An X-ray report arrives with a table of void percentages and a line that reads "maximum void 32 percent". The buyer has to decide whether the lot ships or not, and the number alone does not answer that. A 32 percent void in the centre of a ball and a 12 percent void on the pad interface are judged by different limits, and only one is a reliability risk.
This article explains what the figure measures, which document decides acceptance, how to read the report, and which process causes produce each pattern.
What IPC-7095 does and does not decide
IPC-7095 is a process guide for BGA assembly and inspection. It describes how voids form, how to measure them consistently, and how to set up a control plan. It does not issue the accept or reject call on its own: that comes from IPC-A-610 workmanship criteria and, above both, from the customer drawing.
The figure most programmes start from is a single ball whose void area stays within about 25 percent of the ball area in a 2D projection. High reliability programmes, medical imaging and aerospace in particular, often tighten that to 10 or 15 percent in the contract, and tighten it further on thermal pads. When a drawing is silent, the general figure applies; when the drawing states a number, that number is the only one that matters.
| Document | What it provides | Weight in a dispute |
|---|---|---|
| Customer drawing | The contract limit for that build | Decides |
| IPC-A-610 | Workmanship criteria, voiding as one criterion among several | Default where the drawing is silent |
| IPC-7095 | Process guidance, measurement method, control levels | Supports the disposition |
Location matters more than the number
Stress concentrates at the pad interface, where the intermetallic layer meets the copper land, and at the ball periphery. A void in the middle of a ball reduces the thermal path and the load bearing section, but the joint can survive years of cycling. A void on the interface is a fracture starting point, whatever its area, and a field of small voids in one plane is worse than one large round void in the middle.
| Void type | Where it sits | X-ray signature | Risk |
|---|---|---|---|
| Macro void | Ball centre, from flux outgassing | One dark round region | Thermal path loss, usually tolerable |
| Shrinkage void | Ball centre, irregular outline | Small dark speck | Low |
| Planar microvoid | At the intermetallic layer | Many tiny dots in one plane | High, creates a brittle fracture plane |
| Interfacial void | Ball to pad boundary | Dark band or arc at the edge | High, reject regardless of area |
| Blow hole | Anywhere, explosive shape | Irregular asymmetric region | Critical, usually moisture in the package |
| Head in pillow | Ball never met the paste | Visible only at an angle | Critical, open joint |
2D against CT
A 2D transmission image projects the whole joint onto one plane, so every void along the path is added together. That is why a 2D figure reads higher than a CT slice of the same joint, and why a 2D number from one supplier cannot be compared with a CT number from another. A fair comparison uses the same method, angle and reference area.
For production, 2D is fast and enough for screening when the limit is stated and the worst ball reported. CT earns its time in failure analysis and in qualifying a new package, where the interface has to be seen rather than inferred. A report that gives only a mean hides both cases, because the mean of forty balls says nothing about the one that will crack.
Reading the report line by line
- Worst ball, not the average, and which ball it is: centre, corner or under the die shadow.
- Reference area: ball area or pad area, and whether the number is per void or the sum within one ball.
- View angle: 0 degree for the area figure, an oblique view where head in pillow is suspected.
- Thermal pad called out separately, since the limit for a large pad differs from the ball limit.
- Lot, panel position and serial number, so the trend is followed build to build.
Where the voids come from
Most of the work is upstream of the X-ray. A soak zone that is too short leaves flux volatiles to expand after the solder has started to freeze, which shows as evenly spread centre voids. Paste volume above target on a ground pad traps air under a large thermal land, which is why 3D SPI data is read together with the X-ray. Vias under the pad wick paste away and leave a starved joint, and unfilled via in pad adds a path for gas.
Two causes are easy to miss. Moisture in the package, from storage beyond the floor life allowed by J-STD-033, produces blow holes and popcorn. Finish quality produces planar microvoids: a nickel surface that was over-corroded before the gold went down gives the classic champagne void pattern, which is why the void trend and the finish control belong in the same conversation.
How SUPERB Assembly controls voiding
Assembly runs twelve zone reflow on SAC305 with nitrogen below 500 ppm, 3D SPI on paste volume before placement, and CT capable X-ray on hidden joints interpreted against IPC-7095, with void held below 5 percent on production builds. The same reports carry the worst ball, the thermal pad and the panel position, so a shift in the trend is visible before a lot is rejected.
Where a joint is suspect, a CT slice settles the question of location, and a microsection follows if the cause has to be proved. Component storage and baking follow J-STD-033, workmanship is held to IPC-A-610 Class 2 or Class 3 as the drawing states, and the quality system is ISO 9001 with ISO 13485 for medical builds.
- AOI and X-ray: which defect each method can actually see.
- Class 2 and Class 3: what the class changes in inspection scope.
- PCB testing: the full route from bare board to functional test.
- PCB capability: minimum pad and pitch, which set what X-ray can resolve.
Frequently asked questions
Is a 25 percent void always acceptable?
No. It is a starting figure for a single ball when the drawing is silent. If the void sits on the pad interface, if the ball is under a thermal pad, or if the contract tightens the limit, the same joint can fail. Disposition follows the drawing first and the workmanship standard second.
Why does my supplier report a lower void than my own X-ray?
Different reference areas, different view angles or 2D against CT. Agree the method before comparing figures, and compare worst ball to worst ball rather than mean to mean.
Can voids be reworked?
Small voids are usually left alone. A joint that exceeds the drawing limit is reworked by removal, pad cleaning and reball or replacement, then re-inspected by X-ray. The rework record matters as much as the new image.
Next step
Send the assembly drawing with the void limit and the reference method, the package list and the test requirement. The reply states which limit applies to each package, how the thermal pads will be judged, and quotes the build with X-ray reporting that shows the worst ball rather than an average.
