Copyright. Brie Howley

Transcription

1 Copyright by Brie Howley 2006

2 The Dissertation Committee for Brie Howley Certifies that this is the approved version of the following dissertation: Optical Waveguides for Control of Antenna Arrays Committee: Ray T. Chen, Supervisor Joe C. Campbell Hao Ling Dean Neikirk Grant Willson

3 Optical Waveguides for Control of Antenna Arrays by Brie Howley, B.S.; M.S. Dissertation Presented to the Faculty of the Graduate School of The University of Texas at Austin in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy The University of Texas at Austin May 2006

4 Acknowledgements I would like to express my gratitude to Professor Ray T. Chen for his supervision and advice throughout the research and preparation of this work. Also, I would like to thank my committee members, Professors Joe Campbell, Hao Ling, Dean Neikirk, and Grant Willson, for their valuable time and interest in this research. Special thanks to my colleagues Xiaolong Wang, Yihong Chen, Zhong Shi, Yongqiang Jiang, Lanlan Gu, Qingjun Zhou, and Jinha Kim for their cooperation, assistance, and advice. Finally, I would like to thank my family; my parents for their encouragement and support and most of all my girlfriend, Sandra McGinnis, for her love and patience. Brie Howley University of Texas at Austin March 18, 2006 iv

5 Optical Waveguides for Control of Antenna Arrays Publication No. Brie Howley, Ph.D. The University of Texas at Austin, 2006 Supervisor: Ray T. Chen Many communication and radar applications require antennas with high directionality and narrow beam width. Phased array antennas (PAA) are able to meet these requirements and have the added benefit of agile beam steering without physical movement of the antenna structure. It is generally expected that future PAA systems will be designed to operate across ultra-wide bandwidths. It will be necessary to use true time delay (TTD) steering techniques, rather than the phase shifters, in order to meet these large bandwidth requirements and avoid beam squint. In contrast to optical fiber delay lines, waveguide optical delay lines, defined by photolithography, are able to deliver precise delays for PAA systems. A procedure for fabricating low loss polymer optical waveguides has been developed. Single mode waveguides with propagation losses as low as 0.38 db/cm were fabricated with UV curable polyacrylate materials. A propagation loss measurement technique was invented that decouples the waveguide coupling and bend losses from the propagation loss. Optical switch technology was evaluated for use in an integrated v

6 polymer optical TTD device. Polymer digital optical switch (DOS) and total internal reflection (TIR) switch devices were characterized. An integrated TTD device structure was described. The n-bit delay device was composed of optical waveguide delay lines and 2x2 optical switches. Waveguide trench and offset structures were introduced to reduce the bend loss of the waveguide delay lines. Offsets reduced the mode coupling loss associated with straight to curved waveguide transitions while trenches confined the mode in curved waveguides to prevent radiation losses. Optimization of the trench and offset structures enabled a 50% reduction of the polymer waveguide delay line bend radius. Fully integrated polymer TTD devices were fabricated and packaged. The 4-bit TTD device insertion loss was 14.5 db with a 2.9 db variation in the loss dependant on the activated delay state. Two optical phased array system structures were presented that used the proposed TTD devices to steer the radiation pattern of an array antenna. A wavelength multiplexed optical delay system was implemented to demonstrate the performance of the fabricated 4-bit TTD devices. The far field radiation patterns of 1-dimensional (1D) and 2- dimensional (2D) X-band array antennas were measured and compared to simulations. Transmitting frequencies of and Ghz were steered to angles of -15, 0, and 15 degrees with no beam squint effect. vi

7 Table of Contents List of Tables...x List of Figures... xi Chapter 1: Introduction to phased array antennas Introduction The Phased Array Concept True time delay for phased arrays Overview...5 Chapter 2: True time delay technology for phased array antennas Introduction Electrical phase shifter and delay lines Optical delay line technology Summary...12 Chapter 3: Fabrication of Low-Loss Passive Polymer Waveguides Polymer waveguide loss mechanisms The waveguide fabrication process Optimization of the waveguide structure Summary...26 Chapter 4: Coupling Insensitive Waveguide Loss Measurement Propagation loss measurement techniques A waveguide test structure for measuring loss Experimental results of the WTS loss measurement Summary...32 Chapter 5: Optical Switches Overview of optical switch technology Mach-Zehnder optical switch...38 vii

8 5.3 Thermo Optic Digital Optical Switches Thermo Optic Total Internal Reflection Switches Summary...49 Chapter 6: TTD Device Structure and Optical PAA System Architecture Introduction Time delay requirements Delay device structure Device insertion loss Optical PAA system architecture Wavelength multiplexed system Summary...59 Chapter 7: A 2-bit Polymer True Time Delay Device Introduction Design and Fabrication Measurement Results Summary...65 Chapter 8: Bend Loss Reduction Structures Low index contrast waveguide systems Trench and offset structures Design of trench and offsets for fabrication Experimental procedure and results Summary...81 Chapter 9: An Integrated 4-bit Delay Device TTD Device Design Device loss calculation Device fabrication process Device Packaging Device performance Summary...99 viii

9 Chapter 10: An X-band Phased Array Antenna Demonstration System Structure System Setup and Optimization Two Dimensional (2D) Array Far Field Measurements One Dimensional (1D) Array Far Field Measurements Summary Chapter 11: Future Research Directions & Conclusion Future PAA requirements Conclusion Bibliography Vita 128 ix

10 List of Tables Table 3.1: DOE parameters for optimization of the polymer RIE process Table 4.1: Summary of WTS dimensions Table 7.1: Delays, Losses and Switch States for each Delay Path Table 9.1: Computed Steering Angles and Respective Delays Values for Each Table 9.2: Element Computed Steering Angles and Normalized Delay States for Each Element Table 9.3: Calculated Delay Line Lengths Table 9.4: Simulated losses of waveguide bends Table 9.5: Design Parameters Table 9.6: Switches activated for each delay state Table 11.1: Projected TTD Performance Requirements x

11 List of Figures Figure 1.1: A linear array of antenna elements Figure 1.2: Simulation result showing beam squint for a linear array with 64 elements. Nominal steering angle is 60 from broadside... 4 Figure 3.1: Buried channel waveguide processing steps Figure 3.2: SEM picture, 6000x maginification, of a polymer waveguide structure showing grass features and a trench along the waveguide base Figure 3.3: SEM picture, 9000x magnification, of an improved polymer waveguide. The grass and trench are not present Figure 3.4: Schematic showing how ions may be deflected by the sloped side walls of the photoresist Figure 3.5: Schematic showing how trenches are formed at the base of the waveguide core as the ions etch the polymer material Figure 3.6: A picture from an optical microscope, 800x, showing air bubbles next to the core structure of a polymer waveguide Figure 3.9: (left)patterned hard mask of Cr/Au material Figure 3.10: (right)patterned SiO 2 hard mask Figure 3.11: (top)sem image of a waveguide core etched with the optimized RIE Figure 3.12: parameters using Cr/Au mask material (bottom)sem image of a waveguide core etched with the optimized RIE parameters using SiO 2 mask material Figure 3.7: (left) SEM image showing anisotropy of the waveguide core Figure 3.8: (right) High magnification SEM image showing sidewall roughness Figure 4.1: A schematic of the demonstrated WTS Figure 4.2: Top surface view of 635nm light coupled into set B of the waveguide test structure Figure 4.3(a): Propagation loss measurement results for a correctly coupled WTS Figure 4.3(b): Propagation loss measurement results for an incorrectly coupled WTS. 32 Figure 5.1: Overview of optical switch technologies Figure 5.2: An opto-mechanical switch design Figure 5.3: A 3D MEMs optical switch array (left) and a picture of MEMs mirrors and actuators Figure 5.4: Comparison of technologies competing with TO-polymer switches Figure 5.5: A PLC based Mach-Zehnder switch structure Figure 5.6: The 2x2 DOS architecture. Both the cross (top) and bar states are represented with their respective electrodes Figure 5.7: Schematic of the setup used to measure the optical performance of the switches Figure 5.8: Photograph of the optical switch measurement setup Figure 5.9: Measured bar port optical response of a 2x2 DOS Figure 5.10: The DOS bar state switching speed measurement results Figure 5.11: Structure of a TIR thermo-optic switch and magnified view of the X- junction with a heater xi

12 Figure 5.12: SEM photograph of a portion of the X junction core Figure 5.13: Optical microscope picture of the X-junction region of a fabricated 2x2 TO TIR switch Figure 5.14: TIR switch optical power response to the electrical power consumption. 48 Figure 6.1: Time delay needed as a function of RF frequency for a 4x4 element subarray Figure 6.2: TTD device structure employing waveguide optical switches and delay lines Figure 6.3: Schematic of the antenna system architecture for demonstrating the proposed true time delay devices Figure 6.4: Schematic of a wavelength multiplexed optical PAA system Figure 7.1: Schematic of the 2-bit reconfigurable delay device Figure 7.2: Switching speed measurement results of the 2x2 DOS switch Figure 7.3: Near field images of TE (left) and TM (right) polarized emanating from the 2-bit device Figure 7.4: (Top) Phase versus RF frequency measurement results for each of the four delay states Figure 7.5: (Bottom) Time response of a femtosecond laser pulse through the device Figure 8.1: A diagram of the waveguide core using offset and trench (hatched area) Figure 8.2: structures Mode profile of a straight 6x6 µm channel waveguide (a) and one with a radius of R=3 mm (b). The outline of the waveguide core is superimposed on the mode intensity contours. n Co =1.46, n Cl =1.45, λ=1.55 µm Figure 8.3: Simulated 180º bend loss as a function of the trench separation of 4 waveguide bend radii. The waveguide core is 6x6 µm, n Co =1.46, n Cl =1.45, n Tr =1, λ=1.55 µm, TE polarization Figure 8.4: Figure 8.5: Figure 8.6: Figure 8.7: Simulated optimal offset values for 6x6 µm waveguides with and without trench structures. d=7 µm, n Co =1.46, n Cl =1.45, n Tr =1, w Tr =20um, λ=1.55 µm, TE polarization The BPM simulated bend and junction loss for a 180º bend as a function of waveguide bend radius for cases (1)-(4). The waveguide dimensions are 6x6 µm and the trench separation is optimized at 7 µm with an operating wavelength of 1.55 µm Top view of a waveguide junction employing both a trench and an offset SEM cross-sectional image of the trench structure. Approximate location of the waveguide core is indicated by the square to the left of the trench Figure 8.8: Simulated and measured insertion losses of 180º waveguide bends as a function of the bend radius. Data points are measured values and lines are fitted curves to the simulation data points Figure 8.9: Measured PDL for waveguide bends employing trenches and offsets.. 80 xii

13 Figure 8.10: Comparison between measured and simulated insertion loss values for 180º waveguide bends of cases (1)-(4) Figure 9.1: An array of antenna elements showing the corresponding variables important to design of a TTD device Figure 9.2: TTD device layout with design parameters Figure 9.3: Process flow for building a TTD device Figure 9.4: Schematic of packaged device showing both optical and electrical assemblies Figure 9.5: Schematic of the design PCB for electrical connection to the TTD device Figure 9.6: Four fully packaged TTD devices Figure 9.7: Normalized insertion loss as a function of delay state Figure 9.8: Phase vs. frequency response for each delay state produced by a packaged TTD device Figure 9.9: A comparison between measured and designed delay values Figure 10.1: The wavelength multiplexed PAA system structure Figure 10.2: Photograph of the optical PAA system Figure 10.3: Photograph of the receiving antenna and spectrum analyzer Figure 10.4: Schematic of the electrical circuit used to control the optical switches. 104 Figure 10.5: Tunable resistor bank for minimizing the switch crosstalk Figure 10.6: Plot of S11 for the X-band antenna element Figure 10.7: Measured single element far field pattern Figure 10.8: Measured and simulated far field patterns for a 3x3 element array (0 degree steering angle) Figure 10.9: Measured and simulated far field patterns for a 3x3 element array (15 degree steering angle) Figure 10.10: Measured and simulated far field patterns for a 1x8 element array (0 degree steering angle) Figure 10.11: Measured and simulated far field patterns for a 1x8 element array (-15 degree steering angle) xiii

14 Chapter 1: Introduction to phased array antennas 1.1 INTRODUCTION Many communication and radar applications, such as search, track, guidance, and passive detection radar as well as high bandwidth communication between mobile, fast moving objects, require antennas with high directionality and narrow beam width. Typically, the width and directivity of the radiation pattern of a single radiating antenna element is insufficient for these applications. One method of improving this limitation is to increase the element aperture. A typical approach is to place the radiating element at the focal point of a large parabolic dish. However, for many mobile platform applications this is not a practical approach. With a narrow radiation pattern, the antenna aperture must be steered to direct the signal. For a large dish antenna it may be difficult to position the aperture quickly enough due to inertia effects of the massive antenna. An alternative method for increasing the antenna aperture without the adverse consequences of a large dish structure is to use an antenna array which is an assembly of radiating elements in a given geometrical pattern. The beam pattern of an array antenna can be adjusted by any of five parameters: the geometrical configuration of the array, the displacement between the elements, the excitation amplitude, the excitation phase, and the radiation pattern of the individual elements. As will be shown, by adjusting the phase of the signal delivered to each element it is possible to control the steering angle of the beam without physically repositioning the antenna aperture. Such an antenna is called a phased array antenna (PAA). Because a phased array antenna does not need to be physically repositioned, its reaction time can be several orders of magnitude faster than a large dish antenna. This allows PAA systems to be multifunctional, meaning they can perform many different tasks simultaneously. 1

