ENGINEERING NOTE · BGA VOIDING & X-RAY
Class 2 allows ≤25% voiding in any single joint and ≤40% of total area; Class 3 tightens that to ≤10% and ≤25%. IPC-7095 measures against pad area, so settle two questions before arguing about an X-ray image: which class the board is built to, and whether the percentage is taken on pad area or ball diameter.
The process engineering team · 11 September 2026
THE ACCEPTANCE LIMITS
A voiding dispute is almost never about whether a void exists. It is about which limit applies and what the percentage is measured on. Both answers come from the acceptance class on the customer's drawing, so the class is settled before the first joint is imaged.
| Item | Class 2 | Class 3 |
|---|---|---|
| Single jointlargest void in one joint | ≤ 25% of pad area | ≤ 10% of pad area |
| Total voidingall voids in one joint | ≤ 40% | ≤ 25% |
| Basis of the percentage | Pad area, not ball diameter or ball cross-section | Pad area, calculated the same way |
| Joints in scope | BGA, CSP and QFN joints hidden under the package | BGA, CSP and QFN joints hidden under the package |
"Void-free" is not a specification. IPC-7095 sets an acceptance limit, and a limit you can meet repeatably is worth more than a target you cannot measure.
The basis matters more than most arguments admit. The same void, measured against pad area and against ball diameter, produces two different numbers — and those two numbers are then quoted at each other across the table for an hour. We calculate against pad area, state the acceptance class in the first article documentation, and image the joints at the same magnification for both parties.
WHAT DRIVES VOID FORMATION
A void is a gas pocket that did not escape before the joint solidified. That gives four places to work on it: how hot the joint gets, how long it is held, what the flux gives off, and what comes out of the board underneath.
Mechanism: peak temperature and time above liquidus decide how long volatiles have to leave the joint and how quickly it freezes afterwards. Cut the time above liquidus and gas that is still forming is locked in place as the alloy solidifies. Knob: peak setpoint and time above liquidus on a 10-zone forced-convection profile, checked against the paste supplier's window for SAC305.
Mechanism: the soak brings paste, ball and pad to a common temperature and gives the flux time to activate and to release its volatiles before the joint collapses. A soak that is too short pushes that release into the melting stage, exactly where the gas can no longer leave. Knob: soak length and ramp rate between preheat and reflow.
Mechanism: the flux vehicle boils as the deposit heats, and its volatiles peak at a temperature set by the chemistry. A paste whose volatiles peak late traps gas under a ball that is already collapsing. Knob: paste selection and the profile window that suits it — the two are chosen together, not separately.
Mechanism: gas also arrives from below. Entrained air in a printed deposit, and moisture or plating chemistry released from vias inside the pad, rise into the joint while the alloy is molten. Knob: via filling and plating quality, print quality at the pad, and a bake before reflow for boards that have absorbed moisture.
ATMOSPHERE: AIR VS NITROGEN
Oxidation competes with wetting, and oxygen concentration is the variable that decides how much of it happens during the profile. Air contains 20.9% oxygen; nitrogen reflow is available on our 10-zone convection ovens at 100–1,000 ppm O₂.
Industry-published figures, used here as illustration. They are not our measurements and they are not a promise about any particular board; they are the reason we offer nitrogen as a project option rather than as a default.
Treat the atmosphere as a lever with a cost. Nitrogen buys wetting margin on fine-pitch and high-thermal-mass assemblies, and it also changes how the profile behaves, so it is qualified with the profile rather than switched on top of it. On boards where the voiding is driven by flux volatiles or via outgassing, lower oxygen will not fix the cause on its own.
WHAT 2D X-RAY SHOWS — AND WHERE IT STOPS
A transmission image sums everything the beam passes through. That is an advantage on a single row of joints and a limit the moment two rows overlap.
WHERE THE IMAGE RUNS OUT
On a multi-layer BGA the upper and lower ball rows project onto the same pixels. A void in the upper row and a void in the lower row land in the same place, and no amount of contrast separates them.
X-Ray inspection is configured per project, not applied to every board. When a build needs per-joint void measurement on a stacked package, we say so at quotation and route it to computed tomography instead of quoting a 2D image as if it answered the question.
VOIDING IS NOT HEAD-IN-PILLOW
The two are confused because both appear on the same X-ray image and both are blamed on the reflow profile. They are not the same defect: one is a gas pocket inside a joint that formed, the other is a joint that never formed at all.
| Aspect | Voiding | Head-in-pillow |
|---|---|---|
| What it is | A gas pocket trapped inside a joint that melted, wetted and formed | Paste and ball that never fused; the ball rested on top of the printed deposit |
| On the image | A darker irregular pocket inside a bright ball, with a distinct edge | A round ball outline with no continuous joint beneath it; often confirmed on an angled view |
| Electrical result | Usually none, until the void removes enough load-bearing area | Open circuit, or contact that closes and opens again under thermal or mechanical load |
| Process response | Profile, soak, flux chemistry and pad or via outgassing, judged against the IPC-7095 limits | Warpage and coplanarity control, paste volume, a soak that takes the board and the ball through the melting range together |
| What goes in the specification | Acceptance class and the basis of the void percentage | Coplanarity and warpage limits plus the profile hold, not a void number |
A joint can pass IPC-7095 and still fail electrically: the void percentage says nothing about whether the ball ever fused to its pad. Head-in-pillow ends as an open circuit rather than as a percentage, which is why it is chased with coplanarity data and a revised soak instead of a tighter void limit.
RELATED CAPABILITIES
2D X-Ray for BGA, CSP and QFN hidden joints, AOI after reflow and ICT or flying probe on 2 dedicated test lines — scope set per project.
Seven high-speed SMT lines with 0.4 mm ball pitch BGA placement, 10-zone forced-convection reflow and optional nitrogen at 100–1,000 ppm O₂.
First-pass yield is measured at final inspection before any rework or retest: 98.5% or better on production lots of SMT and THT assemblies.
SEND US THE X-RAY QUESTION
Send the X-ray image, the acceptance class and the pad geometry. We will tell you what the image can and cannot support, and which of the four causes the void pattern points at.
Reply within one business day