Integrated Antennas and Quasi-Optical Device Arrays

Transcription

1 Integrated Antennas and Quasi-Optical Device Arrays (Invited Paper) Robert A. York Department of Electrical and Computer Engineering University of California, Santa Barbara, CA ABSTRACT Integrated antennas have been the focus of increasing research activity, particularly for millimeterwave power combining and quasi-optical arrays. This paper reviews some of the work in integrated antennas for oscillator and amplifiers arrays carried out at UCSB. This includes unusual phenomena involving coupled-oscillator systems that have led to novel beam scanning concepts for cost-effective mm-wave radar and imaging systems. I. INTRODUCTION A number of commercial opportunities are emerging in the wireless communications, imaging, and automotive electronics markets for active, smart antenna array systems. This has spurred interest in integrated antenna structures for compact and cost-effective circuits. In addition, proposed operating frequencies in the millimeter-wave range require creative solutions for combining power from a number of solid-state devices. This has spurred interest in quasi-optical power combining arrays, and integrated antennas are critically important for this technology. The term quasi-optical frequently causes confusion in this context. Ordinarily the term refers to mm-wave and sub-mm-wave systems employing Guassian beam waveguides and associated optical components such as lenses, mirrors, beam-splitters, etc. However, many of the power-combining arrays described as quasi-optical appear in the form of ordinary antenna arrays, without any optical components in sight. In fact, a large number of the published results in quasi-optical power combining have been achieved at low microwave ffequencies where Gaussian beam systems are generally considered impractical. The reason such arrays are described as quasi-optical usually refers to the method used to maintain coherence among the array elements. In the case of a spatial amplifier array, the entire array is illuminated with a coherent beam, whereby each array element amplifies and retransmits a portion of that beam. In the case of an oscillator array, nonlinear interactions between the devices allow them to synchronize (injection-lock) to a common frequency. In both cases there are similarities to well known coherent optical systems like the laser. Naturally the merits of such techniques are likely to have the greatest impact in mm-wave systems; most researchers to date have demonstrated prototypes at lower frequencies for economic reasons. In theory, quasi-optical arrays are limited only by the device technology. A number of research groups across the U.S. have been engaged in research in this area; an overview of the national effort can be found in a recent review paper on the subject [l]. This paper reviews some of the work in coupled-oscillator and amplifier arrays, and associated integrated antennas, that has been carried out at UCSB during the past few years /95/$ IEEE

2 11. AMPLIFIER ARRAYS The most common approach to power-combining at both the device (chip) level and circuit level is shown in figure la; a number of amplifiers circuits are typically paralleled using N-way splitter/combiner networks. These circuits have problems with isolation and/or become too lossy with increasing numbers of devices and increasing frequency. Combining efficiency is particularly a problem in the output combining network. One simple solution is to replace this with an ordinary antenna array (fig. lb), which is an efficient power combiner in the sense that the radiated field is combined in free space with low loss. By our definition, this would not be called a quasi-optical array since coherence is maintained without optical techniques. This approach is most suitable for the microwave and low mm-wave region. For large arrays operating well into the millimeter-wave region, a completely optical system (fig. IC) is preferred. This is sometimes referred to as a spatial feed. In both cases (figs. lb and IC) the combined amplifier and antenna circuit(s) form an integrated antenna if they share a common substrate or are in such close proximity as to require a concurrent analysis/design procedure. amplifier transmission-line /combining network input output (4 amplifier antenna antenna XX antenna array array >I>-( array J output input \ LxxJ xx ourput Figure 2 - (a) Conventional approach to boosting power by paralleling amplifiers, which suffers from excessive losses at high frequencies due to splitter/combiner networks. (b) Increased efficiency using antenna array for combining. (c) Quasi-optical array or spatial combining array, suitable for Gaussian beam system. Integrated antennas used for both (b) and (c). The design of quasi-optical amplifier arrays (fig. IC) is fundamentally a problem in integrated antenna design. Compact and efficient planar radiating structures are required that can be easily integrated with devices on a semiconducting substrate. Other challenges are the limited bandwidth of common 65

