16 element subarray

8/2/2019 16 Element Subarray

1/6

IEEE TRANSACTIONS ON MICROWAVE THEORY AN D TECHNI'JUES, VOL. 44 , NO . I I , NOVEMBER 1996 2093

A 16-Element Subarray for Hybrid-CircuitTile-Approach Spatial Power CombiningMark A. Gouker, Member, IEEE, John T. Delis le , Member, IEEE, and S e a n M. Duf f y

Abstruct- Three designs for a 4-by-4 subarray are describedfor use in a spatial power-combined transmitter. The subarraysare con structed using a hybrid-circuit, tile-approach architectureand are composed of 16 cavity-backed, proximity-coupled mi-crostrip antennas, each fed b y a 0.5 watt a mp lifier. B,oth linearlyand circularly polarized subarrays have been constructed foroperation over a 10% band centered at 10 GHz. The linearlypolarized subarray showed the following peak performance:EIRP greater than 27 dBW, effective transmitter power greaterthan 5 watts, dc-RF efficiency greater than 20%, aind excellentgraceful degradation performance.

I. INTRODUCTIONHE EFFORTS TO develop moderate-power solid-stateT ransmitters at microwave and millimeter-wave frequen-

cies are increasing due to the rapid progress in transistor andintegrated circuit development in recent years. To achievemoderate output power levels (tens of watts), it is necessaryto combine the output from a number of transistors. There aretwo basic techniques to combine the outputs at these powerlevels: in a circuit structure, i.e., circuit power combining,or in free space, i.e., spatial power combining. In the pastseveral years one form of spatial power combining, quasi-optical power combining, has gained considerable attention[11-[31.Som e of the benefits of the quasi-optical approach, how ever,are common to all types of spatial power combining [4]-[7].The most notable benefits are graceful degradation and therealization of significant output powers from modest size tran-sistors. Unlike quasi-optical con figurations which xre spatially-fed, spatially-combined arrays, the basic configuration in thispaper is a circuit-fed, spatially-combined array. The circuit-fedapproach has a number of advantages. It is easier to obtainuniform amplitude and phase at each of the elements. Theloss in the corporate feed network can be hidden by placingdriver amplifiers in the feed. There is mo re latitude in choosingthe circuit layout, amplifiers, and antennas to meet th e systemrequirements.

In a previous paper, we reported on the architecture fora hybrid-circuit, tile-approach spatial power-combined trans-mitter [7]. This design is different from other sp'atial-powercombining efforts in that it gives priority to the issues ofthermal management, maximizing combining effi ' iency , and

Manuscript received March 15, 1996; revised July 22 , 19'26. Thls workwas supported by the Advanced Research Projects Agency L nder ContractFI 9628-95-C-0002.The authors are with the Lincoln Laboratory, M assachusel s Institute ofTechnology, Lexington, MA 02173 USA.Publisher Item Identifier S 0018-9480(96)07921-5.

MICROSTRIP,ANTENNA 7 M M l C

'R F INPUT(4 (b)

Fig. 1. (a) Top and (b) side views of the two RF-level subarray design.

maximizing graceful degradation. In this paper we discussthe design and measured results for three different subarraysappropriate for this architecture. In a complete transmitter anumber of these subarrays would be tiled into an extendedarray.

11. DESIGNAn illustration of the 4-by-4 element subarray is shownin Fig. 1. The design resembles conventional tile-approach

phased arrays; however, the implementation has been tailoredfor the spatial power combining application. Each elementconsists of a monolithic microwave/millimeter wave integratedcircuit (MMIC) amplifier and a proximity coupled microstripantenna. A corporate feed network distributes the input signalto each of th e MMIC amplifiers, and the output of theamplifiers is combined in free space after being radiated fromthe microstrip antennas.

The subarray is specifically designed to be a hybrid con-struction of printed circuit boards and MMIC amplifiers inte-grated onto a single metal carrier. The hybrid approach offersmany design options not available in a monolithic approach.The MMIC amplifiers can be tested before insertion into thecircuit. This insures that 100% of the amplifiers are RF goodand that their large signal amplitude and phase are similar.Populating a subarray with amplifiers which have similarperformance increases the combining efficiency. The hybridapproach allows the freedom of selecting a lower dielectricconstant circuit board for the antenna layers to increase theradiation efficiency. The hybrid approach also allows multipleRF levels (in this case two) for increased isolation andincreased circuit density.

