LINEARIZATION OF A 3.7 GHz MULTI-CARRIER GaN HEMT DOHERTY POWER AMPLIFIER USING DIGITAL PREDISTORTION METHOD

Transcription

1 REFERENCES 1. A. Fard, Phase noise and amplitude issues of a wide-band VCO utilizing a switched tuning resonator, IEEE Int Symp Circ Syst (2005), S. Wu, A low-noise fast-settling PLL frequency synthesizer for CDMA receivers, International Symposium on System on Chip, 2004, pp K. Ang, M. Chia, and D. Li, A process compensation technique for integrated VCO, IEEE Radio Frequency Integrated Circuits Symposium, Fort Worth, TX, 2004, pp P. Sjöblom and H. Sjöland, An adaptive impedance tuning CMOS circuit for ISM 2.4 GHz band, IEEE Trans Circ Syst 52 (2005), J. Holland, Adaptation in natural and artificial systems, MIT Press, Cambridge, MA, A. Somani, P.P. Chakrabarti, and A. Patra, An evolutionary algorithm-based approach to automated design of analog and RF circuits using adaptive normalized cost functions, IEEE Trans Evol Comput 11 (2007), S. Kaitharam, C. Rajagopal, and A. Nunez, SRFCC: Synthesis of RF CMOS circuits, IEEE Midwest Symp Circ Syst (2002), II/ 497 II/ H.-W. Chiu, Y.-C. Chen, and S.-S. Lu, Optimization of CMOSintegrated LC oscillators using the genetic algorithm, Microwave Opt Technol Lett 42 (2004), M. Laumanns, L. Thiele, K. Deb, and E. Zitzler, Archiving with guaranteed convergence and diversity in multi-objective optimization, Genetic and Evolutionary Computation Conference, 2002, pp E. Pires, P. Oliveira, and J. Machado, Multi objective MaxiMin sorting scheme, Lecture Notes in Computer Science, Vol. 3410, Springer, New York, NY, 2005, pp L. Mendes, E. Pires, J. Vaz, and M. Rosário, Automated design of radio-frequency single-ended switched capacitor arrays using genetic algorithms, IEEE Int Midwest Symp Circ Syst (2007), K. Deb, Multi-objective optimization using evolutionary algorithms, Wiley, New York, NY, B. Razavi, Design of analog CMOS integrated circuits, McGraw- Hill, New York, K. Christensen, Design and optimization of cmos switches for switched tuning of lc resonators, IEEE Int Symp Circ Syst (2001), Z. Li and K.K. O, A low-phase-noise and low-power multiband CMOS voltage controlled oscillator, IEEE J Solid State Circ 40 (2005), VC 2010 Wiley Periodicals, Inc. LINEARIZATION OF A 3.7 GHz MULTI-CARRIER GaN HEMT DOHERTY POWER AMPLIFIER USING DIGITAL PREDISTORTION METHOD extracted using a conventional least mean square error algorithm. Using a two carrier signal based on IEEE802.16e, whose PAR is 9.44 db, an ACLR improvement of db is achieved to give an ACLR level of dbc at an offset of 4.79 MHz. The overall efficiency of the Doherty power amplifier is 13.74%, which represents a 2.44% improvement compared with that of a balanced class-ab power amplifier at an average output power of 36 dbm. VC 2010 Wiley Periodicals, Inc. Microwave Opt Technol Lett 52: , 2010; Published online in Wiley InterScience ( DOI /mop Key words: power amplifier; Doherty amplifier; GaN HEMT; digital predistortion; linearization 1. INTRODUCTION Both linearity and efficiency are important performance measures for RF power amplifiers, because the performance of the transmitter system is determined by that of the power amplifier [1 6]. Especially, for the upcoming 4th generation wireless systems, it is necessary to use a higher peak-to-average power ratio (PAR) and higher frequency operation of the digital-modulated signals. Therefore, it becomes harder to secure very good linearity and efficiency at the same time. As it has a high efficiency over a wide output power range, the Doherty power amplifier has been widely used for various wireless communication systems [1, 2]. However, the Doherty amplifier needs to be linearized using additional circuits or systems, such as feedforward or digital predistortion methods, due to the stringent linearity requirements of the base transceiver or relay systems. The digital predistortion method has advantages in comparable linearization techniques due to its low cost implementation and high efficiency of operation compared with the feedforward technique [4, 5]. In