The solder joints of a BGA (ball grid array) sit under the component where the eye cannot see them, so the order is to start with methods that do not damage the board: look at the outer row, test electrically, and X-ray. X-ray shows ball position, bridging and voids, but has trouble with fine cracks and joints that did not bond properly. To confirm which interface a joint broke at, use dye-and-pry (stain, then pull apart); to see what the fracture looks like, cross-section. These last two destroy the sample, so they come at the end and usually need to be agreed with the customer first.
Each method sees different things: a review by NASA’s Jet Propulsion Laboratory notes that X-ray is good at voids, solder volume, bridging and solder balls, while cracks, cold joints and disturbed joints are often invisible to X-ray and need visual inspection or taking the part apart. Joints hidden under the component cannot be seen visually either, which is why dye-and-pry and cross-section are needed.
Screen first, then take apart: an IPC conference paper describes using automated X-ray to screen for head-in-pillow (where ball and paste do not coalesce), with an average detection rate of about 70%. Suspect joints were then re-checked with high-resolution manual X-ray, and the paper notes that some head-in-pillow joints look exactly the same as normal ones.
Agree criteria first: the scope clause of IPC-A-610J states that it does not provide criteria for cross-sections or X-ray, and that X-ray guidance is in Appendix C of J-STD-001. How much voiding passes, and how much dyed area passes, must be agreed with the customer in advance.
For how solder joints are made, see “How Does an SMT Line Work? Printing, Placement, Reflow and Quality Control in One Article” and “How to Choose Solder Paste”; for how acceptance standards differ, see “J-STD-001 vs. IPC-A-610.” This article starts from “the BGA is soldered — how do you know whether the joints are good and where they failed?” It covers how to choose inspection methods and in what order to use them. Why head-in-pillow happens, and how to remove and resolder a BGA, are outside its scope. Each stage covers how to do it and how JIAHON can help; sources and exceptions are in the notes at the end.
Figure 1. BGA joint inspection order: non-destructive first, then destructive, then back to the process. Non-destructive: (1) No disassembly — sees the outer row of balls and continuity; cannot see inner joints, cold joints or cracks. (2) X-ray — sees position, bridging and voiding; cannot see microcracks or cold joints. Destructive (ask the customer first): (3) Dye-and-pry — sees which interface cracked first and at which layer it broke; cannot give strength values or cover other components. (4) Cross-section — sees cracks in the section and solder microstructure; cannot see other planes or areas missed by an off-angle cut. Back to the process: (5) Report — agree criteria, write the report clearly, and note what was not done; trace back to the process. In practice one method is not enough, and they are often chained: X-ray screens suspect joints (screen first, without damaging the board) → dye-and-pry confirms where it broke (faster when the pin is unknown or in the inner rows) → cross-section shows the cause (once you know which ball to cut). (Illustration: each method sees different things and they complement each other.)
Start here: what each of six methods can see
| Method | Can see | Cannot see, or not reliably | Damages the sample? |
|---|---|---|---|
| Visual or side view of the outer row | Gaps between outer balls and pads, whether solder has wetted up, solder bridges | The inner rows; balls hidden by the component | No |
| Electrical test (ICT, JTAG, functional) | Continuity | Cold joints, some cracks, voids; an intermittent joint may happen to be in contact at the moment of test | No |
| X-ray (manual, with tilt and rotation) | Ball position, size, bridging, missing balls, voids; at an angle, gaps between ball and pad | Very fine cracks, cold joints; on double-sided boards, parts on the other side get in the way | No |
| Automated X-ray (AXI) and computed tomography (CT) | Continuous screening of whole boards; automatic void calculation; CT can view layer by layer | Public sources still report defects it cannot see (see Stage 2) | No |
| Dye-and-pry (stain, then pull apart) | Which interface had already cracked before pulling; at which layer it broke | Just the part that was pulled; the result is qualitative, not a strength value | Destroys the component and the cut-out piece of board |
| Cross-section | Solder, pads, cracks and microstructure on one section plane | One plane alone; if the cut is off, the problem disappears | Destroys that part |
Stage 1: Don’t take it apart yet — preserve the evidence, look at the outer row, test electrically
When you get a failed board, do not rework it and do not force it. Once a destructive method is used the evidence is gone, so keep what can be kept. One returned-goods failure analysis procedure JIAHON Engineering has seen runs in this order: numbering and visual inspection, then X-ray, then functional test, after which photos of anomalies go to design and process staff for review and are compiled into a report.