0(118-9480/96$05.00 0 996 IEEE


2/6

2094 IEEE TRANSACTIONS ON MICROWAVE THEORY AN D TECHNIQUES, VOL. 44, NO . 11, NOVEMBER 1996

The entire subarray is integrated on a metal carrier. TheMMIC amplifiers are attached directly to the subarray carrierin order to provide a low thermal resistance path to the heatsink. Multiple subarrays can be tiled to a transmitter baseplate. The base plate houses the R F and dc distribution to thesubarrays, and the entire backside of the base plate is reservedfor the heat sink. The low thermal resistance path from theMMICs to the heat sink allows the possibility of using higheroutput power transistors than other spatial power-combiningapproaches while still maintaining low junction temperaturesfor reliable operation.

The subarray is constructed with two multilayer printedcircuit boards as shown in Fig. 1 . The lower circuit boardcontains the dc power distribution layer and the RF inputcorporate feed network. The top circuit board contains themicrostrip antennas. The height of the air gap between the twoprinted circuit boards is chosen so that the field configurationof the microstrip lines in the input feed network is notperturbed much from a conventional uncovered microstripline. This facilitates the transition into the MMIC amplifiers.The ground plane of the R F output layer acts as a shield toreduce the feedback from the field radiated by the microstripantennas to the RF input network. An early version of thesubarray, consisting of a single RF level design with inputfeed, amplifiers and antennas on the same layer, oscillatesbecause of this feedback path. The two-level approach shownin Fig. 1 has sufficient isolation to prevent oscillation.In this design, the signal is transitioned from the bottomcircuit board to the top circuit board by placing the MMICon a 1 0 degree ramp. Simp le ribbon bonds are made at theinput and output of the amplifier carrier. The ramp approachdecreases the RF discontinuity but at the price of increas-ing the fabrication complexity. Other transition strategiesinclude a long ribbon bond, a RF via, or a coupling aperture.These transition techniques ease the fabrication by keeping theMM IC horizontal but can introduce a significant discontinuity,especially at higher frequencies.

The input feed network is a standard 1-to-16 corporatedivider. All of the dividers are Wilkinson dividers to improvethe isolation between the elements. Thus, if one of the MMICamplifiers fails, the effect on the other elemen ts in the subarm yis minimized. This added isolation aids the graceful degrada-tion performance of the array. A photograph of the input feedlayer and MMIC amplifiers is shown in Fig. 2.

The MMIC amplifiers used in these arrays are Texas Instru-ments TGA 8031-SCC. They have been attached to individualcopper-tungsten carriers with input and output alumina mi-crostrip boards and some filtering elements in the bias linesto insure stable operation. Placing the MMICs on individualcarriers allowed measurement of the devices before insertingthem into the array. This measurement is necessary to selectMMICs with similar phase and amplitude characteristics andto measure their output power into a 50 ohm load for use inthe calculation of the combining efficiency. The selected biasvoltages are VD= 5.2 V, G= -2.3 V , an d V&-~L 0.6 to-1 .2 V. c,,, can be used to adjust the Szl of the amp lifier.The typical pe rformance of the amplifiers with this bias i s Gain= 32 dB.POV~.. 27 dBm, and PAE = 27%.

Fig. 2. Photograph of the RF input circuit board an d MMIC amplifiers.

111. CAVITY-BACKEDNTENNATh e proximity co upled microstrip antennas used in this work

are constructed on a two substrate-layer printed circuit board[8]. The choice of this type of microstrip antenna is drivenby reserving the backside of the transmitter array for thermalmanagement and providing a low thermal resistance path fromthe MMICs to the cooling fins. This antenna has the feed lineand the patch on the same side of the ground plane.

