Characterize EMI from dc-dc converter ringing

Switching power supplies produce radiated and conducted emissions caused by ringing. Oscilloscope and spectrum-analyzer measurements let you see them.

DC-DC converters are ubiquitous in most electronic products. While more efficient than linear regulators, they can also produce considerable amounts of interference that can affect nearby circuits. The measurements in this article show how ringing occurs as a result of switching.

Conducted EMI emissions come from power input through the fast transition of the switching device and ringing of the switched waveform. Harmonic emissions that come from the switched waveform have been covered adequately elsewhere, but It’s this ringing that I’d like to demonstrate in this article.

Switching converter topology

Figure 1 shows a typical buck converter topology. The junction of the switch, diode, and inductor is typically a trapezoidal waveform with sharp rising and falling edges. The fast rise and fall times create harmonic emissions. The capacitive and inductive parasitic elements in the switch and traces can, however, result in ringing on the waveform. This ringing can cause emissions into hundreds of megahertz.

Buck regulator topology — Figure 1. A typical buck converter circuit uses a MOSFET to switch the input to the inductor and capacitor. The switching action creates ringing and thus emissions.

Figure 2 mathematically shows how ringing in the time domain can translate the harmonic peaking in the frequency domain.

Signal ringing — Figure 2. Ringing on the switched waveform can cause an increased emission at the ring frequency. Source: Clayton Paul.

Oscilloscope measurements show ringing

Let’s look at the ring frequency from a 1 MHz dc-dc converter that uses a GaN switch device. Figure 3 shows the test setup. We’ll measure the ringing using the 300 MHz Micsig oscilloscope and show the resulting emissions with the built-in FFT feature. We’ll later compare the Micsig FFT with a more accurate Siglent SSA 3032X spectrum analyzer.

The resulting ringing frequency appears in Figure 4. By adding vertical cursors and aligning with adjacent peaks, we can read off the ringing frequency as “1/Δt”.

Ringing in signals — Figure 4. Expanding the leading edge of the switched waveform reveals a strong ring frequency at 231 MHz.

To observe the frequency domain plot, you should show several switching cycles (Figure 5). Open the FFT on your oscilloscope and adjust the position for best display. I set the display to 100 MHz/div to observe the emission peaks. While the Micsig is just an eight-bit oscilloscope, it displayed a fairly accurate FFT of the ring frequencies.

Time domain and frequency domain in oscilloscope — Figure 5. By adjusting the time base to show more switching cycles, you can best observe the resulting frequency domain plot.

Spectrum analyzer measurements

Now, let’s switch over to the Siglent SSA 3032X spectrum analyzer and compare emission results. We’ll use a Tekbox TBCPP2-750 RF current probe clamped around the power input cable and then around the output load resistor (Figure 6). Any long cables attached to either of these points would likely serve as efficient antennas and potentially cause radiated emission failures. Using an RF current probe, rather than an oscilloscope probe at the switched device, should more closely correspond to potential radiated emissions.

Current probe on a switching power supply — Figure 6. I used a Tekbox TBCP2-750 RF current probe to measure the ring emissions coupled to the power input cable.

The peaks in Figure 7 roughly correspond to the emissions I measured with the oscilloscope. The 500 MHz peak is the second harmonic of the fundamental ringing at 256 MHz. Because we’re measuring the ring frequencies using a whole different method, the peaks may not match those taken with the oscilloscope probe at the switch. These results should, however, provide a good idea of the emissions performance you can expect.

Figure 7. Emissions peaks using an RF current probe on the power input cable (violet) and power output load (blue).

Once we’ve established a baseline for the ringing, we’d start experimenting with improved board layouts or snubbing networks to reduce the effect of the parasitic resonances.

Summary

I was surprised the Micsig could measure these ring-induced parasitic resonances as well as it did. This just demonstrates a moderate-priced oscilloscope with an FFT feature can help with EMC measurements and follow-up mitigation experiments. Of course, for more serious EMC studies, I’d combine the oscilloscope with the spectrum analyzer for accurate ring frequency monitoring.

Because most products have multiple on-board dc-dc converters, this set of experiments also shows the importance of checking for any large ringing on power converter switched waveforms, which can directly translate to radiated emissions, sometimes well into the hundreds of MHz.

References