STENCIL AND SOLDER PASTE INSPECTION EVALUATION FOR MINIATURIZED SMT - - PDF document

Originally published in the Proceedings of SMTA International, Ft. Worth, TX, October, 2013

STENCIL AND SOLDER PASTE INSPECTION EVALUATION FOR MINIATURIZED SMT COMPONENTS

Robert Farrell Benchmark Technologies Nashua, NH USA Chrys Shea Shea Engineering Services Burlington, NJ USA

ABSTRACT New stencil materials and manufacturing technologies promise improved performance over traditional options. To determine the best technologies for a contract electronics manufacturer assembling miniaturized SMT components, a test was designed to assess the performance of 11 different stencils, submitted by 3 different suppliers, using a variety of materials and

• coatings. Performance metrics include print volume

repeatability and transfer efficiency for 0.5mm and 0.4mm pitch BGAs and for 0201 components with area ratios in the challenging range from 0.5 to 0.66. The test also evaluated two methods of automated solder paste inspection to determine the accuracy and repeatability of the different machines and technologies at feature sizes of 10 mil or less. Setup, execution and results of both sets of tests are presented and discussed. INTRODUCTION SMT stencil printing technology continually evolves to keep pace with device miniaturization technologies. Printed Circuit Board (PCB) assemblers have numerous new technology options to choose from, and need to determine the most effective ones to produce the highest quality and most reliable solder interconnections. The objective of these tests was to identify the best stencil technology for high volume production of miniaturized SMT components. The testing grew to include comparison of Type 4 and Type 5 solder pastes, evaluation of SPI accuracy and assessment of aperture wall topographies. The SPI evaluation portion of this study used three different SPI machines from 2 different suppliers. 2 of the machines were from a single supplier but were different models and ages. The other machine was from another supplier. EXPERIMENTAL SETUP Test Vehicle The test vehicle shown in figure 1 was designed in- house for a multitude of PCB assembly tests, including new packages, pad designs, solder paste print performance and process evaluation tests. The devices selected for analysis in these tests included 0.3, 0.4 and .5mm BGAs and 0201s. Their area ratios ranged from 0.5 to 0.75. Locations and names of the specific devices used in the stencil analysis are shown in figure 2. Figure 1. Test Vehicle Figure 2. Features used in stencil analysis Test Design The stencil analysis included:  3 different stencil suppliers  3 different foil materials  3 different manufacturing processes

Originally published in the Proceedings of SMTA International, Ft. Worth, TX, October, 2013

 2 different nanocoatings The experimental design was not a full factorial. Each supplier provided stencils using technologies that were either their top performers

• r

developmental technologies that they wanted to learn more about. Three to five stencils were submitted by each supplier. A total of 11 stencils were print tested. All were created using the same Gerber file, and all were specified at 0.0040” thick. The final test matrix is shown in table 1. Table 1. Stencil test matrix. The three different vendors are designated A, B and C. Abbreviations for the materials in table 1 are as follows:  FG: Datum Fine Grain 301 Stainless Steel (SS)  PhD: Datum 304 SS  FG, Ni Plated: Ni over FG  Laser Ni: Electroformed Ni that has been laser cut  E-form: Electroformed Ni  Nano2: DEK NanoProtek, 2-part wipe-on coating  Nano1: Laserjob thermally cured nanocoating During this first phase of the experiment, all 11 stencils were printed over a two-day span at Benchmark’s Nashua, NH facility on DEK 265 stencil printers with manually placed pin supports, by the same operator. A no-clean, Type 4, SAC305 solder paste was used. The underside of the stencil was wiped between each of ten consecutive prints. PCBs were serialized and printed in the same order for each run. They were cleaned in an in-line, automatic board washer and inspected before each test. In addition to the print tests, an additional, abbreviated leg was added to the DOE that compared the print performance of Type 4 and Type 5 solder pastes. As the experiment progressed, an evaluation of SPI accuracy was planned and executed and an analysis of the aperture wall topography were added. They will be described in a later section. MEASUREMENTS AND METRICS The performance of the stencils was measured using basic print measurement and evaluation techniques. All the paste deposits were measured using a new SPI system that promised improved accuracy over existing

• technology. It is described in a later section as Machine

#2. The ten-print tests produced 1000 data points for each of the 100 I/O device types on each stencil, and 5470 data points for the 547 I/O device. The measured volumes for each device were used to calculate average volumes and Coefficients of Variation (CV = std deviation divided by mean, %) in a spreadsheet. Whereas the average print volumes provide information

• n the stencil’s paste release characteristics, the CVs

provide comparative data on volume repeatability. Aperture sizes were measured with an Acugage model HPL 25-3-LT4. For each device on each stencil, 5 apertures were measured on both the squeegee and PCB side of the stencil. The five measurements from each side were averaged and used in the aperture volume and area ratio calculations. After completing the print tests, measurement coupons were laser cut outside the four corners of the print area. A micrometer was used to measure the thickness of each coupon; the average was used in aperture volume and area ratio calculations. To account for any trapezoidal stencil wall geometries, aperture volumes and area ratios were calculated using formulas for truncated cones and pyramids. Transfer efficiencies (TEs) are calculated as the average solder paste deposit volume divided by the aperture volume that was calculated from the aperture

• measurements. They represent the percentage of solder

paste that was released from the aperture. They are calculated for each of the 5 devices shown in figure 2 for each stencil tested. Area ratios (ARs) are calculated as the area of the stencil apertures’ circuit side opening divided by the area of the apertures’ walls. These were also calculated for each of the 5 devices on each stencil.

