the left the number of inputs the general ring loss in dB the RTT in ns the AE spacing in and the normalized bandwidth can be set These general settings will be used in all the simulated ORRs Next in line is the connection matrix Details on how to fill in the connection matrix are written down extensively in Section 3 3 1 Make sure that the actual size of the matrix in LabVIEW is adjusted to the connection matrix exactly and that no extra empty rows or columns are active To the left of the connection matrix there is a small matrix called Min and Maa delays This
66. ing up to five rings have been precalculated for a large range of bandwidths and delays depending on the size of the delay element Because Matlab is not really suitable for containing large collections of data in a flexible way a data structure consisting of nested structs is used The diagram figure 2 15 can be used as a reference model and should make it fairly easy to get the proper coefficients on demand Since in Matlab arrays cannot be indexed 26 CHAPTER 2 DESIGN AND IMPLEMENTATION OF THE DELAY ELEMENT SIMULATOR Calculated kappa 1 2h Fitted kappa 1 Calculated phi 1 18H Fitted phi 1 Calculated kappa 2 Fitted kappa 2 Calculated phi 2 Fitted phi 2 Calculated kappa 3 Fitted kappa 3 Value kappa phi a e a e e N Normalized delay Figure 2 14 The results of the NLP solver when using the alternative objective function by a user defined value a separate array index is created This array contains all the delays that are precalculated A simple search algorithm will find the proper index 7 for the required delays The ring settings data structure at position 7 returns a set containing two subsets kappas and phis For a delay element containing n rings n coefficient sets in the form of an a4 1 a2 ao are returned Each set contains the coefficients for creating a polynomial function in the
67. lized values and the fysical values Equation 2 9 for delays and Equation 2 10 for bandwidths can be used 2 3 DELAY ELEMENT SIMULATOR IMPLEMENTATION 19 group delay Frequency Figure 2 7 Group delay curve for 3 cascaded ORRs The shaded area denote the parts that are added to the total costs T norm rm 2 9 2 9 Beier 2 10 T can be calculated when the FSR is known The ORRs on the OBFN have a FSR of 14GHz For the reaminder of this thesis all mentionings of bandwidths or delays are normalized unless stated otherwise Symmetry To speed up the process of finding optimal parameters for a given delay and DE con figuration the number of unknowns can be decreased by using symmetry As can be seen in the output window of Figure 2 6 very nice combined output responses can be achieved by a symmetrical distribution of the individual responses of the rings in a DE When using three rings or any other odd number of rings the of ring 1 can be set to 0 since it is always in the center and thus does not have to be calculated The however does have to be optimized The of rings 2 and are identical but opposite so only one of them has to be optimized Also the s of ring 2 and 3 are identical again meaning that only one of them has to be optimized The 6 parameters have now thus been reduced to only 3 Of course for other DE configurations the same reasoning applies 20 CHAPTER 2 DESIG
68. lly over a certain bandwidth The net work analyzer measures a time window that is synchronized with the ramping thereby showing the power and phase response at each frequency Multiple FSRs are included within the frequency range As a result the network analyzer shows one plot contain ing dips due to the round trip losses in the ring at their resonance frequencies and another showing the phase shift of the original signal 5 1 2 Optical chip labelling All the rings heaters inputs and outputs have been labeled by a unique number All input labels start with in the output labels with out and the rings with r The heater channels do not have a prefix and are referred to by only their number These numbers correspond to the sliders of the slider tool as discussed in Chapter 4 The branch labbeled OSBF leads to the Optical Side Band Filter OSBF and has not been used The complete labelling is shown in Figure 5 3 The dotted box surrounding the top left of the OBFN illustrates the part of the chip that was used during all measurements described in this chapter The light from out6 was connected to the EDFA and from there to the optical detector 5 2 Stability and Voltage Levels For the measurements to succeed it is very important that we are working with a stable system Fluctuations of the output responses of the ORRs during long measurements will make those measurements unreliable Two types of measurements have been done an overall
69. ly the sum of the individual responses The result of this is shown in Figure 1 4 as a solid line The so called ripple is the effect that is clearly visible at the top of the concatenated response and is a slight variation of group delay in a certain bandwidth In general the smaller the ripple the better Roughly speaking 6 CHAPTER 1 INTRODUCTION p2 p3 k2 k3 group delay Figure 1 4 Theoretical group delay response of three cascaded ORRs the required number of rings is proportional to the product of the required bandwidth and the maximum delay The optical chip is tuned thermo optically by electrical heating chromium resistors As a consequence of the heat at specific places the optical waveguide s refractive index changes Because of this change either the resonance frequency or the power coupling coefficient of an ORR is altered Thermo optical tuning itself is very well explained in Section 3 4 of 3 1 3 Research organization To have a well defined research project several goals are determined The methodology used to conduct the research and the research questions that have to be answered when this assignment is finished are stated in the following sections 1 3 1 Research goal The research project described in this project has one main goal e The creation of a maintainable and scalable software control system that can automatically tune all the parameters of an OBFN given only the direction of the inc
70. mediate feedback log within the simulator and the fact that there is no need for an additional Java program running in the background The board has to be connected to the same PC where the simulator is running on 3 3 5 Complexity and upscaling A real commercial system could consists of as much as 64 x 64 AEs For a single row a 64 x 1 OBFN could be used Using the exponentially increasing number of rings in each stage the maximum number of ORRs in a delay element is n 2 For n 64 this means a delay element containing 32 ORRs Fortunately as we can see in Table 2 1 the feasible delay as a function of delay element length grows faster than linear which means smaller delay elements can be used The simulator has been built to handle n x 1 OBFN with n an arbitrary number Recall that the only limitation for now is the precalculated approximations for the ring settings for up to 5 rings the coefficients were calculated When analyzing the wiring in the block diagram of the simulator it can easily be seen that the computational complexity of the system or the time it takes to run recalculate a simulation is roughly proportional to the connection matrix meaning O mn 3 4 Simulation results When the OBEN simulator is started all the needed steps to calculate the proper ring settings are performed resulting in a plot as displayed in Figure 3 6 The settings used to run the simulation are displayed in Table 3 4 In the plot we ca
