Warpage Measurement: Design Rules and Process Limits
Warpage is measured as a deviation from a reference plane, and the reference plane is the part of the measurement that is most often left undefined. Two inspectors measuring the same board with different supports will report different numbers, and both will be right. For a figure to mean anything, the support condition, the reference and the measurement points all have to be specified, and they have to be the same at the supplier and at the customer.
What Bow and Twist Actually Describe
Bow is a curvature across the board in one direction, roughly cylindrical, and it is expressed as the maximum deviation from a plane divided by the board length. Twist is a curvature about a diagonal, where opposite corners rise and fall together, and it is expressed in the same way. A board can be bowed without being twisted, twisted without being bowed, and on a large panel it can be both at once.
The distinction matters because the two have different causes and different consequences. Bow usually comes from an unbalanced stack-up or from an asymmetric copper distribution, and it affects how a board sits on a conveyor. Twist usually comes from the laminate’s residual stress and from the panel’s aspect ratio, and it affects how a large board fits a chassis. A warpage measurement report should separate them rather than quoting a single flatness figure.
Reference Plane and Support Conditions
A measurement on three support points gives the deviation from the plane defined by those points, and the result depends on where they are. Measuring on the board’s own outline, on a surface plate with the board free, and on a fixture that clamps the board flat all give different numbers. The convention for unassembled boards is usually a measurement on a surface plate or on three defined points, with the board resting under its own weight.
The reference has to be stated in the specification. A requirement that says the board shall be flat within a limit, without saying how it is supported, cannot be verified and will be argued about at every delivery. The usual form is to state the support points, the measurement points and the calculation, so that any two laboratories can reproduce the figure.
<img src="https://www.gopcba.com/wp-content/uploads/2026/05/X-ray-inspcetion.jpg" alt="Shadow moire fringe pattern across a large panel showing twist” />
Measuring With a Height Gauge
The simplest method uses a surface plate and a height gauge or a dial indicator. The board is placed on the plate, the deviation is measured at a grid of points, and the maximum is taken. The method is accurate for bow and adequate for twist, and it requires the board to rest without any external force.
The limitations are the board’s own weight, which flattens a heavy board, and the difficulty of measuring the underside. Where a board is heavy, the measurement should be made on the support points rather than on a full plate, so that the board takes its natural shape. The grid should be dense enough to find the maximum: a measurement at the corners only will miss a bow that peaks in the middle.
Optical and Shadow Moire Methods
Shadow moire projects a grating onto the board and reads the fringe pattern to produce a height map over the whole surface. It is non-contact, fast, and it gives a complete picture rather than a set of points, which makes it the practical method for a panel with a complex shape. The accuracy depends on the grating pitch and on the calibration, and the board has to be supported in the same way as for a mechanical measurement.
The method is well suited to measuring the change through a process, because the same board can be measured before and after reflow and the difference taken. Where a shop outsources the measurement, the report should state the support condition and the grating used, since a shadow moire result on a clamped board describes the clamp rather than the board.
Measurement Before and After Assembly
A board that is flat when it arrives can be warped when it comes out of the oven, and the change is what matters for assembly. The measurement should therefore be made at both points on the same sample, with the same support, so that the difference is meaningful. Measuring only the finished assembly attributes the whole deviation to the assembly process.
The assembly contribution comes from the thermal excursion, from the copper distribution on the two sides, and from the mass of the components. A board with a large component on one side and none on the other will bow towards the populated side after cooling, because the two sides contract differently. Where the deviation appears consistently after reflow, the stack-up and the copper balance should be reviewed rather than the oven.
Effect of Temperature and Moisture
Laminate expands with temperature and with absorbed moisture, and a board measured warm will differ from the same board measured cold. The measurement should be made after the board has returned to room temperature and after any absorbed moisture has had time to equilibrate, which for a board that has been washed can take hours. A measurement taken immediately after a wash will be affected by the moisture gradient through the thickness.
The temperature during the measurement should also be recorded, because a shop that measures in a warm room in summer and a cool room in winter will see a seasonal variation that has nothing to do with the process. Standard conditions should be stated in the specification, and a board that is measured outside them should be noted as such. The warpage causes and controls notes describe the design and process contributions in more detail.
Acceptance Limits and How to Write Them
The limit for an unassembled board is usually expressed as a percentage of the board’s longest dimension, with a separate figure for bow and for twist, and it applies to the panel or to the individual board depending on where the separation happens. The limit for an assembled board is set by the mechanical interface rather than by a standard, and it should come from the clearance available in the chassis.
The limit should state the temperature, the support and the measurement method. A limit of 0.75 % on a board of 200 mm is 1.5 mm, and whether that is measured on a plate or on three points changes the number by a significant fraction on a board with a natural curvature. Where the board is measured after the depaneling step, the limit should apply to the separated board, because the panel’s constraint hides some of the deviation.
Records and Reaction to Drift
The record should carry the identifier, the stage at which the measurement was made, the support condition, the temperature and the measured values. With those entries, a drift can be attributed to a process change rather than to a measurement change, which is the most common confusion when two departments report different results.
The reaction to a drift should be defined, with a limit for investigation and a limit for rejection. Where the measurement is used to accept a lot, the gauge should have a known repeatability so that the decision is not made on noise. The quality criteria for the product should include warpage alongside the other dimensional requirements, because it is a dimensional property and it should be treated as one rather than as a subjective impression.
FAQ
Can warpage be measured with a straight edge? It can be estimated, and for a quick check on a large bow it is adequate. The reading depends on where the straight edge is placed and on the pressure, so it should not be used for acceptance against a numeric limit.
Does a bowed board always cause an assembly problem? Only where the deviation changes the process: a board that does not sit flat on the conveyor, or that moves during placement, causes placement errors and print defects. A board that is bowed but held flat by the fixture may assemble perfectly, which is why the support condition in the measurement should match the condition in production.
Who should set the limit? The mechanical designer, from the clearance in the enclosure, in consultation with the fabricator who has to meet it. A limit set by the fabricator alone tends to be one that is easy to meet rather than one that the product needs.