Originally published in the Proceedings of SMTA International, Ft. Worth, TX, October, 2013

Plotting TE against AR provides a means of ranking stencil release performance. Detailed descriptions and discussion of the relationship between TE and AR can be found in reference 1. RESULTS & DISCUSSION Print Volumes The average print volumes, standard deviations and coefficients of variation for each device type and stencil are shown in table 2. Stencil B4 was missing the apertures for the 0.3mm device, so no measurements were taken. Table 2. Solder paste deposit volumes (in cubic mils) and CVs. A typical guideline for print volume repeatability is a CV of 10% or less, which was achieved in most cases, except for the 0.3mm BGA. The 7mil aperture is too small for a 4mil foil and a Type 4 paste (AR=0.43), but the feature was analyzed because it pushed the limits of both the printing and measurement processes.

Stencil Component Type Avg of Vol StdDev of Vol CV% A1 0201 A Horizontal 382 25 6.4% A1 100BGA 3Ptich SMD 141 25 17.4% A1 100BGA 4Pitch NSMD 229 27 11.9% A1 100BGA 5Pitch NSMD 421 30 7.2% A1 547BGA 4Pitch SMD 245 23 9.3% A2 0201 A Horizontal 393 24 6.1% A2 100BGA 3Ptich SMD 151 22 14.7% A2 100BGA 4Pitch NSMD 243 24 9.7% A2 100BGA 5Pitch NSMD 438 31 7.1% A2 547BGA 4Pitch SMD 246 23 9.5% A3 0201 A Horizontal 394 24 6.2% A3 100BGA 3Ptich SMD 124 28 22.4% A3 100BGA 4Pitch NSMD 214 34 16.1% A3 100BGA 5Pitch NSMD 413 28 6.7% A3 547BGA 4Pitch SMD 235 24 10.2% A4 0201 A Horizontal 320 42 13.2% A4 100BGA 3Ptich SMD 122 16 12.8% A4 100BGA 4Pitch NSMD 183 19 10.5% A4 100BGA 5Pitch NSMD 329 25 7.7% A4 547BGA 4Pitch SMD 187 18 9.8% B1 0201 A Horizontal 377 25 6.6% B1 100BGA 3Ptich SMD 136 24 17.7% B1 100BGA 4Pitch NSMD 222 29 13.1% B1 100BGA 5Pitch NSMD 422 30 7.1% B1 547BGA 4Pitch SMD 251 24 9.7% B2 0201 A Horizontal 388 25 6.4% B2 100BGA 3Ptich SMD 140 25 18.2% B2 100BGA 4Pitch NSMD 228 28 12.3% B2 100BGA 5Pitch NSMD 424 30 7.0% B2 547BGA 4Pitch SMD 247 25 10.3% B3 0201 A Horizontal 385 26 6.7% B3 100BGA 3Ptich SMD 144 24 16.5% B3 100BGA 4Pitch NSMD 229 27 11.8% B3 100BGA 5Pitch NSMD 425 30 7.0% B3 547BGA 4Pitch SMD 247 24 9.6% B4 0201 A Horizontal 348 21 6.1% B4 100BGA 3Ptich SMD no data no data B4 100BGA 4Pitch NSMD 217 22 10.0% B4 100BGA 5Pitch NSMD 402 27 6.7% B4 547BGA 4Pitch SMD 209 20 9.8% B5 0201 A Horizontal 378 25 6.7% B5 100BGA 3Ptich SMD 101 23 22.7% B5 100BGA 4Pitch NSMD 204 31 15.0% B5 100BGA 5Pitch NSMD 395 25 6.2% B5 547BGA 4Pitch SMD 205 29 14.1% C1 0201 A Horizontal 399 24 5.9% C1 100BGA 3Ptich SMD 159 24 15.2% C1 100BGA 4Pitch NSMD 253 24 9.3% C1 100BGA 5Pitch NSMD 454 27 6.0% C1 547BGA 4Pitch SMD 257 26 10.1% C2 0201 A Horizontal 416 21 5.1% C2 100BGA 3Ptich SMD 164 23 14.3% C2 100BGA 4Pitch NSMD 262 20 7.6% C2 100BGA 5Pitch NSMD 470 26 5.5% C2 547BGA 4Pitch SMD 269 28 10.5% C3 0201 A Horizontal 407 21 5.3% C3 100BGA 3Ptich SMD 161 23 14.3% C3 100BGA 4Pitch NSMD 261 18 6.9% C3 100BGA 5Pitch NSMD 451 22 4.9% C3 547BGA 4Pitch SMD 271 21 7.7%

Originally published in the Proceedings of SMTA International, Ft. Worth, TX, October, 2013

More realistic – and production driven – evaluations are

• n the 0.4 and 0.5mm BGAs, and on the 0201s. Their

area ratios were approximately:  0.4mm BGA (SMD pads): 0.53  0.5mm BGA (SMD pads): 0.60  0.5mm BGA (NSMD pads): 0.65  0201 (NSMD pads): 0.75 Actual area ratios varied from the specification due to variation in aperture sizes and foil thicknesses. Aperture Size Measurements Table 3 shows the average aperture measurements and

• ther statistics for each device.

Table 3. Aperture measurements (in mils) for diameters of circular apertures and sides of square apertures Foil Thickness Measurements Table 4 shows the measured thicknesses of each stencil. No thickness data was available for stencil A2, so the average of the other PhD, No Nano stencils was substituted in the calculations. Table 4. Foil thickness measurements (in mils) The range of measurements and their deviation from the specification are considerable. While the typical aperture size variation within a stencil was less than 2%, the stencil-to-stencil (manufacturing process-to- manufacturing process) differences were as large as 22%. Even small deviations from specification can cause big issues in stencil printing. For example, an aperture specified at 9.8 mils on a 4 mil foil would have an area ratio of 0.61. If that aperture was only 8.5 mils, its area ratio would drop to 0.53, releasing less solder paste, and it’s volume would also drop, resulting in a much smaller (and more inconsistent) deposit than expected. Similarly, if the stencil is only 3 mil thick instead of 4 mil thick, the area ratio increases from 0.61 to 0.81 to release more paste than expected, but the aperture volume drops to 25% less than expected due to the thinner foil. The stencil manufacturing process has a significant influence on the solder paste printing process; therefore, control of the stencil manufacturing process is absolutely critical to performance on the SMT line. Transfer Efficiencies The measured solder volumes were coupled with the measured aperture sizes to estimate actual TEs, which were plotted against actual ARs to generate print performance curves. Both electroformed stencils performed poorly and had significant size and thickness inaccuracies; they were

