A Miniaturized WiMAX Band 4-W Class-F GaN HEMT ... - IEEE Xplore

3184

IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 59, NO. 12, DECEMBER 2011

A Miniaturized WiMAX Band 4-W Class-F GaN HEMT Power Amplifier Module Hae-Chang Jeong, Hyun-Seok Oh, and Kyung-Whan Yeom, Member, IEEE

Abstract—In this paper, the design and fabrication of a miniaturized 4-W power amplifier for the WiMAX frequency band (2.3 2.7 GHz) is presented. The selected active device is a commercially available GaN HEMT chip from TriQuint Semiconducotr Inc., Hillsboro, OR. The optimum input and output impedances of the GaN HEMT at fundamental frequency are extracted using a custom designed tuning jig. A novel output matching network for class-F operation is proposed and designed using the measured impedances. For integration in a small package, the input and output matching networks are implemented using spiral inductors and interdigital capacitors, and their dimensions were determined using electromagnetic simulation. The fabricated power amplifier is 4.4 4.4 mm2 and has an efficiency above 50% and harmonic suppression above 40 dBc for second and third harmonics at an output power of 36 dBm. Index Terms—Class-F power amplifier, gallium–nitride (GaN) HEMT, matching network, WiMAX.

I. INTRODUCTION iMAX services based on the IEEE Standard 802.16 [1] emerged to accommodate the increasing demand for wireless internet access. The frequency band for WiMAX is not fixed, but the 2.3 2.7 GHz frequency band is popular and promising for WiMAX services. Miniaturized power amplifiers are expected to be useful for the deployment of base units in the construction of wireless networks due to their smaller size. A power level of 2 10 W is regarded as appropriate for such networks [2]. Thus, a high-efficiency miniaturized power amplifier of 2 10 W may be a good candidate for base unit application. Gallium nitride (GaN) has a high breakdown voltage due to its wide bandgap property, which is useful for high-power devices [3]. Its high thermal conductivity also provides another advantage for power-amplifier application. However, its electron mobility is somewhat inferior to GaAs, and consequently, an HEMT structure is usually employed. GaN HEMT is assessed to be a promising device for power amplifiers and is expected to be a competitor to LDMOSFETs, which are widely used as base-station power amplifiers for mobile communications. GaN

W

Manuscript received May 31, 2011; revised August 22, 2011; accepted September 13, 2011. Date of publication October 26, 2011; date of current version December 14, 2011. This work was supported by the Korean Government under the National Research Foundation of Korea (KRF-2011-0003851). The authors are with the Department of Radio Science and Engineering, Chungnam National University, Daejeon 305-764, Korea (e-mail: [email protected]; [email protected]; [email protected]). Color versions of one or more of the figures in this paper are available online at http://ieeexplore.ieee.org. Digital Object Identifier 10.1109/TMTT.2011.2169422

HEMTs above 100-W output power have recently been reported and are commercially available [4]. Among the various classes for improving power efficiency, class-F power amplifiers [5] have become very popular in recent years and seem to be a strong candidate for high-efficiency power amplifiers. Various kinds of the highly efficient class-F output matching networks have been studied by many researchers. However, most of them employ transmission lines [6]–[9], which seem to be rather large for miniaturized power-amplifier design. Several ingenious class-F output matching network designs operating with limited order of harmonics using lumped elements have been presented [6], [10]. Some works based on lumped elements are quite general and can be extended to arbitrary harmonics [8], [9], [11]. However, their synthesis basically requires lumped elements replacement by transmission lines due to parasitic elements appearing in the lumped elements. We propose a novel class-F output matching circuit composed of lumped elements. Compared with the previous designs, our class-F matching network includes both matching to load impedance and harmonic suppression. In addition, it provides a more tolerant design for harmonic suppression. In this paper, we present a miniaturized class-F power-amplifier design with a commercially available GaN HEMT chip from TriQuint Semiconductor Inc., Hillsboro, OR. First, the evaluation of the GaN HEMT chip is presented using a fabricated tuning jig to extract the optimum input and output impedances for the selected device. The optimum input and output impedances can be conveniently extracted from the load–pull simulation using the large signal model. However, in most cases, the accurate large-signal model for the selected devices is not timely available and the accuracy of the large-signal model is not sufficiently assessed to be employed in power-amplifier design. In that case, the load–pull and recent -parameter measurements [12]–[14] may be an alternative to shortening design time considering that the development of accurate large-signal model is formidable and requires long time. The presented method provides cost-effective evaluation and sufficient guidelines to design a power amplifier for the selected devices despite the unavailability of complete characterization of the device. The miniaturized class-F amplifier is then designed using the measured optimum input and output impedances. We employed a novel class-F output matching circuit, which is implemented using lumped elements such as spiral inductors and interdigital capacitors for miniaturization. The fabricated amplifier is 4.4 4.4 mm , and has an output power of above 4 W, power gain of about 10 dB, efficiency of about 50%, and harmonic suppression above 40 dBc.

