Class E and Class D -1 GaN HEMT Switched-Mode Power Amplifiers J. A. GARCÍA *, R. MERLÍN *, M. FERNÁNDEZ *, B. BEDIA *, L. CABRIA *, R. MARANTE *, T. M. MARTÍN-GUERRERO ** *Departamento Ingeniería de Comunicaciones Universidad de Cantabria 395, Santander SPAIN **Departamento Ingeniería de Comunicaciones Universidad de Málaga 2971, Málaga SPAIN Abstract: - This paper deals with the design of transmission line class E and class D -1 GaN HEMT power amplifiers (PAs) to be used in polar transmitters. The Vdd-to-AM and Vdd-to-PM characteristics have been measured, being the observed nonlinearities related to device parameters. Measured results under a two-tone excitation support the need for predistorting functions. Key-Words: - Efficiency, nonlinear distortion, polar transmitters, power amplifiers 1 Introduction Polar transmitters are standing as an attractive emerging architecture for simultaneously providing high linearity and efficiency figures. An appropriate design of the high-level modulating stage, a switched-mode PA, is critical for obtaining the desired performance. Polar transmitter residual distortion may be due to both, architecture and circuit nonidealities. A pioneer work on the subject is due to Raab [1], who considered the impact of the finite AM modulator bandwidth and the differential delay between the amplitude modulating component and the phase modulated RF carrier. Finite AM modulator bandwidth was there treated in an approximate way, assuming an ideal brick-wall reconstruction filter, while the differential delay between the AM and PM paths was considered a constant and independent quantity respect to the baseband envelope frequency. However, Raab also recognized the existence of other sources of distortion associated to the switching mode modulating stage: the Vdd-to-AM and Vdd-to-PM characteristics (a notation proposed later in [2]). Milosevic et al. in [3] extended the finite AM bandwidth analysis, considering a general reconstruction filter. The distortion caused by the PA nonlinearities, together with the differential delay, were then experimentally investigated in [4]. Recently, a comprehensive analysis of all these nonlinear distortion generation mechanisms has been proposed in [5], including analytical models for the AM-AM and AM-PM nonidealities. In this paper, the design of transmission line class E and class D -1 GaN HEMT power amplifiers (PAs) is presented. The Vdd-to-AM and Vdd-to-PM characteristics have been experimentally extracted, being the observed nonlinearities related to device parameters. The need for introducing predistorting functions is justified by modulation results under a classical two-tone linearity test. 2 Transmission Line Class E PA A class E PA was first designed at the 9 MHz frequency band, following the transmission line topology proposed in [6]. A 15W GaN HEMT from Cree was selected, the CGH3515F, a device conceived for WiMAX class AB amplifier applications. ISSN: 179-2769 37 ISBN: 978-96-474-7-9
The optimum load impedance value at the fundamental was in a first approach computed as in Eq. 1, while open circuit conditions were assured at the second and third harmonics. When the drain bias voltage is lowered close to V, a residual RF carrier leakage exists together with a significant parasitic phase modulation. 4 Z opt.28 j49 = e ω C out (1) 35 3 25 In the search for the highest power added efficiency (PAE), further Z opt refinements were made through load pull simulations. AM [V] 2 15 1 5 Pin=2dBm Pin=25dBm In Fig.1, a photograph of the implemented PA is shown. The drain biasing path was prepared for inserting a low frequency envelope signal, aimed to modulate in amplitude the PM carrier. A power added efficiency (PAE) value of 71% was obtained at P out = 15W, with the device biased at V GS = -2.8V and V DS = 28V. (a) PM [ ] 5 1 15 2 25 3 V [V] DD -1-2 -3-4 -5-6 -7-8 Pin=2dBm Pin=25dBm (b) -9 5 1 15 2 25 3 V [V] DD Fig.2. (a) Vdd-to-AM and (b) Vdd-to-PM static characteristics. Fig.1. Photograph of the implemented class E PA, based on Cree CGH3515 GaN HEMT. 3 Characterizing the Vdd-to-AM and Vdd-to-PM Nonlinearities Using the traditional experimental procedure, with the PA excited by a CW signal, the drain biasing voltage was varied from to 3V. The evolution of the output signal amplitude and phase components was measured. In Fig. 2, those characteristics may be observed for two different input power levels. At a moderate excitation, P in = 2dBm, saturation of the amplitude characteristic is observed for high V dd values. This behavior, associated to the device abandoning the switching mode of operation, is reduced for a larger drive level. 