Printer Friendly

Planar microwave electro-optic phase shifters.


Most microwave electronic scanning antennas are controlled by ferrite (magnetic) phase shifters. Although they perform well and have low loss, their size and cost limits their practical implementation in cost sensitive applications. There is need for a smaller, lighter weight, lower cost phase shifter. One method toward achieving low cost is to develop low loss, solid-state phase shifters. An alternate method, which is discussed in this paper, is to develop a planar electro-optic phase shifter[1] using a microstrip transmission line.

The considered phase shifter has the form of an electrical delay line, which causes phase shift by controlled variation of the group velocity of the microwave signal. In a ferrite-based delay line, this shift in velocity is caused by a change in permeability of the ferrite. Varying the applied magnetic field to a ferrite will change the permeability, causing corresponding phase shifts. For a phase shifter built using electro-optic materials, electrical delays are caused by a change in permittivity of the materials; the permittivity can be changed by applying a variable electric field. The very high permittivities of electro-optic materials allow them to be smaller than ferrite phase shifters. Figure 1 is a comparison of a 4 GHz ferrite phase shifter element to a 4 GHz electro-optic phase shifter element. It is the much smaller size and the fact that the generation of electric field inherently takes up less space than magnetic field generation that makes the electro-optic phase shifter compatible with planar circuit design.

This investigation uses present-day tools and techniques to determine the viability of electro-optic materials for microwave phase shifter applications. Early (1965) electro-optic waveguide phase shifter development[2] demonstrated phase shift of a microwave signal. However, losses were high (5 dB) and the waveguide structure was both large and expensive. Additional work on that approach has not resulted in a practical device. With recent techniques in preparing barium strontium titanate electro-optic material[3] and an availability of optimizing planar microwave design software, this investigation was initiated to develop practical planar electro-optic phase shifters.

Essential theory on transmission phase shifters, as well as electro-optic materials also are discussed. Practical circuit design is covered, including moding, packaging, effective high voltage biasing and isolation and impedance matching. The construction of an electro-optic phase shifter is presented and the corresponding data are shown.

A first goal is to select an appropriate phase shift circuit configuration that gives good performance. Since the electro-optic effect is E-field dependent, two necessary goals are voltage biasing and isolation. The circuit requirement of high signal transmission demands the additional goals of good impedance matching, low loss metalization and low loss dielectric material. Two other goals traditional for phase shifters are the preference for wideband operation and a high phase shift range. More subtle circuit goals include low phase angle dependence on environmental variables, including temperature, humidity, vibration and time. A final goal is compatibility with existing processes.

Transmission and Reflection Phase Shifters

As shown in Figure 2a, a transmission type phase shifter is a two-port device ideally in which the microwave signal makes a single pass through a variable delay line section. When nonideal conditions caused by assembly or modeling inaccuracies lead to impedance errors at interfaces A and B, nonzero reflection waves result. Two issues are relevant, the wave returned to port one that affects [S.sub.11] and the standing wave set up in the electro-optic material. The standing wave in the electro-optic material is equivalent to an increase in the average time the electromagnetic wave is in the material. This corresponds to the slightly increased effective material length sensed by the electromagnetic wave. The phase delay, then, has a small dependence on the standing wave in the material.

The reflection type phase shifter is shown in Figure 2b. Ideally, with proper impedance matching, the wave travels the length of the electro-optic material twice. Only half the amount of phase delay material is needed compared with the transmission phase shifter. However, small mismatches at interface A cause two errors. One error is the standing wave dependence as it occurs in the transmission phase shifter. The other error is a direct vector error due to the reflection from region 1 back into its path, which vectorially adds to the wave returning from the electro-optic material. This error distorts the phase shifter's voltage dependence in a nonlinear way because of its vector nature. To avoid this distortion, it was concluded that the transmission phase shifter is the preferred topology. A second benefit of the transmission phase shifter is the elimination of the circulator required for the reflection phase shifter.[1]

