A Variable Output Power, High Efficiency, Power Amplifier for the ...

A Variable Output Power, High Efficiency, PowerAmplifier for the 2.45 GHz ISM BandSajjad Ahmed, Ahmed Sayed, Henrique Portela, Olof Bengtsson*, Georg BoeckMicrowave Engineering Laboratory, Berlin University of TechnologyEinsteinufer 25, 10587 Berlin, Germany*Ferdinand-Braun-Institut für HoechstfrequenztechnikGustav-Kirchhoff-Str. 4,12489 Berlin, Germanysayed@mwt.ee.tu-berlin.deAbstract— This paper, presents the design and evaluation of ahigh efficiency, 30 W, two-stage, power amplifier for the 2.45GHz ISM band. The designed amplifier targets an industrialapplication where precise control over the output power withgood efficiency is necessary. A saturated output power of44.7 dBm (30 Watts) was reached at a maximum drain efficiencyof 67.8 %. A full 5 dB control of the output power was achievedby varying the gate bias of the driver stage. A minimum poweradded efficiency of 39.4% was reached over the full controllableoutput power range at the design frequency. The saturatedoutput power over a 150 MHz bandwidth around the designfrequency was more than 44 dBm with a power added efficiencyof more than 60 %.I. INTRODUCTIONRadio frequency (RF) commonly refers to high frequencyradio communication but there is an entirely different categorywhere RF devices are used for non-communication purposes.The frequency band utilized for these applications is known asIndustrial, Scientific and Medical (ISM) band and applicationsin this band are centred at different frequencies ranging from afew MHz to the mm-wave range [1]. Applications at ISMbandwere for a long time considered as non-commercial butrecently these bands have seen a greater commercial interest,mainly because they are now also utilized for communicationlike WiFi and Bluetooth applications. Other applicationswhich fall under the ISM category include RF heating, plasmagenerators, laser exciters, RF lighting, medical telemetry,magnetic resonance imaging and many others [1]. In some ofthe applications, the RF output power (P OUT ) is a few miliwattswhere as it might reach hundreds of watts in otherapplications. Traditionally the vast amount of power used inheating applications have ruled out solid-state solutions butrecent technology development and application demands arenow starting to make transistor solutions a cost efficientalternative [2],[3]. For the heating application this work istargeting, moderate power but controllability of the outputpower is of main concern. The amount of output powerdirectly relates to the temperature of the heated object. Highefficiency is needed over the full output range to reduceoperating cost, size and the cost of additional heat removinghardware.Lately, Gallium-Nitride (GaN) high electron mobilitytransistor (HEMT) technology has emerged as a highefficiency alternative to the mainstream Silicon LDMOStechnology for high power applications in the 10 W to 300 Wpower ranges below 3 GHz. With its high breakdown field,GaN high-power devices can be made small thereby reducingthe parasitic losses and increasing the efficiency [4]. In [5] a50 dBm output power with 15 dB power gain and 63 % drainefficiency (η) is presented where as in [6], drain efficiency (η)of 50 %, output power of 250 W and 15 dB power gain hasbeen reported.In this paper, a two stage, 30 W PA at the 2.45 GHz ISMband is presented. GaN-HEMT technology is used for the highefficiency power stage and Si-LDMOS technology is used forimproved gain controllability in the driver stage. Part IIhighlights the design of the amplifier, concept,implementation and realization. In part III the amplifier isevaluated, under small signal operation for model verificationand under large signal conditions, simulations are compared tomeasurements under all conditions. Part IV is a briefdiscussion of the possibilities with the design and in part V thework is concluded.II. AMPLIFIER CONCEPT AND DESIGNIn the application, the power amplifier is to delivermicrowave power up to 30 W (44.8 dBm) at 2.45 GHz withprecise control of output power with high efficiency at aconstant available input power of 100 mW (20 dBm).Considering these specifications a two-stage amplifier wasdesigned with a Cree AlGaN/GaN HEMT (CGH40025H)transistor for the power stage (PS) and a Freescale Si LDMOS(MW6S004N) transistor for the driver stage (DS). A blockdiagram for the two-stage amplifier is shown in Fig. 1.sourceInputmatchingDriver stageInterstagematchigPower StageFig. 