A New Noise-Suppression Algorithm for Transient Thermal Analysis in Semiconductors over Pulse Superposition.

This article introduces a new measurement method for noise suppression in transient thermal analysis (TTA), which is used to measure the transient thermal impedance of power semiconductors and LEDs. An additional short pulse sequence is added to the standard TTA measurement procedure. The thermal response of this additional sequence is converted into the original single-pulse response of standard TTA by superposition of previous pulses in the sequence. Applying this method, the noise before 1 ms can be significantly reduced without averaging over multiple measurement repetitions that would increase the measurement time accordingly. The noise reduction for times below 1 ms is important to resolve the thermal resistance layers close to the junction of a semiconductor. The measurement method and data analysis algorithm are developed for TTA measurements, but the principle is suitable for all linear time-invariant (LTI) systems where the system response is measured over several time decades. The algorithm was evaluated on simulated and experimental data of a power LED. With an additional sequence of only 100 ms, a signal-to-noise ratio (SNR) improvement of 15.1 dB for simulated data and 14.7 dB for experimental data was achieved. This is equivalent to averaging over 30 repetitions in the standard TTA procedure. Therefore, the new method achieves a reduction of measurement time of 97% for applications where low noise levels are required for times below 1 ms.