Our single-sided test board was 7.5 inches wide, 12.5 inches long and 0.062 inch thick. The pads were bare copper covered by an organic solder preservative.

Three different pad widths, lengths and spacings were tested, for a total of 27 different pad designs. Each pad design was replicated 120 times within a single row. Each row was designated by a three-letter code based on the pad dimensions (see Table 1). For example, pad design ADG was 0.012 inch wide and 0.008 inch long with 0.009 inch of space between pads.

Besides pad-to-pad spacing, we also looked at component-to-component spacing. Four distances were examined: 0.008, 0.012, 0.016 and 0.02 inch. To test component spacing, 30 components were blocked together for each pad design. All pad traces were run out through the ends of the pads, which enabled us to test side-to-side spacing, but not end-to-end spacing.

All totaled, a fully populated board contained 12,960 components.

The test board was designed so that half the components were oriented at 90 degrees to the direction of travel into the reflow oven, while the other half were oriented at 0 degrees.

All solder paste was printed using 0.005-inch-thick stainless steel laser- cut stencils. The stencils were not microetched or plated. A thickness of 0.005 inch was chosen as a compromise between 0.004-inch-thick stencils and 0.006-inch-thick stencils. The 0.004-inch-thick stencil would provide better solder paste release for 0201 paste deposits, but would inherently reduce the solder paste volume available for other surface mount devices that are typically found on most applications. A 0.006-inch-thick stencil would produce unacceptable paste deposits for 0201 components.

Two stencils were manufactured for the project. Stencil 1 was designed for the first experiment. Five different apertures were tested for each pad design. Stencil 2 was designed based on the results from stencil 1. Only one aperture size was used for a given pad design for stencil 2.

One no-clean solder paste and one water-soluble paste were used. Both pastes were 90 percent solids containing Type IV powder. The viscosity of the two pastes was approximately 900 kilocentipoise.

A DEK Infinity screen printer was used to print the paste. The printer was equipped with a metal squeegee, and the squeegee angle was 60 degrees. Print speed was 1 ips; squeegee pressure was 2.3 per inch; separation speed was 0.02 ips; and the print gap was 0 on contact.

All components were placed with a Universal 4796R HSP. Automatic component pickup correction in both the X and Y axes was used at all times. This feature improves the reliability of pickups. The height of the Z axis was fully controlled during pickup and placement to increase pick reliability and ensure that components were placed without excessive or inadequate pressure. All components were fed from tape and reel. Two local fiducials were used for board alignment. Component placement caused minimal displacement of solder paste.

Reflow was performed in a Heller 1800W forced-convection oven. The oven contained eight heating zones and one cooling zone. Two reflow atmospheres were tested: air and nitrogen.

Definitions of Defects

The test boards were inspected manually using a coordinate measuring machine (CMM). We used a CMM because it could easily scan each row while providing adequate magnification for inspecting individual components.

Five types of assembly defects were observed.

Tombstones are a severe defect in which components stand on their ends. There are many causes of tombstones. One of the most common is uneven heating between the two ends of the component. When solder on one side of the component achieves liquidus before solder on the other side, the surface tension of the molten solder pulls the component upright. The amount of time it takes solder to wet the joint is also a major variable in the formation of tombstones. Component terminations and pads with poor solderability will increase the probability of tombstones, as will thermal profiles that heat too fast or do not provide enough time for flux activation.

Solder bridging is the undesirable formation of a conductive path of solder between conductors. Often, molten solder draws components towards each other. This is caused by the surface tension exerted by the molten solder between components during reflow. Solder bridging is caused by poor solder paste deposition. Wet paste deposits that are touching between components before reflow have a high probability of bridging. Other causes of solder bridging include: paste displacement during component placement; poor solderability of terminations and pads; incorrect thermal profiles; and paste-related problems. Poor pad design and large solder volumes will also increase the probability of solder bridging.

Solder balls are small spheres of solder that adhere to the laminate, mask or conductors. Solder balls that are generated by the displacement of solder paste during component placement should not be confused with solder balls that are caused by solderability issues with the paste, components, boards, environment or thermal profile. Displacement of paste from solderable surfaces was the primary cause of solder balls produced in our study. Stencil thickness and stencil opening design, coupled with correct pad design, will eliminate all or most solder balls.

The last two most common defects were insufficient or excessive solder volume. We defined insufficient solder volume as any solder fillet that extended less than 50 percent up the face of the component termination. Excessive solder volume was any solder fillet that produced a convex-shaped solder fillet. Inconsistent paste deposits are the primary cause of these defects. Stencil thickness, stencil aperture design, stencil manufacturing process, print process parameters, powder size and paste viscosity are the main attributes that influence paste deposition.

Results

We performed two experiments. The first experiment examined four processes: no-clean solder paste reflowed in air or nitrogen, and water-soluble solder paste reflowed in air or nitrogen. Six boards were assembled for each process, for a total of 311,040 components. Five stencil aperture sizes and aperture positions were tested for each pad size.

In the second experiment, we ran only three of the four processes. We dropped the test of water-soluble solder paste in nitrogen, because that flux chemistry is rarely reflowed in a nitrogen environment. Only one stencil aperture design was run per pad design. The aperture design was selected based on assembly yield and quality from the first experiment.