0018-9480/$26.00 © 2011 IEEE

JEONG et al.: MINIATURIZED WiMAX BAND 4-W CLASS-F GaN HEMT POWER AMPLIFIER MODULE

TABLE I PERFORMANCES OF TGF2023-01

II. CLASS-F POWER-AMPLIFIER DESIGN A. Evaluation of GaN HEMT We selected the recently released TGF2023-01 GaN HEMT from TriQuint Semiconductor Inc. for the miniaturized power amplifier. TGF2023-01 [15] is supplied as a bare chip and its source terminal is grounded. The gate length and gatewidth are 0.25 m and 1.25 mm, respectively. It can be used up to the -band and its key parameters of performances are summarized in Table I. The optimum impedances for maximum efficiency and maximum power are given in the data sheet at the bias of the quiescent drain current of 125 mA and drain voltage of 30 V. Our dc supply voltage is set to 28 V. Although the output impedance at 3 GHz and 30 V in the data sheet may be enough to design our power amplifier to some extent, the precise performances of the GaN HEMT such as maximum output power and optimum input and output impedances considering the stability at our dc bias are again evaluated. The experience of handling the GaN HEMT, which is sensitive to dc supply and assembly state compared with other devices, can also be obtained through the evaluation. The GaN HEMT is packaged using a ceramic package C580274C from StratEdge, San Diego, CA.1 Fig. 1(a) shows a photograph of the assembled package. In order to connect the GaN HEMT chip to the package, patterned substrates were fabricated using a 10-mil-thick alumina. The fabricated substrate also provides pads for impedance measurement using a wafer probe. The GaN HEMT chip is attached using Au/Sn preform and bonded to the fabricated alumina substrate using 3-mil gold ribbon wire. Fig. 1(b) shows a magnified view of Fig. 1(a). Fig. 1(c) shows a package assembly without the GaN HEMT chip for impedance measurement using a wafer probe. From Fig. 1(c), the measured impedance includes the bonding ribbon wire effects. The bonding wire inductance is estimated, using AnSoft HFSS simulation, to be about 0.23 nH. Fig. 2(a) shows the top view of the tuning test jig for obtaining the optimum impedances. The trimming capacitors are 5052 air trimming capacitor from Johanson Technology, Boonton, NJ.2 The assembled GaN HEMT package is bolted down to the jig, and their leads are soldered to the tuning jig printed circuit board (PCB). Fig. 2(b) shows the mounting of air trimming capacitors. Air-trimming capacitors are vertically mounted on the tuning jig to prevent possible radiation. When they were mounted at the 1C580274C, leaded power amplifier package. [Online]. Available: http://stratedge.com 25502, Air Trimmer Capacitor. [Online]. Available: http://johansonmfg.com

