l[ itp itt - nasa
TRANSCRIPT
NASA-CR-16832619840012669
_ 30-GHz MONOLITHIC RECEIVEMODULE
First Annual Report for Period- November 3, 1982-October 31,1983
- Contract No. NAS3-23356
March 1984
L[_ItP_ItT.OOP7, i-,• '01984
LANGLEYRESEARCHCENTERLIBRARY,NASA
itr_[.i,'.}TO[j,_ylRGI_JI_
- ForNational Aeronautical and Space Administration
- Lewis Research CenterCommunication and PropulsionSection
Honeywell
Physical Sciences Center
k .,.'+f",'_"':/+<',_,.._;,'T_ ,','.+'Q;:?_ :k'_-'+":_'+ ._
r ._:!T .+_._. =
................. t P MT ........! i it'¢HViL_i_il i . _ "_"_ ; .'%il-_-'_J! ;i 1.1i1 _J+...." .P._# i ! .+._l_s £ llJ'---I
• +. r
........... HU I UiqH I I t_. ,....__...... _h__._N_.'k..-__.Tt-'H_-b ............
SE',RCHTITLEDATE r • + " - ,:%_ / i",!+_. __..._
,L_J I,_ _'L,-. I ll_._...'l t _-,.,#I I [ i*._ }
..... :F_' _ 'q'_ i .4-'-'#_ .'MDATE., ;'__,__ _..... ,..,,..
-'=-- '.'n "'- :.'n ,..............'_-' _" SET,_,: ,.,,_,. ,_, ,_..+......... .,.b i i b'-T',_i,_,,
U 1:_-,t-LH Y o_i;z- £.'/ --',/ / m
P GE '" "....- Bg32B"............................_.............................................................• ,:_,-,, i ,:N-I L-I",li+; l+Ji-l_C,-,-"C,,2,;--'-O ;"_d,i i.'-'/L-':L-+- :-+'O FI-IL'_EO +-.--,:+,-.-_--,,_,_,IF iLL--'
r-_ _-.v..,i i _._.1r-- i_ i -r-I ili _ I_1"1_ ill |
U-i ill lli_ d,t;-tmbi.-" ..... ';"; ith, .... ._.,_,,,_ ,,,_.._+,_+,e iLbY; t',.:L--'v, lb+b-f,.-- -; 11 ii i_-_<_.._.i !"-+_i-%. i L. I
,2,1 L--'L+[ • I -;0-2,
....................... :' : _______R.iGFD+.nF£,' +"+_"_":'"":+' _P.
•_tJr..T! .':!_T ''+ r-,,,,n_ ,,._ ^r._- ,_,r- +'._-..+N_:PIiL. -1",_Iib _,.,'Ir+-"..... • _% MLFx/iq.h _t.':i
I'II-Ig-D - i i!_.1%1 i ! _ l, £i1} ! I_L-! Ikk I I I .t _.._i I
I i '- ;'" U{",.I " '_+_- _ ..... " ..........
-'t ..... / .+.,,--+Fi_----................................. _ ................................... :iELD EFFECTl,l ;+,t+--_........... +- +-- ; ._., ++.-++,+. i--L,.I"-._I. I f.'_ : I-.f¢lU 1 ." ,--_L:_ ,_
+_,_,i'+_.'STORS., ...... i FT-Dr-I
LJ i.-: _ .......
_c._. r,._'.'.,,._-,_+,,'_m_;:_, :¢r a 3,.7.+GH..-udA_; .,us+o+Jth+c receive module TC:,.r..................... lUdti Of] d._itefllid +_ru ....... - ._::1,-,uPdceb,-_;riic c:'-_::)iii)iun........... x..... _, -;o+=-,- :i-,i_] j_ =. ;_;.-,_;t ,_..:..... _r;_li,_.,+,J('e df--i
" i F " +_'......... " .............. ' I # .................. t,.-;I -.......................... -ii.l-'.J',.,'_fdlI [_U._!:.,'_ It+-.,uu,e iF_:,ise , 9U L-'+ ...... - +r_ c,- ..... ,.-q.... . .... F:r r+ 1,4),Lfisix levels of +r-.f,--.,_m.--.H_.-.-t..-,+-,-_., ,---,--..,-.,f;-,.--,l - -+,.,...-el,;_ ,-:...... shi+ .... -'................ •............. , ...... , d ._'.'- ::l [ i-'+;d_,,-_ ,LeF.+ .--+;rJ
• . +..:_.: :==_ . - _.=_ _...... tel.:. . t_ m. • !+I- - "m-"vi +itl'+i.,'--'um,_+"+,._-i++ui.+.+P'tlv+iuf 250 ++:[g.T:++ Rr ":.... -'=.'+ ;ur _'aL.i; ;-'1 ....lt;L{. _.... .. ! i.--
. L_,U+.-'+:_v,-+;+;;_, k lL-"b'; IIL'I--_,'--_ d.,'li_; I i ),-_1.+ _,;-',+L %,-++:: ;-v++;, "---+ .-+ +,'-+.J'- .J: + _ J k'_i .+ diiL+" l'x.P
IF c-i b;-m,+-+,+llil_) --=_.=.--.-n-;:ac-..--.ni--.-,_l EAb_._--'[ TOP "li_ .... ._-.,,1 ftet, h.,-+_-.- _ _ _-i.__.+ 1!1_%-1 I _. {..tl %. i-i "--_l'--It L'--_tl it i ::tl
i %.--I'_tLI 1:#/ ! _.l/.s IlUI _-+_ P'P-I_; - ,--.k','%F', _:#.'-'- -,---F-,FI .........+,-.... ;-vl i. fi _,tilJ-!!_, i iili._., ,..,+_ =.,,.-+_.. i "--_ +z_, kilb _l( "-__ _IIIP" iks:"-_L_-tJ
....................... '++'":=,t+'+__..... l ,--+ater"++n tile _,._,,+.,,,_.._+_=:,_-,_,-,-,,-+,,ies.FL. ti,_ '-_,_+_icon--'+_ul, a ....... +_+_ • __-E P--,ITE D :
30GHz MonolithicReceiveModule
FirstAnnualReportforPeriod
-- November3, 1982-October31,1983ContractNo.NAS3-23356
PreparedBY
HoneywellPhysic_SciencesCenter
10701LyndaleAvenueSouth,81oomington,MN 55420V. Sokolov,J. Geddes,P. Bauhahn
_- Phone(612)887-4452
_ For
NASA-LewisResearchCenter
C_municationandPropulsionSection
21000BrookparkRoadCleveland,OH 44135
A1 Downey,TechnicalManager
MailStop54-5
V. Sokolov T.C. Lee
ProgramManager DepartmentManager
47119
1. Report No. 2. Government AccessionNo. 3. Recipient's Catalog No.
r_ASACR-1683264. Title and Subtitle - 5. Report Date
December1983
30 GHz MonolithicReceiveeModule FirstAnnual Report 6.PerformingOrganizationCodefor Period November3, 1982 - October 31, 1983
7. Author(s) 8. PerformingOrganizationReport No.
V. Sokolov, J. Geddes, P. Bauhahn10. Work Unit No.
9. PerformingOrganizationName and Address
11. Contract or Grant No.
HoneywellCorporateTechnologyCenter10701 LyndaleAvenue South,Bloomington,Minnesota55420 NAS3-23356
13. Typeof ReportandPeriodCovered
12. SponsoringAgency NameandAddress Annual, 11-3-82to 10-31-83
NationalAeronauticsand Space Administration 14.SponsoringAgencyCodeWashington,D.C. 25046
Supplementary Notes
ProjectManager,Alan N. Downey,Space CommunicationsDivision,NASALewis ResearchCenter,Cleveland,Ohio 44135
16. Abstract
This report describes the first year's progress on a four year program to develop a 30 GHzGaAsmonolithic receive module for spaceborne communication antenna feed array applications. Keyrequirements included on overall receive module noise figure of 5 dB, a 30 dB RF to IF gain withsix levels of intermediate gain control, a five-bit phase shifter, and a maximumpower consumptionof 250 mW.
RF designs for each of the four sub-modules(low noise amplifier,gain control,phase shifter,and RF to IF sub-module)are presented. Exceptfor the phase shifter,high frequency,low noiseFETs with sub-halfmicron gate lengthsare employedin the sub-modules. For the gain controlatwo stage dual gate FET amplifieris used. The phase shifteris of the passiveswitchedline typeand consistsof 5-bits. It uses relativelylarge gate width FETs (with zero drain to sourcebias)as the switchingelements. A 20 GHz localoscillatorbufferamplifier,a FET compatiblebalancedmixer, and a 5-8 GHz IF amplifierconstitutethe RF/IF sub-module.
A novel approachto the phase shifterfabricationusing ion implantationand a self-alignedgate techniqueis described. This method of phase shifterfabricationholds promise in reducingRF lossesat high frequencies. PreliminaryRF resultsobtainedon such phase shiftersarealso presented.
17. Key Words (Suggestedby Author(s)) 18. Distribution Statement
Ion Implantation 30/20 Communication_MM-WaveMonolithicCircuits SatelliteGaAs MonlithicCircuits Ka-BandReceiver Unclassified- UnlimitedLow Noise FETs MonolithicMixerPhase ShifterSelf-AlignedGate
19. St_ar_y _a_lf.(ofthisreport) j20. SecurityClassif.(ofthis _ge) 21. No. of Pages 22. Price*
Unclassified I Unclassified
* ForsalebytheNationalTechnicallnformationService,Springfield,Virginia22161
NASA-C-168 (Rev. 10-75)
