Power Measurement Technology When Static Measurements Fail in SiC, GaN, and AI Infrastructures

Related Vendors

Taiwan Machine Tool & Accessory Builders' Association (TMBA)

SIMTOS (Seoul International Machine Tool Show)

SiC, GaN, and high loads push classic measurement technology to its limits. However, with the right measurement technology strategy, the efficiency and dynamic switching behavior of WBG semiconductors can be validated without errors.

From EV fast charging and renewable energy systems to AI data centers and grid-connected infrastructures, developers must deliver higher efficiencies, increased power densities, and reliable performance under highly dynamic conditions. This shift is impacting measurement technology. It is no longer sufficient to determine average power or validate systems exclusively under static conditions.

Developers require time-synchronized insights into efficiency, switching behavior, harmonics, transient response, thermal effects, and power quality. If these effects are not precisely captured by the measurement instruments, loss and efficiency calculations can deviate significantly from real operating conditions.

WBG Semiconductors and the Bandwidth Dilemma

With the use of silicon carbide (SiC) and gallium nitride (GaN), the demands on measurement technology are increasing further. Wide-bandgap semiconductors enable higher switching speeds with lower losses compared to conventional silicon components. At the same time, they produce steeper edges, high-frequency noise, and more complex signal shapes. Additionally, system voltages are rising: in e-mobility, the output voltages of fast chargers are moving from 400 to 600 V toward 1 kV, while UPS systems for data centers must handle multi-megawatt loads.

Switching operations occur in the microsecond range, while thermal effects develop over minutes or hours. To prevent errors, conventional bandwidth criteria fall short. Unlike pure signal analyses, the bandwidth for efficiency and power determination is guided by the least distorted signal shape. High-speed switching voltages create triangular distortions in the current waveforms. Therefore, the current path dictates the required bandwidth:

• Five to seven times the switching frequency is often necessary to keep the measurement error below 0.05%.
• If signal waveforms are to be fully reconstructed, this value increases to up to 25 times.

In SiC-based systems with a switching frequency of 20 kHz, a current measurement bandwidth of around 500 kHz is required for the necessary precision. Additionally, there are electromagnetic interferences (EMI): the high-frequency edges emit disturbances that can affect measurement instruments. Instruments must therefore be validated under real switching conditions, which does not always take place under ideal laboratory conditions.

UPS Systems for AI Data Centers

The energy supply of a data center is particularly in focus for measurement technology requirements. AI workloads generate rapid, unpredictable load surges. Engineers at Vertiv's Bologna Power Customer Experience Centre are developing power converters and UPS systems for these high-performance infrastructures.

To absorb millisecond load peaks, the developers rely on input power smoothing with modules that handle up to 2.5 MW in a single module. UPS battery energy is used during power peaks to protect the upstream grid and prevent feedback effects on generators.

Validating this system behavior required a highly complex measurement setup. The UPS systems included AC grid inputs, DC battery inputs, and three-phase outputs. Developers had to safely handle currents up to 5,000 A while simultaneously capturing power flow, signal waveforms, noise, harmonic distortions, and thermal effects during fast switching operations.

By combining precision power analyzers, isolated mixed-signal oscilloscopes, data acquisition systems, and an integrated software platform (in this specific case Yokogawa WT5000, DL950, GM10, and IS8000), engineers were able to evaluate electrical, thermal, and transient data in a time-synchronized manner. This provided evidence that the smoothing algorithms protect generators from dangerous resonances.

Measurement Technology as a Competitive Factor

Whether it's 1 kV fast-charging infrastructure, grid-connected renewable energy, or multi-megawatt AI data centers: fast electrical transients, long-term thermal effects, and strict efficiency standards require integrated measurement ecosystems. High-performance measurement technology is no longer merely a tool at the end of the development process but a central component of system design that determines time-to-market and market success. 

The Problems of GaN and SiC

When Fast Switches Generate New Losses