Power Tuesdays: Automated Switching Loss Measurement Bench

Automated Switching Loss Measurement Bench – Open Science Collaboration

Over the past months, we have developed an open-source automated test bench dedicated to switching loss characterization using the opposition method.
This effort emerged from a two-day hackathon, followed by monthly “Open Power Tuesday” sessions, gathering contributors from LAPLACE, LAAS, OWNTECH Foundation, NXP, and others.

:right_arrow: GitHub Repository: link_here

Figure 1 - All the work we’ve been putting into this project since April 2024


:test_tube: Objective

Traditional double-pulse tests are widely used for switching loss measurement, but they lack flexibility for fully automated multi-parameter sweeps.

The opposition method offers:

  • Higher precision through steady-state energy balance
  • Indirect but more representative measurements
  • Challenges: thermal rise, synchronization, multi-device coordination

Our goal was to fully automate this method using only open tools, from signal generation to instrumentation control and post-processing.


:hammer_and_wrench: System Architecture

The bench is composed of three main layers:

  1. Python Supervisor & Instrument Control (via SCPI / PyVISA)
  2. Microcontroller-Based Command Generation (OwnTech SPIN / 5.4 GHz clock sync)
  3. Power Hardware – Opposed Half-Bridge Cells with Isolated Drivers

Figure 2 - System software architecture (still in French)

Figure 3 - Test bench architecture (still in French though)

:white_check_mark: Current Capabilities

:check_mark: Automated sweep of:

  • DC bus voltage
  • Phase shift between cells
  • Duty cycle offset
  • Switching frequency

:check_mark: Synchronized triggering with hardware sync pulse
:check_mark: Fast segmented memory acquisition on oscilloscope
:check_mark: JSON-based experiment configuration
:check_mark: GUI available for validation and manual runs


Figure 4 - Our GUI

:wrench: Hardware Implementation

We used Infineon EVAL-1ED3122Mx12H half-bridge evaluation boards, initially delivered without power MOSFETs or isolated supplies.

During the hackathon:

  • Components were selected and soldered manually
  • A 48 V Si MOSFET configuration was validated first
  • Later sessions upgraded to higher voltage and thermally improved builds

Figure 5 - Overview of the global test bench

:chart_increasing: Experimental Results

The bench is already delivering fully automated switching energy measurements.

  • Opposed cell control allows separation of turn-on and turn-off losses
  • Energy during switch-off vs Current and Energy vs Voltage curves are now automatically extracted

Figure 6 - Switch off energy for : 5 different peak current levels and 3 different bus voltage levels

Additional captures illustrate phase-shift vs duty-ratio modes:

Figure 7 - Main operating signals over one switching period (phase-shift mode). Parameters: 60 V bus voltage, 80 kHz switching frequency, 50 µH inductance, 150° phase shift, 300 ns dead time.


Figure 8 - Main operating signals over one switching period (duty-cycle offset mode). Parameters: bus voltage reduced to 30 V, switching frequency of 50 kHz, 50 µH inductance, 300 ns dead time, 0.8 % duty-cycle mismatch.

:loudspeaker: Next Steps & Call for Collaboration

Here are the next milestones we invite contributors to join:

  • Full automatic calibration stage (RdsON, inductance ESR, etc.)
  • Thermal modeling & compensation in measurement loop
  • Support for SiC and GaN device characterization
  • Expanded dataset publication in Open Access

:raising_hands: Acknowledgments

Huge thanks to:

  • Nicolas Rouger
  • @luiz_villa
  • @jalinei
  • all contributors from hackathon + Open Power Tuesdays,
  • LAPLACE laboratory
  • LAAS laboratory
  • OWNTECH Foundation

If you want to join the effort, test the bench, or contribute code/hardware:

:backhand_index_pointing_right: Reply in this thread or open an issue on GitHub!