The subarrays discussed in 171 used proximity coupledantennas of a conventional design on a circuit board withdielectric constant of 2.94 (Duroid 6002). In the current work,cavity-backed, proximity coupled patch antennas are used.A diagram of the design is shown in Fig. 3. The cavity isconstructed by placing a ring of plated-through holes aroundeach patch. An additional ground plane placed on the topsurface of the circuit board finishes the formation of thecavity and helps shield the feed network from the radiatedsignal. Constructing the cavity using plated-through h oles doesnot complicate the fabrication since plated through holes arestandard in printed circuit board fabrication.

The purpose of the cavity is to prevent the guided wavepropagation i n the circuit board. This has two benefits. First,it decreases the mutual coupling by eliminating the portionthat travels through the circuit board. Second, the radiationefficiency of the antenna remains high because the loss toguided waves is removed [ 3 ] , 9], [ lo]. Thus cavity-backedpatches can be constructed on relatively thick, moderately highdielectric constant boards. A higher dielectric constant makesit possible to shrink the size of the distributed elements, suchas power dividers and impedance transformers. The cavity-backed antenna enables the circuit boards to be constructedwith high volume cofired ceramic techniques which havedielectric constants in the 6 to 7 range. These substrates offeradvantages over Teflon based boards such as better control ofcircuit feature sizes, integrated lumped elements, easier wirebonding, and CTE closer to GaAs.

The input impedances of the cavity-backed patch antennaare controlled primarily by the dimensions of the patch andfeed. As long as the cavity walls are at least a substrate height


3/6

GOUKER et ul.: A 1&ELEMENT SUBARRAY FOR HYBRID-CIRCUIT TILE-APPROACH 2095

1 TOP GROUND PLANE

PLATED-THROUIGH HOLE

PATCH ANTENNIA

DIELECTR IC FILLED CAVITYFEED

ANTENNA LAYER-FEED LAYER

-\LOWER GROUND PLANE

(b )Fig. 3 .tenna. (a) Top view. (b ) Side view.Illustration

of the cavity-backed, proximity-coupled microstrip an-

away from the patch, the cavity does not significantly alterthe impedance. The impedance locus of the patch can betuned by altering the length and width of the feed line underthe patch and by altering the ratio of feed substrate heightto total substrate height. A more in-depth discussion of thecavity-backed patch antenna will be given in a lorthcomingpaper.

The antenn as in these subarrays are constructed with Duroid6002 , largely because the design has evolved from earlier work[7]. The feed layer thickness is 0.64 mm, and the total circuitboard height is 1.78 mm. The spacing between antennas is0.8X0. The plated-through holes are 0.64 mm in diameter andare placed on 0.89 mm centers. There are approximately 2000plated-through holes in a 4-by-4 subarray.The self and active impedance for a linearly polarizedantenna with a quarter-wave transformer is shown in Fig. 4.The self impedance of the patch is the input impeljance whenthe antenna is isolated. The active impedance is the ratio ofreflected wave to inpu t wave w hile all the antennas in the arrayare radiating. This is the impedance presented to the M M IC un-der normal operation of the subarray, and thus the impedancethat must be considered for proper design of the array. In thiswork the active impedance was found through m easurement ofa 1-by-4 E-plane array. A Hewlett-Packard 8511 test set wasused with appropriate power splitters, bi-directional couplers,and phase trimmers. The mutual coupling between the linearlypolarized elements in the H-plane is negligible. so th e E-plane l-by-4 array is an accurate measurement of the activeimpedance. The curves in the figure are for one af the centerelements in the array.

Iv. MEASURED ESULTSThree different antenna configurations were investigatedin this work, one linearly polarized (LP) and two circularly

polarized (CP). Both CP designs use a power splitter to feedorthogonal sides of the patch and a X,/4 line length differentialto achieve phase quadrature. On e CP design uses (iilkinson

Sl 10

-25

-50 9 10 11SELF IMPEDANCE FREQUENCY (GHr )----- CTIVE IMPEDANCE

Fig. 4.ity-coupled microstrip antenna.Self and active impedances of theLP cavity-backed, proxim-

divider for better axial ratio performance, and the other usesa reactive T-junction divider and a diamond-set arrangementfor space reduction. The layouts of these elements is shownin Fig. 5 . The L P and the diamond-set CP antenna designs aresuch that the entire subarray fits within a 3.2X0 (4 x 0.8X0)area. Thus these subarrays could be tiled into a larger array.Th e CP design with Wilkinson dividers is too large for tilinginto an extended array, but it was investigated for its higherquality CP radiation.