| What to do | How | Why |
|---|---|---|
| Number and photograph | Number each board and record orientation (pin 1 location) and component location; photograph the whole board and the area around the component | So the results of every later method can be traced back to the same joint |
| Look at the outer row | Look at the outer balls at an angle under a microscope: is there a gap between ball and pad, has solder wetted up, is there bridging? | You can see the outer row alone, and just when there is room beside the component; a good outer row does not mean the inside is good |
| Look at the flux | Is the flux around the joints cracked or disturbed? | The figure legend in IPC-TM-650 2.4.53 notes that disturbed flux may indicate the joint or board material has been stressed |
| Match electrical records | Map ICT (in-circuit test), JTAG (boundary scan) and functional test failure records to joints | Electrical tests answer “is it connected”; cold joints, some cracks and voids may not register, and an intermittent joint may happen to be in contact at the moment of test |
| Agree whether destruction is allowed | Ask the customer first: which boards may be destroyed and which must be returned as is | Destructive analysis usually needs the customer’s agreement |
How JIAHON can help:
- High-resolution 7-inch digital microscopes: JH-G1200 with 1–1200x and JH-G1600 with 1–1600x continuous zoom, working distance from 10 mm to infinity, adjustable lens angle, 7-inch touchscreen, built-in photo and video; the advanced accessory kit includes an auxiliary light and a 32 GB memory card for keeping records. Whether you can see the outer balls depends on how much room is left beside the component.
- 4K 3D digital microscope TD-4KHU: built-in measurement system, on-screen viewing, multi-angle rotating display.
- External monitors for microscopes and KVM switches: Full HD resolution, 180° rotating stand; for several people to view and record together.
- Measurements from digital microscopes are mainly for reference, not precision measurement: JIAHON’s “Digital Microscopes | Q&A on Error and Calibration” explains that differences between operators are larger than differences in the equipment itself.
- JIAHON does not offer electrical testers (ICT, JTAG).
Stage 2: X-ray — check position, size, bridging and voids first
X-ray does not damage the board and can see under the component. Decide three things: what angle to view from, what equipment to use, and what to look for.
| Approach | Can see | Limitations |
|---|---|---|
| Top-down | Ball outline, size, bridging, missing balls, voids | Gaps between ball and pad, and cracks, cannot be seen top-down; nor can you tell whether a void is inside the ball or at an interface |
| Angled view with rotation | Ball separated from pad (head-in-pillow), larger cracks; whether a void is inside the ball or at an interface | A 2008 industry article recommends tilting about 55 to 70 degrees and rotating all the way around; very fine cracks are still hard to see clearly; on double-sided boards, parts on the other side get in the way |
| Automated X-ray (AXI) | Continuous inspection of whole boards, automatic void calculation, fast | A trade-off between detection rate and false calls; in the conference paper above, the average detection rate for head-in-pillow was about 70%; the 2008 industry article also says manual X-ray measures voids more accurately than AXI, but more slowly |
| Computed tomography (CT) | Reconstructs a 3D image that can be viewed layer by layer | NASA’s study covered column grid arrays (CGA) and stacked packages, not ordinary plastic BGAs: CT magnification was lower and did not reveal joint damage; that was 2012 equipment, newer equipment may differ, and you should test with your own boards before buying |
What to look for:
- Position and shape: are balls offset or missing, is there bridging, are sizes consistent?
- Voids: the percentage of ball area taken by voids, and whether the void is inside the ball or at the ball-to-component or ball-to-board interface. An IPC conference paper notes that voids at the interface weaken a joint more readily than voids inside the ball.
- What it cannot see: very fine cracks (public articles say microcracks smaller than about 25 µm are difficult for all X-ray systems), cold joints, and some head-in-pillow.
Field experience:
- Low yield across a whole lot: JIAHON Engineering dealt with a lot of BGA modules with low yield across the lot. X-ray of each part showed some balls offset and excess solder, and at the time the low yield was judged to be related to ball offset. A few parts showed no obvious anomaly on X-ray; whether to cross-section those waited for the customer’s agreement.