this article, high efficiency Doherty power amplifier based on GaN HEMTs for 3.7 GHz 4th generation wireless systems is implemented and its performance compared before and after linearization; a conventional class-ab amplifier has also been built using the same device configuration [6]. Full simulation procedures and their results are included for the digital predistortion. The forward and reverse polynomial models for both the amplitude-to-amplitude (AM AM) and amplitude-to-phase (AM PM) characteristics of the power amplifier have been built based on the measured baseband input and output signals of the power amplifier. The reverse polynomial models for both the AM AM and AM PM characteristics are applied for the digital predistorter. Simulations have been carried out for the digital predistortion to predict the linearization performance for the implemented Doherty amplifier in MATLAB. Jonghyuk Jeong, Juho Van, Jaeyong Cho, Min-Su Kim, Hanjin Cho, Kyung-Hoon Lim, Sung-Wook Kwon, Kyonggon Choi, Hyung-Chul Kim, Sungcheol Yoo, Cheon-Seok Park, and Youngoo Yang School of Information and Communication Engineering, Sungkyunkwan University, Suwon, Korea; Corresponding author: yang09@skku.edu Received 17 August 2009 ABSTRACT: In this article, a 3.7 GHz band Doherty amplifier based on GaN HEMTs, which is linearized using a digital predistortion method, is implemented for 4th generation wireless communication systems. The forward and reverse models of the Doherty amplifier for the predistortion consist of simple polynomials whose coefficients are Figure 1 A schematic diagram of the Doherty power amplifier. [Color figure can be viewed in the online issue, which is available at 634 MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 52, No. 3, March 2010 DOI /mop

2 Figure 2 Performances of the implemented balanced class-ab and Doherty amplifiers: (a) IMD3 and (b) PAE. [Color figure can be viewed in the online issue, which is available at Based on the simulation results, the experimental setup has been built for verification. The new digital predistortion method that has been developed has been successfully realized and applied to the implemented Doherty and class-ab amplifiers discussed here. The experimental results after applying the digital predistortion are presented for the Doherty and the balanced class-ab amplifiers that have been produced. 2. DOHERTY POWER AMPLIFIER We designed the GaN HEMT Doherty power amplifier module for the 3.7 GHz band. Figure 1 shows a schematic diagram of the implemented Doherty power amplifier. The power amplifier module consists of two blocks. The first is the drive stage, which comprises an ERA-5SM, AH-1, FP2189, and CGH40010F GaN HEMT whose P 3dB is 10 W. The second is the main stage, in which two CGH40025F GaN HEMTs, whose P 3dB is 25 W, are in parallel. The amplifier has been evaluated with two operation modes: the balanced amplifier and Doherty amplifier modes. For the balanced amplifier mode, the carrier and peaking amplifiers have the same bias condition as class-ab (I DQ ¼ 250 ma). For the Doherty amplifier mode, the peaking amplifier is biased to an experimentally optimized class-c (I DQ ¼ 0 A) point for proper load impedance modulation [1]. The overall performances of the power amplifier are measured under two-tone excitation with a center frequency of 3.7 GHz and a tone spacing of 5 MHz. The amplifier has an overall gain of more than 54 db. The Doherty amplifier mode exhibits 0.5 db higher output power (42.5 dbm vs. 43 dbm) and 3.5% better efficiency (26% vs. 29.5%) at the given IMD3 level of 30 dbc, as shown in Figure 2. A photograph of the implemented power amplifier, whose size is mm 2, is shown in Figure SIMULATION FOR DIGITAL PREDISTORTION Among the various digital predistortion approaches that are available, the polynomial digital predistortion method is adopted for linearization using simple behavioral models of the power amplifier. In the forward modeling, the power amplifier is modeled based on the measured input and output baseband in-phase (I) and quadrature(q) signals, which