Dimensional Verification of Assembled Boards

An assembled board has dimensions, and they matter as soon as it has to fit into an enclosure, sit on a mating connector, or pass a camera module through a hole. Dimensional verification on an assembly is different from dimensional inspection of a bare board, because a connector’s height, a shield can’s position and the flatness of a card edge are created by the assembly process as much as by the fabrication. Measuring them requires a datum scheme, a method, and an agreed definition of what is being measured.

What Is Measured on an Assembly

The features that matter are usually the ones that touch something else: the height of a connector above the board surface, the coplanarity of a row of pins, the position of a module relative to the board outline, the flatness of a card edge, and the clearance between a tall component and the inside of a lid. Each of these is a functional dimension, and each has a tolerance that comes from the mechanical design rather than from the electronic one.

The board itself is part of the measurement. Board flatness, thickness and outline position all contribute, and where the fabrication tolerance uses most of the available budget, the assembly has nothing left. Reviewing the tolerance chain across the fabrication and the assembly before release is far cheaper than measuring parts after they are built, and it is the reason the fabrication notes should state the tolerances that the assembly relies on.

Datums and How They Are Established

A measurement without a datum is a number without a meaning. On an assembly, the practical datum is usually established from the board outline or from a set of tooling holes, because those are the features that the enclosure and the fixtures will use. Where tooling holes exist, they should be specified as datums on the drawing and used consistently by every measurement, including the ones made at incoming inspection.

Where there are no tooling holes, the measurement should be referenced to a defined corner and edge, and the definition has to be unambiguous. A corner that has a radius or a routed profile makes a poor datum because the tangent point is not repeatable. In that case a pad or a fiducial is a better choice, provided it is present on every unit and its position is controlled on the fabrication drawing. The datum scheme should appear on the assembly drawing, not in a measurement procedure that the supplier never sees.

Coplanarity and Connector Seating

Coplanarity is the property that decides whether a connector’s pins all touch their pads at once. It is measured as the distance between the lowest and the highest point of the pin row while the part rests on a reference plane, and it is usually specified as a maximum over the whole row. The measurement is sensitive to how the part is supported: a connector measured while resting on its own pins will read differently from one measured while pressed against a plane.

On the assembly, coplanarity of a soldered connector is a function of the pad height, the solder volume and the seating force during placement. Excess solder on one end lifts that end, and the resulting tilt shows up as a height difference across the row. Where the assembly has a tight coplanarity requirement, the connector should be measured after reflow on a sample, and the measurement should include the board flatness under the connector, because a board that is curved locally will tilt the part. The quality criteria for the assembly should state which of these contributions is being measured.

Coordinate Measurement Methods

A coordinate measuring machine uses a probe to touch defined points and computes positions from them, and it is the reference method for tight tolerances. Its limitation on an assembly is that the probe has to reach the feature, and standing components make many features inaccessible except from above. Touch probing also applies a force, which can move a compliant part or a light connector housing.

Optical methods avoid that problem. A vision system with a telecentric lens measures positions in the plane of the board very accurately, and a laser line scanner measures height across a profile without contact. For coplanarity of a long row, a laser profile in one pass gives a complete picture where a touch probe gives a series of points. The choice should follow the feature: in-plane positions are best measured optically, heights and profiles by laser, and complex three-dimensional geometry by a combination, with the touch probe reserved for features that can take the force.

Tolerance Stack-Up Across an Assembly

A stack-up combines the board thickness tolerance, the pad or land position tolerance, the component’s own dimensional tolerance and the placement accuracy. Even where each is individually reasonable, the sum can exceed the clearance available. The stack-up should be done for the critical dimension rather than in general, and in the worst case rather than in the statistically averaged case, unless the production volume justifies a statistical treatment.

The stack-up also identifies which contributor to control. Where the placement accuracy dominates, an improvement in the placement machine gives the most benefit; where the component tolerance dominates, the answer is a different part or a relaxed requirement. Doing that analysis at the design stage is what prevents a specification that cannot be met by any process, and it is the same reasoning that applies to the assembly documentation. The assembly documentation should record which dimensions are critical and why.

Fixtures, References and Repeatability

A measurement fixture has to hold the assembly the same way every time, and the way it holds it should reflect how the part is used. A connector measured while the board is clamped flat will read differently from one measured while the board is free, and both may be different from the condition in the enclosure. The fixture should therefore reproduce the mounting condition, or the difference should be understood and stated.

Repeatability is checked by measuring the same unit several times, removing and replacing it between measurements. A spread larger than a fraction of the tolerance indicates that the fixture or the datum is not adequate, and no amount of averaging will fix it. Where a measurement system is used for acceptance, its repeatability and reproducibility should be established with a formal study rather than assumed from a single careful measurement. The inspection standard for the product should state the method and the fixture, so that a measurement made at the customer can be compared with one made in production.

Sampling and First Article

The first article is the full dimensional verification: every critical dimension measured and recorded, with the fixture and the method identified. Sampling afterwards should be driven by the process capability for each dimension. A dimension with a wide margin can be checked rarely; one that runs close to its limit should be checked per batch, and one that is at the limit should be checked per unit until the process is improved.

The sample size should follow from the variation, not from a fixed rule. Where the process is stable, a small sample detects a shift only if the shift is large, so the sampling plan should state the shift it is intended to detect. Where that is not known, the honest approach is to measure enough units to estimate the standard deviation and then set the plan from it, which is a one-off exercise that pays for itself in the first batch.

Records and Reaction to Drift

Dimensional data should be recorded per unit or per batch with the identifier, the date and the operator, in the same way as any other process record. The value of the record is in the trend: a dimension that has moved steadily over three batches is telling the process owner something that a single out-of-tolerance reading cannot.

The reaction to a drift should be defined in advance, with a threshold for investigation and a threshold for stopping. Where the dimension is critical to fit, the reaction should be quick, because a batch that is out of tolerance may have to be sorted or reworked. Where the dimension is not critical, the investigation can wait for the next review, but the trend should still be monitored. Writing both thresholds into the control plan removes the debate at the moment when the decision has to be made quickly.

FAQ

Can a height be measured with a calliper on an assembly? It can be used as a check, but the reading depends on where the jaws rest and on how much pressure is applied, and the repeatability is usually much worse than the dimension being controlled. For anything but a coarse check, a profile measurement or a height gauge with a defined reference is required.

How is board flatness included in a connector height measurement? Either by measuring the connector height relative to the board surface directly beneath it, or by measuring the board flatness separately and combining the two. Measuring the height relative to a global reference plane without knowing the local board shape produces a number that cannot be interpreted.

Does a dimensional check replace functional testing? No. A connector that is dimensionally correct can still have a poor electrical contact, and a part that is dimensionally marginal can still work. The dimensional check controls the process; the functional test verifies the product. Both are needed where the dimension affects the function.