HYBRID LINEARIZATION TECHNIQUES FOR BROADBAND 5G AND BEYOND TRANSMITTERS

Abstract

The new era communication systems are desired to support high data rate signals with low latency. The high data rate signals are generated using complex modulation schemes. This also requires RF devices that can support signal transmission. This thesis focuses on removing the hardware impairments mainly due to the power amplifiers in 5G and beyond transmitters. This helps in the linear transmission of RF signals. There are several linearization techniques of which digital predistortion (DPD) linearization technique is widely used in present-day baseband stations as it offers high degrees of linearization. DPD requires feedback signal from the transmission end along with the baseband signal information. The bandwidth of the feedback signal in the conventional DPD is approximately five-to-seven times the input signal. The sampling rate of these feedback signals is further increased by Nyquist criteria. This requires analogto- digital converters (ADCs) and digital-to-analog converters (ADCs) with high data rates as the high bandwidth feedback signal is converted back to baseband and then further converted to RF for transmission. As the bandwidth of the signal increases, the present-day DACs may not be used as they are incapable of supporting the high sampling rate in the feedback loop. This limits the signal transmission process till data converters with high clock rates are available. This motivates us to propose linearization techniques that help in the linear transmission of signals. Analog predistortion (APD) is another linearization technique that supports high bandwidth signal transmission but is mainly used in satellite communications as the degrees of nonlinearity reduction are not at par with DPD. This thesis aims to propose and design linearization techniques that exploit the high bandwidth support feature of the APD and add with the DPD techniques to overcome its limitation and ensure linear transmission of signals.A power amplifier (PA) is one of the important sources of nonlinearity in a communication system. In a PA, there is always a trade-off between linearity and efficiency. The complex modulation schemes that generate high data rate signals produce signals with a high peak-power-to-average-power ratio (PAPR). This forces PA to operate in a high back-off region to ensure linear output. This reduces the efficiency of PA. Linearization techniques can be used to improve efficiency and ensure linear transmission. APD is one technique that may be cascaded to the PA design during the fabrication stage to improve its linearity. The small APD devices proposed in the literature are capable of reducing small degrees of nonlinearity. This finds application in satellite communication which prefers a small footprint. The APD structures which are capable of providing a moderate level of nonlinearity reduction have a bigger footprint as it uses phase shifters, vector modulators, and other analog components. The thesis first investigates a small APD structure that removes the bulky components and still offers significant linearity. The design is fabricated and tested in Ku-band at 15.5 GHz and it finds its application in satellite communications. This design is further improved by adding variable parameters to the APD design. These designs used unequal power dividers and variable power dividers along with different nonlinearity generates. The design is tested with modulated signals and shows significant improvement in nonlinearity reduction. The nonlinearity was measured in terms of an adjacent channel power ratio (ACPR) and normalized mean square error (NMSE). This was used as the basis for another linearizer design which uses a self-adaptive APD linearization method. The self-adaptive APD uses a two-path APD circuit which outputs a predistorted signal to overcome the nonlinearity of the PA. The adaptive mechanism requires small feedback from a few samples from the adjacent channel area. It uses two variable parameters in the baseband, which is very less compared to that of the DPD method. This is tested with long-term evolution-Advanced (LTE-A) signals and achieved a significant improvement in ACPR and NMSE which is acceptable as per the standards for 5G communication. As this used analog circuits for predistortion, the in-band error in the output signal can be further reduced. With the aim of improving NMSE in the in-band signal, the self-adaptive APD process was followed by a bandlimited DPD method. The bandlimited DPD uses a feedback signal of bandwidth equal to that of the input signal. A memory polynomial method was used for predistortion training. As the feedback signal used small bandwidth, the number of coefficients for the DPD shows a substantial reduction. The experimental results show that the NMSE of the signal improves further. This makes the method acceptable for 5G communication systems. The proposed system is also implemented for 2 × 2 MIMO and achieved substantial improvement in linearization with two antenna systems. The RF devices, like analog predistortion circuits used for the implementation of the proposed techniques, are designed and tested at 3.5 GHz. These proposed methods may be extended to other frequency ranges by redesigning these RF circuits.