PCB Manufacturing Tolerances: What a Fabricator Can Hold

Every feature on a printed circuit board is produced by a process that has a natural spread, and the tolerance on the drawing is a statement about how much of that spread the design can absorb. Manufacturing tolerances are therefore not a fixed property of a fabricator but a relationship between what the process can hold and what the design needs, and specifying tighter values than the application requires costs money without adding value.

Board Thickness Tolerance

Finished thickness varies because prepreg flows under pressure and its final height depends on the resin content, the copper area it must fill and the lamination cycle. A typical specification is plus or minus ten percent, which on a 1.6 mm board gives a range from about 1.44 to 1.76 mm. Impedance controlled boards often tighten that to five percent and high reliability boards to three.

This matters because thickness interacts with several things at once: the dielectric spacing that sets impedance, the fit of a card in a guide, the thermal path through the board and the mechanical stiffness. A design that controls thickness tightly for the impedance but ignores the connector fit has solved half the problem, and the dimensions as a whole are covered in the discussion of board outline and mounting design.

Hole Tolerance

Hole tolerance covers both the diameter of the finished hole and its position on the board. Mechanically drilled holes typically hold about plus or minus 0.075 mm on diameter, laser drilled holes about plus or minus 0.05 mm and microvias around plus or minus 0.025 mm. Positional tolerance is a separate figure and is usually expressed with reference to the datum and the panel, not to the individual feature.

The distinction between drill size and finished size matters in practice. Plating adds copper to the wall, so the hole after plating is smaller than the drilled hole, and a press fit pin or a connector that has to fit should be specified against the finished dimension. Getting that wrong produces a board that meets the drill drawing and does not accept the component.

Trace Width and Spacing

Etching removes copper in every direction, so a trace finished narrower than the artwork predicted and the space beside it finished wider. The trace width tolerance on a conventional board is commonly around twenty percent of the nominal width, which is generous enough for most signal and power routing. On HDI boards the figure tightens to about ten percent and on very fine line work to five.

The consequence is not cosmetic. A trace that is narrower than intended has higher resistance and, on a controlled impedance layer, a different impedance. Where the layer is carrying a high speed differential pair, a few percent of width error translates directly into a reflection. Fabricators compensate for the etch factor in the artwork, which is one of the reasons the copper thickness and the finished line width have to be specified together.

<img src="https://www.gopcba.com/wp-content/uploads/2026/08/Final-Inspection.jpg" alt="Measurement of PCB layer registration and drilled hole tolerance” />

Layer Registration and Copper Thickness

Layer registration is the alignment of the patterns on different layers of a multilayer board, and it is one of the hardest tolerances to hold because every lamination cycle adds movement. A conventional multilayer board typically holds about plus or minus 75 microns and an HDI board about plus or minus 25 microns. Where registration drifts, the failure is a via that breaks out of its pad or an inner layer short that only appears after assembly.

Copper thickness is usually quoted to within ten percent, and the figure that matters for reliability is the minimum thickness in a plated barrel rather than the nominal surface value. On a heavy copper board plating uniformity becomes the limiting factor, since the same current has to produce a consistent deposit across a large panel and down every hole.

Tolerances on Mask and Legend

Solder mask registration determines how much of each pad remains exposed and how wide the strip of mask between adjacent pads ends up. A drift of even a few tens of microns changes the wettable area on one side of a pad and reduces the dam on the other, which shows up at assembly as a cold joint next to a bridge on the same device.

Legend is less critical but still has a specification. Print that overlaps a pad interferes with soldering, and print that is displaced far enough to be ambiguous is a documentation defect rather than a functional one, but both are avoidable with a sensible tolerance rather than a tight one.

Standards and Cost

IPC-6012, the performance standard for rigid boards, defines the acceptance criteria for hole quality, annular ring, registration and copper thickness, and it is classified into three levels of increasing severity. The most demanding level is applied to medical, aerospace and automotive products, and citing the standard in the fabrication drawing turns a tolerance question into a documented requirement.

Tighter tolerances cost more because they consume process capability. Impedance control typically adds ten to twenty percent, a high reliability acceptance class fifteen to thirty, laser microvias twenty to forty and fine line etching fifteen to thirty five. Each of those is a yield and inspection cost, and the design should ask for them only where the function requires them.

What Drives the Variation

Material behaviour is the first factor. Laminates expand when heated and shrink as they cool, and different materials do so at different rates, so a stack that combines two of them moves differently from a single material build. Low expansion materials improve stability and cost more, and the dimensional behaviour involved is described in the discussion of dimensional stability.

Equipment and process control are the second. Laser direct imaging removes the film tool and its dimensional error, automatic optical inspection catches drift before it becomes a systematic defect, X-ray verifies registration through the stack and a controlled plating line holds copper distribution. Layer count and design density are the third: a twenty layer server board is inherently harder to hold than a four layer board, and a fine pitch array with microvias narrows every window at once.

Specifying Tolerances Sensibly

Set tolerances from function. A mounting hole needs whatever the hardware needs, an impedance controlled trace needs the value the simulation assumed, and a decorative outline needs very little. Reviewing those choices against the process capability, as described in manufacturable design guidelines, avoids asking for something the process cannot deliver. Applying the tightest available tolerance everywhere increases cost significantly and produces nothing except a more expensive board and a lower yield at the fabricator.

Where a tight tolerance is genuinely required, state the reason on the drawing. It helps the fabricator choose the process, and it helps both parties decide what to do when the measurement lands at the edge of the range but the function is still satisfied, which is a conversation that is much easier to have before the panel is built than after.

FAQ

What are the standard tolerances on a PCB? Roughly plus or minus 0.05 to 0.08 mm on drilled hole diameter, plus or minus ten percent on board thickness, about twenty percent on trace width and about 75 microns on layer registration for a conventional multilayer board.

When should tighter tolerances be specified? When the function requires them: impedance controlled traces, fine pitch packages, connectors that must fit a mating part, or registration on a high density stack. On dimensions that only affect appearance, a standard tolerance is sufficient.

Which standard covers PCB tolerances? The performance standard for rigid boards is the usual reference, supported by the design standard and the visual acceptance standard. It defines acceptance criteria in three classes, with the highest class used for medical, automotive and aerospace work.

Why does a tight tolerance cost so much more? Because it consumes process capability. A narrower window means more setup, more inspection, more rejected panels and a slower process, and each of those is reflected in the price of every board built to that specification.