• mitted from the final plot. The 0201s (AR=0.75)

experienced some fill issues; they were also omitted from the final plot. Figure 3 shows the performance of the best stencils in the key 0.5 to 0.65 AR range. Figure 3. Transfer efficiencies of different stencils The lines in figure 3 represent 3 different variables:  The marker styles represent the different

• suppliers. Suppliers A, B and C are denoted

by squares, circles and triangles, respectively.  The marker and line colors represent the different foil materials. Purple lines represent the 304 SS and orange lines represent the 301 SS.  The line styles represent the coatings. Solid lines indicate no nanocoating treatment, dashed lines indicate the thermally cured

0.3mm BGA 0.4mm BGA 0.4mm BGA 0.5mm BGA 0201 SMD NSMD SMD NSMD NSMD SPEC 7.28 8.86 9.84 10.83 11.6 Average 7.05 8.63 9.59 10.62 11.39 Range 1.56 1.80 1.54 1.60 1.78 Min 6.08 7.62 10.06 9.58 10.56 Max 7.64 9.42 8.52 11.18 12.34

STENCIL MATERIAL

A B C D Average

A1

FG 4.00 4.00 4.00 4.00 4.00

A2

PhD no data no data no data no data 4.08

A3

FG, Ni Plated 4.35 4.40 4.25 4.35 4.34

A4

E-form 2.70 2.75 2.50 2.40 2.59

B1

FG 4.15 4.15 4.15 4.15 4.15

B2

PhD, no nano 4.10 4.15 4.15 4.15 4.14

B3

PhD, Nano 4.15 4.25 4.15 4.10 4.16

B4

E-form 3.60 3.45 3.45 3.70 3.55

B5

Laser Nickel 4.00 4.15 4.00 3.95 4.03

C1

PhD, no nano 4.00 4.10 4.00 4.00 4.03

C2

PhD, nano 4.15 4.05 4.16 4.15 4.13

C3

FG, nano 4.15 4.15 4.20 4.20 4.18 Stencil Thickness - Measured on cut pieces just outside print area

Originally published in the Proceedings of SMTA International, Ft. Worth, TX, October, 2013

nanocoating; dotted lines indicate the 2-part wipe-on nanocoating. All of the stencils plotted in figure 3 performed well. The results are relatively tightly grouped, with little differentiation was noted in overall TE performance, except for one particular stencil that stood out. The stencil that showed the best transfer efficiencies was C2, 304 SS with thermally cured nanocoating. Potential problems that could cause higher print volumes, such as positional inaccuracies or slag on the bottom surface of the stencils were investigated, and the stencil showed no signs of either. In fact, it had the tightest positional accuracy and the smoothest walls with no slag protrusions. The CVs of all prints larger than 0.3mm were all near or less than 10%, also. It was clearly the best performing stencil in these tests. Type 4 vs Type 5 Solder Powders One stencil from each supplier was selected to run Type 5 solder paste. Type 5 solder paste has smaller solder particles in it, ranging from 5 to 15 microns diameter as

• pposed to Type 4’s range of 20 to 38 microns
• diameters. The smaller paste particle size should

enable denser particle packing in the aperture, better release from the aperture, and better volume

• repeatability. The prints were generated and measured

at the end of the original 11-stencil run on Day 2 of testing, using the same techniques as the main

• experiment. The transfer efficiency results are shown

in figure 4. Figure 4. Transfer efficiency comparison of Type 4 and Type 5 solder pastes The Type 5 powder showed better release on stencil B3 across the entire range of area ratios. The Type 5 powder showed a slight release advantage on stencil C2 in the 0.43 and 0.54 ARs, but offered no advantage on the 0.60 and 0.65 ARs, because the stencil demonstrated superior release on the Type 4 paste. Stencil C2 was also the best performer in the previous

• test. Stencil A1demostrated a flat response curve with

Type 5 paste; it is assumed that experimental error, either in the printer setup or the measurement setup, caused the anomalous readings. Volume repeatability is as important, and arguable more important than transfer efficiency. Table 5 compares the average volumes and CVs of the Type 4 and Type 5 pastes for stencils B3 and C2. Table 5. Comparison of print volumes (in cubic mils) and repeatability for Type 4 and Type 5 solder pastes The volumes are generally slightly higher and the CVs slightly lower for the Type 5 solder paste, as expected. The only exception is the performance of stencil C2, which was equivalent (but superior to all others) for both the Type 4 and Type 5 pastes at the higher area ratios. MEASUREMENT SYSTEM EVALUATION The measured volumes and transfer efficiencies appear higher than expected on most of the small devices. One contributor to the difference may be the use of Type 4 solder paste, because most of the comparative data is based on Type 3 pastes, and as previously discussed, the smaller particle sizes enable better print quality. Another contributor is the Solder Paste Inspection (SPI) system itself. Most of the comparative data available is based on phase shift, or Moire, interferometry SPI

• technology. These measurements were taken with a

light and laser-based technology that promises improved accuracy on smaller deposits by capturing more of the actual solder paste deposit in its scan. A photograph of the scanning system is shown in figure 5.

