UNDERSTENCIL WIPING: DOES IT BENEFIT YOUR PROCESS?

Originally published in the Proceedings of SMTA International, Orlando, FL, October, 2012 UNDERSTENCIL WIPING: DOES IT BENEFIT YOUR PROCESS? David Lober, Mike Bixenmen, D.B.A Kyzen Nashville, TN, USA david_lober@kyzen.com; mikeb@kyzen.com Chrys Shea Shea Engineering Services Burlington, NJ, USA chrys@sheaengineering.com Ed Nauss Speedline Technologies Franklin, MA, USA enauss@speedlinetech.com ABSTRACT Understencil wiping has gained an increase in interest over the last several years. Changes in circuit designs, such as miniaturized components, increased density of components, and new stencil technology, as well as changes in and increased attention to employee safety and environmental regulations have driven renewed interest. New understencil wipe solvents have been introduced recently to address these issues. However, little to no quantitative data exists on how these new solvents benefit the printing process, specifically, yields, transfer efficiencies and print volume repeatability. The lack of data has led to slow adoption of understencil wipe cleaning solvents, as there is no proof that the costs justify their benefits. This paper quantitatively examines the effect of various combinations of solvents, wiping frequency, and pastes in a way that is relevant to both process engineers and the bottom line. Key words: Printing, Cleaning, Nano coating, under stencil wipe, solvent INTRODUCTION One of the most critical steps in an electronic assembly process, as defined by the potential to cause the most SMT defects in the final assembly, is stencil printing. Some of the conscious decisions that influence the process are printer configuration, operator skill, training, and solder paste selection. Additionally, there is a myriad of external factors that contribute significantly to the process. The ambient environmental conditions, lot-to-lot variations in materials, and luck are a few examples of external factors that can contribute to the integrity of the process. As with any process that has a plethora of inputs, limiting the variation of the printing process is of utmost importance. As noted above, some inputs are conscious and/or are relatively easy to control while others are less controllable, and as such, preventive measures should be taken to keep their influence on the process to a minimum. At the core of the printing process is the stencil and how the stencil interacts with the circuit board and the solder paste. The goal for the printing process is to deliver a known and controlled amount of solder paste to a location consistently. There are two main steps in transferring the desired amount of solder paste to the desired location: aperture fill and aperture release. In the first step, the aperture is filled with solder paste using a squeegee, and in the second, the paste is released from the stencil and transferred onto the PCB by moving the PCB away from the stencil (Figure 1). In the first step, the solder paste is pumped into the aperture by the squeegee, meaning the paste is forced into the aperture by the mechanical action of the squeegee on the solder paste. In order for the printing process to work effectively, the apertures need to be completely filled with solder paste.

Originally published in the Proceedings of SMTA International, Orlando, FL, October, 2012 Figure 1. Stencil-PCB Separation Process (PCB is underneath stencil and is not visible) Paste release from an aperture is usually incomplete. Some residual paste always remains inside the aperture, and some may migrate to the bottom side of the stencil during the separation process. Errant paste on stencil surfaces is often the root cause of printing problems; stencil cleanliness is a key factor in a successful process. To address stencil cleanliness, many stencil printers have the capability to perform automatic understencil wiping. The understencil wiping process, and the role that solvents play in it, are the primary focus of this paper. paste deposited on the pad, one possible cause is paste pump out, i.e. excess solder paste on the bottom of the stencil that prevents the gasketing of the stencil to the board. This poor gasketing allows solder paste to flow out between the stencil and the board during the filling phase of printing. In the understencil wipe process, paper or synthetic fabric, 1 is wiped across the underside of the stencil in the printing equipment (Figure 2). This wiping paper may have a solvent on it and/or a vacuum applied to it, with the goal of assisting in the removal of any solder paste from the underside of the stencil and the inside the aperture. Figure 3. Example of Paste Buildup on Aperture Walls Many printers have the option to add a solvent and/or an auxiliary removal step, typically involving a vacuum, to the wiping material to either prevent or reduce these two problems. The solvents are typically isopropyl alcohol (IPA) or an engineered solvent blend. Figure 2. Understencil Wipe System Failure modes include insufficient or excess amounts of solder. In the event that there is an insufficient amount of solder paste deposited on a pad, one of the possible causes is that an aperture has become partially occluded or the aperture wall has developed a film of solder paste (Figure 3). In the event that there is an excessive amount of solder IPA has been traditionally solvent used in essentially all applications involving unreflowed solder paste. Historically the choice of IPA made sense, as most solder flux formulations were based on IPA. However, solder paste manufactures are moving away from IPA-based fluxes for several reasons, the most notable of which are safety considerations and new flux platforms. IPA is a flammable solvent with a flash point (the minimum temperature required for a substance to produce flammable vapors) of 12 C (54 F), which can easily lead to fires and/or explosions.