- Not seen on X-ray does not mean nothing is wrong: in another defect analysis JIAHON Engineering has seen, a lot of test failures was examined by both X-ray and cross-section. For one group, both methods showed no wetting on the component side; for another group, neither method showed any joint anomaly, so the cause had to be sought beyond the joints.
How JIAHON can help:
- JIAHON does not have X-ray inspection machines or a product page for X-ray inspection services. If you are buying a machine or sending samples out, ask the equipment maker or lab to run samples with your own boards and check three things: whether it can tilt to the angle you need, whether the resolution is enough to see the defects you are looking for, and whether parts on the other side of a double-sided board get in the way. The IPC conference paper also notes that manual X-ray’s ability to see head-in-pillow improves with resolution, contrast and tilt capability.
- Catching solder volume before reflow can reduce ball offset and excess solder: the SPI implementation evaluation is an evaluation report JIAHON compiled, not a piece of equipment; for troubleshooting printing defects, see “Troubleshooting Solder Paste Printing Defects.”
Stage 3: Dye-and-pry (stain, then pull apart) — confirm which interface it broke at
IPC’s current name for the dye-and-pry test is Dye and Pull; it used to be called Dye and Pry. The sample is immersed in dye and put under vacuum so the dye penetrates any existing cracks; after drying, the component is pulled off the board and you look at which surfaces are stained. Stained surfaces were already cracked before pulling; unstained surfaces broke during this pull. The table below follows IPC-TM-650 2.4.53 (2017 edition).
| Step | Key points in the standard | Why it matters |
|---|---|---|
| Inspect and photograph | First check for signs of mechanical damage and whether the flux is cracked | After pulling, this evidence is gone |
| Cut out the sample | Leave about 19 to 38 mm of board around the component; a diamond saw is recommended, and no force should be applied to the joints | Applying force while cutting can create cracks of its own |
| Remove flux | Soak for at least 1 hour in a cleaning solution that dissolves the flux; alcohol will not do | If flux is not fully removed, dye cannot penetrate, giving a false “looks fine” result |
| Dye and vacuum | Immerse the whole sample in dye; vacuum 67.7 kPa (about 20 inHg) for 3 to 4 minutes, release a little and pull again, with a total soak of at least 30 minutes; do not exceed this vacuum | Exceeding it makes the dye boil; dye that is too thick will not penetrate cracks and should be replaced |
| Dry | Oven no higher than 100 °C; overnight if possible | Dye that is not dry smears when pulled, causing misjudgment |
| Pull | Bond a nut to the top of the component and apply an even pull with a pull fixture | The standard requires an even pull; the old name was “pry,” now “pull” |
| Examine the fracture surfaces | Pair the fracture surfaces of both sides; stained surfaces were already cracked | Judged by dye indications, not a strength measurement |
| Record | Stain location and the layer where it broke; record the percentage of stained area if needed | Criteria are agreed between the lab and the customer |
Where it broke is a clue to the cause, not a conclusion:
- Component-side interface: ball separated from the component pad.
- Through the solder: broke within the ball itself.
- Board-side interface: solder separated from the PCB pad.
- Pad lifted: the pad came off along with the material under it — the laminate cracked (cracking of the board material under the pad, called pad cratering), which is a different matter from whether the solder bonded well.
Limitations:
- The standard says it is for observation: IPC-TM-650 2.4.53 states that the method is for observing dye indications, not a strength test.
- Results depend on technique: flux not fully removed means missed cracks; dye not dry means misjudgment.
- Labs differ: in older records JIAHON has referred to, some labs’ methods differ from the current IPC version (with higher bake temperatures), so check which version a report follows.
- What reports look like: lab reports JIAHON has seen commonly include the test purpose, sample source and description, the standard followed, equipment, procedure, and for each BGA a grading of fracture area with photos.
How JIAHON can help:
- Laboratory vacuum pump (diaphragm pump): oil-less; pumping speed 7.0 L/min, ultimate vacuum below 130 mbar (absolute); for gases and vapors alone. IPC specifies a vacuum of 67.7 kPa (about 20 inHg) and says not to exceed it, so use it with a pressure regulator and vacuum gauge. JIAHON has no product pages for vacuum chambers, vacuum gauges or dyes; please ask.