1 Block diagram for two stage power amplifierOutputmatchingLoadThe high power density of GaN makes the device size verysmall which corresponds to high impedances at source andload compared to other devices which in-turn makes matchingeasier [7]. The selection of Si LDMOS transistor for the driverstage is primarily because of its high gain and high drainvoltage characteristics. To achieve the objective of high978-1-4244-5357-3/09/$26.00©2009IEEE 821

efficiency the transistors are biased towards class B operation.The amplifier was designed for a drain supply (V D ) of 28V.A. Design ProcedureLarge-signal models were available for both transistorsfrom the manufacturers. Although their validity could bequestioned they were later shown to well model the behaviourunder the conditions used in this design. Initially harmonicbalance,load-pull simulations were conducted in Agilent-ADS to establish a high power added efficiency (PAE) at therequested output power for the power stage. The trade-offbetween output power, gain, efficiency and quiescent pointwas observed in the simulations and an optimal quiescentcurrent of around 10 % and 5 % of the maximum saturateddrain current for both power and driver stage respectivelywere determined to achieve the desired objectives of outputpower and efficiency. Independent load and source-pullsimulations were then performed for both stages. Initially,load and source-pull simulations were performed with idealcomponents and ideal optimum load and source impedanceswere found. Independently both transistors were stabilized tounconditional stability by fulfilling the Rollet’s stabilitycriterion [8] (K > 1). This was achieved by adding a parallelresistance with capacitance in series at the input, effectivelyreducing the low frequency gain. λ/4 transmission lines wereused in the biasing networks to improve isolation at RFfrequency. Load and source pull simulation were performedagain to achieve optimum impedances for the two stages. Theoptimum load and source impedances found are summarizedTable I. A maximum optimum driver stage quiescent current(IDQ (DS) ) was found to be 32 mA at an optimum power stagequiescent current (IDQ (PS) ) of 123 mA.TABLE IOPTIMUM LOAD AND SOURCE IMPEDANCE AND QUIESSENTCURRENTPower stage(PS)Driver stage(DS)Z LOAD[]Z SOURCE[]I DQ[mA]6.2 – j10.3 6.7 – j26.5 1236.0 – j6.6 5.0 – j30.0 32B. Design ImplementationThe matching networks were realized using micro-striplines together with open stubs in low pass form. Inputmatching was designed to achieve good return loss. It wasimplemented using a tapered line to compensate for some ofthe series capacitance introduces by the capacitor in thestabilizing network. The inter-stage matching network wasdesigned to minimize mismatch losses and for maximumtransfer of power between the two stages. The outputmatching network was designed for minimum loss to the loadand a simulated insertion loss < 0.1 dB was achieved for afrequency range of 200 MHz with centre frequency of2.45 GHz. Momentum simulations were used to verify theimpedances created by the linear matching networks.C. Design RealizationThe amplifier was built on Rogers RO4003 substrate with adielectric constant ε r of 3.66 and thickness of 0.51 mm. Fig. 2shows a prototype for the designed PA.Fig. 2 Power amplifier prototypeVia holes were manually fabricated. Thermal control wasestablished using an aluminium heat sink with a cooling fan.Bias was separately controllable for the two stages