Several improved active disturbance rejection control (ADRC) methods, including input modified ADRC (MADRC), input delay ADRC, Smith predictor-based ADRC, and Smith-like based ADRC (SLADRC), have been proposed to tackle $\boldsymbol{K/(Ts+1)^{n}}$-type high-order processes. This article aims to investigate their control performance theoretically, such as the input disturbance rejection and measurement noise rejection capabilities. It is proved that these improved ADRCs with the same parameters have the same measurement noise rejection capability. Besides, quantitative tuning rules of MADRC and SLADRC are derived based on their desired dynamic equations. Compared with proportional-integral controller and regular ADRC, their effectiveness is verified by comparative simulations and experiments, where MADRC obtains the best control performance with good robustness. Finally, the successful field applications of MADRC on the main steam pressure process further illustrate its application potential in complicated industrial processes.