Voids Under a QFN Thermal Pad: Causes and Limits

A quad flat no-lead package conducts most of its heat into the board through the exposed pad on its underside, and that pad is soldered to a matching copper area on the board. The joint is large, thin and covered by the package, which makes it the ideal place for flux gas to be trapped. The result is a field of voids under the thermal pad that no cleaning step can reach and no rework can easily correct.

Voiding under a thermal pad is normal to a degree, and the engineering question is how much is acceptable for the thermal and mechanical duty of the product. Answering it requires an understanding of where the gas comes from, how the paste pattern and the stencil shape it, and how the measurement is made. Treating the void percentage as a single pass or fail number hides all three.

Why the Thermal Pad Voids

The void is the space left by gas that could not escape before the alloy solidified. Sources include the solvent in the paste, the volatiles released by the flux as it activates, air carried into the deposit during printing, and moisture on the pad or in the laminate. Under a large pad the gas has to travel a long distance to reach an edge, and the paste at the perimeter solidifies first and seals the route.

The geometry makes the problem worse than it looks. A thermal pad may be several millimetres on a side, so the distance from the centre to the edge is more than a millimetre, and the joint height is only 50 to 80 um. The gas has to move through a channel that is far longer than it is tall, and the alloy is solidifying at the same time. Reducing the path length is the most effective single action available.

Paste Volume and Deposit Pattern

The deposit on a thermal pad is a compromise between volume and escape routes. A single large deposit gives the highest volume and the longest path, and it produces the largest voids. A divided pattern gives up a small amount of volume for a set of shorter paths, and it is the standard answer for pads above about three millimetres on a side.

The total volume still has to fill the joint and leave a small fillet. Too little paste produces an incomplete joint with unfilled regions that look like voids on an X-ray image and are mechanically worse, because the solder is not in contact with the pad. The solder paste volume is calculated from the pad area and the target bond line thickness, and the stencil thickness is chosen to suit it.

Stencil Design for the Pad

A window pane pattern divides the pad into a grid of openings separated by thin webs of stencil. The webs leave channels in the printed deposit along which gas can escape during reflow, and they disappear as the alloy melts and flows together. Typical grids use webs of 0.2 to 0.4 mm, with openings sized so that the total area is 50 to 70 percent of the pad area and the printed volume matches the target.

The webs have to be strong enough to survive printing and cleaning over thousands of cycles. Where the pattern is finer than the stencil thickness can support, the webs deform and the paste bridges across them, which defeats the purpose. The aperture side of the design is covered in the notes on paste volume measurement.

Reflow Profile and Outgassing

The profile decides how much gas is produced and how much time it has to escape. A soak that raises the board slowly through the flux activation range gives the volatiles time to leave before the alloy becomes liquid, and it is the single most effective profile change. A fast ramp into the peak traps the gas under a skin of molten solder and produces a large central void.

The time above liquidus has to be long enough for the gas that remains to travel to the edge, and short enough to limit intermetallic growth. A window of 45 to 75 seconds above 217 °C with a peak between 235 and 245 °C suits most tin-silver-copper pastes. Vacuum assisted reflow, which lowers the pressure around the joint while the alloy is liquid, is used where the void requirement cannot be met otherwise.

Via-in-Pad and Wick-Out

Thermal vias under the pad are often used to move heat into the inner layers, and they introduce a second path for the paste. If the vias are open, the paste wicks down the barrels during reflow and the joint under the pad loses volume, leaving voids at the top. The vias have to be plated over or filled and capped before printing, and the flatness of the fill has to be good enough that the stencil seals.

Where the vias are filled, the paste stays on the pad and the void behaviour is controlled by the pattern and the profile. Where they are open, no stencil pattern can compensate for the loss. The options and their tolerance implications are compared in the notes on via in pad processing.

X-Ray Measurement and What It Reports

Voids are measured on a two dimensional X-ray image as a fraction of the pad area, and that number is a projection rather than a volume. A void that is tall and narrow and one that is flat and wide can project the same area while behaving differently thermally. The measurement settings, the threshold used to define a void and the magnification all change the number, so a specification without a measurement method is not reproducible.

The conventional method sums the void areas and divides by the pad area, with the largest single void reported separately. Slice computed tomography gives a true three dimensional picture and is used for failure analysis rather than for production. The conventions behind the number are set out in the notes on void measurement and in the background on solder void causes.

Thermal Performance and Void Fraction

The effect of voids on thermal resistance is smaller than their area fraction suggests, because the remaining solder still forms a continuous path and the voids are filled with gas that conducts poorly but does not break the metal. Measurements on real packages typically show a modest rise in junction to board resistance at a void fraction of 20 to 30 percent, and a steeper rise only when the voids coalesce into a layer that spans the pad.

What matters more is the position and the shape. A void that covers a corner of the pad removes the mechanical support at that corner, and a void that spans the pad in a continuous band isolates a region of the joint. That is why specifications usually limit both the total area and the size of the largest single void, and sometimes prohibit voids at the corners.

Acceptance Criteria

A common commercial criterion limits the total void area to 25 percent of the pad with no single void above 10 percent, and stricter criteria of 15 percent total are applied to power devices. The criterion should be derived from a thermal and a mechanical test on the actual package rather than copied, because the sensitivity depends on the pad size, the power density and the temperature cycling the product will see.

The criterion also has to state the measurement conditions. A 25 percent limit measured on a two dimensional projection at one magnification is not the same specification as a 25 percent limit measured with a different threshold, and a supplier and a customer using different methods will disagree about the same board. Writing the method into the specification is what makes the limit enforceable.

Process Control and Records

The controls are the stencil pattern and its condition, the printed volume, the profile with its soak and time above liquidus, and the state of the vias under the pad. Each is recorded per lot, and the X-ray result of a sample is kept with the record so that a change in voiding can be traced to a change in one of them.

Where voiding rises, the order of investigation is the printed deposit first, then the profile, then the stencil condition, and only then the paste. A stencil that has accumulated paste in its webs and a stencil that is due for replacement produce the same defect, and both are visible on the print inspection record. With those records the void problem usually resolves to a maintenance action rather than a material change.

FAQ

Can voids under a thermal pad be eliminated? Not completely in a normal convection oven. The aim is to keep them below a limit that has been shown to be acceptable for the product.

Does a window pane stencil reduce the joint volume? It reduces the printed volume slightly, and the openings are sized so that the melted alloy still fills the joint with a small fillet.

Is vacuum reflow always better? It reduces voiding substantially and adds cycle time and equipment cost. It is used where the thermal requirement cannot be met by pattern and profile alone.