Intermittent Tombstoning on an 0402 Capacitor
A six-layer control board carries 240 identical 0402 100 nF decoupling capacitors. After a change of contract manufacturer, 3.2 % of the C41 position tombstone - one termination lifts clear of its land. C41 has one pad tied to the internal ground plane through four 0.3 mm thermal spokes, and the other pad on a short 0.2 mm signal track ending at another small component. Both lands are 0.55 × 0.60 mm with unchanged stencil apertures (0.12 mm foil). The new reflow profile ramps at 2.8 °C/s from 25 °C, soaks at 150–190 °C for 60 s, peaks at 243 °C and spends 55 s above 217 °C. The previous supplier used the same paste with a 1.2 °C/s ramp and a 110 s soak. Moisture-sensitive parts are baked and the paste is in date.
1. Explain the mechanism by which the thermal relief and the signal track produce tombstoning on this component.
Read the worked answer
Tombstoning is a torque about the component, generated when the two terminations are not simultaneously molten.
- The ground-plane pad is thermally tied to a large copper mass through four spokes. The spokes limit heat flow but do not stop it: that land lags the rest of the board in temperature.
- The signal-side land is connected only to a thin track and a small neighbour, so it has little thermal mass and reaches paste melting temperature first.
- While the signal side is liquid and the plane side is still solid, the molten fillet's surface tension pulls that termination down and wets up the component end. The body pivots about the still-solid end.
- A 2.8 °C/s ramp and a 60 s soak give the assembly less time to equalise, so the pad-to-pad time lag is larger than with the 1.2 °C/s / 110 s profile - hence the failure appeared only after the process change.
Note the direction of the logic: the problem is a thermal imbalance between the two lands during ramp and soak, not a paste chemistry or a component problem.
2. Two fixes are proposed: (i) revert to a 1.2 °C/s ramp with a 110 s soak; (ii) delete the thermal spokes and connect the pad solid to the plane. Evaluate both and state your recommendation.
Read the worked answer
(i) Slower ramp and longer soak - sound. The defect is driven by the difference in time at which the two pads cross the paste melting point. A gentler ramp and a longer soak let the plane-connected pad catch up with the signal pad before either goes liquid, so the imbalance shrinks. It costs throughput, not hardware, and it is reversible. This is also the change that is known to have worked: the previous supplier ran essentially this profile.
(ii) Solid connection to the plane - wrong direction. Removing the spokes increases the thermal coupling from the land to the plane. The land becomes a better heat sink, so it heats even more slowly relative to the signal pad: the lag increases and the tombstoning rate should rise, not fall. It also makes hand rework of C41 much harder.
Recommendation: adopt (i) first, and add a secondary balancing measure so the fix is not profile-only - for example reduce the spokes to two, or widen the signal-side track/pad slightly, so the two lands have more similar thermal mass. A component of this size is normally also helped by an aperture tweak, but that is a second-order lever here: the asymmetry is thermal, not volumetric. Verify with a reflow profile on the two pads before releasing the change.
3. What measurement would confirm the diagnosis before committing to the change?
Read the worked answer
Measure the thermal histories of the two pads, not of the board or the oven.
- Attach a fine-gauge thermocouple (or a two-channel profiler with fine bead probes) to each C41 land, bonded with a small spot of high-temperature adhesive, and run the oven profile with a representative populated board.
- Record the time at which each pad crosses the paste liquidus (217 °C) and the peak temperature difference between the pads.
- Diagnosis is confirmed if the plane-side pad crosses liquidus measurably later than the signal-side pad (of the order of a second or more) and if that lag shrinks when the slower ramp/longer soak is used.
Supporting checks: verify paste deposit volume is equal on both apertures (a stencil or print inspection), confirm there is no shadowing or airflow difference at that board position, and re-inspect the defect rate after the change over a statistically useful sample (a few thousand placements), since 3.2 % is well above the normal ppm level and a single panel proves nothing.