DC-DC Converter Layout and Routing Guidelines
A switching regulator converts voltage efficiently because it never operates the pass device in its linear region. The cost of that efficiency is a node that switches between the input rail and ground at hundreds of kilohertz to several megahertz, and every trace connected to that node behaves as part of a loop antenna. The performance of a converter, including its efficiency, its output ripple and its emissions, depends far more on the layout than on the choice of controller.
How the Converter Makes Its Noise
The regulator chops the input voltage into a high-frequency pulse train using a MOSFET pair and stores energy in an inductor and capacitor, then rectifies and filters the result into a stable output. Switching frequencies typically range from a few hundred kilohertz to several megahertz. That range gives the converter its small size and high efficiency, and it also means the loop carrying the switched current changes its current abruptly at every switching event.
The rate of change of that current, combined with the area of the loop it flows in, determines how much field the board radiates. This is the single most important geometric relationship in a power supply layout, and it is the same relationship that governs radiated EMI in switching regulator layout.
<img src="https://www.gopcba.com/wp-content/uploads/2026/08/hdi-pcb-design-guidelines-11-b6900f71.webp" alt="DC-DC converter layout with a tight switching loop” />
The Switching Loop Comes First
The high-current loop runs from the input capacitor, through the high-side switch, through the low-side switch or diode, and back to the capacitor. The inductor current is continuous and relatively benign; it is the pulsed current that matters. The loop must therefore be drawn as a small, closed shape, and the input capacitor must be the physical neighbour of the switch node rather than a component placed on the other side of the board.
Placement order follows from that requirement. The input capacitor goes first, adjacent to the device pins, with the shortest possible connection to both the input rail and the ground. The inductor comes next, close to the switch pin. The output capacitor follows the inductor, and only then are the remaining components arranged around what is left. A layout that places the controller first and then fits the capacitor nearby has already lost the argument, and no amount of copper added afterwards will recover the performance.
The ground connection between the input capacitor and the device ground pin deserves separate attention. That connection is part of the pulsed loop, so it should be a short, wide trace or, better, a direct connection to a local ground plane under the device. Connecting it through the general board ground plane adds inductance and defeats the purpose of the tight loop.
Grounding and the Analog Ground Pin
Most controllers provide a separate signal ground pin for the internal reference and the feedback divider. This pin carries almost no current, and its reference must not be disturbed by the pulsed ground currents of the power stage. The correct arrangement is to connect the signal ground pin to a quiet point on the ground network, close to the output capacitor, and to keep the power stage currents on their own path back to the input capacitor.
A single continuous ground plane is usually the simplest way to achieve this, provided the components are placed so that the two current paths do not overlap. Where a split is used instead, the connection between the two grounds should be a single defined point rather than a broad overlap, and the consequences of getting that wrong are described in this discussion of power plane splitting.
Thermal Vias and Copper Area
A switching regulator loses power in the switches, the inductor and the copper, and the loss must leave the board. Controllers and integrated switches in small packages rely on the copper underneath them as their heatsink, which means the layout must provide both area and a path to the other side of the board.
Thermal vias are that path. A group of vias placed in the thermal pad of the device carries heat to a copper region on the opposite layer, where it spreads and dissipates. The vias should be small enough that solder does not wick away during assembly, and they should be arranged so that the paste is not starved; the same geometric rules that apply to pads in general apply here, as described in this article on PCB pad design standards. The copper region on the far side should be as large as the available space allows, and it should be connected to the ground plane where the design permits.
Feedback and Compensation Routing
The feedback trace carries a small, precise voltage from the output to the controller, and it is the most sensitive analog connection on the board. It should be routed away from the switch node, away from the inductor and away from the diode, and it should pick up its reference at the point where the output voltage is actually measured, which is normally the output capacitor rather than a convenient via.
The compensation network belongs immediately adjacent to the controller pins, not distributed around the board. Any capacitance introduced by long traces between the compensation components and the controller changes the loop response and can destabilise a design that was stable on paper. The feedback divider should be placed so that the resistor connecting to ground returns to the quiet ground point rather than to a point carrying pulsed current.
Practical Checks
Before the layout is released, trace the pulsed current on the artwork with a pencil and confirm that the loop it forms is as small as the design allows. Confirm that the input capacitor connects directly to both the input pin and the ground pin of the device, that the thermal pad has an adequate via array and copper area, that the feedback trace is short and clear of the switching node, and that the compensation network is adjacent to the controller. Taken together, those four checks account for most of the difference between a quiet supply and one that fails an emissions test. Where the supply feeds a sensitive analog section, the additional layout rules in this discussion of ground current and harmonic distortion should be applied as well.
FAQ
Why does the input capacitor placement matter more than the output capacitor? Because the input capacitor carries the pulsed current of the switching loop, which is the current with the fastest edge rate and the largest field. The output capacitor sees the much smoother inductor current, so its position affects ripple but has far less influence on emissions.
Should the converter ground be split from the system ground? Usually not. A single continuous plane with careful component placement works better, provided the power stage currents are kept on their own path back to the input capacitor. If a split is used, join the grounds at one defined point rather than over a broad area.
How many thermal vias does a switching regulator need? Enough to carry the heat out of the package without starving the solder joint during reflow. The vias should be arranged across the exposed pad with spacing that leaves room for paste, and the copper on the far side should be as large as the layout allows.