NASA Annual Report
.-- Table of Contents
I. SUMMARY S-i
II. PROGRAM PROGRESS BY TASKS 1
1.0 ReceiveModule Design--TaskI 1
I.I CircuitDesignand TestPlan 11.1.1 Overall Receiver 2
1.1.2 Phase Shifter Design 5-- 1.1.3 Gain Control 10
1.1.4 RF/IF Submodule 121.1.5 30 GHz Low Noise Amplifier Submodule 15
1.2 Device Design 161.2.1 SAG FET for Phase Shifter 16
1.2.2 100 Micron Low Noise FET 18
2.0 Phase Shifter Fabrication and Evaluation -- Task II 31
2.1 Results for 180 ° SAG Phase Shifter Design Using Shunt FETs 32
2.2 Results From Second Mask Set Using SAG Series Switches 333.0 Gain Control Fabrication -- Task III 35
3.1 100 Micron FET 37
_- 3.2 Dual Gate Amplifier 37
4.0 RF/IF and LNA Submodule Fabrication and Evaluation -- Tasks IV and V 38
5.0 Submodule Integration and Receive Module Development -- Tasks VI and VII 38
'" 6.0 Product Assurance and Reporting -- Tasks IX and X 38
III. REFERENCES 70
oooIII
NASA Annual Report
List of Figures
Figure Page
S-1 Submodule Functions of Receive Module S-8
S-2 Block Diagram of Overall Monolithic Receiver S-8
S-3 Layout of 180° Bit Using Four Series FET Switches S-9
S-4 Photomicrographs of SAG Phase Shifter Circuits on a Processed Wafer S-10
S-5 Calculated and Measured Differential Phase Shift for 180° Phase Shifter S-11
S-6 CALMALayout for 100 Micron Gate Width Low Noise FET S-11
S-7 Low Noise FET Design S-12
S-8 Design for Monolithic Five-StageAmplifier S-12 -
S-9 CALMALayout for First Two Stages of Gain Control Amplifier S-13
S-10 Design for Monolithic 2-Stage Dual-Gate Amplifier S-13
1 Block Diagram of Overall Monolithic Receiver 39
2 Interconnected Receive Module and RF Test Fixture 39
3 Simplified Equivalent Circuits for ON and OFF FETs 40
4 Insertion Loss and Isolation versus Gate Width 40 -.
5 Chip Layout for 180 ° Bit 41
6 Layout of 180° Bit Using Four Series FET Switches 41
7 Computer Calculated Results for 180 ° Phase Bit Using Series Switches 42
8 Calculated Differential Insertion Phase for 3-Bit Phase Shifter(180°, 90°, 45° Bits) Using Series FET Switches 43
9 Calculated Insertion Loss "Envelope" for Eight States of 3-Bit _.Phase Shifter 44
10 Gain Control Design 44
11 Design for Monolithic 2-Stage Dual-Gate Amplifier 45 -
12 RF/IF Submodule with Rat Race Hybrid 45
13 Diode Structure for High MixerPerformanceand Fabrication Simplicity 46
14 Noise Figure and GainCirclesfor First Stage of 22 GHz Amplifier 47
iv
NASA AnnualReport
List of Figures (Continued)
Figure Page
15 Design for Monolithic 22 GHz Amplifier 48
16 Design for Monolithic Five-Stage Amplifier 48
17 Channel Sheet Resistance versus Gate-to-Channel Voltagefor Different Implant Energies 49
18 Source-Drain Resistance versusChannel Sheet Resistance 49
19 100 x 0.25 MicronLow Noise FET Design 50
20 Gain and Noise Circlesfor Low Noise FET Input Plane- No Cdg or Ls for Noise Circles 51
21 Gain and Noise Figure Circles for Low Noise FET Input PlaneIncluding Cdg and Ls 52
22 Gain Circles for Low Noise FET Output Plane 53
23 Effect of Gate Resistance on 29 GHz Gain and Noise Figure 54
24 Effect of Source Resistance on 29 GHz Gain and Noise Figure 54
25 Insertion Phase for 180 ° Phase Shifter (TR061A)As Measured on the Network Analyzer 55
26 Test Fixture and Measured Insertion Loss for the Two Statesof the 180 ° Phase Shifter 56
27 Calculated and Measured Differential Phase Shift for 180 ° PhaseShifter (TR061A) 57
28 Measured RF Results for SAG 180 ° 1-Bit Phase Shifter 58
29 Comparison of Measured and Calculated Insertion Loss for SAG Phase Shifter 59
30 Photomicrographs of SAG Phase Shifter Circuits on a Processed Wafer 60
31 Calculated and Measured Differential Phase Shift for 45° Phase Shifter 61
32 Calculated and Measured Differential Phase Shift for 90° Phase Shifter 61
33 Calculated and Measured Differential Phase Shift for 180° Phase Shifter 62
34 180° Phase Shifter 63
35 45° Bit Insertion Loss 63
36 45° Bit Return Loss 63
NASA AnnualReport
List of Figures (Concluded)
Figure Page
37 CALMA Layout for 100 Micron Gate WidthLow Noise FET 64 -
38 CALMA Layout for First Two Stages of Gain Control Amplifier 64 -
39 Profile of Ion Implant 65
• List of Tables
Table Page
S-1 NASA's Key Performance Goals S-7
1 NASA's Key Performance Goals 66
2 Proposed Initial Low-Noise FET Design 66
3 Calculated Parameters for FET Design 67 --
4a S Parameters with No Cdg or Ls 68
4b S Parameters with No Cdg, But with Ls 68
5 S Parameters for Complete Model, Including Ls and Cdg 69
vi
I.SUMMARY
The objectiveof the four year programis to developa 30 GHz monolithic
-- receivemodulefor communicationantennafeed arrayapplications,and to
deliversubmodulesand 30 GHz monolithicreceivemodulesfor experimental
evaluation.Key requirementsincludean overallreceivemodule noisefigure
__ of 5 dB, a 30 dB RF to IF gainwith sixlevelsof intermediategaincontrol,a
5 bitphaseshifter,and a maximumpowerconsumptionof 250roW. In addition,
_ themonolithicreceivemoduledesignaddressesa costgoalof lessthan $i,000
(1980dollars)per receivemodule in unit buys of 5,000 or more, and a
mechanicalconfigurationthat is applicableto a space-bornephasedarray
. system. Theserequirementsare summarizedas performancegoalsin TableS-I.
o- The designof the monolithicmodule (TaskI) includesa partioningof the
receiverintofoursubmodulefunctions:the low noise amplifier(LNA),the
5-bit phase shifter (PS), gain control (GC) and the RF/IF frequency
down-conversionfunctionas shownin FigureS-I. The PS, GC, RF/IF and LNA
functionsembodiedin the monolithicreceivemoduleare alsodevelopedas
_ individualsubmodules,each fabricatedon a separate chip (Tasks II-V
respectively),anddesignedso as to permittheirinterconnectionto formthe
interconnectedreceivemodule(TaskVI).
p,
Duringthe firstyear of the programwork has concentratedon deviceand
circuitdesign (TaskI) and the fabricationand evaluationof the phase
shifter(TaskII)as wellas the initialdevelopmentof thegaincontrol(Task
_ III)submodules.Consequently,SectionII of thisreportdiscusses,for the
61
most part, the details of the progressmade on the first three tasks.
Nevertheless,sincemanyof the taskshaveoverlap,especiallyin the design
area, the work accomplishedin the first year is also applicableto the
developmentof theothertwo submodules.Forexample,in the caseof thegain
controlamplifier,thefirstmask set includesi00microngatewidthlownoise
FETs which are also requiredfor both the RF/IF and the LNA submodules.
Highlightsof specificaccomplishmentsduringthe firstyearare outlinedin
the following..
OverallMonolithicReceiveModuleDesian- FigureS-2 showsthe functional
configurationfor the monolithicreceivemoduleandalso indicatesnominal
gains, noise figures,dc power consumption,and overallnoise and gain
performancefor each su_noduleelementand the completereceivemodule. The
lownoiseamplifierprovidessufficientlyhigh gain (32 dB) and low noise
operationto be themajordeterminingfactorin establishingthe required5 dB
noisefigurefor theoverallreceiver.To achievethistypeof performanceit
is expectedthatsub-halfmicrongatelengthFETswith i00microngatewidths
willbe required.
The phase shifteris a 5-bit passivedeviceand utilizesself-alignedgate
(SAG)switchingFETsto achievelowinsertionloss. Theswitchesare used to
switchbetweentransmissionlinesof differentelectricallengthsto achieve
true time delayphaseshiftingforthelargerphaseshifts(180o,90° and 45°
bits),and to switchin reactiveloadingon the lines to accomplishthe
smallerphaseshifts(22° and iio bits). A SAGFETapproachutilizingseries
switcheshas been chosenas the base line approachto the phase shifter
_2
development.
Gain control is accomplishedby a dualgateamplifierin conjunctionwitha
passiveswitchedattenuator.The formerutilizes100 micron low noise dual
gate FETs,while the switchedattenuatorusesthe switchingFET technology
developedin theswitchedlinephaseshifter.
• Finally,the RF/IFsubmoduleincorporatesSchottkydiodesin a balancedmixer
configurationto accomplishthefrequencytranslationto IF (5.5- 8 GHz).
Again,a low noiseamplifieris usedto providea suitableLO signalat 22 GHz
to drivethe balancedmixer. A uniquefeatureof the design is that the
Schottky diodes are FET compatible in that fabrication of the diodes as well
as the FETs in the LO and IF amplifiers can be done on the same chip.
To address the cost goal for the monolithic receive module the basic
fabrication approach is ion implantation into semi-insulating GaAs.
Design and Init_l Fabrication of the Multi-Sit Phase Shifter - Figure S-3
_ showsa representativedesignof a 180obit usingseries,400microngatewidth
FET switches.The seriesswitchdesignreducestherequiredGaAsareaoveran
equivalentshuntdesign (testedearlyin theprogram)by nearlya factorof
two. It also introducesless parasiticreactancedue to a more compact
layout.FigureS-4 showstheactual4-bitphaseshifterfabricatedby theSAG
techniqueon the 0.15mm thickGaAswafer. Thefirstthreebits (180o, 90°,
45o) are switchedlinephaseshiftersand the lastbit (22o) isa loadedline
- phaseshifter.
FigureS-5 showsthemeasureddifferentialinsertionphaseforthe two states
of the 180Obit across a 5 GHz band (27.5-32.5GHz). Exceptfor a constant
offsetof about7o, (whichcanbe easilycompensatedfor in the finaldesign)
the measured curve follows exceptionally well the ideal time delay
characteristiccorrespondingto 6.3O/GHz.The insertionlossmeasuredon this
firstrunof SAGphaseshiftersusingseriesFETsis about3.5-4dB (excluding
fixtureloss). Thisis about1 dBworsethanthebestSAG resultsobtainedon
earlierversionsof the 180obit usingshuntFET switches.
Similarresultswere obtainedon the 90o and 45° bits. It is expected,
however,thatwithan optimizedSAG process,an insertionlossof about 1.5 -
2 dB per bitwillbe achieved.
Designof a 100 MicronGate-Width,Sub-Half-MicronGate Length FET - To
achieve the requirednoiseperformanceof the 30 GHzLNA,a FETwitha noise
figureof notgreaterthan4 dB andwithan associatedgain of at least 6 dB
must be developed. Based on extensive device design and circuit
considerations,a 100microngate-width,and sub-half-microngate lengthFET
has been designed.FigureS-6 showsthe computergeneratedlayoutof the100
micronFET. The gateconsistsof two 50micronfingers,each having a gate
length of about 0.25 microns. Thisconfigurationwas determinedto be best
suitedforachievinglownoiseandhighgainoperationat 30 GHz and yet be
consistentwith impedancematchingschemeswhicharerealizablein monolithic
form.
8-4
From an analysisof the specificfabricationrelateddesignparameters,an
equivalentcircuitshownin FigureS-7was derivedfor the i00micronFET. As
shown in the figure,thepredictedperformanceof theFET,as calculatedfrom
theequivalentcircuit,includesa maximomavailablegainof 8 dB and a noise
figureminimumof 2.5 dB at 32 GHz.
_- With the equivalentcircuitestablished,a preliminarydesignof a fivestage
LNAwas completedusingthe100micronFETsina single-endedcascadeas shown
in FigureS-8. The computeroptimizedamplifierdesignpredictsa gainof
about32.5d8 from27-31GHz.
_ Designand InitialFabricationof the Gain ControlSubmodule- The gain
control (GC) is accomplishedby incorporatinga two-stagedual gate FET
°.4 amplifierand controllingthe secondgate voltageto vary the RF gain. A
passiveswitchedline attenuatorfollowsthe amplifierto addan additional
attenuationbit. To date,workhas beeninitiatedon the developmentof the
dualgateFETamplifier.
.i FigureS-9 showsthe initiallayoutof thetwo-stagegaincontrolamplifier,
as generatedby theCALMAcomputerizedlayoutsystem. The dualgateFETshave
i00micronwideby 0.3micronlonggatestripes,andare basedon the design
and fabricationof thesinglegatei00micronFETsusedin thefront-endLNA.
On-chipsiliconnitridecapacitorsare usedto RF groundthe secondgateof
eachdualgateFET,yet permitapplicationof a DC controlvoltage for the
gain control function. All DC bias distributionand filteringis also
- includedon thechiplayout.
Finally,FigureS-10showsthe computedresponseof the two-stagegaincontrol -
amplifierin itsmaximumgainstate. The insetalsoshowsthedetailsof the
circuitschematicwhosecorrespondinglayoutis shownin the previousFigure,
S-9. Overthebandof interestthecomputedgainis 20.3+ .2dB.
Clearly, the designs developed during the first year are either starting
points (as in the gain controlsubmodule)or interimdesigns (as in the case
of the phase shiftersubmodule)which will be updatedas the experimentaldata
base is obtained. Nevertheless,the resultsof the theoretical and computer
calculations discussed above serve as guidelines for further device and
circuitdevelopment which will lead to the final designs of the receiver
submodulesand the monolithicreceivemodule.
The followingSectionII presentsthedetailsof theprogressmade during the
firstyearof the programby tasknumbers.
Table S-1. NASA's Key Performance Goals
-- Design Parameter PerformanceGoal
RF Band 27.5- 30 GHzIF Center Frequency Between 4-8 GHzNoise Figure at Room Temperature <_5 dBRF/IF Gain _-30 dB at highest level of gain controlGain Control . At least six levels; 30, 27, 24, 20, 17 dB
and Off.Phase Control 5 bits;each bit +_3° at bandcenterModule PowerConsumption 250 mWin all states except OFF. In OFF
state, 25 mW.Phase and Gain Control Operate on digital input.Mechanical Design Fully monolithic construction; compatible
with 30 GHz spaceborne phased arrayapplications.
"_- Unit Cost Less than $1000 (1980 dollars) in unit buysof 5000 or more
DIGITALCONTROLLINES(TrL)
¢ \/ \
/ \
/ \ ---
/ \\
z \
/ %
I_.,. ! I- - "_-- I I...... I -I i i I II '_,, ,,. I I I I I |RF ..J LNA" , PS i = GC ' _ IF-I .'" i I = I i RF IF I
IN i .-" , I I I I OUTIIIIII
L_ •REFPHASEINPUTAT LO FREQUENCY
Figure S-1. Submodule Functions of Receive Module
TIL LOGIC SIGNALSIFOR DIGITAL CONTROLI
11 iDtGITAL CONTROL
,.f'--AI 22 GHe SIGNAL
r ,r - - -',II LOwNOISEAMPLIFIliR II II P. SeSHlteR I I GAINCONTROLI "1v_ /LO AMP IFAMP III i I I I ' i L +6dBm I
INPUT I [>1[ [r _>,4 __ l l [_ II (_)____> ' ouTPutz75:3o l I l I l 1 5.58.0GHi t .J L J L_ .... / L j c.H_
LO OlOOE_OR 6STAGE FET AMP PS LOSS DUAL GAT.,_._E, AM..__P MIXER If AMP
NF qdBI 4 4 4 6 47 _ 14 _ IO 5 6 4
GAIN (de1 6 6 6 _7 l -<_ -I _ 12 24 --6 5
VOLTAGE IVl :I 3 3 _1 3 4 3 3
CURRENT ImAI 4 4 4 4 4 I 5 26 8
SURMOOULE F = o0 (JB F..,= 14048 +f w` - IF _ F_m m 10 dBNOISE FIGURE f_, . 1OdUB Gm
F II 48dBMONOEIIHI(: HFC|IVI MOI}ULIr
[DIAl I)l: POW[R 250 mW IINCLUDING H illW FOR DIGiTAl CONTROLIOVERALL NOISE FIGURE 4 9 dB IMAX GAINI 5 0 dB IMIN GAINIRf III (:ONV| RSION GAIN 35 lIB I,II_l. _l._lb, 2:1 dH IMIN GAINIMINIMUM D[ I [ C |AH| | SIGNAL IO dl:b.DYNAMI( HANG[ :10 dR IMAX GAINI 4:.111bl IMIN (;AIN,
Figure S-2. Block Diagram of Overall Monolithic Receiver
S-8
f_
L
_o .//---400 MicronFETs
- GatePad--
l 1IN OUT
Figure S-3. Layout of 180° Bit Using Four Series FET Switches
--. S-9
(a)
J --.(b)
FigureS-4.PhotomicrographsofSAG PhaseShifterCircuitson aProcessedWafera) The fourbitphaseshifter(45°,90°,180o,22o)usingseries
FET switchesb) Enlargedviewof180°bit
S-IO
,---- 15o I I I I I i I27 28 29 30 31 32 33
Frequency (GHz)
" Figure S-5. Calculated and Measured Differential Phase Shift for 180° Phase Shifter
s0u Ej[i 91 "i-'i_'i'i'i'i'i.i..91.9
50UREE
Figure S-6. CALMALayout for 100 Micron Gate Width Low Noise FET
__ S-ll
a b 2.7 .01 11.3 : -
,,. I° -•, t Jl"2 -_ ,_[_ . N • b
• BUFFER '5.4 Units
Gate length Lg = 0.25 microns gm = 16.9e'J'"_mS _ -Unit gate width w = 50 microns _ = 2.5 ps pFChannelthickness a = 0.15 microns _=0.07 nH
Dopingdensity N = 2.5 x 1017
Totalqate width Z = 100 microns _[ _Gate-sourcespacing Lsg = .75 microns
Gate metalthickness h = 0.5 microns I 9.3dB @ 27 GHzSpecificcontactresistance rc _-i0"6 ohm.cm2 MAG = 78dB@ 33GHzaI = 0.22
a2 = O.15
N1 = 2.5 x 1017
N2 = 2.5 x 1017
Calculated n(wse figure = 2.5 dB at 32 GHz
Figure S-7. Low Noise FET Design
50t3 46 ° 79 ° . 63 ° 46 ° 79 ° 63 ° 40 ° 57°
=" Stage 1 2 "=- 3 4 "-- 5 = _
TRL's have 7_o= 909Reference Frequency = 29 GHzAllBlocking Cap's = 1PF J-'
34 All FETs 100#m Gate WidthChip Layout ~ 1.58 x 1.25 mm2
_, 33
._R 32
31
3O
29
I I I I I I I25 26 27 28 29 30 31 32
Frequency (GHz)
Figure S-8. Design for Monolithic Five-Stage Amplifier _.