3 planar antenna structures, isolating input and output signals to prevent oscillation, and constraints imposed by the optimum cell size for efficiency. Designs based on the patch antenna [2,6,7], integrated horns [4], and short dipoles [5] have all produced bandwidths of less than 3%, and alternative designs are being sought. There are several possibilities for broadband antennas on high dielectric constant substrates [ The bandwidth of a patch antenna can be increased by adding parasitic resonant elements [10,11], which is analogous to the design of broadband filter networks. Travelling-wave or complementary antennas can also be used [8,12], but these typically occupy a large area, and in the case of the bow-tie, have an undesirable radiation pattern. Folded-slot 20-, ,{ : pa<2-w-f-/ '\...., _, Transmission, Device Off ground metal ~ ~ 3.8 " " " 4.0 ' " " " 4.2 " " " " 4.4 ' ~ Frequency, GHz (4 (b) Figure 2 - (a) Polarization-rotating amplifier cell using a resistive feedback FET amplifier and folded-slot antennas. (b) Measured EIPG (power gain x directive gain) for the amplifier cell, in both transmission mode and reflection mode. Amplifier cells based on wideband patch antennas and wideband slots seem promising. The folded slot [3] is attractive because it is easily integrated with three-terminal devices, can be fabricated in a single mask step, and is relatively easy to analyze with commercial software packages. An early 4 GHz prototype amplifier cell using folded slots is shown in figure 2a. This circuit incorporates a simple wideband resistive feedback MESFET amplifier between the orthogonally polarized input and output antennas. The measured effective isotropic power gain (EIF'G) in free space is shown in figure 2b as a function of frequency. The EIPG is written as GampG$,nt, where Gamp is the amplifier power gain and Gant is the gain of a single antenna. No polarizers were used in this measurement, and since the folded slot radiates both above and below the substrate, the measured gain in transmission is -6 db lower than is achievable with polarizers. For use as a reflection amplifier, the circuit was placed in front of a mirror, which significantly increases the directivity and hence EIPG of each antenna. The measured gain in transmission with the device unbiased is also shown. The 3dB bandwidth of this circuit was approximately 12% on Rogers Duroid 6010 (et = 10.8). A prototype 4 x 4 array based on the folded slot cells of figure 2 was fabricated and tested to explore suitable biasing schemes, and ensure that neighboring cells would not strongly couple and lead to parasitic oscillation [3]. A photograph of this array is shown in figure 3. The element spacing is approximately a half-wavelength (in air) in both directions. The measured frequency response for

4 Figure 3 - A hybrid 4 x 4 amplifier array on Duroid 6010 (E, = 10.8) using the folded-slot cell of figure, with 0.25 pm packaged GaAs MESFETs [3]. This array operated at 4 GHz with a 10% 3dB bandwidth. this array was essentially the same as for a single element, but with a larger directivity. The output power was 16 times that of the single element for the same incident power density suggesting high combining efficiency, but the 3dB bandwidth of 9% was slightly less than the single element. This is probably a result of poor etching tolerance in the fabrication process, which led to nonuniform antenna dimensions across the array. When biased for class-a operation, no oscillation was observed, but parasitic oscillation did occur for a small range of gate bias below the class-a point, which indicates either excessive coupling between neighboring elements (which could be occurring through substrate modes), or an improper matching network to a conditionally stable device. Accurate electromagnetic modelling of integrated antenna structures is crucial to successful design. Most available circuit design software does not have any capability to model radiative structures, and even commercial electromagnetic modelling codes often have trouble with such open structures. At UCSB, a numerical code based on the Finite-Difference Time-Domain method has been developed for analyzing planar circuit and antenna structures [14]. This code is based on a direct discretization of Maxwell s equations with no approximations, so all important electromagnetic phenomena (edge effects, substrate mode excitation, losses, radiation) are incorporated. This is crucial for monolithic circuit development. Using the FDTD code, a new antenna structure has been developed which is a variant of the foldedslot design, created by adding additional parasitic slots. Using simple antenna theory [15], it can be shown that the driving point impedance of this antenna scales as Zi, = Zslot/N2, where N is the number of slots, and Zslot is the impedance of a single-slot antenna. This impedance scaling property was verified by simulation and direct measurement (fig. 4a). By choosing the correct number of slots and slot dimensions, a 50w input impedance can be engineered. These antenna structures are also 67