71. n be below 1 as we have seen in Section 2 3 4 using the coaxial delays we can prevent the ORR to be tuned below a normalized delay of 1 alltogether In theory when the MZIs of the ORRs are produced perfectly the minimum normalized delay is exactly 1 which means a minimum total path delay equal to the number of ORRs within that path When we have Ar 40ps RTT 2 and a minimal path delay in each path of the OBFN of 0 1 1 2 2 3 3 4 RTT respectively a compensation has to be added that compensates for these minimum delays This would enable us to receive a broadside signal AOA 40 CHAPTER 3 DESIGN AND IMPLEMENTATION OF THE OBFN SIMULATOR 0 by tuning all the rings to the minimal delay Added to that we need a normalized RTT 2 between all paths in the case of a maximum negative AOA The total coaxial offset is now obtained by summing these two values The results are displayed in Table 3 3 Although we are able to continuously tune a delay element between 0 and some maximum delay the demands on the total delays of the delay elements are increased by these coaxial offsets resulting in higher ripple and thus a more distorted signal Table 3 3 Additional coaxial delays Path Rings Ring comp Max neg AOA Total coax delay norm Input 1 0 4 3 5 4 3 5 7 5 Input 2 1 3 3 3 3 6 Input 3 1 3 2 5 3 2 5 5 5 Input 4 2 2 2 2 2 4 Input5 2 2 1 5 24 1 52 3 5 Input 6 3 1 1 1 1 2 Input 7 3 1 0 5 1 0 5 1 5 Input 8 4 0 0 0 0 0
72. n immediate effect on the output windows The second tab block is used for automatically calculating the ring setting by using the MMSE calculation and the approximation algorithm both described in chapter 2 Tab Load amp Save The Load and save tab shows two buttons which allows you to save and restore ring settings After opening a previously saved file all the ring settings will be restored Also the number of rings in the Settings tab will be adjusted according to the number of rings the settings were saved for Tab Values array The Values array tab contains the values of each ring that are used for drawing the group delay output response in the top left window Values can be individually changed here if needed Changes are immediately processed The chart array data is just for error checking purposes and can be ignored during normal operation A 2 MANUAL TT Tab Settings Several default settings are available for the delay element simulator LabVIEW offers the possibility to store default values for all input fields which can be used to store custom settings for later use The settings are e Number of rings The number of rings to simulate e hound trips The number of round trips used to model an ORR A higher value will result in a better response at the cost of speed e Input amplitude A constant 1 signal to model the input signal e Output amplitude The output amplitude with respect to the input amplitud
73. n loss and given maximum delay Now r follows directly from filling in the loss factor and Tq from the lossless case which is 2 2 DELAY ELEMENT SIMULATOR DESIGN 13 0 dB loss 1 dB loss 3 dB loss 10 dB loss 0 95 09r 0 85 08r 075 07r Compensated kappa 0 65 06r 0 55 0 5 1 1 5 2 2 5 3 3 5 4 4 5 5 Maximum normalized delay Figure 2 2 Compensated as function of loss per round trip and maximum delay derived from Equation 2 2 The new kappa that now compensates for the ring loss is simply determined by 2 compensated 1 Closs 2 4 The method described is only usable when losses are not too high lower than 1 dB For higher losses the compensated curve does not match the lossless curve anymore as we can see in Figure 2 3 Kappa limitation An ORR consists of a straight waveguide and a circular waveguide next to it See Figure 2 4 The coupling section is in fact a Directional Coupler DC The relation between the fields at the inputs and outputs is given by E E 1 j k 2 5 E jV amp V l k E The value of the power coupling coefficient to the ring amp which controls the height of the delay response is limited due to the fabrication process of the optical chip The best value for a single Kg of one of the rings on the optical chip currently achieved is 0 465 according to 6 where the ideal value
74. n see the needed response of all the paths within an 8 x 1 OBFN for an arbitrarily choosen of 0 5 which is equivalent to an AOA of 53 degrees In Figure 3 7 the results for a negative AOA of 53 degrees is shown Again the time delay A7 between the paths around the normalized frequency of 0 is 0 5 Part of the OBFN has been reused in 10 The simulator created there covers the complete path of a real data signal being sent and received by a PAA The delays for the beam forming process were calculated using the LabVIEW code described in this and the previous chapter Results show a nice gain and the transmitted signal could be restored without errors 42 CHAPTER 3 DESIGN AND IMPLEMENTATION OF THE OBFN SIMULATOR Normalized group delay 3 1 05 0 0 5 Normalized frequency Figure 3 6 Simulation results for an 8 x 1 OBFN width a of 0 5 equivalent to an AOA of 53 degrees Normalized group delay 3 1 0 5 0 0 5 Normalized frequency Figure 3 7 Simulation results for an 8 x 1 OBFN width a of 0 5 equivalent to an AOA of 53 degrees 3 5 MANUAL 43 Table 3 4 Simulation settings Property Value Number of inputs 8 Ring loss 45 dB RTT 0 08 ns AE Spacing 0 015 m Normalized bandwidth 0 09 Connection Matrix See Table 3 2 3 5 Manual A user manual and API Documentation are included in appendix B The manual contains the steps that need to be taken in orde
75. ne of the greatest challenges in NLP is that some problems have local optima that is solutions that satisfy the requirements of the constraint functions Algorithms that propose to overcome this difficulty are called Global optimization Global opti mization would prologue the necessary time to precompute solutions and is therefore not applied Good initial values should be guessed in order to prevent halting in a sub optimal state MMSE A Minimum Mean Sqared Error MMSE estimator describes the approach which min imizes the mean square error An example an error function based on this technique is shown in equation 2 8 Note that this function is not used to precalculate the ring settings but serves merely as an example In the equation the represents the combined responses of several ORRs and the target delay D is subtracted of it Next the result is squared and integrated for all the frequencies between fmin and which results in the total error In Figure 2 7 the top part of the combined output response of a DE with 3 ORRs is shown The ripple is clearly visible The total error is the square of the sum of the areas of all the shaded areas for the bandwith of interest In this case the bandwidth is limited to B fmaz fmin fm ax je D af 2 8 fmin 2 3 3 Normalization Both the bandwidth and delays are normalized throughout the system and throughout this thesis To convert between the norma
