By the APTPCB SMT & Quality Engineering Team · Updated 1 October 2026 · Standards: IPC-7095, IPC-A-610, J-STD-001, J-STD-033, IPC-9252.
Most BGA voiding articles correctly state the 25% IPC limit. Almost none address the three things that actually determine whether you pass or fail field reliability: void location (center vs. interface vs. planar), Head-in-Pillow detection limitations of 2D X-ray, and the upstream SPI data that predicts X-ray outcomes before reflow. This guide covers all three, and the BGA X-ray inspection service page lists the 2D, 2.5D, and 3D CT methods APTPCB runs.
Key takeaways
- IPC Class 2 and Class 3 use the same 25% limit for total void area — but they are not equivalent. Class 3 adds documentation requirements and typically includes OEM-specific restrictions on interfacial voids (which IPC-A-610 alone does not prohibit by percentage).
- 2D top-down X-ray cannot reliably detect Head-in-Pillow. The ball and paste deposit overlap in the Z-axis; they appear connected even when separated by an oxide layer. Angled X-ray (45–70°) or 3D laminography is required — yet most production lines run only 2D AXI.
- Void location changes the verdict more than void area. A 20% center void in a BGA ball is generally benign. A 5% interfacial void (at the pad surface) is a crack initiation site and grounds for rejection regardless of area — this distinction is absent from most published guides.
- Nitrogen reflow measurably reduces voiding and is underused. Published data shows 5–15% absolute void reduction under N₂ atmosphere vs. air. For high-reliability products where vacuum reflow is not justified, N₂ reflow is the cost-effective middle step.
- SPI (Solder Paste Inspection) is the earliest X-ray predictor. Stencil aperture blockage and paste volume < 80% of nominal are the two leading indicators of both voiding and HiP defects — both visible before placement, not after reflow.
What IPC actually requires — and what it leaves to OEM contracts
| Standard | Void area limit (single ball) | HiP classification | Notes |
|---|---|---|---|
| IPC-A-610 Class 2 | ≤ 25% of ball cross-section | Process indicator — should be investigated | IPC text: "process indicator"; does not say "reject" |
| IPC-A-610 Class 3 | ≤ 25% of ball cross-section | Process indicator — higher documentation burden | Same percentage; difference is in process records and OEM overlays |
| IPC-7095 (BGA process guide) | References IPC-A-610 limits | Identifies warpage-driven HiP as a separate root cause | Design and assembly guide, not a pass/fail standard |
| Customer-specific (automotive, aerospace) | Often ≤ 10–15% for signal balls | Reject — zero tolerance | Contractual; not in IPC. Must be captured in work instructions |
| Thermal pad (QFN/LGA) | Up to 50% in some customer specs | N/A | Large thermal pads have different void-reliability physics |
The gap most engineers miss: IPC-A-610 calls HiP a "process indicator" — not an automatic reject. In practice, every responsible OEM treats any confirmed HiP as a reject, because HiP means no metallurgical bond was formed. If your work instruction reads "process indicator, document and continue," it is non-conforming to the intent of Class 3 production.
What is IPC-7095?
IPC-7095 (often searched as IPC 7095 or IPC7095) is the IPC guide titled Design and Assembly Process Implementation for BGAs. It explains how BGA design, paste printing, reflow, inspection, and rework interact, and it describes void types and their likely causes. It is a process guide: the pass/fail call on a finished assembly comes from IPC-A-610 or the customer contract. Queries often name a revision letter (for example IPC-7095D), so always confirm which revision your customer SOW cites. For how IPC-A-610 classes and acceptance terms work across all joint types, see our IPC-A-610 acceptance criteria overview.
| Question | IPC-7095 | IPC-A-610 |
|---|---|---|
| What is it? | BGA design and process guide | Acceptance standard for electronic assemblies |
| Used for | Root cause, process setup, void type classification | Accept/reject decisions by Class 1/2/3 |
| BGA voiding | Explains void types, locations, and causes | Sets the void-area acceptance limit |
| Head-in-Pillow | Links HiP to warpage and paste volume | Classifies the joint condition |
| Who uses it | Process, NPI, and design engineers | Inspectors and quality engineers |
Void types: not all voids are equal
The 25% rule applies to total void area, but the location of the void determines its failure mode and criticality.