Type 4 Type 5 Type 4 Type 5 B3 0201 A Horizontal 385 406 6.7% 5.6% B3 100BGA 3Ptich SMD 144 163 16.5% 11.6% B3 100BGA 4Pitch NSMD 229 252 11.8% 7.9% B3 100BGA 5Pitch NSMD 425 445 7.0% 6.9% B3 547BGA 4Pitch SMD 247 258 9.6% 8.7% C2 0201 A Horizontal 416 429 5.1% 4.6% C2 100BGA 3Ptich SMD 164 168 14.3% 14.7% C2 100BGA 4Pitch NSMD 262 276 7.6% 5.9% C2 100BGA 5Pitch NSMD 470 476 5.5% 5.8% C2 547BGA 4Pitch SMD 269 270 10.5% 10.7% Avg of Vol CV% Stencil Component Type

Originally published in the Proceedings of SMTA International, Ft. Worth, TX, October, 2013

Figure 5. Simultaneous light-based 2-D and laser- based 3-D solder paste inspection. To compare the volume measurement accuracy of the new technology against the Moire technology, a PCB was printed and baked for 1 hour at 125°C to stabilize the paste deposits by evaporating the liquid portion of the paste to leave only the metal. The board was measured 10 times with each machine. A variety of deposit sizes and pad configurations were tested. Figures 6 through 9 show the results for Machine #1, the existing measurement technology, and Machine #2, the new measurement technology. Figure 6. Average volume readings in initial accuracy test Figure 7. Average area readings in initial accuracy test Figure 8. Variation in volume readings in initial accuracy test Figure 9. Variation in area readings in initial accuracy test Machine #2 consistently measured higher volumes and higher areas than Machine #1, with comparable variation down to the 8.8 mill aperture size. At the 7 mil aperture size it showed far less variation in its

• readings. Smaller apertures will naturally produce

more variation in deposit volumes due to their tighter area ratios; however the divergence in CVs between the two machines indicates considerable differences in the measurement techniques.

Originally published in the Proceedings of SMTA International, Ft. Worth, TX, October, 2013

SPI ACCURACY ASSESSMENT The two technologies

• ffered

widely ranging differences in measurement values. Which one was correct? Conversations with SPI manufactures indicated that accuracy is typically assessed by measuring various sized metal cylinders of known

• volumes. However, the disadvantage to this test is that

the cylinders are symmetrical, shiny, and smooth, versus solder paste deposits which are irregularly shaped and comprised of metal spheres that are suspended in a liquid. Solder paste deposits would reflect light differently than metal cylinders, which could affect the results. The volume assessments referenced in Figures 6 through 9 provided insight but used dried (liquid evaporated) paste deposits which are not representative of production conditions. The

• bjective of this work was to devise a test using actual

paste deposits to asses which SPI machine was more accurate. Paste Weighing Experiment This test printed the solder paste on labels, used the SPI machines to inspect the prints, then removed the labels and weighed them. The paste then was removed from the labels, and they were weighed again to calculate the mass of the paste that had been printed them. The labels were weighed on an analytical balance with a 5- digit gram scale at Custom Analytical Services in Salem, NH. Stencil B2 was chosen for this test because it uses common, low cost materials and manufacturing processes, and it performed well in the previous round

• f print tests.

The density of the solder paste was determined by Benchmark’s Nashua, NH analytical laboratory by weighing a precise volume of the paste. The density was not available from the paste manufacturer partially because the density can change based on allowable variations in the metal load. The calculation was made

• n the two separate lots of paste used in Run 1 and 2.

The readings were .000000071 and .000000072 grams per cubic mil indicating the paste density between lots was consistent and the calculations were accurate. The average volume of each deposit was then calculated using the mass, density and I/O count of each device. Figure 10 shows the PCB with the labels affixed; figure 11 shows a close up of the printed labels. Again, a variety of device sizes were tested. 8 samples were taken for each device. Figure 10. PCB with labels ready for print test Figure 11. Printed labels prior to weighing. The results of the weight tests showed consistency in the method of determining actual print volumes. Figures 12 through 14 depict the results from the weight tests in the 0.5 to 0.65 AR range. Figure 12. Weighing test results for 0.5mm BGA 100s

Originally published in the Proceedings of SMTA International, Ft. Worth, TX, October, 2013

Figure 13. Weighing test results for 0.4mm BGA 620s Figure 14. Weighing test results for 0.4mm BGA 100s Machine Comparison to Weighing Results In the first round of comparisons, a new, state-of-the-art Moire technology machine (Machine #3) was installed at the evaluation site, and Machine #1 was no longer used in the tests. Machine #3 originally returned readings significantly below the weight test. Machine #2 was no longer at the evaluation site, so the readings taken from a prior set of measurements were substituted and were closer to the weighed results for every device. In the second round of comparisons, Machine #3 underwent a hardware upgrade on-site and was reprogrammed once. It then produced somewhat better results in a second round of tests, with its best performance after a 3rd round of programming. Machine #2 also underwent a hardware upgrade while it was off-site, and initially gave poorer results upon reinstallation (run #2). It was reprogrammed and then returned good accuracy results on run #3. Table 6 summarizes the overall results. Table 6. Summary of paste weighing experiments  “Good” is considered within 15% of weighed and calculated volumes  “Better” is considered between 15% and 30%

• f weighed and calculated volumes

 “Bad” was higher than 30% deviation from weighed and calculated volumes Overall, the most consistent, fool-proof measurement system was the weight test method. It is impractical as a regular production control method, but can serve as an excellent test for equipment calibrations, periodic process control checks, or troubleshooting. Insights gained from the accuracy tests include:

• Machine-to-machine variation influences volume

readings

• Programmer-to-programmer variation influences

volume readings

• Use extreme caution when comparing

datasets generated on different platforms, different machines or by different programmers

• SPI manufacturers and users tend focus more on

repeatability than accuracy, because repeatability is easier to measure and considered more relevant in a production environment

• None of the machines seemed to capture as much

volume as weighing; they all came in 10-65% below the weighed amounts, depending on how they were programmed.