in theinitial tests but was later fixed for the power stage andintegrated with a temperature-gain control circuitry. Thematching networks were made on separate substrates so thatlinear performance of the networks could be measured andverified individually.III. AMPLIFIER EVALUATIONThe design process was totally based on a model approach.Small signal evaluation was therefore used to verify thecorrectness of the models used thereby validating the designprocess.A. Small Signal PerformanceThe small-signal s-parameters of the full amplifier are showin Fig. 3.S-Parameters [dB]40200-20-40S21simS11simS22simS21measS22measS11meas2.0 2.2 2.4 2.6 2.8 3.0ƒ [GHz]Fig. 3 Measured and simulated small signal s-parameters at V D = 28 V andmaximum drain current for the driver stage.It can be observed that a small-signal gain of 27 dB withinput and output reflections less than -15 dB in the operatingfrequency is achieved. The K factor was measured to > 4 overthe frequency band from 1 kHz – 10 GHz. This is mainlybecause both the stages were individually stabilized beforecascading. Compared to simulations, gain measurements2009 SBMO/IEEE MTT-S International Microwave & Optoelectronics Conference (IMOC 2009) 822

exhibit some degradation although the main behaviour is verywell tracked by the simulations. The small discrepancy is acombination of many factors, transistor model in-accuracy,tolerance of the used devices and tolerance in the fabricatedboards.B. Power Performance at Maximum BiasA signal generator together with a pre-amplifier was usedas power source in order to generate desired maximumavailable input power (P AVS ) of 20 dBm during the powermeasurements. Initially the maximum operating performancewas tested at maximum driver stage biasing, I DQ(DS) = 32 mA.The measured results of output power, transducer power gain(Gain), drain efficiency and power added efficiency are shownin Fig. 4 and Fig. 5.P OUT[dBm], Gain [dB]50403020Gainsim [dB] Gainmeas[dB]Poutsim [dBm] Poutmeas [dBm]105 10 15 20 25Fig. 4 Measured and simulated gain and output power performance at V D =28 V, I DQ(DS) = 23 mA and I DQ(PS) = 123 mA.80C. Power Performance over BandwidthAlthough the amplifier was designed for a fixed frequencyuse at 2.45 GHz the ISM band allows some frequencymodulation of the signals. The amplifier was thereforemeasured over a limited bandwidth around the designfrequency. Fig. 6 shows the power performance at saturationover a 150 MHz frequency range.PA Performance806040200Gainsim [dB] Gainmeas [dB]PAEsim [%] PAEmeas [%]Poutsim [dBm] Poutmeas [dBm]2.35 2.40 2.45 2.50ƒ [GHz]Fig. 6 Output power and PAE for a frequency range of 150 MHz at V D = 28V, I DQ(DS) = 23 mA and I DQ(PS) = 123 mA.An output power of more than 29 W and a PAE of morethan 60 % could be measured over the full 150 MHzbandwidth. Very good agreement was also found over theband between simulated and measured results.D. Power Performance over Gain ControlThe gain of the amplifier is controlled by setting the gatevoltage of the driver stage LDMOS thereby adjusting I DQ(DS) .Fig. 7 shows the output power controllability of the amplifier.46Efficiency [%]604020P OUT[dBm]4442sim [%] PAEsim [%]meas [%] PAEmeas [%]05 10 15 20 25Input PowerdBmFig. 5 Measured and simulated drain efficiency and PAE at V D = 28 V, I DQ(DS)= 23 mA and I DQ(PS) = 123 mA.A saturated output power of 44.8 dBm (30 W) and a gain of25 dB was achieved. The maximum drain efficiency is 67.8 %at saturated power along with an almost similar PAE due tothe high gain of the amplifier. Very good agreement betweensimulation and measurements were observed.40P OUT(meas)P OUT(sim)380 10 20 30 40IDS_Driver [mA]Fig. 7 Measured and simulated power control at f = 2.45 GHz. V D = 28 V,and I DQ(PS) = 123 mA.As shown the controllability tracks very well the simulatedvalues. By varying the current of the driver stage between3 mA and 33 mA at a constant input power (P IN ) at 100 mW(20 dBm) a measured tuneable range of 5 dB was achieved2009 SBMO/IEEE MTT-S International Microwave & Optoelectronics Conference (IMOC 2009) 823