- Flux removers: JH64, JH46 and JH37, designed for no-clean and rosin fluxes. Whether they will fully clean the flux on your samples — please give the flux model and process conditions when asking for a quote, and try a small piece first.
- To view stained surfaces and keep photo records, use the digital microscopes in Stage 1.
- JIAHON does not offer pull fixtures, pull testers or diamond saws.
Stage 4: Cross-section — see cracks and microstructure in the section
Cross-sectioning cuts the joint open, grinds it flat and examines the section under a microscope. It shows things clearly, but it is labor-intensive and shows just one plane.
| Approach | Can see | Limitations |
|---|---|---|
| Vertical cross-section | Whether there is a gap between solder and pad, the path of cracks, the solder microstructure, whether the pad has lifted; you can then section the next row | Destructive; labor-intensive; one plane alone — if the cut is off or ground too far, the problem disappears |
| Planar grinding | Grind layer by layer to the target height and look down at the ball-to-pad bond | Also destructive; you must grind to the right height |
- Know which ball to cut first: the IPC conference paper says that before cross-sectioning you must know which pin has the problem, while dye-and-pry is faster when the pin is unknown and the problem is in the inner rows. So first use X-ray and electrical records to find the suspect joints.
- For finer microstructure: for example, intermetallic compound (IMC) thickness needs a higher-magnification metallurgical microscope or a scanning electron microscope (SEM); a public BGA inspection article also used SEM to confirm cracks seen on X-ray.
- Be specific when sending to a lab: which ball to cut, from which direction, and what to look at, with the X-ray images and electrical records attached.
How JIAHON can help:
- After cross-sectioning, the digital microscopes in Stage 1 can show the general shape of the section and the location of cracks; measurements are for reference. To measure IMC thickness, use a lab’s metallurgical microscope or SEM.
- JIAHON does not offer sectioning saws, grinders or mounting resin.
Stage 5: Agree the criteria, write the report, then go back and fix the process
The purpose of inspection is to improve the process; if the report is unclear or the criteria were never agreed, the earlier work is wasted.
| Agree with the customer first | Content | Basis |
|---|---|---|
| Voids | How much void area per ball passes; whether voids inside the ball and at interfaces are judged separately | The IPC-7095C table of contents includes void identification (7.5) and voids in solder balls (4.8.2); a 2006 paper notes that IPC-7095A and IPC-A-610D treated voids differently at the time, and both allowed buyer and supplier to agree otherwise |
| Dye | How much dyed area counts as a problem; whether to map each ball’s location | IPC-TM-650 2.4.53: criteria are agreed between the lab and the customer |
| Who reads X-ray | X-ray readers should be trained | The J-STD-001J table of contents includes “X-ray-specific personnel proficiency” (1.10.1) |
| Order of destruction | Non-destructive methods first, destructive after; ask the customer before destroying | JIAHON Engineering’s cases and failure analysis procedure |
Put at least these items in the report:
| Item | What to include |
|---|---|
| Sample | Number, board number, component location and orientation |
| Method and basis | Which edition of the standard; method parameters (X-ray voltage and angle, dye and vacuum, drying conditions) |
| Images | X-ray images (with the angle noted), photos of both stained surfaces, cross-section photos and magnification |
| Results | Location and grade of each joint; a dye location map (IPC-TM-650 2.4.53 has an example marked by coordinates) |
| What was not seen | Methods not used and defects that cannot be seen, written as “not done” or “not visible,” not as “none” |
What you see, and where to look first (a starting point for finding the cause, not a conclusion):
| What you see | Where to look first | Related article |
|---|---|---|
| Ball offset, excess solder, bridging | Paste printing and placement: stencil, squeegee, placement position | “Troubleshooting Solder Paste Printing Defects” |
| Ball and paste not coalesced, incomplete soldering | Whether the profile met requirements; whether board and component warped in reflow, or had absorbed moisture | “How to Attach Thermocouples for Reflow Profiling,” “White-Light Structured-Light Thickness and Deformation Measurement” |
| Excessive voids | Solder paste, reflow profile, stencil apertures | “How to Choose Solder Paste” |
| Cracked, bulging or delaminated components | Moisture and baking | “What Is the Popcorn Effect,” “Storage and Baking Conditions for Moisture-Sensitive Devices” |
How JIAHON can help:
- Choosing a reflow profiler: JIAHON lays out the specifications of three common reflow profiler brands side by side, and explains which specifications matter in which situations.