3185

top of the PCB, the loss was significant and showed some degree of instability due to radiation. The locations of air trimmer capacitors were determined using the optimum impedance given in the data sheet [15]. Fig. 3 shows the measurement setup for tuning. The driver amplifier is needed because the signal generator does not provide enough power to drive the GaN HEMT. The driver amplifier was designed and fabricated in our laboratory; its measured output power is above 25 dBm and the second and third harmonic suppression are below 40 dBc, which is sufficient to drive our GaN HEMT. In order to prevent the change in the output power of the driver amplifier during tuning, the isolator ADC250CSQH from Admotech Inc., Daejeon, Korea.3 was inserted as shown in Fig. 3. The test jig output power is split in two ways: one is applied to a power meter and the other is applied to a spectrum analyzer. The spectrum analyzer was employed to observe any possible oscillations. The air-trimming capacitors from input to output were then tuned for maximum power at a frequency of 2.5 GHz. The maximum output power was achieved after two cycles of tuning from input to output. Our tuning was carried out for maximum output power rather than for maximum efficiency. Generally, maximum output power is dominantly dependent on fundamental impedance, while harmonic impedances affect the maximum efficiency. One reason for tuning for maximum output power is that the output power is a primary specification. Another reason is the consideration of the tuning capability of our tuning jig. The harmonic impedances cannot be tuned to yield the desired harmonic impedances using our tuning jig. The maximum efficiency condition achieved using fundamental impedance tuning alone may also cause the problem of reproducibility due to the difference in harmonic impedances. Two samples were prepared for repeatability check. Fig. 4 shows the results of the output power and the power-added efficiency (PAE) for the two samples. Sample #2 provides higher power, but lower efficiency than sample #1. For the two samples, the measured output power was above 37 dBm, the smallsignal gain was about 17 dB, and the efficiency was observed to be above 50%. The GaN HEMT is biased to allow a flow of 125-mA drain current with no RF input. When RF input is applied, the current increased to about 300 mA. It was concluded through the measurement that the selected GaN HEMT is the adequate device for a 4-W power amplifier at 2.5 GHz. The load–pull simulation at the same dc bias using recently released large-signal model from TriQuint Semiconductor4 was also carried out for comparison at the input power of 22.5 dBm. The same value of the input stabilization resistor of 3.9 in the tuning jig in Fig. 2(a) was included in the input during the simulation. Although no oscillation was observed in the simulation, the oscillation was observed in the real measurement without the input stabilization resistor. The input and output impedances yielding the maximum output power were obtained after iterative source– and load–pull tuning. Using the input and output 3ADC250CSQH, Coaxial Circulator. [Online]. Available: http://admotech.com 4TriQuint EEHEMT model implemented in ADS and AWR For TQT 0.25-m 3MI GaN on SiC process 1.25-mm discrete FET: 28 V @ 100 mA/mm @ 3-14 GHz, Aug. 2010.

3186


Fig. 3. Measurement setup.

Fig. 1. Photographs of package assembly. (a) Package assembly, (b) magnified view, and (c) assembly for measuring impedance using wafer probe. Fig. 4. Measured output power and efficiency using the tuning jig and those from load–pull simulation at a frequency of 2.5 GHz for the two samples.

Fig. 5. Photograph of measurement setup for optimum impedances.

Fig. 2. Tuning jig. (a) Top and (b) sectional views to show assembly of air trimmer capacitor.

impedances, the output power and PAE were then computed for input power change and plotted in Fig. 4. As shown in Fig. 4, the simulated output power is higher than the measured output power by about 1 dB, while the PAE is between the PAEs of the two samples. The difference in the output power of about 1 dB is believed to be caused by the tuning jig loss. Thus, the TriQuint Semiconductor Inc. large-signal model is assessed to be accurate to some degree.

In order to find the input and output impedances, the test jig was disassembled and the impedance measuring package in Fig. 1(c) connected. The optimum impedances were measured using a wafer probe. Fig. 5 shows a photograph of the measurement. The measured impedances are listed in Table II. The real parts of the measured admittances are found to be somewhat widespread according to the PAE change, while the imaginary parts are close for the two samples. A similar phenomenon is found in the data sheet. It also should be noted that the impedances in Table II include the bonding wire inductances, while the latter is not included in the load–pull simulated impedances. However, the two results obviously provide the output power of above 4 W and the average (65 /0.7 pF) of the real and imaginary parts of the admittances was selected as the design


TABLE II MEASURED INPUT AND OUTPUT IMPEDANCES

3187

at the fundamental frequency. Defining the as

be matched to and ratios

and

TABLE III COMPARISON OF MEASURED AND LOAD–PULL SIMULATION IMPEDANCES

the values of given by [16]

,

, and

(1)

at the fundamental frequency are

(2a) and

(2b)

The signs of and are chosen as positive, and that of is chosen as negative in which case the value of can be determined as (3) Denoting , , and as the angular frequencies of the fundamental, second, and third harmonics, respectively, the admitof the branch at the fundamental frequency, tance which is resonant at the third harmonic, is given by (4)

Fig. 6. Proposed class-F output matching network.

impedance. The correction of the bonding wire inductance of at 2.5 GHz) resulted in approximately same 0.23 nH ( average values of the output impedances, while the imaginary parts of the input impedance increases by 3.6 . The load–pull simulated impedances are compared with the averaged impedances in Table III and are found to be between the two sample impedances. This implies indirectly that it is reasonable to select the impedances for the power-amplifier design by averaging the measured impedances of the two samples.

