Abstract

The increasing demand for broadband wireless transmission in the modern internet has led to the proposal and standardization of the fifth-generation (5G) mobile communication system, which offers massive device connectivity, high bit rates, low latency, and cost sustainability. However, maintaining a high transmission rate as well as low latency is difficult to achieve simultaneously, which requires some state-of-art fronthaul transmission techniques. Therefore, radio over fiber (RoF) with different approaches like digital RoF (D-RoF), analog RoF (A-RoF), and delta-sigma modulation based RoF (DSM-RoF) for 5G fronthaul transmission has been introduced. Those RoF techniques may significantly reduce complexity and power consumption at base stations, but the extra electric to optic (E/O), optic to electric (O/E) converters and power amplifiers could introduce extra nonlinearity into the system. Moreover, ultra-broadband or multi-band ultra-broadband signal is introduced in 5G to further increase the transmission rate, which further increases the impact of the nonlinearity. Therefore, broadband linearization techniques are necessary for RoF fronthaul transmission systems due to the fragile of the signal and the inherent nonlinear distortions introduced by RoF link. To reduce the degradation of nonlinearity for RoF link, digital predistortion (DPD) techniques have been extensively researched to address these challenges.

In a multi-band or multi-dimensional RoF system, multi-band DPD is required. Multi-dimensional DPD should be able to suppress the internal distortion within each band/dimension but also inter-distortion between different bands/dimensions. Unfortunately, the dimension higher than 3 causes a high calculation complexity to get the DPD function coefficients. There have been lots of efforts that have been made to obtain less-complexity DPD with better accuracy for multi-band or multidimensional signals. However, very limited DPD techniques have been proposed in simplifying the fundamental linearization function for bands exceeding four. Thus, the multi-band/multidimensional DPD has not been really got in used in commercial products because of the high complexity, high cost and high-power consumption. Thus, a simplified linearization approach for multi-band DPD is still needed.

In this thesis, a new low-complexity multidimensional DPD is introduced. This proposed DPD introduces a simplified DPD function, which evolves from the conventional memory polynomial function. Compared with the conventional multi-dimensional DPD, this proposed approach has lower complexity increased with the increase of signal bands or dimensions, nonlinearity orders, and memory effect depth. For example, the conventional DPD function needs a total of 40040 coefficients for the 6-band signals with a nonlinearity order of 10 and a memory depth of 5. However, this proposed low-complexity DPD function needs 640 coefficients. A substantial reduction in complexity is clearly observed.

The performance of the proposed DPD is evaluated by both simulation and experiments. An up to 6-band 64-QAM orthogonal frequency division multiplexing (OFDM) signal with each band of 200 MHz in simulations and an up to 5-band 20 MHz 64-QAM OFDM signal in experiments are used. The performance is evaluated in the means of error vector magnitude (EVM) of the received signal. The average improvement of EVM in simulation for 3-band, 4-band, 5-band and 6-band signals is 19.97 dB, 18.65 dB, 16.64 dB and 15.44 dB, respectively. The average improvement of EVM in experiments for 4-band and 5-band signals is 5.67 dB and 8.1 dB, respectively. The above results prove that the proposed DPD can significantly reduce the complexity and provide good linearization.