Circuits Assembly Online Magazine - Correlating 2-D and 3-D SPI

News

As pads get smaller, the correlation between the various area measurements weakens.

Solder paste inspection systems can be classified into two major categories: in-printer inspection and end-of-process inspection. This inspection system can be broken down to 2-D (area coverage) or 3-D (height/volume coverage) inspection. Both processes have advantages and disadvantages. The area coverage data (2-D) can be obtained from true X-Y paste coverage information or through a height map from a 3-D inspection. A literature survey shows little information available in the public domain that correlates a true 2-D area measurement with an area measurement through height-generated data. A recent study by Speedline and CyberOptics¹ details discussion of both methods. This column excerpts those findings.

Solder paste area measurement is highly influenced by the technique used to collect the solder paste deposit image. Since solder paste deposits are rarely shaped like perfect cylinders or cubes, area measurements based on 2-D images typically will be slightly larger than area measurements based on 3-D images. Likewise, since 3-D area measurements represent an area of a cross-section of the deposit, 3-D area measurements typically will be smaller than 2-D measurements. Could they be correlated in a sound statistical manner that will provide a reliable prediction model?

2-D inspection techniques are capable of providing transfer area information only, while 3-D inspection techniques provide additional information such as height and volume transfer. While the volumetric transfer is considered to be more informative, there is not a well-known supply of data comparing and contrasting the two techniques.

Our study attempted to correlate the area transfer data obtained from a true 2-D measurement technique with the area measurement obtained from a 3-D height map through a series of controlled experiments. In addition, we attempted to correlate area measurement with volume measurement to provide a tool to determine when area measurement can be used to monitor and control the printing process.

Experimental methodology. The test vehicle was a 10" x 8" x 0.062", four-layer FR-4 board with ENIG surface finish. It included a range of commercially available components and packages: BGA 225, BGA 256, PLCC 28, QFP 160, QFP 208, R 0603, R 1206 and R 1201. An MPM Accela printer was used to conduct all printing and in-printer 2-D area inspection. CyberOptics SE300 SPI system was used to inspect the boards offline to provide 3-D area map, height and volume data. Four boards were printed using a 0.003" Efab stencil and Type 4 solder paste. To reduce print direction noise, both forward and reverse squeegee strokes were incorporated into this experiment. Prior to the experiment, a Gage study was conducted on both the printer inspection system and SPI to ensure measurement repeatability.

Results. Based on the R2 value, which is the square of the correlation coefficient, it can be concluded there is a weak correlation between 2-D and 3-D area measurement. Usually, a strong correlation exists when R2 > 90%. With larger pad sizes, we see the correlation is stronger. We also observed the correlation strongly depended not only on pad size, but also aperture shape. However, the 3-D to 2-D area data ratio can change depending on shape; this will be a topic of further study.

As predicted, there is an inevitable bias between 2-D and 3-D area measurements, with 2-D consistently measuring a slightly larger area. It should be noted that, for very small pads, the results showed a different trend. As pads get smaller, other effects such as pixel quantization, edge irregularities and the actual size of the solder balls at the edge of the pad begin to affect measurement. We have seen evidence of this phenomenon for much smaller pad size and will report on it in the future.

The correlation between 2-D area and volume appears to be stronger than the 2-D area versus 3-D area correlation.

Some bimodal structure in the correlegram showed one board possibly having a different print characteristic, a possible investigation for future work.

References