- Thermocouple wire (type K): PTFE-insulated, 0–260 °C, and fiberglass-insulated, 0–480 °C; fine 0.127 mm wire with fast thermal response, used for spot-welding onto boards to measure reflow profiles.
- For board and carrier warpage, see the non-contact measurement methods summarized in “White-Light Structured-Light Thickness and Deformation Measurement”; for controlling solder volume, see the SPI implementation evaluation in Stage 2.
- If you need digital microscopes, microscope monitors, laboratory vacuum pumps, flux removers, reflow profilers or thermocouple wire, please contact JIAHON. JIAHON does not offer X-ray inspection machines, or cross-section or dye-and-pry equipment and services.
Related reading
To cover the related process steps, continue with these articles:
- “How Does an SMT Line Work? Printing, Placement, Reflow and Quality Control in One Article”
- “J-STD-001 vs. IPC-A-610: Process Standard and Inspection Standard Compared”
- “How to Choose Solder Paste: 5 Validation Stages from Data Sheet to Pilot Run for Your Candidates”
- “Troubleshooting Solder Paste Printing Defects: Work Back from the Print to the Consumable to Change”
- “How to Attach Thermocouples for Reflow Profiling: Five Methods Differ by 9 °C, and the Risk of Tape Lifting Components”
- “What Is the Popcorn Effect: Why Moist ICs and Boards Crack in Reflow”
- “Tray vs. Tape and Reel: 8 Comparisons, and How to Load and Store BGAs and QFPs”
- “What an SMT Process Audit Asks: 6 Stages of Line Documents and Records, Starting at Solder Paste Printing”
Notes: sources and exceptions
These notes are for readers who want to check the sources; you can follow the steps above without them. For the text of standards, refer to the official originals; where your customer’s specifications or a lab’s methods differ from this article, they take precedence.
Sources for each practice
| Practice | Content | Source |
|---|---|---|
| What each method sees | X-ray is good at voids, solder volume, bridging and solder balls; cracks, opens, cold and disturbed joints and flux residue are judged more accurately by visual inspection; hidden joints can be seen by X-ray alone; delamination inside the package cannot be seen by X-ray and needs acoustic scanning | NASA/JPL (Ghaffarian, 2012), Table 2.3-1 and Chapter 2 |
| Computed tomography (CT) | For the column grid arrays and stacked packages studied, 2D X-ray could not see microcracks; 3D CT had lower magnification and showed just column tilt, not joint damage; the conclusion is that X-ray must be complemented by other methods | NASA/JPL (Ghaffarian, 2012), Chapter 5 |
| Pros and cons of traditional methods | Microscopes (including 90° prisms) see the perimeter alone and rely on judgment; cross-sections show cracks and microstructure but are destructive, labor-intensive and two-dimensional; electrical tests cannot see cold joints, some cracks and voids; X-ray cannot see cold joints and fine cracks | NASA NEPP presentation (Denning; older, so prices and equipment capabilities have changed; cited for the pros and cons of methods alone) |
| Head-in-pillow inspection flow | Electrical tests may miss intermittent contact; automated X-ray averaged about 70% detection, with suspects re-checked by manual X-ray; side-view microscopes see the outer row alone; dye-and-pry is faster when the pin is unknown and in the inner rows; you must know the pin before cross-sectioning | IPC APEX 2011 conference paper (Choo et al.); the authors’ field experience, not a standard |
| X-ray angle and resolution | Tilt about 55 to 70 degrees and rotate; head-in-pillow separations smaller than 1 mil (about 25 µm) can be seen; microcracks smaller than 1 mil are difficult for all X-ray; cracks of 3 mil (about 76 µm) or more are easy to see | Industry article in the July 2008 issue of SMT magazine (Feng et al.; an industry article, not a standard) |
| Void location and standards | Voids are inside the ball, at the ball-to-component interface or at the ball-to-board interface; interface voids weaken joints more readily; IPC-7095A and IPC-A-610D treated them differently, and both allowed buyer and supplier to agree otherwise | IPC APEX 2006 conference paper (Balsavar) |
| Dye-and-pry (Dye and Pull) | Procedure, vacuum, drying temperature, flux removal, pulling, report items and example figures; the method is for observation | IPC-TM-650 2.4.53 (August 2017 edition, full text read) |
| BGA inspection sections | 7.3 inspection techniques (X-ray, acoustic microscopy, ball height, optical inspection, destructive analysis), 7.5 void identification, 3.5.1 limits of visual inspection, 4.8.2 voids in solder balls | IPC-7095C (table of contents) |
| Acceptance standards and X-ray | IPC-A-610 scope clause: does not provide cross-section or X-ray criteria, with X-ray guidance in J-STD-001; the J-STD-001J table of contents has X-ray-specific personnel proficiency and Appendix C “X-Ray Guidelines” | IPC-A-610J (table of contents and scope clause), J-STD-001J (table of contents) |
| Returned-goods failure analysis and cases | Order of failure analysis; X-ray inspection of a lot with low yield; the group where neither X-ray nor cross-section showed an anomaly; common contents of lab reports | Older records JIAHON has referred to (single cases, with names and figures removed) |
This describes the general case; individual situations may differ.