76. ng to calculated ring settings By combining more ORRs into a delay element larger bandwidths can be delayed Because of the large amount of heater elements and the influence that one heater element has on another a sophisticated control system is needed that is capable of automatically calculating the correct settings and tuning all the heater elements given only the direction of arrival of the incoming satellite signal To achieve the goal of creating the automatic control system two simulators were written in LabVIEW to see if underlying calculations would work in theory The first of the two simulators was specifically designed to simulate the delay response of delay elements with a variable amount of rings The settings for the rings were obtained by using an approximation algorithm with pre calculated values Several effects and their compensations have been incorporated The end result is a scalable simulator capable of simulating delay elements containing a variable amount of rings The second simulator was an additional layer around the code of the first simula tor thereby creating a tool that can simulate an entire OBFN The distribution of the delays across the rings and the calculation of the voltages is all done within this simulator The connection to a previously designed amplifier board made it possible to apply these calculated voltages to the actual lab setup Finally as a proof of concept the simulator has been tested in the lab
77. ntire OBFN The distribution of the delays accross the rings and the calculation of the voltages is all done with this simulator The connection to the previously designed amplifier board makes it possible to apply these calculated voltages to the actual lab setup Finally as a proof of concept the simulator has been tested in a lab environment to see if the apprach taken could work and would be a feasible candidate for further research The first measurements using the voltages calculated by the control system look very promising Also the system is capable with very small adjustments of tuning future chip designs or using other tuning methods than thermo optical To keep this report as generally applicable as possible no specific applications were kept in mind when performing simulations or measurements 71 72 CHAPTER 6 CONCLUSIONS AND FUTHER RESEARCH 6 2 Further research During this research project some interesting questions arose and things came to mind that could possibly improve the system as a whole e Negative AOA are now fully handled by additional coaxial delay lines These ad ditional lines however put a higher demand on the delay that must be achieved by each ORR Perhaps a more symmetrical OBFN design would solve this problem Since more rings create a higher level of tuning complexity some trade off must be found e The delays are now calculated for a flat linear PAA In real life PAAs are often curved or
78. nversion The relation between kappa and the actual heater response to a certain voltage can be see as a raised cosine function see Figure 2 5 To properly operate the OBFN controller the are converted according to equation 2 6 The converted values can then be used for applying voltages in a similar fashion as the The converted values will be denoted as coupier from now on When no subscript is used 2 2 2 Structure The delay element simulator is built according to the event based programming model Event based programming or event driven programming is a programming paradigm in which the flow of the program is determined by events i e sensor outputs or user actions mouse clicks key presses or messages from other programs or threads When there are no such events the program simply waits without using any resources A decrease from 10096 CPU usage when using user event catch loops to less than 1096 using the event based paradigm proves the usefullness of this approach In the simulator there are a few CPU intensive operations e Changing the delay by turning the delay button as seen in Figure 2 6 e Updating the screen with new information and drawing the graphs 16 CHAPTER 2 DESIGN AND IMPLEMENTATION OF THE DELAY ELEMENT SIMULATOR Phi i coupler Figure 2 5 Relation between coupler For these two operations separate events were built Now when the user inputs the d
79. oming beam Of course to reach this goal it has to be broken down into smaller more com prehendable pieces First the problem will be modelled and simulated Next the model can be used to implement a working system using already available hardware Finally to verify the correctness of the system we have to check if the simulation results correspond to real life measurement data 1 4 THESIS ORGANIZATION 7 1 3 2 Methodology A methodology is used to identify distinctive actions taken in the process of this as signment 5 The following steps are taken in a more or less sequential order e Literature study Related papers theses and books must be studied to get ac quainted with the subject e Defining research question In this Thesis a research goal is set which must be achieved e Requirements analysis In order to present a proper design and implementation of a control system the requirements have to be made clear e Architecture design The design of the controlling system is presented based on which a prototype can be built e Prototype implementation A prototype implementation is developed to test and verify the design and to provide input for further research e Performing measurements To test the prototype and the underlying model for correctness the output of the system should be tested against expected results e Results and conclusions The results are evaluated Research questions will be
80. or a lossless ring remain the same Since the surface area below the response of a single ORR is constant the time delay response curve is almost solely determined by the highest point which has a delay of Tmax Note that this is only the case for low losses as we will see later on The current optical chip batches have been produced with a loss between 0 1 and 0 3 dB per cm and a ring length of 1 5 cm The loss lies somewhere 0 15 and 0 45 dB per round trip These losses are low enough to use the following method of compensation for the loss afterwards Using this Tmax the corresponding can be recoverd for the lossless case In case there is loss the Tmar must remain the same and a new corresponding has to be calculated The maximum normalized group delay follows from Equation 3 16 in 2 r max 2 1 yee cer _ where 1 1079 2 with a the loss in dB In the lossless case would be 1 and the formula simplifies to l e max 1 1 2 2 T lossless 2 2 When we fill in c in Equation 2 2 we acquire the maximum delay for the lossless case To calculate the c for the case there is loss we need to rewrite Equation 2 1 as a function of the loss factor r and the maximum delay 75 44 Tox Tmax EN TAA 272 12 TA T Ar 2 3 Closs 2 r Tmax Equation 2 3 has been plotted for several losses in dB see Figure 2 2 With this equation the c can be determined for a give
81. other path delays The total path delay is acquired by Toa TOS Tone Toart F U I 3 2 32 CHAPTER 3 DESIGN AND IMPLEMENTATION OF THE OBFN SIMULATOR 2 OBFN Simulator General settings V 2 1 Calculate Delta Delay Convert the DOA to Delta Delay using the AE spacing Coaxial delay offsets Delta delay Number of inputs 2 2 Calculate total path delays Total path delays 2 4 Calculate distributed delays For each output calculate the right amount of delay and incorporate the coaxial delay lines For each of the delay elements calculate the appropriate amount of delay Distributed delays 1 1 Get ring settings calculate the ring setting 1 2 Delay element simulator For all the distributed delays per delay element Simulate the delay elements and sum the responses per path See Flow Diagram 1 Simulation responses 2 7 Create plot data Convert the responses to plottable data ID Datastore General settings Number of inputs Min and max delays 2 3 Determine overflow Give a warning when too small or too large delays are being used True false 0 Q ight indicators Min and max delays Number of inputs Total path delays Number of inputs fin and max delays Connection matrix Ring el 2 8 Convert ring settings to voltages The ring settings need to be converted to voltages 2 9 Com
