<p>The trade-off between estimation accuracy and noise sensitivity poses a practical limitation for active disturbance rejection control (ADRC). To deal with this problem, a new estimator that combines a finite impulse response filter-based observer (FIRFO) and a reduced-order extended state observer (RESO) is presented in this work. The proposed approach is formulated in discrete time and consists of three stages. First, the FIRFO estimates part of the total disturbance while filtering sensor noise. These tasks are accomplished via a sampling-based method, which facilitates observer tuning. Second, the RESO estimates the residual disturbance and the remaining system states from the control input and noise-filtered output. Here, the performance and robustness of the structure are improved by using a RESO instead of a full-order extended state observer (ESO). Finally, the total disturbance and the state vector are reconstructed to implement an ADRC scheme. The characteristics of the new estimator are supported by a frequency-domain analysis and validated through experiments performed on a power electronic converter. Comparisons with single and cascaded ESO show the superiority of the proposed strategy in terms of disturbance compensation and noise reduction.</p>