4 Transmission Line Class D -1 PA A class D -1 current mode power amplifier [7] was also designed in the above mentioned frequency band, using in this case two units of the same GaN HEMT, the CGH3515F from Cree referred above. In Fig. 3, details of the amplifier implementation, using a Tx line topology, are presented: Fig.3. Photograph of the implemented class D -1 PA, based on Cree CGH3515 GaN HEMT. ISSN: 179-2769 371 ISBN: 978-96-474-7-9
The optimum load impedance at the fundamental was obtained through load pull simulations, while open and short circuit conditions were set at the second and third harmonic respectively, combining transmission line output matching networks and an Anaren balun. Under similar biasing conditions, V GS = -2.8V and V DS = 28V, an output power level of 22.9W was measured with a PAE of 64%. The experimental procedure for extracting the Vddto-AM and Vdd-to-PM characteristics was also employed with this PA. The measured results for two RF excitation power levels are shown in Fig. 4: 5 Device Factors Limiting the Modulation Linearity. Different works have dealt with the origins of the Vdd-to-AM and Vdd-to-PM nonlinearities. 5.1. Feedthrough In [8], Raab gave some insight into the amplitude-modulation characteristics of a BJT amplifier. The feedthrough was there associated to the capacitive coupling between the base and collector, which produces a minimum output signal that is present even when V CC = V. The amount of carrier leakage could be estimated by considering it to be the result of applying the fundamental frequency component of the baseemitter voltage to a series arrangement of the base-collector capacitance and the collector load impedance. Recently, a comprehensive study was employed to accurately describe the Vdd-to-AM and Vddto-PM profiles of a PHEMT based class E PA [5]. The nonlinear gate-to-drain capacitance in the device equivalent circuit was shown to be responsible for the observed behavior. (a) (b) Fig. 4. (a) Vdd-to-AM and (b) Vdd-to-PM static characteristics with V GS =-2.8V. The expected nonidealities in the amplitude modulation process, associated to feedthrough and the operation in current source mode, may be perfectly observed. However, for a very high excitation level, P in = 36dBm, a residual nonlinearity still remains. In both implemented GaN HEMT switching mode PA, the measured amplitude and phase characteristics may also be reproduced with the aid of C gd. Since the feedthrough is coupled through a reactance that is generally larger than output load resistance, the feedthrough signal is generally in phase-quadrature with the amplified component at the output. This causes a phase variation in the carrier (Vdd-to-PM conversion) close to 9º at low modulating levels. 5.2 Vdd-to-AM saturation In [5], the saturation of the AM/AM curve for high biasing voltages has been associated to the device I/V characteristic. For a linear amplitude modulation, the maximum drain biasing voltage and the input power level should be selected to guarantee that the drain current gets saturated in the transition to the linear region. As the saturation value is almost linearly proportional to V DD for a fixed load, the sought linear modulation would be assured. ISSN: 179-2769 372 ISBN: 978-96-474-7-9
