A small joint with retained solder-particle outlines is a graping-like joint worth investigating, not a diagnosis. The appearance alone does not prove oxidation, exhausted flux, an inadequate peak temperature, defective paste or any other root cause. First establish whether the particles are part of the joint, then compare print and local thermal evidence against unaffected locations before authorizing a controlled change.
For loose solder balls, beads and the broader investigation of uncoalesced particles around an assembly, see Solder Balls in SMT Reflow. A graping investigation starts with a narrower decision: what the morphology and available evidence say about coalescence at a particular joint.
First decision: what is the suspicious feature?
Record the board and joint locations before moving or reworking the board. Responsible quality personnel should decide whether further inspection or a hold is needed before rework or release under the project requirements. A close image of a shiny sphere is not enough to classify it; the question is where the material sits relative to the intended pad, component termination and joint.
| Observation | Working classification | What to check next |
|---|---|---|
| Particle outlines are retained within the solder at a small joint | Graping-like joint | Compare the joint with the matching print, neighboring joints and thermal evidence. |
| A sphere is separate from the intended joint on the board surface | Loose solder ball | Check its position, size and distribution separately from joint coalescence. |
| A larger sphere sits beside a component or joint | Possible solder bead | Check whether it is separate from, attached to or emerging from the joint. |
| Solder has not spread or bonded as expected on a pad or termination | Possible non-wetting | Inspect the exposed surface and interface; a joint can show both poor wetting and retained particles. |
| A joint appears insufficiently fused, with evidence that its local thermal exposure may have been inadequate | Possible underheating | Compare measured thermal evidence with the materials’ applicable requirements, recognizing the measurement location. |
These are working classifications, not acceptance decisions. Reflections, residue and limited viewing angles can obscure the joint boundary. Use scaled images from more than one useful angle where feasible. An authorized section may help reveal the joint structure and interfaces, but it will not reconstruct the full temperature or flux history.
For the distinction between interface wetting conditions, see solder non-wetting vs dewetting.
An image alone cannot prove joint quality. Electrical continuity alone cannot prove it either: a conductive path may exist in a joint whose mechanical or metallurgical condition still needs evaluation.
Build a location-linked evidence record
Start with the affected joint, then select unaffected locations that are genuinely comparable: preferably the same component and joint design on the same board and on other boards, if available. Include nearby joints that differ, because that contrast may be more informative than a collection of defect photographs.
Before rework, record:
- Board identity and revision, build or lot identification, panel position, and component reference and joint location.
- Scaled images of affected and unaffected locations, including enough surrounding area to show the joint boundary.
- Paste’s exact product and batch, plus available storage, opening, preparation, handling and exposure records.
- Stencil revision and actual print evidence mapped to the locations where possible—not merely the intended aperture design.
- Relevant as-built thermal measurements, instrument attachment locations and conditions, board population, oven recipe and run identification.
- Any missing records, prior handling or rework that could limit the comparison.
If a record, suitable comparison location or measurement does not exist, mark it unknown. Do not substitute a nominal recipe or an assumed process condition for observed evidence.
Print evidence is not a post-reflow result
A mapped pre-placement image or solder paste inspection (SPI) result can help establish whether paste was present and, depending on the available inspection, describe the deposit. It cannot establish that the final joint coalesced after reflow.
Compare the actual deposit at an affected location with deposits at matched unaffected locations. Consider whether the evidence shows a repeatable print difference or only a perceived difference in the final joint. If the inspection coverage or location mapping is incomplete, that comparison remains incomplete too.
Stencil design can be relevant, but an aperture drawing does not establish what was actually printed. Nor should an aperture ratio be treated as a universal pass/fail rule without the applicable design and product requirements.
Setpoints are not joint temperatures
Oven setpoints describe the programmed environment; they are not measured joint temperatures. A correctly placed thermocouple on a representative populated board records the temperature at its actual attachment location and condition. It does not measure every hidden joint.
If an as-built profile exists, check the board population, attachment integrity and proximity to the locations under investigation. Compare the trace with the paste, component and PCB manufacturers’ applicable temperature and exposure limits. If a relevant profile does not exist, the thermal hypothesis remains unresolved; arrange an authorized measurement if a suitable board and equipment are available.
For measurement setup and attachment evidence, see thermocouple attachment for reflow profiling.