Some of the operational equations behind the transmission type phase shifter are considered. By calculating the electrical length for a material at two different effective dielectric constants [[Epsilon].sub.rh] and [[Epsilon].sub.rl], at a frequency f and in a material of length L, the phase shift due to a shift in dielectric constant is found to be [Delta] [Phi] = 360 Lf/C ([square root of [[Epsilon].sub.rh] - [square root of [[Epsilon].sub.rl]) ([degrees]) where c = the speed of light The dielectric loss factor for microstrip[4] is [Alpha] d = 27.3L [Mathematical Expression Omitted] By forming the ratio of these two expressions, a performance criterion for microstrip phase shifters in which the dielectric loss dominates in the device [Delta] [Phi]/[Alpha] d [Mathematical Expression Omitted] The units are maximum phase shift per dB of insertion loss with [[Epsilon].sub.rh] and [[Epsilon].sub.rl] taken at their extremes in the electro-optic material.

A microwave concern is to avoid resonance modes along the phase shifter signal path. This is an important concern, since [Ba.sub.x][Sr.sub.1-x][TiO.sub.3], the material considered in this paper, has a high dielectric constant on the order of 200 to 6000 depending predominantly on the composition and to a lesser extent on temperature. Two possible modes are surface mode in the dielectric that can occur when the thickness is on the order of a quarter wavelength, and transverse mode that can occur when the effective width of microstrip is a half wavelength.

Another microwave concern is that current microstrip equations[5] are derived for dielectric constants of <50. Actual circuit measurements have shown them usable but with increased error. The data presented later in this paper use a metalized rod or truncated microstrip in which the effective constant was obtained empirically

Material Properties

Electro-optic materials such as [Ba.sup.x][Sr.sub.1-x][TiO.sub.3], have two main regions of behavior as a function of temperature. These are known as the ferro-electric and para-electric regions with the transition temperature named the Curie point. A solid-to-solid phase transition or martensitic transition distinguishes these two regions from a crystal form without center of symmetry in the ferro-electric region, to a crystal form with a center of symmetry. In electrical terms this means that there is a built-in dipole in the ferro-electric region only. Analogously to ferromagnetic materials, domains form in the ferro-electric material in which dipoles are all locally aligned. These domains can move, grow or shrink when under a biasing electric field. The ferro-electric region has hysteresis effects with electric field. From a practical perspective, domain movement and hysteresis would yield nonpredictable phase shifting behavior with respect to time and bias field, respectively. Therefore, a material operated in the para-electric region is preferred.

When working in the para-electric region, one must consider how close to the Curie point a phase shifter can operate. Several factors need to be considered. The solid-to-solid phase transition that occurs at the Curie point induces local stress inside the material, which leads to microcracks and a slight degradation of the material with each crossing of the Curie point. Also, the exact temperature of the Curie point can shift several degrees Celsius depending on the strength of the applied E-field bias. A good estimate is to operate at least 10 [degrees] C above the Curie point in order to avoid these adverse transition effects. This can be accomplished by keeping the barium fraction to 0.59 or less for operation over 0 [degrees] C or < 0.45 for operation over- 50 [degrees] C.[2]

Material options include either a single crystal or polycrystalline ceramic pressed powder. A single crystal would have a dielectric constant that depends on crystallographic orientation,[6] while a ceramic's dielectric constant is uniform (homogenous). Ceramic materials were used in this investigation.

In order to minimize the effects of humidity on the ceramic material's properties, it is necessary that the starting powder be pressed to a high density. The high density results in a closed pore structure that reduces water vapor diffusion along grain boundaries. Alternatively, a highly humidity dependent material can be used if one is willing to provide encapsulants to block the moisture.

Some standard preparation methods include cold press and fire, hot isostatic press and fire, tape cast and sol gel. Hot pressing yields a high density sample with low humidity dependence. Tape cast material is potentially humidity dependent but lends itself to large batch and large area fabrication. Since the density can be tailored, there is more flexibility in developing materials with lower dielectric constants.

In the para-electric region, the dominant dependence of the dielectric constant on the bias voltage is caused by the deformation of the index ellipsoid, known as the quadratic electro-optic effect.[7] Deviations are due to higher order effects of the interaction between the electric field with various lattice parameters. In the ferro-electric region, a different voltage dependence holds between dielectric constant and applied electric field, thus giving an additional motivation for operating within the para-electric region.

