Shear Test and Pull Test Methods for Solder Joints

Shear and pull tests are the two mechanical measurements that turn a visual opinion about a solder joint into a number. They are used to qualify a paste, a surface finish, a reflow profile and a component, and they are used in failure analysis to decide where a joint actually broke. Both are easy to perform badly, and a badly performed test produces a number that looks authoritative and means nothing. Choosing the right test method for the question being asked is the first decision, and it is often the one that gets skipped.

What the Two Tests Measure

A shear test pushes the component sideways while the board is held, loading the joints in the plane of the board. A pull test lifts the component away from the board, loading the joints in tension. The two load the joint differently and therefore detect different weaknesses, which is why a qualification programme usually includes both. A joint that is strong in shear can still be weak in tension, and a single test method will not reveal it.

Neither test measures the strength of a single joint in isolation. The measured force is the sum of the joints that carry the load, and the failure occurs at the weakest point in the assembly, which may be the joint, the pad, the component or the board.

Shear Testing

Shear testing is the standard method for a chip component, a package with a small footprint or a ball array. A tool is brought against the side of the component at a defined height above the board and driven at a defined speed until the component separates.

The tool height is critical. A tool that contacts the component too high applies a bending moment as well as a shear load, and the component may fail before the joints do. A tool that contacts the board itself adds friction to the reading. The height should be stated in the procedure and held constant across the series, because it is the single parameter that most often changes the result between two operators.

Pull Testing

Pull testing applies a tensile load to the component or to an individual joint. For a leaded package the load can be applied to a lead, and for an area array package it is applied to the die or to the package body through an adhesive fixture.

A ball pull test applies the load to a single ball after the package has been removed, which localises the measurement to one joint. It is slower than the shear test but it gives a per joint figure rather than an assembly figure, which makes it more informative for a bonding or a pad cratering investigation. The trade is a slower test and a more delicate preparation.

Specimen Preparation

The specimen has to represent the process. A board soldered with a different paste, a different reflow profile or a different surface finish is a different specimen, and mixing them in one series destroys the comparison.

The board also has to be supported rigidly during the test. A board that flexes under the load absorbs part of the energy and produces a lower reading, so the fixture should clamp the board close to the component being tested. The support arrangement is part of the test method rather than a detail of the bench. Our land pattern notes describe how the pad geometry affects the load the joint sees.

Test Speed and Its Effect

Solder is strain rate sensitive, which means it behaves differently when loaded quickly than when loaded slowly. A fast test produces a higher force and a more brittle looking failure, while a slow test allows the alloy to deform and produces a lower force with more ductile failure.

The speed therefore has to be specified and held. Comparing a result taken at one speed with a specification written for another is a common error, and it produces arguments that cannot be resolved from the data. The speed should therefore be written into the procedure with the tolerance the machine can actually hold.

Failure Mode Classification

The number is only half of the result. The failure surfaces have to be examined and classified: bulk solder, solder to pad interface, pad to laminate interface, component metallisation, or a mixed failure.

The classification is what tells you where to look. A failure in the bulk solder points at the alloy or the profile, a failure at the pad interface points at the surface finish, and a failure in the laminate below the pad is pad cratering, which is a board or a handling problem rather than a soldering one. The proportion of each mode in the sample is often more useful than the force figure itself.

Interpreting the Numbers

A single force figure has little meaning without the distribution it came from. Solder joint strength varies between nominally identical joints because of small differences in the fillet, the voiding and the cooling rate, so the result should be reported as a distribution with a mean, a spread and a minimum.

The minimum is usually the figure that matters for reliability, because a product fails at its weakest joint rather than at its average one. A high mean with a wide spread is a worse result than a slightly lower mean with a tight spread, because the spread is what the weakest joint on the product will experience.

Correlation With Assembly Defects

The tests become powerful when they are correlated with what inspection sees. A lot with a higher void rate under X-ray should show a lower shear force, and a lot with a marginal reflow profile should show a different failure mode. If the mechanical result does not follow the inspection result, one of the two measurements is not measuring what it claims.

Our BGA inspection notes describe how the internal condition of a joint is assessed, and our solder defect guide groups the failures by their point of origin.

Specifying the Test in a Qualification

A specification should state the test type, the tool geometry, the contact height, the test speed, the sample size, the acceptance figure and the failure mode criteria. Without those details the test cannot be reproduced, and a qualification that cannot be reproduced is a snapshot rather than a control.

At gopcb the mechanical results are recorded with the process conditions that produced them, and the acceptance of the finished joint is judged in our quality documentation, so that a change in the process can be linked to a change in the measured strength.

FAQ

Should shear or pull be used? Both, if the budget allows. Shear is faster and suits small chip components, while pull gives a per joint figure and is better for area array packages and for interface investigations.

What sample size is needed? Enough to characterise the spread, which for a variable process is usually twenty or more joints. A sample of three tells you almost nothing about the minimum.

Can the test be run on production boards? It is destructive, so it is run on coupons or on samples from a lot rather than on the shipped product. The process control version is usually a coupon that is built and tested on a defined frequency.