This should be equal to given in (2b) for matching and the and can be computed as values of and The admittance pressed as

(5)

at the fundamental frequency can be ex-

(6) B. Class-F Output Matching Circuit Fig. 6 shows our class-F output matching network, which provides a short for second harmonic and an open for third harand monic frequencies. Series connected branches are established to be resonant and become short at second and third harmonic frequencies, respectively. Since is a short at the third harmonic frequency, the impedance seen can be made open with the proper selection of element from values to satisfy . In the case when , makes it difficult for the impedance seen the loss due to to be driven to at the third harmonic frequency. from and represent drain and load conductances. Capacitor and inductor represent drain capacitance and dc supply inductance, respectively. Each corresponding susceptance of the Pi-shaped output matching network is represented by . at the reference For an appropriately selected plane, as shown in Fig. 6, the drain and load impedances can

The admittance seen from the drain at the third harmonic frequency should be zero, and is given by (7)

Solving (6) and (7) for

and

results in (8) (9) (10)

3188


Fig. 7. (a) Transmission characteristics for impedance at n : .

= 05

n

= 0:2; 0:5; and 0:8. (b) Drain

TABLE IV CALCULATED ELEMENT VALUES FOR n : , AND : , WITH Cp : pF

05

08

=07

= 0:2,

Fig. 7(a) shows the transmission characteristics for S for the parameter and . The calculated values using (1)–(10) are listed in Table IV. becomes smaller, more harmonic suppression results As and the bandwidth at the fundamental frequency becomes narrower. Fig. 7(b) shows the impedance seen from the drain . The impedance is matched at the fundamental for and becomes short and open exactly at the second and third harmonic frequencies, respectively. Our matching network shown in Fig. 6 obviously includes the drain capacitance. The formulas for the matching network element values can similarly be derived when the drain bonding

Fig. 8. (a) Transmission characteristics and (b) drain impedance at when drain bonding inductance 0.2 nH is included.

n

= 0:5

inductance is included in the matching network. This is, however, too complex, which necessitated employing an optimization technique. The results from (1)–(10) provide a good initial point for optimization. Optimization can be successfully carried out using the constraints of matching at the fundamental frequency, short at the second harmonic, and open at the third harmonic frequencies. Fig. 8(a) shows the transmission characteristic and the drain impedance. From Fig. 8(b), it can be found that a successful class-F matching network is achieved. However, it should be noted that the transmission zero does not occur at the second harmonic frequency. The values of the circuit elements are listed in Table V. Fig. 9 shows the comparison of transmission characteristic . We sewith other researches [6], [9] for for our matching network. Many designs are lected available; however, most of them are based on transmission lines that will not make for a proper comparison. Thus, we selected the design of two researchers. The results of Kuroda et al. [9] in Fig. 9 shows a narrower short and open at the second and third


TABLE V VALUES OF OUTPUT MATCHING CIRCUIT WITH Lp

3189

= 0:2 nH

Fig. 10. Class-F amplifier schematic. TABLE VII VALUES OF INPUT AND OUTPUT MATCHING CIRCUIT ELEMENTS

Fig. 9. Comparison of the transmission characteristics with other researches. TABLE VI CALCULATED ELEMENT VALUES WITH OTHER RESEARCHERS

harmonic frequencies, although their design can be extended to arbitrary harmonics. The results of Grebennikov [6] show similar characteristic to ours, which is also slightly narrower than ours. The calculated elements values of the three circuits are listed in Table VI. The two designs of the other researchers require matching networks. These are realized using constant load impedances, and their values are stated in Table VI. Practically, when the matching network is implemented, it can disturb the frequency response shown in Fig. 9. Thus, careful matching network design is required to preserve the class-F characteristics. on GrebenFurther investigation of the effects of finite nikov’s output matching networks and ours was carried out. The of second and third harmonic resonators was altered from 10 to 1000. Our matching network is sensitive to the lowering of the of the second harmonic series resonator, while Grebennikov’s is sensitive to that of the third harmonic peaking resonator. The second harmonic series resonator is close to the drain in the case of ours, while third harmonic peaking resonator is close to the drain in the case of Grebennikov’s network. This implies the of the resonator close to the drain should be higher to minimize possible loss effects. In addition, the lower affects the fundamental matching in the case of Grebennikov’s network,

Fig. 11. Simulated drain voltage and current using harmonic-balance simulation at an output power of 36 dBm.