15 1.2 THE PHASED ARRAY CONCEPT For a linear array of radiating elements the far field radiation pattern is expressed as a function of time, t, and angle from the broadside direction, Φ, as N E( φ, t) = A exp( iω t) exp[ i( ψ + nk Λ sinφ)] (1.1) n= 1 n m n m where A n is the radiation pattern of the individual element, n, ω m is the operating microwave frequency, k m =ω m /c is the wave vector, Ψ n is the nominal phase shift and Λ is the distance between radiating elements. 1 It can be seen from this equation that the direction of the radiation peak can be controlled by manipulating the phase between the array elements. By varying the successive phase front between elements, the peak can be orientated in the desired direction. For example, to point the peak at an angle Φ 0, the phase shift, Ψ n, is set to the value ψ n = nk nλ sinφ 0. (1.2) where the nominal wave vector is equal to the operating wave vector (k n =k m ). Figure 1.1 shows a linear array of antenna elements. The wave front propagation direction is controlled by the constructive interference of each individual element radiation pattern. By applying a linear phase excitation across the array, the wave front can be steered from broadside (Φ 0 = 0º) to any desired angle. 2

16 Φ o Figure 1.1: A linear array of antenna elements. 1.3 TRUE TIME DELAY FOR PHASED ARRAYS It has been shown that the peak direction of the beam deviates as a function of frequency according to ω m = tan φ (1.3) ωn φo 0 where ω n is the nominal frequency, Φ 0 is the nominal pointing direction, and Φ 0 is the change in the pointing direction for a change in the operating frequency, ω m. 2 There is no shift in beam direction for a broadside steering angle (Φ 0 =0 ), but as the steering angle changes to either side, the beam direction shifts from the desired steering angle based on the frequency of operation. This phenomenon, referred to as beam squint, is undesirable for most applications. A simulation of beam squint for a linear array is shown in Figure 1.2 where the number of antenna elements is 64 and the desired steering angle is 60 degrees from broadside. For a deviation of 1 GHz, there is approximately 10 degrees of beam squint. 3

17 Figure 1.2: Simulation result showing beam squint for a linear array with 64 elements. Nominal steering angle is 60 from broadside. It is generally expected that future phased array antennas will be designed to operate across ultra-wide bandwidths. It will be necessary to use true time delay (TTD) steering techniques, rather than the phase delay techniques described above, in order to meet these large instantaneous bandwidth requirements and avoid beam squint. With TTD, the steering angle, Φ 0, is decoupled from the frequency so that the peak can be controlled independently without beam squint while changing the frequency of operation. TTD functions by lengthening the microwave feeds which introduces a path difference between radiating elements. The microwave exciting the (n+1) antenna element propagates through an additional line with a length of D n =-nl(φ 0 ). 3 The length of this delay line introduces a time delay, t n, of 4

18 ( nλ sinφ0 ) t n ( φ0 ) = (1.4) c for the (n+1) element. Ψ n is given by ψ n = ω t ( φ 0 ). (1.5) m n With such a delay design, when the second phase term of equation (1.1) is changed due to the frequency change, the first term is adjusted to compensate so that the radiation peak will be directed towards Φ 0 at all frequencies. 1.4 OVERVIEW In the next chapter a survey of the various methods of generating true time delays will presented, including optical and electrical approaches. Optical waveguide delay lines, which are the focus of this dissertation, will be explained in detail. A comparison of material systems will be discussed. Chapters 3 and 4 will discuss the fabrication procedure for low loss polymer waveguide and a new method for measuring the loss of these waveguides. As any optical delay device needs a method of selecting the appropriate delay path, optical switches will be discussed in Chapter 5. Chapter 6 will present an integrated waveguide delay device structure as well as an optical PAA system structure. Chapter 7 will show results from a proof of concept, 2-bit polymer delay device. In order to obtain more functional delay devices and to improve device performance while decreasing the device size, bend loss reduction structures are presented. Simulations and measurements show that these structures are effective for 5

19 reducing the minimum bend radius of a low contrast optical waveguide. Chapter 9 will present a monolithically integrated 4-bit polymer delay device which uses bend loss reduction structures. A 2-dimensionally controllable optical phased array antenna system is then constructed using these 4-bit delay devices. Results from the system demonstration are presented in Chapter 10. Finally, Chapter 11 discusses future research targets. 6

20 Chapter 2: True time delay technology for phased array antennas 2.1 INTRODUCTION This chapter will provide an overview of methods used to provide true time delays (TTD), both in the electrical and optical domains. A comparison of the advantages and disadvantages of each approach to polymer waveguide delay lines will be shown. Conventional antenna array feed lines use bulky, heavy metallic waveguides or coaxial cable components. As higher frequency array operation is developed, the antenna element spacing will become increasingly compact, raising the concern of waveguide congestion. These waveguide and coaxial transmission lines can also be susceptible to substantial loss, crosstalk and electromagnetic interference (EMI). Additionally, the electrical transmission lines can be dispersive leading to restricted operating bandwidths. Optical fiber and optical waveguide transmission lines are excellent candidates for the transmission and control of RF signals in future wide bandwidth PAA systems. Optical waveguides and fiber provide low propagation loss of RF and microwave signals, immunity to electromagnetic interference and reduced system size and weight. In contrast to fiber delay lines, waveguide optical delay lines, defined by photolithographic methods, are able to deliver precise delays with sub-picosecond resolution for phased array antenna systems. Additionally, optical waveguide delay lines can be formed in a compact planar lightwave circuit (PLC) with waveguide-based optical switches. This technique has the potential for occupying less space than systems using optical fibers, such as MEMs and acousto-optic based delay systems. Several material systems have been used to fabricate waveguide optical delay lines, the most notable being silica, 4 silicon on insulator (SOI), 5 and polymers. 6 7 In

21 comparison to the alternative material systems, polymer waveguides are easily fabricated on almost any substrate of interest thereby reducing manufacturing costs and increases the potential for single chip integration with active PAA components, such as lasers, modulators, and detectors. The refractive indices of polymer waveguide materials can be adjusted widely to address the compromise between coupling losses and bending losses. Additionally, the thermo-optical coefficient, n/ T, of polymer materials can be more than an order of magnitude greater than that of SiO 2, 7 while the thermal conductivities of polymers are a fraction of SiO 2 and silicon. Because of these properties, polymers can be used to make thermo-optic switches with power consumptions less than silicon or oxide based devices ELECTRICAL PHASE SHIFTER AND DELAY LINES Conventional microwave electrical phase shifters use a non-planar ferrite rod structure inside a waveguide or microstrip transmission line. 9 By applying a current to a wire coil surrounding the structure, the magnetic dipole moment of the ferrite rod can dynamically influence the phase of a microwave or RF signal traveling through the device. Because this technology is mature, the cost of these devices is relatively low. The drawback is that this is a pure phase shift approach which results in low system bandwidth. There are also several types of electrical delay line techniques which combine microstrip transmission lines with PIN diodes, 10 FET switches, 11 or MEMs switches. 12 These approaches boast a compact device size and larger bandwidths of linear phase control than the ferrite phase shifter approach. However, the loss can vary greatly over large operating bandwidths due to the switch design and microstrip transmission lines. 8

22 Additionally, the MEMs approach requires expensive hermetic device packaging to insure device reliability. 2.3 OPTICAL DELAY LINE TECHNOLOGY The use of photonic systems for PAA s offers unique features for high performance antenna systems while at the same time meeting the challenging weight and size requirements Recent advances in optical communications technology have resulted in reliable, high performance components such as optical waveguides, lasers, modulators, switches, detectors, and optical fiber. 8,27-44 These optical components are suitable for system implementation targeted at the optically fed true time delay for PAA applications. There are a large number of optical true time delay technologies. The technologies can be divided into broad band techniques and narrow band techniques, including acoustic-optic (AO) integrated circuits 45,46 and Fourier optical techniques. 47,48 As the true interest in optical systems for antenna delay lines is the benefit of ultra wide bandwidths, a few of the relevant optical broad bandwidth technologies will be described in detail Substrate-guided wave structures A promising set of optical TTD structures based on substrate-guided wave fanouts have been demonstrated With these structures, an RF modulated optical signal is split and collimating fibers are then used to couple the light into a glass substrate by either a volume hologram or a mirror. Due to total internal reflection (TIR), the wave bounces through the substrate and is coupled out by a volume hologram placed at a prescribed location on the substrate surface. The delays are generated by the total path 9

23 length that the light travels within the substrate. Depending on the design, the substrate guided wave structures generate either a discrete pool of delays or a continuously variable delay path driven by a wavelength tuning and volume hologram dispersion mechanism. The disadvantage of the discrete delay pool type modules is the time delay intervals are limited by the thickness of the substrates and ultimately the diameter of the optical beam spot. For the continuously tunable type modules, it is not possible to scale the system to practical PAA dimensions due to the narrow bandwidth of the dispersive holograms Optical fiber based delay systems Wavelength Division Techniques Molony 49 used a Bragg fiber grating to implement a TTD system. Equally spaced, highly reflectivity Bragg gratings of different central wavelengths were distributed along a single fiber. The location in the fiber where the light was reflected was determined by the optical wavelength. A drawback of this approach is that multiple gratings need to be fabricated on one fiber. Furthermore, small delay increments are not possible because the gratings must be prevented from overlapping. Corral 50 designed a chirped fiber Bragg grating in order to implement a continuously tunable TTD system. The delayed light waves were separated by a wavelength division demultiplexer and then fed to each antenna element after photodetection. The system is complex due to the need for a separate tunable light source for each antenna element. Additionally, the delays generated by fiber Bragg gratings drift if the fiber is not temperature controlled due to the thermal expansion of the fiber. 10

24 Wavelength Dispersive Techniques Esman 17 demonstrated a fiber optic prism TTD antenna feed. A wavelength tunable optical carrier modulated with an RF signal was passed through a set of nominally equal delay fibers with differing net dispersion. The time delay was generated by tuning the optical carrier wavelength. The drawback to this technique is the extremely long lengths of dispersive fiber needed to achieve sufficient delay. Jiang 51 improved on the dispersive fiber approach by using photonic crystal fiber with a dual core structure in place of conventional dispersive fiber. The photonic crystal fiber has a dispersion of 600ps/km nm, which is six times greater than the dispersion of index guiding fiber resulting in a more compact structure. However, the ultimate limitation of dispersive fiber techniques is the tuning speed of the wavelength tunable source which limits the PAA steering angle speed Optical waveguide based delay systems In contrast to optical fiber delay lines, waveguide optical delay lines, defined by photolithographic methods, are able to deliver precise delays with sub-picosecond resolution for phased array antenna systems. Additionally, optical waveguide delay lines can be integrated with lasers, modulators, switches, and detectors to form a planar lightwave circuit (PLC). Such a PLC is ideal for reconfigurable delay lines due to its compact size. III-V and silica Sullivan 52 proposed a GaAs based optical waveguide TTD system. The TTD device consisted of GaAs waveguide delay lines and 2x2 EO two-mode interference (TMI) switches. The drawback of GaAs material based waveguide devices is the large propagation loss of the GaAs waveguides in the C-band communication window. 11

25 Additionally, the coupling losses to optical fiber will be large due to mode mismatch of the high index waveguides. Horikawa 53 proposed a photonic-switched TTD beam-forming network using silica waveguide circuits. Once again the programmable delay lines were composed of 2x2 switches and fixed length delay lines. However, the author did not mention the loss of the optical switches or waveguides, so a performance comparison is not possible. Polymer Waveguides Tang 6 proposed using polymer waveguides and surface relief gratings to produce true time delays for phased array antennas. By placing the gratings at the desired position of a long polymer waveguide, light can be coupled vertically out of the waveguide and collected by collimating fibers. This type of device requires an extremely complicated packaging design in order to effectively couple out a large number of beams simultaneously. Other work on polymer waveguide based TTD devices has concentrated on the design of large scale integration of delay lines. However, to date, there have been no reports of functional, fabricated polymer devices. 2.4 SUMMARY A survey of electrical and optical based techniques for producing phase delays and true time delays has been presented. Though requiring a more complicated system structure, optical true time delay techniques have the advantage of ultra wide instantaneous bandwidths. Several optical based methods for generating time delays have been described including substrate guided wave, fiber based, and optical waveguide based designs. While each technique has merit, polymer optical waveguides are 12

26 relatively easy to fabricate and are uniquely suited for integration with other active optical components. 13

27 Chapter 3: Fabrication of Low-Loss Passive Polymer Waveguides 3.1 POLYMER WAVEGUIDE LOSS MECHANISMS In order to develop an optically controlled true time delay system based on waveguide delay lines, low loss waveguides must be fabricated. The most limiting factor for the practical implementation of optical waveguide true time delay networks is the propagation loss encountered in optical components. Because of the need to boost system performance by lowering the loss and increasing the system gain, system designers naturally prefer not to forfeit this performance by using high loss optical components. Polymer materials offer attractive features for the production of optical waveguides. In comparison with other waveguide materials, polymers are relatively easy to process because expensive tooling, other than conventional clean room equipment, is not involved. Additionally, the near IR absorption of recently developed fluorinated polymer materials allows these polymers to compete with typical waveguide materials such as SiO 2 and silicon. In fact, new fluorinated polymers could potentially exhibit lower losses than the high purity glass currently used in telecommunication optical fibers. 54 There are two categories of loss from waveguide structures: intrinsic and extrinsic. Intrinsic loss is attributable to the inherent properties of the material used for the waveguide. Intrinsic loss includes absorption and volume scattering effects. Material absorption is a function of the wavelength of the light transmitted and the nature of its interaction with the molecular structure of the material. For polymer materials, this absorption at near infrared wavelengths is caused by excitation of the C-H bonds comprising almost all organic materials. Absorption peaks are formed at the fundamental 14