Design Techniques

The requirement for a biasing network is met by using a radial stub at the end of a high impedance quarter wave transmission line, instead of using a rectangular stub at the end of a high impedance quarter wave line. Radial stubs offer broader bandwidth and a localized zero-impedance reference node, and are 20 to 30 percent shorter than rectangular stubs.[8] The center of the transmission line that contains the electro-optic material is connected to the high impedance quarter wave line and radial stub bias network. The electro-optic material is biased between the top transmission line through the material to the ground plane.

The potential applied to the electro-optic material can be as high as 1000 V across a 4 Ghz sample that is 508 [Mu] m thick. This high potential is not compatible with most systems or with sensitive or expensive test equipment. For example, the high potential will destroy the test ports of a vector network analyzer. The requirement to block the high voltage (HV) from the input and output ports of the phase shifter cannot be met with the use of chip capacitors. They are lumped-element devices with good frequency response, but exhibit low voltage breakdown ratings, typically 100V across the package. A new approach at preventing HV from damaging external equipment is to create a ground plane bias gap.[9,10] With this design, a zigzag 76 [Mu]m wide gap is etched in the ground plane before and after the electro-optic circuit region to form a voltage island around the material. This ground plane bias gap presents a passband of 15 percent with an insertion loss of 0.2 dB. The gap is then covered with silicone to raise the voltage breakdown of the gap from that of air, approximately 200 V to 3000 V (with the silicone).

This configuration performs well under normal conditions, but because of the large perimeter and high number of metal points in the zigzag pattern, the probability of a voltage breakdown is increased. The sharp points in the pattern are regions of high electric field concentration, and, therefore, are the locations of voltage breakdown. To increase the reliability of the HV block, another approach that uses quarter wave coupled transmission lines, as shown in Figure 3a, is preferred. Through the use of tightly coupled lines, an HV DC block can be designed to yield 20 percent bandwidth with 0.5 dB of loss with a return loss of 22 dB.[11] Without any coating, this design has a maximum block of 200 V. The HV breakdown point is increased from 200 to 4000 V by coating the coupled lines with silicone, as shown in Figure 3b.[11] The line widths must be adjusted to compensate for this microstrip with overlay dielectric. This allows for a large factor of protection from damaging either the expensive or sensitive test equipment, or the system it is embedded in. This method of DC block is currently used. A module with and without silicone is shown in Figure 4.

The electro-optic materials typically have high permittivities, thus presenting a low characteristic impedance transmission line. The electro-optic materials can present a transmission line impedance of 10 [Omega] or less. A system's characteristic impedance is typically 50 [Omega]. The challenge then becomes how to deal with such a large impedance step. The first attempt to match was a quarter wave impedance transformer, shown in Figure 5a. This design has an SWR of 3.1 with a 10 percent bandwidth. This method has an inherent problem of metal discontinuity between the wide conductor pattern on the low dielectric constant microstrip and the narrow conductor on the high dielectric constant electro-optic microstrip. One of the design rules formed was that there must never be a transmission line conductor discontinuity at the interface of the low dielectric constant and high dielectric constant material. Hence, the line width on the low dielectric constant material must be the same width as the line width on the high dielectric constant material.

Consequently, the next approach to impedance matching was single open circuit shunt stubs (SOCSS), shown in Figure 5b. Once the material's permittivity is found, the impedance is calculated. First-hand calculations are performed to find the approximate matching network, which yields stub lengths and distance from the load. Next, a microwave CAD program is used to optimize the network. Finally, the circuit is fabricated and tested on a VNA. This SOCSS design yields an SWR of 2.33 with 15 percent of bandwidth. This result is better than the quarter wave impedance transformer. Also, a dual open circuit shunt stub matching network, shown in Figure 5c, was designed and tested, resulting in an SWR of 1.93 with 16 percent bandwidth. It has been reported[8,12] that radial stubs yield a better broadband impedance match than rectangular stubs. A single radial open circuit shunt stub, shown in Figure 5d, was designed, built and tested to yield an SWR of 2.2 with a bandwidth of 20 percent. This is broader banded than the SOCSS, but gave a higher SWR. Finally, a dual radial open circuit shunt stub matching, shown in Figure 5e, is implemented. This design gave the best performance to date with an SWR of 1.9 and 20 percent of bandwidth. This impedance matching configuration is successful for materials with dielectric constants ranging from 50 to 6000 by the tests.

