SMT assembly defects - where they originate and which inspection catches them

Published:

A board passes solder paste inspection without a flag. After reflow, AOI reports a bridge between two leads of a QFP. The print matched the program, so the printer is not at fault. The cause sits further upstream: the BOM now calls for an alternate part whose lead foot is shorter than the one the pad was designed for. This scenario, described in a paper presented at IPC APEX EXPO 2017, is a good way to read quality control on an SMT line. Each inspection method sees a specific class of defects, and the point where a defect is detected is rarely the point where it was created.

This article maps the defects to the inspection stages: what SPI, AOI and X-ray can detect, what they cannot, and where to look for the root cause.

Paste printing gets most of the blame, but SPI only judges the deposit

Greg Smith of FCT Assembly, a stencil and solder materials manufacturer, opens his IPC APEX EXPO 2017 paper with: "It is commonly accepted that the solder paste printing process causes the highest percentage of yield challenges in the SMT assembly process". The industry also circulates specific percentages of defects attributed to printing. We could not trace them to an original source, so we do not repeat them here.

3D solder paste inspection (SPI) measures the volume and position of every deposit right after printing. It is the earliest point at which the line can stop a board before components are placed on it. How much paste stays on the pad depends largely on the stencil aperture geometry. IPC-7525C (November 2021) defines the area ratio as the ratio of the aperture opening area to the area of the aperture walls. In Smith's experiment, measured on a production 3D SPI system, transfer efficiency rose linearly with area ratio. He found it a better predictor of paste release than metrics that also account for copper thickness and pad finish.

The widely quoted 0.66 threshold comes from revision B of the guideline (2011), which Smith cites. We have not verified what revision C says on this point. What is worth knowing is how IPC-7525C describes itself: as a document "intended as a guideline only", because "printing performance depends on many different variables and therefore no single set of design rules can be established". An area ratio threshold is a starting point for your own SPI measurements, not an acceptance criterion. Smith says so directly: manufacturers should compare their own SPI results against area ratio to decide what minimum transfer efficiency produces acceptable joints.

SPI has a limit built into its principle. The machine compares each deposit with a program generated from the paste layer. If the paste layer is reproduced correctly on the stencil but does not suit the component that actually lands on the board, SPI reports a good print.

A bridge after reflow that was absent after printing starts with the pad-to-lead relationship

Smith describes this mechanism for gull-wing leads, as on QFP packages. A lead exposed to convection in the oven heats faster than the paste on the pad. If the foot is much shorter than the pad, the paste melts first at the foot and wicks up the lead. Once it cannot climb any further, molten solder pools around the foot before the rest of the pad has heated. At fine pitch, that pool spreads to the neighbouring lead.

Smith gives a concrete stencil correction rule. When the foot is 25% shorter than the pad, reduce the printed paste volume by 10%. At 50% shorter, reduce by 25%, and at 70% shorter, by 40%. The reduced aperture must be centred on the foot, not on the pad. This is the author's recommendation from his experience as a stencil designer, not a requirement of any standard.

The lesson for purchasing and design engineering is less obvious than the rule itself. Smith points to a common cause: the pad was designed for one component, which was later replaced with a similar part with a shorter foot. An alternate in the BOM, even one that is electrically equivalent, can therefore require a stencil change. SPI will not catch it, and AOI only sees the result.

Tombstoning on small passives is usually a geometry and temperature problem, not a placement problem

Tony Lentz and Greg Smith covered the whole family of component-movement defects at SMTA International 2021: tombstoning, draw-bridging, bill-boarding and skewing. They mainly affect small passives, and the authors name 1206, 0805 and 0603 as examples. The common mechanism is unequal wetting forces of molten solder at the two ends of the part, amplified by temperature differences between the pads and by placement errors.

Most of the causes the authors describe lie outside the pick-and-place machine:

  • One pad set for two package sizes, for example 0805 and 0603. The pads can then be too close together for the larger part and too far apart for the smaller one.
  • A pad at one end of the part connected to a ground plane. The plane draws heat away, so the paste on the other, ordinary pad melts first and can lift or skew the part.
  • A stencil designed to the pads instead of the terminations. When the pads are spaced wider than the terminations, the paste lands beside them. When the terminations sit on the inner side of the deposits, the risk of tombstoning and draw-bridging rises. When they sit outside the centre of the deposits, skew and mid-chip solder balls appear.

As remedies, Lentz and Smith list anti-tombstoning paste, stencil modification and adding a soak to the reflow profile, often in combination. In his 2017 paper Smith recommends a reverse U-shape aperture that removes paste from the outer part of the pad and so lowers the wetting force there. He also stresses a precondition for the analysis: you need the copper and solder mask layers from the Gerber files, not just the paste layer.

AOI detects these defects easily because they are visible. It cannot tell you which cause was at work. The authors also note that these defects can usually be reworked and do not end up as scrap, but rework costs money and delays delivery.

AOI can only judge joints that have a visible surface

Packages with terminations under the body are a separate category. IPC-7093B, published in January 2026 and superseding revision A from October 2020, covers bottom termination components (BTC) such as SON, DFN, QFN and LGA, "with or without wettable side terminations or flanks". It defines a wettable side flank as a perimeter termination partially plated to improve solderability.

This distinction sets the limit of AOI. In its chapter on production inspection, the IPC-7093B table of contents lists four stages: SPI, AOI, automated X-ray inspection (AXI) and visual inspection. The role of AOI is illustrated with the assessment of side flank wetting, while the detection of missing solder is shown on X-ray images. A joint under the body, including the thermal pad, stays invisible to the camera. On a BGA no joint is visible from above, and the outer row of balls can at best be viewed from the side.

