PCB Flatness Measurement Methods and Acceptance Limits
Flatness is one of the few board properties that is easy to feel and hard to measure. A panel that rocks on a table is obviously warped, but the same panel measured with a height gauge may pass, because the result depends entirely on how the board is supported and where the measurement is taken. PCB flatness measurement is therefore as much about the method as about the instrument, and the acceptance limits are meaningless unless the measurement conditions are defined with them.
Why Flatness Matters
An assembly line expects a board to sit flat on the conveyor, on the printer table and on the reflow support. A board that is warped lifts at the corners, which changes the paste deposit height, the placement height and the contact with the support pins, and the resulting defects appear as inconsistent printing rather than as a flatness problem.
Flatness also matters after assembly. A warped board can prevent a connector from mating, can crack a solder joint at a corner, and can cause a heat sink or a shielding can to sit unevenly. Where the product stacks boards or slides them into a chassis, the flatness of the bare board sets the tolerance available for everything that follows, and a board that is at the limit of its own specification can still be unusable in that assembly.
Warpage, Bow and Twist
Warpage is the general term for the distortion of a board from a plane. Bow describes a curvature in which the board takes a cylindrical shape, similar to a shallow dish, while twist describes a condition in which the corners are alternately high and low so that the board will not sit on three points without rocking.
The two behave differently in assembly and are measured differently. Bow is usually expressed as the maximum deviation divided by the length of the board, while twist is expressed as a maximum vertical difference between corners. A limit that quotes a single percentage without stating which condition it applies to is not a usable specification, and quoting one without the other is a common cause of disputes between a fabricator and a customer who measured differently.
The Reference Plane Problem
All flatness measurement depends on a reference plane, and the reference plane has to be defined before a number can be quoted. A board can be measured in free state, resting on its own weight on a surface plate, or constrained on the datum points used in assembly, and the three measurements will differ substantially.
For most products the meaningful measurement is the one made in the condition the board sees in use. For a bare board order, the usual convention is a free state measurement with the board resting on a flat surface, and that convention should be stated on the drawing or in the specification that applies.
Shadow Moire and Optical Methods
Shadow moire measures out of plane displacement by projecting a grating onto the board and observing how the pattern distorts with height. It gives a full field map rather than a few points, which makes it well suited to detecting twist and to visualising a bow, and it can be repeated on the same board before and after a process step.
The method has limits. Very shiny surfaces can interfere with the pattern, and highly reflective or transparent coatings may need a preparation step. The equipment also has its own calibration and reference requirements, so a measurement is only as good as the setup that produced the image, and the same board measured on two systems can produce two different maps unless the reference is agreed.
Contact Measurement and Surface Plates
A surface plate with a dial indicator measures the height of defined points relative to the plate and remains the reference method for many shops. It is simple, it is traceable and it produces numbers that can be recorded, and its main limitation is that it only measures the points that are actually touched.
Where the board is large or flexible, the support itself affects the result. Measuring a thin board on a surface plate allows it to conform to the plate, which can hide a bow that would appear if the board were supported only at the corners. The support arrangement is part of the measurement and should be described with the result.
Measurement Conditions and Thermal State
Temperature affects the result because the materials in the board expand at different rates and because the laminate absorbs moisture. Measuring a board immediately after a hot process gives a different result from measuring it after it has returned to room temperature, and measuring it in a humid room gives a different result again, particularly on a thin panel.
Conditioning before measurement is therefore part of the method. The usual approach is to allow the board to stabilise at a defined temperature and humidity for a defined time, and to record the conditions with the result. Without that control, comparing measurements taken on different days is not meaningful, and a trend that looks like a process change may simply be a change in the weather.
Acceptance Limits and Standards
Acceptance limits are normally expressed as a maximum bow and twist as a percentage of the diagonal or of the board length, with the measurement condition stated. The limit for a bare board is not the same as the limit for the assembled product, and both may differ from the requirement of the equipment that will eventually hold the board.
The limit should be derived from what the product needs rather than adopted from a table. A board that is placed in a connector with a tight clearance needs a tighter flatness limit than one that is mounted on standoffs, and the assembly process may require a different limit again. The tolerances that apply to the outline are described in the guide to board outline tolerance rules.
Effect of Design and Materials
Copper distribution is the largest single design influence on flatness. A board with a large plane on one side and sparse copper on the other will behave differently through lamination, and the imbalance is normally corrected with thieving or by rearranging the layers. The practice is described in the guide to laminate material properties.
Thickness, layer count and material also matter. Thin boards warp more easily for the same imbalance, while a high layer count tends to be more stable because the stack is more symmetrical. Where flatness is critical, the design should be discussed with the fabricator before the first build rather than after the panels are measured.
Controlling Flatness in Production
Production control depends on measuring at defined steps rather than only at the end. Flatness after lamination, after pattern plating and after the finish all give information about where a deviation is introduced, and that information is what allows the cause to be corrected rather than merely reported.
The records should include the measurement method, the conditions and the number of boards sampled. A sample of one passing panel says very little, while a distribution across a panel and across a lot shows the capability of the process, and the position of the measurement points should be fixed so that successive readings are comparable. The same principle of defining the criteria before measuring is set out in the guide to judging PCB quality.
FAQ
What is an acceptable bow and twist for a PCB? It depends on the product and on how the board is supported in assembly, and the limit is usually quoted as a percentage of the diagonal with the measurement condition stated. A common starting point for a bare board is a fraction of a percent, tightened where the application demands it.
Can a warped board be flattened after assembly? Sometimes, by clamping or by heating under load, but the board will tend to return to its previous shape and the process stresses the joints. The better answer is to control flatness at fabrication and to design the assembly so that it tolerates the flatness the process can deliver.
Why do two measurements of the same board disagree? Almost always because the reference plane or the support arrangement differed. Free state measurement, three point support and constrained measurement all give different values, so the method and the conditions have to be recorded with the number.