Construction and implementation

There are several options to laying out the electro-optic material. One method is to create a rod of electro-optic material, shown in Figure 6a, with the same height as the surrounding dielectric and to cut it to the same width as the microstrip transmission line, with metalization on top and bottom. This is modeled as a truncated microstrip configuration.[13] The relatively high dielectric constant of the electro-optic material contains most of the electro-magnetic wave energy in the rod. The rod structure also reduces the possibility of generating low frequency transverse modes that might absorb energy. The disadvantage of this method is the difficulty in impedance machining and assembling the dielectric rods, especially for higher frequencies or higher permittivities. For example, with a permittivity of 800, at 10 Ghz, the line widths become on the order of 100 [Mu] m. The rod then becomes fragile, and installation into the Duroid(*) based matching and biasing network becomes very difficult. The rod technique is usable up to frequencies of 5 Ghz with dielectric constants up to 1000.

Assembly of the Bulk Rod

Once the rod is machined to proper size, the Duroid microstrip circuit is fabricated, and a brass fixture with SMA coaxial connectors is fabricated. The assembly of the rod into the circuit occurs next. First, the bulk rod is aligned into its proper orientation with the microstrip transmission line using a 50 power stereo microscope: The area where the electro-optic rod will rest must have the Duroid dielectric cut away, leaving 1 mm on each end for later fine trimming. The area removed is wider than the width of the electro-optic rod to allow ease of installation and alignment of the rod. Next, the brass fixture is heated on a hot plate and the circuit is soldered down without the electro-optic rod. The brass block and circuit are cooled and cleaned, following the soldering operation. Under the microscope, the fit of the electro-optic rod in the dielectric hole is adjusted by removing small amounts of Duroid until a snug fit is achieved. Then, a thin layer of silver paste is put on one side of the electro-optic rod. The ground plane (brass block) in the Duroid recess is cleaned and a thin layer of silver paste is laid down on the brass ground plane. The electro-optic rod, silver pasted side down, is placed against the ground plane and aligned with the transmission line. Slight pressure is applied to assure proper seating of the electro-optic rod to the ground plane. If there is any gap between the electro-optic and Duroid substrate, it can be filled using one of three approaches, including filling with dielectric paste made from high permittivity material powder and low loss adhesive; putting pieces of trimmer dielectric in the gaps and applying pressure; and deforming the Duroid slightly to fill the gap by pushing down on the edge where it contacts the electro-optic. With no gaps at the interface of the electro-optic and the Duroid dielectrics, applying a thin layer of silver paste on top of the electro-optic rod and up to and onto the microstrip circuit on the Duroid will bridge the interface. Next, the silver paste is baked for annealing. The brass fixture and circuit are cooled and excess silver paste is removed from the sides of the electro-optic rod.

A second method, desirable for higher frequencies, of configuring the electro-optic material in a transmission line is to create electro-optic microstrip, as shown in Figure 6b. The material is plated with metal and then a microstrip circuit is etched onto the material. The advantages of this method are smaller, more precise line widths and ease of machining and assembly. The lines can be made as small as the available lithographic process, thus allowing the characteristic impedance of the line on the electro-optic material to be raised. The disadvantage of this method is that low frequency sideways substrate modes can occur if the electro-optic substrate is too thick.

