Owing to the varying working speed， the vibration signal of gearboxes inherently exhibits non-stationary characteristics.The resultant fuzzy spectrum is not conducive to feature extraction and fault diagnosis. The order tracking method， as an effective analyzing method for non-stationary signals， is realized by the signal transformed from time domain to angle domain. It’s helpful to improve spectrum discrimination. The widely used computation ordering tracking methods sample signal in equal-angle increment according to the extracting rotating speed or the phase demodulation method are limited by the integral cumulative error or the small tracking speed. According to the equal peak-to-peak angle interval of gear vibration signal， this paper proposes a novel tacholess order tracking method based on searching extremums of time signal. The proposed method doesn’t require the integral process of the rotating speed， and simultaneously allows a wide range of tracking speed. The error caused by the signal transforming process is decreased. The result shows a high concentration in spectrum energy and great discrimination of spectrum characteristics. It is verified by the theoretical analysis， simulation， and experiment. The proposed method can be applied to the gearbox signal analysis and fault diagnosis under varying speeds.