5 inherently wideband; a typical return loss measurement is shown in fig. 4b showing a 25% fractional bandwidth at the conservative -10dB points. This topology has been used for direct integration of commercial 50-ohm MMIC chips to make X-band amplifier array [15]. 5 0 F9 a * -5 m B E U LI a 'i -15 a C H -20 Frequency range : 5 GHz to 15 GHz ( Frequency, GHz (b) Figure 4 - (a) Measured impedance scaling property of cpw-fed multiple slot antennas, for 2,3,4,and 5 slots, on 25 mil Alumina substrate. (b) For a given substrate, the number of slots and slots dimensions can often be chosen to give a relatively broadband 50 w load impedance, suitable for integration with MMIC chips. 68

6 111. COUPLED OSCILLATOR AND SCANNING ARRAYS Oscillator arrays share two common requirements: all the devices must operate at the same frequency, and with a prescribed phase relationship (usually all-in-phase for broadside beamforming). Several approaches to oscillator arrays have been developed, and these have been classified by Itoh [16] as shown in figure 5. Quasi-Optical Oscillator Arrays,, Wave-Beam Type ActiveAyyType Grid Oscillator Fabry-Perot Resonator Mutual External Injection Synchronization Locking Strong Coupling Coupling Line Guided Wave Structure Weak Coupling Radiative coupling Figure 5 - Classification of quasi-optical sources (after Itoh [16]. The approach used at UCSB is the mutually synchronized active array type, first suggested by Stephan [17]. In this technique, the oscillators are coupled together through a coupling network described by common N-port circuit parameters. This coupling allows the oscillators to mutually synchronize through the phenomenon of injection-locking [18] which Stephan termed inter-injectionlocking. Each oscillator feeds an integrated planar antenna so that the power is combined optically. In some cases the oscillator also interact optically [19], either by radiative coupling between antennas,or through an external Fabry-Perot cavity, in which case this would be a true quasi-optical combiner. An enormous variety of active antenna configurations have been developed which are suitable for coupled-oscillator arrays. Designs using both two and three terminal devices have been reported based on the patch antenna [20,21,23], coplanar-waveguide-fed slot antenna [22,25,26], and taperedslot or notch antenna [24]. Two of the more common examples using patch antennas and cpw-fed slots are shown in figure 6. The design of such elements is straightforward, but requires accurate modelling of the antenna impedance and antenna-circuit coupling. The antenna itself is often used as the resonator or frequency-selective feedback element in the oscillator design to save space, although this typically results in a low-q oscillator or motion detector, where the frequency is sensitive to environmental disturbances. A more stable and reproducible design that is well-suited to monolithic implementation has been developed at Cornel1 [22] using a resonant cpw tee to determine the oscillation frequency. The design of high-efficiency patch oscillators has also been explored, with DC-to-RF conversion efficiencies of 44% [23]. Most recently, monolithic quasi-optical slot oscillators using the topology of figure 6b have been demonstrated at 155 GHz and 215 GHz using pseudomorphic HEMTs [26]. 69