while ours is almost unaffected. This may be an advantage of our matching network. C. Power-Amplifier Design Fig. 10 shows our class-F amplifier circuit. The input and capacitor , and matching circuit consists of inductor matches the transistor input impedance at the fundamental is included for stabilization of the frequency. Resistor amplifier. The gate of transistor Q1 is biased by resistor . in the input and output matching circuits are All capacitors is selected dc block and bypass capacitors. The value of as 22 pF. The role of other elements in the output matching network is the same as those in Fig. 6. In the output matching network design, we first approximated the bonding wire inductance to zero, assuming that the bonding wire length can be minimized. For the chosen drain impedance and pF, the computed values of of the output matching circuits are listed in Table VII. The transmission characteristic resembles a low-pass filter frequency response. The insertion loss of the output matching network is

3190


Fig. 13. (a) Drain impedance and (b) transmission characteristic up to third harmonic frequencies including bonding-wire inductance. Points A–C correspond to the second harmonic frequencies of 2.3, 2.5, and 2.7 GHz, respectively.

Fig. 12. (a) Layout, (b) EM simulation layout for input, and (c) output matching circuits.

close to zero around the fundamental frequency, and the transmission zeroes are formed at the second and third harmonic

frequencies. Fig. 11 shows simulated drain voltage and current using the large-signal model supplied by TriQuint Semiconductor Inc. at a frequency of 2.5 GHz. Due to the limiting of the number of harmonics, the waveform does not clearly show class-F operation. However, the drain current is almost zero when the drain voltage shows peak. Thus, the waveform in Fig. 11 shows approximate class-F operation. The inductors and capacitors in Fig. 10 were implemented using spiral inductors and interdigital capacitors. The initial geometries of the spiral inductors and interdigital capacitors were determined using ADS schematic simulation. Fig. 12(a) shows a layout for the amplifier shown in Fig. 10. The substrate is a 5-mil-thick alumina with a permittivity of 9.9 [17]. The size of the alumina substrate is 4.4 4.4 mm . Single-layer capacitors of 22 pF are employed for dc blocks and bypass capacitors. is also implemented using a Input matching capacitor single-layer capacitor because its value is too big to be implemented with interdigital capacitor. The patterns 1–1’, 2–2’, 3–3’, and 4–4’ are for wafer probing, and thus the -parameters of input and output matching circuits can be measured


3191

Fig. 16. Measured and EM simulated S -parameter of the fabricated power amplifier. The drain voltage is 28 V and the quiescent drain current is 125 mA.

Fig. 14. Comparison of the simulated output powers and PAEs for the power amplifiers composed of the lumped elements in Table VII and EM simulated layouts. (a) Simulated output power and PAE for the frequency at the input power of 22.5 dBm. (b) Output power and PAE for the input power at the frequency of 2.5 GHz. Fig. 17. Comparison of measured transmission characteristic and re-simulation results with the conductor width of the interdigital capacitors reduced by 2 m. Points A–C correspond to the second harmonic frequencies of 2.3, 2.5, and 2.7 GHz, respectively.

Fig. 15. (a) Package assembly and (b) magnified view of power-amplifier assembly.

using wafer probes before assembly. Fig. 12(b) and (c) show layouts for electromagnetic (EM) simulation using Agilent’s ADS Momentum. Single-layer capacitors including bonding wires are separately simulated using Ansoft HFSS, and they are represented by -parameter data items. The simulated bonding-wire inductance is estimated to be below 0.1 nH. Due

to the low inductance value, the second harmonic transmission zero is close to that without the bonding inductance and can be approximately used as the direct measure for the second harmonic short for class-F operation. Input and output matching networks are simulated using co-simulation with Momentum. In the case of the output matching network, the element values were tuned iteratively to yield the desired impedances at the fundamental, second, and third harmonic frequencies. The dimensions of the spiral inductors and interdigital capacitors were finally determined through tuning based on co-simulation with EM simulations. Conductor thickness and loss were considered in the EM simulation. Fig. 13 shows the final EM simulation results for the output matching circuit. Fig. 13(a) shows the impedance seen from the drain terminal. The fundamental impedance is close to the desired 65 ; however, the loss is observed to appear at the second and third harmonics. Fig. 13(b) shows the transmission characteristic, which is close to the circuit simulation results shown in

3192


TABLE VIII MEASURED PERFORMANCES OF POWER-AMPLIFIER MODULE

Fig. 18. Simulated: (a) output power and PAE for the frequency at the input power of 22.5 dBm and (b) the output power and PAE for the input power at the frequency of 2.5 GHz.

