PLASMA STENCIL TREATMENTS: A STATISTICAL EVALUATION

Transcription

1 PLASMA STENCIL TREATMENTS: A STATISTICAL EVALUATION Matt Kelly, P.Eng. 1, William Green 2, Marie Cole 3, Ruediger Kellmann 4 IBM Corporation 1 Toronto, Canada; 2 Raleigh, NC, USA; 3 Fishkill, NY, USA; 4 Mainz, Germany ABSTRACT As printed circuit board complexities continue to increase, packing more functionality into smaller physical dimensions has become increasingly important. As a result, a wider variety of electronic components are being incorporated into product bills of materials. Large body ASICs (logic) and sub-system docking connectors generally drive large SMT pad designs, while fine pitch devices such as flip chip QFNs, 0402, and 0201 chip passives are now commonplace within electronic circuitry. Integration of these large and small body components onto a single printed circuit board assembly (PCBA) with ever increasing population densities and tighter placement spacings, drives the need for consistent solder paste print deposits to ensure maximum first pass assembly yields and highest product quality / reliability levels. Balancing solder paste printing of large and small print deposits has been reported to be enhanced using various surface treatments on laser cut stencils. This study focused on examining the effects of a plasma coating compared to conventional stainless steel (SS), laser cut technology. Printing performance on a variety of components was evaluated including five different BGAs, flip chip QFNs, SMT electrolytic capacitors, 0805, 0402, and 0201 chip passives. Lead-free no clean and water soluble paste chemistries were included, along with two different aperture ratio (A/R) design points for all components studied. For assessing the durability of the plasma coating, production volume cleaning simulations were conducted with twenty four different solvents. Statistical analysis was conducted to evaluate any observed differences between conventional stencil technology and a plasma treated alternative. A design of experiments (DOE) was conducted to evaluate main effects and interactions, helping to make data generated decisions to answer the question: Do plasma treated stencils offer benefits over conventional technology stainless steel laser cut stencils? Key words: Plasma treatment, Stencil printing, Stencil cleaning, Design of experiments, Design for Six Sigma INTRODUCTION A good assembly process starts with a good solder paste printing process. As electronic designs continue to become more dense, integrating a wider variety of both large and small SMT components, stencil designs, stencil material selection, and stencil fabrication methods must be evaluated to ensure optimal solder paste deposition. This effort ultimately helps to maximum first pass assembly yields as well as overall product reliability performance. With ever increasing assembly cost pressures, suppliers have been offering lower cost stencil technology alternatives. Nano, or plasma coated laser cut stainless steel stencils are now available, claiming to offer improvements over conventional SS stencils and cost advantages over efab build up technology. Many studies have already been conducted in this area [1] [2] [3] [4], examining the benefits and drawbacks of surface treated stainless steel stencil technologies. This study is intended to expand the industry s body of knowledge by (1) statistically evaluating print performance of plasma coated, laser cut, stainless steel stencils compared to non-plasma treated, laser cut stainless steel stencils and (2) evaluating production life performance of plasma treated stencils using a wide variety of cleaning solvents. SCOPE & INTENT The scope of the study includes printing of SMT components found on Server and Storage class hardware including fine pitch discretes down to 0201 formats, large body 3,777 I/O LGA hybrid sockets, and a variety of SMT connectors including DIMM sockets and mezzanine style formats. The intent of the study was to answer the question: Do plasma treated printing stencils offer improved print performance and improved first pass print yields for a wide variety of SMT interconnect devices? Numerous marketing claims have been made by suppliers of stencils with state-of-the-art surface treatments. Claimed benefits of the technology include (among others): Improved paste release Reduced print errors More consistent performance Extended stencil life Easier to clean Limited reaction to cleaning solvents The study was therefore designed to evaluate such claims and to identify any significant differences between plasma treated and non treated SS laser cut stencils using design for six sigma (DFSS) statistical tools and methods.