| Void type | Location | X-ray signature | Criticality | Likely root cause |
|---|---|---|---|---|
| Macro void | Center to mid-ball | Dark circular region in ball interior | Low-medium — reduces thermal conductance | Flux volatile outgassing trapped as solder freezes inward |
| Shrinkage void | Ball center (small, irregular) | Small dark speck, irregular outline | Low — shrinkage artifact during solidification | Fast cooling rate; solder alloy composition |
| Planar microvoid / champagne void | At intermetallic layer (IMC) | Many tiny dots at pad interface | High — creates fracture plane; brittle failure | Plating contamination; pad surface chemistry issues |
| Interfacial void | Ball-to-pad or ball-to-component interface | Dark band or arc at pad boundary | High — crack initiation site; reject regardless of % | Pad oxidation; flux activity insufficient; ENIG "black pad" |
| Blow hole / popcorn | Anywhere — irregular, explosive shape | Irregular, asymmetric dark region | Critical — component delaminated | Moisture in BGA package; MSL exposure exceeded |
| Head-in-Pillow | Vertical separation between ball and paste | Snowman shape at angle; top-down looks normal | Critical — no bond; intermittent open | BGA or PCB warpage separating ball from paste during reflow |
Head-in-Pillow: why 2D X-ray misses it and what to do instead
The detection problem
In a top-down 2D transmission X-ray image, the solder ball sits above the solder paste deposit. Both are dense alloy — both block X-rays. From above, they appear as a single dense mass even when separated by a thin oxide or non-wetted surface.
A 2D image cannot resolve vertical separation. The HiP joint looks identical to a good joint in top-down view.
What works
| Method | HiP Detection | Voiding Detection | Cost/throughput impact |
|---|---|---|---|
| 2D AXI (top-down) | ❌ Unreliable | ✅ Good for macro voids | Baseline — fastest |
| 2.5D / angled X-ray (45–70°) | ✅ Good — shows snowman shape | ✅ Good | +15–30% cycle time |
| 3D laminography (CT slice) | ✅ Best — shows separation plane | ✅ Best — shows depth and location | +200–400% cycle time; NPI and sampling |
| Cross-section (destructive) | ✅ Definitive | ✅ Definitive | Destructive; coupon-based |
Head-in-Pillow X-ray checklist
- Inspect corner and edge balls first; warpage-driven HiP starts there.
- Use an angled (2.5D) view or 3D CT on every BGA where HiP risk is flagged; top-down only screens voids.
- Look for the "snowman" outline or a visible boundary between ball and paste.
- Compare the suspect ball with its neighbours; HiP balls often look slightly larger or non-collapsed.
- Confirm suspected HiP with cross-section or 3D CT before you change the process.
- Record serial number, package, ball position, and image for traceability.
Production recommendation: Run 2D AXI for voiding screening on all boards. Add 2.5D angled scan for all BGA packages ≥ 15 mm diagonal and for any lot where warpage is suspected. Reserve 3D CT for NPI first article and field failure investigation.
The "snowman" signature
At 45–70° angle, a HiP defect presents as two distinct round shapes — the BGA ball on top, the reflowed paste deposit below, with a visible separation line between them. A good joint at the same angle shows a single fused shape with no internal boundary.
Upstream control: SPI as the X-ray predictor
X-ray finds defects after reflow. SPI (Solder Paste Inspection) finds the conditions that create defects before placement. These two metrics are predictive of each other:
| SPI finding | Predicted X-ray outcome | Action |
|---|---|---|
| Paste volume < 75% of nominal on BGA pad | High risk of HiP (starved joint) | Stop and clean stencil; investigate aperture blockage |
| Paste volume > 130% nominal | High risk of bridging | Check stencil registration and print pressure |
| Paste offset > 25% of pad width | Alignment risk; voiding on offset side | Check stencil registration; re-print if > 25% |
| Paste height variance > 20% across board | Non-uniform wetting → non-uniform voiding | Investigate print support; check board planarity before print |
| Missing paste on any BGA pad | 100% risk of open or HiP | 100% rework before placement |
Implementation: Feed SPI data into your SPC system. Set a hard stop for paste volume < 75% on any BGA pad — do not reflow, rework paste. This prevents the defect rather than detecting it.