Assembly of Bulk Electro-optic Microstrip

Once the electro-optic material is machined to the proper height, width and length, the Duroid microstrip circuit fabricated and brass fixture machined, the assembly process can begin. A transmission line is etched into the metal plated electro-optic material with a photolithographic process. The circuit area where the electro-optic microstrip will rest, must have the Duroid removed, leaving 1 mm on each end for later fine trimming. The brass fixture is heated on a hot plate and the circuit is soldered down without the electro-optic microstrip. Then the circuit is aligned to the SMA connector centers. The brass block and circuit are then cooled and cleaned. The fit of the electro-optic microstrip in the dielectric hole is adjusted under a microscope by removing small amounts of Duroid until a snug fit is achieved. The microstrip and the ground plane (brass block) are cleaned. A thin layer of silver paste is laid down on the brass ground plane. The electro-optic microstrip is placed silver side down against the ground plane and aligned with the Duroid transmission line. Slight pressure will assure proper seating of the electro-optic substrate to ground plane. If there are any gaps between the electro-optic and the dielectric, these can be filled by one of the three discussed methods. Once the electro-optic microstrip is installed down to the ground plane, the top etched circuit can be connected to the circuit on the Duroid dielectric. This can be accomplished with silver paste.

One future possibility is to use in situ thin-film deposition techniques for forming the electro-optic phase shifter. This will make it possible to have smaller dimensions, and thus, higher frequency operation. This technique could make production line assembly more feasible.

Encapsulates serve many packaging functions. They reduce the humidity dependance of the electro-optic material. They prevent vibration and thermal shock from destroying the electro-optic material. If a potting material can be found with a low loss tangent, low permittivity and high dielectric strength, it could be used in place of the silicone over the coupled line HV DC block. Such a material would allow potting the whole circuit in one step, thus saving a processing step of applying the silicone to the coupled lines. Possible materials that meet this requirement include silicone-based or polyester-based compounds. The encapsulate is applied under vacuum to eliminate air bubbles.

Figure 7 is a schematic diagram of the realized electro-optic phase shifter. The main parts are the DC block, impedance matching section and bias pad. These are built on top of the low loss substrate and the metal trace that is formed on top of the electro-optic section. Figure 8 is a photograph of the electro-optic phase shifter. Its features include a broadband matching network and biasing network in a 80 g 6 X 3 X 0.6 cm package operating at 4 Ghz. Figure 9a depicts the transmission and reflection performance data for the 4 Ghz electro-optic phase shifter. Figure 9b depicts the phase shift vs. frequency for potential applied from 0 to 2.4 V/ [Mu] m. The electro-optic phase shifter's performance characteristics at 4 Ghz, including accompanying circuit and SMA connectors, using a 6550 X 420 X 508 [Mu] m size of [Ba.sub.0.45][Sr.sub.0.55][TiO.sub.3] electro-optic material with a permittivity of 800, and low loss Duroid microstrip with a permittivity of 2.2, include a return loss of 15 dB, an insertion loss of 2.5 dB over a 20 percent bandwidth and a phase shift of 35 [degrees] with 1.2 V/ [Mu] m applied at an ambient temperature of 25 [degrees] C.

Remaining issues

The temperature dependance of the permittivity for electro-optic material over wide changes must be considered. One method to compensate for this is to adjust the composition of the material such that the material is far into the para-electric region of operation, and thus, the permittivity is more temperature stable. Additional material work could focus on minimizing temperature dependance of the dielectric constant. The material used thus far had large loss tangents of 0.09. More material research must be conducted to lower this loss. Direct metalization and lithographic methods have not been attempted and should prove to be useful in the future.


The microwave electro-optic phase shifter is potentially superior in size and cost to ferrite phase shifters. However, high losses and low temperature stability of electro-optic phasers need to be resolved. A planar microwave electro-optic phase shifter was designed, built and demonstrated using the material [Ba.sub.0.45][Sr.sub.55][TiO.sub.3]. Ongoing material research in the area of electro-optics promises to yield materials with higher shift in permittivity with applied potential and lower loss tangents. With the use of such materials, the future of planar microwave electro-optic phase shifters and their role in electronic scanning antennas looks promising. (*) Duroid is a trademark of Rogers Corp.