• The delta between the machine reading results and

the weighing results increased as feature size decreased APERTURE SURFACE ANALYSIS A final analysis of the stencils attempted to correlate aperture wall surface topographies with transfer efficiencies. Experimental Method Stencil sections were laser cut from each stencil used in the study. They sections exposed the long side of an 1812 aperture, as shown in figure 15.

Originally published in the Proceedings of SMTA International, Ft. Worth, TX, October, 2013

Figure 15. Diagram of aperture wall cut for surface analysis. The samples were mounted vertically (figure 16) and scanned with a Cyber Technologies CT-300 at the Aculogic test facility in Peabody, MA. The scanner uses a confocal white light sensor with a measurement range of 0.6mm and a resolution of 0.02 microns. Figure 16. Vertically mounted stencil aperture sample and direction of scan. For each of the three exposed aperture walls, a 1x0.3mm area was scanned. From that data, the area selected for analysis was 1x0.075mm to exclude bottom side slag from the wall roughness calculation, as shown in figure 17. Figure 17. Scanned area, analysis area and bottom side slag Results Bottom side slag was found on several stencils, and in some cases was severe, as seen in figure 18. Figure 18. Photograph and scan of stencil wall with peak slag height of 3 mils on a 4 mil foil. Slag protrusions are on the PCB side of the foil. Not all stencils demonstrated excessive slag, which is likely related to laser cutting parameters. Table 7 lists the findings for all the stencils in the study. Table 7. Scan results for stencils  Ra, or surface roughness, is measured in µmeters  Thickness and slag measurements are also expressed in µmeters. Note that overall thickness measurements do not include slag, which is reported separately where applicable. The laser cut stencils from supplier A showed the roughest walls and greatest amount of slag. The laser cut stencils from supplier B showed some slag, and the laser cut stencils from supplier C showed the cleanest walls and no slag. Note that stencil C2 was the top performer in the print tests. Electroformed stencils do

Originally published in the Proceedings of SMTA International, Ft. Worth, TX, October, 2013

not form slag in their plating process, but did not perform well in print tests. Gage Repeatability and Reproducibility (GR&R) An industry accepted practice to assess the repeatability

• f SPI platforms is a Gage R & R study which entails a

single board with paste deposits placed in an oven at 105 to 125°C for approximately 1 hour to drive off the liquids and solidify the paste deposits. The board is then run through the SPI platform multiple times including rotations of 90, and 180 degrees. The summary of these results for Machine #2 and #3 appears below: In running a GR&R as defined by the Automotive Industry Action Group (AIAG) and Measurement Systems Analysis (MSA) guidelines, findings indicated that both Machine #2 and Machine #3 performed acceptably

• ver

all angles, neglecting process

• tolerances. When applying 50% tolerance margins to

the analysis, it was discovered that most conclusions of repeatability were significantly different, favoring Machine #3. SUMMARY In print performance tests based on transfer efficiencies and volume repeatabilities, stencil C2, the 304 SS with thermally cured nanocoating, was the top performer. Characteristics that could impact paste release, such as wall roughness, bottom side slag protrusions, and positional accuracy were all checked; the stencils showed the cleanest cuts and the best positional

• accuracy. They are; however, more costly and have

longer lead times than typical laser-cut SS stencils, but may provide the

• ption

for high-performance, miniaturized print requirements. Considerable differences were noted in the aperture sizes and foil thicknesses, affecting area ratios and print

• volumes. Electroformed stencils had the most widely

varying thicknesses and gave the poorest print performance. SPI systems demonstrated significant differences in accuracy, both from platform to platform, and from programmer to programmer. As deposit sizes get smaller, accuracy will become increasingly important. The SPI machines always returned values lower than the weight test results. No direct correlations were drawn between wall roughness and release, partially due to the influence of bottom side slag. The protrusions will likely create gasketing problems that cause print quality issues in production environments. None were noted during this test, but bottom side protrusions were noted to cause inflated TEs on a related test. ACKNOWLEDGEMENTS The authors would like to thank the numerous participants that made these analyses possible:  Joe Crudele, Paul Bodmer, Bruce Tostevin and Rey Molina of Benchmark Electronics for technical support  Jeremy Saise for technical support and assistance in execution of the tests  Ray Whittier of Vicor for technical support and assistance in execution of the tests  Chris Tibbetts of Analogic for technical support and Cyberscans of the aperture walls  Bruce Guttman of Custom Analytical Services for technical support  Stencil providers for stencils and technical support  SPI equipment companies for machines and support  Matt Holzmann of CGI Americas for funding REFERENCES [1] Shea, C. and Whittier, R., “Evaluation of Stencil Foil Materials, Suppliers and Coatings’” Proceedings of SMTA International, 2011

Stencil and Solder Paste Inspection for Miniaturized SMT Components

Chrys Shea Shea Engineering Services Burlington, NJ Robert Farrell Benchmark Electronics Nashua NH

SMTA International – Fort Worth, Texas October 13-17, 2013

• Original objective was to determine the best stencil

technology for printing small deposits

– Better solder paste transfer efficiencies and print volume repeatabilities indicate a better performing stencil

• However, two Solder Paste Inspection (SPI) machines gave

very different results for paste volume

– As a result, it was decided to include an accuracy assessment of the SPI platforms used in this study

• Platform to assess stencil wall topography was made

available and a decision was made to include this in the experiment to determine if there was a correlation between wall topography and print performance

Overview & Study Development

Stencil Experiment

• Purpose: Choose the best stencil technology for

production

• 0.4 & 0.5mm pitch µBGA
• 0201s
• Type 4 no-clean solder paste
• Side leg on Type 4 vs Type 5
• Tests used:
• Benchmark Electronics test vehicle
• 3 stencil suppliers – 12 stencils total
• Same solder paste
• Same 10 PCBs
• 10 consecutive prints off of same printer and tooling, dry wipe after each

print

• All printed same day
• SPI Machine # 2 to measure print volumes

Objective: Identify the best stencil technology for volume production of miniaturized populated SMT components

