A recent assembly case involved solder paste loss on the signal pads of a QFN device. The defect rate was approximately 75%.
Figure 1 | Defect condition
X-ray inspection confirmed that, after the solder paste melted, it was pulled toward the central thermal pad. As a result, the signal pad could not form a reliable solder joint.
In this discussion, the surrounding small pads are referred to as signal pads, while the large central pad is referred to as the thermal pad.
The component condition before soldering and the corresponding local enlargement are shown below.
Figure 2 | Defect location before soldering and local enlargement
1. Root-Cause Analysis
The component has 48 pins, but the problem occurred only on the grounded signal pad. This signal pad and the thermal pad belong to the same net and are connected by a copper trace. The other signal pads are not connected to the thermal pad by traces.
The connecting trace reduces the physical separation between the signal pad and the thermal pad. When the solder paste melts, the two solderable areas can come into contact through the reduced gap. Because the two pads have significantly different areas and therefore different surface-tension behavior during reflow, solder on the signal pad is pulled toward the thermal pad. The signal pad consequently fails to form a reliable solder joint.
1.1 Effect of the Connecting Trace on Pad Spacing
A local enlargement of the red ellipse in Figure 2 shows the difference between the two conditions. Without a connecting trace, the distance between the signal pad and the thermal pad is equal to the solder-mask dam plus twice the solder-mask opening width. With a connecting trace, the effective pad-to-pad spacing is reduced to the solder-mask dam width.
Figure 3 | Local enlargement of the red ellipse in Figure 2
The relevant pad and component-library parameters are as follows:
- The nominal inner spacing between the thermal pad and the signal pad is 0.3 mm.
- The solder-mask opening extends 0.075 mm beyond the pad on each side.
- The solder-mask dam width is 0.15 mm.
Without a connecting trace, the pad spacing is 0.3 mm, which means that the actual spacing is the same as the designed spacing, as shown on the left side of Figure 4.
With a connecting trace, the pad spacing is 0.15 mm. The trace and solder-mask opening effectively extend the pad along the trace direction. After this extension, the pad becomes continuously connected to the solder-mask dam. The effective pad spacing is therefore reduced to the solder-mask dam width, as shown on the right side of Figure 4.
The pad, component-library, and PCB designs complied with general design practices and did not contain an obvious design-rule violation. The issue resulted from the interaction between the design geometry and the subsequent manufacturing data-processing steps.
1.2 Effect of Manufacturing Data Processing
Design data cannot be used directly for PCB fabrication. Before production, the data must undergo engineering processing based on equipment requirements, process capability, and manufacturing compensation rules. The original geometry may be enlarged, reduced, or otherwise adjusted. The exact compensation values vary among suppliers.
For supplier A, the pad compensation was 2 mil per side, while the solder-mask clearance from the pad was 2.5 mil. These values were provided by the supplier.
After this processing, the pad spacing changed as follows:
- Without a connecting trace: approximately 0.2 to 0.3 mm.
- With a connecting trace: approximately 0.08 mm.
After compensation, the solder-mask opening was approximately 4.5 mil, or about 0.11 mm, larger than the original pad on each side. The resulting solder-mask dam width was therefore:
0.3 ? 2 × 0.11 = 0.08 mm
This 0.08 mm value became the effective spacing between the signal pad and the thermal pad.
1.3 Summary of the Geometry Change
- In the original design, the inner spacing between the thermal pad and signal pad was 0.3 mm.
- After the two pads were connected by a trace, the inner spacing was reduced to 0.15 mm.
- After manufacturing data processing, the effective inner spacing was further reduced to approximately 0.08 mm.
2. Technical Conclusion
The primary cause of the soldering defect was excessive reduction of the effective pad spacing. The signal pad and thermal pad also differed substantially in size. When the solder paste melted, the difference in surface-tension behavior caused solder to move from the smaller signal pad toward the larger thermal pad, preventing a reliable joint from forming.
As with most manufacturing defects, the failure was not necessarily caused by a single factor. However, the contributing factors have different levels of importance. Once the primary cause is eliminated, the remaining secondary factors may no longer be sufficient to trigger the defect.