82. pensate for crosstalk Multiple the phi values with the crosstalk matrix to compensate for crosstalk Chart data Voltages ar gus Reponse Voltages per graphs channel Figure 3 2 Dataflow for the OBFN simulator 3 2 OBFN SIMULATOR DESIGN 33 Wave front Beam direction AEs Figure 3 3 A general uniform linear PAA where 7 is the number of the input and is the required inter arrival time between the AEs For example when the maximum normalized coaxial delay is determined to be 3 5 and is for some specific AOA is determined to be 0 5 the total path delays can be calculated using the above formula As an example the total path delays have been calculated for a 4 x 1 OBFN see Table 3 1 The results in the second last column are the delays that must be realized by the path to fully compensate all the additional delays caused by the coax cables For instance path 2 must realize a normalized delay of 2 with its one ORR If we sum the fixed coaxial delays with the newly calculated delays the resulting values displayed in the last column indeed show a of 0 5 More information on how to calculate the coaxial delays is available in Section 3 3 3 Table 3 1 Calculation of the total path delays Input Coax mMaX Teoar Tcoari i 1 AT Tpathi Total 1 3 5 0 0 0 3 5 2 2 1 5 0 5 2 4 3 1 5 2 T 3 4 5 4 0 3 5 1 5 5 5 Calculate distributed delays Using t
83. r The controller board is 53 54 CHAPTER 5 MEASUREMENTS Figure 5 1 The inside of the styrofoam box Jez Ajeue YJOMON optical EDFA detector Stab Figure 5 2 Measurement setup TEC Temperature Controller for the Laser Curr Current controller for the laser lts current is sweeping controlled by the Network Analyzer DUT Device under test the optical chip in the photo between the laser and the Contr block Mod modulator PC is the com puter interfacing the control system Contr EDFA Erbium Doped Fiber Amplifier Electrical wires are represented by dotted lines and optical wires by solid lines 5 2 STABILITY AND VOLTAGE LEVELS 55 in its turn connected to an ordinary PC to get its instructions The modulator mod superimposes a RF signal of 100Mhz onto the optical carrier and is then fed into the optical chip The signal as it is leaving the optical chip is amplified using an Erbium Doped Fiber Amplifier EDFA after which it is detected using an optical detector and returned to the network analyzer The network analyzer shows the time domain of each current sweep on the x axis which we will explain later and the received power and phase shift as two separate traces on the y axis Measurements are done using a laser current ramping technique The laser current is ramping between two values much like a sawtooth wave As an effect of the current ramping the laser frequency changes gradua
84. r delay elements having a number of ORR between 1 and 5 error plots have been created Two of them are shown in Figure 2 10 and Figure 2 11 To prevent the system from curve fitting a function in a range that is not useful at all only the part with minimal error is used The range determination heavily depends on the required bandwidth When larger bandwidths are required the error will dramatically rise and the near errorless range of delays is reduced An example for a normalized bandwidth B 0 16 is shown in 2 11 Plots such as the ones shown are created for all 5 rings for B 0 09 for which feasible delay ranges are constructed For now all minimum delays are gt n where n 22 CHAPTER 2 DESIGN AND IMPLEMENTATION OF THE DELAY ELEMENT SIMULATOR Calculated kappa Fitted kappa Calculated phi Fitted phi 12 Error total cost Normalized delay Figure 2 10 Fitted polynomials for a delay element with 2 ORRs for a normalized band width of B 0 09 Number of rings Minimum delay Maximum delay 1 1 0 2 0 2 2 0 7 0 3 3 0 14 0 4 4 0 25 0 5 5 0 30 0 Table 2 1 Theoretically feasible delays per delay element for B 0 09 is the number ORRs in the delay element The results are shown in table 2 1 Due to the fact the the curve of large delay elements is not fittable any more with low order polynomials and inversed polynomials used in 6 all curves are fitted with a high
85. r to properly setup the simulator The API provides more insight in the functional hooks that can be used to extend the simulator or to be called from other programs 3 6 Summary and conclusions A fully functional OBFN simulator has been built that meets the predetermined re quirements The simulator is again built in LabVIEW and uses the simulator from the previous chapters in its core With this simulator OBFNs of different sizes and layouts concerning the number of rings used in each stage can easily be simulated In theory the simulator can be extended to larger OBFNs just by adding extra precalculated coefficients 44 CHAPTER 3 DESIGN AND IMPLEMENTATION OF THE OBFN SIMULATOR Chapter 4 Design and Implementation of the Microcontroller Software Creating a dedicated software program running on the ARM chip itself creates more flexibility towards the controlling aspect of the OBFN system as a whole The PC would serve as just a user interface for settings some parameters and the desired AOA This way the user interface part can easily be ported to for instance a mini PC for direct integration with the rest of the OBFN system 4 1 Overview of the control system The beam forming system is controlled by a control system This control system consists of a combination of already existing hardware 4 and partly new developed software A schematic of the control system that has been used during this project is shown in Figure 4
86. rdware floating point support However a soft ware library can be included during compilation that enabled a transparent use of floating point values anyway A separate section in the Starting the micro controller environment doc document described the exact settings that need to be set in order to make use of this functionality Appendix D OBFN layout Diagram of the board for reference purposes FLYBx1 5 4 o r So ALY 3 y cm Co Figure 0 1 8x1 FlySMART chip layout 91 92 APPENDIX D OBFN LAYOUT Appendix E Ring channel and waveguide data Below an enumeration is given of the voltage sign per ring their channels for and tuning and the used input waveguide 1 channel 1 k channel 2 input waveguide 7 2 channel 15 channel 16 input waveguide 5 3 channel 3 amp channel 4 input waveguide 5 4 channel 7 channel 8 input waveguide 2 5 channel 17 channel 18 input waveguide 3 6 channel 5 amp channel 6 input waveguide 4 T channel 21 channel 22 input waveguide 2 8 channel 23 channel 24 input waveguide 2 93 94 APPENDIX E RING CHANNEL AND WAVEGUIDE DATA Appendix F Rowley Crossfire Licenses 020025 OMO060 9JQ6G IAJOX O8WL2 FCOSU BC3CN KGFB2 PKL5S JOAFQ UUZ7M M5664 QMXZY U