2 PROJECT DELIVERABLES There were seven key deliverables for the project: 1. Increase understanding of plasma treatment application and limitations using SS laser cut printing stencils Identify benefits / drawbacks of using plasma stencils 4. Assess incoming stencil fabrication quality of plasma treated vs. non treated stencils 5. Compare solder paste print performance of plasma treated vs. non treated SS laser cut stencils 6. Assess print capability using a variety of component types including: 0201, 0402, 0805, FC-QFN, BGA, and SMT electrolytic capacitors 7. Assess print capability using SAC405 no-clean and water soluble solder paste flux chemistries 8. Assess stencil life (wear out) of the stencil plasma treatment when using a wide variety of industrial cleaning solvents and equipment types APPROACH The study was divided into two phases. Phase 1 was designed to evaluate resulting print performance of no-clean and water-soluble lead-free solders using a Server product test vehicle utilizing a variety of large and small SMT components. The objective of this phase was to directly compare print performance of plasma coated vs non-coated stencils; assuming highest quality plasma application. The second phase of study was created to assess any degradation in the plasma coating, resulting from any potential cleaning solvent interactions (leaching, chemical reaction, mechanical spray removal, etc). Phase 1: Print Performance Process Flow Figure 1 shows the general process flow used for Phase 1 testing. For each stencil configuration, nine test vehicle replicates were printed and analyzed. In total, four different stencils were included within the design of experiments. Figure 1: Printing Process Flow The first step in the process was to perform incoming quality inspections on stencils. Plasma coating quality was measured using wetting inks (Figure 3) and coupon (Figure 5) contact angles. Foil thickness was measured at four corners and one middle location of the image. Next, aperture dimensions were measured and compared to specifications. Lastly, aperture registration accuracy measurements were made; comparing actual vs specified x/y locations. Eight unique print set-ups were required using a full factorial design of experiments. Three different variables were included: plasma vs laser, production vs reduced aperture designs, and no clean vs water soluble lead free solder paste. For each test cell, three primer cards were printed, then six replicate test cards. The reason for using primer cards was to ensure the printing process was running at steady state to help reduce start-up print variation that might be observed. Resultant solder volumes and heights for each card were then measured using production scale equipment. All solder print deposits on the card were recorded and stored for subsequent statistical analysis. Phase 2: Solvent Cleaning Process Flow With the nature of today s global hardware operations, it was important to ensure that a wide variety of stencil cleaning equipment and commercially available cleaning solvents be considered. Phase 2 of the study focused on determining plasma treatment quality after exposure to repeated spray and ultrasonic cleaning processes using twenty-four different commercially available cleaning solvents. Figure 2: Solvent Cleaning Process Flow Figure 2 illustrates the process flow used for Phase 2 testing. Three different stencils were required. Two were used to compare any plasma treatment degradation using spray vs ultrasonic cleaning processes; using a single cleaning solvent. The third stencil was used to evaluate a next generation solvent (with faster drying times) using a spray cleaning process. The first step for Phase 2 testing was to measure incoming stencil contact angles with inks and contact angle test coupons. Figure 3 shows the inks used for quick assessment of the condition of the plasma coating. Two ink lines 1 cm in length were drawn on each specimen. The surface tension was determined by iteration between the maximum value at which the test ink flowed and the minimum value at which the ink contracted. Surface tension ratings of the three pens were 18, 20, and 30mN/m respectively.

3 exposure for the life of a production stencil. Contact angle measurements were taken every 50 cycles (Figure 6); monitoring any changes in surface energy of the plasma coating. Figure 3: Inks used for Surface Assessment For greater accuracy, contact angles were also measured. Ten identical coupons were designed into the perimeter of each stencil border such that each could be removed after multiple cleaning cycles. Figure 4 shows the location of the coupons relative to the card print image. Figure 5 shows a single coupon in greater detail. Apertures incorporated within coupon were selected from actual locations found within the card print image. Figure 6: Example Contact Angle on Coupon TEST VEHICLE The test vehicle used for the study was selected for the variety of SMT components found within the bill of materials. The PCB was 15.2 x 5.4 x ; with six signal and ten power/ground layers. Only the top side of the card was utilized with focus on 0201, 0402, 0805 passives, FC-QFNs, BGAs, and electrolytic capacitors found on the card. Figure 7 shows top and bottom side card layouts. Card orientation and squeegee print direction for all trials is shown in Figure 8. Figure 4: Contact Angle Coupons on Stencil Image Figure 5: Contact Angle Locations on Single Coupon To determine surface energy, 3 µl drops of water, diiodomethane and ethylene glycol were deposited on the surface that was to be tested. The dispersal and polar shares of the surface energy from the measured contact angles were calculated accordingly. Figure 7: Product Test Vehicle used for Study Stencils were subjected to 200 cleaning cycles (10 minute durations) which was determined to be a representative

