Somehow you will always need to feed power to your system, this means you have a cable going to your system. Now, if you have noise generated from your system on this cable, it will start radiating. So the goal is to keep this line as clean as possible. Typically you see on this line all the noise generated by the first DC/DC, in the example above the 24 V to 5 V buck regulator plus some noise (typically higher frequency) that will find the way to the input.

Typically the noise you see here has some lower frequency components from the DC/DC such as the switching frequency plus harmonics, as well as high frequency components. The low frequency parts can be typically eliminated effectively with a Pi filter, L3, C42 and C44 in the example in Figure 2-4. Here, you need to be a bit careful to not build a resonant tank that makes things worse, so a lossy aluminum capacitor (with significant ESR) is a good choice. Power Stage Designer offers a filter designer, that can help here. (POWERSTAGE-DESIGNER)

The high frequency components can ignore this filter and use the parasitic capacitance of this inductor to pass through the filter. Therefore it is useful to add a ferrite bead in addition, this can reduce the noise in the two and three digit MHz range.

Lastly, there can also be some common mode noise, you cannot eliminate this way, but an additional common mode choke directly at the input connector can reduce this.

If all these measures are really needed to pass EMI conformance tests is hard to judge upfront, but better place them on the first PCB, if it is not needed, you can remove them in the final product. With a current clamp you can measure whether the noise you see on the lines is a common or differential mode noise.

Figure 2-4 External Power Supply

For the layout of such a filter the layout can be simplified into three stages, the high frequency filter, the low frequency filter and the common mode noise filter.

For the high frequency filter, a ferrite bead, similar rules apply as for the internal supply rails. You want to keep the noise where it is generated and not have it travel everywhere. This is done as before, by placing a ferrite bead close to the noise source. Figure 2-5 shows how this is implemented in the layout.

A buck converter always generates more noise on the input side than on the output side, whereas with a boost converter it is the other way round, due to the topology LC combination. So the buck converter needs to have a filter, it draws fast rising currents from its input capacitor, that is already filtering a lot, especially the lower frequencies but it is not able to do much at higher frequencies. Looking at the frequency domain, the noise consists of the switching frequency, all the harmonics and noise caused by parasitic capacitances and inductances. As the rise times are designed to be as fast as possible (to reduce switching losses) the harmonics can go up to three digits MHz. As example a buck converter with 1 MHz switching frequency has a rise time in the single digit nano second range, resulting in several 100 MHz of noise spectrum. The input capacitor of 4.7 µF already has more than 100 mΩ impedance at this frequency, as the parasitic series inductance takes over. The ferrite bead can do a good job in keeping this away from your power line.

Besides that, of course you need to continue to use the general design rules for a DC/DC converter, otherwise the ferrite bead cannot save you.

Figure 2-5 DC/DC Filter Layout

Second step, the previous mentioned filter has eliminated a lot of the high frequency noise, but the fundamental switching frequency and some harmonics are still left on the line. Here the Pi filter as shown in the schematic can help. The placement is not that critical anymore, as we are taking about lower frequencies. In Figure 2-6, the filter is shown, try to minimize capacitive coupling from the input to the output and connect capacitors in a way the current has to pass them.

Figure 2-6 Pi Filter Layout

Also shown in Figure 2-6, is the common mode choke. A common mode choke should reduce all common mode noise, meaning all noise that is the same on both supply lines. This is typically higher frequency noise that couples in a capacitive way into supply planes and from there to the supply traces. As this originates for example the ground plane, it is necessary to cut out this plane where the common mode choke is located and the traces to the connector are. Otherwise the effect of the common mode choke can be reduced.

Table 2-1 is a quick checklist to verify a schematic and layout.