87. rdware software communication 454 Floatmig point op rations lt x ecas 9 o o Ad 50 s don sacose Re we ee ee we 5 Measurements Gl System OVERVIEW sesede eaa ok dox E y Bee 5 1 1 Measurement 123 cp labelling uoo d ERE aE 5 2 Stability and Voltage Levels 521 stability lt se roa PRS EES EES 522 Vokagelevels gt 5c soroa ew Re oy e XR 53 Chip characterization uu x o RE XR 4 hx Xo douce dead Bol 222259 9 XS EX REA x xS Dod Pheslibration cs oee dt nyd ERS Te ee 5 4 Crosstalk iua saccos eac eu ee Ae whe ee ee 2 Wis 5 4 1 Measurement CONTENTS VII Dr RUE 4 he eo RO RE RO SU 65 5 5 Delay measurements 66 5 5 1 Determining group delay offsets 66 5 5 2 Single 66 5 5 9 amp xl OBFN unu uso go rusaka no eRe Oh Ee 68 5 6 Summary and conclusions sooo e a 68 6 Conclusions and Futher Research 71 OUEST 0 oo ecs kk bd parea REAR ERDAS 71 D Pirie ieseni oda owe BR CROCO OUR y EEE ee 72 A Delay Element Simulator Documentation 75 AJ General e eS xc 75 D s o0 Uses Beh Gry Se ee ew eee 75 A 2 1 Tour of the interface
88. roximation algorithm is used in the form of a NLP solver that precalculates proper estimations 6 This section will first describe the general theory of an NLP solver followed by the implementation of such a solver in this specific case NLP solver A NLP is a problem that can be stated as follows there is one scalar valued function f of several variables r here is a vector that we want to minimize subject to one or more other functions that serve to limit or define the values of these variables f is called the objective function or cost function while the other functions are called the constraints Of course the minimization function could be replaced by a maximization function Formally we have min f z 18 CHAPTER 2 DESIGN AND IMPLEMENTATION OF THE DELAY ELEMENT SIMULATOR where X CR Basically several solutions of each parameter within a range of possibly suitable values are tried The algorithm then calculates the costs or error using the objective function and depending of the result of the error in comparison with previous results the solver tries a different possible solution This process repeats itself until a large portion of the parameter values within range have been evaluated within the bound aries Depending on the complexity of the problem to solve this could take a while When done the NLP solver returns the parameter values for which the evaluation of the cost function was minimal O
89. s The first graph shows the responses from all the inputs For the second graph the coaxial delay for each individual input are prepended to the responses creating the final result In this plot the vertical space between successive cumulative path responses should be equal to Ar Convert ring settings to voltages All calculated and s have to be converted to voltage levels that can be sent directly to the amplifier boards The conversion from the ring parameters to voltage levels is pretty straightforward and will be discussed in Chapter 5 where the interaction with the controller board is described 1A subVI is equivalent to a function subroutine or method in other programming languages and useful for encapsulating code that will be reused multiple time A subVI is also used to develop hierarchical programs 3 3 OBFN SIMULATOR IMPLEMENTATION 35 Compensate for crosstalk The simulator has the ability to use a crosstalk matrix to compensate for linear crosstalk effects These effect are caused for the most part thermally meaning that the heat of heater a has an effect on some other heater b Again details about the compensation method will be discussed in Chapter 5 3 3 OBFN simulator Implementation 3 3 1 Connection matrix One of the requirements of the OBFN simulator is scalability In theory the simulator we have created is only limited to the number of rings within the delay elements for which precalcula
90. several AOAS 67 4x1 OBFN simulated 68 4x1 OBFN measured response 69 Screenshot of the delay element simulator at startup 76 Screenshot of the OBFN simulator at startup 82 8x1 PHSMABI chip layout lt 4 04 pd ees ea oo 91 List of Tables Ll 2 1 3 1 3 2 3 3 3 4 5 1 5 2 5 3 5 4 5 9 5 6 B 1 Subset of the original requirements for the SMART project Theoretically feasible delays per delay element for B 0 09 Calculation of the total path delays rss Connection matriz Ex OBEN iu sus xor ER Ev US Additional coaxial delays dox oo Re XO mulation Settings lt n c coco xo ko RRR RO RR Voltage levels used for measuring output responses by switching between coaxial cables outside the styrofoam box Example measurements for determining the voltage relationship of 09485 70 P a BE ee Coefficients for the voltage curve Coefficients for the voltage curve fit aooaa a Crosstalk matrix The numbers on the top and on the left denote the heater numbers suus uoo o9 om v RS X ox v RO RE RS xi LIST OF TABLES Abbreviations AE AOA API CLI DAC DC DE DFD DFS DLL DUT DVB s EDFA FPU FSR GUI MMSE MVC MZI
91. stability test of one ORR and another test to verify the output voltages of the controller board with the values that were sent to it 5 2 1 System stability Stability tests can be very time consuming All the measurements described in this chapter do not take longer than an hour so we are interested in knowing the level of 56 CHAPTER 5 MEASUREMENTS OSBF out6 Figure 5 3 Labelling of all heaters inputs and outputs of the optical chip stability in the course of that period For exactly one hour at a regular interval of 2 minutes the output response of a single ORR r1 was measured and saved During the measurements nothing of the setup was touched or moved and all measurements were performed with the blinds shut to prevent direct sunlight and windows closed for minimizing draft The current ramping was set to a ramp time of t 0 624708 seconds which roughly means having a coverage of t T FSR 32Ghz with an FSR of 14GHz The results of all these responses are shown in Figure 5 4 As we can see the curves are almost exactly aligned meaning that for at least the measurement period the system can be labeled as stable To further investigate the stability the fluctuations of the resonance frequency and the changes of the maximum delay during the hour were examined In Figure 5 5 we see that the resonance frequency drift is almost negligible and does also not show any trend The stability of the maximum delay
92. standard form a4z a 12 az where represents the coefficient 2 4 Manual A complete system user manual and Application Programmers Interface API Docu mentation are included in appendix A The manual is written in such a way that it can be used independently of this report The API provides more insight in the functional hooks that can be used to extend the simulator or to be called from other programs 2 5 Summary and conclusions When we look at the predetermined requirements we were able to meet a substantial amount of them The simulator is programmed in LabVIEW combined with Matlab for parts of the code Both programming environments are the de facto standard when it comes to developing simulation tools Scalability a very important requirement is met The simulator simulates delay elements of basically an infinite number of rings 2 5 SUMMARY AND CONCLUSIONS 2T Ringsettings for index n Kappas Figure 2 15 A representation of the data structure containing the polynomial coefficients Using higher powered computers the coefficients of the approximation algorithm of larger delay elements can be calculated in less time bringing the simulation of larger OBFN in the near future to be feasible The simulator is build using an event based programming model The speed acquired by the use of this technique is tremendous and also keeps the computer available for other tasks wh