The equivalent isotropic radiated power (EIRP) and radi-ation patterns were measured for all of the subarrays. Theeffective transmitter power, dc-RF efficiency and combiningefficiency were calculated from the measured EIRP. Theeffective transmitter pow er [111 is defined as th e EI RP dividedby the directivity of the array. Using the directivity instead ofan estimate of the gain of the array has several advantages.First, the directivity of the array is unambiguous. For auniformly excited array, as in this work, the directivity canbe calculated from the physical aperture. Second, the loss inthe array is properly taken into account by using th e directivity.The unwanted phase and amplitude variations among the arrayelements and the radiation efficiency of the patch antennasimpact the effective transmitter power. Finally, the effectivetransmitted power is roughly eq uivalent to the measured outputpower of a circuit combined transmitter.

A photograph of the fully assembled LP subarray is shownin Fig. 6. The EIRP, effective transmitter power, and dc-RFefficiency for the LP subarray are shown in Fig. 7. Theseresults are for the amplifiers operating in saturation. The EIRPpeaks at 27.2 dBW at 9.7 GHz with a corresponding effectivetransmitter power of 5.0 watts and dc-RF efficiency of 21%.The 3 dB drop in EIRP in the center of the band is caused byinsufficient isolation between the radiated field and the inputfeed network. While the leakage is not great enough to causeoscillations, a portion of the radiated field is still picked upby the input feed and causes interference of the input signalat the input of the amplifiers.The combining efficiency versus frequency is shown inFig. 8.The peak combining efficiency is 64% at 9.6 GHz.The combining efficiency is taken as the effective transmitterpower divided by the sum of the available power from eachof the MMICs. The available power was measured with theMMIC on its individual carrier driving a SO ohm load. This


4/6

2096 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 44, NO. 11, NOVEMBER 1996

(a ) (b) (c )Fig. 5 .set with reactive T-junction dividers.Two-by-two illustrations of the antenna layout in the three subarrays designs: (a) LP design, (b) CP with Wilkinson dividers, and (c) CP diamond

Fig.

9 9.5 10 10.5 11FREQUENCY (GHz)

Fig. 7.subarray.EIRP, dc-RF efficiency, and effective transmitter power for the LP

6. Photograph of the fully assembled LP subarray.

is a conservative definition of the combining efficiency thatincludes the major loss mechanisms in the array: the imperfectfeeding of the subarray, the nonoptimum load presented to theamplifiers, the radiation efficiency of the patch antennas, andthe phase and amplitude variation among the array. Typicalreported combining efficiencies only include loss due to phaseand amplitude variations among the array elements. It is feltthe definition used here more accurately reflects the systemperforman ce and provides a better comparison to the efficiencyof circuit combined approaches. Far-field patterns for thesubarray are shown in Fig. 9.The unevenness in the sidelobesis due to phase and amplitude variations in the array.

The m easured EIRP, effective transmitter power and dc-R Fefficiency for the circularly polarized subarray with W ilkinsondividers [Fig. 4(b)] are show n in Fig. 10. The EIRP peaksat 26.9 dBW at 10.1 GHz with a corresponding effectivetransmitter power of 4.3 watts and dc-RF efficiency of 17%.The decrease in the EIRP and the other figures-of-meritcalculated from it is caused by the loss in the power dividerand extra line length in the feed from the amplifier to th eantenna. The measured far-field patterns for the subarray areshown in Fig. 11. The plot shows a vertical and horizontalcut of the dominant right-hand circularly polarized wave. The

bwL! 40a *OzmI00

Z 60

Uw

9 9.5 10 10.5 11FREQUENCY (GHz)

Fig. 8. Combining efficiency, available power, and effective transmitterpower for the LP s u b m a y .

cross polarization cut is the left-hand CP wave. The far-fieldpatterns for the CP subarrays were calculated from near-fieldmeasurements. The axial ratio of the radiated field was lessthan 2 db across a 10% band centered at 10 GHz.