[1.] V.K. Varadan, D.K. Ghodgaonkar, V.V. Varadan, J.F Kelly and P. Gilkerdas, "Ceramic Phase Shifters for Electronically Steerable Antenna Systems," Microwave Journal, Vol. 34, No. 8, August, 1991, pp.102-123. [2.] G. Harrison, L. Lavedan and M. Donaldson, "Microwave Ferroelectric Phase Shifters and Switches," Final Report, Army Contract No. DA 36-039-AMC-02340 (E), March, 1965. [3.] S. Jang, A. Bhalla and D. Dube, "Studying a Ferroelectric at Microwave Frequencies," Ferroelectrics, Vol. 87,1988. [4.] James D. Woermbke, "Soft Substrates Conquer Hard Designs," Microwaves, January, 1982. [5.] James J. Lev, "Synthesize and Analyze Microstrip Lines," Microwave and RF, January, 1985, pp. 111-116. [6.] Shyh Wang, Fundamentals of Semiconductor Theory and Device Physics, Prentice Hall, Englewood Cliffs, NJ, 1989. [7.] E. Fatuzzo and W.J. Merz, Selected Topics in Solid State Physics, VII, Ferroelectricity, John Wiley & Sons, 1967, New York. [8.] V. Sadhir and I. Bahl, "Radial Line Structures for Broadband Microwave Circuit Applications," Microwave Journal, Vol. 34, No. 8, August, 1991, pp. 102-123. [9.] T.E. Koscica, "Wideband Ground-plane DC Block and Bias Feed," IEEE Trans. Microwave Theory Tech., Vol. MTT-38, 1990,pp.805-806. [10.] T.E. Koscica, "High Voltage Microwave DC Block for Microstrip Ground Planes," Elec. Lett., Vol. 26,1990, pp. 1287-1288. [11.] T.E. Koscica, "Microstrip Quarter-wave High Voltage DC Block," IEEE Trans. Microwave Theory and Tech., under review. [12.] H.A. Atwater, "Microstrip Reactive Circuit Elements," IEEE Trans. Microwave Theory Tech., Vol. MTT-31, 1983, pp. 488-491. [13.] C.E. Smith and R. Chang, "Microstrip Transmission Line With Finite-Width Dielectric," IEEE Trans. Microwave Theory Tech., Vol. MTT-28, 1980, pp. 90-94.

Richard W. Babbitt received his BS degree in engineering physics form Lehigh University in 1958. Currently, Babbitt is a senior project engineer in the Microwave and Signal Processing Division of the US Army Electronics Technology and Devices Laboratory. He is a member of the IEEE. He was a recipient of a 1982 and 1989 Army R&D Achievement Award and received a third place award at the 1982 Army Science Conference. In addition, Babbitt has more than 30 patents related to magnetics and microwave device, mm-wave control devices and electronic scanning techniques. Currently, he is coordinating the development of superconducting devices for insertion into military systems and the investigation of ferro-electrics for phase shifters and microwave lens applications.

Thomas E. Koscica received his BS, BA and MS degrees in electrical engineering, physics and solid-state electrical engineering in 1987 and 1989, respectively, from Rutgers University. Currently, he is working towards his PhD in solid-state electrical engineering at Rutgers. He has been at US Army LABCOM for four years and has been involved with designing novel passive microwave circuits, and more recently with electro-optic materials.

William C. Drach received his BS degree in electronics engineering in 1991 from Monmouth College. Currently, he is working towards his Master's degree in electron devices and electro-optics at Monmouth. in July of 1989, he joined the Electronics Technology and Devices Laboratory, US Army LABCOM. His research interests include innovative approaches to test and evaluation equipment. He holds one US patent.
COPYRIGHT 1992 Horizon House Publications, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 1992 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Author:Babbitt, Richard W.; Koscica, Thomas E.; Drach, William C.
Publication:Microwave Journal
Date:Jun 1, 1992
Previous Article:A 10 kHz to 400 MHz synthesized frequency converter.
Next Article:Active-passive bistatic surveillance for long range air defense.

Terms of use | Privacy policy | Copyright © 2021 Farlex, Inc. | Feedback | For webmasters |