Practical EMI Guidance for Power Converters

SiC and GaN devices allow converters to switch faster while improving efficiency and power density. But nanosecond switching makes parasitic inductance and capacitance much more important because high dv/dt and di/dt can turn the package, PCB, DC-link, and busbar into high-frequency noise paths.

Switching frequency is not the only cause of electromagnetic interference (EMI). For instance, a converter operating at tens or hundreds of kilohertz can still generate significant energy into tens or hundreds of megahertz because of fast edges, overshoot, and parasitic ringing. These effects impact the emissions spectrum that engineers observe during conducted or radiated testing.

EMI testing needs to do more than show if a limit has been exceeded. Engineers need to connect the measured disturbance to the switching event that created it, trace how it propagated through the converter, and confirm that the probe or measurement setup itself has not distorted the waveform.

Connect switching behavior to the EMI spectrum

The first step in debugging EMI is to connect what appears in the spectrum with what is happening during the switching transition.

Switching frequency and switching transition speed describe different aspects of the behavior. The switching frequency determines how often the waveform repeats and where the fundamental and lower-order harmonics occur. The rise and fall times determine how much high-frequency content each individual transition contains (Figure 1).

These transitions happen in just a few nanoseconds, so even a converter switching at tens or hundreds of kilohertz can generate noise well into the megahertz range. The converter’s parasitic inductance and capacitance then shape that high-frequency noise.

Rapid current changes can result in voltage across parasitic inductance according to v = L(di/dt). This can happen in commutation loops, PCB traces, DC-link connections, and busbars. The result may be voltage overshoot or ringing.

Fast voltage changes cause a similar effect through stray capacitance. High dv/dt can drive current from a switching node into nearby structures such as the heatsink, chassis, or PCB planes. Even a few picofarads can carry noticeable current when the voltage changes very quickly.

These effects can often be seen in both the waveform and the EMI spectrum. For example, ringing after turn-on or turn-off may appear as a peak or rise in spectral energy at roughly the same frequency.

Engineers can characterize switch-node voltage and gate voltage by analyzing rise and fall times, overshoot, ringing amplitude, and ringing frequency and then comparing these observations with the conducted or radiated EMI spectrum.

A matching frequency is also useful information. If a turn-off waveform rings near 75 MHz and the emissions scan also shows a strong feature near 75 MHz, the next step is to disturb the suspected mechanism. Changing gate resistance, adding or modifying a snubber, or altering a parasitic path should consistently change the time-domain and spectral signatures if the diagnosis is correct.

Trace the noise path

Once a spectral feature has been linked to a switching event, the next step is to determine how the noise propagates through the converter.

A useful way to do this is to think in terms of source, coupling path, and receiver. The source might be a fast-switching node, a high-current commutation loop, or ringing created during turn-on or turn-off. The coupling path is the route the noise takes through the system. The receiver is where that noise is measured or causes a problem.

The path may not be obvious from the circuit schematic because a switching node can couple current through stray capacitance to a heatsink, chassis, or PCB plane. In a motor drive, common-mode voltage can also drive current through the motor, cable shield, and chassis. The semiconductor may create the noise, but a cable or metal structure can become the main path.

It also helps to decide if the noise is differential-mode or common-mode. Differential-mode noise flows through the normal power conductors, while common-mode noise flows through paths such as chassis, earth, or stray capacitance. The difference can help engineers understand the part of the converter that requires attention.

Measurement tools can be used to confirm the suspected path; for example, near-field probes can locate noisy switching nodes or current loops, current probes can show high-frequency current on cables, and CM/DM separation can show which noise mode is dominant.

The main question is where the high-frequency current flows and how it returns. Once engineers understand the path, they can choose a more targeted fix instead of relying on trial and error.

When the measurement lies

Fast SiC and GaN switching is difficult to measure accurately. A high-side gate signal may be only a few volts, while the source node can move by hundreds of volts in a few nanoseconds. A differential probe may reject this common-mode voltage well at low frequency but perform less effectively as frequency rises (Figure 2). The oscilloscope can then show a gate spike or ringing that is partly caused by the probe.

The probe connection can also affect the result. Long ground leads add inductance and increase the size of the measurement loop, which makes it easier to pick up nearby magnetic fields. Probe capacitance can also change the switching node. These effects may increase existing ringing or create ringing that was not there before. Current measurements can cause similar problems if a shunt or probe changes the commutation loop.

Bandwidth also has significance on what appears on the screen because too little bandwidth can hide overshoot and ringing, while too much can add noise that makes the waveform harder to read. If voltage and current are measured together, the probes also need to be time-aligned. Even a small delay can make one event appear to happen before or after the other.

A simple mechanism to check the measurement is to change the setup without changing the converter. Engineers can shorten the probe connection, reduce the loop area, try an isolated probe, or change the bandwidth setting. If the spike or ringing changes significantly, the measurement setup is affecting the result. Engineers should confirm this before redesigning the converter around a doubtful waveform.

Tracing the EMI problem

The faster switching of SiC and GaN makes parasitic effects and measurement errors more important. Engineers therefore need to trace an EMI problem back to the switching event, see how the noise travels through the converter, and make sure the measurement itself is accurate before deciding on a fix.

Cover image: Rohde & Schwarz

The post Practical EMI Guidance for Power Converters appeared first on Power Electronics News.