28 frequency of the bond oscillation and also at the overtone frequencies that lie close to the 1550nm center wavelength of the C-band telecommunication window. A common method used to reduce these absorption peaks is to introduce fluorinated polymer materials where a majority of, if not all, C-H bonds are replaced with C-F bonds which do not exhibit the same resonance. 54 Scattering effects can be part of both the intrinsic and extrinsic loss. Volume scattering occurs when small inhomogenities in the material lead to distortions in the refractive index causing the light to be refracted. If the refraction angle is greater than the acceptance angle of the waveguide, the light will escape the waveguide confinement and contribute to the waveguide loss. This effect is present to some extent in any nonideal material. From a device fabrication point of view, the choice of material used in the fabrication is the only means of affecting the intrinsic loss. Great care must be taken in the selection of a high quality polymer material suitable for the amount of acceptable loss at the wavelengths of interest. In fact, even though considerable steps have been made in reducing the intrinsic loss of optical polymers, many of these polymers are not adequate for use in polymer waveguides with sufficient loss values THE WAVEGUIDE FABRICATION PROCESS As compared to the intrinsic losses, care in the fabrication process can minimize the extrinsic losses. Absorption, surface scattering, or coupling losses can all cause extrinsic losses. The fabrication process is briefly described below to provide a context for discussing extrinsic losses. A schematic of a typical fabrication process for a buried channel polymer waveguide is shown in Figure 3.1. A polymer bottom cladding material is spin coated 15

29 onto a clean substrate with the thickness of the layer determined by the spin speed (a,b). After UV and thermal curing, a second layer of polymer is spun which serves as the core layer (c). This core layer has a slightly higher refractive index for guiding the light by total internal reflection. A suitable thickness of a hard masking material is then deposited (d) followed by photoresist which is defined by photolithography (e,f). The hard mask is then patterned by either a wet or dry etching method depending on the hard mask material. Once the hard mask is properly defined, reactive ion etching (RIE) is used to form the channel waveguides in the core material (g). The remaining hard mask is then removed (h) and a polymer top cladding layer is spin coated and cured (i). The substrate is then diced and the end faces are polished to ensure effective coupling into and out of the waveguide. Figure 3.1: Buried channel waveguide processing steps. 16

30 3.3 OPTIMIZATION OF THE WAVEGUIDE STRUCTURE In order to minimize extrinsic losses, great care must be taken in forming the core structure of the waveguide, as this is where the majority of the optical power is concentrated for a single mode waveguide. The waveguide sidewall smoothness is very important to reducing extrinsic surface scattering losses Additionally, impurities cannot be introduced into the polymer material. The masking material that contacts the polymer core layer must not diffuse into the polymer during processing. Proper tolerances of the core dimensions will insure that the mode field diameter of the radiating light is as close as possible to that of the single mode optical fiber used to couple light into and out of the waveguide Removal of grassy formations During the process development for constructing a polymer waveguide, many obstacles were encountered which resulted in waveguides with losses much greater than would be expected from the intrinsic loss values alone. Figure 3.2 is a SEM picture of an early waveguide. The picture was taken after the RIE process and before the top cladding was deposited. The grass-like structures surrounding the waveguide core are a result of sputtering of surrounding aluminum material. 62 It was originally thought that the aluminum anodization of the bottom electrode of the RIE tool was sufficient to prevent sputtering of the aluminum metal by the accelerated ions in the chamber. However, repeated attempts to obtain a grassless topology proved futile until the bottom aluminum electrode was covered with a silicon wafer. An eight-inch silicon wafer was clamped to the bottom electrode and the sample was placed on the silicon wafer. In this manner, the eight-inch wafer served as the bottom electrode. Figure 3.3 shows an SEM of the 17

31 resulting waveguide structure. The grass features are not present because sputtering from the aluminum electrode was no longer possible. By removing the grass, potential scattering sites in the cladding of the waveguide were removed. Figure 3.2: SEM picture, 6000x maginification, of a polymer waveguide structure showing grass features and a trench along the waveguide base. Figure 3.3: SEM picture, 9000x magnification, of an improved polymer waveguide. The grass and trench are not present. 18

32 3.3.2 Elimination of trench formation during RIE Additional surface scattering sites can be formed when the top cladding is deposited. If the sidewalls of the core structure are undercut or if a trench is present at the base of the waveguide core, air bubbles become trapped when the top cladding is spun. If these bubbles cannot escape before the top cladding is cured, they will act as surface scattering sites attached to the waveguide core. Figure 3.2 also shows that a trench has formed at the base of the core. The trench is approximately 1µm wide and greater than 5µm deep. High viscosity polymer typically will not fully fill this trench during the spin coating process. Based on this work and that of others, 59,63 we hypothesize that the profile of the masking material plays an important role in the formation of the trench. Originally the mask consisted of an aluminum layer, 1500 Å thick, coated with a photoresist layer, 0.9 um thick, which was used to pattern the aluminum layer. However, by removing the photoresist layer with an acetone soak procedure, and thus using only the aluminum layer as the RIE masking material, the trench was completely removed, as seen in Figure 3.3. Figures 3.4 and 3.5 show a possible mechanism for the formation of the trench when the photoresist is present. Because the sidewalls of the photoresist are not perfectly vertical, impinging ions may be reflected by the sidewalls of the photoresist and impact the core polymer layer near the base of the aluminum mask. Because more ions come in contact with this small region at the base of the mask structure, this material is etched at a greater rate than the surrounding polymer. Once a small well is etched to a sufficient depth, ions inside the well may become confined, causing them to deflect inside the well and further increase the etching rate. In this way, very deep, narrow trenches could be formed at a rate much faster than the overall bulk etch rate. 19

33 Figure 3.4: Schematic showing how ions may be deflected by the sloped side walls of the photoresist. Figure 3.5: Schematic showing how trenches are formed at the base of the waveguide core as the ions etch the polymer material. Figure 3.6 shows a top down view of waveguides with trenches. Small air bubbles can be seen next to the waveguide cores. Propagation loss values for waveguides 20

34 identical to those shown in Figures 3.2 and 3.3 were measured. For waveguides with grassy features and trenches, the average propagation loss is 1.80 db/cm. By removing the grass and trenches, the loss is reduced to a value of 1.51 db/cm. Nonetheless, this value is larger than expected considering that the intrinsic material loss specified by the polymer supplier is only 0.35 db/cm at the 1550 nm wavelength. Bubbles near waveguide core 20um Figure 3.6: A picture from an optical microscope, 800x, showing air bubbles next to the core structure of a polymer waveguide Hard mask material selection Several hard mask material systems were examined to determine their effect on the overall propagation loss. Choices of hard mask materials included aluminum (Al), 21

35 chromium/gold (Cr/Au), and silicon dioxide (SiO 2 ). Aluminum significantly degraded the performance of the polymer. Waveguides formed with aluminum hard masks exhibited extremely high polarization dependant loss (PDL). The exact mechanism of this degradation has yet to be determined but values for the propagation loss of the TE polarized light was 1.55 db/cm while that for a TM polarization was in excess of 20 db/cm. One possibility is that the aluminum diffuses into the polymer material and is not fully removed during the aluminum wet etch. Because the aluminum is concentrated near the top surface of the waveguide core, TM polarized light is highly absorbed while the TE polarization is relatively unaffected. A combination of a thin layer of chromium and a thicker gold layer improved the results. The measured propagation loss with this material system was as low as 0.45 db/cm with no measurable PDL. However, it was observed that the edges of the patterned Cr/Au mask were not smooth. The rough mask edge pattern was transferred to the polymer through the RIE process resulting in rough waveguide sidewalls. This roughness was caused by the wet chemical etching of the metal hard mask. In order to combat this problem, silicon dioxide (SiO 2 ) was employed as the hard mask material. SiO 2 has the ability to be dry etched with an RIE plasma. The dry etch process yields smoother hard mask edges thereby reducing the roughness of the polymer waveguide sidewalls. Figures 3.9 and 3.10 are SEM images of the patterned Cr/Au and SiO 2 hard masks, respectively. Figures 3.11 and 3.12 are SEM images of waveguide cores formed using Cr/Au and SiO 2 as the hard mask, respectively. The figures show that the roughness of the waveguide sidewall is greatly reduced by using SiO 2 as the hard mask. 22

36 Figure 3.9: (left)patterned hard mask of Cr/Au material. Figure 3.10: (right)patterned SiO 2 hard mask. 23

37 Figure 3.11: (top)sem image of a waveguide core etched with the optimized RIE parameters using Cr/Au mask material. Figure 3.12: (bottom)sem image of a waveguide core etched with the optimized RIE parameters using SiO 2 mask material. 24

38 3.3.4 RIE process optimization An optimization study was performed on the polymer RIE processing parameters in order to further decrease the waveguide propagation loss. The RIE parameters that were studied were process temperature, pressure, and RF power. Table 3.1 shows the design of experiment (DOE) parameters that were used. Table 3.1: DOE parameters for optimization of the polymer RIE process Parameter High Value Low Value Pressure (mtorr) Temperature ( C) 25 0 Power(Watt) The measured effects of the DOE were the etch rate, anisotropy, and sidewall edge roughness. Through experimentation and analysis it was determined that a combination of high temperature, low pressure, and high power resulted in channel waveguides with the smallest anisotropy and sidewall roughness. Figures 3.7 and 3.8 are SEM images of the waveguide core. Figure 3.7 shows a profile of the core and the markers indicate the dimensions used for the anisotropy measurement. Figure 3.8 shows the sidewall roughness at high magnification. Propagation losses of polymer waveguides formed with a SiO 2 hard mask were as low as 0.38 db/cm when an optimized etch process was used. This is near the lower limit set by the material absorption, measured from the planar waveguide propagation loss to be 0.35dB/cm. 25

39 Figure 3.7: (left) SEM image showing anisotropy of the waveguide core. Figure 3.8: (right) High magnification SEM image showing sidewall roughness. 3.4 SUMMARY Intrinsic and extrinsic loss mechanisms can be controlled to yield low loss polymer waveguides. Intrinsic loss is managed by choosing a high quality polymer with low loss in the optical communication window of 1550 nm. Extrinsic loss is minimized by careful processing in order to remove scattering sites near the waveguide core. Grassy features were removed by replacing the aluminum bottom electrode of the RIE chamber with a silicon wafer, thus prohibiting sputtering and re-deposition of aluminum. Undesirable trenches, which caused air bubbles to be trapped near the waveguide core, were removed by stripping the photoresist prior to etching the core. Hard mask materials were studied and silicon dioxide was shown to yield waveguides with smooth sidewalls. A DOE was performed for the RIE process parameters and low pressure, high power conditions were shown to yield waveguides with low surface roughness and anisotropy. By combining all of these processing techniques, polymer waveguides were fabricated with propagation losses as low as 0.38 db/cm. 26

40 Chapter 4: Coupling Insensitive Waveguide Loss Measurement 4.1 PROPAGATION LOSS MEASUREMENT TECHNIQUES In order to accurately characterize optical waveguide devices, it is necessary to distinguish the waveguide propagation losses from coupling and bending losses. There are several well known methods for measuring the propagation loss of optical waveguides. Some of the most common propagation loss measurement techniques include the cut-back method, 64 the sliding prism-coupling method, 65 the scattered light measurement method, 66 the Fabry-Perot Cavity technique, 67 and the ring resonator method. 68 However, each of these measurement techniques has a drawback resulting in inaccuracies. The cut-back, sliding prism, and Fabry-Perot methods are inaccurate because the output of the sample must be coupled multiple times at several interfaces. Each time the sample is cleaved and polished or the prism location is moved, the coupling efficiency changes slightly. For waveguides with low propagation losses, the coupling losses can be similar in magnitude to the total propagation loss. As a result, changes in the coupling efficiency during the measurement will have a significant effect on the measured propagation loss. The scattering loss measurement method exhibits measurement variations attributable to the mechanical motion of the detector. Distance variation between the waveguide and the detector results in an unwanted perturbation in scattered photon collection. For the ring resonator measurement technique, the propagation loss cannot be entirely isolated from the bending loss of the ring. Additionally, this method requires a highly stable laser with very narrow spectral control and large tuning ability to achieve accurate results. The negative effects of each of the above mentioned measurement 27