93. sured voltage show an upward trend This could be corrected for directly by a multiplier constant either in the simulator or in the microprocessor Examples of compensated voltages are also shown in the two figures where a multiplier constant of 1 0131 has been chosen for both channels The maximum absolute voltage difference now drops to less than 0 1 volt 5 3 Chip characterization To properly convert calculated ring settings to voltages the properties of the individual ORRs must be known This section describes the measurements performed to get the relation between ring parameter and voltages for the individual rings During one day the chip has been re characterized a few times to see if any changes occurred No significant changes were observed 5 3 1 Kappa calibration For the simulator to calculate the proper voltage levels for a particular value of the relation between kappa and voltage must be determined We use to control the height of the group delay peak where usually lies anywhere between 0 5 and 1 Lower values of amp create excessive delays which are undesirable The value of amp is converted to dcoupler using Equation 2 6 since in theory it is proportional to V 60 CHAPTER 5 MEASUREMENTS Voltage percentage Absolute voltage difference uncompensated Relative voltage difference uncompensated Absolute voltage difference compensated Relative voltage differenc
94. tab These are the values in volts that will be sent to the proper channel as set in the Mapping tab Next to the final voltages array a few switches and buttons are visible When the Enable hardware 84 APPENDIX B OBFN SIMULATOR DOCUMENTATION Property Value Baud rate 115200 Data bits 8 Parity None Stop bits 1 Flow control None Delay before read 500 Table B 1 COM port settings in for the OBFN simulator write switch is active commands are sent to the controller board When disabled only the new voltages are calculated but no commands are sent One can choose between two write types One uses a single combined command and is thus faster The other uses single commands for each voltage to be set The latter can be used for debugging purposes The last switch on the tab determines which connection type to use Either directly via COM or indirectly using the Console Debug Tool For the fastest results the combination wrchall and TCP must be used Tab Angle The Angle tab contains a knob to simulate the angle of the incoming satellite signals Note that this simulator is currently suitable for a one dimensional antenna array and thus one angle suffices The angle can be set between 60 and 60 degrees corresponding to the system specifications as shown in Table 1 1 In the lower left the normalized delay difference between the antenna elements is shown The two indicators in the top corners warn the user when the del
95. ted parameter coefficients exist In our case we are thus limited to OBFNs containing delay elements no larger than five rings Each delay element is given a unique ID denoted by DE where zx is a increasing number For the case of the 8 x 1 OBEN we have an exponentially increasing number of rings in each stage as shown in Figure 1 3 To label the delay elements in an orderly fashion we use a DFS algorithm Formally a DFS is a search that progresses by expanding the first child node of the search tree that appears thereby going deeper and deeper until a goal node is found or until it hits a node that has no children Then the search backtracks returning to the most recent node it has not finished exploring A depth first search can be performed on many types of graphs In our case we have a binary tree and thus infinite recursion cannot occur A short formal algorithm in pseudo code is given in Listing 3 1 8 To make things more clear the 8 x 1 OBFN has been traversed using the DFS algorithm The order taken is accordingly to the DFS algorithm and is displayed in Figure 3 4 This gives us the following 7 delay elements e DE Delay line with 1 ring e DE Delay line with 2 rings e DE Delay line with 1 ring e Delay line with 4 rings e Delay line with 1 ring e Delay line with 2 rings e DE Delay line with 1 ring To represent this tree we use a matrix Every row represents a path and every column represents on
96. tellite and use the elevation and azimuth information to calculate the correct tunings of the ORRs The ORRs are tuned in such a way that there is constructive interference of the Radio Frequency RF signals coming from the desired direction During the course of this research a system will be developed that uses a simplified scenario consisting of linear array and therefore dealing with only 1 variable angle Description Value Frequency range 10 7 12 75 GHz K band Scan angle 60 to 60 degrees Selectivity lt 2 degrees continuous tuning No elements 8 Element spacing 1 5cm or 40ps Maximum delay 2ns Table 1 1 Subset of the original requirements for the SMART project 4 CHAPTER 1 INTRODUCTION AEs 1 E O OBFN gt Rx angle 3 control Figure 1 2 A high level overview of the system from input to processed output Stage 0 Stage 1 Stage 2 Figure 1 3 A 8x1 binary tree OBFN for a transmitter phased array antenna with 8 inputs 1 output and 8 optical ring resonators 1 2 3 Optical Beam Forming Networks OBFN The optical chip in this system is manufactured using planar optical waveguide tech nology by LioniX B V 1 It consists of the following building blocks waveguides Mach Zehnder Interferometers MZIs couplers and ORRs ORRs are chosen because they provide True Time Delay TTD so beam squinting will not occur Beam squint usually occurs when workin
97. tely long theoretically One possibility to achieve delays smaller than one is to set at a proper value and change the phase shift in such a way that the lower parts of the curve are within the bandwidth region of interest or put differently shifting away from resonance gradually However the NLP solver persistently finds another set of optimal parameters where there is a sudden phase shift from on to off resonance and the region of interest is exactly between two resonance peaks see Figure 2 12 Optimal parameters have been determined using the same MMSE method as ex plained before and again curve fitted for DEs containing 1 to 5 rings Because of the relatively small degree of the fitted polynomial a sudden change on the transition from 1 to gt 1 would give rise to serious errors Therefore the ring settings cal culation method chooses the correct data file containing the parameters according to the required delay In Figure 2 12 an ORR response is shown with the parameters set for a normalized delay D of 0 5 The region of interest is centered around 0 In 24 CHAPTER 2 DESIGN AND IMPLEMENTATION OF THE DELAY ELEMENT SIMULATOR Resonance frequencies bun ct N Normalized group delay i 05r 1 1 1 1 Lj 0 2 0 15 0 1 0 05 0 0 05 0 1 0 15 0 2 0 25 Normalized frequency Figure 2 12 Response for a normalized delay of 0 5 for a delay element containing 1 ORR Figure 2 13
98. the University of T wente the development of a broadband integrated optical beamformer based on ORRs in CMOS compatible waveguide technology has been done The next step is controlling this optical chip in a manageable way The 1 2 BACKGROUND 3 SMART system has a long list of requirements The early prototype implements a subset of these requirements shown in Table 1 1 The main advantages of the SMART concept are e Low loss and large instantaneous bandwidth e Continuous tunability high resolution Relatively compact and light weight realization e Inherent immunity to EMI e Potential for integration with optical distribution network 1 2 2 System overview A full system overview of the SMART system is shown in Figure 1 2 When used at the receiving end the AEs collect radio waves coming from a satellite These signals are converted from the electrical domain to the optical domain E O block by intensity modulation and afterwards fed into the OBFN The OBFN is used to apply appropriate delays on each optical input and combining them After combining the signals a strong optical signal is acquired which can then be converted back to the electrical domain O E block Finally a receiver can process the signal for example a Digital Video Broadcasting via Satellite DVB s set top box The OBFN shown in the system overview is managed by a control system Ideally the control system would automatically track a specific sa