7 grounded substrate ground metallization (4 (b) Figure 6 - Two examples of active antenna oscillators for arrays. (a) Feedback oscillator using patch antenna [23] (b) FET oscillator using a slot antenna [25,26]. The earliest quasi-optical coupled-oscillator arrays relied on the inherent (weak) mutual coupling between neighboring antennas in an array for mutual synchronization [19]. These were especially useful as a vehicle for understanding how to establish the correct phase relationship in an array of coupled oscillators. The weak radiative coupling meant that the array elements were not strongly perturbed from the free-running state, which allowed a description of the mutual synchronization using a modified version of Adler's equation [27]. It was found that the phase of the mutual interaction was crucial in determining the stable mode of operation. Stronger inter-element coupling can be accomplished using transmission-line circuits connecting the array elements. This requires suitable modifications to the coupled-oscillator theory so that the stronger oscillator interactions can be properly controlled to insure the desired phase relationships. This has been addressed using the circuit model shown in figure 7, where a collection of N independent sinusoidal oscillator circuits are connected via an arbitrary N-port coupling network, described by Y-parameters.. Figure 7 - General model for a coupled oscillator array. Each individual oscillator is modelled by a very simple RLC negative resistance circuit, which leads to the van der Pol equation. Coupling between the oscillators is described by Y-parameters, which gives a simple set of differential equations describing the array dynamics [28]. Analysis of the phase dynamics can be simplified by assuming nearest-neighbor interaction and identical oscillators. The mutual interaction between adjacent oscillators is described using a complex coupling coefficient 6 exp( -J@), which can be related to the g- or z-parameters of the coupling circuit

8 that connects the oscillators [28]. The phase dynamics of a system of N coupled oscillators can then be derived as [27,28] where w,, and 8% are the frequency and instantaneous phase, respectively, of oscillator i, and Q is the Q-factor of the oscillator embedding circuits. This is obviously just a generalization of Adler's equation (1). When the free-running frequencies, w,, are similar enough, then the oscillators can lock to the same frequency so that do,/dt = w in the steady-state. For beam-scanning a constant progressive phase shift of Aq5 is required, represented mathematically as 8, - O,-l = A4 where i = 2... N. Substituting this condition into (3) and assuming CP = 0" leads to a set of conditions on the free-running frequencies [29] wo/ [l - Aw, sin Aq5] if i = 1 if 1 < i < N w0/ [1+ Aw, sin Ad] if i = N (4) where WO is the desired steady-state synchronized frequency, and Aw, = ~wo/2q is the locking range. This indicates that a constant phase shift can be programmed by slightly detuning only the end elements of the array in opposite directions. This is a relatively simple technique for creating cost-effective mm-wave phased arrays. Although (4) seems to imply that any phase shift A$ can be obtained, a stability analysis of (3) puts limits on this quantity [27]. For the special case of CP = 0", the limits are -90" < A4 < 90". In a typical array with element spacing of d = X0/2, this phase shift is sufficient to scan the beam over a k30" range. In reality the dynamics are complicated by many factors that have been left out of (1), such as amplitude dynamics and non-uniformities, frequency-dependent coupling networks, nonnearest-neighbor interactions, non-uniform tuning profiles of the VCOs, and frequency-dependent device characteristics. I I I patch I coupling line Drain bias vias Figure 8 - Illustration of a linear MESFET oscillator and patch antenna array showing transmission line coupling network [28,30]. An example of a mutually-synchronized scanning array is shown in figure 8, where a number of simple MESFET oscillators with patch antenna loads are coupled via resistively-loaded one-wavelength Ill-I 71