The widest spacing between pads-0.015 inch-was also dropped from the second experiment. This reduced the total number of different pad designs from 27 to 18. Data from experiment 1 showed that a pad spacing of 0.015 inch produced more open solder joints than smaller pitches. A total of 50 boards were assembled for each of the three processes for a total of 1,116,000 components.

Of the three processes, the no-clean solder paste reflowed in air produced the fewest defects, 66. The water-soluble paste reflowed in air produced the next lowest number of defects, 1,499. The no-clean paste reflowed in nitrogen produced the greatest number of defects, 5,665.

Solder bridging varied by paste and reflow atmosphere. Water-soluble paste reflowed in air produced the lowest percentage of solder bridges at 7 percent, followed by the no-clean paste reflowed in nitrogen at 15 percent. No-clean paste reflowed in air produced the largest percentage of bridges at 21 percent.

Solder bridging was directly related to component spacing for the three assembly processes. No bridging was recorded for any of the processes at a spacing of 0.012 inch. No-clean paste reflowed in air produced the fewest solder bridges, 14. Water-soluble paste reflowed in air produced the next largest number of bridges, 99. No-clean paste reflowed in nitrogen produced the greatest number of bridges, 866.

Twelve pad designs out of 18 did not produce any bridges at the smallest component spacing of 0.008 inch for the no-clean paste reflowed in air. Ten pad designs out of 18 did not produce any bridges at the smallest spacing of 0.008 inch for the water-soluble paste reflowed in air. Six pad designs out of 18 did not produce any bridges at the smallest spacing of 0.008 inch for the no-clean paste reflowed in nitrogen.

Pad design AEG, which resulted in the largest distance between paste deposits of 0.016 inch, produced the fewest solder beads. Solder beads are reduced when the distance between paste deposits is increased. The amount of paste displaced by the component during placement is reduced when the distance between paste deposits is increased.

An analysis of paired samples was done to determine if component orientation significantly influenced assembly yield. For no-clean paste reflowed in air, we found no significant difference in yield with component orientation. The lower flux activity of no-clean paste when reflowed in air does not increase the risk of tombstones. In contrast, the higher flux activity of water-soluble paste produced a significant increase in tombstones for parts that were oriented at 90 degrees to the heat source. Similarly, nitrogen increased the number of tombstones among parts that were oriented at 90 degrees to the heat source. The vast majority of open joints were on the component termination that was reflowed second-the trailing termination. Nitrogen increased the wetting speed of the molten solder and thus produced open joints at a significantly higher rate for parts oriented at 90 degrees compared with those oriented at 0 degree.

Seven of the 18 pad designs (BDH, BEG, BFG, BFH, CDH, CEH and CFH) did not produce any defects. However, based on pad size, printing difficulty, and the quality of solder joint shapes, we prefer designs BEG and CEH. The smallest pad designs require small stencil apertures that clog faster than larger apertures. Stencils that are 0.004 inch thick will reduce clogging, but other surface mount devices that require more solder may receive insufficient paste. Moreover, the smallest pad designs did not produce the desired concave-shaped solder fillet. The largest pad designs are good for paste release from the aperture and also produce acceptable fillet shapes. However, the larger pad designs require more space on the circuit board.

Water-soluble paste reflowed in air produced defects on all pad combinations with either component orientation. Pad CEG produced the fewest assembly defects. Pad CDH did not produce any defects in the 0 degree orientation, but did produce a relatively high number of defects in the 90-degree orientation. Pad CEG produced good solder joint shapes and does not occupy as much board space as larger pad designs. Paste clogging the apertures was not a problem with the CEG pad.

No-clean paste reflowed in nitrogen produced defects on all pad combinations and component orientations. Pad CEG produced the fewest defects with this process.

We also analyzed defects by pad width. These data were generated based on the optimal pad designs with respect to each process. Pad length and spacing were held constant, while pad width was varied across all experimental levels. For the three assembly processes, yield improved as pad width increased. For both water-soluble paste reflowed in air and no-clean paste reflowed in nitrogen, the fewest defects were produced at a pad width of 0.018 inch.

This trend changes slightly for no-clean paste reflowed in air, where the best yield was actually produced at a pad width of 0.015 inch. However, due to the limited number of defects found across the boards built with this process, the difference in defect levels between pad widths of 0.015 inch and 0.018 inch was not statistically significant. The yield produced by no-clean paste reflowed in air is least sensitive to pad width, while the no-clean paste reflowed in nitrogen was the most sensitive to pad width.

Pad length was analyzed in the same way. Pad width and spacing were held constant, while pad length was varied across all experimental levels. Our results indicate that the optimal pad length was 0.012 inch for all three assembly processes. The no-clean paste reflowed in nitrogen was more sensitive to pad length than any other process. No defects were observed on any boards assembled with no-clean paste in air at pad lengths of 0.012 inch and 0.016 inch.

When pad width and length were held constant, and pad spacing was analyzed, all three processes yielded similar results, with more solder joint failures occurring at a pad spacing of 0.012 inch. No-clean paste reflowed in nitrogen was the most sensitive process to changes in pad spacing. No-clean paste reflowed in air was the most resistant to defects attributed to pad spacing.

In summary, of the three processes we tested, no-clean solder paste reflowed in air produced the fewest number of tombstones and solder bridges. This process also yielded the most pad designs that were free from defects. This process was the least likely to produce defects across a variety of pad designs.