Junction temperature is an important factor affecting the reliabilities of semiconductor devices. Usually, the method of measuring the junction temperature is not tested on-line. However, due to the fact that neither contact thermal resistance nor thermal resistance varying with temperature is taken into account, there exists an error in the off-line measurement. A way to solve the problem of off-line measurement is to measure the junction temperature on-line. In this paper, we propose an electrical method of measuring the temperature rise of high-power bipolar transistor in the working condition. The measurement method is based on a good linear relationship between base-emitter voltage (Vbe) and temperature during the steady-state. Taking the model 2N3055 of bipolar high power transistor for example, in this paper we study the relationship between base-emitter voltage (Vbe) and temperature under the conditions of constant collector-emitter voltage (Vce) and collector-current (Ice). During the experiment, the device is placed in a thermostat. A voltage is applied to the device collector, a current is applied to the base, and the emitter is earthed. Before the device is measured, we set different temperatures and make sure that the equipment is in a steady state. In order to avoid the effect of self-heating, the pulse current is used in the experiment. The pulse width and the period are 500 μups and 1 ms, respectively.The research result shows that the base-emitter voltage (Vbe) decreases linearly with temperature increasing and the base-emitter current is changed below 4% when the temperature is in a range of 40 ℃-140 ℃. In this paper we also deduce the mathematical expressions for base-emitter voltage (Vbe) and temperature under a steady state. It is proved that the Vbe-temperatrue curve is linear and temperature error is less than 0.5 ℃ when the changes of base current value does not exceed 4%. Therefore, in this paper we deduce a new method of testing the junction temperature in the speeding up measurement experiment. By measuring any of the reference points on the calibration curve under certain experimental conditions, the junction temperature can be calculated quickly according to the proposed formula.Finally, the phase 11 is used to verify the proposed method. We measure the thermal resistance upper the case for the junction of model 2N3055 and the thermal resistance under the case for the junction of model 2SD1047. The measurement results of phase11are compared with the junction temperature calculated using the test formula. The results show that the error of junction temperature between the two methods is less than 0.7%, which is corresponding with the needs of practical application. It proves the correctness and feasibility of the method.