0.4mm µBGA

Materials and Mfg Processes

• Materials - 3
• Stainless steel optimized for laser cutting
• Stainless steel with smaller grain size
• Electroformed nickel
• Manufacturing Processes - 3
• Laser cutting
• Electroforming
• Nickel plating over Stainless Steel
• Nano-coatings - 2
• Can be applied only by stencil manufacturer
• Can be applied by manufacturer or user

Experimental Matrix

• Three different suppliers

– A, B, C

• All used same Gerber file
• All specified 4mil thick foils
• Abbreviations for materials:

– FG: Datum Fine Grain 301 SS – PhD: Datum 304 SS – FG, Ni Plated: Ni over FG – Laser Ni: Electroformed Ni that has been laser cut – E-form: Electroformed Ni – Nano2: Dek NanoProtek – Nano1: Laserjob Nanocoating

Stencil Material(s)

A1 FG A2 PhD A3 FG, Ni Plated A4 E-form B1 FG B2 PhD, No Nano B3 PhD, Nano2 B4 E-form B5 Laser Ni C1 PhD, No Nano C2 PhD, Nano1 C3 FG, Nano1

Test Vehicle

• Designed in-house by Joe Crudele
• Contains many tests
• SPI measured most features
• Analyzed for 0.3, 0.4,

0.5mm BGAs and 0201s

• Gave ARs from 0.50 to 0.75

100BGA 4Pitch NSMD 100BGA 5Pitch NSMD 100BGA 3Pitch SMD 0201 A Horizontal 547BGA 4PitchSMD

7.2” 5.2”

Basic Metrics in Stencil Printing

• Aperture Area Ratio (AR)
• Paste Transfer Efficiency (TE)
• Simple statistics

– Mean – Standard deviation – CV% (std deviation as % of mean)

Area Ratio, AR

Area of aperture walls Area of circuit side opening = AR

Transfer Efficiency, TE

Volume of paste deposited Volume of stencil aperture = % TE x 100

Transfer Efficiency & Area Ratio

ARs and TEs can be theoretical or actual:

• Theoretical are based on specified dimensions
• Sufficient for paste or print parameter tests that use the same stencil
• Actual are based on measured dimensions
• Needed when different stencils are used
• Shortcut AR formula: AR = D/4t

where D= circle’s dia or square’s side, t = foil thickness

This experiment used Actual Area Ratios and Transfer Efficiencies

Transfer Efficiency & Area Ratio

At separation, the forces holding the deposit to the pad must overcome the forces holding the deposit to the stencil walls Stencil

PWB

After the aperture is filled, the solder paste sets up and sticks to both the stencil walls and the pads. Depending on area ratio, a portion of the paste will release to the PWB, while some will stay in the aperture PCB Pad Paste

• The smaller the AR, the lower the TE
• Higher TE’s indicate better paste release

Results

Foil Thickness (mils)

• Measurement coupons cut from just outside print area
• Four measurements per coupon were averaged
• Average thicknesses used in AR and TE calculations
• No data for A2; used average of other PhD, no nano thicknesses

STENCIL MATERIAL

A B C D Average

A1

FG 4.00 4.00 4.00 4.00 4.00

A2

PhD no data no data no data no data 4.08

A3

FG, Ni Plated 4.35 4.40 4.25 4.35 4.34

A4

E-form 2.70 2.75 2.50 2.40 2.59

B1

FG 4.15 4.15 4.15 4.15 4.15

B2

PhD, no nano 4.10 4.15 4.15 4.15 4.14

B3

PhD, Nano 4.15 4.25 4.15 4.10 4.16

B4

E-form 3.60 3.45 3.45 3.70 3.55

B5

Laser Nickel 4.00 4.15 4.00 3.95 4.03

C1

PhD, no nano 4.00 4.10 4.00 4.00 4.03

C2

PhD, nano 4.15 4.05 4.16 4.15 4.13

C3

FG, nano 4.15 4.15 4.20 4.20 4.18

Stencil Thickness - Measured on cut pieces just outside print area

Aperture Sizes (mils)

• 5 of each aperture size were measured on each stencil

– Measured with an Acugage model HPL 25-3-LT4 – Aperture size variation within each stencil <2%

– Stencil-to-stencil variation as much as 22%!!!!

• Average measurements used in AR and TE calculations

0.3mm BGA 0.4mm BGA 0.4mm BGA 0.5mm BGA 0201 SMD NSMD SMD NSMD NSMD SPEC

7.28 8.86 9.84 10.83 11.6

Average 7.05 8.63 9.59 10.62 11.39 Range 1.56 1.80 1.54 1.60 1.78 Min 6.08 7.62 10.06 9.58 10.56 Max 7.64 9.42 8.52 11.18 12.34

Paste Volume Measurement

• Taken with Machine # 2 SPI system
• 10 consecutive prints

–Identical Print Set Up for Each Run – 1000 data points for each 100 I/O BGA & 0201s – 5470 data points for the 547 I/O BGA

• Exported to Excel

– Pivot tables to extract average volumes and standard deviations – CV calculated