S-12
0.65 mm
< 1.9 mm >
- Figure S-9. CALMALayout for First Two Stages of Gain Control Amplifier
50t2 52° 87.5 ° 47.3 ° 1PF 1.5 nH
22 - AUTRLshaveZ.= 90t2,:, ReferenceFrequency = 29 GHz
21 _ AllFETs = 100 #m GateW_th
20- f • :/_ 18o
17
16-
15 . I I I I I I26 27 28 29 30 31 32
Frequency(GHz)
Figure S-IO. Design for Monolithic 2-Stage Dual-Gate Amplifier
----- S-13The reverse of this page is blank.
If. PROGRAMPROGRESSBYTASKS
The objectiveof the 4 yearprogramis to developa 30 GHz monolithicreceive
moduleforcommunicationantenna feed arrayapplications,and to deliver
submodulesand 30 GHzmonolithicreceivemodulesforexperimentalevaluation.
Key performanceand designgoalsareshownin TableI.
1.0 ReceiveModule Design - TaskI
A major portionof the firstyear'swork was directedtowardsdeviceand
circuitdesign. Thisincludesthe detaileddesignof a 30 GHz monolithic
receivemoduleand foursubmodules.The submodulesare a lownoiseamplifier
(LNA),a phaseshifter(PS),a gaincontrol(GC),andan RF/IFsubmodule.The
submodulesare designedto functioneitherindependentlyor interconnectedto
forma completefunctioningreceivemodule. Accomplishmentsincludethe
design of low noise 30 GHz FETs, switching FETs for phase shifter
applications,amplifierandmixerdesignsas well as the designof switched
line multi-bitphase shifters. To addressthe costgoalof lessthan$i000
permonolithicreceivemodule,ion implantationis the baselinematerials
technologyapproach.
i.iCircuitDesianandTestPlan
The circuitdesign effortaddressedthe designof eachsubmoduleincluding
interfacespecifications,as wellas theoverallreceivemodule. Since the
phase shiftersubmodulesare the firstto be scheduledfor deliveryin the
secondyearof theprogram,a more comprehensivedesignwas completedforthis
submodule than, for example,for the RF/IF or LNA submoduleswhich are
scheduledfordeliverytowardstheend of thethirdyear of the program. In
fact, two fabricationiterationsof the phase shifterdevelopmentWere
completedand evaluatedduringthe firstyear. In addition,a test plan for
testingeachof the submodulesas wellas thereceivemodulein accordancewith
thedesignparameterswas alsocompletedand approvedby the NASA Technical
Manager.
The followingsubsectionsdeal with the specificcircuit designdetails
completedforTask I of theprogram.
i.i.i OverallReceiver
The functionalconfigurationfor the monolithicreceivemoduleis shownin
Figure1 which also indicatesnominal gains, noise figures, dc power
consumption,and overallnoiseand gainperformancefor eachsubmoduleelement
and the completereceivemodule.
The firstmodule functionto be performedis low noise RF amplification,
followedby RF phaseshiftingand gain controlusingsubmicronGaAs FETs.
Frequencydown-conversionand amplificationat the IF is performedwith a
balancedmixer,whichconsistsof a pairof monolithicallyintegratedSchottky
barrier diodesand a GaAs FET IFamplifier.The localoscillatorsignalis
suppliedby the LO amplifier,whichraisesthe reference22 GHz LO signal to
+6 dBm. This LO power is sufficientto drivethepairof mixerdiodes. To
maintaina lowlevelof LO noise,the LO amplifieris designedfor a 5 dB
noisefigure.
The phaseshifteris a passivedeviceandmakesuseof Self-alignedgate (SAG)
FETsto reduceresistivelossesin the channels.Sinceeachsubmodulemustbe
capableof individualcharacterizationandmust thereforefunctionas a
separateunit,thegaincontrolfunctionis implementedat RF (incontrastto
an approachthatwouldrealizepartof thegaincontrolfunctionat IF).
The LNA consistsof five FET stages with an overall noise figure of
approximately4.8 dB and a gain of 32 dB. TheLNA gainis highenoughthat
itsnoisefiguredeterminesto within0.2 dB the noise figureof the total
receive module. This is necessarydue to the lossesand noise figures
associatedwith thephaseshifter,gaincontrolandRF/IFfunctions.
The gain controlfunctionfollowsthe phaseshifterbecauseit is expected
thatthenoisecontributionof thegaincontrolduringminimumgain setting
will be greaterthan that of the phase shifter. Specifically,thephase
shifteris budgetedfor 8 dB loss while the gain controlamplifierduring
minimumgainsettingshas a 1 dB lossanda 14 dB noisefigure. Thisapproach
r_ is verifiedby Friis'formula(aswell as common engineeringsense),which
shows that the overallnoise figureof the two functionsis lowestif the
phaseshifteris placedaheadof thegaincontrol. Itmay be noted that the
higher noisein thegaincontrolis dueprimarilyto thenoisydual-gateFETs
duringthelowerlevelsof amplification.
o
The local oscillatorfrequencyis at 22 GHz,with a resultant5.5 to 8.0 GHz
IF. Withthischoiceof LO frequencytheimagebandis in the rangefrom14.0
to 16.5 GHz. Good image rejectionis achievedbecausethe imagebandis
locatedbelow the cutoff frequencyof the WR-28 waveguide(fcutoff= 21.1
GHz). In the caseof a WR-28waveguideinputto theLNA (asduringsubmodule
RF testing),the imageband will be automaticallycutoffby thewaveguide.
Furthermore,the imagebandis sufficientlyremovedfremthe passbandof the
LNA thata 30 d8 or greaterrejectionwillbe accomplisheddue to the passband
characteristics.
Submodule Intearationand Cb_p LayoutDesig_ - The four submodulesare
interconnectedon a single3 x 9 mm2 carrierplateto formthe interconnected
receivemodule,as depictedin Figure2. Connectionsbetweenthe four chips
aremadewithbeamleadsand thermocompression,or ultrasonicbonding.
Formoduleand suhmoduletestingthe rf inputinterfaceis a WR-28 waveguide
(26.5-40GHz),whiletheLO inputinterfaceis a WR-42waveguide(18-26GHz).
Fin-linewaveguide-to-microstriptransitionsare used at both the RF and LO
interfaces. The IF outputis takenoff the chipby a 50 ohmmicrostripline
to an SMAoutputconnector.BiasandTTL signalsare routedfremthe chip to
externalconnectorsvia appropriateprintedcircuitboards.
An importantfeatureof the testfixtureis that the chip is enclosedin a
waveguide below cutoff. This providesinherentrf shieldingand good
isolationcharacteristicsbetweentheinputandoutputas well as imageband
rejection.
_ The ultimate goal of the program is to achieve a fully monolithic 30 GHz
receiver chip. This approach, if practical, will be implemented towards the
4 end of the programandwillbe basedon the designsand resultsof the inter-
connectedmodule.
1.1.2 PhaseShifterDesi_
_I The phase shiftersubmodulemakesuseof switchedtransmissionlinelengths
and passiveFET switches.Two phaseshiftercircuitdesigns(aswell as two
fabricationapproaches- discussedin SectionII-l.2and II-2.0)weretested.
Initiallya test circuitincorporatingFET switches in shunt with the
transmissionlineswere used. Thisdesignwas employedat firstbecauseit
was considered a more conservative approach and served as an experimental
basis for the following design iterations. The second approach uses FET
switchesin serieswiththe transmissionlines. It has two advantagesover
the former designin that top-sideRF groundingcan be avoidedand the
microstripcircuitlayoutrequireslessGaAsarea.
2_ - An important designconsiderationis the dissipativeloss
r associatedwith the switchesand the microstriplines. As shown in the
following,the dominantlossesoccurin thepassiveFETs.
Microstrip-lineloss is calculatedto be about0.15dB perwavelengthat 30
GHz basedon metallossesonlyanda goldthicknessof 1.5_m for a 50 _ line
on 0.2n_nthicksemi-insulatingGaAs [4].
For theswitchingFETs,however,RF lossesaremuchgreaterand are primarily
associatedwith the parasiticresistancesof the device. Figure3 shows
simplifiedequivalentcircuitsforan ON and OFF FET,namelya singleresistor
for the ON (low impedance)stateand a seriesresistor-capacitorcombination
for the OFF (highimpedance)state [2-5]. The capacitorCoff is the total
pinchoffcapacitancebetween sourceand drain,and the resistorrs is the
totalseriesresidualresistanceat pinchoff(undepletedchannelresistance
plus sourceand drain contactresistance).To increasetheRF impedanceof
theOFF state,an inductorL parallelresonatestheRC combination.For this
case, the effective RF OFF resistance at resonance,Roff,is closely
approximatedby Roff = Q2crs= (_rsC)-2rs = (C)-2rs-I. NotethattheratioA
Roff/Ron is to firstorder,justthe figureof meritQ as definedby Kurokawa
and Schlosserfor a switchingdeviceswitchingbetweentwo impedancestatesZ1
and Z2.
It can be shownthat
_(Ronrs)-i(_C)-2= Roff/Ron [2,3]
It is clearthatthe figureof meritdegradeswithfrequencyand has smaller
valuesat millimeter-wavefrequenciesthanat lowermicrowavefrequencies.To
maintainthe figureof merithigh it is essentialto reduceRon and rs (andC
if possible).
A specificexampleof an idealRF switchincorporatingconventionalpowerFETs
I
fortheswitchingdevicescanillustratethelossesincurredat 30GHz. Fora
typical 300 #m power FET, Ron = ii-14_ and with C = 0.07 pF and rs = 5_,
Rof f is about I000_ at 30 GHz. By choosing these ON and OFF resistance
-" valuesand usingthemin an idealSPDTswitchconfigurationas shown in the
inset of Figure4, we can calculatethe idealinsertionlossand isolationof
a singleswitchat centerfrequency.Furthermore,by scalingRon, rs,andC
for other gate widths, an estimate can be made of the best compromise between
isolation and insertion loss. Note that this leaves the figure of merit Q
_- invariant. These considerations are summarized in Figure 4 where the
insertion loss and isolation at 30 GHz are calculated as a function of gate
width for two different Q values. Note that for the dashed curves which
represent a reduction in Ron and rs (14_ to 4.7_ and 5_ to 2.4_,
respectively), the loss is reduced by 0.6 dB at a gate width of 300 #m, for a
single SPDT switch. For a four-bit switched line phase shifter this amounts
to nearly 5 dB less insertion loss. As shown in Section II-l.2.1 such
reductionsin the parasiticresistanceare possibleusingthe self-aligned
fabricationprocess.
First Desiqn:180° - 1 Bit PhaseShifter(ShuntFETs)- Figure5 sh_s the
chip layoutfor the 180° bit. Four switchingFETs are used to RF switch
betweentwomicrostriplineswhosedifferentialelectricallengthis equalto
180oat centerfrequency.TheFETsarepassiveinthesensethatno dcbias
is appliedto thedrainandonlya switchingvoltage (0,-6V)is used at the
gate [4-6]. The circuitutilizesa pairof SPDTswitchesrealizedby four,o
300_m gatewidthFETsin shuntacrossthe50_ transmissionlineswith each
FET located at a distance of _g/4 (0.8 mm at 30 GHz on GaAs) from either the
7
inputor outputT-junction.
The circuitlayoutis conservativein thatthe arearequired(chipdimensions
are 3 x 2.67x 0.2mm3)hasnot beenminimized.Thiswas doneto ensurethat
the couplingbetweenadjacentsectionsof transmission lines would be
negligibletherebyfacilitatingevaluationof initialRF performanceresults.