Originally published in the Proceedings of SMTA International, Orlando, FL, October, 2012 In addition to the flammability considerations, there are concerns about the toxicity of IPA. IPA can exhibit toxic effects at high and or prolonged exposure. IPA is also becoming an inefficient solvent for modern solder pastes. When natural rosins were the primary constituents of solder paste flux, IPA was an excellent choice as a solvent material because rosin is highly soluble in IPA, the IPA readily evaporates, and the IPA was extremely affordable. Current fluxes, however, especially no-clean formulations, have materials in them that are not as soluble in IPA as rosin and require more specialized solvents. The solubility character of today s solder pastes require an engineered solvent designed to match up with the flux s composition. When considering the functions the engineered solvent needs to perform, it must: Solubilize the flux package Release the solder spheres Dry rapidly Leave zero residue Not interact with materials used within the process. Additionally, the engineered solvent needs address the safety concerns: it must be either nonflammable or have a very high flash point, and exhibit fewer negative health effects than IPA. One question that continuously arises is how often should an understencil wipe be done? Many factors influence wipe frequency requirements. Generally speaking, miniaturized, high density designs require more frequent wipes because they present more opportunity for errant paste to remain in the stencil s apertures or stick to the stencil s bottom surface after separation. Wipe frequencies can range from every print on a highly miniaturized product to every 10-20 prints on a low density design. Another question that often arises regards the wipe sequence. Stencil printers equipped with automatic underwipe systems offer users the option to program the order and speeds of the wiper passes using dry, vacuum, or solvent modes. Most wipe solvents, frequencies and sequences are set by engineers or technicians based on observations. Some are derived by DOE s. Most selections are product-specific, based on the solder paste formulation, PCB configuration and yield history, and only one comprehensive, recently published study has been identified. 2 METHODOLOGY The complex process of depositing solder paste on a circuit board for SMT assembly is critical to the success and profitability of the assembly line. Many primary factors such as paste formulation and machine setup affect the process quality, but a multitude of external factors such as ambient environment, operator skill level and lot-to-lot material variation are also highly influential. The key to a successful printing process is limiting its variation. Primary factors are relatively easy to control, but external influences are not. The strategy behind a highyielding process is two-fold: control what can be controlled, and become robust against what can t be controlled. Experiment The stencil underwipe investigation studied the effects of wipe frequency and chemistry on the output quality of the solder paste printing process. Six pastes were selected for the underwipe study, a no-clean and water-clean formulation from each of three leading solder paste suppliers. Due to time constraints, 4 of the 6 paste formulations (3 no-clean and 1 water-clean) were tested. Also due to time constraints, print optimization studies were not performed. Pastes were tested at typical print parameters for highly miniaturized, densely populated PCBs. Two laser-cut fine grain 4mil stainless steel stencils were used; one had a wipe-on release-modifying coating applied. Print tests took place in the Speedline Technologies applications laboratory in Franklin, MA. Paste was printed using an MPM Momentum printer, and measured using a Koh Young Aspire solder paste inspection system. Four solvent options were investigated, labeled: dry (no solvent), IPA (isopropyl alcohol), K1 (developmental solvent) and K2 (commercial solvent). Wipe frequencies of 1 print per wipe (ppw) and 3 ppw were tested. The test vehicle was a very densely populated production PCB. The PCBs were numbered 1-20, printed in consecutive order for all runs, and cleaned in an ultrasonic bath between runs. For each combination of paste, stencil and solvent, multiple test prints were taken and measured. For wipe frequencies of 1 ppw, the sample size was the standard 10 board run, for 3 ppw the sample size was expanded to 12. RESEARCH OBJECTIVE The overall objective of the under wipe research project is to understand the role of engineered solvents in controlling production-level solder paste printing processes. The goal of this specific study was to determine the effect of various understencil wipe solvents and process parameters on the output of the printing process. Experimental Design The experiment was designed to systematically identify key factors that influence the robustness of the printing process. The variables that were examined were the solder paste type, the solvent, the stencil and the wipe sequence. The DOE matrix is shown as Table 1. The response that was measured was the solder paste volume of the 0.5 mm BGA features. All prints were made on one particular model of