41 methods cause the measured waveguide propagation loss to be dependent on the loss measurement technique A WAVEGUIDE TEST STRUCTURE FOR MEASURING LOSS A new propagation loss measurement technique is described which is both quick and easy to perform and also accurate and repeatable. The technique uses a lithography defined waveguide test structure (WTS) to segregate the propagation loss from other losses in the sample. Additionally, this test structure provides a loss measurement method that is independent of the coupling condition. Repeatable waveguide loss data can thus be acquired. A WTS was devised that could ensure that the propagation, coupling, and bending losses for multiple waveguides would be equal. A schematic of the WTS used to make the propagation loss measurement is shown in Figure 4.1. The WTS consists of three waveguide sets, each set consisting of a test line and a calibration line. Each test line has a different propagation length and consists of straight waveguide sections and a single 180-degree bend. Because each of the waveguides terminates at a common surface only one cleaved and polished interface is used in the measurement. By using a fiber array with a pitch that matches to the test structures, all waveguides exhibit the same coupling loss under the fabrication condition that the waveguides have equal dimensions. Additionally, each set of test lines has the same bending radius and bend propagation length enabling equal bend losses. The calibration lines, all of equal length and bend radius and consequently equal insertion loss, are used to ensure the sample is properly aligned to a fiber array. 28

42 Figure 4.1: A schematic of the demonstrated WTS. A large fiber array with 250 µm pitch was used to couple light into and out of each of the three test sets. The total propagation lengths are 44.21, 19.61, and mm for the test lines of sets A, B, and C, respectively and 11.4 mm for the three calibration lines. Note that each set represents two groups of waveguides with two different curvatures. The radius of curvature for each of the test lines is 3.75 mm and that of the calibration lines is 3 mm. 4.3 EXPERIMENTAL RESULTS OF THE WTS LOSS MEASUREMENT To evaluate the new method, a polymer waveguide chip was fabricated on a silicon substrate. The waveguides formed with this mask structure were single mode polymer channel waveguides, each with core dimensions of 7 µm width, measured by optical microscope, and 6.8 µm height, measured by profilometer. The polymer material used was Zen Photonic s ZPU12 series UV curable polymer. Table 4.1 summarizes the 29

43 dimensions of the WTS. Figure 4.2 shows the top surface of set B of the WTS with 635 nm light coupled into the waveguides through a glass fiber array. Table 4.1: Summary of WTS dimensions Test Lines Calibration Lines Bend Radius (mm) Total Length (mm) Core Height (µm) Core Width (µm) Set A Set B Set C Set A Set B Set C Figure 4.2: Top surface view of 635nm light coupled into set B of the waveguide test structure. 30

44 Figure 4.3(a) shows the plot of insertion losses for each of the test lines as a function of their corresponding propagation length. A line was fitted to the data using linear regression. The slope of this line represents the propagation loss/cm and the y- intercept is the combination of the coupling and bending losses. The propagation loss for the polymer waveguides was measured to be 0.88 db/cm and coupling and bending losses were 1.76 db. Additional measurements were made with an intentionally misaligned fiber array. The measurements results are shown in Figure 4.3(b). In this case the combination of coupling and bending loss understandably increased to 2.22 db due to the misalignment; however the propagation loss value was unchanged. It is to be noted that the shift of waveguide coupling loss due to misalignment was the same for all waveguides since all were coupled simultaneously using a 250 µm linear fiber array which assured an equal displacement for all channels. The bend radius of each group of test lines is equal. Therefore, the actual value of the bend radius plays no role in the propagation loss measurement results. In this work we measured three sets of waveguides to provide an accurate slope for each line shown in Figures 4.3(a) and 4.3(b). 31

45 Insertion Loss(d y = 0.88x (a) Insertion Loss(dB) y = 0.88x (b) Waveguide Length(cm) Waveguide Length(cm) Figure 4.3(a): Propagation loss measurement results for a correctly coupled WTS. Figure 4.3(b): Propagation loss measurement results for an incorrectly coupled WTS. 4.4 SUMMARY A waveguide test structure (WTS) was proposed and demonstrated which is capable of providing accurate propagation loss measurements, regardless of the coupling condition, using only one coupling end face. The testing method is both simple and reliable and the test structure provides the ability to decouple the propagation loss from the coupling and bending losses. Using the measurement technique, a polymer channel waveguide loss was determined to be 0.88 db/cm. The test method reported here is suitable for channel waveguides made out of any material. The accuracy of the loss measurement is dependant on the number of sets of waveguides measured. More sets will result in more data points and increased accuracy of the fitted line. 32

46 Chapter 5: Optical Switches 5.1 OVERVIEW OF OPTICAL SWITCH TECHNOLOGY Optical switches are essential components of an optical time delay device. Optical switches must redirect light to the desired optical delay line to control the steering angle of the antenna. There are several design requirements that an optical switch must meet for satisfactory performance of a TTD system, the most important of which include: size, loss, switching speed, reliability, polarization independent operation, cross talk, power consumption, and the ability for integration with passive waveguides that form the delay paths. The performance of an optical switch device is dependant on the switching mechanism and material used to fabricate the device. Figure 5.1 shows a breakdown of commercially available optical switch technologies. Figure 5.1: Overview of optical switch technologies. 33

47 5.1.1 Liquid crystal switches Liquid crystals (LC) can be used in switches as well as the more familiar use in displays. In both cases, the LC is a thin layer between a pair of parallel glass plates. Applying a voltage across the LC layer alters the LC molecule orientation between one state that rotates the polarization of light passing through it, and a second state that does not affect polarization. With suitably arranged polarizing optics, the LC can function as a switch. Liquid crystal switches separate the input light into two polarizations, which are reflected from liquid crystal plates. Applying a suitable voltage changes the polarization of the reflected light, which the polarizing optics deflect differently than if its polarization is not rotated, switching it between output ports. The switching speed of an LC device is limited by the molecular movement of the liquid crystal material. Additionally, LC switches require complex optics and are highly polarization dependant Opto-mechanical switches Opto-mechanical switching depends on the mechanical motion of switch components using any of a variety of switch designs. The moving component may be a mechanical slider, solenoid, stepper motor, mirror or lens. In each design, the moving part directs light into different output fibers. Opto-mechanical switches operate over a wide wavelength range with typical switching times of 10 ms to 25 ms. These devices are the most mature and inexpensive of switching technologies. Because of the potential error associated with moving parts, opto-mechanical switches are used mainly where signal switching capability is vital but where switching is infrequent. Important 34

48 examples include protection switches to route signals around broken cables. Figure 5.2 shows the working principal of an opto-mechanical switch. Figure 5.2: An opto-mechanical switch design MEMs switches Optical micro electro-mechanical systems (MEMS) are semiconductor-based devices that use mounted mirrors to direct light waves. These devices are manufactured using common integrated circuit techniques. In most 2D arrays, a mirror flaps up and down at the intersection of various input and output ports. When the mirror is down, the light beam passes over the mirror and into an output port. When the mirror is up, the signal is deflected and sent to a second output port. In 3D arrays, two or more mirrors are used to redirect the light. Figure 5.3 shows the working principal of a 3D MEMs optical switch array and a portion of a mirror array with actuators. MEMs optical switches have become popular because of their low insertion losses and the ability to latch thereby reducing power consumption. 70,71 Switching speeds can be in the microsecond range. The main disadvantage of MEMs optical switches is the difficulty in integrating them with other waveguide based devices to form a planar lightwave circuit. 35

49 Figure 5.3: A 3D MEMs optical switch array (left) and a picture of MEMs mirrors and actuators PLC optical switches Planar lightwave circuit (PLC) optical switches are composed of optical waveguides fabricated on a substrate using integrated circuit fabrication techniques. Because optical fibers and bulk optics are not used, PLC switch designs have a clear advantage for integration with waveguide delay lines. Due to mature fabrication procedures, planar lightwave circuit (PLC) technology has become competitive for making reliable, low cost, optical waveguide switches. 72 Several PLC switch architectures have been demonstrated, including a digital optical switch (DOS), 73 a Mach-Zehnder interferometer (MZI) switch, 74 a directional coupler switch, 75 and TIR switches The optimal architecture for a given application depends on the waveguide material system used to fabricate the switch and the modulation mechanism. The modulation mechanism is the optical effect that transfers the light to the desired output waveguide. Examples of modulation mechanisms include: the thermo-optic (TO) effect, in which a change in the material temperature causes a change in the refractive index; the electro optic (EO) effect, in which an electric field applied across a material causes a change in the refractive index. A material s EO and 36

50 TO coefficients characterize its response to electric fields and changes in temperature, respectively. For the application of a polymer waveguide TTD device, polymer PLC waveguide switches are preferred. Polymer, electro-optic waveguide switches typically have poor reliability due to the degradation of the chromophores in the polymer that provide the refractive index modulation. Additionally, EO active polymers usually have high loss as compared to passive (non-eo active) polymer materials. These factors suggest the use of a polymer thermo optic waveguide switch. Polymer materials are best suited for such a switch because they have large TO coefficients (dn/dt~-10-4 ) compared to SiO 2 which has a dn/dt value of approximately These large dn/dt values means lower temperatures are needed and hence less power consumption for a polymer thermo-optic switch than for a similar oxide based switch. Figure 5.4 summarizes the comparison between TO-polymer based switches and the switch structures described earlier in this chapter. 37

51 Figure 5.4: Comparison of technologies competing with TO-polymer switches. There are currently three switch architectures that are suitable with polymer materials using the TO effect; Mach-Zehnder, DOS and TIR. These architectures will be explained in the following sections. 5.2 MACH-ZEHNDER OPTICAL SWITCH A Mach-Zehnder optical switch consists of a Mach-Zehnder interferometer with a thin-film phase shifter deposited on the waveguide arms of the interferometer. The phase of the optical waves traveling in the interferometer can be tailored through the TO effect by heating the waveguide or through the EO effect by applying a voltage across an EO active waveguide material. The resulting phase of the light determines the output port. Figure 5.5 shows the Mach-Zehnder waveguide structure. Incoming light passes through a directional coupler that splits the light into two waves. The phase of the optical waves in one (or both) arm(s) can be modified by thin-film phase shifter(s). If the phase shifter 38

52 is not activated, the two waves recombine in-phase and exit the switch at output port 1. If the phase shifter is activated, the two waves recombine out-of-phase and exit at output port 2. Figure 5.5: A PLC based Mach-Zehnder switch structure. Although the Mach-Zehnder structure is mature, it has several drawbacks. First, because the Mach-Zehnder switch works by the interference of optical waves, the optical bandwidth of the device is small. Additionally, because the active mechanism is relatively weak, the interaction length of the arms of the interferometer must be quite long, resulting in a large device. For these reasons the polymer TO DOS and TIR switch structures were selected for use in a polymer waveguide TTD device. The following sections describe these switch structures and their performance. 5.3 THERMO OPTIC DIGITAL OPTICAL SWITCHES x2 DOS structure 39

53 Figure 5.6 shows the 2x2 DOS structure. The switch is composed of two input and two output ports as well as four Y-junctions (or adiabatic splitters). There are eight electrodes placed near the arms of the Y-junctions. Electrodes 1 through 4 are connected within one closed circuit. When a current flows through electrodes 1 through 4, heat is produced which changes the refractive index of the polymer below the electrode. This causes the light from either input 1 or 2 to be directed straight through to the corresponding output channel. This switching state is known as the bar state. When current passes through electrodes 5 through 8, which are connected through a separate closed circuit, the light is switched to the opposite output channel. This is known as the cross state. Unlike Mach-Zehnder and directional coupler structures, as the electrical power is increased beyond the switching point, the optical power remains constant in the same output channel and does not oscillate between the output channels with a further increase in power. Figure 5.6: The 2x2 DOS architecture. Both the cross (top) and bar states are represented with their respective electrodes. 40

54 5.3.2 Measurement of DOS performance The setup used to measure the power consumption, extinction ratio, and insertion loss of the optical switch devices is shown in Figure 5.7. A randomly polarized broadband light source with a wavelength range of µm was connected to a single mode fiber. The single mode fiber was aligned and butt coupled to one of the input arms of the 2x2 switch. A single mode fiber array with a pitch equal to the separation of the output arms was butt coupled to the output side of the switch. Two fibers in the array were aligned to the two output arms of the switch. These two fibers were then connected to photodiodes that were monitored by separate power meters. A DC power supply was connected by probe needles to the switch electrodes. Figure 5.8 is a picture of the measurement setup. Figure 5.7: Schematic of the setup used to measure the optical performance of the switches. 41

55 Figure 5.8: Photograph of the optical switch measurement setup. The optical power reaching each of the photodetectors was measured as the DC power supply was ramped. The results are shown in Figure 5.9. It can be seen that the optical power values plateau at an electrical power of 300mW, the required power consumption to reach the fully switched state for the DOS device. The optical power difference between the two channels at this plateau is the crosstalk, which was measured to be -48 db. An interesting characteristic of the digital optical switch architecture can also be seen in Figure 5.9. When electrical power is not supplied, both output channels exhibit the same optical power throughput. But as the electrical power is ramped to activate bar port switching, the bar port optical power in not constant, but rather increases by 6 db. This is an effect of the position of the electrodes over the Y-branches. Normally these Y- branches each exhibit a 3 db loss, but because the heat changes the refractive index of the 42

56 two Y-branches that the light passes through, this 3 db loss is canceled and there is a net 6 db gain. dbm Power(mW) Cross Port Bar Port Figure 5.9: Measured bar port optical response of a 2x2 DOS. The insertion loss is measured using the same setup described above. The optical power through the output arms are measured in both the cross and bar states. The switch is removed and the input single mode fiber is then aligned and butt coupled directly with the fiber array and the optical power is measured again. The difference between the two measured values is the insertion loss. For both the bar and cross states, the insertion loss measurement was 8.88dB without power supplied to the electrodes and 2.88 db with 350 mw of DC power. Loss was measured with the polarization aligned both vertically (TM) and horizontally (TE) with respect to the substrate plane. The difference between these two measurements is referred to as the polarization dependant loss (PDL) and was measured to be less than 0.1 db for these switches. 43