9 transmission lines [30]; the array demonstrated excellent correlation with the theory when the endelements were bias-tuned in opposite directions. Several such arrays have since been constructed. A recent version was built in collaboration with researchers at the Jet Propulsion Laboratory, operating at 8.4 GHz. It used a novel VCO based on a varactor-tuned patch antenna, and produced over 1 Watt with an 8-element array, and had a total scanning range of 45' (-15" to +30" around broadside). Typical radiation patterns are shown below e Measured m U z a.- 9 i I - m d L- m n Angle, degrees Angle, degrees Figure 9 - Results from an 10-element MESFET VCO array using patch antenna. The array operated at 8.4 GHz, produced over 1 Watt of power, and has a total scan range of 45" [31]. One limitation of the injection-locked or coupled-oscillator topologies is the limited range of phase shifts that can be synthesized. This can be solved using the topology shown in figure 10. A frequency doubler circuit is used at each array element which effectively doubles the inter-element phase shift, which extends the theoretical scanning range to full hemispherical coverage [32]. The signals can then be amplified (for additional power and efficiency) and fed to a planar radiating element. V V V V Vmtchantenna Q Quplingq oscillator scan control DR-stabilized oscillator scan control Figure 10 - An improved version of the scanning oscillator concept uses frequency doublers to increase the scan range, and a low-noise source to stabilize the array, as well as power amplifiers to buffer the oscillators and increase output power. This technique has some other potential benefits: the oscillators can be designed at a lower frequency (half the desired output frequency), which is useful because oscillators are sensitive to parasitic reactances in hybrid circuits, and also because oscillator design is simpler when the device has high 72

10 gain, which is more easily achieved at lower frequencies. For 545 scanning in a frequency-doubled array (half-wavelength spacing), the relationship between oscillator tunings and interelement phaseshift is quite linear, which makes calibration and scan control circuit design much simpler. The doublers (and possibly amplifiers) following the oscillators provide a desirable measure of isolation between the antenna and oscillator. Using FET-based doublers, a stable broadband load impedance is presented to the oscillators, which will greatly improve the performance of the array. A version of this array is currently under construction in the millimeter-wave region in collaboration with Hughes Research Laboratories. IV. ACKNOWLEDGEMENTS This work was supported by the US. Army Research Office, the Hughes Research Laboratories in Malibu, CA, and the Jet Propulsion Laboratory in Pasadena, CA. Active devices used in the work were donated by NEC and California Eastern Labs, and substrate material was donated by Rogers Corp. VI. REFERENCES 1) R.A. York, Quasi-Optical Power-Combining Techniques in Millimeter and Microwave Engineering for Communications and Radar, J. Wiltse, ed., vol. CR54, pp , SPIE Press: Bellingham, Washington, ) H.-S. Tsai and R. A. York, Polarization-rotating quasi-optical reflection amplifier cell, Electronics Lett., vol. 29, pp , Nov ) H.-S. Tsai, M.J.W. Rodwell, and R.A. York, Planar amplifier array with improved bandwidth using folded-slots, IEEE Microwave Guided Wave Lett., vol. 4, no. 4, pp , April ) C-Y. Yu and G.M. Rebeiz, A quasi-optical amplifier, IEEE Microwave Guided Wave Lett., vol. 3, pp , June ) N.J. Kolias and R.C. Compton, A microstrip-based unit cell for quasi-optical amplifier arrays, IEEE Microwave Guided Wave Lett., vol. 3, pp , Sept ) T. Mader, J. Shoenberg, L. Harmon, and Z.B. Popovic, Planar MESFET transmission wave amplifier, Electronics Lett., vol. 29, pp , Sept ) J.A. Benet, A.R. Perkons, and S.H. Wong, Spatial power combining for millimeter-wave solid state amplifiers, IEEE MTT-S International Microwave Symp. Digest (Atlanta), pp , June ) D.B. Rutledge, D.P. Neikirk, and D.P. Kasilingam, Integrated Circuit Antennas, in Infrared and Millimeter Waves, vol. 10, chap. 1, pp. 1-90, K.J. Button, Ed., New York: Academic Press, ) G.M. Rebeiz, Millimeter-Wave and TeraHertz Integrated Circuit Antennas, Proc. IEEE, vol. 80, pp , NOV ) D.M. Pozar, Microstrip Antennas, Proc. IEEE, Special Issue on Antennas, vol. 80, pp , Jan ) J.R. James and P.S. Hall, eds., Handbook ofmicrostrip Antennas (two volumes), Peter Peregrinus Ltd., London, ) K.S. Yngvesson et al., The tapered slot antenna - a new integrated element for millimeter-wave applications, IEEE Trans. Microwave Theory Tech., vol. MTT-37, pp , Feb ) D.M. Pozar, Considerations for millimeter-wave printed antennas, IEEE Trans. Antennas Propagat., vol. AP-31, pp , Sept I

