Switching noise PCB layout is where most converter interference problems are won or lost. The noise is generated by fast current changes flowing around physical loops, and the size of those loops is set on the board rather than in the schematic. Adding filters to a badly laid out converter treats the symptom while the source keeps radiating.
The loops that matter are the ones with discontinuous current
In any switching converter some current paths carry a smoothly varying current and some carry current that starts and stops abruptly each cycle. The abrupt ones are what generate interference, because a large rate of change of current in a loop produces a correspondingly large magnetic field and a voltage across any inductance in that path. In a buck converter the critical loop runs from the input capacitor through the high-side switch, through the low-side switch and back to the capacitor, and current in it turns on and off within nanoseconds. Identifying these loops before routing is the single most useful analytical step available. Once identified, the rule is simple: make their enclosed area as small as physically possible, and do not let anything else share their return path.
Why small loops matter more than good components
Every loop has inductance roughly proportional to its enclosed area. That inductance multiplied by the rate of change of current gives a voltage that appears wherever the loop touches the rest of the circuit, producing ringing at the switch node, overshoot on the switching devices, and a magnetic field that couples into anything nearby. Halving the loop area roughly halves those effects, and it costs nothing but attention. By contrast, adding a filter downstream cannot undo a field that has already coupled into a neighboring trace, and adding capacitance in the wrong physical place adds another loop rather than removing one. This is why converter manufacturers publish recommended layouts and why deviating from them in ways that grow the hot loops produces boards that fail compliance testing.
Placement decisions that do most of the work
Put the high-frequency input capacitors immediately adjacent to the switching device pins, with the shortest possible connections, before considering anything else. Keep the switch node copper small enough to carry the current comfortably but no larger, because it is the highest dv/dt node on the board and every extra square millimeter is an antenna. Place the feedback network away from the switch node and the inductor, and route the feedback trace as a quiet signal referenced to the output ground rather than running it past noisy copper. Keep the gate drive loop tight, since gate current also switches abruptly. Give the converter a contiguous ground reference directly under its own components, and do not route unrelated signals through that region.
Where filtering does belong
None of this makes filtering unnecessary; it makes filtering effective. Once the generating loops are small, an input filter can meaningfully reduce conducted emissions back onto the supply, and an output filter can reduce ripple reaching a sensitive load. Both work best when they are physically positioned so that the noisy side and the clean side do not share a return path, because a shared return lets noise bypass the filter entirely. A common and disappointing outcome is a well-chosen filter mounted so that its input and output grounds connect through the same piece of copper, giving the interference a route around the component. Spatial separation of the noisy and quiet regions, with a single deliberate connection between them, is what makes a filter perform as its datasheet suggests.
Measuring what you actually have
Measurement technique dominates results here. A standard oscilloscope probe with a long ground lead forms a large loop of its own and will show ringing that is partly its own creation, so use a short spring ground or a probe designed for the purpose. Measure at the device pins rather than at a convenient test point. A small hand-wound loop probe connected to a spectrum analyzer, or even to a scope, is an inexpensive way to find which parts of the board are radiating, and moving it around a working converter will identify the hot loops quickly. Compare readings before and after a layout change rather than relying on absolute numbers, and remember that the enclosure, the cables and the load all participate in the final result.
What to check before you act
- Identify every loop carrying discontinuous current before routing begins.
- Place high-frequency input capacitors at the switch pins, not merely nearby.
- Keep the switch node copper as small as the current allows.
- Probe with a short ground connection at the device pins.
Common questions
Will more decoupling capacitance reduce switching noise?
Only if it shortens the fast current loop. Capacitance placed far from the switching pins adds inductance in series with itself and may do nothing useful at the frequencies that matter.
Does slowing the switching edges help?
It reduces high-frequency content and often helps compliance, at the cost of higher switching loss and heat. It is a legitimate tool but a poor substitute for tight layout.
The practical takeaway
Find the loops where current changes abruptly, shrink them ruthlessly, and only then add filtering. Follow the converter manufacturer’s recommended layout where one exists, measure with a short-ground probe at the pins, and treat any layout change as something to verify rather than assume.
References and further reading
- FCC: electromagnetic compatibility and radio frequency safety questions answered
- FCC: radio frequency safety background and exposure guidance
- Department of Energy: how electronics losses turn into wasted energy and heat
Numerical scenarios are illustrative unless identified otherwise. Follow the exact product instructions; component ratings and local installation requirements can differ.