4 Table 1 shows aperture geometries, production design points, 20% reduced aperture ratio designs, and a percentage reduction summary. All dimensions are in mils; all volumes are in cubic mils. Table 1: Test Vehicle Aperture Designs Figure 8: Test Vehicle Orientation and Print Direction STENCIL APERTURE DESIGN Four different stencils were fabricated for testing as follows: Stencil 1: Stencil 2: Stencil 3: Stencil 4: Plasma treated, production apertures Plasma treated, reduced apertures Stainless steel, production apertures Stainless steel, reduced apertures All stencils were fabricated at the same supplier, at the same time, with a nominal thickness of 4 mils. Stainless steel stencils were laser cut with no plasma coating. Production aperture designs used typical 1:1 designs from copper pad geometries found on the raw card. Reduced aperture designs targeted a 20% reduction from conventional 1:1 designs. An example of a stencil loaded into the printer is shown in Figure 9. Figure 10 shows the same A/R design points graphically. The general A/R range of interest was ; a common range found on server class products. Figure 10: Aperture Ratio Comparisons SOLDER PASTE Two different lead-free solder pastes were included as variables within the DOE matrix: No-clean SAC405, type 4 Water-soluble SAC405, type 4 Figure 9: Example Test Stencil Loaded in Printer Aperture design reductions were included in the study to determine if the use of plasma treated stencils could help improve solder deposition in cases where non-optimized aperture ratios (A/R) were present. Worst case A/R for the study was set at 0.6 for 0201 discretes. Note that all A/R manipulation was achieved by varying aperture openings only, keeping the stencil foil thickness constant at 4 mils. PCB FIXTURE A custom vacuum plate PCB fixture was designed and fabricated for the study as shown in Figure 11. The intent of the fixture was to minimize PCB warpage in the printer; helping reduce an external source of variation of resultant print deposits. The fixture included mechanical tooling pins for alignment/polarity and an array of vacuum nozzles to ensure the card remained flat during processing.

5 PCB Site Flatness Local site flatness measurements were taken to identify source of variation contributions originating from the PCB fabrication process. DESIGN OF EXPERIMENTS The experiment consisted of three variables of interest: Figure 11: Custom Printing Vacuum Plate MEASUREMENT MAPPING In total, 6,949 solder pads were measured on a single card. A break down by component type is provided below, and a mapping is shown in Figure 12. Component , ,040 QFN 152 BGA 4,597 SMT elec caps 10 Number of Pads per Card 1. Stencil type 2. Aperture design 3. Solder paste flux chemistry type A two level, full factorial design (3+6) (2 3 ) was used to understand variables of interest primary effects and interactions, shown in Table 2. Table 2: Design of Experiments Matrix Setup Stencil A/R Paste Replicates NC (3+6) = 9 1 Plasma Production WS (3+6) = 9 NC (3+6) = 9 2 Plasma Reduced WS (3+6) = 9 NC (3+6) = 9 3 Laser Production WS (3+6) = 9 NC (3+6) = 9 4 Laser Reduced WS (3+6) = 9 Stencil types consisted of two groups: stainless steel, laser cut, with and without plasma surface treatments. Aperture Ratios (A/R) consisted of conventional 1:1 aperture to copper pad design points and 20% reduced aperture design points for all components of interest. Two different lead-free solder pastes were evaluated including no-clean and water soluble flux chemistry types. Figure 12: SMT Solder Pad Mapping MEASUREMENT ACCURACY To ensure measurement accuracy of data collected, machine measurement capability analysis and incoming card site flatness evaluations were conducted. These verification measurements were taken to minimize external sources of variation during DOE printing trials. Measurement Capability Analysis The first verification test determined measurement capability of the solder volume analyzer equipment. This was conducted to ensure the gauge was in statistical control. The goal was to confirm that the variability of the gauge was small compared with the variation of the parts measured. The target limit was set at 10%. To achieve this, a nested study was created; printing nine cards, three times each, totaling twenty seven total measurements for analysis. Replicates were divided into two groups: three primer cards were printed followed by six test replicates. The reason for this was to further examine how quickly the printing process achieved steady state. If differences were observed, analysis would be conducted only on six test replicates. If no differences were found, all nine replicates could be included in the analysis. Constants for the experiment included: stencil thickness, PCB board fabrication, stencil supplier, printing process, printing equipment, and solder paste alloy. The primary output response for the test was solder paste volume for all deposits monitored on the test vehicle.

