Linearization and Efficiency Enhancement of RF Power Amplifiers based on Advanced Measurements and Transistor Modeling
MetadataVis full innførsel
In recent years, the development of more spectrally efficient modulation schemes, has challenged the power amplifier (PA) designers to develop more complex designs. The class AB amplifier has been the "working horse" in PA design for many years, but will not do the job of amplifying complex signals with high efficiency. To design a high performance PA to meet today’s requirements; a more advanced design is required, which includes some kind of linearity- or efficiency technique, or both. According to the author’s interests, this thesis involves mostly practical measurements with advanced measurement setups. A governing idea in this work is to do real-life measurements, and not only rely on theoretical simulations. This involves, of course, a lot of practical challenges, which would have been avoided by doing theoretical experiments. However, valuable experience in RF/microwave design and measurements, together with a better understanding of limitations of theoretical models and assumptions is obtained with this approach. The work presented in this thesis covers several topics involved in a practical power amplifier design. The basic building block in the power amplifier is the transistor. In order to design a high performance power amplifier, either an accurate transistor model is needed or the design can be performed by means of advanced measurements. Both these design strategies are addressed in this work. In the first chapter of this thesis, a motivation for this work is given. A brief summary of the main challenges in power amplifier design is given here, together with a summary of the contents of the thesis and the main contributions of the work. Chapter 2 presents the most relevant background and theory for the work presented in this thesis. Important signal properties for the power amplifier design are presented, and the most common linearity- and efficiency definitions are reviewed. A great range of different linearizationand efficiency enhancement methods exist, with varying complexity and enhancement potentials. A few of these methods which are relevant for the work presented in the thesis are reviewed in this chapter. The most important experiences gained from the practical measurements are presented in chapter 3. For the author, the ability to control instruments and acquire measurement data from Matlab represented a big breakthrough in how to make advanced measurements effectively. A brief introduction to remote control and automatic measurements is given here. A few guidelines regarding the design of calibration standards in order to characterize transistors are also presented. Traditional load-pull measurements and extension including time-domain measurements are also covered in this chapter. The basic formulas for measurements with instrumentation for time-domain measurements are presented, and also a method to ensure good quality of the measurements performed by this instrument. Finally, some methods for advanced simulation of a PA using Advanced Design System (ADS) are suggested. In chapter 4 an introduction to the transistor modeling concept is given. The convergence issues in nonlinear simulations of transistors in older simulation software was the motivation for developing own and more robust transistor models. Two different transistor models are used to model an HBT transistor, VBIC and FBH. Parameters for these models are extracted based on standard DC- and RF measurements. It is shown that the FBH model, which is a model created for HBTs, more accurately predicts the s-parameters and a single-tone power sweep, than the more general VBIC model. A detailed procedure for parameter extraction of a discrete pHEMT transistor based on the Angelov model is also given. The procedure is based on existing methods. Good agreement between measurements and simulations is achieved. Different linearity- and efficiency enhancement techniques are analyzed in chapter 5 based on simulations and measurements. An active bias circuit is suggested to linearize a MMIC HBT power amplifier. The very simple on-chip method improves the linearity of a class A PA with minimum of extra area consumption. The rest of the work presented in this chapter is based on evaluation of a discrete 1 W pHEMT transistor in a different power amplifier designs, and the investigation of important power amplifier design issues. A measurement setup for power amplifier design based on measurements is developed. The transistor is mounted on an in-house test card and inserted into the measurement setup. The effect of varying impedances at baseband is well known in the literature to degrade the linearity in a PA. A method for measuring the baseband effects by modifying the bias networks without affecting the RF impedance is developed. Memoryless digital predistortion (DPD) is included in the setup and it is demonstrated how a simple memoryless predistorter cannot compensate for the memory effects introduced by the bias network. The choice of bias- and load conditions for a power amplifier with different linearity criteria, with and without digital predistortion (DPD), is investigated. It is shown that in order to design a high performance power amplifier the application, and whether DPD will be implemented or not, should be taken into consideration as early as possible in the design process. A method for device characterization for supply-modulation techniques based on measurements and post-processing of data is developed. The device is evaluated for envelope-tracking and shows promising potential to amplify complex signals with high efficiency. An overall power added efficiency (PAE) of 59.5 % at 25.7 dBm output power, assuming 80 % efficiency of an ideal envelope amplifier, could be obtained by the use of envelope tracking for a 16 QAM signal with peak-to-average power ratio (PAPR) of 6.41 dB. The maximum PAE for constant supply voltage is 37.7 % for the same signal and the same allowed distortion. The characterization method can also be easily extended to load-modulation applications. Finally a method for estimating transistor output parasitics and intrinsic waveforms by time-domain waveform measurements and load-pull is suggested. This very simple method seems to give a good estimate of the intrinsic waveforms of a 10W GaN transistor. To prove the concept the load is tuned for class F and inverse class F operation based on the intrinsic waveforms. The resulting efficiencies were measured to 78.5 % and 84.5 %, respectively and are comparable to the highest efficiencies reported for this kind of device.