The 0.2ma substratethicknesswas chosenas a compromisebetweenincreased
microstrip transmissionline loss and extraneouscapacitiveparasitics
(thinnersubstrate),and increasedcircuitlayout dimensions (thicker
substrate). Pads are providedat the edgeof the chipfor grounding.For
thesetests,groundconnectionis accomplishedby a seriesresonantcircuit
consistingof an external0.ipF capacitor(bottomplatesolderedto ground)o4
and a mesh wire inductorconnectingfrom the capacitortop plate to the
groundingpad on the chip. To increasethehighimpedancestate(V=-6V),a
shortsectionof lineprintedon the chip is used to resonatethe pinchoff
capacitance betweendrainand source (C_ 0.07 pF) for each FET. High
impedanceto the gateterminalis providedby bendingwireinductance.
SecondDesign:Multi-BitPhaseShifter(SeriesFETs)- To improvetheexpected
performance of our multi-bitphase shifter,a series switchdesignwas
implemented. A representativelayoutfor a 180° bit is shownin Figure6.
Four,400microngatewidthunbiasedFETswith five80 micronfingersper FET,
are used in the seriesconfiguration.The areaper bit is almosthalvedfrom
thatof a shuntdesign. Figure7 shows the calculateddifferentialphase.
Over a 4 GHz band the differentialphasecharacteristicis within2° of the
idealcurve. Foran assumedON resistanceof 9_ anda pinch-offresidual
8
series resistanceand capacitanceof 4.5_ and .09pF respectively,the
lossis about2.3 dB per bit.
The designof a 3-bit phase shifterusingseriesFETswitchesis similarto
the 180o bit exceptthatdifferentiallinelengthsof 450,90°, and 180° arep
used. To improveinsertionlossandphasetrackingaccuracyover the band,
-- impedancematchingelements (a shortsectionof high impedancelineand a
shuntstub)are usedfor the45° and 90° bits. All threebitsuseexclusively
seriesswitches.Thelayoutfor all 3-bitsis realizedin an area of 2.5 x
4.1ram2.
Figure8 showsthe calculateddifferentialinsertionphaseresultsforall
eightstatesoverthe 25 to 33 GHz frequencyband. The calculatedvalues
(shownas discretepoints)are shownrelativeto the idealcharacteristics,
i.e.,timedelayscorrespondingto thevariousphasestates.The resultsshow
thatoverthe bandof interestthecalculatedvaluesarewithin2 or 3 degrees
of the idealcurve. For thesecalculationsthecomputerprogramCDMPACT was
used. The circuit model included 12 switching FETs, interconnecting
°- transmissionlinesandmatchingnetworks.For theOFF state,the FETswitches
were representedby a 4.5_ resistorin serieswitha pinch-offcapacitance
-- of .09 pF, and the combination parallel resonated by a high impedance 90
transmission line. The ON states were modeled by a simple series resistance
of 9_ (and the 90_ transmission line in parallel).
Figure9 shows the "envelope"for the insertionlossforall eightstates.
All lossesin the switchingFETsand lossesin the transmissionlineswere
9
includedin the calculations.Overthe bandof interestthe insertionlossis
about6.5 to 7 dB,witha spreadof about0.5to .75dB.
1.1.3 _
_e gaincontrolsubmoduleis requiredto providefivelevelsof RF to IFgain
(30,27, 24,20 and 17 dB)and an off state. To maintainthe noisefigure of
the receivemoduleat 5 dB, theworstcasenoisefigurefor the gaincontrol
submoduleis budgetedat 14 dB.
Two criticalrequirementsof this submoduledictateto a largeextentits
design:the requireaentfor a 1 dB attenuationstatewitha noisefigureof 14
d8 or better,and a +12 d8 gainstatewithan allowableinsertionphasechange
of no more than +5o over the full 13 dB gain range. Also,the uppergain
level is limitedto a maximum of +12 dB to ensurea 30 dB dynamicrangefor
theoverallreceiver.Highergainsand subsequenthigherdrive levelswould
resultin overdrivingthemixer,thusreducingthe 1 dB gaincompressionpoint
of the receivemodule.
The requirementscitedabovewill be met with dual-gateFETamplifiersby
controllingthe secondgatevoltage.Tne dualgateFETwitha 0.5microngate
length is expectedto providea maximum gainof aboutI0 dB at 30 GHz. By
controllingthesecondgatevoltage,thegaincanbe adjustedat least over a
20 dB dynamicrangeto an attenuationstateof -i0 dB.
10
Two problemsarisein usingonlyone dual-gatedevicefor gaincontrol:noise
_ and phase shift. Liechti [7] and othershaveshownthatthenoisefigure,
althoughnotmuch degradedfrom a singlegate deviceduringmaximumgain,
_ degradesdrasticallyas the deviceis driventowardan attenuationstate. For
example,at I0 GHz,Liechti showsan ii d8 degradationfroma 4.1 dB noise
figureat an associatedgainof 12.8d8, to a 15 dB noisefigureat a gainof
2 dB. The insertionphaseshift,as a functionof the secondgate voltage,
also variesconsiderablyas the dual-gateFETis broughtfroma gainstateto
-- an attenuationstate. Thus,a singleamplifier,althoughhavingconsiderably
greaterthan13 dB dynamicrangein gain,is inadequatefroma noiseand phase
shiftpointof view.
To overcomethesedeficiencies,a two-stagedual-gateFETamplifier,followed
by a variablepassiveattenuator,is usedforthegaincontrolsubmodule,as
shown in FigureI0. Thefirststagehas gainstatesof 3 and 6 dB, whilethe
- secondstageamplifierhas threegainstatesof 2, 5 and 8 dB. These states
are achievedby applyingproper controlvoltagesto the secondgate. The
switchedattenuatormakesuseof thepassiveSPDTswitch(describedin earlier
sections)to form two switchablepathsof unequalloss. Themajorityof the
lossin thepathof minimumattenuation(-2 dB path) residesin the series
_ unbiased FETs. In the secondpath, a 4 dB increasein attenuationis
required.Thiscanbe realizedeasilyby incorporatingeithera thin film
seriesresistorevaporatedonto the substratefor a sectionof the50 ohm
line,or highresistivityselectiveion implantscanbe incorporatedbeneath
the microstriplineto forma lossysectionof line. Notethatthe tolerance
of the lossis notverycritical,i.e.,on theorderof 1 dB, sincechangesof°
this magnitudecanbe compensatedby thedualgateamplifiers.Someunwanted
phase shift in the dual gate amplifiers can likewise be compensated for by the
switched attenuation, by altering the length of one of the paths.
Preliminary Design - For initial development, a two-stage dual-gate FET
amplifier design is employed. The design is based on the i00 micron gate
width FET discussed in Section II-1.2.2. A cascode connection of two such
FETs are used for the equivalent circuit of a dual gate FET. (Common source
firststage,comaongatesecondstage[7]).
Figure ii shows the circuit schematicand the computeroptimizedgain
characteristicfor the twostagedualgateamplifier.From 27 to 30 GHz the
gain is 20.4 + 0.i dB. The inputand interstagenetworksare on-chipwhile
the output impedance matching network includes a bond wire inductor. This
last feature is incorporated to facilitate initial rf circuit evaluation.
Developmentand testingof thepassiveswitchedattenuatorport_onof the gain -
controlis beingdonein parallelwith theinitialphaseshifterdevelopment
discussedearlier.Subsequentiterationsof thegaincontrolwill includethe
passiveattenuatoron thesamechip. 4
1.1.4 RF/IF Submod_le
TheRF/IFsubmodulehas threefunctions:
o Convertthe 27.5 to 30 GHz inputRF signalto the IF (5.5GHz) in a
mixer.
o Amplify an external reference signal to provide a local oscillator
12
input(22GHz) to themixer.
-- o Amplifyand bufferthemixerIFoutputto drivethe IF amplifiers.
The technicalapproachesforachievingthesefunctionsare discussedbelow.
BalancedSchottkvDiodeMixer- Our baselinemixerdesignapproachwill be the
balancedSchottkydiodemixerwitha rat racehybridwhichis shownintegrated
with the other elementsof the RF/IF submodulein Figure 12. The input
signal,insertedon the rightsideof thechip,is shiftedin phaseby 90° to
one mixer diode, and 2700 to the other. The diodes are connected 180° out of
phase to suppress noise from the local oscillator amplifier and reference
signal, while combining the IF signals in phase. The diodes are connected to
IF frequency and dc ground by vias through the substrate; open circuited stubs
are used to ensure a high quality ground for the signal and local oscillator.
The mixer employs a pair of surface oriented Schottky barrier diodes so that
°- fabricationis compatiblewith that o_ the FETs. Effortsto developsuch
diodeshavebeensuccessfulin thepast. Thebestdiodes,usingthickactive
layers,have cutoff frequenciesgreaterthan200GHz [8,9]. Recentresults
[9]wereobtainedusingmany smalldiodes in parallelto lower the overallF.
seriesresistance.The diodesto be usedin thisprogramwillbe an extension
- of thesetechniquesas indicatedin Figure13. By decreasingthecapacitance
of the individualdiodeswhilemaintainingan approximatelyconstantvalueof
seriesresistancethesemethodsshouldbe usefulat 30 GHz.
22 GHz LO _mplifier- The requirementsforthe 22 GHz localoscillatorinclude
a gain of 24 dB and a noise figure of 5 ds. Other than frequency,the
requirementsare similarto theRF low noise amplifierused for the "front
end" of the receiver. Consequently the i00 micron gate width FETs discussed
in Section II-1.2.2 which are designed for the RF LNA are also used for the LO
amplifier.
m
To assess the noise figure and gain trade-offs for the i00 micron FET at 22
GHz, the noise figure and gain circles were calculated and plotted on the
Smith Chart as shown in Figure 14. The 3 stage amplifier shown in Figure 1
and 15 is designed to achieve the required 24 dB total gain and yet realize
the source impedance (e.g., z s = .6 + j 1.2) as seen by the first stage
especially, to be in the region of minimum noise using Figure 14 as a
guideline.
The theoreticalcalculationof the total gain over the frequencyrange from 20
GHz to 24 GHz is sh(_n in Figure15, and is about 26 dB. The overall gain
flatnessis within 0.3 dB, and the layoutestimatedarea about 2.44 x .53ram2. -
-Tne IF amplifier is primarily needed to buffer the mixer output
to drive lower impedance transmission lines. The requirements include 5 dB
gain from 5.5 to 8 GHz. A preliminary design using a 150 micron FET
equivalent circuit in a single stage common source configuration was
developed. The circuit design makes use of the so-called "lossy" match
technique, i.e., frequency dependent resistive loss is used to obtain flat
gain over a 60% fractional bandwidth. A gain of about 6 dB is achieved from 5
to 9 GHz.
14
1.1.5 _0 GHz Low NoiseAmplifierSubmodule
Five stagesof GaAs FET amplificationare usedin theLNA. The firstthree
stagesare designedfora noisefigureof 4 d8 and an associatedgainof 6 dB,
while the last two stagesare budgetedfor slightlyhighergainsand noise
figuresas was shownpreviouslyin Figurei. The numberof LNA stagesis
chosenlarge enoughto achievesufficientfront end gain to minimizethe
effectsof noise figurecontributionsand lossesof followingsubmodule
-- componentson the totalnoisefigureof thereceivemodule,whichis to be 5
dS.
Because of thelow dc powerconsumptionbudget(3Vand20 mA for the complete
LNA)and the theoreticallypredictedimprovementof devicenoise figurewith
smallergatewidths,theFETsusedin theLNA havegatewidthsof i00microns
(andsub-half-microngatelengthsas discussedin SectionII-1.2.2).Although
pk even smallergate widths are desirablefroman efficiencyandnoisefigure
point of view, a lowerlimit on gate width exists becauseof impedanceL
matchingconsiderations.In particular,too smalla gatewidthresultsin
high impedancelevelswhichare difficultto matchto 50 ohms (especiallyfor
the FET outputcircuit). This is due to designconstraintsimposedby the
.... physicallimitationsof monolithicallyfabricatedreactivecircuitelements.
The most direct consequenceof thisdifficultyisbandwidthreductiondue to
-- largeimpedancetransformationratios.
An initialdesignfor thefivestageamplifieris shownin Figure16. TheFET
equivalentcircuitdiscussedin SectionII-1.2.2is usedforthe FET devices.
15
As shown in the figure the gain is the required32 d8 _+.5dB from27 to 31
GHz. Inputimpedancematchingtakes into accountthe gain/noisefigure
trade-off(seeSectionII-1.2.2)especiallyfor thefirststage. As shown in
the figurethe designutilizessimilarcircuitelementswhichare relatively
easyto realizemonolithicallyon a 0.15rsnthicksubstrate.
It shouldbe noted thatthe designactivityandresultingcircuitschematics
presentedabove are strictlystartingpointsfor the developmentof the
various receiverfunctions.Theyserveto definethedesignmethodologyfor
subsequentdesigniterations.As experimentaldata is obtainedthe various
equivalent circuits will be up-dated and revised for final design
implementation.
1.2
Duringthe first year devicedesigncenteredaroundthelow noisei00micron -
gatewidthFET,and the 400micronswitchingFET for implementationin the
passive phase shifters. For theswitchingFET the primarygoalis to reduce
thechannelresistanceand resistiveparasiticsof the deviceto minimizethe--4
insertionloss of the 30 GHz SPDT switchesusedin the switchedlinephase
shifters.For the100micronFETsthegoalis to achievea deviceperformance
of at least 4 dB noisefigurewith an associatedgainof 6 dB at 30 GHz. As
shownin thissectionit is expectedthatsub-half-microngatelengthswillbe
required.
