Abstract

The benefits of rapid wireless communication development are continuously being enjoyed by people all over the world. However, the increasing number of wireless applications leads to surge in demand of spectrum resources which conflicts with the limited band resources. One solution is increasing the available band resources, but a more effective way is improving the utilization of the existing bandwidth. In the Fourth Generation (4G) wireless communication and Wi-Fi standard, the orthogonal frequency division multiplexing (OFDM) technique is introduced. As an OFDM signal is a non-constant envelope signal and usually has a high peak-to-average power ratio (PAPR), it is very sensitive to the distortion caused by the nonlinear response of inline components in transmission systems.
Within a system, the nonlinearity is basically due to two areas. First, RF power amplifier (PA) is a main contributor. For efficient operation, the PA is set to the saturation zone where a high amount of nonlinearity is present. Consequently, the nonlinearity in the power amplifier will be unbearably high. This nonlinearity is called quiescent nonlinearity. Moreover, with the increasing of signal bandwidth and transmission rate, the memory effect may occur. So, in the 4G and 5G wireless system, both phenomenon should be considered.
Second, when radio over fiber (RoF) transmission is incorporated into front-haul transmission systems, it is capable of simplifying the remote radio units (RRU) from complex signal processing, thus potentially cost-effect. However, optical sideband modulation is used in RoF, and nonlinear distortion is induced to the signals. Therefore, RoF is another contributor to high quiescent nonlinearity.
The nonlinearity from RF PA, RoF system, and other devices will cause out-of-band spectral regrowth and in-band distortion. These problems lead to performance degradation. Some techniques to overcome the limitation of the above nonlinearity were proposed such as feed-forward, feed-back, pre-distortion, and post-distortion. Among these techniques, pre-distortion methods have been widely adopted for its flexible and robust mechanism in the modern ultra-wideband communication systems.
Pre-distortion is divided into two branches: Analog Pre-distortion (APD) and Digital Pre-distortion (DPD). The basic idea of APD is using dedicated nonlinearity devices, like a diode or field effect transistor, to generate a certain nonlinearity that can compensate a PA or RoF system. With the APD method, the odd order of intermodulation distortion (IMD) is usually compensated, but APD devices cannot fix the memory effect induced by PA. Different from the APD method, the DPD linearizes a communication system by digital means. With some digital processing methods, DPD is capable of compensating both memory effect and quiescent nonlinearity. However, the baseband DPD cannot suppress out-of-band distortion, while RF DPD requires costly RF detectors and A/D converters with a high sampling rate
In view of the fact that both APD and DPD have their advantages and disadvantages, a new pre-distortion method of combined APD and DPD is proposed. This new method can both suppress the quiescent nonlinearity and memory effect in a transmission system or PA. Here, APD is used to compensate the third order Intermodulation Distortion (IMD3) induced by an RoF system and the DPD is used to suppress the memory effect that mainly comes from the PA, in addition to in-band nonlinear distortion.
The validity of proposed pre-distortion method is proved in both simulation and experiment. In the simulation, an RoF transmission system model is built to induce both quiescent nonlinearity and memory effect. The RoF model is based on measured data. The PA model is an evolved Wiener-Hammerstein model that can induce both long-term and short-term memory effect. The APD circuit is designed and simulated in software Advanced Design System (ADS). In the simulation, the performance of the proposed pre-distortion method is evaluated by the improvement of Error Vector Magnitude (EVM) and IMD3. After determining the optimal memory depth and optimal RF input power in terms of EVM and IMD3, three cases are considered in the simulation. The first case is a general two-band LTE signal. Two LTE signals located at 800 MHz and 900 MHz pass through the simulation system. It is shown that an average 21.6-dB improvement of EVM, an average 4.6-dB improvement of IMD3, and a 23.8-dB improvement of Adjacent Channel Power Ratio (ACPR) are obtained. In the second case, a two-band LTE signal is located at 800 MHz and 840 MHz. It is shown that a 11.3 dB improvement of EVM, a 9.8 dB improvement of IMD3, and a 9.3 dB improvement of ACPR are achieved. For the third case that is a three-band test, three signals are set to 800 MHz, 850 MHz and 900 MHz. It is shown that a 16.7 dB improvement of EVM, a 21.8 dB improvement of IMD3, and a 18.4 dB improvement of ACPR are got. In the first two cases, 2D-DPD is used as DPD method, and for the third case the RF DPD method is used. The APD circuit used in all three cases is an anti-parallel Schottky diodes based circuit that was designed and fabricated by a former student. The results of the proposed pre-distortion method is compared to traditional DPD-only and APD-only. It is revealed that the proposed pre-distortion method results in higher improvement in EVM, ACPR, and IMD3. This means that the advantages of APD and DPD were combined successfully.
The experiments further verify the advantages of proposed linearization method. Like the simulation, three cases are also considered in experiments. For case one with two-band LTE signal located at 800 MHz and 900 MHz, it is shown that a 11.0 dB improvement of EVM, a 15.0 dB improvement of IMD3, and 19.4 dB improvement of ACPR are obtained. For case two with two-band LTE signal located at 800 MHz and 840 MHz, it is shown that an 8.2 dB improvement of EVM, a 16.8 dB improvement of IMD3, and a 4.6 dB improvement of ACPR are achieved. For case three with three LTE signals at 800 MHz, 850 MHz, and 900 MHz, it is shown that a 10.1 dB improvement in EVM, a 16.9 dB improvement in IMD3, and an 8.6 dB improvement in ACPR are achieved. Moreover, the proposed pre-distortion method leads to better performance in EVM, IMD3, and ACPR than APD-only or DPD-only method.
At the end of the thesis, another evolved 2D-DPD method is proposed based on the mathematical model of combined analog and digital pre-distortion method. The high order nonlinear memory effect is ignored in evolved 2D-DPD. Therefore, the number of coefficients is reduced while the performance remains almost the same. This relationship between the coefficients number and nonlinearity order is discussed. Finally, a simple simulation is done which proves the benefits of evolved 2D-DPD.