Sam Jaffe, Applications Engineer, >30V Buck Converters and Controllers, Texas Instruments
Electromagnetic interference (EMI) presents constant challenges in automotive power end equipment. With the rise of mild hybrid electric vehicle (MHEV) solutions, EMI becomes even more challenging, as the battery voltage shifts from 12 V to 48 V in many electronic circuits in the system.
Most engineers designing automotive circuits understand how to reduce EMI through filter design, layout guidelines and management features such as spread spectrum, flip-chip packaging and more. However, there are some less-known tips that can significantly improve EMI in a buck converter (and other topologies) without redesigning the board. These tips may mean the difference between passing EMI tests in 10 minutes and needing to spin a new board.
Tip No. 1: Rotate the power inductor
A buck converter uses an inductor-capacitor filter to convert a switching waveform (switch node: VSW) into a DC waveform (output voltage: VOUT). Figure 1 is a simplified schematic of a buck circuit.
As Figure 1 shows, one of the inductor terminals connects to VSW, and as such generates EMI noise. The voltage on this node is a rectangular wave with fast edges, swinging from 0 V to the battery voltage. This could be as high as 48 V in some MHEV designs. The other terminal connects to VOUT, which in terms of EMI is electrically quiet, being nearly DC. Good printed circuit board (PCB) layout practice requires the minimization of switch-node surface area to reduce capacitive coupling to the ground plane that would otherwise cause large common-mode noise and result in poor EMI performance. It is possible to apply the same capacitive-coupling reduction idea to the inductor. EMI performance will vary, depending on the inductor’s construction and orientation.
Power inductors are essentially just wires wrapped around a core material, as shown in Figure 2. They can be wrapped from top to bottom in a single layer or wrapped winding-over-winding with multiple layers – but the important piece for this analysis is that an inductor is never perfectly symmetrical from one terminal to the other. Simply rotating the inductor 180 degrees changes which inductor terminal connects to the high-noise switch node. This creates different EMI results.
Click image to enlarge
Figure 2: Simplified inductor construction – single layer (left) and multiple layers (right)
It’s possible to reduce the capacitive coupling for a single-layer inductor by connecting the noisy switch node to the terminal that starts winding at the bottom (Figure 2, L1, terminal B). The bottom sits physically closer to the board, which will have better shielding by the GND planes on the board compared to the terminal connected to the top of the winding. To reduce the capacitive coupling for a multilayer inductor, connect the noisy switch node to the terminal that starts winding from the inside out (Figure 2, L2, terminal D). That way, the noisy parts of the coil are effectively shielded by the quiet VOUT parts of the coil on the outside of the winding.
The dotted terminal on the schematic (Figure 1, L) generally signifies the beginning of the winding at the inside of the coil. This is the terminal that should connect to the switch node for reduced EMI. That said, the dot convention is not always true for every vendor or every inductor; however, you can simply rotate the inductor 180 degrees to see if the results improve.
Results: Rotating the inductor on a 13.5-VIN, 5-VOUT, 3-AOUT, 400-kHz two-layer board using the Texas Instruments LMR33630-Q1 resulted in an 8-dBµV improvement in the FM band. Results for average detection improved from 15 dBµV at 108 MHz (3 dBµV below the Comité International Spécial des Perturbations Radioélectriques [CISPR] 25 Class 5 limit) to 7 dBµV (11 dBµV below the CISPR 25 Class 5 limit).
Tip No. 2: Remove the capacitor at the supply terminals of the EMI filter
EMI filters often consist of an inductor, ferrite bead and sometimes a common-mode choke, as shown in Figure 3. These three components will have capacitors placed before, after and/or between these filter components. A filter often starts with a small ceramic high-frequency capacitor connected to the supply terminals farthest from the buck (Figure 3, CHF1). The idea is simple: reduce the ripple on the supply terminals by adding a capacitor. This typically results in reduced differential-mode EMI, but there are cases that do not result in improved EMI performance.
As shown in Figure 3, CHF1 (the left-most capacitor) provides a low resistance path from the battery +12 V and GND (IN+ and IN-), with a parasitic inductance from the physical attributes of the harness connecting the battery to the capacitor (battery to J1). Low resistance paths with inductance and capacitance resonate at an angular frequency inversely proportional to the square root of the product of inductance and capacitance. A 0.1-µF capacitor only needs 0.022 nH of inductance to resonate at 108 MHz (the high end of the FM band, a notoriously difficult frequency to pass when testing EMI).
Click image to enlarge
Figure 3: Example schematic showing an EMI filter and LMR33630-Q1 buck converter
Depending on the capacitor selected and the layout, it is possible to see an improvement by simply removing the capacitor.
Results: Removing CHF1 in Figure 3 resulted in a 3- to 5-dBµV improvement in the average detection within the FM band. In some cases the capacitor helps, but removing the capacitor will often improve high-frequency results.
Tip No. 3: Change the load resistor placement
An EMI test consists of operating the power circuit at a typical input voltage and maximum output current. The test requires a load in order to run the converter at this output power, which is usually a resistive “dummy” load (see RLOAD in Figure 1). You must consider what type of load to use (such as a wire-wound resistor or a noninductive resistor), what heat sink to use (a larger heat sink will act as an antenna for capacitive coupling, but an undersized heat sink may cause the resistor to overheat and break before the test is done) and shielding (a grounded shield will reduce capacitive coupling but will increase the required load/heat-sink size due to the trapping of hot air).
Another important aspect of the load-resistor considerations is how the load resistor connects to the output. This tip is similar to tip No. 2. Outputs with only ceramic capacitors may resonate with the parasitic inductance of the connection from the output capacitors to the load. Little or no damping may cause this resonance to result in a failing EMI test. The easiest way to ensure this resonance is not a large source of EMI is to solder the load directly to the ceramic output capacitors, which will minimize the parasitic inductance and reduce the resonance, or shift the issue area to a higher frequency. Either way, you will know if the resonance is causing EMI.
Results: Moving the load from the VOUT terminals directly to the output capacitors in one board resulted in a 10-dBµV improvement in the average detection within the FM band. The results improved from 22 dBµV (4 dBµV over the limit) to 12 dBµV (6 dBµV below the limit).
EMI mitigation is both a science and an art. There are many articles, application notes and trainings about best practices for EMI. While it is important to give a design the best chance of passing right from the start, there’s no guarantee that the board will pass on the first run. That’s why it’s important to have a strategy to improve EMI without making drastic changes to the design. The tips presented in this article take little time to implement and can make the difference between passing now and passing after countless hours of redesign and retesting. Use these tips to improve your EMI and pass that test.