57 The device length was 30 mm, which is relatively long for a PLC switch. In order to reduce the device length, the Y branch angle must be increased. However, this will cause an increase in the device loss as well as the power required to fully switch the optical signal. The optical response time was also measured. A 25 Hz square wave electrical signal was applied to the switches and the optical output fall and rise time responses were measured. Both the bar and cross state switching speeds were measured. Figure 5.10 shows the optical response time measurement for the DOS bar state. For both switching states, the rise and falls times were measured as 2.1 and 3.1 ms, respectively. Figure 5.10: The DOS bar state switching speed measurement results. 5.4 THERMO OPTIC TOTAL INTERNAL REFLECTION SWITCHES x2 TIR switch structure Among various PLC switch architectures, total internal reflection (TIR) switches have the merits of compact size, wavelength insensitivity, and multimode tolerance. 76,77 Figure 5.11 shows the structure of the 2x2 TIR switch. Two input waveguides of a 44

58 suitable pitch are connected to waveguide bends. The waveguide bends have a radius of curvature, R, which is large enough to provide negligible bending loss and guided mode perturbation. The waveguide bends are then connected by two straight waveguides to form an X-junction. Horn structures are introduced near the junction to reduce the cross talk and to operate the switch by a temperature gradient induced by a thin film heater formed by a layer of gold film deposited above the top polymer cladding layer. The input and output sides of the switch are symmetric. Without a current applied to the thin film heater, light launched into input 1 travels straight through the X-junction to output 2 (the cross state). The temperature of the heater increases when a sufficient driving power is applied causing a temperature gradient in the polymer below the heater. The refractive index of the heated polymer material is lowered and a TIR effect causes the light to reflect at the crossing point of the X-junction and propagate to output 1 (the bar state). Figure 5.11: Structure of a TIR thermo-optic switch and magnified view of the X- junction with a heater. 45

59 For the case of the TO TIR switch described herein, the pitch of the input/output waveguides is 250 µm and R is 10 mm. The most critical parameters of the switch are the taper length l, the junction width D, and the half branch angle θ. These parameters have been adjusted, through simulation and experimentation, to achieve the desired operating characteristics for the delay device. For the results reported, the values of l, D, and θ are 1442 µm, 50 µm, and 4, respectively Measurement of the TIR switch performance Figure 5.12 is an SEM photograph of the core of the fabricated X-junction and Figure 5.13 is a top down optical microscope picture of the X-junction region of a fabricated 2x2 TO TIR switch. Figure 5.12: SEM photograph of a portion of the X junction core. 46

60 Figure 5.13: Optical microscope picture of the X-junction region of a fabricated 2x2 TO TIR switch. Figure 5.14 shows the optical power in the bar and cross ports as a response to the electrical drive power supplied to the thin film heater. The tested switch has a cross talk of -31dB in the cross state and a power consumption of 0 mw. The zero static power consumption is a beneficial feature since it can reduce the average driving power in real applications. With an increase in electrical driving power, the optical power in the cross port decreases while the bar port optical power increases with the switch eventually reaching the bar state. The bar state power consumption is the driving power resulting in maximum optical power in the bar port, which is 44 mw while the bar state cross talk is - 32 db. 47

61 Figure 5.14: TIR switch optical power response to the electrical power consumption. The switching time of the TIR switch is determined by the thermal conductivity and thickness of the polymer. The total polymer thickness is approximately 20 µm. A 200 Hz square waveform with an amplitude of 1.3 V and an offset of 0.65 V was used to drive the switch heater. The optical response in the two output channels, as shown in Figure 5.15, demonstrates a delay of 1.5 ms for t rise and 2 ms for t fall. The total insertion loss (single mode fiber in, single mode fiber out) of the fabricated switch was 2.8 db in the bar state and 3.4 db in the cross state. The variation in the insertion loss values is due to the interaction of the light with X-junction and thermal gradient induced by the heater. The reflection of light from the thermal gradient causes less scattering than if the light passes directly through the X-junction. The device length was 19 mm and the PDL was measured to be less than 0.2 db in either switch state. 48

62 Figure 5.15: Time response of the TO TIR switch. 5.5 SUMMARY Optical switches are a critical component in the implementation of an optical time delay device. Optical switch technologies available for use in a polymer TTD device were summarized. These technologies include liquid crystal, opto-mechanical, MEMs, and PLC optical switches. Polymer thermo optic PLC switches are most suitable for integration with polymer optical delay lines in a TTD device. The performance of a polymer TO DOS device was evaluated. The DOS switch was found to have an insertion loss of 2.88 db, crosstalk of -48 db, PDL less than 0.1 db, and a response time less than 4 ms. The electrical power required to activate the switch was 350 mw. 49

63 The performance of a polymer TO TIR switch was also evaluated. In the bar state the TIR switch had 2.8 db insertion loss, a cross talk of -32 db, and 44 mw power consumption. In the cross state the switch had 3.4 db insertion loss, cross talk of -31 db and 0 mw power consumption. The response time of the switch was 2 ms and the PDL was less than 0.2 db. Because of their lower electrical power consumption and smaller device size, TO TIR switches are more suitable than TO DOS switches for integration in a waveguide TTD device. 50

64 Chapter 6: TTD Device Structure and Optical PAA System Architecture 6.1 INTRODUCTION As discussed in Chapter 2, waveguide transmission lines are excellent candidates for the transmission and control of RF signals in wide bandwidth phased array antenna (PAA) systems. This chapter introduces an n-bit true time delay device structure composed of waveguide delay and reference lines and 2x2 optical switches. Hypothetically, any waveguide material system could be used for the proposed device structure. However, because of their unique properties described in Chapter 2, polymers are ideally suited for an integrated, high performance device. In addition to the TTD device structure, an optically controlled phased array antenna system architecture will be presented later in this chapter. 6.2 TIME DELAY REQUIREMENTS For a rectangular array of antenna elements, the time delay required for each element of a PAA is given by t m, n Λ y sinθ sinφ Λ x sinθ cosφ = ( n 1) + ( m 1) (6.1) c c where (m,n) is the index of each element, Λ x and Λ y are the distance between elements in the x and y directions, respectively, c is the speed of light in vacuum, and θ and Φ are the beam steering angles in the elevation and rotational directions, respectively. Figure 6.1 shows the maximum time delay requirements for a 4x4 sub-array of a PAA with a ±45 51

65 scanning coverage in the elevation and azimuth directions over the frequency range of GHz (X, Ku, and K-bands). It is assumed that the antenna element spacing is fixed at one half the RF wavelength. Figure 6.1: Time delay needed as a function of RF frequency for a 4x4 element subarray. 6.3 DELAY DEVICE STRUCTURE An optical TTD device structure is described which provides the required time delays to each element of a PAA sub-array. The TTD device structure is based on thermo-optic polymer switches and polymer waveguide delay lines. The structure for the waveguide delay module is shown in Figure 6.2. A single mode input waveguide is coupled to a 2x2 optical switch (n=1). Two different lengths of polymer waveguides are positioned at the output ports of the n=1 switch. Depending on whether the bar state or cross state of the switch is chosen, light is delivered to waveguide reference line with 52

66 length, L, or a waveguide delay line with length, L+ L. The output of these waveguides are connected to another 2x2 optical switch (n=2) and the switch s output ports are coupled to two more waveguides of lengths L and L+2 L. This sequence is continued with lengths of the reference waveguides remaining at a length of L and the waveguide delay lines increasing in length according to L+ L 2 (n-1). The time delay, t, provided by these fixed delay waveguides is given by l ng t = (6.2) c where l is the sum of the waveguide lengths through which the light travels, c is the speed of light in vacuum, and n g is the group index of the waveguide. The last switch (n+1) of the n-bit delay device is controlled to deliver the optical signal to the output waveguide. 53

67 Figure 6.2: TTD device structure employing waveguide optical switches and delay lines. A key feature of this device structure is the layout of the delay lines, which are each composed of two straight segments of the appropriate length connected by a single 180 degree bend. This design minimizes the total length of curved waveguides thereby minimizing the total amount of device bending loss. Additionally, there is an excess path loss compensating effect with this layout. Longer delay paths will induce extra path loss on the optical signal. However, because the reference lines have a smaller bend radius, the bending loss will be greater for optical signals traveling through the reference paths. By choosing an appropriate bend radius for the reference line, the excess loss from the appropriate delay line can be offset so that all delay states will experience a uniform insertion loss. 54

68 6.4 DEVICE INSERTION LOSS The insertion loss is a critical parameter for such a TTD device. For a fully integrated device there will only be optical coupling and reflection losses between the input and output fibers and the planar lightwave circuit. For an n-bit TTD device, the total insertion loss will be the sum of the fiber to waveguide coupling losses (CL), the input and output reflection losses (RL), the loss from the n+1 optical switches (SL), the bending loss (BL) from the n waveguide delay lines, and the total path loss (PL) of the appropriate delay line length, l. The resulting equation for the insertion loss of an n-bit channel waveguide based TTD device is IL = 2 ( CL + RL) + ( n + 1) SL + n BL + PL (6.3) where c t Pr opagationloss PL = (6.4) n g The polymer waveguide is the basic building block of the TTD device. An improperly fabricated waveguide will significantly increase the propagation loss, causing both the overall path loss (PL) as well as the switch loss (SL) to increase. Additionally, the bending loss (BL) may also increase due to scattering effects if the waveguide edges are rough. Because of the critical need to minimize the total insertion loss of the TTD device, it is important that the waveguide efficiency be maximized and that an appropriate bend radius be chosen. 55

69 6.5 OPTICAL PAA SYSTEM ARCHITECTURE The system architecture for the demonstration of a phased array antenna based on these delay modules is shown in Figure 6.3. In this system, an electrical microwave signal is modulated onto a CW laser source operating at 1550 nm wavelength. At the output of the modulator, an erbium doped fiber amplifier (EDFA) is used to amplify the signal in order to compensate for the loss from the modulator. This amplified signal is then split via a 1xN optical splitter, where N is the number of elements or subarrays of elements to be delivered a TTD signal. The waveguide delay modules described above are then placed on each of these N branches. The optical signals from the output of the delay modules are fed into N photodetectors of the appropriate bandwidth for operation in the desired frequency spectrum. The photodetectors act as low bandpass filters, allowing only the microwave frequencies to pass, now with the appropriate time delays. The emanating electrical signal is then amplified by electrical microwave amplifiers and finally fed into each of the elements or subarrays of the antenna head. 56

70 Figure 6.3: Schematic of the antenna system architecture for demonstrating the proposed true time delay devices. A computer and control interface can be used to control the large number of optical switches in this system. Thus the delays needed for a desired steering direction can be computed and the appropriate switches activated automatically. Using a computer and control electronics that are sufficiently responsive, the steering angle response time of the system will be dictated by the optical switch response time. 6.6 WAVELENGTH MULTIPLEXED SYSTEM An alternative system architecture is now introduced which requires fewer delay devices for a given array size. If the antenna array is of a size N x M, the system will only require N + M delay devices rather than N x M devices as in the previous system 57

71 architecture. The drawback of this architecture is that a multi-wavelength optical source or multiple single wavelength lasers are required where the number of wavelengths is N. Additionally M+1 wavelength demultiplexers (DMUX) and one wavelength multiplexer (MUX) are also needed. A schematic of the system architecture is shown in Figure 6.4. In this system, an electrical microwave signal is modulated onto the multiple wavelength source. After the modulator, a single DMUX is used to separate the wavelengths. The first set of TTD devices is then inserted, one device for each of the N wavelengths. A MUX is then used to recombine the delayed wavelengths followed by an EDFA to compensate for the loss of the system up to this point. A 1 x M optical splitter is then used so that each wavelength is equally split among M different channels. The second set of M TTD devices is then followed by the final M DMUX to again separate the wavelengths. For the wavelength multiplexed system, the first stage of delay devices controls azimuth steering angle while the second stage delay devices control the elevation. Figure 6.4: Schematic of a wavelength multiplexed optical PAA system. 58

72 6.7 SUMMARY The design of an n-bit optical TTD device structure based on optical waveguides and 2x2 optical switches has been presented. By controlling the switch states, an optical signal can be appropriately delayed. Two antenna system architectures were presented which are suitable for optically controlling phased array antennas using the n-bit delay devices. The first system architecture uses a delay device for each element of the antenna array. The second system architecture employs wavelength multiplexing to reduce the number of delay devices but requires multiple lasers. 59