are converted to their polar forms to build the dynamic AM AM and AM PM characteristics [4]. The digital predistortion function can be obtained using reverse modeling, as shown in Figure 4 [4, 5]. A simple memoryless polynomial has been used for both the forward and reverse modeling using two carrier orthogonal frequency Figure 3 A photograph of the implemented Doherty power amplifier. [Color figure can be viewed in the online issue, which is available at Figure 4 Digital predistortion using forward and reverse modeling method. [Color figure can be viewed in the online issue, which is available at DOI /mop MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 52, No. 3, March

3 Figure 5 Measured (grey) and modeled (black) responses for the Doherty amplifier: (a) forward AM-AM, (b) forward AM-PM, (c) reverse AM-AM, and (d) reverse AM-PM characteristics division multiplexing (OFDM) signal based on IEEE802.16e as the input signal. The forward and reverse models of the power amplifier module are expressed as follows: AM F ða in;amplitude Þ¼A out;amplitude ¼ XNAF a n A n in;amplitude ; (1) PM F ða in;amplitude Þ¼A out;phase A in;phase ¼ PF 0 þ XNPF b n A n in;amplitude ; (2) AM R ða out;amplitude =GÞ ¼A in;amplitude ¼ XNAR c n ða out;amplitude =GÞ n ; PM R ða out;amplitude =GÞ ¼A in;phase A out;phase (3) ¼ PR 0 þ XNPR d n ða out;amplitude =GÞ n ; (4) where AM F and PM F are the forward polynomials, and AM R and PM R are the reverse polynomials for the amplitude and phase responses, respectively, N AF, N PF, N AR, and N PR are the orders of each polynomial and G is the forward gain of power amplifier. The modeled results are shown in Figure 5, where Figures 5(a) and 5(b) show the modeling results of the forward AM AM and AM PM models for the implemented Doherty amplifier, respectively. Figures 5(c) and 5(d) present the modeling results for the reverse AM AM and AM PM models of the implemented Doherty amplifier, respectively. These reverse models can be directly used as a predistorter as shown in Figure 4. For the MATLAB simulations, the linearized output amplitude and phase signals are obtained as: A out;amplitude ¼ AM F ðam R ðpd in;amplitude ÞÞ; (5) A out;phase ¼ PM F ða in;amplitude ÞþPM R ðpd in;amplitude ÞþPD in;phase ; (6) 636 MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 52, No. 3, March 2010 DOI /mop

4 Figure 6 Simulated linearization results using digital predistortion: (a) AM-AM and (b) AM-PM characteristics where PD in,amplitude and PD in,phase are the linear input amplitude and phase signals, which are scaled corresponding to the predistortion function. The simulated AM AM and AM PM characteristics after the digital predistortion are compared with the original measured AM AM and AM PM characteristics of the implemented Doherty amplifier in Figure 6. The AM AM and AM PM characteristics can be transformed to the power-spectral density (PSD) using the Welch function, provided in the MATLAB, as shown in Figure 7. Finally, the overall procedure for the modeling and linearization of the power amplifier is summarized as a flowchart in Figure 8. For synchronization between the measured input and output signals of the power amplifiers, the signal generator and the spectrum analyzer are connected with a 10 MHz reference signal and event trigger ports. A residual delay difference can be 4. EXPERIMENTAL RESULTS For both the balanced and Doherty operation modes, digital predistortion linearization is applied. The digital predistortion setup has been built as shown in Figure 9. A two carrier OFDM signal based on IEEE802.16e has been downloaded to a signal generator, viz. Agilent s E4438C, and the raw I/Q data have been extracted using a combinational setup consisting of a spectrum analyzer, E4440A, and a vector signal analyzer, Figure 8 The procedure of the modeling and linearization. [Color figure can be viewed in the online issue, which is available at Figure 7 Simulated PSDs before and after linearization Figure 9 The experimental setup for the modeling and digital predistortion. [Color figure can be viewed in the online issue, which is available at DOI /mop MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 52, No. 3, March