3. Corrective Action
The thermal pad was changed from a NSMD structure to a SMD structure. The solder-mask opening was pulled back by 0.2 mm on each side.
An SMD pad structure is not a perfect solution. Solder-mask thickness is more difficult to control and may have some influence on component coplanarity. Nevertheless, the change represents a practical trade-off: it addresses the more serious risk of solder being drawn away from the signal pad.
4. Incorporating the Corrective Action into Design Rules
The corrective action was incorporated into the design guidelines so that the lesson from this failure could be applied to future products.
Before the change, the design guideline required the inner spacing between the thermal pad and signal pad to be at least 0.2 mm, with NSMD thermal-pad construction preferred.
After the change, the guideline specified the following:
- The inner spacing between the signal pad and thermal pad should be between 0.2 and 0.3 mm: 0.2 ≤ d ≤ 0.3 mm.
- An SMD thermal pad is preferred.
- The solder-mask opening should be pulled back from the inner edge of the pad by 0.1 to 0.2 mm.
Connecting traces between adjacent pads are a routine PCB design practice and would not normally be expected to affect downstream assembly. However, this case demonstrates that their influence can become significant when pad size, solder-mask geometry, supplier compensation, and molten-solder behavior interact.
5. Additional BGA Design Implications
The effect is not limited to QFN components. In another BGA design example, traces connected between pads caused substantial distortion of pad size and shape.
Figure 4 | BGA pad-shape distortion caused by connecting traces
- The pads marked in green have no connecting traces and retain a regular circular shape.
- The pads marked in red have connecting traces and become irregularly shaped.
- IPC-7095C specifies a connecting trace width of no more than 0.2 mm for NSMD pads and limits the number of connecting traces to fewer than two.
- IPC-7095D specifies a connecting trace width of no more than 0.2 mm for NSMD pads and removes the requirement concerning the number of connecting traces.
6. Lessons from the Case
A design can comply with conventional rules and still encounter a serious quality problem during production. This type of incident is only one of the many issues that can emerge in a manufacturing process. A complete manufacturing system, including personnel, equipment, materials, methods, and environment, generates large amounts of process data every day.
During defect analysis and corrective action, valuable information is accumulated through inspection results, process adjustments, engineering judgments, and manufacturing experience. Much of this information, however, remains scattered across computers, documents, or individual engineers' knowledge. Without a systematic method for collecting and applying it, useful experience can eventually disappear into the overall flow of manufacturing data.
The concept of manufacturing feeding back into design, also known as reverse engineering in this context, describes a continuous iteration process in which manufacturing experience drives design optimization. Its purpose is to use design defects, process limitations, and quality problems discovered during production to improve and refine future product designs.
Manufacturing experience is not limited to individual know-how. It is the accumulated result of observing, analyzing, and consolidating data generated by the entire manufacturing system. The essential idea is to replace a one-way design-to-production process with a two-way feedback loop. Practical manufacturing experience then becomes an important input to the next design iteration.
Understanding this concept is relatively easy. The more difficult task is turning it into a repeatable working method.
Design and manufacturing are often separated organizationally. After completing the drawings, the design team may transfer them to manufacturing and consider the work complete. When production discovers a problem, the manufacturing team may rely on a temporary process adjustment to keep production moving. If the underlying problem data is not fed back, or is communicated only through an isolated, case-by-case message, it cannot be systematically analyzed and converted into a design rule.
As a result, similar problems can recur across different products. In that situation, "manufacturing feedback" remains only a concept rather than an effective engineering mechanism.
For this reason, defect information should be captured in a form that design teams can use. A useful feedback process should connect the observed defect with its physical cause, the affected design feature, the manufacturing condition, the corrective action, and the revised design guideline. In the QFN case, the key lesson was not simply to change one pad. It was to recognize that a connecting trace could alter the effective pad geometry after solder-mask and PCB data compensation.
Data is a resource in modern manufacturing. Broad access to reliable process data, combined with the ability to convert that data into engineering improvements, allows manufacturing experience to become a continuing source of design quality and process capability.