73 Chapter 7: A 2-bit Polymer True Time Delay Device 7.1 INTRODUCTION This chapter describes a proof of concept, 2-bit true time delay device composed of polymer waveguide delay lines and polymer thermo-optic switches. A previously reported 2-bit integrated polymer delay unit with delays of 47 and 94 ps had an insertion loss greater than 34 db, despite having glass optical fibers for two of the delay lines. 79 The delay device architecture was similar to the one discussed in this chapter. Such a high loss is impractical for actual antenna applications. In order to reduce the insertion loss, smaller waveguide propagation losses must be obtained by using passive polymer materials. Additionally, optimized waveguide bending radii must be used to minimize the propagation distance while maintaining appropriate time delays. The insertion loss and device size of the delay device reported in this chapter are reduced by using passive polymer materials and by reducing the bending radii, and thus the propagation length, of the waveguide delay lines. Using a rectangular channel waveguide structure rather than rib or strip loaded waveguide structures enables the mode profiles of the waveguides and fibers to be similar in shape and size, thus reducing the coupling losses. The single mode polymer channel waveguides have a numerical aperture (NA) of 0.17, which matches well with the telecommunication fiber employed (NA = 0.14 ). Full characterization results of the delay device are presented. 7.2 DESIGN AND FABRICATION Polymer waveguides were fabricated using a UV curable perfluorinated acrylate material. The refractive index of the core and cladding materials after curing was

74 and 1.45, respectively. The channel waveguide cores, 7 um wide and 6.5 um high, were formed by reactive ion etching. The waveguides exhibited single mode behavior with a measured propagation loss of 0.64 db/cm at the wavelength of 1.55 um. These polymer channel waveguides were used to form passive delay lines and 2x2 digital optical switch (DOS) structures. Figure 7.1 shows a schematic of the 2-bit reconfigurable device including the 2x2 DOS. Light enters one input port of the first switch and is adiabatically split by a Y- branch. The large thermo-optic effect ( /K) of the polymer material is employed to lower the refractive index of one of the arms of the Y-branch, causing the light to propagate in the opposite arm. The 2x2 switch operates in the bar state when current is applied to the inner four heaters. By contrast, the switches operate in the cross state when the outer four heaters are activated. 80,81 Depending on the state of the first 2x2 switch, light is guided through one of two polymer waveguide delay lines. A second 2x2 optical switch follows these polymer waveguides. A fiber array is then coupled to the opposite side of the chip that holds the 2x2 switches and the waveguide delay lines in order to complete the delay device. In this device the first two delays are generated by the polymer waveguides and the second two delays are generated in the output fibers of the fiber array. The length difference of these output fibers is designed to yield a 40 ps delay while the difference in polymer waveguide lengths is intended to provide 160 ps of delay. The bend radii of the inner and outer waveguide delay lines are 3 and 3.75 mm, and total propagation lengths are and mm, respectively. 61

75 2-Bit Delay Device DOS Fiber Delay Lines Waveguide Delay Lines DOS Fiber Figure 7.1: Schematic of the 2-bit reconfigurable delay device. 7.3 MEASUREMENT RESULTS The switching speed of the DOSs were measured. A 25 Hz electrical signal with a 50% duty cycle was applied to the switches and the optical output fall and rise time responses were measured. Both the bar and cross state switching speeds were measured to be less than 4 ms. Figure 7.2 shows the switching speed measurement results. The insertion loss of each DOS was 2.8 db with less than 0.2 db difference between cross and bar switching states. The extinction ratio of both switches was greater than 40 db with 360 mw of applied power. 62

76 Figure 7.2: Switching speed measurement results of the 2x2 DOS switch. The total insertion losses, fiber-in to fiber-out, corresponding to each of the four delay combinations, are shown in Table 7.1 along with the switching state of each DOS. The insertion losses varied between 8.12 and 9.81 db depending on the delay path chosen. The device s polarization dependant loss (PDL) was also measured. The device insertion loss with TM polarized light was 0.04dB greater than when using TE polarized light. Figure 7.3 shows near field images of TE and TM modes emanating from the device. The mode patterns and intensities are nearly identical. Table 7.1: Delays, Losses and Switch States for each Delay Path Delay Path Switching State #1 Switching State #2 Coupling Loss(dB)* Insertion Loss(dB) Measured Delay(ps) Designed Delay(ps) 1 Cross Bar Cross Cross Bar Cross Bar Bar * Coupling losses were calculated from the mode mismatch between the delay lines, the DOSs and the fiber array. 63

77 Figure 7.3: Near field images of TE (left) and TM (right) polarized emanating from the 2-bit device. Exact time delay values generated by the 2-bit device were measured. A CW laser, operating at 1.55 um, was modulated by a LiNbO3 modulator fed by an HP8510C network analyzer. The output of the modulator was input to the delay device. The output fibers of the fiber array were connected to a 2x1 coupler and an erbium doped fiber amplifier (EDFA) was used to amplify the signal. A photodetector (PD) covering the X- band frequency range was used to convert the modulated optical signal to an electrical signal that was fed back to the network analyzer. Figure 7.4 shows the measured microwave phase versus the frequency sampled from 2-14 GHz. The shortest delay path, delay path 1, served as the reference. The time delays for each delay path were calculated from the slope of the fitted lines. Table 7.1 lists the delays associated with each delay path. The measured time delays were 0, 37.8, 160.4, and ps. The experimental uncertainty in the delay measurement is ±1 ps due to jitter in the network analyzer s measurement of the phase. Figure 7.5 shows the time response of a femtosecond laser pulse propagating through the device. The four delays are clearly seen as shifts in the peak s position. 64

78 Phase Shift (Degrees) Frequency (GHz) 199.2ps Pulse Response (mv) ps 160.4ps Delay (ps) Figure 7.4: (Top) Phase versus RF frequency measurement results for each of the four delay states. Figure 7.5: (Bottom) Time response of a femtosecond laser pulse through the device. 7.4 SUMMARY The 1.69 db variation in insertion loss, depending on the delay path chosen, is smaller than the 2.11 db that would be expected from the measured propagation loss and length differences of the polymer waveguides. The insertion loss measurements of the 65

79 DOSs show that the switching state has less than 0.2 db effect on insertion loss. The waveguide delay line is designed such that the bend radiation loss of the inner waveguide delay line is large enough to partially compensate for the added loss of the longer waveguide delay line. This is a desirable effect for maintaining a constant radiated power from the PAA, regardless of the steering angle chosen by the delay path. The demonstrated 2-bit polymer waveguide delay unit provides up to ps of optical true time delay. By using polymer thermo-optic DOS switches, the steering angle response time of the PAA is less than 4 ms which is acceptable for the Navy s 10 ms requirements of PAA s onboard submarines. 82 The use of passive polymer channel waveguides and thermo-optic polymer switches ensures that the polarization dependant loss is minimal. While the stable sub-10 db insertion loss of the polymer delay device is currently larger than what is required for actual PAA systems, the fundamental loss of the 2-bit device is less than 3 db with small incremental increases in the loss as the device scales to higher numbers of bits. These losses could be compensated with the potential integration of a polymer waveguide amplifier. 83 Three factors contribute to the majority of the loss in this device; propagation, bend, and switch loss. New highly fluorinated polymers have been reported with losses in the C-band of less than 0.1 db/cm 84 and in fact the theoretical propagation loss limit of polymer has been calculated to be less than that of glass optical fiber in this wavelength range. 85 The bend loss can be managed by increasing the bend radius at the expense of the device size or alternatively introducing bend loss reduction methods as will be reported in the next chapter. The long propagation length of the DOS is responsible for all but approximately 0.5dB of the switch loss. By using low loss polymers, the switch loss could be reduced to the excess loss introduced by the Y-junctions. 66

80 Chapter 8: Bend Loss Reduction Structures 8.1 LOW INDEX CONTRAST WAVEGUIDE SYSTEMS Low index contrast waveguide material systems, such as polymers, provide low fiber to waveguide coupling losses and low scattering losses due to the small difference of refractive index between the core and cladding. However, a major drawback of any low index contrast system for a PLC application is the large footprint attributable to the requirement of large waveguide bend radius for low loss. 86 Polymers have the advantage of a widely tunable refractive index 87 but this still leaves a compromise between coupling loss and bend loss. The large bend radii requirement for low index contrast systems is not practical for devices requiring large scale integration such as optical delay lines 88 and is detrimental to the device yield. Air trenches have been proposed to reduce the radii by increasing the index contrast of a curved waveguide segment By increasing the index contrast within the waveguide bend region, the mode is tightly confined in order to prevent bend radiation losses without significantly affecting the propagation losses. Another modification to the waveguide design used to decrease bending losses is the offset. Waveguide offsets shift straight waveguide segments laterally with respect to curved waveguides to decrease the mode mismatch. 93,94 This in turn minimizes the junction loss between the curved and linear waveguide segments. In this chapter an experimental comparison is reported between fabricated polymer waveguides and simulation results to determine the practicality of trench and offset structures for polymer PLC s. A general explanation of the waveguide trench and offset is presented in Section 8.2. Section 8.3 describes the design of the trench and 67

81 offset structures for fabrication, and Section 8.4 presents a comparison between experimental and simulated results followed by the concluding remarks. 8.2 TRENCH AND OFFSET STRUCTURES Trenches Although an asymmetric index distribution is an effective means for reducing bend loss, 95 it is difficult to fabricate a cladding with multiple refractive index values in only a curved section. A curved waveguide with an air trench structure is a relatively simple way to confine the mode through the bend. Yamauchi 96 describes the configuration parameters of a trench and waveguide bend and explains the expected effects with the use of an equivalent index transformation. Figure 8.1 illustrates a top view of a step index channel waveguide junction between straight and curved waveguide segments. The structure utilizes both an air trench, (n Tr =1), and an offset. The core and cladding indices are n Co and n Cl, respectively, and the waveguide has a bend radius of R. The width of the trench, w Tr, and the separation between the inside radius of the trench and the outside radius of the waveguide core, d, are also labeled. By placing the air trench sufficiently close to the waveguide core, (reducing d) the evanescence tail is reduced and a decreased bend loss is expected. 68

82 Figure 8.1: A diagram of the waveguide core using offset and trench (hatched area) structures Offsets In a straight channel waveguide (R= ) the electric field intensity pattern of the fundamental mode is symmetric about the center of the waveguide core, Figure 8.2a. In contrast, the fundamental mode in a curved waveguide, with the same core cross section, has the field peak displaced laterally toward the outside of the bend and the mode is asymmetric with a width different from that of the straight waveguide, Figure 8.2b. Because the mode fields are mismatched, transition losses will occur at the junction of two waveguides with different radii causing a reduction in the power transfer between the two waveguides. A lateral offset at the junction of the two waveguides may be used to reduce the mode mismatch. To a first approximation, it can be concluded that the loss 69

83 between waveguide segments of different radii will be minimized when they are offset to make their field peaks coincide. 97 Subramaniam 97 compared experimentally measured loss values of offset rib waveguides to simulation results. However, to date there has not been a comparison between simulations and fabricated waveguides utilizing trenches with offsets. The following section provides an overview of the design of these trench and offset structures, followed by a comparison of the simulation results and fabricated structures. a) b) Figure 8.2: Mode profile of a straight 6x6 µm channel waveguide (a) and one with a radius of R=3 mm (b). The outline of the waveguide core is superimposed on the mode intensity contours. n Co =1.46, n Cl =1.45, λ=1.55 µm. 70

84 8.3 DESIGN OF TRENCH AND OFFSETS FOR FABRICATION Trenches In order to properly design the waveguide, trench, and offset structures to work with our polymer material system, we simulated the performance characteristics using a 3-dimensional, semi-vectorial beam propagation method (BPM). All simulations were based on core and cladding materials with refractive indices of 1.46 and 1.45 respectively, and that the core was 6 x 6 um square in order to maintain single mode behavior at the wavelength of 1.55 µm. The first group of parameters determined were the trench width (w Tr ) and the separation distance (d) between the air trench and waveguide core. A trench width of at least times larger than the core width is needed to prevent the evanescent tail from spanning the trench.[9] A wider trench also reduces the height to width aspect ratio to maintain vertical sidewalls during the etching process. However, an excessively wide trench may interfere with compact placement of other PLC structures. A trench width of 20 µm was chosen to maintain an aspect ratio of one. Figure 8.3 shows the simulated bend loss as a function of the separation distance between the waveguide core and the air trench for 180º bending radii of 1, 1.5, 2, and 3 mm. Several features are notable in these results. First, oscillations of the bend loss that occur as the trench separation is increased. Consistent with the work of Yamauchi, 96 this is an expected result caused by the curved waveguide guided mode coupling to a quasiwaveguide mode in the area between the core and trench, resulting in a peak in the bend loss. As the trench separation increases, higher order quasi-waveguide modes are formed which repeatedly match the propagation constant of the curved waveguide guided mode, resulting in oscillations of the bend loss. A second notable feature of Figure 8.3 is that as the bend radius increases the oscillations become smaller and first occur at larger trench 71

85 separations. This is due to the index contrast of the waveguides adequately confining the propagating mode at larger bend radii. To suppress these severe oscillations of the bend loss, the trench separation should be kept small, especially for smaller bend radii. However, there was a concern that having the low index trench in close proximity to the core would cause high polarization dependant loss (PDL) through the bend. Typically, structures with high cross-sectional asymmetry along the principal polarization directions exhibit polarization sensitivity. 92 With these considerations in mind, a trench separation of 7 µm for all radii was chosen with the understanding that radii of 1 mm and less would exhibit 5 db or more bend loss with this separation distance. Figure 8.3: Simulated 180º bend loss as a function of the trench separation of 4 waveguide bend radii. The waveguide core is 6x6 µm, n Co =1.46, n Cl =1.45, n Tr =1, λ=1.55 µm, TE polarization. 72