• Avg volumes (cu mils) used in TE

calculations

Stencil Component Type Avg of Vol StdDev of Vol CV% A1 0201 A Horizontal 382 25 6.4% A1 100BGA 3Ptich SMD 141 25 17.4% A1 100BGA 4Pitch NSMD 229 27 11.9% A1 100BGA 5Pitch NSMD 421 30 7.2% A1 547BGA 4Pitch SMD 245 23 9.3% A2 0201 A Horizontal 393 24 6.1% A2 100BGA 3Ptich SMD 151 22 14.7% A2 100BGA 4Pitch NSMD 243 24 9.7% A2 100BGA 5Pitch NSMD 438 31 7.1% A2 547BGA 4Pitch SMD 246 23 9.5% A3 0201 A Horizontal 394 24 6.2% A3 100BGA 3Ptich SMD 124 28 22.4% A3 100BGA 4Pitch NSMD 214 34 16.1% A3 100BGA 5Pitch NSMD 413 28 6.7% A3 547BGA 4Pitch SMD 235 24 10.2% A4 0201 A Horizontal 320 42 13.2% A4 100BGA 3Ptich SMD 122 16 12.8% A4 100BGA 4Pitch NSMD 183 19 10.5% A4 100BGA 5Pitch NSMD 329 25 7.7% A4 547BGA 4Pitch SMD 187 18 9.8% B1 0201 A Horizontal 377 25 6.6%

Transfer Efficiencies & CVs

• When comparing different stencils, TE for each stencil is the only way to normalize the data

– Comparing volumes only does not account for differences in aperture size or foil thickness – Comparing TEs based on specified aperture sizes and foil thicknesses instead of actual does not account for variation among stencils, either

• A4 eliminated due to varying thicknesses, B4 did not have apertures for 0.3pitch

A1 A2 A3 A4 B1 B2 B3 B4 B5 C1 C2 C3 MIN MAX SPEC

AR 0.61 0.58 0.56 0.58 0.58 0.59 0.69 0.56 0.61 0.60 0.61

0.56 0.69 0.62

TE 84% 82% 74% 82% 82% 80% 85% 71% 83% 87% 84% 71% 87% AR 0.44 0.43 0.42 0.43 0.43 0.44 no data 0.40 0.45 0.43 0.45

0.40 0.45 0.46

TE 70% 75% 57% 65% 68% 67% no data 53% 72% 77% 71% 53% 77% AR 0.76 0.74 0.76 0.73 0.73 0.74 0.87 0.73 0.77 0.74 0.74

0.73 0.87 0.73

TE 73% 73% 64% 70% 72% 71% 74% 73% 71% 76% 71% 64% 76% AR 0.53 0.53 0.50 0.52 0.52 0.53 0.62 0.48 0.54 0.52 0.54

0.48 0.62 0.55

TE 77% 80% 67% 72% 74% 72% 83% 72% 77% 83% 78% 67% 83% AR 0.65 0.64 0.64 0.63 0.65 0.65 0.77 0.60 0.67 0.64 0.65

0.60 0.77 0.68

TE 94% 94% 82% 91% 91% 89% 101% 91% 93% 100% 91% 82% 101%

547BGA 4Pitch SMD 100BGA 3Ptich SMD 0201 A Horizontal NSMD 100BGA 4Pitch NSMD 100BGA 5Pitch NSMD

A1 A2 A3 A4 B1 B2 B3 B4 B5 C1 C2 C3 MIN MAX SPEC

AR 0.61 0.58 0.56 0.58 0.58 0.59 0.69 0.56 0.61 0.60 0.61

0.56 0.69 0.62

CV 9.3% 9.5% 10.2% 9.7% 10.3% 9.6% 9.8% 14.1% 10.1% 10.5% 7.7% 8% 14% AR 0.44 0.43 0.42 0.43 0.43 0.44 no data 0.40 0.45 0.43 0.45

0.40 0.45 0.46

CV 17.4% 14.7% 22.4% 17.7% 18.2% 16.5% no data 22.7% 15.2% 14.3% 14.3% 14% 23% AR 0.76 0.74 0.76 0.73 0.73 0.74 0.87 0.73 0.77 0.74 0.74

0.73 0.87 0.73

CV 6.4% 6.1% 6.2% 6.6% 6.4% 6.7% 6.1% 6.7% 5.9% 5.1% 5.3% 5% 7% AR 0.53 0.53 0.50 0.52 0.52 0.53 0.62 0.48 0.54 0.52 0.54

0.48 0.62 0.55

V 11.9% 9.7% 16.1% 13.1% 12.3% 11.8% 10.0% 15.0% 9.3% 7.6% 6.9% 7% 16% AR 0.65 0.64 0.64 0.63 0.65 0.65 0.77 0.60 0.67 0.64 0.65

0.60 0.77 0.68

CV 7.2% 7.1% 6.7% 7.1% 7.0% 7.0% 6.7% 6.2% 6.0% 5.5% 4.9% 5% 7%

547BGA 4Pitch SMD 100BGA 3Ptich SMD 0201 A Horizontal NSMD 100BGA 4Pitch NSMD 100BGA 5Pitch NSMD

Transfer Efficiencies (TE)

Higher TE indicates better paste release

0.4mm NSMD (9mil) 0.4mm SMD (10mil) 0.5mm NSMD (11mil)

Type 4 vs Type 5 Paste

Type 5 solder paste showed a TE advantage

• Stencil C2: ARs below 0.6.
• Stencil B3: all ARs
• Stencil A1: data indicates systematic error in printing or measuring

C2 shows advantages in:

• Better overall release
• No need for T5 paste at

ARs 0.6 & higher

TE Values – Seem Very High

• 75% at AR~0.52? (9mil aperture/4mil foil)
• 80% at AR~0.60? (10mil aperture/4mil foil)
• 90% at AR~0.65? (11mil aperture/4mil foil)

Measurements were not taken with traditional structured white light SPI system that we are accustomed to. They were taken with a new type of color light & laser combination measuring system (Machine # 2)

Measurement Accuracy

• Different SPI machines measured different area and volume values on

same PCBs

• Which instrument is more accurate?