(39 dBm to 44 dBm). This was somewhat less than the 6 dBexpected from simulations. The achieved PAE during powercontrol is shown in Fig. 8Gain [dB], PAE [%]70605040302010Gain (meas)PAE (meas)Gain (sim)PAE (sim)39 40 41 42 43 44P OUTdBmFig. 8 Measured and simulated gain and PAE during power control atf = 2.45 GHz. V D = 28 V and I DQ(PS) = 123 mA.From this figure it can be seen that a PAE of almost 40 % ismaintained for the two-stage amplifier over the full tuneablerange.IV. DISCUSSIONAlthough this amplifier was designed for heatingapplications at ISM band where power generated at otherfrequencies is of little concern but the topology used may bevery suitable for more frequency sensitive applications [2]within the ISM band. This covers e.g. applications withincommunication and microwave enhanced chemical reactions.Fig. 9 represents the two-tone IMD performance of thedesigned amplifier.Output Power 2-tone [dBm]806040200-20-40-60P3(meas)P2(meas)P1(meas)P3(sim)P2(sim)P1(sim)IMD3 ~65 dBcV. CONCLUSIONSIn this paper the design of a two stage 30 W poweramplifier with controllable gain for the ISM band is presented.Power control and high efficiency over the controllable outputpower range were the main design objectives. A model baseddesign approach was used and verified in the work. High gainrequirement forced it to be a two-stage amplifier where ashigh efficiency over the full controllable power range wasachieved using GaN HEMT technology in the power stageand Si-LDMOS technology in the driver stage. A gain of25 dB was measured for the two stage amplifier at themaximum output power of 30 W. In compression a drainefficiency of 67.8 % and PAE of 67.2 % was achieved. Theamplifier performs well over a full bandwidth of 150 MHzwith a saturated output power of more than 44 dBm at a PAE> 60% for the full band.The output power is controllable within a 5 dB range bysetting the bias of the driver stage. More than 40 % PAE ismaintained over the full controllable output power rangewhich makes the amplifier well suited for the heatingapplication it was designed for.ACKNOWLEDGMENTSThe authors would like to thank Cree for providing themodel for the GaN-HEMT transistor used and Freescale forproviding the model for the Si-LDMOS used.REFERENCES[1] “The ISM Revolution: The Next Big Thing”, By Iboun Taimiya Sylla,Texas Instruments, 9th Feburary, 2009, www.rfdesignline.com[2] Wojciech Wojtasiak, Daniel Gryglewski and Wojciech Gwarek,' “AlOOW ISM-2.45GHz BAND POWER TEST SYSTEM,” *Institute ofRadioelectronics, Warsaw University of Technology, Poland. Journalof telecommunications and information technology, 2005.[3] Isao Takenaka, Hidemasa Takahashi, Kazunori Asano, JunkoMorikawa, Kohji Ishikura, Mikio Kanamori", Masaaki Kuzuhara andHiroaki Tsutsui, “High efficiency S-band 30W power GaAs FETs”ULSI Device Development Laboratories, NEC Corporation. 9-1,Seiran 2-Chome70tsu,Shiga 520,Japan.[4] A. H. Jardnal, “Large signal modelling for GaN devices for high poweramplifier”, PhD., University of Kassel, Germany, Nov. 2006.[5] Nitronex Corporation, “GaN Power HEMT for 2.5 GHz WiMAXApplications”, NPT25100, 2305 Presidential Drive. Durham, NC27703.[6] K. Krishnamurthy, M. J. Poulton, J. Martin, R. Vetury, J. Brown, and J.B. Shealy, “A 250 W S-Band GaN HEMT Amplifier”, InfrastructureProduct Line, RF Micro Devices, Inc., Charlotte, NC 28269, USA.[7] J. Shumaker, M. Ohoka, and N. Ui, “Design of Power AmplifiersUsing High Breakdown GaN HEMT Devices”, Eudyna Devices USAInc, and Eudyna Devices Inc Japan.[8] S. Cripps, “Advanced Techniques in RF Power Amplifier Design”, 1stEd., Artech House, 2002.-20 -10 0 10 20 30Input Power 2-tone [dBm]Fig. 9 Two-tone IMD performance of the amplifier for a tone-spacing of 100MHx at the deigns frequency 2.45 GHz at V D = 28 V, I DQ(DS) = 23 mA andI DQ(PS) = 123 mA.2009 SBMO/IEEE MTT-S International Microwave & Optoelectronics Conference (IMOC 2009) 824