Fig. 7. The transmission zeroes are closely formed at the second and third harmonic frequencies. The insertion loss at the passband is about 0.6 dB. Using the -parameters of the lumped circuit in Table VII and the EM simulated layout, the output power and PAE is simulated using a TriQuint Semiconductor Inc. large-signal model. Fig. 14(a) shows the comparison of the output power and PAE performances of the power amplifiers composed of the lumped elements and EM simulated layouts at the input power of 22.5 dBm for the passband frequency. The output power does not show uniform decrease of about 0.6 dB, which is the passband insertion loss. The decrease of the output power is larger near the lower passband edge. The PAE is accordingly degraded in proportion to the output power decrease. When the output power is 4 W, the 0.6-dB output power decrease causes a PAE decrease of about 11% assuming the same dc power dissipation. Thus, the PAE of the power amplifier with EM 15 . simulated layout shows a decreased PAE of about 10 It should also be noted that the PAE of the lower band edge is poorer compared to that of the higher passband edge. The frequency dependence of the output impedance is the main reason for such PAE trends. The output impedance locus of

Fig. 19. Output spectrum at a 2.5-GHz input power. The spectrum is obtained at an output power of 36 dBm.

higher band edge appears at higher PAE contour, while that of the lower band appears at lower PAE contours. On the contrary, the impedance locus appears within equi-level output power contour. Fig. 14(b) shows the output power and PAE for the input power. This also shows the degradation of PAE and output power due to passband loss. III. FABRICATION AND MEASURED RESULTS Fig. 15 shows a photograph of the fabricated power amplifier. A GaN HEMT chip and substrate were attached to the package using Au/Sn preform. This was the same package used to evaluate the GaN HEMT chip. Single-layer capacitors are attached using silver epoxy. All bonding wires are 4-mil ribbons. The assembled package was mounted on a 50- line test jig. The test jig is connected to the setup shown in Fig. 3 for testing. Fig. 16 shows the measured -parameters for the fabricated amplifiers. The measured -parameters show a slight shift to higher frequency and the gain at the fundamental frequency is


3193

TABLE IX COMPARISON OF MINIATURIZED GaN HEMT POWER AMPLIFIERS

also slightly lowered. Due to this shift, the second transmission zero occurs at a higher frequency. However, the overall frequency response is observed to follow the simulated frequency response. The reason for the shift to a higher frequency was investigated by measuring the -parameters of the output matching circuit. In order to convert the measured data to the frequency response of the output matching network, the drain capacitance and HFSS simulated bonding wire inductance is added and the drain side reference impedance is changed to 65 . Fig. 17 shows the resulting measured transmission characteristic, which shows the shift to higher frequency. The shift to higher frequency was found due to the fabricated interdigital capacitor patterns. The widths of the interdigital capacitor conductor patterns were measured to be reduced by 2 m. We re-simulated considering the reduction of the conductor pattern width in the interdigital capacitors. This is also shown in Fig. 17. The simulated results show good agreement with the measured results. Such errors are found to be within process errors. Fig. 18 shows the simulated and measured output powers and PAEs. The simulation data is for the power amplifier composed of EM simulated layouts. The measured output power at the center frequency is above 36 dBm, and the measured PAE is about 55%. The best performance of power and efficiency are observed to appear at 2.7 GHz rather than at the center frequency. However, for the frequency range of 2.3 2.7 GHz, the output powers are above 36 dBm. The performance are listed and summarized in Table VIII. The output power is the maximum power and the corresponding input power, power gain, and PAE at the maximum power are listed. Fig. 19 shows the harmonics of the fabricated power amplifier. The measured harmonic suppression for the frequency span 2 8 GHz is above 40 dBc for the second and third harmonic frequencies. IV. CONCLUSION We demonstrated the design of a 4-W class-F power amplifier for the WiMAX band, which is suitable for base units. The impedance yielding the maximum power was measured using a specially fabricated tuning jig. A novel lumped-element class-F output matching circuit is proposed. The circuit provides a good harmonic suppression, which is shown to be tolerant to design parameter changes. The fabricated amplifier shows an output power above 36 dBm; efficiency is about 50%, and harmonic suppression is above 40 dBc for the WiMAX frequency band. The size of the amplifier alone is 4.4 mm 4.4 mm.

