BGA Rework: Equipment, Profiles and Yield

A ball grid array hides its joints underneath the package, so it cannot be inspected or repaired in the way a leaded part can. When one of those joints fails, the board is not necessarily scrap: it can be taken back to the reflow state locally, the device lifted, the site prepared and the part replaced. It is a process with a narrow window, and the equipment exists to hold the assembly inside that window.

This article describes what BGA rework involves, the subsystems a rework station provides, the thermal profile that makes the process repeatable, and the practical limits that decide whether a board should be repaired or replaced.

What BGA Rework Involves

The operation is a localised reflow. The board is supported and preheated from below, the area around the device is brought to a temperature just under the melting point of the alloy, and the device itself is heated from above until its solder spheres liquefy. The part is then lifted with a vacuum nozzle while the alloy is still molten.

Replacing the device follows the reverse path, with two extra steps in the middle: the residual solder has to be removed from the pads and the site dressed back to a flat, clean surface, and the replacement device has to be aligned before the alloy freezes. Both are done while the board is still hot, because the sequence has to finish before the assembly cools below the point at which the alloy can still wet.

The Subsystems of a Rework Station

A station that can do this repeatably is built from four parts. An upper heater delivers the local reflow energy, usually by forced hot air. A lower heater preheats the whole board so that the local heater does not have to fight the thermal mass of the assembly. A thermocouple or two measures the actual temperatures being reached. An optical alignment system lets the operator see the sphere pattern and the pads as a single image.

The distinction between a station that is adequate and one that is not usually lies in the control loop rather than in the heater power. A machine whose measured temperature can be held within a couple of degrees at the joint is a machine that will not lift pads, and it is the measured temperature at the joint rather than the set point on the front panel that matters.

Why Optical Alignment Matters

The device has to be placed with its spheres sitting on the pads, and the tolerance is a fraction of the pitch. An operator cannot see the interface directly, so the station has to superimpose the image of the sphere pattern on the image of the pads and let the operator bring the two into register before the part descends.

Alignment and thermal behaviour are connected, because a device that is placed slightly off centre will be pulled into position by surface tension only if the alloy on every sphere melts at the same moment. Where the board is not uniformly preheated, the pull is uneven and a part that was aligned correctly will drift as it settles.

Thermal Profiles and the Local Reflow Window

A rework profile has the same stages as an oven profile: a ramp, a soak, a short excursion above the liquidus, and a controlled cool. What makes it different is that the excursion is local, so the board sees a steep gradient between the device and the area a few centimetres away, and that gradient is what stresses the laminate.

The alloy decides the window. A lead-free alloy melts higher and needs a longer time above liquidus than a tin-lead alloy, so the local excursion is hotter and lasts longer, which is one more reason to understand which of the two the assembly uses. The comparison of the two alloys sets the numbers the profile has to reach.

Preheating, Moisture and Board Protection

The lower heater does two jobs. It reduces the temperature difference across the board so that the local heater can reach the joint without cooking the surface, and it drives moisture out of the laminate before the temperature rises. A board that has absorbed water and is heated quickly will push that water out as steam at the worst moment, which is how a repair turns into a blister.

Support of the assembly matters for the same reason. The board has to be held flat and restrained while it expands, because the warpage that a local hot spot causes can be enough to change the effective coplanarity of the joint. Where the design includes a large thermal pad, the choice between a via in the pad and a plated through hole decides how much heat the rework has to supply.

Site Preparation and Reballing

Once the device has been lifted, the pad site is not ready. It holds residual solder in various shapes, sometimes with a sphere still attached to a pad that should be flat, and the amount left on each pad varies across the array. A flat, even surface is what the replacement device needs, and it is produced by wicking or vacuuming the old alloy away and then cleaning the site.

Devices that will be reused need new spheres. That is done with a stencil and a paste or with preformed spheres, and the important detail is that the sphere height has to be uniform, because a device with one low sphere will have an open joint that passes every check short of an X-ray.

Yield, Inspection and the Limits of Rework

A successful rework is one whose joint cannot be distinguished from the original, and the way to know that is inspection rather than confidence. Visual inspection of the periphery, X-ray of the array and, for critical products, a cross-section are the tools; the ones that cost least are used on every repair and the expensive one occasionally.

The process is worth having only where the board is worth more than the repair. On a low-cost board that can be replaced in an hour, rework is not economical; on a board carrying expensive parts, a design that has been laid out to make the repair possible is what gives the operation a chance. The quality characteristics of a well-made board include that kind of foresight, and the movement of components during reflow is the failure mode that a good profile avoids.

FAQ

Can any board be reworked? No. The part has to be accessible for the nozzle, the board has to fit the fixture, and the surrounding parts have to tolerate the local temperature rise. A design that placed the device flush against tall components may not leave room for the operation.

How many times can one site be reworked? Each cycle adds thermal stress to the pads and to the laminate. Two or three cycles are common in practice, and the pad adhesion, not the device, is usually what fails first.

Does rework affect product qualification? For regulated products it can, because the process is no longer the qualified one. The rework procedure has to be documented and the repair recorded if the product is to remain traceable.