6 RESULTS Test results recorded and analyzed included incoming stencil quality assessment, solder volume machine capability measurement performance, incoming card flatness, printing performance assessment, and cleaning solvent compatibility. Incoming Stencil Quality A total of eighteen different component apertures were measured on a single stencil, comparing actual vs specified dimensions. Figure 13 shows that both laser and plasma treated stencils were built to the same 10 m range. As expected, there were no significant differences found with aperture range capability between SS laser cut and plasma treated stencils. Figure 13: Aperture Size Deviation ( m) Next, actual x/y aperture locations were compared to specified locations for all four stencils. Measurements were taken after aperture laser cutting and electro polishing. Figure 14 confirms that both stencil types were built to within a (0.05mm) range. Again, as expected no significant differences were observed with x/y aperture location accuracy between stencil types. Incoming verification measurements were considered important to ensure correct solder volumes were delivered to pads with adequate paste release from apertures. Measurement Capability Analysis A gauge study was conducted to understand the suitability of the gauge to detect true variation in solder volume using nested ANOVA models. Analysis was conducted to identify and quantify significant sources of variability for each component type of interest as shown in Table 3. For example for 0201 devices, the contribution of measurement to total variation was only 0.4%. The majority of the variation was shown to come from either within card or card-to-card variation; which was the desired outcome for the study. Table 3: Sources of Variability by Component Type Sources of Variation in % Measurement to Card to within Card Measurement Card Pad to Pad Total Component 3 cycles 9 cards Figure 12 % SMT elec Step 6mil Step 7 mil WB-QFN WB-QFN FC-QFN FC-QFN FC-QFN BGA-R Finally a nested ANOVA analysis covering all 6,949 pads was conducted examining variation contribution of the card ID, pad identifier, and repeat measurement cycle. As shown in Figure 15, very little variation contribution came from the repeat measurement cycle itself (0.07%). Figure 15: Variation Contribution By combining these results, it was determined that the solder volume measurement equipment was very capable of delivering accurate measurements and would not inject significant variation into DOE output responses. Figure 14: x/y Aperture Position Deviation (mils)

7 PCB Local Site Flatness Figure 16: 0201 Local Site Flatness Verification Figure 16 shows an example result of local site flatness measured on 0201 locations on the test vehicle. A maximum height of 15 m was measured on copper trace to via structures near to target print pads. This was considered an excellent result, showing that the variation contribution coming from the PCB design and fabrication was minimal. Similar results were obtained for the other components of interest, but are not shown. Print Results Analysis was conducted starting with the two smallest components of interest; 0201s and FC-QFNs. Figure 18: 0201 Pareto of Standardized Effects Figure 18 shows the same data in pareto format. Aperture ratio design had a 4X effect when compared to stencil type. Main effects and interactions were examined and are shown in Figure 19 and Figure 20. Examining this output shows that plasma treated stencils and no-clean lead-free pastes delivered only marginally higher mean volumes, while aperture ratio design had the most significant impact Package Analysis Solder paste volume output was analyzed using statistical software examining main effects and interactions based on analysis of data means and variance. The 88 solder pads per card x 9 test replicates tallied 792 data points collected for each test cell. Analysis (provided below) includes 6,336 data points collected across all eight test cells (Table 2). Statistically significant main effects were analyzed using a iterative reduction model technique for eliminating stepwise unnecessary model terms. Figure 17 shows that stencil type was not the most significant effect on solder volume delivery. Aperture ratio (factor B) was found to be the most dominant variable of significance deviating substantially from the straight fit line. Figure 19: Main Effects 0201 Printing Figure 17: 0201 Significant Factors Figure 20: Interactions 0201 Printing

8 FC-QFN Package Analysis Once 0201 analysis was completed, the next component of interest was examined; FC-QFN packages. The 152 solder pads per card x 9 test replicates tallied 1,368 data points collected for each test cell. Analysis included over 10,900 data points collected across all eight test cells (Table 2). The SMT pad structure of FC-QFNs is not typical, and is shown in Figure 24. The package consists of three different pad sizes (Table 1), with different orientations. Analysis was conducted only on pads one and two within the package. Sample sizes for pad three (single large pad) were insufficient to conduct a statistical analysis. Figure 21: 0201 Pad Orientation Effect During analysis, an unexpected trend was observed. Figure 21 shows that a pad orientation effect was identified. The same effect was observed for all three variables of interest. The 0201 sites printed in a parallel orientation yielded higher print volumes with significantly less variability. Figure 22 shows the preferred 0201 pad orientation relative to print and card transport directions. Figure 24: FC-QFN Pad Structure and Print Deposits The same approach used for 0201 analysis was conducted for FC-QFN packages. Pad 1 analysis revealed that the magnitude of the deviation from the straight fit line shows that aperture ratio (factor B) was the most important factor (Figure 25). Figure 22: 0201 Parallel vs Series Print Orientation Lastly, steady state print run order was analyzed. As previously stated, replicates were divided into two groups: three primer cards and six test replicates. Analysis of the data (Figure 23) revealed that no statistical difference was found between these groups; no settling time was found. The analysis included pad orientation effects described above. As a result, all nine replicates for each test cell were included in data analysis described within this section. Figure 25: FC-QFN Pad 1 Significant Factors Figure 26 shows the same data in pareto format. Aperture ratio design had a 7X effect when compared to stencil type. Figure 23: Run Order with Pad Orientation Included

