System agility and increased bandwidth – one focus at the European Microwave Week 2017
Fig 1: Supply voltage against time for a two-level class-G modulation with increased minimum pulse width, maintaining signal integrity (top). Peak-to-average power ratio for low- and high-voltage power supply (PS) level and measured efficiency (without class-G modulator driver losses) vs. relative modulator switching frequency for an OFDM signal (bottom)
Fig 2: Reconfigurable packaged GaN amplifier using thin-film BST and GaN-HEMT: Schematic (top) and photograph of the PA (bottom)
Fig 3: Schematic of the proposed pulsed RF measurement system including ET system used for demonstration (top), comparison of achievable ENOB for an 8 bit and a 14 bit ADC in the DPD paper. Lower values are without averaging and oversampling, higher values with optimum averaging and oversampling
FBH’s RF Power Lab was well represented with four papers at this year’s European Microwave Week (EuMW2017) in Nuremberg. All presented papers show the direction towards future more agile telecom-systems, operating at much larger instantaneous bandwidth and at a more limited power budget.
The first work deals with class-G, i.e. discrete level supply modulated systems. It investigates how efficiency improvement achievable through this approach depends on maximum switching frequency of the supply modulator. As could be expected for high modulation bandwidths, a fast class-G modulator is desirable in order to achieve maximum efficiency enhancement. However, for class-G, contrary to continuous envelope tracking (ET) systems, it is possible to operate the modulator at switching frequencies far below the modulation bandwidth and still achieve a significant efficiency enhancement, as illustrated in Fig. 1 (top) [1]. The results show that for systems with two supply voltage levels no further significant improvement in efficiency is obtained with modulator switching frequencies higher than 85% of the modulation bandwidth; see Fig. 1 (bottom). For lower frequencies, the efficiency decreases almost linearly. Furthermore, by increasing the amount of supply voltage levels from two to four at constant switching speed, it is found that the bandwidth at constant efficiency can be increased by a factor of 3.4. In combination with digital predistortion, no significant influence of the switching frequency on linearity is observed. The results are very promising for the applicability of the class-G topology in future 5G systems with large instantaneous modulation bandwidth. This work is an extension of a project financed by the DFG.
The second topic presented concerns agile systems based on a reconfigurable packaged GaN power amplifier, where the tuning elements are based on barium–strontium-titanate (BST) thin-film varactors. The schematic of the amplifier (top) and the amplifier mounted in the customized package (bottom) are shown in Fig. 2 [2]. By tuning the input and output matching networks, several advanced functionalities, e.g. efficient back-off operation, frequency agility, VSWR protection, and post tuning of mismatched loads can be realized. Two prototypes were manufactured and measured on a probe station and in a microstrip test fixture. The amplifier was characterized in the frequency range from 1 to 2.7 GHz for varactor tuning voltages up to 24 V and delivered a maximum output power of 37.4 dBm and 58% PAE. Using the available degrees of freedom allows configuring the packaged power amplifier (PA) for a variety of applications and even changing situations during operation. A designer only needs to provide drain and gate supply for the packaged PA and the four control voltages of the varactors to obtain a very compact 50 Ω matched system which can be adapted to any environment just by changing the control voltages. An excellent linearity of the thin-film BST varactors is demonstrated, since the tunable PA shows no difference in IMD compared to a bare GaN HEMT cell over the entire tuning range. Thus, using BST varactors allows realizing very flexible and compact PAs, which are capable of fulfilling the requirements of present and future communication systems.
Two further papers presented at EuMW document that measurement technology for advanced PA systems is gaining in importance. The first work covered a pulsed RF characterization of a buck-converter-based ET system. In contrast to continuous wave (CW) measurements, the proposed method allows ET system characterization over its full dynamic range without exceeding the safe operating area with regard to dissipated power. It includes dynamic characterization of the individual subsystems, i.e., the envelope amplifier (EA) and the RF PA with all signals at their interface. This involves also nonlinear and secondary effects like EA load impedance shift. Applicability of the test setup is demonstrated by characterizing a 57 W pulsed peak output power ET system based on a reverse buck converter and a floating-ground RF PA operating at 1620 MHz. For this ET system it is shown that using pulsed RF measurements the available source power can be increased by 8 dB compared to CW measurements. Fig. 3 (top) illustrates the proposed measurement system [3].
The second work in this field covered benchmarking of RF measurement systems for digital predistortion. In this study, iterative learning control (ILC) linearization is used to characterize the performance of a measurement system for modulated measurements and digital predistortion of RF power amplifiers. Without such linearization it is not possible to evaluate the dynamic range and effective bit resolution (ENOB) of the measurement system because of nonlinear and linear distortions. ILC allows linearization of periodic signals without requiring a predistorter model. With this information it is possible to determine in advance how much signal processing, e.g. averaging and oversampling of measurements, will be necessary to fulfill the linearity requirements. It is shown that applying signal processing and ILC the effective resolution of a system using an 8 bit analog-to-digital converter on the receiver side can be increased by 4.3 bit without the need for additional hardware. Besides that it is found that special attention has to be devoted to the receiver nonlinearities to avoid projecting the nonlinearities to the device under test. A comparative investigation of a nominal 14-bit system and the evaluated 8-bit method was conducted, see Fig. 3 (bottom).
Overall, EuMW shows that the work in the RF Power Lab at FBH is second to none in this field, paving the way for highly efficient wideband modulated RF systems based on discrete supply modulation.
References - EuMW 2017
[1] The Efficiency/Bandwidth Trade-Off in Class-G Supply-Modulated Power Amplifiers
[2] Reconfigurable Packaged GaN Power Amplifier Using Thin-Film BST Varactors
[3] Pulsed RF Characterization of Envelope Tracking Systems for Improved Shaping Function Extraction
[4] Benchmarking of RF Measurement Systems for Digital Predistortion Using Iterative Learning Control