The practical conclusion for designers: choosing a BTC package with wettable flanks gives AOI a surface it can assess. Without it, judging those joints falls entirely on X-ray.

The status of this document matters too. Revision A was titled "Design and Assembly Process Implementation…", revision B is "Design and Assembly Process Guidance…". Both revisions refer acceptance criteria to J-STD-001 and IPC-A-610 in their scope. IPC-7093B tells you how to inspect BTC joints, but not which ones to reject.

X-ray under a BGA does not settle every defect

The hardest case is head-on-pillow (HoP), often also called head-in-pillow. IPC-7095E (August 2024) defines it as a solder joint comprised of two metallurgically distinct masses formed from a BGA ball and reflowed solder paste, with incomplete or no coalescence. Lars Bruno and Benny Gustafson of Ericsson identify the cause as dynamic warpage of the package in the oven, especially with large BGAs and thin fine-pitch packages. The ball loses contact with the paste and flux for part of the process.

Their study, published in the Journal of Microelectronics and Electronic Packaging in 2019, shows the limits of X-ray clearly. The authors deliberately induced head-on-pillow, assessed the joints on X-ray images, then pulled the components and compared the results. Many joints classified as head-on-pillow on X-ray showed no trace of it after pulling. In their conclusions they state that this defect, a necked joint and a cracked joint cannot easily be told apart on an X-ray image. They recommended that the next revision of IPC-A-610 should treat any necking in a BGA joint as a defect.

An X-ray result pointing to head-on-pillow is therefore a trigger for further analysis, not a verdict. Confirmation requires a destructive method. Among its referenced documents IPC-7095E lists IPC-TM-650 method 2.4.53, the dye and pull test.

The second limit concerns criteria. IPC-A-610J (March 2024) states in clause 1.1: "This standard does not provide criteria for cross-section evaluation or x-ray inspection. There are x-ray guidelines located in J-STD-001." In the current J-STD-001J this is Appendix C, "X-Ray Guidelines". It is therefore best to agree the X-ray image acceptance criteria in the contract, rather than assume they follow directly from the acceptance standard. Voiding criteria are covered in our article on voids in BGA solder joints.

Moisture damages components where no line inspection looks

The damage happens in the oven, but its cause lies in storage and at the feeder setup station. IPC/JEDEC J-STD-020F (December 2022) describes the mechanism in its introduction. The vapour pressure of moisture inside a non-hermetic package rises sharply at reflow temperature. The result can be internal delamination, internal cracks that may or may not reach the outside of the package, and bond damage. In the most severe case the package cracks externally with an audible pop, hence the name popcorning. The standard notes that SMDs are more susceptible than through-hole parts: they see higher temperatures in reflow because they are soldered on the same side of the board they are mounted on.

To detect delamination, J-STD-020F uses acoustic microscopy. That is a laboratory method, not one of the inspections run on an assembly line. AOI will at most see a crack that has reached the package surface. Internal damage can pass the whole inspection chain unnoticed.

The defence is procedural. J-STD-020F assigns a component its moisture sensitivity level (MSL), and IPC/JEDEC J-STD-033D (April 2018) specifies how to handle it: moisture barrier bags, desiccant, humidity indicator cards, the allowed floor life on the shop floor, and baking parts that have exceeded it. The effectiveness of this system is checked by auditing the procedures, not by looking at a camera image.

Which inspection sees which defect

The table sums up the cases above. It is not a complete catalogue of SMT defects, only examples that show the limits of each method.

Defect Where it usually originates What detects it What this inspection cannot settle
Insufficient, excess or offset paste Printing: stencil aperture geometry, printer settings SPI Whether the programmed volume suits the component that will land on the pad
Bridge on gull-wing leads that was absent after printing Foot length versus pad length, for example after a switch to an alternate part AOI after reflow SPI passes the board because the print matched the program
Tombstoning, draw-bridging, skew Pad geometry, stencil, pad temperature difference, placement, profile AOI after reflow Which of the causes was at work
Missing solder under a BTC or thermal pad Paste printing, pad and stencil design X-ray (AXI) AOI can at most judge a wetted side flank
Head-on-pillow under a BGA Package warpage in the oven X-ray, with limitations Distinguishing it from necking and cracking. Confirmation needs a destructive method
Delamination and cracks inside the package Storage and exposure time of moisture-sensitive components Acoustic microscopy in a laboratory Line inspections see an external crack at most

The question to ask an EMS provider: where does defect information go back to

A list of inspection machines says less about assembly quality than it seems. Each machine stops a defect at a different point, and several of the cases above have their cause in the PCB layout, the stencil or the stores. When choosing a supplier or reviewing a process, ask about things that do not appear in a line specification:

  • Does the AOI result feed back into stencil design? A recurring tombstone or post-reflow bridge should end with an aperture correction. Reworking boards removes the effect and leaves the cause in place.
  • Does switching to an alternate part trigger a stencil review? The QFP bridge example shows that an alternate with a different foot is a process change, not just a purchasing decision.
  • How are the X-ray image acceptance criteria agreed? The current IPC-A-610J does not contain them, so they should come from the contract or a referenced document.
  • How are moisture-sensitive components controlled? This means tracking floor life and baking in line with J-STD-033D.
  • Does the DFM review cover the copper and mask layers, not just the paste layer? Without them the pad-to-lead relationship cannot be assessed.

At EAE Elektronik we run SMT assembly with solder paste inspection (SPI) and 3D AOI. We check joints hidden under packages on a General Electric Microme|x X-ray system, which can view the board at angles of up to 70°. If you have a design in which one of these defects keeps recurring, or you want the layout reviewed for assembly before production, send us your documentation for a quote.

Let's talk!

Take the first step! Write or call and let's talk about your project.
We will present you our offer and choose the optimal solution.

contact us
eae elektronik montaż smt