There is no universal soak, ramp, peak, atmosphere or exposure setting for this investigation. A change that helps one joint may create a different problem elsewhere or exceed a material’s limits.
Compare explanations before choosing a change
Two thermal states need particular separation:
- Underheating: the local solder may not have reached the melting conditions needed for the joint.
- Failure to coalesce despite melting: some or all of the solder may have melted, yet the final joint retained separate particle outlines.
The second state is not established merely because a nearby thermocouple trace crosses the paste alloy’s stated liquidus. The hidden joint can differ from the attachment point, and the trace does not directly show what each particle did. Conversely, an unattractive final shape does not prove the solder never melted.
The following table turns appearances into competing checks rather than a list of presumed causes.
| Observed evidence | Competing hypothesis | Discriminating check | Limitation |
|---|---|---|---|
| Retained particle outlines at a small joint | Local solder was underheated; alternatively, it melted but did not fully coalesce | Review a relevant measured profile against the paste alloy information and applicable product limits. If feasible and authorized, use direct in-process observation on a representative setup. | A thermocouple reports its attachment point; observation may not see a hidden production joint. Neither automatically proves the history of every particle. |
| Graping-like appearance at some joints but not matching joints | A location-specific print difference contributed; alternatively, placement or local heating differed | Compare mapped pre-placement deposits, component positions and local surroundings at affected and unaffected locations. | Missing or unmatched images leave the comparison unresolved; a normal-looking print does not prove coalescence. |
| Several affected joints cluster on one board or panel position | Board position or local thermal conditions contributed; alternatively, a shared print or handling event did | Map joint results by board, panel position and run; compare with any relevant measured profiles and print records. | Joints on one board are not independent evidence of a board-level cause. A small cluster may have several explanations. |
| Suspicious joints occur after a paste handling interval | Paste condition or flux behavior may have contributed; alternatively, the interval coincided with another process change | Review product and batch handling records; if authorized, compare with a controlled paste condition using matched locations. | Appearance does not prove oxidation, flux exhaustion or poor paste. A comparison may still be confounded by time or setup. |
| Solder remains off part of a pad or termination | Non-wetting may be present; alternatively, limited solder delivery or placement affected apparent coverage | Compare print and placement evidence; examine the interface or an authorized section if appropriate. | Surface appearance alone may not identify the mechanism or establish acceptance. |
| A nearby profile indicates melting conditions, while the joint remains granular | Local melting and incomplete coalescence are plausible; local underheating remains possible | Review attachment location and condition, then seek a closer representative measurement or direct observation if practical. | A nearby trace cannot certify the hidden joint’s temperature or prove a single cause. |
Particle size and atmosphere may be relevant to a particular approved process, but neither should be changed on appearance alone or assigned a universal setting. The useful next step is the check most likely to distinguish the leading explanations while preserving evidence.
Indium Corporation's Minimizing Graping discusses deposit geometry, oxidation, flux and heat exposure as interacting factors. AIM Solder's W20 SAC305 case study describes a change to both paste and profile; it cannot isolate either factor or validate another assembly. Check the actual paste, component and PCB documentation and project requirements before a trial.
A fictional comparison: count the same locations
The following example is invented to show how to structure a comparison. It is not a customer result, a real test or a claim of yield improvement or causality.
An investigator defines the unit as one named joint location. The baseline uses three boards in panel positions A, B and C; the trial uses three matched boards in the same panel positions. Four corresponding joint locations are assessed on every board, giving 12 locations per group. Each location is classified visually as G, graping-like, or C, no graping-like appearance observed. This classification is not a quality disposition.
| Matched board position | Baseline: locations 1–4 | Baseline G count | Trial: locations 1–4 | Trial G count |
|---|---|---|---|---|
| A | G, G, C, C | 2 of 4 | G, C, C, C | 1 of 4 |
| B | G, C, C, C | 1 of 4 | G, C, C, C | 1 of 4 |
| C | G, G, C, C | 2 of 4 | G, C, C, C | 1 of 4 |
| Total | 5 of 12 | 3 of 12 |
The fictional records identify the same board revision, stencil revision, paste product and batch for both groups. Handling is recorded for both. Mapped pre-placement images show paste present at all 24 assessed locations, but no deposit-volume measurements are available. That supports a presence comparison, not a conclusion that the prints were equivalent.