99. the output of the network analyzer the voltage was adjusted to match the expected phase response Finally the coupier equivalent of is calculated The results of this process are shown in Figure 5 9 Note that the phase shift detected by the network analyzer shows an offset This offset was determined using the method that will be described in Section 5 5 1 This offset tends to fluctuate and should be repeated when doing comparisons of measurements At this moment the cause of the fluctuations is unknown and should be further investigated The relations are almost linear in V but because the slightest change of has a large effect on the actual group delay especially for large delays only the best possible fit is good enough To determine what the impact will be on the loss of precision an error measure is determined The error value is calculated by taking the absolute difference between the lineair approximation and the measured value The maximum error based on measurements of the first three rings r1 r2 and r3 is 0 0512 rad The small curvature of the measured slope has not been further examined and for now the resulting marginal error is not taken into account in further calculations Notice that the slope of the characterization of the third ring is negative This is because the natural delay the delay when 0 volt was applied caused by the ring was too high Because of this the characterization was done in the range of m to
100. tor described in the previous chapter serves as a building block Only the most 3 2 OBFN SIMULATOR DESIGN 31 Ring Section Figure 3 1 UML model of the layered architecture for the OBFN simulator important activity is displayed being a change in the required AOA by the user The activity is aimed at calculating the required voltages per channel for the amplifier boards with only an AOA as input Details about the intermediate steps are described below The flow of data passes a few stages mentioned in the following paragraphs Calculate 7 When we assume that the antenna array is not curved and the AE spacing is constant then for a specific AOA the time between the arrival of the satellite signal Ar at AE and AE is also a constant This constant can be calculated as follows 180 sin X d 3 1 where a is the AOA d is the AE spacing and c is the speed of light roughly 3 10 The idea is depicted in Figure 3 3 Of course the A7 can be normalized by dividing it by the RTT which we will use for the remaining part of this thesis Calculate total path delays Because of the fact that coaxial delay offsets are used to provide the means to tune for both positive as negative AOAs the actual delay that has to be realized by the OBFN itself must be determined Therefore a small algorithm is used that determines the highest coaxial delay offset and uses this offset as a reference point for determining the
101. trical crosstalk depending on the polarity of the voltages In our measurements a total of 3 rings needed to be tuned having a total of 6 heaters Fortunately a large part of the crosstalk effects can be ignored Firstly the effect of any heater on a MZI can be ignored The two branches of the MZI are so close together that the spreading heat causes both of them to warm up roughly the same amount The resulting phase difference is not influenced by this Visual determination using the network analyzer confirms that there is no effect This means that half of the crosstalk matrix values can be left blank Secondly because of the large distance between ring r2 and rings ri and r3 see Figure 5 3 the effects of the heat are almost unnoticeable Compensation requires only change of a few hundredths of volts For the sake of simplicity these small effects will be ignored For our proof of concept the effects of the amp heaters on the other heaters were not taken into account Note that they do have an impact on the final output response and should be considered in following projects 5 4 1 Measurement execution The influence of each heater element to another is measured using the steps below 1 Characterize all ORRs on the optical chip once and 5 4 CROSSTALK 65 9 10 Set all rings to 0 volt Pick a ring i Tune ring to 27 Do not take offsets into account Pick a ring j the crosstalker
102. trolled by a few dozen parameters which are all set by hand one by one This is not only error prone but also too time consuming A better more rigid and less time consuming solution is thus needed A generic piece of control software will not only 1 2 CHAPTER 1 INTRODUCTION Figure 1 1 Example of a phased array antenna the Cobra Dane Radar in Sheyma Alaska built in 1976 help current researchers work with the OBFNs but will also greatly reduce the time and effort future researchers will have to spent on yet to be created OBFNs based on ORRs Besides the direct results of this work future use of these types of integrated sys tems consisting of smart antennas and intelligent software could provide new ways of communicating between moving objects As a result the project as a whole could bring an interesting new technology and new exiting applications one step closer to consumers and companies 1 2 Background To get a better understanding of the complete system of which the controller software will be part of an overview is given in this section 1 2 1 SMART project The SMart Antenna systems for Radio Transceivers SMART project is aimed at pro viding live television services on airplanes through DV B s by developing a novel antenna for airborne reception of satellite signals using a broadband conformal phased array antenna The SMART project is a collaboration of different companies and research institutes At
103. urn values e coefficients The optimal coefficients for the given bandwidth and delay e error The error comparable to the ripple error phasefun2 The actual function to be optimized for a set of unknowns 80 APPENDIX A DELAY ELEMENT SIMULATOR DOCUMENTATION Parameters e Denotes an array of parameters to solve for Return values e mu The error for this function for the parameters tested Appendix B OBFN Simulator Documentation B 1 General information The LabVIEW OBFN simulator has been built using LabVIEW 8 5 on a Microsoft Windows XP Professional operating system The simulator depends on several Matlab routines that were developed in Matlab 7r14 so a proper installation of Matlab is also required For LabVIEW to find the proper Matlab routines Matlab should have a path reference file menu set path pointing to the directories simulator OBFN and coefficients lookup table Due to caching of code changes in MatLab code are not immediately effective in LabVIEW The best way to circumvent problems related to cached code is to completely restart LabVIEW B 2 Manual This section will describe the OBFN simulator from a user s perspective B 2 1 Tour of the interface A complete overview of the system interface is shown in Figure B 1 Unfortunately the GUI does not fit on a standard screen resolution of 1280 x 1024 which makes scrolling necessary We see three main parts The top part consists of tab contain
