Ball Shear and Pull Testing for Solder Joint Strength

A solder joint that looks perfect can still be weak. Voiding, partial wetting, brittle intermetallic layers and contaminated pad finishes all hide beneath a smooth fillet, and none of them show up reliably under a microscope. Mechanical testing exists to expose those hidden weaknesses before a customer does. This article explains how ball shear and pull testing are applied in production, what each method actually measures, and how to read the numbers so they drive real process decisions rather than comfort.

Why Joint Strength Needs Two Different Tests

Shear and pull load a joint along different axes. A shear test pushes the ball sideways and loads the interface in a way that is dominated by the bond between solder and pad. A pull test lifts the ball or the component away from the board, which loads the joint, the pad and the surrounding laminate in tension. A joint that survives one can fail the other.

That is why serious reliability programmes run both. Measuring only shear gives a confident picture of the metallurgical bond while saying almost nothing about how the assembly behaves when a board flexes in service. Measuring only pull says little about the pad interface. The pair together brackets the behaviour of the joint far better than either test alone, and the extra cost is modest.

How the Ball Shear Test Works

A shear tool with a flat face is brought alongside the ball and driven horizontally at a controlled height above the substrate. The tool must clear the pad without touching it, so the setup height becomes one of the most important variables in the whole measurement. The load rises until the ball separates, and the peak force is recorded in grams-force or newtons.

Tool alignment matters more than most operators expect. If the shear face is not parallel to the substrate, the ball is loaded unevenly and the recorded peak force drops even though the joint itself has not changed. A quick way to detect this is to check whether the failure mode stays consistent across a coupon: a sudden change in mode usually indicates a fixture problem, not a material problem.

How the Pull Test Works

In a ball pull test, a hook or clamp engages the ball or a stud attached to it, and the load is applied perpendicular to the board. For area array packages, the same principle is scaled up with a fixture that grips the package body or a bonded stud while the board is restrained. The measurement is again a peak force, often accompanied by the extension at failure.

Clamping is the weak point in practice. A jaw that slips produces a low reading that has nothing to do with the joint, and a jaw that bites too hard introduces a notch and a premature failure. Documenting the clamp type and torque alongside every dataset is the simplest way to keep results comparable between operators. Useful background on tolerance and reliability explains why that consistency matters downstream.

Fixtures, Clamps and Alignment

Every mechanical test is really a measurement system, and the measurement system must be verified before the samples are. That means running a known reference coupon, checking that the recorded force is repeatable within a few percent, and confirming that the load cell has been calibrated recently. Skipping that step makes every subsequent number suspect.

Board support is the next consideration. A board that is allowed to flex during a pull test absorbs part of the load and lowers the measured force. Rigid backing plates and consistent edge support remove that variable. Where the board must be tested as it sits in a fixture, the fixture itself should be described in the report so that another laboratory can reproduce the setup.

Test Speed and Its Effect on Results

Solder is strain-rate sensitive. Pulling a joint quickly produces a higher peak force than pulling it slowly, because the alloy behaves more elastically at short time scales. A test run at ten millimetres per second and a test run at one millimetre per second can differ by a meaningful margin on identical joints, which is why the speed is normally fixed by an internal standard.

The same logic applies to the failure mode. Fast tests encourage brittle fracture through intermetallic layers, while slow tests allow more ductile deformation of the bulk solder. Neither result is wrong. They simply answer different questions, so choose the speed that reflects the loading the product will actually see and state it clearly with the data.

Intermetallic Growth and Strength Loss

At every solder interface, copper and tin react to form intermetallic compounds. A thin, continuous layer is desirable because it proves wetting occurred. A thick, irregular layer is a problem because the compound is brittle and the interface between compound and solder becomes the preferred path for a crack. Ageing and repeated thermal cycling both thicken that layer.

Because of this, strength measurements taken on fresh assemblies and on aged assemblies tell different stories. A common and useful protocol is to measure as-built parts, then measure parts after a defined bake or thermal cycle, and compare both the peak force and the failure mode. A drop in force accompanied by a shift from ductile to brittle fracture is one of the clearest signals that a joint will not survive field life.

Reading Failure Modes in Fracture Surfaces

The force number alone is incomplete. Inspectors classify each failure as bulk solder, pad lift, intermetallic fracture, or a mixed mode, and record the percentage of the fracture area belonging to each. Those classifications convert a scatter of numbers into a pattern that points at a specific process step when something goes wrong.

Bulk solder failures usually indicate that the joint geometry or the reflow profile is the limiting factor. Pad lift failures implicate the laminate or the pad adhesion, which is a fabrication issue rather than an assembly one. Intermetallic failures point to excessive thermal exposure. Comparing mode distributions between two builds often identifies the cause faster than any statistical treatment of the forces. Reviewing solder defect patterns side by side with the fracture log sharpens that judgement.

Weibull Analysis and Sample Size

Joint strength is variable, and a single measurement says very little. Thirty samples per cell is a common minimum for a meaningful distribution, and far more are needed if the goal is to estimate a low-percentile strength with confidence. Testing five balls and reporting the average is a common but essentially meaningless practice.

Weibull analysis is often applied to the resulting dataset because it handles the weakest-link behaviour of brittle interfaces well. Plotting force at failure against cumulative probability produces a shape parameter that describes scatter and a scale parameter that describes characteristic strength. Watching the shape parameter is particularly useful: a drop in scatter usually means the process has become more consistent.

Turning Measurements into Process Limits

A specification limit without a distribution behind it is a guess. Once a qualified baseline exists, lower control limits can be derived from the observed distribution, with a margin that reflects how much the process normally drifts. New lots are then compared against that baseline rather than against an arbitrary number inherited from a supplier datasheet.

Finally, keep the measurement itself under control. Periodic re-testing of a retained reference coupon catches drift in the tester, the fixture and the operator at the same time. When a failure investigation eventually occurs, that retained evidence is often the only way to prove whether the joints changed or the instrument did. Companion checks such as solderability testing round out the picture of incoming material quality.

FAQ

Which test is better, ball shear or pull testing? Neither replaces the other. Ball shear is highly repeatable and directly sensitive to the pad and intermetallic interface, which makes it the better choice for monitoring a stable process. Pull testing adds information about the board, the package and the laminate under tensile load. Most reliability programmes run a shear series for process control and a pull series to qualify a design.

What is a typical acceptable joint strength? There is no universal number, because the result depends on ball diameter, alloy, pad finish, test speed and fixture. The right approach is to establish a baseline on a known-good build, verify that the distribution is tight, and then set limits relative to that baseline. Absolute figures taken from another product or another laboratory are not transferable.

How many samples should be tested per lot? For routine monitoring, twenty to thirty joints per condition gives a usable distribution at reasonable cost. For qualification or for estimating low-percentile strength, sixty or more are usually required, and the samples should be spread across the board rather than clustered in one corner, since corner joints often see different thermal and mechanical conditions.