1.2.1 SAGFET forPhaseShifter
16
By usinga n+ implantself-alignedFET [i0]switchin the phaseshifterstwo
objectivescanbe achieved:First,realizationof a lowvalueof ON resistance
for minimizing insertion loss, and second,low power consumptionfor
minimizingtheoverallDC powerconsumptionof thereceivemodule. The latter
objectiveis met becausethe switchingFETsare basically"unbiased"FETs,
whereonlya controlvoltageis appliedto thegateelectrodes.
The principlefactor limitingthe reductionof the "ON"resistanceof the
self-alignedstructureis thesheetresistanceof thechannel.The calculated
sheetresistanceunderthe gateas a functionof gate-to-channelvoltagefora
relativelyheavyimplantof 5 x 1012siliconatc_s/cm2is shownin Figure17.
Thiscalculationincludestheeffectof thevaryingdepletiondepthunder the
gate. The maximum electricfield in the device for completedepletion
(pinchoff)is approximately480Kv/cm,assuminga Gaussiandopingprofile.
The sheet resistanceshownwas calculatedusinga mobilityof 4000cm2/v-sec
independentof doping. Increasingthe implantenergydecreasesthe channel
_ sheet resistanceat the price of a largerpinchoffvoltage.Somereduction
canbe achievedby forwardbiasingthegatebut the sheet resistancevalves
-- shownin Figure17 are probablycloseto thebestwhichcanbe achieved.
_r
Combiningthechannelresistancewith ohmic contacts (lineresistance)and
self-alignedn+ regionthe source-drainresistanceshownin Figure18 is
calculated. The resistancegivenin thisfigureis conservativein thatn+
sheetresistancesof 100 ohms/squareand lineresistanceslessthan0.iohm-_m
17
are expectedto be achievedin an optim_-nSAGprocess.For example,witha 5
source-drainresistance(theON state)it is expectedthatthe insertionloss
per bitshouldbe between1.5 and 2 dS.
1.2.2 i00MicronLow NoiseF_T
The I00microngatewidth.25microngatelengthlow noiseFET is thecritical
elementin the low noise receiverfrontend. This sectiondescribesthe
design of the i00 micron gate width FET. The description of the FET design is
separated into three subsections: I) Rationale for Low Nois_ FET EW$i_n which
describes the basic reasoning used to determine the requirement for a .25
micron gate length and i00 micron gate width, 2) Low Noise FET Equivalent
which describes the calculations used to arrive at an equivalent
circuitfor theFETand 3) GainandNoiseCircleCalculat_on_whichdescribes
the calculationof tradeoffsin circuitmatchingforgainand noisefigure
basedon theequivalentcircuitmodel.
RationaleforLow NoiseFETDesign- The FET requiredfor use in the lownoise
amplifiermust have a 4.0 dB noise figurewith a 6 dB associatedgainper
stageto achieveoverallreceivemodulenoisefigureof 5 dB.
The devicedesignmustconsidermaterialand geometricparametersthatimpact
thefollowingrequirements:
o Low noisefigure
o Highfrequencygain
o Low powerconsumption
18
Devicedesign requirementsfor a low noise figurecan be identifiedfrom
-- Fukui'snoiseequation[ii,12]which is known to give a good relationship
betweendeviceparametersandan experimentallymeasurednoisefigure.
r i0oi011E3- w- o. w2hL i12 (I)
1.8Ls- /0.18r_ 1/211/2
where
F = noisefigure(dB)
L = gatelength(microns)
k = noisecoefficientdependenton material(typically0.033to 0.040)
Lsg = gateto sourcespacing(microns)
rc = specificcontactresistance(10-6 ohm-om2)r-
P = gatemetalresistivity(10-6ohm-ore2)
- w = unitgatewidth(nm0
a = channelthicknessbeneaththe gate (microns)
- aI = channelthicknessbeneaththesource(microns)
a2 = channelthicknessin region2 (microns)
N = carrierconcentrationin theactivechannel(1016cm-3)
_ N1 = effectivecarrierconcentrationbeneaththesource(1016cm-3)
N2 = effectivecarrierconcentrationin region2 (1016cm-3)
h = gatemetalthickness(microns)
19
f = frequency (GHz)
The various thickness regions (aI, a2) , identified in Figure 19 are due to
recessetching.Surfacedepletionmustbe takenintoaccountwhen determining
a2. The firsttwo termsin bracketsin equation(i)are dueto thegatemetal
resistance.The firsttermis thedc gateresistance;thesecondterm is due
to skin effect. Term three is due to the material resistance in region two.
The fourth term is the contact resistance. Equation (i) can also be written
F = i0 lOgl0 [i + C(Rs + Rg)I/2] (2)
where Rs is the source resistance, Rg is the gate resistance, and C is a
constant dependent on material, frequency, gate length, doping density, and
channel layer thickness.
Examiningequation (i) shows that the devicerequirementsfora low noise -
figureare:
o Gate lengthshouldbe as shortas possible.
o Gate metal resistanceshouldbe low so Rg <.Rs. Thisimplieslow
metalresistivityandthickgatemetal.
o Source resistance should be as low as possible. This implies low
contact resistance and increased doping density and thickness in the
regions between the source and gate.
Requirements for high gain can be identified from the equation for the
unilateral gain, Gu, of a FET [13].
20
Gu= [gm/(2zrCgs)] 2 1(2f)2 _ds(Rg+Rs)+2zrftRgCdg] (3)
where
gm = transconductance
Cgs = gate-sourcecapacitance
Cdg= drain-gatecapacitance
Gds = outputconductance
ft = unitycurrentgainfrequencygm/(2_ Cgs)
Tne transconductanceandgate-sourcecapacitancecanbe expressedin terms of
materialand geometricparameters.
So that
gm Vs
2_Cgs 2"_"L"
where
vs = saturationvelocityof electronsin the channel
and
Gu = s . 1 . (4)
_. 16zr2L2 ds(Rg+Rs) RgCdg
Equation (4) indicatesthat some of the requirementsforhighgainare the
same as for a low noise figure,i.e.,gate length shouldbe as shortas
possibleandgate and sourceresistancesshouldbelow. In addition,bothGds
and Cdg shouldbe madeas low as possible.
21
Powerconsumptionis another considerationin the FET design for this
application. The power consumptionis given by the productof the drain
sourcevoltageand the draincurrent.A reasonablevaluefor thedrainsource
voltageis 3 volts. If the draincurrentis 4 milliamps,the Idss of theFET
will be I0 timesthe low noise bias value, or 40 milliamps. _e Idss is
related to material and geometric parameters by
Idss! q N vs Z(a-d)
whenthe FEToperatesin thevelocitysaturatedmode.
where
d = zerobiasdepletionwidth= 0.067micron
a = channel thickness = 0.15 microns -
vs = 107 cm/sec
N = 2.5 x 1017 cm-3
So the total gate width Z must be:
Z S 120 microns
Since 120 microns is too long for a unit gate width, the gate will be broken
up into two sections. We use a i00 micron total gate width, with a unit gate
width of 50 microns. This gate widthwill givea reasonableimpedancefor
22
matching networks.
Combining the requirements for low noise, high gain and low power consumption
gives the FET constraints that impact the final design. Tne gain is increased
and the noise figure is reduced by minimizing the gate length.
The factorswhichlimittheamountof gatelengthreductionare:
o E-beamlithographyprocessinglimitsforreproducibility.
o Channel thickness and doping control implied by channel dimensions for
best performance.
Consideration of these factors with present processing technology makes a 0.25
micron gate,a channellayer 0.15 micronthick,and 2.5x 1017cm-3 doping
reasonablechoicesfor thedevicestructure.
Gain and noise figureare also enhancedby minimizingthegateand source
resistance.The gate resistancecan be reducedby using increasedmetal
thickness and low resistancemetalsto fabricatethe gate. The source
resistanceismost stronglyaffectedby the dopingbetweenthe sourceand
_ gate, the geometryof this regionandohmiccontactquality. Initially,we
will usea source-gatespacingof 1 micronto easealignmentrequirements,but
-- it is expectedthat advancesin technologywill permit reductionof this
dimension.Othertechniqueswhichwill be employedto minimizethe source
resistanceare (i)recessedgate, (2) optimizeddopingprofilesand (3)n+
surfacelayerswithadvancedcontactingmetallurgyforoptimumohmic contact
characteristics.
FET designparametersanda calculatednoisefigurebasedon thisdiscussion
are givenin Table2.
A final design considerationis the output conductanceand drain gate
capacitanceof the FET. Althougha formulafor the output conductancewas
developedby Pucel,Haus,andStatz[14],it predictsa Gds muchlowerthanis
actuallymeasuredon FETs. It is believedthisis dueto substrateconduction
[15, 16] near the high field region. In an ion implantedFET,the output
conductanceis undoubtedlyinfluencedby implant distribution and the
substratequality,althoughthe outputconductancecan be decreasedby theuse
of an AlxGal_xAsbufferlayer[16]or an insulatingregionbetweenthechannel
layerand substrate[17]. We do not believe the increasein processing
complexityis warranted.
FET performanceis influencedby a complexrelationshipbetweenequivalent
circuitelementvalues that are affected by materials and processing -
limitations. Therefore,the final FET designwill be achievedthrough
experimentalmeasurementsand computeraidedmodeling.
Low Noise FET EquivalentCirCuit- As discussedin the preceedingsection,a
i00 micron gate width FET with a .25 micron gate length is required to meet
the gain and noise figure goals for the rf receiver. To design the multistage
amplifier an equivalent circuit model for the FET is needed. This section
will describe the equivalent circuit model for the FET and the procedure used
to arrive at the element values used in the model.
24
The equivalent circuit model for the FET is shown in Figure 19b. Some of the
-- element values are calculated from theory while others are determined by
measurement or scaling. The FET gate width is chosen to be I00 microns to
ease input matching requirements and minimize the effects of phase shift of
the 30 GHz input signal along the gate.
_ The circuitelementsthatcanbe calculatedfrommaterialparametersand
device geometryare: the gate resistanceRg, the inputcapacitanceCgs,the
-- transconductancegm, the sourceresistanceRs, thedrainresistanceRc, and
thephaseshiftfactorr . These elementvalueswere calculatedfrom the
followingequations:
pW --
Rg = _- 2. 7fl
.- Rs = Rc + Rsg = 5.4-fZ
Rc = _= 1.8_Z
- Ps2Lsg - 3.6_2Rsg - Z
25
PS1 = sheetresistanceof semiconductormaterialbeneaththe ohmic
m
_ I = 325ohmsper squareNq_aI
q = electroncharge= 1.6x 10-19coulcmbs
= electronmobility= 3500cm/v-sec
p_ = sheetresistanceof materialin source-gateopeningexcluding _1
- - 476ohmsper squaresurfacedepletedregion Nq#a2
Rd = Rc + Rdg = 11.3ohms
Rc=_Pc_ 1.8ohmsz
Ps2LdgRdg= Z =9.5ohms
Ldg = 2 microns
(,o,,qN 1/2
eo = 8.854x 10-14 fd/cm
26
€r = 12.5
#hi= builtin potential= .8V
_ Vg = appliedgatevoltage= 0V
_ L = Leff = effectivegatelength= physicalgatelengthplus
activelayerthicknessto correctfor fringingcapacitance=
0.4microns
gm = VsZ qNe°er']= 16.9mmho_2(q_bi-Vg)J
VS = Saturationvelocity= 107 cm/sec
L_"=--=2.5psec
_ V s
The remainingequivalentcircuitelements:thedrain-gatecapacitanceCdg, the
drain-sourcecapacitanceCds, the resistanceRi andthe outputresistanceRo
cannotbe accuratelymodeled.Our approachis to useexistingpublicationson
quarter micron gate FETs [18-20]togetherwith COMPACTanalysisof the
equivalentcircuitto estimatevaluesfortheseelements.For theseelements
we havechosenthevalues:
Ro = 700 ohns
27
qg-.01pf-.04
Ri= 6 ohms
The complete equivalent circuittogetherwith a tableof criticalFET
parametersare givenin Figure19. The calculatedmaximumavailablegainis 8
dB at 32 GHz. Fukui'snoise equationpredictsa noisefigureof 2.5dB for
the FETmodel. All designequationsarebasedon a uniformdopingmodelwhich
is typicalfor FETsmadewithVPE material.The useof ionimplantedmaterial
shouldnot significantlychangethesevalues.
Gain and Noise Circle CalculatioDs- UsingCOMPACTtheS parametersfor the
FET equivalentcircuitwere calculatedas givenin Table3. Noiseparameters
for the FETcan be calculatedaccordingto a paperby Fukui[21,22] using:
Fmin = 1 + .016f Cgs _g+ Rs
• gm
Rn = 0.8/gm
Xop = 160/(f Cgs)
Rop = 2.2 [4-_m+Rg+RsJ
where
f = frequencyinGHz
28
Cgs = capacitancein pf
gm = transconductancein ohms
Rg = gateresistancein ohms
Rs = sourceresistancein ohms
Fukui'smodel, on which thesecalculationsare based,doesnot considerthe
effectof sourceinductanceor gatedraincapacitanceon highfrequencynoise
figure. The effectof the feedbackelementson noisewere calculatedusinga
programfor theHP41Cwritten by Suter [23]. Gain circlesare calculated
usinganotherHP41Cprogram,GCIR.