86 8.3.2 Offsets 2-dimensional mode overlap integrals were performed to determine the optimum offset distance and junction loss between straight and curved waveguide sections for each waveguide bend radius. Fig 8.4 shows the simulated optimal offset values for bends with and without trenches. The extreme variability of the simulated offset values at small bend radii is attributed to the straight input/output waveguides coupling to the waveguide radiation mode of the bend Simulated Optimal Offset Values No Trench With Trenches Offset (um Bend Radius (mm) Figure 8.4: Simulated optimal offset values for 6x6 µm waveguides with and without trench structures. d=7 µm, n Co =1.46, n Cl =1.45, n Tr =1, w Tr =20um, λ=1.55 µm, TE polarization Bend Loss Simulations were performed to evaluate the bend and junction loss versus bend radius of four cases: (1) standard bends employing neither offsets nor trenches, (2) 73

87 waveguide junctions employing optimal offsets without trenches, (3) waveguides with trench structures, and (4) both optimal offsets and trenches. The 180º bends were simulated with a BPM equivalent index transformation in order to avoid paraxial effects. This method has been shown to be accurate for bends with radii much greater than the core width dimension. 93 Figure 8.5 shows the results of these simulations. For case (1), there is an exponential increase in the bend loss as the radius decreases, as would be expected for any dielectric waveguide. 86 For case (2), the curve shifts down because the offsets decrease the loss of the two straight to curved waveguide junctions. The simulated reduction in loss for this case, in comparison with case (1), increases as the bend radius decreases. In case (3), there is a notable improvement in the loss for bend radii less than 3 mm. Between approximately 1.5 and 3 mm the loss remains flat. However, as expected from Figure 8.3, the loss rises sharply with smaller bend radii because of the relatively large trench separation. Finally, due to the use of offsets, case (4) shows an improvement over case (3) for bend radii ranging from 2.5 to 1.5 mm. A 90% transmission value (0.45 db insertion loss) for a 180º waveguide bend can be used to quantify the performance of the offset and trench features. The 90% transmission value occurs at bend radii of approximately 3.3, 2.9, 3.1, and 1.4 mm for cases 1, 2, 3, and 4, respectively. 74

88 Loss (db Simulated bend and junction loss vs. Radius (1)=No Offset, No Trench (2)=Offsets, No Trench (3)=No Offsets, Trenches (4)=Offsets & Trenches Bend Radius (mm) Figure 8.5: The BPM simulated bend and junction loss for a 180º bend as a function of waveguide bend radius for cases (1)-(4). The waveguide dimensions are 6x6 µm and the trench separation is optimized at 7 µm with an operating wavelength of 1.55 µm. 8.4 EXPERIMENTAL PROCEDURE AND RESULTS A UV curable perfluorinated acrylate material was used to fabricate low loss polymer waveguides. A viscous perfluorinated acrylate cladding layer was spin-coated onto a silicon wafer substrate with the layer thickness controlled by the spin speed. After spinning, the polymer was cured by exposure to an intense UV light source for several minutes. A subsequent core layer material was spun and cured in a similar fashion. The refractive index, measured with a Metricon prism coupler, of the core and cladding materials after curing was and 1.450, respectively. A SiO 2 hard mask was deposited with low temperature plasma enhanced chemical vapor deposition (PECVD) and was patterned by AZ5209 photoresist and an i-line contact lithography tool. Reactive 75

89 ion etching (RIE) was used to etch the core material to the bottom cladding layer, after which the hard mask was removed and a top cladding material was spun and cured. A second hard mask layer was deposited and patterned in a similar fashion to form the trenches. RIE was used to etch the cladding material in the trenches to the silicon substrate. The channel waveguide cores were 6 µm wide by 6 µm high and the trench depth was measured by an alpha-step profilometer to be 21 µm deep. Figure 8.6 is an optical microscope photograph of a fabricated waveguide core with offset and trench structures. The trench corners are rounded to prevent the effects of stress in the polymer. Figure 8.7 shows a cross-sectional SEM image of the trench structure. The location of the waveguide core is outlined by the square on the inside radius of the trench. Figure 8.6: Top view of a waveguide junction employing both a trench and an offset. 76

90 Figure 8.7: SEM cross-sectional image of the trench structure. Approximate location of the waveguide core is indicated by the square to the left of the trench. Straight single mode waveguides were measured to have a propagation loss of 0.45 db/cm at the wavelength of 1.55 µm. The coupling loss associated with mode size mismatch and reflection loss between the polymer waveguides and a single mode optical fiber was measured to be 0.7 ± 0.1 db per facet. This is consistent with the coupling loss value of 0.75 db calculated from a 2-D mode overlap integral simulation and reflection loss calculation. 180º waveguide arcs with two millimeter straight waveguide segments at each end were patterned for measurement of the waveguide bending loss. The 2 mm straight waveguide segments were intended to provide ample length for mode transformation to and from the optical fiber as well as a way to test the performance of the offset structures. A fiber array was used to couple 1.55 µm light into and out of the waveguide bends. Each of the fibers of the fiber array were tested and found to exhibit less than 0.1 db loss. Insertion loss measurements of the waveguide arcs were performed to evaluate the effectiveness of the fabricated trench and offset structures. The measurements were performed on arcs with bend radii ranging from 500 µm to 5 mm in 500 µm increments. 77

91 Figure 8.8 shows the insertion loss measurement results of cases (1) through (4). Each data point is the average of four measured samples. The lines in Fig 8.8 represent the calculated insertion loss values based on the simulated bend and junction losses for cases (1)-(4). The calculations were made by adding the waveguide propagation loss (0.45 db cm -1 ) and the coupling losses of the input and output single mode fiber array (2 x 0.70 db) to the simulated bend and junction loss. The simulated and measured values suggest that for bend radii greater than 3 mm the use of trenches and offsets have an insignificant effect on the insertion loss. However, for bend radii of less than 3 mm, trenches improve the bend loss performance significantly. The use of trenches, or the combination of both trenches and offsets, can result in smaller radius bends having less insertion loss than bends with greater radii because of smaller propagation losses. For the polymer waveguides tested, the bend radius can be reduced to approximately 1.5 mm without significant increases in the loss, a 50% bend radius reduction over the standard waveguide bend. The use of offsets with trenches also reduces the bending loss further although not as strongly as suggested by simulations. 78

92 Loss (db Comparison of measured and simulated insertion losses (1)=No Offset, No Trench (2)=Offsets, No Trench (3)=No Offsets, Trenches (4)=Offsets & Trenches Measured (1) Measured (2) Measured (3) Measured (4) Bend Radius (mm) Figure 8.8: Simulated and measured insertion losses of 180º waveguide bends as a function of the bend radius. Data points are measured values and lines are fitted curves to the simulation data points. The PDL (TM-TE) was also measured for each bend radius of the waveguides employing both trenches and offsets. Measurement results are shown in Figure 8.9. For bend radii greater than 1 mm the PDL was positive and less than 0.1 db. For the 0.5 and 1.0 mm bends the PDL was and db, respectively, with TM polarization exhibiting slightly more loss than TE. The trend for the magnitude of the PDL to increase as the bend radius decreases is caused by an increased interaction of the TM mode with the roughness of the trench side wall. 79

93 PDL (db Polarization Dependant Loss (TM-TE) Bend Radius (mm) Figure 8.9: Measured PDL for waveguide bends employing trenches and offsets. Figure 8.10 separates the four cases of simulated and measured data to evaluate the trends more clearly. The error bars of Figure 8.10 represent ±1 standard deviation in the measured data. The measured insertion loss values reasonably match the simulated values. However, the measured values have a greater standard deviation at small bend radii. This could be attributed to random RIE induced defects on the sidewalls of the cores and trenches in some samples, which if present cause scattering of light in the bends. This phenomenon would be more prevalent in bends of lesser radii because of higher field intensities near the trench and core sidewalls. It is expected that this effect would become more severe as the trench separation is decreased. 80

94 25 20 Standard (1) Measured (1) Offsets (2) Measured (2) Loss (db Loss (db Bend Radius (mm) Bend Radius (mm) Trenches (3) Measured (3) Offsets & Trenches (4) Measured (4) Loss (db Loss (db Bend Radius (mm) Bend Radius (mm) Figure 8.10: Comparison between measured and simulated insertion loss values for 180º waveguide bends of cases (1)-(4). 8.5 SUMMARY Polymer channel waveguide samples were fabricated and characterized. Insertion loss measurements of 180º bends revealed the dependence of the bending loss on the bend radius, as predicted with BPM simulations. Simulations and measurements show that bend loss could be slightly reduced with offsets. With bend radii less than 3 mm, air trenches were more effective at reducing the bending loss. However, as the radius is decreased below 1.5 mm the bend loss increases quickly even if air trenches are used. 81

95 It may be possible to maintain low loss and decrease the bend radii further by decreasing the separation between the trench and waveguide core. However, experimental data values show that there is a large variation in the measured insertion loss at small bend radii. This is attributed to random defects induced by the core and trench etching processes which have more of an effect as the bend radius decreases. Additionally, decreasing the trench separation further is expected to increase the PDL of the waveguide bend. By choosing the appropriate trench and offset design parameters, the bend radius of low index contrast polymer waveguides can be reduced by 50% or more. This bend radius reduction will improve the design and fabrication of compact, high yield PLCs. 82

96 Chapter 9: An Integrated 4-bit Delay Device 9.1 TTD DEVICE DESIGN Delay values The antenna system criteria need to be specified in order to select the correct design parameters for a delay device; namely, the maximum required steering angle, the antenna element spacing, the number of antenna elements, and the element configuration, i.e. a 1-D or 2-D pattern. Once these criteria have been specified, Equation (6.1) for a 2D array can be used to determine the maximum time delay needed for a specified system and the time delay values required for each element to achieve a desired steering angle. t y+ x d = ( n 1) y sinθ0 sin Φ 0 d x sinθ0 cosφ + ( m 1) c c 0 (6.1) For a 1D array the 2D equation simplifies to: d sinθ t = ( n 1) c (9.1) Equation (6.1) can be solved as a transcendental equation, allowing θ 0 and Φ 0 to vary within the desired steering space and determining a time delay value for each of the (m,n) array elements. Because only integer values for t y+x are possible ( t, 2 t, 3 t, etc), a finite number of θ 0 and Φ 0 exist. Figure 9.1 is a schematic of a rectangular array of patch antenna elements with each of the variables in Equation (6.1). For the purpose of a 4-bit TTD device design, a 4x4 element array (n = m = 4) can be used with an element spacing of cm in both 83

97 the x and y directions. A steering space of 0 θ0 45º and 0 Φ0 < 360º is also used. By substituting these values in Equation (6.1), all possible steering angles and the corresponding delays for each element are obtained. Table 9.1 lists the solutions. The smallest non-zero delay magnitude is 11.8 ps so this value is chosen as t. Table 9.2 shows the delays for each element as positive integer multiples of t. θ y x m=4 Φ d y n=4 d x Figure 9.1: An array of antenna elements showing the corresponding variables important to design of a TTD device. 84

98 Table 9.1: Computed Steering Angles and Respective Delays Values for Each Element 85

99 Table 9.2: Computed Steering Angles and Normalized Delay States for Each Element Delay line lengths From Figure 6.2 it is seen that the longest delay line will have a length of 2 (b-1) l where b is the number of bits of delay that the device can generate. For a 4-bit device (b=4) the delay lines will have lengths of 8 l, 4 l, 2 l, and l corresponding to delays of 8 t, 4 t, 2 t, and t, respectively. BPM simulations show that the group index (n g ) is for a 7x7 um waveguide similar to the low loss waveguide developed in Chapter 3. Using Equation (6.2), t = L n c g the lengths of each of the device delay lines can be calculated. Table 9.3 lists the results. 86

100 Table 9.3: Calculated Delay Line Lengths Delay line Time Delay (ps) Delay line length(um) t t t t Bend radii The next step in the design is the selection of an appropriate bend radius. The results of Chapter 8 show that by employing waveguide offsets and trench structures the bend radius and the overall bend insertion loss can be decreased simultaneously. Table 9.4 is a summary of the simulated loss results of waveguide bends with trenches from Chapter 8. The 2x2 TIR optical switches presented in Chapter 5 have a pitch of 250 um. With all reference lines using the same bend radius, the waveguide reference lines use a bend radius that is 250 um smaller than the waveguide delay lines. By choosing a bend radius of 1.5 mm for the reference lines and 1.75 mm for each of the waveguide delay lines, the device size can be minimized without incurring significant increases in bend loss. 87

101 Table 9.4: Simulated losses of waveguide bends Bend radius (mm) Loss without offset (db) Loss with offsets (db) Optimal offsets (µm) Calculated Insertion Loss with Optimal Offsets* Offset 1 Offset Table 9.5 lists the design parameters of the TTD device and Figure 9.2 shows a layout of the TTD device with the parameters labeled. Using these values, the total device area to contain these structures is estimated to be 2.1 cm long by 1.5 cm wide. Table 9.5: Design Parameters Parameter Value (um) R R L1 500 Lin Lout 6484 Lswitch 3016 t 809 2t t t