Quantitative effect of voiding on thermal resistance
The failure mode of BGA voiding is not just mechanical — it is thermal. Published data (IPC/JEDEC stress testing) shows:
- Each 10% of void area in a 1 mm-pitch BGA ball increases thermal resistance of that joint by approximately 5–8% relative to a void-free joint.
- A 25% void in a power BGA (where the ball carries both signal and heat) is measurably different from a 25% void in a signal-only ball.
- For BGAs used as flip-chip thermal paths (die-on-board with solder bumps), the IPC 25% limit may be contractually tightened to 10–15% by thermal engineers — not quality engineers.
Practical implication: Apply tighter void limits selectively to BGAs in the thermal path (large exposed pads, thermal balls, CSPs on power stages). Use the standard 25% limit for signal arrays.
Nitrogen reflow: the underused voiding reduction tool
| Atmosphere | Typical total void % (macro voids) | Cost impact | Best for |
|---|---|---|---|
| Air | 8–20% (process-dependent) | Baseline | Standard commercial products |
| Nitrogen (N₂, O₂ < 100 ppm) | 5–12% | +$0.10–0.30/board est. | High-reliability; power BGAs |
| Vacuum reflow | < 3% | +$0.50–1.50/board est.; lower throughput | Aerospace; medical implant; server power modules |
N₂ atmosphere reduces pad and ball surface oxidation during reflow, improving solder wetting. Better wetting allows gas bubbles to escape the molten solder before solidification. It does not eliminate voiding caused by paste chemistry or profile problems — but it reduces the baseline by 5–15 percentage points, which is often the margin between passing and failing a tight customer spec.
Full inspection and acceptance ruleset
| Parameter | Accept | Conditional review | Reject |
|---|---|---|---|
| Total void area (single ball) | ≤ 25% | 20–25% — document, monitor trend | > 25% |
| Largest single void | ≤ 20% of ball | 15–20% — root cause investigation | > 20% |
| Interfacial void (pad surface) | 0% | — | Any presence |
| Planar microvoid / champagne void | 0% | — | Any presence |
| Head-in-Pillow | 0% | — | Any confirmed HiP |
| Solder bridge | 0% | — | Any bridge |
| Solder ball diameter variance | ±15% of nominal | ±10–15% — check warpage cause | > ±15% |
| Corner ball elongation | < 25% elongation vs. center balls | 15–25% — warpage investigation | > 25% |
| Starved joint (volume) | > 75% of nominal | 65–75% — SPI review | < 65% |
| Popcorn / blowout | 0 | — | Any occurrence |
| Component offset | < 25% of pad width | 15–25% — placement check | > 25% |
Implementation workflow
Set class and OEM-specific limits first. Load IPC Class 2 or 3 into AXI software. Layer in any customer overlay (e.g., ≤ 15% for power balls, 0% interfacial voids). These must be in the work instruction, not just memory.
Calibrate with a golden board. Run a known-good board to baseline grayscale thresholds for solder, copper, laminate, and void. Teach all BGA component positions from the P&P file.
Run 2D top-down for void screening. Auto-flag any ball > 25% void area. Manually review any ball 20–25% for trend patterns.
Run 2.5D angled scan for all BGAs ≥ 15 mm diagonal. This is the minimum for HiP detection in production. Inspect corner balls first — HiP rates are highest at corners due to warpage.
Classify void location on any flagged ball. Use depth-focus or 3D reconstruction to determine center void vs. interfacial void. Interfacial = reject. Center = document and hold for trend review.
Correlate defect map with board position in oven. Systematic voiding concentrated at one region of the board indicates an oven zone issue (cold spot → insufficient TAL) or a support/fixture issue creating uneven thermal contact.
Log every X-ray result with board serial number and AXI image. Required for Class 3 traceability and field failure investigation.
Troubleshooting
High void percentage (> 25%), distributed uniformly across BGA
- Root cause: Insufficient soak time — flux volatiles not outgassed before solder melts.