Using these HP41Cprogramsthe gainand noisecirclesfor the FETinputand
outputplanescanbe calculated.For theFukuinoiseanalysisthe equivalent
circuit does not includeCdgor Ls. By usingtheS parametersfor themodel
with no Cdg or 5s (Table4a) with no CdgbutwithLs, (Table4b),andwith
both Ls and Cdg (Table5), the noise circlesfor the input plane can be
calculatedas shown in Figures20 and 21. These figures illustratethe
tradeoffbetweengain and noisefigure. Thenoisefigurereductionfor the
- case includingCdg andLs may be in errorsincethecalculationmodelassumes
Cdg is connectedin shuntfromthe inputterminalto the outputterminal.In
-- any case,the modellinggivesa good indicationof the gain-noisefigure
tradeoffsince there is not a lar°gechangein the gain and noise figure
circles. In a similarway the gain circlesfor the FET outputplane are
calculatedandthe resultsare shownin Figure22.
A finalitemof interestthatcanbe calculatedis thevariationof gain and
noise figure with changesin sourceand gate resistance. This type of
calculation has been used by Yamasaki and Schellenberg [24] to show that the
gate resistance has a stronger influence on high frequency gain that the
source resistance. Since the maximumavailablegain Gmax is affectedby
stabilityconsiderationstheyuse theunilateralgain,Gu, forcomparison.We
havecalculatedthe effectof gateand sourceresistanceon maximumavailable
gain and noise figureandplottedthe resultsas shownin Figures23 and 24.
Themaximumavailablegainwas calculatedusingCDMPACTand the noise figure
using Fukui's noise equation. These graphs indicate that both source and gate
resistances have a significant affect on gain and noise figure. High source
resistance can cause the FET to become unstable. In summary, these
calculations of gain and noise circles based on a model for a .25 micron gate
length, i00 micron gate width, GaAs FET indicates that the noise figure and
gain requirements for the front end of the NASA low noise receiver can be met
with reasonable matching networks.
30
2.0 l_aseShifterFabricationand Evaluation- TaskII
During the first year results were obtained on two phase shifter mask set
designs; the single 180o bit phase shifter using shunt FETs and a more
comprehensive mask set that includes multi-bit phase shifters using series FET
switches. Two fabrication approaches were also tested; one uses standard
-- power FET fabrication and a second approach employs a new self-aligned gate
(SAG) fabrication technique for reduced parasitic resistive losses in the
switching FETs. Much of the work, which is still incomplete, has been to
develop a reproducible and reliable SAG fabrication process. Nevertheless,
preliminary RF results have been obtained on SAG phase shifters which have
_ been sufficiently encouraging to commit our final design and fabrication
approach to the SAG technique.
To put theSAG resultsin perspectivewe presenttypicalphaseshifterresults
for the 180o bit using conventionallyfabricateddevices. For one run,
typicalFET ON resistance,Ro'n, as measuredon the curve tracer,ranged
between 12.5 and 15.5 ohms on dc good phase shifterchips. For Circuit
_ modelinga value of Ron = 14_ , a residualseriesresistance,r = 7_ ,
and a pinch-offcapacitanceof .07pF was usedto comparewith the measured
-- results [2]. Figures25,26aand 26b showthemeasuredinsertionphase,the
rf testfixture(coverremoved),and themeasuredinsertionloss (including
1.4 dB fixtureloss)respectivelyforone of the phaseshifterchips.
The RF test fixturehas in-lineWR-28 waveguideinput and outputports.
Antipodelfinlinetransitionsfabricatedon 0.25 mm thick RT/duroidare
employed to transition from waveguide to microstrip. The 1.4 d8 fixture loss
includes mismatch loss ( .4-.6 dB) as well as dissipative loss in the total
circuit comprising the transitions and interfaces to the GaAs chip. Figure 27
shows the phase data plotted together with the calculated results. Note the --
calculated results are in closer agreement with the measurements when a 0.i pF
shunt capacitance is included for each of the four FETs in the circuit model
(see reference [2] for details). This parasitic capacitance is associated
with the layout metallization of the FETs and can be decreased as will be
shown below in the series FET switch design.
2.1 Results for 180o SAG Phase Shifter Using Shunt FET_
Figure 28a shows the encouraging results obtained for the insertion loss for
the two phase states across the 27.5 to 32.5 GHz band. As seen from the CRT
traces, the insertion loss including test fixture loss is between 4 and 5 dB
in the band of interest, 27.5 to 30 GHz. This is in contrast to 6 to 8 dB
loss for the corresponding measurements made on the conventional phase shifter
chip as was shown above. Figure 29 shows the comparison between the
calculated and measured results including the 1.4 d8 test fixture correction.
As shown in the figure the calculated values are based on the values of the
FET parameters measured off test FETs on the SAG wafer. The reasonable
agreement obtained justifies the use of our simple equivalent circuit model
used for the ON and OFF FETs. Additionally, it should be noted that this
wafer demonstrated a limited amount of surface conduction (conversion) which
could account for some of the discrepency between measured and calculated
values as the circuit model does not include such effects.
Finally,Figure28b shows the measureddifferentialinsertionphasefor the
SAG phase shifter,from 27.5 to 30 GHz. Overthisbandthephaseis within
20o of the calculatedcurve. Note,again,the tendencyfor thedifferential
phaseto decreasetowardstheupperhalfof theband.
-- 2.2 Results From Second Mask Set Using SAG Series Switches
A comprehensivemask set has been developedthat includesmulti-bitphase
shifters.Figure30 showsphotomicrographsof someof the circuitpatternson
the finishedwafers. Figure 30a shows the four bit phaseshifternetwork
_ (45o,90°, 180°, 22°) while Figure30b is an enlargedviewof the 90 bit.
These phase shiftersemploy 400 microngate width series FETs for the
switchingdevices.The 180°, 90° and 45° bitsare switchedlinetypes,while
the 22° bit is of the loadedlinetype. To date,one SAG fabricationrunhas
beencompleted.As seenin thefollowing,theRF insertionphaseresultsfor
the individualbits (180o, 90° and 45o)supportthepredictedadvantageof the
seriesswitchcircuitdesign. Unfortunatelyour firstrun had low yield and
_ did not attainthe low resistancevalues (<8_)we were hopingfordue to
severalfabricationrelatedproblems.Thesearepresentlybeingcorrectedfor
"- the secondrun. Nevertheless,phase shiftdataobtainedon the individual
bitsis quiteencouraging.
Figures 31-33 show the measured differential phase for the three bits across
the 27.5-32.5 GHz band and the comparison with the ideal time delay
characteristics. The measured data is taken with the H.P. network analyzer
usingthe 26-40GHzwaveguidereflection-transmissiontest set. Although
displaceddownwardsby about70,themeasureddatafollowscloselythedesired
slopeof thecharacteristic.Notetheimprovementfrompreviousphasedata
using the shunt design. This improvementis primarilydue to the more
optimizedlayoutand the reductionin reactiveparasitics.Figures34 through
36 showthe actualmeasureddatafor theinsertiondifferentialphase,the
insertion loss, and the return loss respectively from 27.5 to 32.5 GHz for the
180 ° and 45o bits. In Figure 35 the insertion loss includes the fixture loss
of about 1.4 dB and is better than the results obtained with conventionally
fabricated FETs but worse than the best SAG result obtained on the 180o bit
using shunt FETs.
34
3.0 Gain Control Fabrication - Task III
The maskdesignfor thefirstiterationof thegaincontrolis complete. The
layoutcontainsthe followingcircuitsand devices:
o i00microngatewidthlownoiseFET.
o Low noiseFETwithstubinputmatching.
-- o Low noiseFETwith capacitormatching
o Dualgate100microngatewidthFET.
o DualgateFETwith capacitormatching.
o Two stagedualgateamplifierforgaincontrol.
o Testpatternsfor in-processtesting.
Eachdeviceor circuitis expandedintoan arrayor reticleof devicesin the
" actual layout. For example,the i00microngatewidthFETdeviceis expanded
into a 4.87 by 4.87 mm arraycontaining198 FETs and 1 test pattern for
in-processtesting. EachFETdie is 300_m by300#m. The layoutfor the
_-- singlegateFETis shownin Figure 37. A secondexample is the two-sta°ge
amplifier.
The two-stageamplifierreticleoccupiesthe sameareaas the the100micron
FETreticle.The two-stageamplifierreticlecontainsonly twelvetwo-stage
amplifiersandone in-processtestpattern.The two-stageamplifiershownin
Figure38,has a 1.9_nx .65ramchipsize. Thesingleand dualgate FETs with
Capacitormatchingare combinedin a singlereticleso thereis a totalof
fiveseparatereticles.Two of the reticles,thei00micronFET and the 100
- micron FET with stub inputmatching,do not requirenitrideor airbridge
-- 35
processing.The remainingthreereticlesall requirenitrideand air bridge
processing. Since the processingwithout nitrideor air bridgesrequires
fewerstepsit makessenseto separatethesereticlesintotwo groups; i) the
100 micron FET group,and 2) the two-stageamplifiergroup. For the100
micronFETgroup all layersexcept the finalpad metal layer are done by
e-beam lithography. The two-stageamplifiergroup requiresfouroptical
levelsaftere-beamprocessing:I) thecircuitmetallevel,2) the dielectric
patterninglevel,3) the secondlevelmetaland 4) the airbridgelevel. The
opticalmasksfor theselevelswillbe madeon theCambridgee-beamsystem.
i-
3.1 i00 Micron FET
The first fabricationruns will includeonlythei00micronFETgroup. The
100microndeviceswill be testedto check the validityof the FET model
discussedin Section II-1.2.2. The initialrun will be fabricatedon ionr
implantedLEC substratematerial. Thematerialhas beenimplantedwith5 x
-- 1012ions/cm2of siliconat 120Kevand annealedfor 30 minutesat 850ocin an
arsenicoverpressure.The measuredsheet resistanceof the implantedand
annealedmaterialis 600 ohms per squarecomparedto a calculatedsheet
resistanceof 538ohmsper square. This indicatesthat the activationis
high, on the order of 90 percent. However,CV profiling (SeeFigureo39)
indicatesthatthepeakcarrierconcentrationis lower than the theoretical
valueand thereis somespreadingof the implantdistribution.
3.2 DualGateAmplifier
Thelayoutforthe dualgateFETamplifierhas beencompletedaccordingto the
designdiscussedin Section1.1.3. Fabricationof the dual gate amplifier
_ willbeginaftercompletionof thefirsti00micronFET runs.
4.0 RP/IFandIRASuhmoduleFabricationandEvaluation- TasksIVandV
TasksIV andV are scheduledto beginin the secondyearof theprogram.
Developmentof thelownoiseamplifierhasactuallystartedduringthe first
yearwith thedesignandinitialfabricationof theI00microngatewidthFET
includedin the gaincontrolmaskset (discussedin SectionII-1.2.2and
II-3.1).
5.0 SubmoduleIntegrationandReceiveModuleDevelopment- TasksVI andVII
Tasks VI and VII are scheduled to begin in the third and fourth year of the
program. The designs for both the integrated module and the monolithic module
was discussed in Section II-l.l.l.
6.0 ProductAssuranceandReporting- TasksIXandX
A product assurance program has been implemented in accordance with the
requirements of Section 3.4 of the RFP. Log books concerning device and
circuit fabrication and development have been utilized from the outset of the
program.
Reports have included an updated work plan, monthly technical progress
reports, as well as monthly and quarterly financial and management reports
(533M, 533Q, 533P) as required under the contract.
38
IlL LOGIC SIGNALS
tFOR DIGITAL CONTROLI
- " I I 01GITAL CONTROL iI - 18 d_ PHASE REFERENCE SIGNAL__'-AI 22 GHt
,r !. V ,,, rG,,,NCON,,oLV,,VLO." ,,..,,I LOW NOISE AMPLIFIER I I PHASE SHIF I I --
" I I I I I L ._6d13m I
'"++ 't>"_"_"_oI I I I ss 8.0GH! I,--. ,.J " 1-- .... .J L _j GH,
--" LO DIODE
5OR 6$TAGE FET AMP PS LOSS DUAL GATE AM__P MIXER IJFAMP
NF 4d81 4 4 4 6 _ _ 14 _ IO 5 6 4
GAIN ICiBJ 6 6 6 7 ; "4_ -I _ 12 24 -6 S
VOLYAGE IVl 3 3 3 3 3 4 3 3
CUqREN! ImAI 4 4 4 4 4 I S 26 8
_" SUfMODULE F _ _I.B f_.. 14Od6 .fo" I_-- f _ f_ _m 10dlB
NOISE FIGURE F_, m 10dB Gm
:_= 48d6MONOt I ! HIC Hff C| IV[ MODUt t
rO|AI I IC POW£R 250 mw IINCtUIIING g mW I-UH DI(;IIAL CONIHOLI
_" OVERALL NOISE FIGURE 4 9 dB IMAX GAINI _) 0 d_ IMIN GAINI
NFIIF CONVENSION GAIN 3_ tJtl Ifl_ll _l_ll ;)3 _ lldl|N GAINI
MINIMUM DE IEC | ABt E NIGNAL /0 lll_t_
OYNAMI(: RANGE 30 _IH IMAX (_AINI 43 (ll_ IMIN (;AIN,
Figure1.BlockDiagramofOverallMonolithicReceiver
LO WAVEGUIDE (WR-.42)"_ INPUT
IFIF OUTPUT OUTPUT
_ . (SMA) LO INPUT
TTLINPUT RF/IFCONNECTOR
,,...-.
