Home › Guides › PCB Stackup Design

PCB Stackup Design Explained: The Layers Your Board Is Built From

A printed circuit board is not a single sheet of copper with tracks drawn on it. It is a sandwich of copper layers, insulating dielectric layers, solder mask and silkscreen, laminated together into one rigid board. The stackup is the recipe for that sandwich: which layers are copper, what each copper layer is used for, how thick each dielectric is, and in what order the layers sit. The stackup decides whether signals stay clean, whether the board warps during manufacture, whether high-speed signals arrive intact, and whether the board passes electromagnetic compatibility testing. This guide explains how stackups work, how to choose one, and the decisions that matter most.

The board-design workflow around this guide is covered in our KiCad tutorial for beginners and PCB design with AI, and the parts that sit on the board in reading datasheets. For the signals themselves, see electronics circuits basics.

What a stackup actually is

Start with the simplest board: two copper layers separated by a dielectric core, with solder mask on the outside faces and silkscreen marking the component positions. That is a two-layer stackup, and for simple circuits it is entirely sufficient. The top layer carries components and tracks, the bottom layer carries tracks and often a ground pour, and vias — small plated holes — connect tracks between the layers.

A four-layer board adds two inner copper layers. The usual arrangement is signal on the top, ground on the first inner layer, power on the second inner layer, and signal on the bottom. The inner layers are laminated inside the board, invisible from the outside, and they transform what the board can do: the ground plane gives every signal a close return path, the power plane distributes supply rails with low impedance, and the two signal layers gain routing room because power and ground no longer consume track space on the outer layers.

Beyond four layers, the same pattern extends: pairs of signal layers separated by plane layers, with the total stackup kept symmetrical top to bottom. Symmetry matters for a reason that has nothing to do with electronics: a board whose copper is heavily weighted to one side can warp during lamination and soldering, because the materials expand differently. Manufacturers will warn you about an unbalanced stackup, and heeding that warning is cheaper than receiving warped boards.

Choosing a layer count

Layer count is a trade-off between cost, routing room and electrical performance, and the right answer depends on the circuit, not on ambition. Two layers suit simple, slow circuits: a microcontroller board with a handful of peripherals, a sensor board, a power supply with modest currents. Routing is a puzzle on two layers, because every crossing needs a via, but the puzzle is solvable, and the board is cheap.

Four layers are the usual step up, and the step is worth taking when any of these appear: routing that will not fit on two layers without heroic effort; a solid ground plane needed for noise control; controlled impedance for high-speed signals; or a mix of analogue and digital sections that need a quiet ground. The cost increase from two to four layers is real but modest compared with the cost of a two-layer board that fails its noise tests or cannot be routed at all. Six layers and beyond are justified by density — fine-pitch parts, many high-speed signals, or several supply rails — and at that point the stackup design becomes a specialist task in which the manufacturer's capabilities and the signal integrity requirements drive every choice.

A useful rule: choose the smallest layer count that gives you a solid ground plane and comfortable routing, and no more. Extra layers bought for comfort add cost and manufacturing complexity without improving the circuit, while too few layers force compromises — split grounds, long detours, cramped routing — that show up as noise, crosstalk or failed tests.

Signal, ground and power planes

Every copper layer in a stackup has a job, and the jobs fall into three kinds. Signal layers carry tracks between parts. Ground planes are solid copper areas connected to ground, and they are the most important single feature of a good stackup. Power planes are solid or split copper areas carrying supply rails.

The ground plane earns its importance through return currents. Every signal current flows out along a track and returns to its source along the ground, and the return current follows the path of least impedance — which, at high frequencies, is the path directly beneath the signal track, on the nearest ground plane. A solid, unbroken ground plane beneath a signal layer gives every track a tight return loop, which keeps radiated noise low and keeps the track's impedance predictable. Break that plane with a slot, a split or a large cutout, and the return current must detour around the gap. The loop area grows, the noise grows, and the impedance changes where the track crosses the gap. That is why the advice "never split the ground plane" is repeated so often: splits are sometimes necessary, for example to separate noisy digital ground from quiet analogue ground, but every split is a cost paid in noise, and it should be a deliberate decision with a single controlled connection point, never an accident of routing.

