Contact Us

Tombstoning Prevention

Tombstoning Prevention

A 0402 capacitor that stands on one end after reflow is one of the cheapest defects to correct and one of the most expensive to ignore. The part is found at AOI, reworked by hand, and the line moves on until the same position fails again on the next lot. The cause has usually been blamed on the paste, the oven or the operator.

Tombstoning prevention is a process question with a small number of physical causes. Once the mechanism is clear, each cause has one or two countermeasures that can be checked on the line inside a shift, instead of a list of settings to try one after another.

What actually lifts the part

During reflow both terminations sit in molten solder. Each joint pulls the chip down with a wetting force, and the two forces are meant to be equal and opposite. A tombstone happens when one joint melts and wets first while the other is still solid, or still covered in paste that has not liquefied. The molten side rotates the chip about its own edge, and the body stands up before the second joint can pull it back.

That is why the defect concentrates on the smallest bodies. An 0402 or 0201 has almost no mass to resist the torque, and the same imbalance on an 0805 usually shows as a slight tilt that nobody reports.

The pad and aperture geometry that decide it

Three geometry items set the balance before the board reaches the printer. The first is pad symmetry. If one pad connects to a ground plane and the other runs to a thin trace, the plane works as a heat sink, the joint on that side melts later, and the chip stands toward the cooler side. Thermal relief spokes, or a balanced copper area under both pads, remove most of it.

The second is mask. A pad partly covered by solder mask has less exposed metal, so the paste wets a smaller area and the pull on that side is weaker. The stencil aperture should match the exposed copper rather than the nominal pad on the drawing.

The third is paste volume. Unequal deposits produce unequal forces even when the timing is right. Paste that has slumped, an aperture that has clogged after a few hundred prints, or a squeegee leaving more paste on one side of the panel is enough to start it. Volume is measured by 3D SPI on the first boards and at intervals, not judged by eye.

Reflow profile effects

Heating rate and soak decide how much of the flux has done its work before the alloy melts. A fast ramp drives solvent out of the paste in a burst, and the escaping gas lifts a small body. A soak long enough to bring the whole assembly close to thermal equilibrium leaves both joints melting within a second of each other, which is exactly what the balance needs.

Tombstoning prevention, cause by cause

What the panel showsMost likely causeFirst thing to change
Failures sit on positions tied to one copper plane, other positions passThermal imbalance between the plane pad and the trace padThermal relief on the plane pad, or balanced copper under both
Failures appear and disappear with the paste lotPaste viscosity or slump after storage and thawCheck thaw and open time, verify deposit volume with SPI
Failures start mid shift on one printerAperture clogging or poor gasketingUnder wipe and aperture check, then measure the deposit
Only the smallest bodies fail, larger ones tilt slightlyPlacement offset plus low body massPlacement accuracy check, aperture matched to the body
Failures cluster on one side of the panelUneven heating or board warp in the ovenSupport pins, panel orientation, profile measured on both edges
Failures began when the paste changed to lead freeHigher melting point and a narrower windowRe-profile with a longer soak, then add nitrogen if needed
Failures sit on pads with vias underneathPaste wicking into the via during printVia fill and cap, or aperture adjustment on that pad only

Placement and board support

Placement offset of a quarter of a pad width is enough to start the rotation on an 0201, because the body pivots about the edge that has paste under it first. A nozzle that is too large for the body can drag it on release. Feeder pitch and tape pocket condition are worth checking when one position fails and its neighbours do not.

Board support is the item most often missed. An unsupported panel sags in the heat and the parts near the middle see a different curve from the parts at the edge.

Catching it before the customer does

3D SPI catches volume asymmetry at the printer, which is the only place the cause can still be corrected without touching the board. AOI after reflow catches the lifted body, but AOI is confirmation rather than control: by the time it sees the defect the panel has already been built. X-ray adds little here, because the joint that failed is open and the open is visible from above.

How SUPERB PCBA controls it

Assembly runs on seven SMT lines with 01005 placement, 3D SPI after print, 3D AOI after reflow, and a 12-zone reflow oven running SAC305 under nitrogen held below 500 ppm oxygen. Class 2 and Class 3 work is held to IPC-A-610 under ISO 9001, and medical builds under ISO 13485.

First articles are profiled with thermocouples and the curve is kept with the build record, so a repeat order starts from a measured profile rather than from an oven setting. Paste deposit is measured at the start of the run and at intervals, and the numbers sit with the SPI record. Where an aperture needs tuning, the change is made on the stencil and noted against the panel, with the published area ratio guidance as the reference. The reflow oven is profiled after any maintenance that touches a zone.

When a position keeps failing, the build is held and the position is measured instead of reworked.

Questions buyers ask

Can tombstoning be eliminated completely?

No. It can be held to a level where AOI finds a few units per thousand instead of per hundred. What remains is process variation, and it is measured rather than assumed.

Does lead free paste make it worse?

The window is narrower because the alloy melts higher and the flux is used up sooner. The same countermeasures work, but the soak band matters more and the ramp is less forgiving.

Next step

Send the Gerber set, the BOM, and the position that fails, with a photograph of the defect and the panel location. The reply names the probable cause, the countermeasure, and whether it sits in the stencil, the profile or the copper. If the build is already running, the lot can be held and the deposit at that position measured so the numbers reach you before the next panel starts.