5 Figure 10 Measured PSDs of the implemented power amplifiers: (a) PSDs for the balanced class-ab amplifier and (b) PSDs for the Doherty amplifier. [Color figure can be viewed in the online issue, which is available at adjusted using the cross-correlation function in MATLAB. Using a two carrier OFDM signal based on IEEE802.16e, more than 20,000 samples of the input and output baseband I/Q signals are downloaded from the instruments for the linearization simulation. The forward and reverse models are obtained using a 7thorder polynomial for AM F, 5th-order polynomial for PM F, 11thorder polynomial for AM R, and 5th-order polynomial for PM R. Power amplifiers for each operation mode were modeled for the sake of comparison. Finally, the predistorted I/Q data have been downloaded to the E4438C and applied to the power amplifier. Figure 10 shows the PSDs for both operation modes. The improvement in the ACLR performance of the balanced power amplifier is dbc with a PAE of 11.3% at an average power of 36 dbm (5.51 db improvement). The ACLR performance of the Doherty power amplifier is dbc with a PAE of 13.74% at the same average power of 36 dbm (10.74 db improvement). 5. CONCLUSION A Doherty amplifier module has been built using GaN HEMTs for 4th generation wireless communication systems operating in the 3.7 GHz band. For two-tone excitation, the Doherty amplifier shows both better ACLR and PAE at the same time than the conventional balanced class-ab amplifier. To linearize the implemented Doherty amplifier, a predistortion technique based on the indirect learning method has been adopted. Simple polynomial models for the forward and reverse AM AM and AM PM characteristics have been built using the measured baseband input and output signals. Then, the reverse model is used as the predistorter. The simulation procedure has been described and its setup is implemented within the MAT- LAB environment; it is hoped that via this form of simulation setup, the linearization performance can be quickly and easily predicted. For the two carrier OFDM signal based on IEEE802.16e, the implemented Doherty power amplifier exhibits a 2.44% higher PAE than the balanced power amplifier at an average output power of 36 dbm. An ACLR performance of dbc at 4.79 MHz offset has been achieved using the simple digital predistortion method. This results show that the Doherty power amplifier and digital predistortion method can be a promising combination to enhance the efficiency and linearity for the upcoming 4th generation communication systems. ACKNOWLEDGMENTS This work was supported by the Korea Science and Engineering Foundation (KOSEF) grant funded by the Korean government (MOST) (No. R ). REFERENCES 1. Y. Yang, J. Cha, B. Shin, and B. Kim, A fully matched N-way Doherty amplifier with optimized linearity, IEEE Trans Microwave Theory Tech 51 (2003), S. Jung, Y. Xi, H. Park, S. Kwon, J. Van, K. Lim, M. Kim, H. Cho, J. Jeong, and Y. Yang, An optimized Doherty power amplifier using an unequal quadrature input splitter, Microwave Opt Technol Lett 50 (2008), Y. Yang, J. Yi, J. Nam, B. Kim, and M. Park, Measurement of two-tone transfer characteristics of high-power amplifiers, IEEE Trans Microwave Theory Tech 49 (2001), C. Eun and E.J. Powers, A new Volterra predistorter based on the indirect learning architecture, IEEE Trans Signal Process 45 (1997), S. Hong, Y.Y. Woo, J. Kim, J. Cha, I. Kim, J. Moon, J. Yi, and B. Kim, Weighted polynomial digital predistortion for low memory effect Doherty power amplifier, IEEE Trans Microwave Theory Tech 55 (2007), J. Jeong, J. Van, J. Cho, M. Kim, H. Cho, K. Lim, S. Kwon, K. Choi, H. Kim, S. Yoo, and Y. Yang, A 3.7 GHz GaN HEMT Doherty power amplifier using digital predistortion linearization, Presented at Asia-Pacific Microwave Conf. Dig., CD-ROM, Hong Kong, Dec VC 2010 Wiley Periodicals, Inc. 638 MICROWAVE AND OPTICAL TECHNOLOGY LETTERS / Vol. 52, No. 3, March 2010 DOI /mop