9 (83%) solvents submitted were fully approved for use. Four materials were approved with exposure limitations. Table 4: Approval Status of Commercial Solvents Figure 26: FC-QFN Pareto of Standardized Effects Extremely similar results observed during 0201 analysis were found for FC-QFN pad 1 printing and are shown in Figure 27 and Figure 28. FC-QFN pad 2 was examined separately, again with very similar results observed with pad 1 analysis. Surface tension analysis via contact angle and ink measurements of three plasma treated stencils, with two selected cleaning solvents, and two different cleaning processes (spray in air and ultrasonic) showed no significant changes compared to initial values after 200 cleaning cycles. Testing confirmed that the surface tension of the plasma treated coatings (Figure 29) was not degraded after multiple solvent wash cycles, simulating production wash exposures. Figure 27: Main Effects FC-QFN Printing Surface Tension [mn/m] Plasma Teatment Performance Over Multiple Wash Cycles 30.0 Variable SOL1 Spray SOL2 Spray 27.5 SOL1 Ultrasonic Cycles Figure 29: Plasma Treatment Performance Multiple Cycles Figure 28: Interactions FC-QFN Printing Cleaning Solvents Eight different suppliers and 24 cleaning solvents were evaluated for compatibility with the plasma treatment. Results are shown in Table 4 indicating that 20 of the 24 SUMMARY & CONCLUSIONS In summary, results from the study show that these particular plasma treated printing stencils offered only marginal performance improvements when compared to conventional stainless steel, laser cut stencils. For the SMT components examined, representing an expected range found on Server and Storage class hardware, aperture ratio (A/R) design points were the single most important factor contributing to solder volume delivery, variation, and control. If plasma stencils from suppliers come with a cost

10 premium, the study shows that the marginal print improvements offered may not justify a potential cost increase. Future work is recommended to focus on even more difficult aperture ratios < 0.5, such as flip chip wafer bumping applications during first level module assembly. Differences in print performance with plasma treatments may become more apparent with these types of applications. The following conclusions were drawn, resulting from the statistical methods and analyses applied: Incoming quality of plasma treated stencils was found to be on par with conventional stainless steel stencils. This was an expected result, since front-end processing of both stencil types are similar. Foil thickness, aperture dimensions, and x/y registration were all within acceptable limits Results for other components of interest including 0402, 0805, and BGAs are not presented here, however the conclusions made for 0201 and FC-QFNs apply to these devices as well The plasma coating under evaluation was generally very stable and was shown not to degrade for a wide variety of commercially available cleaning solvents. This indicates that the treatment is compatible with many of the industry s leading cleaning chemistries. It is important to note that confirming compatibility is key for the longevity of the plasma coating, which can react negatively to some types of solvents Plasma coating wear-out was not observed after 200 cleaning cycles using spray and ultrasonic cleaning processes. This indicates that the plasma treatment is capable of withstanding typical production environment Successfully minimized external sources of variation to cleaning life cycles ensure DOE output could be more easily analyzed REFERENCES o Incoming printed circuit board (PCB) quality 1. Effect of Nano-Coated Stencil on Printing, R. was excellent, local site flatness was minimal Mohanty et al, Speedline Technologies. IPC-APEX o Custom PCB fixture helped ensure card 2. Qualification of stencil printing with nanocoated flatness during printing trials SMT-stencils, M. Rösch et al, FAPS Institute Erlangen-Nürnberg. o Solder volume equipment used was extremely 3. Evaluation of Stencil Foil Materials, Suppliers and accurate and repeatable; only 0.07% Coating, C.Shea, Shea Engineering Services and contribution to data variance R.Whittier, Vicor Corporation. SMTAi October Stencil Printing of Small Apertures, William E. Statistical analysis of 0201 and FC-QFN components Coleman, Photo Stencil. IPC-APEX revealed: o o o Plasma treated stencils only offered marginal improvements to printed volume means and variation reductions compared to conventional stainless steel stencils; findings consistent across multiple components Aperture ratio (A/R) design points dominated solder paste print performance; 4X-7X effect The no-clean lead-free solder paste evaluated printed marginally better than the watersoluble lead-free formulation tested 0201 SMT pad orientation effect found. Pads aligned in parallel to the print direction yielded higher overall print volumes and lower variability. This effect can be included in design for manufacturing (DfX) guidelines to help maximize print volumes for these small devices No start up (settling time) printing effects were observed during testing. As a result all nine print replicates were used for statistical analysis; further improving confidence levels of conclusions drawn