Table IX shows a comparison of miniaturized GaN HEMT amplifiers. The center frequency of our power amplifier is the lowest, but the size is also the smallest among the hybrid-type power amplifiers. Our power amplifier is especially fabricated using conventional thin-film technology. Other power amplifiers use multilayer ceramic fabrication technology or a highpermittivity substrate for miniaturization. Most of the designs are implemented using distributed elements, except the design in [20], while our design uses lumped elements. The PAE of our power amplifier is found to be considerably good compared with other miniaturized power amplifiers. REFERENCES [1] The IEEE 802.16 Working Group on Broadband Wireless Access Standards, IEEE Standard 802.16, 2009. [Online]. Available: http://ieee802.org/16 [2] U. H. Andre, E. J. Crescenzi, R. S. Pengelly, A. R. Prejs, and S. M. Wood, “High efficiency, high linearity GaN HEMT amplifiers for WiMAX applications,” High Freq. Electron., vol. 6, no. 6, pp. 16–29, Jun. 2007. [3] U. K. Mishra, P. Parikh, and W. Yi-Feng, “AlGaN/GaN HEMTs—An overview of device operation and applications,” Proc. IEEE, vol. 90, no. 6, pp. 1022–1031, Jun. 2002. [4] W. Nagy, S. Singhal, R. Borges, W. Johnson, J. Brown, R. Therrien, A. Chaudhari, A. Hanson, J. Riddle, S. Booth, P. Rajagopal, E. Piner, and K. Linthicum, “150 W GaN-on-Si RF power transistors,” in IEEE MTT-S Int. Microw. Symp. Dig., Jun. 2005, pp. 483–486. [5] F. H. Raab, “Class-F power amplifiers with maximally flat waveforms,” IEEE Trans. Microw. Theory Tech., vol. 43, no. 11, pp. 2007–2012, Nov. 1997. [6] A. V. Grebennikov, “Circuit design technique for high efficiency class F amplifiers,” in IEEE MTT-S Int. Microw. Symp. Dig., 2000, vol. 2, pp. 771–774. [7] S. Chen and Q. Xue, “A class-F power amplifier with CMRC,” IEEE Microw. Wireless Compon. Lett., vol. 21, no. 1, pp. 31–33, Jan. 2011. [8] M. Wren and T. J. Brazil, “Experimental class-F power amplifier design using computationally efficient and accurate large-signal pHEMT model,” IEEE Trans. Microw. Theory Tech., vol. 53, no. 5, pp. 1723–1731, May 2005. [9] K. Kuroda, R. Ishikawa, and K. Honjo, “Parasitic compensation design technique for a C -band GaN HEMT class-F amplifier,” IEEE Trans. Microw. Theory Tech., vol. 58, no. 11, pp. 2741–2750, Dec. 2010. [10] C. Trask, “Class-F amplifier loading networks: A unified design approach,” in IEEE MTT-S Int. Microw. Symp. Dig., 1999, vol. 1, pp. 351–354. [11] J. D. Rhodes, “Output universality in maximum efficient linear power amplifiers,” Int. J. Circuit Theory Appl., vol. 31, no. 4, pp. 385–405, Jul. 2003. [12] M. S. Hashmi, F. M. Ghannouchi, P. J. Tasker, and K. Rawat, “High reflective load–pull,” Microw. Mag., vol. 12, pp. 96–107, Jun. 2011. [13] G. Simpson and R. Pollard, “Automated microwave tuner system simplifies transistor characterization,” in ARFTG Conf., 1987, vol. 11, pp. 66–89. [14] J. Verspecht and D. E. Root, “Polyharmonic distortion modeling,” Microw. Mag., vol. 7, pp. 44–57, Jun. 2006. [15] “TGF2023-01, 6 Watt discrete power GaN on SiC HEMT,” TriQuint Semiconduct. Inc., Hillsboro, OR, 2010. [Online]. Available: http:// triquint.com