11 14) H.-S. Tsai and R.A. York, FDTD analysis of folded-slot and multiple slot antennas on thin substrates, to appear in IEEE Trans. Antennas & Prop., October ) H.-S. Tsai and R.A. York, Multi-slot 50 Q antennas for quasi-optical circuits, IEEE Microwave Guided Wave Lett., vol. 5, no. 6, pp , June ) T. Itoh, presented at 1993 Workshop on Millimeter-Wave Power Generation and Beam Control, (Huntsville, Alabama), sponsored by U.S. Army Research Office, Sept ) K. D. Stephan and W. A. Morgan, Analysis of Tnter-Injection-Locked Oscillators for Integrated Phased Arrays, IEEE Trans. Antennas Propagat., vol. AP-35, pp , July ) K. Kurokawa, Injection-Locking of Solid-state Microwave Oscillators, Proc. IEEE, vol. 61, pp , Oct ) R.A. York and R.C. Compton, Quasi-optical power-combining using mutually synchronized oscillator arrays, IEEE Trans. Microwave Theory Tech., vol. MTT-39, pp , June ) H.J. Thomas, D.L. Fudge, and G. Morris, Gum source integrated with microstrip patch, Microwaves & RF, pp , Feb ) K. Chang, K.A. Hummer, and G.K. Gopalakrishnan, Active radiating element using FET source integrated with microstrip patch antenna, IEEE Trans. Microwave Theory Tech., vol. MTT-31, pp , Sept ) M. Vaughan and R.C. Compton, Resonant-tee cpw oscillator and the application of the design to a monolithic array of MESFETs, Electronics Lett., vol. 29, no. 16, pp , Aug ) R.D. Martinez and R.C. Compton, High efficiency FET/microstrip patch oscillators, (preprint), IEEE Antennas and Propagation Magazine, Dec ) J.A. Navarro, Y-H. Shu, and K. Chang, Broadband electronically tunable planar active radiating elements and spatial power combiners using notch antennas, IEEE TI-ans. Microwave Theory Tech., vol. MTT-40, pp , Feb ) H. P. Moyer and R. A. York, Active cavity-backed slot antenna using MESFETs, IEEE Microwave Guided Wave Lett., vol. 3, pp , April ) B.K. Kormanyos, S.E. Rosenbaum, L.P. Katehi, and G.M. Rebeiz, Monolithic 155GHz and 215 GHz quasi-optical slot oscillators, (preprint), submitted to IEEE MTT-S Int. Microwave Symp. (San Diego), June ) R.A. York, Nonlinear analysis of phase relationships in quasi-optical oscillator arrays, IEEE Trans. Microwave Theory Tech., special issue on quasi-optical techniques, vol. MTT-41, pp , October ) R.A. York, P. Liao, J.J. Lynch, Oscillator array dynamics with broadband N-port coupling networks, IEEE Trans. Microwave Theory Tech., vol. MTT-42, pp , November ) P. Liao and R.A. York, A new phase-shifterless beam-scanning technique using arrays of coupled oscillators, IEEE Trans. Microwave Theory Tech., special issue on quasi-optical techniques, vol. MTT-41, pp , October ) P. Liao and R.A. York, A six-element scanning oscillator array, IEEE Microwave Guided Wave Lett., vol. 4, no. 1, pp , Jan ) P. Liao and R.A. York, A 1 Watt X-band power-combining array using coupled VCOs, IEEE MTT-S International Microwave Symposium Digest (San Diego), pp , June ) A. Alexanian, H.C. Chang and R.A. York, Enhanced scanning range in coupled oscillator arrays utilizing frequency multipliers, 1995 IEEE Antennas and Propagation Society Symposium Digest (Newport Beach, CA). 74