Power planes distribute supply rails with low resistance and low inductance, which matters when parts draw fast-changing currents. A power plane also forms a capacitor with the neighbouring ground plane, which helps supply decoupling at high frequencies. Where several rails exist, a power layer is often split into regions, one per rail, with the splits arranged so that no high-speed track crosses a split on an adjacent signal layer. Crossing a split in the reference plane beneath a fast track has the same effect as splitting the ground: the return current detours, and the signal degrades.

Impedance control basics

At low speeds, a track is just a wire: its resistance matters, its shape does not. At high speeds, a track is a transmission line, and its characteristic impedance — set by the track width, the copper thickness, the distance to the nearest plane and the dielectric material between them — determines whether a signal arrives cleanly or reflects back and corrupts itself. Controlled impedance means the manufacturer builds the board so that a specified track has a specified impedance, within a tolerance, by controlling exactly those dimensions.

The designer's job is to specify the target impedance and the tracks it applies to, usually from the requirements of the interface in use: common targets exist for USB, Ethernet, HDMI and memory interfaces, and the interface specification or the chip datasheet states them. The manufacturer's job is to adjust the track width and dielectric thickness to hit the target, and to verify it with test coupons — small test structures built on the same panel as your board and measured after manufacture. You do not achieve controlled impedance by guessing a track width; you achieve it by specifying the target, letting the manufacturer tune the geometry, and checking the coupon results.

Two practical points follow. First, impedance depends on the distance to the nearest plane, so a track's impedance changes if it moves to a layer with a different dielectric thickness above its plane. Keep controlled-impedance tracks on the layer they were designed for, and route them over an unbroken plane. Second, the dielectric material matters: standard FR-4 is adequate for many designs, but at higher frequencies its losses and its variation with frequency become significant, and a lower-loss material may be needed. That is a conversation with the manufacturer, driven by the frequencies your design actually uses, not by a desire for premium materials.

EMI: keeping the board quiet

Electromagnetic interference is the stackup's report card: a board that radiates too much, or is too easily disturbed, fails compliance testing and cannot be sold. The stackup decisions that most affect EMI are the ones already covered: a solid ground plane close to signal layers, short return loops, no tracks crossing plane splits, and decoupling capacitors placed close to the parts they serve, with short connections to power and ground planes. These are not exotic techniques; they are the basic discipline of a good stackup, and they account for most of the difference between a board that passes first time and one that needs shielding, ferrites and rework to scrape through.

Two further habits help. Keep fast, noisy signals — clocks, switching converter outputs, high-speed data — away from board edges and away from connectors, where their fields escape most easily. And keep analogue sections, with their small signals, physically separated from digital sections, with their own quiet ground region connected to the main ground at a single point. Neither habit costs anything at design time, and both are expensive to retrofit after the board is built.

A worked example: a four-layer sensor board

Consider a modest sensor board: a microcontroller, a few sensors, a wireless module and a USB connector. On two layers, routing is possible but cramped, and the wireless module's high-frequency signals would run over a ground pour broken by tracks, which is a recipe for noise. On four layers, the stackup becomes: top signal, ground plane, power plane, bottom signal. The microcontroller and sensors sit on top, their tracks route on the top layer and drop to the bottom layer through vias where they must cross. The ground plane is solid and unbroken beneath both signal layers. The power plane carries the main rails, split into regions for the digital supply and the sensor supply, with no fast track crossing a split. The wireless module's antenna area is kept clear of ground copper, as its datasheet requires, with the keep-out respected on all layers. This board costs more than the two-layer version, but it routes cleanly, its signals are quiet, and it passes testing without rework — which is the real measure of a good stackup choice.

Frequently asked questions

Related guides

Complete the board-design cluster: KiCad tutorial for beginners for the design workflow, reading datasheets for choosing parts, and PCB design with AI for the wider workflow. The rest of this cluster covers loop diagrams explained. Then browse all guides for the wider LOTICI library.