If the RF excitation is not enough or the drain biasing voltage too high, the case of P in = 2 dbm in Fig. 2, the device operation may get into the saturated region where it works in current source mode. In such operating condition, the drain current does not linearly follow the drain voltage, but it tends to a constant. 5.3 Secondary factors Other factors, as the ON state resistance variation with drain biasing voltage, R on (V DD ), may also lead to deviations in the Vdd-to-AM characteristic. This variation is related to the existence of device frequency dispersion effects, due to self-heating or trapping, and may be responsible for the remaining nonlinearity observed in Fig. 4 at a high excitation level. 6 IMD Behavior under TWO-TONE Excitation In order to describe the impact of the above characterized Vdd-to-AM and Vdd-to-PM nonlinearities, on the IMD performance of a polar transmitter, the measurement system of Fig. 5 was implemented. Two vector signal generators, controlled from Matlab, allowed exciting the class E GaN HEMT PA with a phase modulated carrier and the modulating amplitude component. A linear stage was used in order to accommodate the envelope signal to the required levels at drain terminal for an optimum modulation performance of the switching stage. V GS PM Modulated CW Fig.5. AM Component RF in Class E PA Envelope amplifier v DD (t) DC RF out Spectrum Analyzer Vector Signal Analyzer Implemented characterization system. Using the classical linearity test, with two tones at a frequency spacing of 1 khz, the output spectrum was captured with the aid of a vector signal analyzer. P out [dbm] 4 3 2 1-1 -2-3 -4 898 898.5 899 899.5 9 9.5 91 91.5 92 f [MHz] Fig.6. Measured output spectrum, as captured from a vector signal analyzer under two-tone excitation. As observed from Fig. 6, a carrier to intermodulation distortion (CIMD) rejection of more than 3 db was obtained, mainly due to the nearly linear AM characteristic. Using a slightly higher input power level and incorporating predistorting functions, both for the phase and amplitude components, this result could be enhanced. 7 Conclusion An experimental procedure has been proposed for accurately characterizing the Vdd-to-AM and Vdd-to- PM nonlinearities in a class E PA, designed over a 15W GaN HEMT from Cree. The observed nonlinearities may be also described using the device mechanisms discussed in [5], and the extracted derivatives may help optimizing the operating conditions for linear modulation. Measurement IMD results of a polar transmitter architecture support the need for correcting the modulating stage nonlinearities. 8 Acknowledgement The authors want to thank the support received from MEC and CICE-JA through Projects TEC25-7985-C3-1 and P7-TIC-2649. References: [1 F. H. Raab, Intermodulation distortion in Kahntechnique transmitters, IEEE Trans. Microwave Theory and Tech., Vol. 44, No. 12, pp. 2273-2278, Dec. 1996. [2] A. Dient, C. Berland, M. Villegas and G. Baudoin, EER Architecture Specifications for OFDM Transmitter Using a Class E Amplifier, ISSN: 179-2769 373 ISBN: 978-96-474-7-9
IEEE Microwave Wireless Comp. Lett., Vol. 14, No. 8, pp. 389-391, Aug. 24. [3] D. Milosevic, J. van der Tang and A. van Roermund, Intermodulation products in the EER technique applied to class-e amplifiers, Int. Symp. on Circuits and Syst. Dig., vol. I, pp. 637-64, Vancouver, May 24. [4] N. Wang, X. Peng, V. Yousefzadeh, D. Maksimovic, S. Pajic, and Z. Popovic, Linearity of X-Band class-e power amplifiers in EER operation, IEEE Trans. Microwave Theory and Tech., Vol. 53, No. 3, pp. 196-112, Mar. 25. [5] J. C. Pedro, J. A. García, and P. M. Cabral, Nonlinear Distortion Analysis of Polar Transmitters, IEEE Trans. Microwave Theory and Tech., Vol. 55, No. 12, pp., Dec. 27. [6] T. B. Mader, Z. Popovic, The Transmission-Line High Efficiency Class-E Amplifiers, IEEE Trans. Microwave Theory and Tech., Vol. 43, No.5, pp.29-292, Sept. 1995. [7] H. Kobayashi, J. M. Hinrichs, and P. M. Asbeck, Current-mode Class-D Power Amplifiers for High-Efficiency RF Applications, IEEE Trans. Microwave Theory and Tech., Vol. 49, No.12, pp.248-2485, Dec. 21. [8] H.L. Krauss, C.W. Bostian, F.H. Raab, Solid State Radio Engineering, John Wiley & Sons, 198. ISSN: 179-2769 374 ISBN: 978-96-474-7-9