PCB Trace Design: Types, Materials, and Width Rules
Every circuit is only as good as the copper that connects it. A PCB trace carries both power and signal, and its geometry decides whether a design performs as intended or behaves differently on every board. The parameters involved, width, spacing, copper weight, and length, are all chosen early and are all difficult to change later.
The types of trace
Signal traces carry analog or digital information, and their width and routing are set by frequency and by signal integrity requirements. Power traces carry larger currents and are wider and more carefully planned. Ground traces form the return path, and in high-frequency designs they are usually replaced by a solid plane rather than a set of individual conductors.
Differential pairs are routed as a matched two-conductor line carrying equal and opposite signals, and they are used by interfaces such as USB, HDMI, and Ethernet. Microstrip and stripline describe where the conductor sits rather than what it carries: a microstrip runs on an outer layer, which makes it easier to probe and adjust, while a stripline runs between reference planes on an inner layer, which is preferable when electromagnetic requirements are strict.
Materials
Copper type matters more than it appears. Electrodeposited copper is inexpensive and used in the majority of rigid boards, while rolled and annealed copper is more ductile and is specified for flexible circuits and applications with repeated movement.
The base material then decides the electrical behaviour. Standard FR-4 is cost-effective and adequate for most circuits, but its loss rises above roughly a gigahertz. Low-loss laminates and polytetrafluoroethylene are used for high-frequency, radio, and microwave work, and polyimide is chosen where high temperature tolerance is required, including flexible and defence applications.
Surface finish completes the picture. An immersion gold finish offers flatness and oxidation resistance for dense and high-speed boards at a premium over hot air levelling, which remains the economical choice for general designs. An organic solderability preservative suits lead-free processes but has a short storage life, which is a scheduling constraint as much as a technical one. Our guide to trace width and current calculation takes the geometry further.
Width, spacing, and copper weight
Width is chosen against current for power traces and against impedance for high-speed signals. Signal traces commonly run from 4 to 10 mil, which is roughly 0.1 to 0.25 mm, while power traces can exceed 100 mil. As a working reference on one ounce copper, a 10 mil outer-layer trace carries about one ampere safely.
Spacing is set by the voltage between conductors and by the risk of shorting or breakdown. Six mil is a reasonable minimum for low-voltage circuits, while high-voltage circuits follow the applicable international design standard and can require more than 30 mil.
Copper weight is the third input. One ounce, about 35 micrometres, is the default. Two ounces serves medium and high current requirements, and three ounces or more is used for automotive and power designs. The combination of width and copper weight is what determines current capacity, so specifying one without the other leaves the calculation open.
Width and copper weight are one decision. Doubling the copper has the same effect on current capacity as roughly doubling the width, at a different cost.
Width calculation in practice
Width is normally taken from a current-capacity chart derived from the applicable IPC design standard, with temperature rise as the input. As a working reference on outer layers, one ampere is served by about 10 mil of one ounce copper, three amperes by about 35 mil of two ounce copper, and five amperes by about 100 mil of three ounce copper.
Two corrections matter. Inner-layer traces dissipate heat less effectively than outer-layer traces, so their width should be increased by roughly half for the same current. And a margin of 20 to 30 percent should be added on top to absorb current variation and ambient temperature rise.
A worked example makes the method concrete. A 5 V rail carrying 3 A is a candidate for two ounce copper on an outer layer with a trace around 35 mil wide, with the margin applied on top. Choosing one ounce copper instead would roughly double the required width, which may not fit the available area.
Impedance, length, and via count
Above roughly 100 MHz, the design stops being about resistance and becomes about impedance. The trace width, spacing, dielectric thickness, and copper weight have to be calculated together to reach the target, commonly 50 ohm for a single-ended line and 100 ohm for a differential pair.
Length and via count then decide how well that target survives. Short, direct traces reduce delay and reflection, and every additional via introduces a discontinuity and a return-path question. A high-speed net should be planned as a path with a defined reference from end to end, not assembled from convenient segments.
Routing practices
Four habits cover most of the quality difference. Replace 90 degree corners with 45 degree segments to limit radiation and avoid the acid traps that etching leaves in a sharp corner. Plan the ground early, using a star point or a large ground plane so that return currents do not share a path with noisy circuitry.
Route differential pairs in parallel with equal length and equal spacing, so the two conductors see the same environment and arrive together. Isolate high-speed signals from their neighbours, using a ground guard where spacing alone is not enough, and keep clock and radio-frequency paths as short as the layout allows. The geometry behind guard traces is covered in our guide to the 3W rule for crosstalk, and the corner and pair geometry that preserves impedance is examined in right angle routing of differential traces.
Return path is the parameter that is easiest to leave unplanned and most expensive to fix, because it is decided by the plane rather than by the trace.
Common mistakes
A trace that is too narrow for its current heats up and drops voltage, and the failure appears as an intermittent fault rather than as a clean break. Insufficient spacing causes breakdown or shorts, particularly where the voltage is high, and it is a design error rather than a manufacturing one.
Ignoring the return path is the third mistake, and it produces interference and waveform distortion in circuits whose traces look perfect. Mismatched differential pairs are the fourth: an unequal pair produces skew and distortion on interfaces such as USB, low-voltage differential signalling, and HDMI, and it is invisible until the link fails.
High-frequency and radio-frequency techniques
Four techniques carry most of the additional performance. Controlled impedance routing is applied to any signal above roughly 100 MHz, where an uncontrolled line distorts the waveform. Length matching is applied to parallel buses and serial links such as memory, PCIe, and Ethernet, where bit arrival times have to line up. Guard traces are added beside analog and radio-frequency signals to raise their immunity. And termination resistors are placed to absorb reflections at the end of a line.
The common thread is that each technique replaces a rule of thumb with a calculation. High-frequency work does not have a simpler version of good practice; it has less tolerance for its absence.
Choosing design tools
The tooling matters less than the data it produces. A full-featured suite supports stackup definition and impedance simulation, which is useful when the design has controlled-impedance nets. Open-source packages handle ordinary two and four layer boards adequately. Current-capacity calculators based on the applicable design standard are useful for the width decision, provided the assumptions they use are the ones the board will actually see.
Whichever tool is used, two outputs decide the manufacturing outcome: a clear stackup definition and a fabrication drawing that states what the traces are supposed to achieve. Everything else can be reworked; those two are the inputs the fabricator cannot improve.
FAQ
How narrow can a trace be? Conventional processes hold about 4 mil as a minimum, and advanced shops reach 2 mil. Narrower features cost more because they demand tighter imaging and etching control.
Can a trace carry 10 A? Yes, with a very wide conductor, heavier copper, or multiple layers in parallel, and usually with additional heat sinking. It is a thermal problem as much as an electrical one.
Does trace width affect signal speed? It affects the impedance and the capacitance of the line. Resistance falls as width increases, but capacitance rises, so width is chosen to meet the impedance target rather than for resistance alone.