The figures-of-merit for the circularly polarized subarraywith reactive T-junction dividers and diamond-set arrangementare about the same as the other CP subarray. The onlyexception is the axial ratio of the radiated field which hada best value of 5 dB in the 10% band of interest. The poorquality C P was caused primarily from an unoptimized design,


5/6

GOUKER et u l.: A 16-ELEMENT SUBARRAY FOR HYBRID-CIRCUIT TILE-APPROACH 2097

I LP Subarray 27.1 dBWTABLE I

SUMMARY OF PERFOR MANCE FOR THE THREEUBARRA' IS AVERAGEDV E R A 10%BANDWIDTHIT H NOMINALENTER FREQUENCYF 10 GH zEffectiveTx Power dc-RF Efficiency CombiningEfficiency

4. 3 w 17.2% 54 %CP Wilkinson 26.6 dBWCP DiamondSet 26.5 dBW

3. 6 W 15.2 % 4 4 %3.6 W 14.3 % 46 %

__ E-PLANEH-PLANE

-1 0

-90 -70 -50 -30 -10 10 30 50 70 90ANGLE (deg)

Fig. 9. Farfield patterns for the LP subarray.

I-

9.5 10 10.5 11l 'o- 'FREQUENCY (GHz)

EIRP, dc-RF efficiency, and effective transmitter power for the CPig. 10.subarray with Wilkinson dividers.

but it also reflects that an orthogonally fed patch using areactive T-junction requires a more precise design and hasinherently narrower bandwidth than one using a Wilkinsondivider. For this subarray the EIRP is calculated for the totalradiated power not just the power in the desired RHCP wave.Clearly this antenna design would have to be iterated beforeuse in a real system, but the results calculated from thetotal radiated power show the potential performance for anoptimized antenna design.

The graceful degradation performance of tt e subarraysis shown in Fig. 12 . The theoretical curve in the plot isthe calculated decrease in EIRP based on simply removingelements from the array. Thus, it represents the maximumgraceful degradation performance. The measured results wereobtained by turning off random elements in the array. Theagreement between the curves indicates that the input signalis not significantly changed under the failure of devices. It also

-70 -50 -30 -10 10 30 50 70ANGLE (deg)

Fig. 11. Farfield patterns for the CP subarray with W ilkinson dividers.

II I I I I I I I

2 3 4 5 6 7 8NUMBER OF ELEMENTS FAILED

0 1

I I I I I I0 10 20 30 40 50

PERCENT OF ELEM ENTS FAILEDGraceful degradation performance of the LP an d CP subarraysig. 12.

indicates that the change in load impedance presented to th eamplifiers which arises from the change in active impedanceof the antenna is not significant for failure of up to 40% ofthe elements.

The results for all three subarrays are summarized in TableI. The EIRP, effective transmitter power, dc-RF efficiency,and the combining efficiency averaged over a 10%bandwidthare shown in the table. The band for averaging the LP arrayresults is centered at 10.0 GHz while the band for both ofthe CP arrays is centered at 10.2 GHz. It is believed that thisaveraging is a fair representation of the array performan ce andthat centering the optimum p erformance at the desired 10GHzcould be accomplished with further iterations of the design.


6/6

2098 IEEE TRANSACTIONS ON MJCROWAVE THEORY AN D TECHNIQUES, VO L 44, N O 11, NOVEMBER 1996

V. CONCLUSION [9] F. Zavosh and J. T. Aberle, Single and stacked circular microstrip patchantennas backed by a circular cavity, IEEE Trans. Antennas Propagat.,vol. 43, pp . 746-750, July 1995.[lo] S. M. Duffy and M. A. Gouker, Experimental comparison of the radia-tion efficiency for conventional and cavity backed microstrip antennas,

1996.