104. ve and restore settings would be of great value Not only does this improve the operating speed but also prevents mistakes to be made in each initial simulation setup 5 Generic Although the simulator is tailored to the available 8 1 OBFN chip the simulator itself must be capable of simulating future chip designs with different tuning etc For instance the new liquid crystal based version which will become available in the near future 6 Allowance for reusability Since the Ring Section Simulator RSS will be part of a bigger software system later on the software needs to be reusable T Operating system independence Because of the wide variety of operating systems commonly used nowadays it would be nice to make use of programming envi ronments that are available on different platforms 2 2 Delay element simulator design With the requirements in mind we can begin to design the simulator The simulator will be built in National Instruments LabVIEW LabVIEW is a graphical programming environment that enables the rapid development of test measurement and control applications LabVIEW is also capable to comprise with all of the aforementioned requirements and therefore the tool of choice This section describes the steps taken to design the simulator starting with the initial design the flow of data and the software structure 2 2 1 Dataflow Dataflow is a software architecture based on the idea that changing the value of
105. ween any PC and the microcontroller using conventional networks A good candidate would be the NXP 4 3 4 Floating point operations The microcontroller that was used does not have a Floating Point Unit FPU for processing floating point values like 2 1283 Floating point values are needed for sce narios 2 and 3 as previously described However using external libraries a software implementation can be used that mimics the operations of the missing FPU To enable this feature some specific compiler options must be added to the compiler within the CrossStudio development environment The steps that need to be taken are explained in Appendix C 52 CHAPTER 4 DESIGN AND IMPLEMENTATION OF THE MICROCONTROLLER SOFTWARE 4 4 Summary The software now running on the microcontroller has improved stability and is altered for further extension Three scenarios were described of which the first is used during the rest of this thesis Handles and functions for implementing the other two scenarios have partly been implemented API documentation is provided in the Appendix Chapter 5 Measurements As mentioned earlier in Chapter 3 the OBFN simulator is also capable of controlling the real OBFN system by sending commands to the controller board This chapter describes measurements that were performed to properly initialize the simulator for the specific OBFN under test After that measurements done to verify the correctness of the simulations
106. when using other kinds of topologies the overflow algorithm might need some adjustments 38 CHAPTER 3 DESIGN AND IMPLEMENTATION OF THE OBFN SIMULATOR Listing 3 2 Pseudo code overflow algorithm Parameters totalpathdelays contains the total delays of all paths minmaxdelays matrix of the minimum and maximum delays achievable for delay elements of different lengths connectionmatrix the connection matrix numberofinputs the number of inputs of the OBFN Algorithm Initialize a matrix delays with zeros do for each input i delayleftover totaldelay totalpathdelays i do for each delay element j of this path if connectionmatrix i j is set and delays i j is 0 then do if delayleftover larger than 0 delayleftover totaldelay sum of delays in row i if delayleftover is smaller than maxdelay newdelay delayleftover else newdelay maxdelay end if Then force this new delay to all the positions in the same column j where the connectionmatrix is not 0 end if end if if delayleftover is larger than 0 then throw an overflow error end do end do 3 3 OBFN SIMULATOR IMPLEMENTATION 39 Full DOA compensation delay compensation Optical Beam Forming Network 2 out E EO OG i amp 7 8 Figure 3 5 Coaxial delay lines prepended to the OBFN The length of the lines is propor tional to the required additional delay
107. within the OBFN For the simulation a path instance is created containing multiple delay element in stances The creation of the delay element instances is done according to the value on position i j in the connection matrix where the i row is the representation of the path and the value of position 7 7 the length of the delay element The parameters for each delay element are obtained using the same methods as described in Chapter 2 The input signal is fed through all the newly created instances where the results are added When examining the connection matrix multiple non empty values in a column indicate a shared delay element among different paths For example DE is being used by four paths The delay of the fifth path solely depends on which means that the value of DE becomes a constant when the target delay for the fifth path is known The required A7 delay difference between the fifth and sixth path thus have to be realized by just one ring To calculate the distribution of delays while keeping the shared paths in mind an overflow algorithm is created The code is given in listing 3 2 We made use of the fixed structure of the binary tree structure and the DFS algo rithm to create the connection matrix One of the useful properties of the connection matrix when using this approach is that the number of rings in each row of the matrix is decreasing from left to right from which we can benefit now Note that this also means that
108. y of changing the number of inputs changing the physical layout of the chip changing the coaxial delays from the antenna elements to the actual chip and last but not least changing the different ORR ring parameters like loss and circumference 3 Resource usage The simulator must work fluently even on an every day computer A proper design of the simulator makes the most out of the available CPU cycles and thereby maximizing the speed and responsiveness of the simulator 4 Ease of use The simulator s GUI must be readily usable for anyone who has 29 30 CHAPTER 3 DESIGN AND IMPLEMENTATION OF THE OBFN SIMULATOR knowledge of OBFNs Clutter and unimportant input elements must be either hidden or not be shown at all 5 Generic Although the simulator is tailored to the available 8 x 1 OBFN chip the simulator itself must be capable of simulating future chips with different types of tuning For instance the new liquid crystal version that will become available somewhere in the near future 6 Operating system independence Because of the wide variety of operating systems commonly used nowadays it would be nice to make use of programming envi ronments that are available on different platforms 3 2 OBFN simulator design The OBFN simulator is again for the most part created in LabVIEW For additional functionality and specific algorithms we will revert to MatLab Appendix B contains a list of the functions that were written in

Modelling, Simulation and Implementation of an Optical Beam

Contents

Download Pdf Manuals

Related Search

Related Contents