Originally published in the Proceedings of SMTA International, Orlando, FL, October, 2012 stencil printer and the paste volumes were measured by automated Solder Paste Inspection (SPI). Solvent Stencil Coating? Prints/ wipe 1 Y 3 None (Dry ) 1 N 3 1 Y 3 IPA 1 N 3 1 Y 3 K1 1 N 3 1 Y 3 K2 1 N 3 Table 1. DOE matrix for each solder paste formulation. Two pastes were subjected to the full set of tests and two were tested with coated stencils only. Flux Selection A solder paste s flux formulation is a major factor in the printing process. To characterize the impact that the flux has on the effectiveness of the understencil wipe process four solder paste formulations were examined: three noclean fluxes and one water soluble flux. The emphasis was placed on no-clean fluxes due to their notoriety of generally being more difficult to clean solder paste after reflow. Due to this, it was inferred that the un-reflowed solder paste should be difficult to clean as well, and experiments in the laboratory confirmed that. The solder pastes came from several major solder paste manufactures, which provided flux samples as well as expertise. The fluxes were designated as A, B, C, and E. Fluxes A, B, and C were leading no-clean fluxes, and flux E was a leading water soluble flux. Due to time constraints, pastes D and F were not tested, and pastes C and D were tested with coated stencils only. Solvent Selection As the main goal was to characterize the effects that solvents, specifically, engineered solvents, have on the printing process four solvents were selected. The first one was no solvent used, meaning the wiping paper was dry. The second solvent selected was IPA, to serve as a baseline. The next two solvents were leading engineered solvents, designated K1 and K2. The two engineered solvents were designed to be drop in IPA replacements. Stencil Selection Two different stencil types were selected for this experiment. Both were laser cut, fine grain 4 mil stainless steel stencils. The only difference between them was one was coated with a nano coating and the other had no nano coating. The two stencil types were chosen as the nano coating changes certain properties of the surface that may influence the under wipe process. The area ratio of the BGA aperture was 0.68. Test Vehicle: The test vehicle shown in Figure 4 was a densely populated PCB that runs in high-volume production, and the subject of previous print studies and baseline data. 4 The test vehicle had 14,466 solder deposits, of which 6,912 were 0.5mm BGA36 with 6 devices per board and 32 boards per array. The paste volumes were measured for all pads on the PCB, but only the BGAs were used to calculate the mean and Coefficient of Variation (CV) of the volume. Twenty test vehicles were obtained and labeled 1-20 and were printed in numerical order for all runs. Between runs, the PCBs were cleaned in an ultrasonic tank. The theoretical volume, 100% transfer efficiency, for the BGAs was 366 mils 3. For the purposes of this study, 80% transfer efficiency and 10% CV (standard deviation expressed as a percentage of the mean) is considered a robust, in control process.

