The difference between the actual value and the estimated value of the system variable was large in the initial stages of observation using linear extended state observer (LESO). Since the gain of LESO was large, there was obvious peaking phenomenon, which affected the measurement of the disturbance. The larger the gain of the observer, the more severe of the peaking phenomenon. In order to solve the problems, a timevarying gain extended state observer (TESO) was designed. Its gain was a time varying function, which was a smaller value at the initial time, and then the function value increased with time to a larger constant. The parameters tuning formula was given based on Lyapunov transform and differential algebraic spectral theory. Replacing the extended state observer in the linear active disturbance rejection controller (LADRC) with TESO, the controller was applied to the speed control of permanent magnet synchronous motor regulation system. The effectiveness of the controller design was verified by the simulation and experimental results.