- SPI correlation: Paste volume and height normal — profile is the issue.
- Fix: Extend soak zone by 15–30 seconds; verify TAL ≥ 40 s for large BGAs.
- Long-term: If standard profile cannot be extended, switch to vacuum reflow or N₂ atmosphere.
Voids concentrated at BGA corners only
- Root cause: Corner areas have different thermal mass than center → they reach liquidus at different times, creating non-uniform outgassing windows.
- Also indicates: Board or component warpage causing uneven standoff height, which changes heat transfer.
- Fix: Flatten the thermal gradient across the board; verify board warp < 0.5% before assembly.
HiP confirmed on angled X-ray, not visible on 2D
- Root cause: BGA package warped convex during reflow (or PCB warped concave), lifting ball from paste at peak temperature. They re-contact on cool-down but without coalescence.
- SPI correlation: Check paste volume and height — if paste was thin at the failed joints, investigate stencil aperture blockage.
- Fix: Increase paste volume (step-up stencil or aperture enlargement); slow cooling rate to 1–2 °C/sec; bake BGAs per J-STD-033 before assembly.
Champagne voids / planar microvoids at IMC layer
- Root cause: Surface contamination of pad (organic residue, oxidation, or ENIG corrosion "black pad").
- Detection: Requires 3D CT or cross-section to confirm planar distribution vs. random bubbles.
- Fix: Evaluate pad surface finish; replace any ENIG from suspect plating bath; check incoming PCB storage (OSP shelf life).
Solder bridging under BGA
- Root cause: Paste volume too high; stencil aperture not reduced (standard: 80–85% area ratio for fine-pitch BGA pads); paste slump.
- SPI correlation: Will show paste volume consistently above 120% before reflow.
- Fix: Reduce aperture by 10–15%; verify stencil is clean on underside; confirm aspect ratio ≥ 0.66.
Design factors that determine X-ray outcomes
Via-in-pad (open vs. filled) Open vias under BGA pads wick solder into the barrel, creating starved joints above and voiding below. Filled and capped vias (VIPPO) prevent solder loss. Our guide on VIPPO BGA void control covers fill, planarity, and capping checks.
NSMD vs. SMD pad definition NSMD (Non-Solder Mask Defined): solder mask opening > copper pad. Solder wraps the pad edge. Better wetting, lower void risk, good self-alignment. SMD: mask overlaps copper. Less solder contact area. Used at very fine pitch but increases void risk and stress at mask edge.
Component clearance for angled X-ray If tall components (connectors, shields, capacitor arrays) are placed within 3 mm of a BGA, they can shadow the X-ray detector at the required 45–70° angle, making HiP detection impossible. Design clearance requirements into the placement rules for any BGA requiring Class 3 inspection.
Stencil aperture geometry Square apertures with rounded corners (radius ≥ 20% of aperture side) release paste more completely than pure circles. Electro-polished stencil vs. laser-cut: electro-polished walls reduce paste adhesion, improving release and reducing paste-volume inconsistency across the BGA field.
FAQ
Does IPC Class 3 require tighter void limits than Class 2? The IPC-A-610 percentage (25%) is the same for Class 2 and Class 3. Class 3 is stricter in documentation, process records, and traceability — and most Class 3 OEM contracts add zero-interfacial-void clauses that are not in the base IPC standard. Always check the customer SOW, not just IPC.
Can 2D AXI detect Head-in-Pillow? Rarely and unreliably. Top-down 2D shows the ball and paste as a combined mass regardless of whether they are bonded. HiP detection requires angled X-ray (2.5D) at 45–70° or 3D CT. Published research (Meridian/SMTA) confirms false-negative rates > 60% for HiP in 2D-only inspection lines.
What is a "champagne void" and why is it more dangerous than a regular void? Champagne voids are many tiny voids clustered at the intermetallic compound (IMC) layer between solder and copper pad. They create a planar fracture surface — instead of the solder tearing under stress, it shears cleanly along the void plane. The failure is brittle, with almost no plastic deformation warning. Caused by plating chemistry problems or surface contamination, not reflow profile.