Exceptions and limitations
- Age: the public articles and papers date from 2006, 2008, 2011 and 2012. X-ray equipment keeps improving, so the detection rates and visible sizes in them are for reference, not current equipment capability.
- CT: NASA’s results are for column grid arrays and stacked packages using 2012 equipment, and cannot be taken directly as conclusions for ordinary plastic BGAs and current equipment. This article found no public CT comparison for ordinary BGAs, so it does not say what CT can see.
- Detection rate: the roughly 70% is the average from a single paper, a single plant and a single automated X-ray system, not a general rule.
- Dye-and-pry: this article follows the 2017 edition of IPC-TM-650 2.4.53; earlier editions and labs’ own methods will differ; the method is for observation, not a strength test.
- Criteria values for voids and dye: this article gives no percentages; for the values in IPC-7095 and IPC-A-610, see the official originals and agree them with your customer. For IPC-7095C, IPC-A-610J and J-STD-001J, we read the tables of contents and scope clauses.
- Scope: this article does not cover delamination inside the package (which needs acoustic scanning), the causes of head-in-pillow, BGA rework and reballing, or prediction of pad cratering.
- Equipment: JIAHON does not offer X-ray inspection machines, sectioning saws, grinders, or dye-and-pry staining and pull test equipment; JIAHON has no product pages for dyes, vacuum chambers or vacuum gauges.
- Product information: specifications are as stated on each product page. The diaphragm pump’s ultimate vacuum (below 130 mbar absolute) is deeper than the 67.7 kPa vacuum IPC specifies (about 34 kPa absolute), so a regulator and vacuum gauge are needed to control it; this is a reminder from comparing the two figures, not a statement on the product page, so please confirm with your lab’s equipment.
Standards and references
Refer to the current version published by each issuing body for the standards cited here.
- IPC-TM-650 2.4.53, Dye and Pull Test — test method for staining and pulling BGA joints (August 2017 edition). Issuer: IPC (Global Electronics Association)
- IPC-7095C, Design and Assembly Process Implementation for BGAs (table of contents). Issuer: IPC (Global Electronics Association)
- IPC-A-610J, Acceptability of Electronic Assemblies (table of contents and scope). Issuer: IPC (Global Electronics Association)
- J-STD-001J, Requirements for Soldered Electrical and Electronic Assemblies (table of contents). Issuer: IPC (Global Electronics Association)
- JPL Publication 12-21, 3D X-ray CT for BGA/CGA Workmanship Defect Detection (2012). Issuer: NASA Jet Propulsion Laboratory (JPL) / NEPP
- Denning, BGA Inspection — presentation on BGA inspection methods. Issuer: NASA Electronic Parts and Packaging Program (NEPP)
- APEX 2006 conference paper E16/S16-02 (Balsavar): location, criteria and standards for BGA voids. Issuer: IPC technical resources / conference
- APEX 2011 conference paper E6/S10_02 (Choo et al.): inspection methods and field experience for head-in-pillow. Issuer: IPC technical resources / conference
This article summarizes the general case. It is not a restatement of any standard, not an endorsement of any lab or piece of equipment, and not a promise about your process. Follow your customer’s specifications, the edition of the standard you use, the lab’s report and your own validation; individual situations may differ.