DC-DC Inductor Placement and Radiated Emissions

A switching supply that fails radiated emissions is usually attacked by changing the inductor, adding filter capacitors and improving the ground plane. Those measures address the symptoms, and they often produce only a few decibels of improvement. In one representative case, a 60 volt to 12 volt stage exceeded the limit across the 50 to 300 MHz band with the worst point more than ten decibels over at 200 MHz, and replacing the inductor, adding input and output capacitance and reworking the ground gave roughly five decibels in total.

The cause turned out to be a placement decision rather than a component choice: the output inductor was on a different layer from the switching regulator. Once the inductor was moved onto the same layer as the switch node it connects to, the excess disappeared with margin to spare, without any additional filtering or shielding. The example is worth understanding because the mechanism generalises, and because it is invisible in a component list.

What a Power Loop Actually Is

A switching converter stores energy in the inductor and releases it to the output, and the current that does this circulates in a loop bounded by the input capacitor, the switch, the inductor and the output capacitor. That loop is the primary radiating structure of the supply. Its area, not its component values, determines how much field it produces, and radiation rises steeply with loop area rather than linearly.

The correct inductor placement follows from that fact. The device, its input capacitor, the switch node and the output capacitor should sit on the same layer, as close together as the thermal and mechanical constraints allow, so that the loop can be closed in copper on one plane with wide, short connections. When the inductor is placed on another layer, the loop must leave the layer and return through vias, which lengthens it in the third dimension as well.

Three Effects of Cross-Layer Placement

The first effect is geometric. Cross-layer connections require vias, and a via pair carrying the switching current adds both length and inductance to the loop. What was a compact few square centimetres on one layer becomes a structure several times larger once the vertical excursions are included, and the radiated field grows with the area.

The second effect is that the cross-layer trace behaves as an antenna. The inductor produces a strong alternating magnetic field, and a long conductor connected to it radiates that field directly and can also couple interference from other layers. In the 50 to 300 MHz range, where board traces form resonant structures at relatively short lengths, the cross-layer connection is a particularly efficient radiator, which is why conventional shielding and filtering measures make so little difference.

The third effect is that the filter stops working. Input and output capacitors suppress high frequency ripple only if their connection to the device and the inductor has low impedance. Cross-layer routing increases the equivalent series inductance and resistance of that connection, so the capacitor that was added to fix the problem contributes less than expected, and each additional capacitor yields less improvement than the last.

Layout Rules for a Switching Cell

The rule that follows from the case is simple to state: the device, the output inductor and the input and output capacitors belong on the same layer. In practice that means choosing the layer at the placement stage, before the routing is committed, and treating the power loop as a constraint rather than as a preference. The switch node should be a compact copper shape with no unnecessary via, and the connection from the inductor back to the output capacitor should be wide and short.

Ground return deserves the same attention. A continuous plane beneath the cell gives the return current a low impedance path directly under the outgoing trace, which minimises the loop without any additional component. If the second layer is used for signal routing instead, the return current is forced to divert around the breaks, and the effective loop area grows for exactly the reason described in DC-DC converter layout and routing. Where the application also has thermal constraints, the copper that spreads heat can usually be arranged to serve as the return path as well, following the practice in thermal management PCB design.

Two details are worth checking before release. The first is the plane structure under the switch node: a plane that is split for another purpose can force the return current around an obstacle, and the general rules for keeping a reference continuous are covered in the power plane design guide. The second is the position of the filter. Input and output capacitors placed at the far side of a connector do not filter the loop current, and the suppression measures that work on a switching cell are described in EMI suppression principles.

Verifying the Change

The verification method matters as much as the fix. Radiated emissions from a supply are best checked with a near field probe swept over the board, which identifies the cell that radiates before the measurement is repeated in a chamber. Comparing the scan with the switching current waveform confirms that the radiator is the power loop rather than a cable or a connector, which would need a different solution.

A final measurement should be made on the assembled product rather than on a bare board, because the enclosure, the cables and the position of the board inside the product all change the result. Where the margin after the change is small, the responsible step is to look again at the loop rather than to add a shielding can, since the can addresses the symptom and leaves the design sensitive to any later change in layout or load.

Choosing the Layer Before Routing

Layer decisions are usually made when the stackup is defined, long before the switching cell is placed, and by then the constraint has been forgotten. A better sequence is to identify the components that form each high current loop, decide which layer they will occupy, and then confirm that the stackup supports that choice. Inductor placement is a stackup question as much as a layout question, because the answer determines whether the loop can be closed in a single layer or has to pass through vias.

Where the board has several switching supplies, the same reasoning applies to each one separately. Sharing a layer between two cells is acceptable only if each loop remains compact; placing the inductors of two converters next to each other so that their fields overlap creates a coupling path that no filter addresses. Grouping the cells by function and keeping the sensitive analogue area away from all of them is the arrangement that holds up in a chamber measurement.

FAQ

How much does placement alone affect emissions? In the case described, moving the inductor to the regulator layer removed more than ten decibels of excess at the worst frequency, which is more than the combined effect of replacing the inductor and adding filtering.

Can a shielded inductor compensate for a poor loop? Partly. A shielded part reduces the field it radiates directly, but the loop formed by the connections still radiates, and the cross-layer routing remains. Shielding treats one contributor and leaves the geometry unchanged.

Does the same rule apply to a boost converter? Yes. Any topology with a switching loop has the same requirement: keep the loop compact, keep it on one layer where possible, and give the return current a continuous path directly beneath it.