102 Figure 9.2: TTD device layout with design parameters. 9.2 DEVICE LOSS CALCULATION In order to calculate the device insertion loss, accurate values are required for each of the variables of Equation (6.3). The coupling loss from the optimized channel waveguide of Chapter 3 to a single mode glass optical was calculated to be 0.46 db. Additionally, the Fresnel reflection loss, between the 1.46 index of the waveguide core and the 1.0 air index, was calculated to be 0.15 db. In Chapter 5 2x2 TIR switches were fabricated with SL values of 2.04 db. The bend loss (BL), according to the simulation results of Table 9.4, is and db/bend for 1.75 and 1.5 mm bend radii, respectively. From the TTD device parameters listed in Table 9.5, the total waveguide 89

103 length of the longest delay state is 8.99 cm. For the shortest delay state the length is 4.34 cm. The TTD device insertion loss can then be calculated as db for the longest delay state and db for the shortest. 9.3 DEVICE FABRICATION PROCESS The waveguide layer, consisting of the core, bottom and top cladding layers is first patterned on a silicon substrate as previously described in Chapter 3. The core mask layer contains input/output waveguides, optical switch tapers and x-junctions, as well as delay and reference lines with offsets at the boundary between straight and curved waveguide sections. Figure 9.3 provides an overview of the steps for the rest of device fabrication process. Once the waveguide layer (a) has been patterned, two layers of metal are deposited by electron beam evaporation (b). A 5 nm layer of chromium acts as an adhesion layer between the polymer and a 250 nm thick gold film serves as the switches electrode structure after patterning. A layer of AZ5209 photoresist is then spun and soft cured at 90ºC for 1 minute on a hotplate. An electrode layer mask is then aligned to the embedded waveguide core layer and the photoresist is exposed with an i-line contact lithography tool (c). The developed photoresist layer then serves as a soft mask for wet etching the gold film with a TFA gold etchant. The photoresist is then removed leaving the patterned gold electrode layer (d). It is important that the chromium remains across the entire surface of the wafer as it is also needed as an adhesion layer between the polymer and a 100nm SiO 2 layer which is deposited by plasma enhanced chemical vapor deposition (PECVD) at 200ºC (e). Another photolithography process is performed so that the trench layer can be aligned and patterned (f). The SiO 2 hard mask layer is RIE etched in a CHF 3 /Ar plasma along with the 5 nm chromium layer in the trench window 90

104 (g). It should be noted that the trench mask layer not only opens the trench windows but also the electrode bonding pads. It is necessary to electroplate these bond pads with at least 3 um of gold in order to provide an adequate surface for the gold wire bonding process performed during device packaging. Once the bond pads have been electroplated, the remaining photoresist is stripped with acetone and the wafer is placed back into the RIE to perform the 20 um deep polymer trench etch (h). After the trench has been formed the remaining SiO 2 is stripped with a BOE etchant and finally the remaining chromium is removed with RIE (i). Figure 9.3: Process flow for building a TTD device. 91

105 9.4 DEVICE PACKAGING After the wafer has been processed in the clean room it is diced with a wafer saw to separate the devices. The wafer saw also has a polishing effect on the waveguide end faces so that additional polishing is not required. A device package was designed to optically connect the input and output waveguides of the delay device and to also provide a versatile electrical interface. Figure 9.4 shows a schematic of the package. Oz Optics optical patch cords for integrated optics packaging were used. A single optical patch cord is composed of a 1 meter long, 9/125 µm Corning SMF-28 fiber with a 900 µm outside diameter (OD) jacketed tubing. The fiber is terminated with an FC/PC connector on one end and a flat polished rectangular glass ferrule on the other end. The dimensions of the glass ferrule are 1.4 x 2.5 x 7.5 mm. A gold plated Kovar ferrule with a 2.5 mm OD is used to support a strain relief boot. The fiber between the kovar ferrule and glass ferrule has a 250 µm OD acrylate coating and a length 12.0 mm. The bare exposed fiber at the kovar ferrule front end face is encapsulated with epoxy. 92

106 Figure 9.4: Schematic of packaged device showing both optical and electrical assemblies. The device chip was first bonded to a glass substrate with epoxy. A Newport Autoaligner station was used to align the input and output patch cords with the device waveguides. UV curable epoxy was then used to bond the input and output optical patch cords to the device chip and the glass substrate. The alignment and bonding of the optical patch cords to the device was critical as was the appropriate UV curable epoxy. After the patch cords were bonded to the device, the optical assembly was then epoxied to a PC board (PCB) to provide the electrical interface. Figure 9.5 shows the dimensions of the PCB. The board contains a 6-pin 90 degree connection header to 93

107 interface with an electrical ribbon cable. Buried copper leads from the connection header connect with gold plated pads on the other side of the board. The electrical connection between the PCB and the device was made by wire bonding between the gold pads on the device and the gold pads on the PCB. A 7700E wire bonding tool was used to perform the ball to wedge wire bonds. Figure 9.6 shows four fully packaged TTD devices complete with both optical and electrical connections. Figure 9.5: Schematic of the design PCB for electrical connection to the TTD device. 94

108 Figure 9.6: Four fully packaged TTD devices. 9.5 DEVICE PERFORMANCE The TTD device was expected to be dependant on the delay state. Although relatively low, the propagation loss of the polymer waveguides in the device is not negligible. Consequently, longer path lengths should induce higher losses. Additionally, as was noted in Chapter 5, the insertion loss of the TIR switch is also dependant on the switch state activated; bar port states having roughly 0.6 db lower loss than the cross port state. So delay states with more switches activated in the bar state (heater turned on) are expected to have lower loss than delay states with fewer switch heaters turned on. Despite these device properties, for a TTD device to be useful the insertion loss and insertion loss variation should be kept as low as possible. To evaluate the performance of the fabricated delay device, the insertion loss of a fully packaged 4-bit delay device was measured for each of the delay states. The insertion loss of the shortest delay state, 0 t, was measured and found to be 14.5 db. The insertion loss of the other 95

109 delay states were then measured individually. Table 9.6 lists the switches that were activated to produce each delay state and the designed delay values. Table 9.6: Switches activated for each delay state Time Delay ( t) Delay Value (ps) Switches Activated 0 0 none shortest path , , ,2,3, , , , , , ,2,4, , ,2,3, ,3,4, ,3,4, , ,5 longest path Figure 9.7 shows the normalized insertion loss values as a function of the delay state. The overall variation in insertion loss is 2.9 db with 3 t having the lowest insertion loss. Delay states where four switches are activated have higher throughputs than those states with just two or no switches activated. Additionally, larger delays also have lower throughputs due to the extra propagation loss. The polarization dependant loss (PDL) was measured as 0.7 db for this device. 96

110 0 Insertion Loss vs Delay Normalized Insertion Loss No Switches 2 Switches 4 Switches Delay State ( t) Figure 9.7: Normalized insertion loss as a function of delay state. The delay values of the packaged TTD device were measured for comparison with the designed values. The phase versus frequency response of each delay state was obtained as described in Chapter 7. Figure 9.8 shows linear responses indicative of the TTD effect. Each line is composed of 225 measurement values. The measured delay values were derived from the slope of the respective measured data set and are listed in the legend of Figure 9.8. Figure 9.9 plots the measured delay values with respect to the designed delay values. There is less than 0.25 ps variation between any of the measured and designed values. 97

111 Phase (deg 4-bit Delay Measurement Frequency (GHz) 0 t=0ps 1 t=11.7 ps 2 t=23.5 ps 3 t=35.4 ps 4 t=47.4 ps 5 t=59.1 ps 6 t=70.9 ps 7 t=82.7 ps 8 t=94.4 ps 9 t=106.2 ps 10 t=118.0 ps 11 t=129.8 ps 12 t=141.8 ps 13 t=153.6 ps 14 t=165.1 ps 15 t=176.9 ps Figure 9.8: Phase vs. frequency response for each delay state produced by a packaged TTD device Measured Delay Value Designed Delay Value 120 Delay (ps) Delay State (delta t) Figure 9.9: A comparison between measured and designed delay values. 98

112 9.6 SUMMARY The design of a integrated 4-bit optical TTD device has been presented. Calculations were used to determine the possible steering angles and the required delay values for a prescribed antenna configuration. Bend radii were chosen and delay line lengths were calculated to determine the delay device dimensions. An insertion loss calculation was presented and the loss was estimated to be between db and db depending on the delay state chosen. The device fabrication process was described and a package design was presented. A fully packaged 4-bit delay device was measured for insertion loss uniformity, PDL, and delay accuracy. The maximum variation of insertion loss was found to be 2.9 db and the PDL was 0.7 db. The measured delay values had less than 0.25 ps deviation from the designed delay values. 99

113 Chapter 10: An X-band Phased Array Antenna Demonstration 10.1 SYSTEM STRUCTURE This chapter reports the demonstration of an optically controlled phased array antenna system using the 4-bit TTD devices discussed in Chapter 9. The TTD devices were designed to provide steering angles within the steering space of 0 θ 0 45º and 0 Φ 0 < 360º for a 4x4 element X-band antenna array. A wavelength multiplexed PAA system structure, described in Chapter 6, was implemented to reduce the number of TTD devices for this 2-dimensional (2D) array. The multiplexed system structure for the 4x4 array is shown in Figure Figure 10.1: The wavelength multiplexed PAA system structure. Four tunable wavelength lasers supply the carrier wave of the optical delay system. The four wavelengths are combined with a 1x4 polarization maintaining combiner to a single optical fiber. The continuous wave optical signal is modulated with a 40 Gb/s optical modulator. The analog microwave signal driving the modulator is provided by an HP 8510C network analyzer. The modulated optical signal is then 100

114 demultiplexed with a four channel wavelength division demultiplexer (DMUX). The wavelengths of the four lasers are tuned so that each occupies one DMUX channel. The laser wavelengths are , , , and nm. After demultiplexing, the light enters the first stage of TTD devices. Each wavelength travels through a single TTD device where the optical signal is delayed by an appropriate amount. The delayed signals are then multiplexed with a four channel wavelength division multiplexer (MUX) with the same channels as the DMUX. An erbium doped fiber amplifier (EDFA) is used to amplify the signal of all four optical wavelengths. A 1x4 optical splitter then splits the optical signal in then equal ratios and a second stage of TTD devices is encountered. In the second TTD stage, all four wavelengths propagate through each of the four TTD devices. A set of four DMUX separates the four wavelengths coming out of each of the four TTD devices so that a total of 16 optical signals, each with its own delay value, can be delivered to a photodetector bank. The photodetectors convert the modulated optical signal into an analog electrical signal. The appropriately delayed microwave signals are then amplified and input to the X-band antenna elements. A horn antenna, located in the far field region, receives the propagating microwave signal and a microwave frequency spectrum analyzer measures the intensity and frequency radiated by the antenna array. To measure the far field pattern, the antenna array is rotated by a computer controlled rotational stage and the received microwave signal intensity is recorded at designated angles. Figure 10.2 shows the optically controlled PAA system and Figure 10.3 shows the receiving antenna horn and spectrum analyzer setup. 101

115 Figure 10.2: Photograph of the optical PAA system. Figure 10.3: Photograph of the receiving antenna and spectrum analyzer. 102

116 10.2 SYSTEM SETUP AND OPTIMIZATION Delay equalization Prior to testing the system performance, the optical and electrical path lengths for each element were equalized to insure that all delays were generated only by the TTD devices. Wherever possible, components with similar optical fiber or electrical cable lengths were used in the system. However, the optical channel path lengths for each MUX/DMUX device were found to vary by several meters. To compensate for this path length difference, the delays of all element feed lines were measured and optical fibers were fabricated with the correct lengths to compensate for these delays. Additionally, tunable electrical delay lines were used to fine tune the feed line lengths so there were no delays TTD device control A computer controlled relay switch network, seen in Figure 10.2, simultaneously controlled the switching state of all optical switches in the system. Figure 10.4 is a schematic of the electrical circuit. A single power supply, with a fixed voltage of 4.25 V, provided the current for all optical switches. The power supply was connected to the common line of the relay switch network. Each relay switch was connected in series with a tunable resistor and one of the optical switches in the TTD devices. The thin film heaters of the TIR optical switches are represented by a pure resistance in Figure These heaters were connected in parallel and the return line was connected to the ground of the power supply. 103

117 Figure 10.4: Schematic of the electrical circuit used to control the optical switches. The appropriate relay switches were programmed to close for a predetermined steering angle, causing a current to flow through the corresponding tunable resistor and thin film heater. This activated the associated optical switches allowing the optical signal to travel through the desired delay path. The tunable resistors compensated for defects in the optical switches introduced by the TTD device fabrication procedure. Figure 10.5 is a photograph of the tunable resistor bank used in the system. 104

118 Figure 10.5: Tunable resistor bank for minimizing the switch crosstalk. Ideally, each thin film heater would have identical dimensions resulting in equal resistances. However, the wet etch process used to pattern the thin film heaters did not yield uniform heater dimensions causing the heater resistances to vary by as much as 50% across all devices. Because the heater resistances were not uniform, their currents were also not uniform. If the optical switches had a greater thin film heater resistance, the current would be less, resulting in an intermediate optical switch state, neither fully bar state nor cross state. This condition would result in large cross talk and high optical signal loss through the TTD device. By using the tunable resistors, the current for each optical switch could be adjusted independently to obtain an optimized optical switch response TTD device difficulties As explained previously, the system was originally designed to provide delays for a 4x4 element array using eight TTD devices. However, two devices were damaged in the optical packaging of the delay devices. With six functional TTD devices, a maximum 105