For the trial, product engineering authorizes one oven-zone setpoint change for investigation; the other recipe settings are held as recorded. A populated measurement board is run for each condition. In both runs, the trace at its documented attachment point passes the paste alloy’s stated liquidus; the trial trace differs from the baseline trace. The applicable paste, part and PCB limits are reviewed for the trial. These traces do not establish the temperature of each of the 24 assessed joints.
The team also checks 12 matched adjacent joint locations per group for visible bridging: 0 of 12 in the baseline and 0 of 12 in the trial. That is a limited visual observation, not evidence that no other defect or reliability tradeoff exists.
The trial has two fewer graping-like appearances: 5 minus 3 equals 2. But the comparison includes only three boards per group, with multiple locations on each board. Board-level correlation, run order, unmeasured print differences and hidden-joint temperatures remain possible influences. The change correlates with a lower count in this fictional sample; it does not prove the sole cause or establish a release decision.
Validate a controlled change
Choose a leading hypothesis and state beforehand what observation would support it, what would weaken it and what would remain unmeasured. For example, if local thermal exposure is the leading concern, define the comparison locations and measurement attachment before adjusting a recipe. If print variation is the leading concern, preserve mapped print evidence before changing the stencil or printer setup.
Use a matched baseline and, where feasible, make one authorized change at a time. Keep track of other differences that cannot be held constant, including paste exposure time, board population, panel position and run order. A single setpoint change can still alter several aspects of the thermal trace; record the resulting profile rather than describing the trial only by the setting changed.
Assess the same named locations and use the same observation method in baseline and trial. Also monitor relevant adjacent defects and potential tradeoffs, such as visible bridging, changes to other joints or possible thermal exposure concerns. A result at one small joint must not be purchased at the expense of an unexamined area of the board.
Repeat or expand the comparison when appropriate and possible, accounting for the fact that joints share boards, panels and process runs. If evidence conflicts or a critical measurement is unavailable, preserve the uncertainty. An improvement after a change is useful evidence, but correlation alone does not establish that the changed factor was the only cause.
Separate investigation, engineering and quality decisions
- Investigator: preserve and organize the evidence, classify observations, compare hypotheses, report missing information and propose discriminating checks.
- Product engineering: decide which measurements or experimental changes are technically appropriate and authorize them within applicable material and product constraints.
- Quality: evaluate the resulting evidence and disposition the product under the applicable customer requirements. Document any release decision and its basis.
Do not treat a visual classification, continuity reading or experimental improvement as an automatic release.
RFQ or investigation handoff
For a PCB assembly discussion or quote, provide what is available and label what is missing:
- Board identity and revision, build identification, panel position, component references and exact joint locations.
- Scaled pre-rework images of affected and unaffected joints, with the intended joint boundaries identified.
- Paste exact product, batch and available storage and handling history.
- Stencil revision, print setup records and mapped actual print or SPI evidence, including inspection coverage.
- As-built profile files or measurements, board population, thermocouple attachment locations and conditions, plus oven recipe and run identification.
- Applicable paste, component and PCB requirements and the customer’s acceptance and disposition requirements.
- Any proposed controlled comparison, relevant neighboring defects, prior rework and known evidence gaps.
FAQ
Is every granular-looking joint solder-paste graping?
No. Retained particle outlines make graping-like coalescence a reasonable description, but residue, reflections or an incomplete view can mislead. Establish where the material is relative to the pad and termination, and distinguish joint material from separate balls or beads.
Does graping prove that the reflow peak was too low?
No. Underheating is one possibility. Solder may also melt and still fail to fully coalesce. Use relevant measured thermal evidence, and recognize that a thermocouple does not report every hidden joint.
Can SPI confirm a good joint after reflow?
No. SPI describes the pre-placement print within its inspection coverage. It can support a comparison of deposits, but post-reflow inspection and any other required evaluation are separate.
Should the oven recipe be changed immediately?
Not solely because a joint looks granular. Preserve evidence, review the applicable material limits and compare the leading hypotheses first. Product engineering can then authorize a controlled change with a matched baseline and monitoring for tradeoffs.
What if there is no profile or usable print record?
Record those gaps and leave the associated hypotheses unresolved. If suitable resources are available, arrange an authorized measurement or comparison rather than treating oven setpoints or stencil design as proof of what happened.