Three different designs for a circuit-fed, spatially-combinedsubarray have been presented, Two of these designs areThe measured results show EIRPs in excess of 27 dBWappropriate for the subarrays into an extended array. presented at the IEEE AP.S (nt , Symp, , Baltimore, MD, July 21-26,with correspondlng effective tranvmitter powers of 5 wattsand dc-RF efficiencies greater than 20%. The most indicativefigure-of-merit, however, is the combining efficiency whichpeaked at 64% for the LP array For the spatial pow er-combined architecture described in this paper, the combiningefficiency would remain essentially constant no matter howmany subarrays are tiled together. This is fundamentally different from a circuit comb ined approach where the combiningefficiency decreases as the number of amplifiers in the systemincreases.

[ I l l M A Gouk er, Toward ctandard figures of merit for spatial and quasi-optical power-combined arrays, Tranc. Microwave Theorv Tech , vol43, pp 1614-1617, July 1995

Mark A . Gouker (S84 M92) received the B Sdegree in physics from Emory University, Atlanta,GA, and the M S and Ph D degrees from theGeorgia Institute of Technology, Atlanta, in 1983,1985, and 1991, re5pectivelyFrom 1985 to 1987 he was a Research Engi-neer at the Georgia Tech Research Institute work-ing on millimeter wave integrated circuit antennasSince 1991 he h a been with Lincoln Laboratory,Macsachusettc Institute of Technology, Lexington,MA His research interests are i n microwave and

ACKNOWLEDGMENTThe authors would like to thank R. Magliocco, L. Hill, and

S. Robertson for assembly of the arrays. They are gratefulto D. Snider and R Bauer for their continued support andencouragement.millimeter wave Integrated circuit5 and antennas

L11

121[31

REFERENCESJ. W. Mink, Quasi-optical power combining of solid-state millimeter-wave sources, IEEE Trun.5. Micrownve Theory Tech., vol. 34, pp .273-219, Feb. 1986.R. M. Weikle 11, et al. Transistor oscillator and am plifier grids, Proc.IEEE, vol. 80, pp. 1800-1809.R. A. York, Quasi-optical power combining techniques, in Millimeterand Microwave Engineeringjor Communication and Radar, J. C . Wiltse,Ed., Bellingham, WA, SPIE, vol. CRS4, pp. 63-97.J. Lin and T. Itoh, A 4 x 4 spatial power-combing array with stronglycoupled in m ulti-layer structure, in IEEE M 7 7 - S Int. Microwave Symp.Dig. , Atlanta, GA, June 1993, pp. 607-609.J. A. Bent, et al , Spatial power combining for millimeter wave solidstate amplifiers, in IEEE MTT-S lnt. Microwave Symp. Dig., Atlanta,G A, June 1993, pp. 619-622.A. Balasubramaniyan and A. Mortazawi, Two-dimensional MESFET-based spatial powcr combiners, IEEE Microwave Guided Wave Lett.,vol. 3 , pp. 366-368, Oct., 1993.M. A. Gouker, R. G. Beaudette, and J . T. Delisle, A hybrid-circuit tile-approach architecture for high-power spatial power-combined transmit-ters, in IEEE M77- S In t. Microwave Symp. Dig., San Diego, CA, May,1994, pp. 1545-1548.G. Splitt and M . Davidovitz, Guidelines for design of electromagneti-cally coupled microstrip patch antennas on two-layer substrates, IEEETruns. Antennas Propugat., vol. 38, pp. 1136-1 140, July 1990.

John T. Delisle, (S87-M89) received the B.S.degree in electrical engineering from RensselaerPolytechnic Institute and the S.M. in electrical en-gineering from the Massachusetts Institute of Tech-nology, Lexington, M A .He has been with MIT Lincoln Laboratory since1991 where he currently designs microwave andmillimeter wave circuits for EHF satellite groundterminals.

Sean M . Duffy, received the B.S. and M.S. degreesin electrical engineering from the University ofMassachusetts, Am herst, in 1993 and 1995, respec-tively, and is currently working toward the Ph.D.degree there.

Since 1995 he has been with MIT Lincoln Labo-ratory, Massachusetts Institute of Technology, Lex-ington, MA. His current research interests are inthe design and analysis of printed antennas andmillimeter wave multichip modules.

16 element subarray

Documents