IAS DISTR.
METAL.... DC BIAS CARRIER
GAIN PLATERF INPUT CONTROL
RF WAVEGUIDE (WR-28)INPUT LNA
PHASESHIFTER
RF TEST FIXTURE INTERCONNECTED RECEIVE MODULE
-- Figure2.InterconnectedReceiveModuleandRF TestFixture
39
' ON =ET OFF FET
Near Resonance
_ = C=r,Qc= (_r,C:,)-'_o = Center Band RadianFrequencyL - (_0_,)-'. = External Inductor
Switch Figureof Merit, (_;
ROFF Example:(_I10GH,= 738--__--[(oaCoFF)2rsRo.]-' --
Ro. (_l_., = 82
Figure 3. Simplified Equivalent Circuits for ON and OFF FETs
F_ = 4.7t_ 1
_o r, = 2.49 _'.... _°
C = .07pF| For300 Micron)- FET and ScaJed
"_ --_-R° f f 1RL: R° '_/ or'R"= 14H"_==59 _,_' _ forOtherGatewidths -_ _ _._ .O7pI-J J
_.......-.--,
mm ._---_ ----_ 30
"o_ 0.5 - 4-_-_ _ _''_ .... _ _ _ _ _ _ -25 U_o
O / _ --"- O.J 1.0 I / _ --20 :
.9 __--_-----_ "_•:: 1.5 - _ --15 m
= 2.0-- i
-- 10
r2. 5 I I I I I I I I 5150 300 450 600 750 900 1050 1200
FETGateWidth (Microns)
(_-.g/4 = 855 Microns for 50_ Line at 30 GHz)
Figure 4. Insertion Loss and Isolation versus Gate Width
4O
-- Figure 5. Chip Layout for 180 ° Bit
400 Micron FETs
Gate Pad
- IN [ ] OUT
r
0 005"
-- Figure 6. Layout of 180 ° Bit Using Four Series FET Switches
190 -
180 _ Calculate
I
170 -- l / / " "
_ / //__Tdeal curve for 180o bit160 /
/
150 -
,140 I I I I I I I I26 27 28 29 30 31 32 33 34
Frequency (GHz)
Figure 7. Computer Calculated Results for 180 ° Phase Bit Using Series Switches
1 ', I ,' , I ' ) ( / ( , I
...1I"
..... Ideal Characteristic _.._ "• CalculatedPoints ,.,,- "
...." -- 180 °
_} 160 .
a [40 135° -----
C !.20 ----_".£
---
-- 1.oo 90° --J'---
,-,--- .__._
60
450 __l.- .......
40q, "--
I I I I I I ' i'5 "_, ? 1 2R 29 "_J 31. _2 3_
Frequency (GHz)
f360 /
//
.//.
,/
315 ° .../_" 32o
O / / /300 --" /"
a,, //" I /
" " 2_'o / 270o. I"._o
_60 / _ j_
/I( _
• ' 240 _ _"
i5 225 ° _. f "te"
• 220 . _
200 - "
Frequency (GHz)
Figure 8. Calculated Differential Insertion Phase for 3-Bit Phase Shifter(180°, 90°, 45° Bits) UsingSeriesFET Switches
-- 4.3
0 AssumedValues R= = 99forSwitching Co.= .09pFFET Parameters r, = 4.5fl
2
I I I I I I I -25 26 27 28 29 30 31 32 33
Frequency (GHz)
Figure 9. Calculated Insertion Loss "Envelope" for Eight States of 3-Bit PhaseShifter
UNBIASED ,Ser,'t sGATE 2 CONTROL VOLTAGES FETS
CORRESPONDINGOVERALL RECEIVER
GAIN STATES (dB) RF/IF GAIN (riB)
q, 3+2-6= -1 17IN OUT (_ 3 + 5 - 6 = +2 20
(3) 3 + 5 - 2 = +6 24(4) 6+5-2= +9 27
-2dB ('_ 6 + 8 - 2 = +12 30GATE GATE _ ~15 -15 - 6 =FET FET - 36 OFF
GAIN STATES (G) (G2) WORSI CASE NOISE FIGURE (STATE (1))(dB) 3,_ 2,5,8
F2-1 L=-IASSOCIATED FMAX = F1 + + - 13.6dB
G1 GIG 2 ---NOISE FIGURE (F1) (F2)14,10,10
(dB) 10,9 BEST CASE NOISE FIGURE (STATE (_)
VOLTAGE(V) 4 4 FMIN = 10.1
MAX. CURRE'NT(mA) 8
Figure 10. Gain Control Design
44
509 52° 87.5 ° 47.3 ° 1 PF 1.5 nH
TF; o ;°°AllTRLshave_ = 90922 -ReferenceFrequency = 29 GHz
_. 21 AllFETs = 1O0/.EnGateWidth
_ _ 19,
18-
17 -
16-
- 15 I I I I I I26 27 28 29 30 31 32
Frequency(GHz)
Figure 11. Design for Monolithic 2-Stage Dual-Gate Amplifier
I _., 2.7mm ..._1
,-- I- {] -I
, ' lI LO REFAMP I
_" "1 I I , _.1AMP I I I viA"7
I I I t"_" L._. 2.1me
IF MIXER DIODESFILTER
I _1 SIGNAL INPUT./VIA
F -
RAT RACEMIXER
-- Figure 12. RF/IF Submodule with Rat Race Hybrid
45
ws
OHMICCONTACTLEAD
x,..SCHOTTKY CONTACTLEAD _
SCHOTTKYCONTACT _-
DIELECTRIC
OHMIC ""CONTACT,
o- n-GaA:
J ;aAsOHMIC ;.I. GaAs
DEPLETIONSCHOTTKY REGIONCONTACTTHROUGH
DIELECTRIC OF SELECTIVEIMPLANT _
SCHOTTKY LEAD ON TOFOF DIELECTRIC
Figure 13. Diode Structure for High Mixer Performance and Fabrication Simplicity . _
46
! _ k! T _ _ _ _ ! _ J 1 / _ l \
-- NoiseFigure SourceimpedanceZ,-- - Gain
Figure 14. Noise Figure and Gain Circlesfor First Stage of 22 GHz Amplifier
62.4 ° 51.6 ° 1PF • 73.1 ° 60.7 ° 1PF 51.2 °8zo __
7., AllTRUshaveZ,, = 909Reference Frequency = 22 GHz
28 AllFETs = 100#m Gate Widths
A 27
26 _ ,
c....-
25-
I I I I
20 21 22 23 24
Frequency(GHz)
Figure 15. Design for Monolithic 22 GHz Amplifier
--- Stage 2 ---- 3 4 --- 5 _ __L.
AllTRL'shave 7--0= 909Reference Frequency = 29 GHzAllBlockingCap's = 1PF
34 AllFETs 100#m Gate Width
_" 33 Chip Layout - 1.58 x 1.25 mm2
32 --
31
30
29
I I I I I I I25 26 27 28 29 30 31 32
Frequency(GHz)
Figure 16. Design for Monolithic Five-Stage Amplifier _ _
8
40.0 Dose 5xl 0'_Activation70°0
"- 200 /
.i- 10.0 /
/- d 5.0. , {,_
c /-_- 4.0•_ /i¢1
._ 2.0m
&.°;-/1.0 ///
J/
f
0.5 I I I0 -2 -4 -6 -8 -10
. Figure17. ChannelSheetResistanceversusGate-to-ChannelVoltage for Different Implant Energies
7
- - Une Resistance.2 £--mmn. SheetResistance 200 L)sGate Width400_
E Source--Drain_ o Spacing4/.{
6ue-
I 5
4500 600 700 800 900 1000
-- Figure 18. Source-DrainResistance versus Channel Sheet Resistance
49
a b 2.7 .01 11.3
o:: t,a2 _ _'--F'" N
• BUFFER _'_ .4 Uni ts
Gate length Lg = 0.25 microns gm = 16.9 e'J_mS _ fi _
Unit gate width w = 50 microns I = 2.5 ps pF
Channel thickness a = 0.15 microns 0.07 nH
Doping density N = 2.5 x 1017
Total qate width Z = 100 microns
Gate-source spacing Lsg = .75 microns
Gate metal thickness h = 0.5 microns _9.3dB @ 27GHzSpecific contact resistance rc = 10-6 ohm.cm2 MAG = (7.8dB @ 33GHz
aI = 0.22
a2 = 0.15
NI = 2.5 x 1017
N2 = 2.5 x 1017
Calculatednoise figure = 2.5 dB at 32 GHz
Figure 19. 100 x 0.25 Micron Low Noise FET Designa) Device parametersb) Equivalent Circuit
50
012 _I$Ost
0._II ,',. 5t
Conjugatematch '°8.6 dB gainat 29 GHz
8 dB7 dB6 dB
._ 5dB
NoiseFigureMinimumF = 2.25 dB .."Rn= 47.34 ohms£,:_t= .63 < 50.53
/Figure 20. Gain and Noise Circlesfor Low Noise FET Input Plane
- No Cdg orLs forNoiseCircles
012 0.15O_t
Conjugatematch °" °"8.6 dB gainat 29 GHz
8dB7 dB
.._".<: 6dB5dB
2.5 dB
NoiseFigureMinimumF = 2.06 dB
Rn = 33 ohmsropt= .58 < 64.43
;_,,/-t,,7-,_./ -7-4__iFigure 21. GainandNoise Figure Circlesfor Low Noise FET Input Plane
Including Cdg and Ls
! !
O.IZ C.15o J4
ConjugateMatch °" o.,. ....8.6 dBGain '0 LowNoiseat 29 GHz * ConjugateMatch
-t . 0 !I
R, = 5.4t'1 - 3.5F = 29 GHz "-
10.-
3.0
,,, £"o 9. --
0 2.5
8. m
2.0
7. --
I I I 1.52. 3. 4. 5. --
R, (_3)
Figure 23. Effect of Gate Resistance on 29 GHz Gain and Noise Figure
o Max.AvailableGain11.-
Rg= 2.7[2
F = 29 GHz o10.
O- 3.0
-rl
9. _ o.O3
2.5
8.
_ - 2.0
7. _
I I I I I I I I 1.52.5 3.0 4.0 5.0 6. 7. 8. 9. 10.
R, ([2)
Figure 24. Effect of Source Resistance on 29 GHz Gain and Noise Figure
54
\
State 1
State 2
27.5 (GHz) 30
Vertical: 450 IdivHorizontal: 500 MHzldiv
30 (GHz) 32.5
Figure 25. Insertion Phase for 1800 Phase Shiftet(TR061A)As Measured on the Network Analyzer
•• RT/duroidI Substrate
Part of~Fin-Line
Transition
Insertion Loss
Vertical: 5 dB/divHorizontal: 26.5 - 37 GHzReference: Center Line
(Test Fixture and Chip)
Figure 26. Test Fixture and Measured Insertion Loss for theTwo States of the 1800 Phase Shifter
J J J J J J J J 1 j
. . , . . . ,
/ IdealJ
f
J
200 / NoShuntCapacitance
190
°- _ lqO"o
O.1 pg ShuntCapacitance(4 plcs)- _. 170
160a) _ Calculated
_ Experiment
I I I I I I I I_ 26 27 28 29 30 31 32 33 34
Frequency (GHz)
Figure 27. CalculatedandMeasuredDifferentialPhase Shift- for 180°Phase Shifter (TR061A)
57
0100
Insertion Loss
Vertical: 5 dB/divReference: Center UneHorizontal: 500 MHz/div: 27.5 - 32.5 GHz
Differential Insertion Phase
Vertical: 45°/divHorizontal: 250 MHz/div; 27.5 - 30 GHz
Figure 28. Measured RF Results for SAG 1800 1-Bit Phase Shifter
) 1 ) } J 1 / ) f
I
0
Ron : 7.8_ Circuit _iodelFET Parameters
_- I - rs = 3.9_ (TR-O8-2ASAG)
2 _ ' C°ff = .07pF
A 3 • m_1mm"D •
.- _ 4 - State28.._ 5 . •
-6 Experiment*
- _ Calculated
m
*Corrected for 1.4 dB fixtureloss.
f
, I I I I I I I26 27 28 29 30 31 32 33
Frequency(GHz)
Figure 29. Comparison of Measuredand Calculated Insertion Lossfor SAG Phase Shifter
- 59
(a)
(b)
Figure 30. Photomicrographsof SAG Phase Shifter Circuitson a Processed Wafera) The four bit phaseshifter (45 °, 90°, 180°, 22°)b) Enlargedview of 90° bit
6O
f
6O
---- Ideal- _ 50
• 1.56°/GHz
_ 30-
-_ I I I I I I27 28 29 30 31 32 33
Frequency(GHz)
Figure 31. Calculated andMeasuredDifferential Phase Shift for 45° Phase Shifter
°__
Ideal100
3.13°/GHz
,. ](N 80 I"e-
_ .=_ ]_Q
zo [60 I I I I I I
27 28 29 30 31 32 33
Frequency(GHz)
- Figure 32. Calculatedand Measured Differential Phase Shift for 90° PhaseShifter
61
Ideal
200 -
J _t 6.26°/GHz
190
_) 180
._ 17o
i5 _
16O
150 I I ! I I _ I27 28 29 30 31 32 33
Frequency(GHz)
Figure 33. Calculated and Measured Differential Phase Shift for 180 ° Phase Shifter
62
Figure 34. 180° Phase Shifter Vertical -- 45°/div.Horizontal-- 27.5 ° 32.5 GHz, 500 MHz/div.
