Experimental Procedure to Optimize Out-Of-Band Terminations for Highly Linear and Power Efficient Bipolar Class-AB RF Amplifiers M. Spirito 1 , M. P. van der Heijden 2 , M. Pelk 1 , L. C. N. de Vreede 1 , P.J. Zampardi 3 , L.E. Larson 4 and J. N. Burghartz 1 1 HiTeC Laboratory, Delft University of Technology, Mekelweg 4, 2628 CD Delft, The Netherlands 2 Philips Research Laboratories, Prof. Holstlaan 4, 5656 AA Eindhoven, The Netherlands 3 Skyworks Solutions, Inc., Newbury Park, CA 91320 USA 4 University of California at San Diego, La Jolla, CA 92093 USA Abstract — An optimization procedure based on load-pull measurements to obtain both highly-linear and highly-efficient class-AB operation is presented. This procedure can be applied without any foregoing device characterization; therefore it is an excellent method to compare the linearity performance of different bipolar technologies. The presented approach provides the optimum out-of-band terminations and quiescent current yielding IM3 improvement of more than 15 dBc at 3 dB back-off compared to traditional design techniques. Optimum power- added efficiency is achieved at the same time. Index Terms — Load-pull, linearity, out-of-band termination. I. INTRODUCTION Class-AB or inverse-class AB amplifier operation is preferred for power amplifiers (PA’s) in linear wireless communication systems, since it offers a workable trade-off between linearity and efficiency. When considering pure class-AB amplifier operation, it is customary to cancel the harmonics at the output of the active device to achieve maximum collector efficiency, ideally 78.5% for pure class B [1]. When considering the input matching conditions of the active device, it has been shown that at low power levels (large power back-off conditions) the use of proper out-of-band terminations at baseband and second harmonic frequencies can yield dramatic improvements in linearity without compromising gain or DC power consumption [2,3,4]. These out-of-band termination techniques applied at the input of a bipolar device are based on the cancellation of the 3 rd -order intermodulation (IM3) products. This is achieved by compensating the IM3 products resulting from the 3 rd -order nonlinearities, by the 3 rd -order (IM3) products which result from secondary mixing of the even-order components with the fundamental over the nonlinear base-emitter junction. Only by proper selection of the input out-of-band terminations at the baseband (BB) and the second harmonic it is possible to control this secondary mixing process so that perfect IM3 cancellation can be achieved. This has successfully been demonstrated for various low power circuits such as LNA’s [2], [3], mixers [4] and a differential driver stage [5]. Recently, the selection of the optimum quiescent current and proper out-of-band terminations was demonstrated theoretically also at higher power levels [6]. An extended Volterra series analysis supported by circuit simulations was used for that purpose [7]. A shortcoming of that approach was, however, that a complete device characterization and model extraction were required prior to the linearity characterization. In this paper we present an extended experimental procedure that facilitates linearity optimization on devices without any previous transistor measurement and modeling. Furthermore, the PAE improvements, close to compression, are greatly enhanced compared to the prior work. In Section II, we discuss the in-house developed active harmonic load-pull system that facilitates an accurate control of the base-band impedance. In Section III we utilize this feature to provide a straightforward experimental linearity optimization procedure for BJT devices. First the optimum quiescent current, base- band and 2 nd harmonic ohmic impedance, are found. These are characteristic for a given bipolar device and provide the highest linearity over a large bandwidth in back-off conditions. Then, while maintaining the high linearity (Sections IV and V), the sweet spot is shifted towards higher power levels. This is done by changing the quiescent current and the 2 nd harmonic termination in a pre-described way. As last point, a relation that links the optimum linearity quiescent current and BB source impedance, is used to reduce the static power dissipation and improve the PAE of the device, not affecting the achieved linearity. II. HARMONIC LOAD-PULL SYSTEM FOR PA OPTIMIZATION The measurement system shown in Fig. 1 facilitates BB DC f 0 Loop 2f 0 Loop DUT a1f0 a12f0 b 1f0 b 12f0 b2f0 b22f0 a 2f0 a 22f0 a 1BB a 1f0 a 12f0 b 1BB b 1f0 b 12f0 a 2BB a 2f0 a 22f0 b 2BB b 2f0 b 22f0 HP 8510 Receiver unit a 1 b 1 a 2 b 2 PSA BB DC f 0 Loop 2f 0 Loop RF Synth ESG Synth High Power Path Calibration Path Fig.1. Simplified block diagram of the implemented active Load-Pull system. 112 0-7803-9309-0/05/$20.00 ©2005 IEEE IEEE BCTM 7.3