Accuracy Tests

• SPI manufacturers typically use metal cylinders of known volume for accuracy assessment

– Cylinders are a smooth, shiny, and symmetrical metal – not representative of solder paste deposits, which have irregular shapes and are comprised of metal spheres suspended in a liquid – A low cost test was devised to assess accuracy using actual paste volumes

• “Referee” between the machines
• Weigh deposits, determine paste density and calculate volumes

Calculating Paste Volumes

• Cut labels to cover component footprints on TVs
• Printed PCBs with stencil B2
• Measured with SPI machine
• Weighed labels, scraped paste off, weighed again to

determine paste weight

• Converted to volume using density measured in lab

Weighing Test Results

Results were extremely consistent, indicating that printing was consistent and weighing is a robust measurement method

Machine Accuracy Results

• First run:

– Machine #1 taken out of experiment and replaced with newer model Machine #3. Reads much lower volumes than weighing test returned – Machine #2 not available for test, reference measurements from GR&R substituted and were much closer to weighing results

• Second run:

– Both machines get hardware upgrades – Machine #3 reads somewhat closer to weighing results – Machine #2 reads lower volumes than machine #3 or weighing results. First time for machine #2 reading shiny labels, so it will get reprogrammed and tested again.

• Third run:

– Machine #3 reprogrammed for shiny white background and provides good readings close to weighing results – Machine #2 also reprogrammed and also provides good reading close to weighing results

Both machines reported volumes 10-50% less than the method of weighing deposits

Machine Accuracy Results

Test Method Run #1 Run #2 Run #3 SPI Machine #2 Good Bad Good SPI Machine #3 Bad Better Good Weighing Good Good Good

Insights from results:

• Machine-to-machine variation influences volume readings
• Programmer-to-programmer variation influences volume readings
• Use caution when comparing datasets generated on different machines or by

different programmers

• Weighing method is unique and most consistent of all; easy if 5 digit scale

is available

• Weighing paste method is more true-to-life than the metal cylinders on

equipment manufacturers’ calibration plates

Good: Within 15% of weighing, Bad = greater than 30%

Insights From Weighting Results (cont’d)

• SPI manufacturers and users focus more on repeatability than accuracy.

– Repeatability is easier to measure and considered more relevant in production.

• None of the machines seemed to capture as much volume as weighing.

– Came in 10% to 65% below the weighed amounts depending on how they were programmed.

• Delta between weighing and SPI results increased as the paste deposits

became smaller

• Weighing is not practical for regular production control

– Can serve as an excellent test for equipment calibration, periodic process control checks, or troubleshooting.

• Limitations on weighing:

– White labels changed color and topography of board – Glossy labels minimized liquid absorption into the paper.

Gage Repeatability and Reproducibility GR&R

• An industry accepted practice to assess the repeatability of

SPI platforms is a Gage R & R study

– A single board with paste deposits placed in an oven at 105 to 125°C for approximately 1 hour to drive off the liquids and solidify the paste deposits – The board is then run through the SPI platform multiple times including rotations of 0, 90 and 180 degrees.

• In running a GR&R as defined by the Automotive Industry

Action Group (AIAG) and Measurement Systems Analysis (MSA) guidelines, findings indicated that:

– Both Machine #1 and Machine #2 performed acceptably over all angles, neglecting process tolerances. – When applying 50% tolerance margins to the analysis, it was discovered that most conclusions of repeatability were significantly different, favoring Machine #3.

Aperture Surface Analysis

Objective: Investigate correlation between aperture wall topography and print performance

• Stencil sections were laser cut from stencils used in the

study

• Long edge of 1812 apertures were examined
• Edges were trimmed to expose stencil walls
• Stencil sections were then vertically mounted

Measurement Device

• A Cyber Technologies CT-

300 was used to scan the stencil walls.

• The sensors is a confocal

white light sensor with a measurement range of 0.6mm and a resolution of 0.02 microns.

• Good sensor for

transparent materials, like Nanocoating

Scan and Microscope Views of Slag

Stencil Wall Slag

A3 - Laser Cut Stainless Steel Stencil

Surface Roughness Results – Ra

1 5

(µm)

The lower the Ra (um), the smoother the stencil wall

Surface Roughness Discussion

• Laser cut stainless steel from supplier A showed excessive

slag on apertures

– Can cause gasketing issues during printing – May produce higher paste volumes due to gasketing problems – Did not increase average deposit height

• Laser cut SS from supplier B showed some walls with

slag, some without

• Laser cut SS from supplier C showed the smoothest SS

walls and no slag. Also, they were the top performers.

• Electroformed stencils showed no slag
• No correlation between aperture topography and stencil

performance observed. May be the result of noise caused by slag.

Overall Discussion

• Laser cut PhD with Nanocoating 1 (Stencil C2) showed the

best overall performance

– Checked positional accuracy: 0.01mil in X, 0.6mil in Y – Best release with T4 and T5 pastes – Release T4 just as well as T5 – No slag – Longer lead times and higher cost

• Big differences in aperture sizes and foil thicknesses

– Electroform had widely varying thicknesses based on aperture density, too hard to control

• Big differences in SPI system accuracy

– Platform-to-platform and programmer-to-programmer – Values always less than weighing method

• Electroform had smooth walls, laser cut had rougher walls and

slag, depending on laser cutter

• No direct correlation drawn between wall roughness and

release

Acknowledgements

The investigators would like to thank:

• Joe Crudele, Paul Bodmer, Bruce Tostevin, and Rey

Molina of Benchmark Electronics for technical support

• Bruce Guttman of Custom Analytical Services for

technical support

• Chris Tibbetts of Analogic for Aperture Surface Analysis
• Stencil providers for stencils and technical support
• Jeremy Saise for technical support
• SPI equipment companies for machines and support
• Matt Holzmann of CGI Americas for funding

Chrys Shea chrys@sheaengineering.com Bob Farrell

Robert.farrell@bench.com

Questions?

Thank You!