Figure 35. 45° Bit Insertion LossVertical = 5 dB/div.Horizontal = 27.5 - 32.5 GHzReference -- Center line
Figure 36. 45° Bit Return LossVertical = 10 dB/div.Horizontal -- 27.5 - 32.5 GHzReference -- Center line
63
SOUREE
DR AI N _
• !! iii;iiliiiii;i
SOUaCE x_ I _
Figure 37. CALMALayout for 100 Micron Gate Width Low Noise FET
,< 1.9 mm >
Figure 38. CALMALayout for First Two Stages of Gain Control Amplifier _.
64
4-.
3-
2 - Measured_ 850°C 30 min.E As
€-- ._o Si Implant
•,- 0_7 5xl 0_2--120KeV" lxl -r-O 8 -o Calculated--90%activationim '-2
L_L_
m
B
f
2 l I I I I0 0.5 .10 .15 .20 .25 .30
Depth_microns
Figure 39. Profile of Ion Implant
65
Table 1. NASA's Key PerformanceGoals
DesignParameter PerformanceGoal
RF Band 27.5-30GHz --IFCenterFrequency Between4-8GHzNoiseFigureatRoom Temperature <5 dBRF/IF Gain __30 dB at highest level of gain controlGain Control At least six levels; 30, 27, 24, 20, 17 dB
and Off.Phase Control 5 bits; each bit +3° at band centerModulePowerConsumption 250mW inallstatesexceptOFF. InOFF
state,25roW.PhaseandGainControl Operateondigitalinput.Mechanical Design Fully monolithic construction; compatible _
with 30 GHz spaceborne phased arrayapplications.
Unit Cost Less than $1000 (1980 dollars) in unit buysof 5000 or more
Table 2. Proposed Initial Low-NoiseFET Design
GATELENGTH L= 0,25MICRONS --
UNITGATEWIDTH w = 50MICRONS
CHANNELTHICKNESS A = 0,15MICRONS --
DOPINGDENSITY N = 2,5x 1017
TOTALGATEWIDTH Z = 100MICRONS --
GATE-SOURCESPACING LsG = 0,75 MICRONS
GATEMETALRESISTIVITY p = 4 x 10-60HM-CM _
SPECIFICCONTACTRESISTANCE Rc = 10-60HM-CM2
A1 : 0,22MICRONS
A2 : 0,15MICRONS
SEEFIGURE19 N1 = 2,5x 1017
N2 = 2,5 x 1017
CALCULATEDNOISEFIGURE= 2,3DBAT32GHz
66
r Table 3. Calculated Parametersfor FET Design
SParameters(29GHz)r
S,, = .75 /-- - 70°S,== 1.06/- 79.8 °S_ = .086 z_.79.7 °S= = .73/--- - 45°
NoiseParameters(29 GHz)F=_= 2.6 dBR, = 48 ohms
-- ]'o = .64/_ 43.9 °r, = .67/_ 51.1°G,= 5dB
GainParameters(29 GHz)K = 1.14G_= = 8.63 dBr. = .87 z..80°r,_ = .87 z..57°
67
Table 4a. S Parameters with No Cdg or Ls
POLAR S-PRRAI_ETER$ IN 50.0 OHM :SYSTEMFREQ. _ II ._21 ::_12. :'-'22 $2 t K
(MRGM_.RMGL) (MRGN<AtIGL) ( MAGM<:AI'IGL>(MRL;M<'RMGL) DB FACT.
:_1500.00 .95< -46 1.25< 116.0 .020< 120.2 .87< -2.8 1.95 -.0323000.00 .94< -49 1.23< I11.8 .022< 118.2 .87( -30 1.80 .0124500.00 .93< -52 1.21{ 1"07.7 .025< 116.3 .87< -:32 1.64 .05•:'600,).01) .93< -54 1.1:3<: 10:3.7 .02,":'< 114.3 .86( -2.4. 1.47 .10::'.50¢'.00 .9;'< -57 1.16< 99.? .0:30< 112.3 .:36-': -36 1.3,) .15:'--"31JO,).Oi) ....91( -60 1 14( 95.8 0:3_( 110.2. 86 !,. -38 1. I 1 .2.*'|:3051)|).00 .9,)( -63 I. If< 92.0 .035< 108.._ .85< -40 .'._ .25:3_.000.00 .90.( -65 1.1:i9< :38.3 .037< 1136.2 ::35:: -42 .73 . :31:'::_500. I)l'.l .8'_( -68 1.06<: _4.6 .040( I('4.2 .:_5<: -44 .5:._ .:37
POLAR COORDIr'IATE_ OF :._IMULTAMEOU:::COM_iUL;RTEMATCH
Table 4b. S Parameters with No Cdg, But with Ls
POLAR :_-PAF'Ar'IETER-._IM 50.0 OHM :_.'r':_TEMFF'Et_. _:11 :_21 :_.12 "=-22 :"::='1 K
(.I'IA|3;'t(AI'tGL> (MAL_N-:.':Ar4i3L) ( MA6N.:.;AHGL:, _MRGM<AMGL..', DB FAC:T.
:'-'l_l:li'.l. i_il:l . :3:_.( --45 1 • _4::_ i I I • 5 . 1"140{i 174.5 . :34:{ -26 1• :35 -. 06•:':3':!1)0. 00 .:37< -4:3 1.2.2.{ 1*:iT.4 .04," .'.: 174.0 . :34<." -27 1.70 -. 07::'4500.00 .86:. -50 1.20'::: 10"_.5 .055<. 173.2. .84.( -29 1.55 -.0;_•"-"601)0. OI) ._5.( -5:3 I. 1_( '_'_.6 • 06:3< 17_..:_ • 8:3::: -:31 1.40 -. U.c'•:'7500._)i) .83_: -56 I. 15< '.:'5.'_. 073< 171.3 .8:3< -3:3 I._.5 --.09:_9000.00 .821 -59 1.1:3< 92..2. .08:3<. 170.1 .83": -:34 1.0;3 -.10:':1:'500. O0 .:31( -61 1.11( 88.7 .0'.:'4 :: 16:3.7 .:33< -:36 ._2 -.1032000.00 • :3':i'( -64 1.09_. .'3:5.2. . 106,: 167..3 . :32"( -::::3 .75 -. 10:3:::51)0.00.79< -67 1.07{ :31.8 .I19{ 165._3 .'_32<:-4.') .57 -. 10
POLAR COOFeDII'IATE_::OF :'_"IMULTA/_EOU'_COM JUGATE MATCH
68
Table 5. S Parameters for Complete Model, Including Ls and Cdg
POLAR _-PARAMETER._ Ill 50.0 rIHM SY._TEMFREQ. _II _21 S12 $22 _21 K
(MIaGN<F_MGL) (MR_SN<RMGL) (MRGN<_MGL) (MR6H<FIMGL) DB FRCT.
21500.00 .:._2< -55 1.17< 99.9 .080< 68.9 .77< -35 1.39 .90_:3000.00 .80< -58 1.15< 95.6 .081< 70.0 .76< -37 1.22 .9624500.00 .79< -61 1.13< 91.4 .082< 71.7 .75< -39 1.05 1.0126000.00 .77< -64. 1.11< 87.4 .08:3< 73.9 .74< -41 .87 1.0627500.00 .76< -67 1.08< 83.5 .084< 76.5 .74< -43 .69 1.11_9000.00 .75< -70 1.06( 79.8 .086< 79.7 .73< -45 .51 1.1430500.00 .73< -73 1.0,$< 76.2 .088. 83.2 .72< -47 .33 1.1732000.00 .72: -76 1.02< 72.8 .091< 87.0 .72< -49 .15 1.1733500.00 .71< -79 1.00{ 69.4. .095< 90.9 .71< -51 -.04 1.16
POLF_R I:(](]RDIMFtTE_OF :_IMULTF_NEnU_ COMJUGFtTE MF_TCH
F :_OURCE REFL. COEFF. LO_D _EFL. ¢OEFF. GMR×MHZ MRGr'I.<RNL_LE MRGM..':RML_LE DB
•SUO.') 1.11a_rlIl'lL_:POTErtTIFtLLYUM'::TRBLECIRCUIT21500.0 .:-::6( 67 .82< 52 11.67
• 000.0 I,:AF;'tIr4L_: POTEMTIFtLLY ur'I:.'..TFIBLECIR,:UITr ,_30n0.0 .91<: 70 .89< 53 II.52
.-'4500.0 .96 _ 73 .97 54 10.69•%900.0 .91< 75 .90< 54 9.69
- _.750,).0 .88< 78 .87< 55 9.09:-"._,),:sO.0 .87< 80 .86 < 57 8.633,)500.0 .86:: :_:3 .86< 58 8.'_.732000.0 .86": :_6 .86<: 59 7.'.:'8_500.0 .:_6" :-38 .86< 61 7.76
FILE rtFtrtE. T_,:.:[MFO • ' r_'LIrtE'; • .LRrtL_E", 'r4Etd' OG, "_'_U[T"
-_ 69
HI. REFERENCES
i. K. Kurokawa and W.O. Schlosser, "Quality Factor of Switching Diodes for
Digital Modulation", Proc. IEEE, vol. 38, pp. 180-181, January 1970.
2. V. Sokolov,J. Geddes,A. Contolatis,P. Bauhahnand C. C_ao,"AKa-Band
GaAsMonolithicPhaseShifter",IEEETrans.MicrowaveTheoryTech.,vol.
MTI_-31,December1983.
3. H.A. AtwaterandR.W.Sudbury,_dseof SwitchingQ in the Designof FET
MicrowaveSwitches",in MTI_SInt.MicrowaveSymp.Dig., pp. 370-372,
June1981.
4. W.V. McLevige and V. Sokolov, "Resonated GaAs FET Devices for Microwave
Switching", IEEE Trans. Electron Devices, vol. ED-28, pp. 198-204,
February 1981. --
5. Y. Ayasli, A. Platzker, J. Vorhaus, and L.D. Reynolds,Jr., "A
MonolithicSingle-ChipX-Band Four-BitPhaseShifter",IEEE Trans.
MicrowaveTheoryTech.vol.MIT-30,pp.2201-2206,December1982.
6. R.A. GaspariandH.H.Yee, "MicrowaveGaAsFETSwitching",in 1978IEEE
Int.MicrowaveSymp.Dig.,pp. 58-60.
7. C.A. Liechti, "Performance of Dual-Gate GaAs MESFETs as Gain-Controlled
Low-Noise Amplifiers and High-Speed Modulators", IEEE Trans., Microwave
70
Theory and Tech., pp. 461-469, June 1975.
8. A.K. Wortman and E.E. Kohn, "A Monolithic Integrated Schottky Diode for
MicrowaveMixers",SolidStateElectronics,18,pp. 1095-1098,December
1975.f
9. W. Kellner,et.al., "DesignConsiderationsfor PlanarSchottkyBarrier
Diodes",SolidStateElectronics,23,pp. 9-15,January1980.
r
i0. N. Yokoyama,T. Oknishi,K. Odani,H. OnoderaandM. Abe, "Ti/WSilicide
GateTechnologyForSelf-AlignedGaAsMESFETVLSI's",Tech.Digest IEDM
80 (December1981).
ii. B.S. Hewitt, et.al., "Low-Noise GaAs MESFETs: Fabricationand
Performance",Inst.Phys.Conf.Set.No. 33a,pp.246-254,1977.
12. H. Fukui, "OptimalNoiseFigureof MicrowaveGaAsMESFETs,"IEEETrans.
Elect.Dev.,26,pp. 1032-1037,July1979.
13. C.A. Liechti,"MicrowaveField-EffectTransistor- 1976",IEEETrans.
- M_T,24,pp. 279-300,June1976.
14. R.A. Pucel, H.A. Haus, H. Statz, "Signaland Noise Propertiesof
GalliumArsenideMicrowaveField-EffectTransistors",in Advances in
electronicsand ElectronPhysics,38,p. 195,1975.
71
15. P.G. Borden, "Substrate Conduction in GaAs MESFETs", Electronics Letter,
15, pp. 307-308, 1979.
16. L.F. Eastmanand M.S. Shur, "SubstrateCurrentin GaAsMESFETs",IEEE --
Trans.ElectronDev.,26,pp. 1359-1361,September1979.
17. J.C. Voker, et.al., "NovelMicrowaveGaAs Field-EffectTransistor",
ElectronicsLetter,15,pp. 627-629,September1979.
18. C.H. Oxley, et.al., "Q Band (26-40GHz)GaAsFETs",1981IEEMDigest,
pp. 680-683. --_
19. P.W. Chye and C. Huang, "QuarterMicronLow NoiseGaAs FETs",IEEE
ElectronDeviceLetters,13, pp. 401-403,December1982.
20. C.H. Oxley, et.al., "Q Band (26-40 Ghz) GaAs FET Single Stage
_mplifier",ElectronicsLetters,18,pp. 260-262,March18,1982.
21. H. Fukui, "Designof MicrowaveGaAs MESFETsforBroad-BandLow-Noise
Amplifier",IEEETrans.MIT,27,pp.643-650,July1979.
22. H. Fukui,Addendumto above,IEEETrans.MTT,29, October1981.
23. W.A. Surer, "Feedback and ParasiticEffectson Noise",Microwave
Journal,p. 123,February1983.
72
=
24. H. Yamasakiand J.M. Schellenberg,"Designof Millimeter-WaveFETs",
Proc.GaAsandRelatedCompounds,1981,Olso,Japan.
73Thereverseof this pageis blank.
FF-