Does nitrogen reflow really help? Yes, measurably. N₂ atmosphere prevents pad and ball surface re-oxidation during reflow. Better wetting lets gas bubbles escape molten solder. Published production data shows 5–15 percentage point reduction in total void area vs. air reflow with identical paste and profile. For products requiring < 15% total voiding, N₂ is the most practical first step before considering vacuum reflow.
What is the difference between IPC-7095 and IPC-A-610 for BGA voiding? IPC-7095 is the BGA process guide: it describes void types, their causes, and how to prevent them. IPC-A-610 is the acceptance standard that inspectors apply. When a customer asks for "IPC-7095 BGA voiding criteria," clarify whether they mean the guide's process targets or the IPC-A-610 accept/reject limit, and write the agreed limit into the work instruction.
Can I rework a BGA that fails X-ray? Yes — BGA rework (IR/hot-air removal, site cleaning, re-paste, reballing or new component, re-solder) is a defined process in IPC-7711/7721. Practical limits: pad damage typically limits rework to 2 cycles maximum; reballed BGAs need re-qualification for Class 3 products. Track all reworks per serial number for traceability.
How does SPI data connect to X-ray outcomes? Paste volume < 75% on a BGA pad is the primary leading indicator of HiP defects. Paste volume > 130% correlates with bridging. Paste offset > 25% correlates with alignment-induced interfacial voiding on the offset side. SPI data from the paste printer is the earliest process control gate — X-ray is the verification, not the primary control.
What is the typical cost of 100% X-ray for BGA boards? For 2D AXI on a standard SMT line, cycle time adds 30–90 seconds per board depending on BGA count. For high-volume products, 100% AXI (2D) is standard. 2.5D angled scans add 15–30% cycle time. 3D CT is NPI and sampling only — one board can take 5–20 minutes at full resolution.
Related pages
- SMT and THT Assembly — SMT assembly with vacuum reflow option for high-reliability BGAs
- BGA, QFN and Fine-Pitch Assembly — BGA placement, reflow, and X-ray release for fine-pitch packages
- X-Ray Inspection — Equipment and inspection capabilities at APTPCB
- Warpage Control During Assembly — Root cause analysis for warpage-driven HiP defects
- DFM Guidelines — Via-in-pad, stencil, and pad definition rules for BGA assembly
Glossary
| Term | Definition |
|---|---|
| BGA | Ball Grid Array — SMD package using a grid of solder balls as I/O |
| Head-in-Pillow (HiP) | Defect where the BGA ball and solder paste never coalesce — no metallurgical bond |
| Voiding | Gas or air pockets trapped inside a solder joint during solidification |
| Interfacial void | Void located at the ball-to-pad or ball-to-component interface — crack initiation site |
| Champagne void | Multiple tiny voids clustered at the IMC layer, creating a planar fracture surface |
| IMC | Intermetallic Compound — the Cu-Sn bond layer between solder and copper pad |
| Laminography | 3D X-ray method generating cross-sectional slices at selectable depths |
| AXI | Automated X-ray Inspection — production X-ray system for inline BGA inspection |
| SPI | Solder Paste Inspection — optical measurement of paste volume, height, and offset after printing |
| NSMD | Non-Solder Mask Defined — pad definition where solder mask opening is larger than copper pad |
| VIPPO | Via In Pad Plated Over — filled and capped via under BGA pad; prevents solder wicking |
| TAL | Time Above Liquidus — duration the solder is fully molten during reflow |
| IPC-7095 | Design and Assembly Process Implementation for BGAs — process guide, not a pass/fail standard |
| N₂ reflow | Reflow under nitrogen atmosphere (O₂ < 100 ppm) to reduce surface oxidation and voiding |
The 25% void limit is the starting point, not the complete answer. Whether a 20% void is acceptable depends on where it is (center vs. interface), what the component does (signal vs. thermal path), and what your OEM contract says. HiP defects require angled or 3D X-ray to detect — running 2D-only and calling it BGA inspection is a quality gap. Fix the inspection method first, then trace systematic defects back to the SPI data that predicted them.
For NPI BGA qualification or a production process audit, APTPCB's engineering team can review your reflow profile, stencil design, and SPI control limits before your first build.