3194


[16] S. C. Cripps, RF Power Amplifiers for Wireless Communications. London, U.K.: Artech House, 1999. [17] “Design guidelines, thin-film circuits design rules,” Applied Thin-Film Product (ATP), Fremont, CA, 2007. [Online]. Available: http:// thinfilm.com [18] S. Gao, C. Sanabria, H. Xu, S. I. Long, S. Heikman, U. Mishra, and R. A. York, “MMIC class-F power amplifiers using field-plated AlGAN/GaN HEMTs,” in IEEE Compound Semicond. Integr. Circuit Symp., Nov. 2006, pp. 81–84. [19] K. Yamanaka, Y. Tuyama, H. Ohtsuka, S. Chaki, M. Nakayama, and Y. Hirano, “Internally-matched GaN HEMT high efficiency power amplifier for space solar power stations,” in Proc. Asia–Pacific Microw. Conf., Dec. 2010, pp. 119–112. [20] L. Rui, D. Schreurs, W. De Raedt, F. Vanaverbeke, and R. Mertens, “A low cost compact LTCC-based GaN power amplifier module,” Integr. Nonlinear Microw. Millim.-Wave Circuits, pp. 1–4, Apr. 2011. [21] L. Rui, D. Schreurs, W. D. Raedt, F. Vanaverbeke, J. Das, M. Germain, and R. Mertens, “Integrated AlGAN/GaN HEMTs in MCM-D technology,” in Electron. Compon. Technol. Conf., Jun. 2010, pp. 1562–1567. [22] H. C. Jeong, H. S. Oh, Y. S. Heo, K. W. Yeom, and K. M. Kim, “Design of a GaN HEMT 4 W miniaturized power amplifier module for WiMAX band,” J. Korea Electromagn. Eng. Soc., vol. 22, no. 2, pp. 162–172, Feb. 2011. [23] D. M. Snider, “A theoretical analysis and experimental confirmation of the optimally loaded and overdriven RF power amplifier,” IEEE Trans. Electron Devices, vol. ED-14, no. 12, pp. 851–857, Dec. 1967. [24] A. Grebenikov, “Effective circuit design techniques to increase MOSFET power amplifier efficiency,” Microw. J., pp. 64–72, Jul. 2000. [25] P. Colantonio, F. Giannini, and E. Limiti, “HF class F design guidelines,” in 15th Int. Microw., Radar, Wireless Commun. Conf., 2004, vol. 1, pp. 27–37. [26] B. C. Wadell, Transmission Line Design Handbook. Norwood, MA: Artech House, 1991. [27] E. Cristal, “Tables of maximally flat impedance-transforming networks of low-pass-filter form,” IEEE Trans. Microw. Theory Tech., vol. MTT-13, no. 5, pp. 693–695, Sep. 1965. [28] G. L. Matthaei, L. Young, and E. M. T. Jones, Microwave Filters, Impedance-Matching Networks and Coupling Structures. New York: McGraw-Hill, 1964. Hae-Chang Jeong was born in Cheonan, Korea, in 1984. He received the M.S. degrees in radio science and engineering from Chungnam National University, Daejeon, Korea, in 2010, respectively, and is currently working toward the Ph.D. degree in radio science and egineering at Chungnam National University. His research interests are in the design of hybrid and monolithic microwave circuits and microwave systems.

Hyun-Seok Oh was born in Seoul, Korea, in 1979. He received the M.S. and Ph.D. degrees in radio science and engineering from Chungnam National University, Daejeon, Korea, in 2007 and 2011, respectively. His current research interests are in the design of hybrid and monolithic microwave circuits and microwave systems.

Kyung-Whan Yeom (M’95) was born in Seoul, Korea, in 1957. He received the B.S. degree in electronics from Seoul National University, Seoul, Korea, in 1980, and the M.S. and Ph.D. degrees in electrical engineering from the Korea Advanced Institute of Science and Technology (KAIST), Daejeon, Korea, in 1982 and 1988, respectively. From 1985 to 1991, he was a Principal Engineer with LG Precision. He has been with the Microwave Integrated Circuit (MIC) Team as a Team Leader and involved subsequently in military electronics division for electronic warfare (EW) equipment. From 1991 to 1995, he was with LTI, where he was involved with power amplifier modules for analog cellular phones. In 1995, he joined Chungnam National University, Daejeon, Korea, as an Assistant Professor. He is currently a Professor with the Department of Radio Science and Engineering, Chungnam National University. From 2004 to 2006, he was the editor-in-chief of The Korean Institute of Electromagnetic Engineering and Science. His research interests are in the design of hybrid and monolithic microwave circuits and microwave systems. Prof. Yeom is a member of the Korean Institute of Electromagnetic Engineering and Science (KIEES). He was the recipient of the IR-52 Jang Youg-Sil Prize of the Ministry of Science and Technology (MOST), Korea, for his work on cell phone power amplifier in 1994. He was also the recipient of the KIEES Academic Award from for “design and fabrication of a novel 60 GHz GaAs pHEMT resistive double balanced star MMIC mixer” in 2004.