Originally published in the Proceedings of SMTA International, Orlando, FL, October, 2012 Figure 4. Test Vehicle Wipe Frequency The production printing process of the test vehicle uses a wipe frequency of one, meaning the stencil is wiped after every print. In these tests, wipe frequencies of one and three prints were chosen. For tests conducted with one print per wipe, ten boards were printed giving a sample size of 69,120 solder deposits. For tests conducted with three prints per wipe, there were twelve prints made, and the sample size was 82,944 solder deposits. DATA FINDINGS A typical measurement of a printing process effectiveness is the repeatability of the paste deposits that it produces. While different solder pastes may exhibit different release characteristics, the consistency of the process output is the key to its success and profitability in a production environment. draw any conclusions. The data collected for the variation in solder paste volumes are shown in Figure 6. Once again there is very little variation in the CV for all conditions. The solvent K2 with an uncoated (no nano coating) stencil at one print per wipe showed slightly less variation than all other conditions tested, but again was not substantially different. All measured CV values were either below 10% or within a very small margin of error. The processes that used Paste A met the criteria of greater than 300 mil 3 of paste release. This was the last in the series of print tests, and the wear and tear on the PCBs began to induce variation in the data. Despite the additional experimental error, the process maintained CV s of less than 10% in all but one or two cases. It also appeared to have a relatively negative response to solvent K1, as evidenced by the spike in CVs for that series of wipes. The 0.5mm BGA deposits used to measure process repeatability are the finest feature on the PCB. They are printed from stencil apertures nominally 10.8mil diameter in a 4mil foil, with an area ratio of 0.68. Theoretical aperture volume is 366mils 3. For comparative purposes, 300 mil 3 = 82% transfer efficiency and provides a convenient reference point. The volume of each paste deposit was measured and recorded in a database, which was exported to Excel. In spreadsheet format, the files were manipulated to extract the print volumes of the BGAs. The average, standard deviation, and CV were calculated and plotted. No-Clean Solder Pastes The solder paste volumes for solder paste A with all conditions tested are shown in Figure 5. The mean solder paste volume for all solvents was approximately the same, and per the criteria set for the experiment all was acceptable. When wiping without a solvent, the solder paste volumes were marginally higher, but not substantially enough to Figure 5. Solder Paste A Data Findings

Originally published in the Proceedings of SMTA International, Orlando, FL, October, 2012 Figure 6. Coefficient of Variation for Solder Paste A Solder Pastes B and C were run on a nano coated stencil only. The results for paste B are shown in Figure 7. Paste B printed consistently regardless of the wipe chemistry or frequency. This indicates that the nano-coated stencil provides good paste release. It met the test criteria of at least 300mil 3 of solder paste with a CV of less than 10%. The use of underwipe solvents appears to lower variation and provide some added degree process control at one print per wipe. Figure 8. Solder Paste C Data Findings Water Soluble Paste The water soluble paste, E, was completed in full. The mean solder paste volumes are shown in Figure 9 and the CV data is shown in Figure 10. Paste E provided robust output regardless of underwipe solvent or frequency. The solder paste volume averaged near 300 mil 3, and CV are all below 10%. Its response to solvents appears negligible for print volumes, and negligible or slightly positive with respect to limiting variation. Figure 7. Solder Paste B Data Findings Solder Paste C met the minimum volume criteria, transferring just under 95% of the theoretical aperture volume, but also showed higher levels of variation in print volumes, barely meeting and sometimes slightly exceeding the 10% CV criteria. The solder paste seems to react negatively to the introduction of any solvent to the process, especially K2. This is a key indicator that the success of a solvent wiping process requires a cleaning fluid that matches up (readily dissolves) the flux present in the solder paste. Figure 9. Solder Paste C Data Findings

Originally published in the Proceedings of SMTA International, Orlando, FL, October, 2012 Print Yields Print yield is the coarsest measurement available, and the easiest to compute. Print test yields are measured at the PCB level, not the per-deposit level. If all 14,466 deposits per print fall within their volume control limits, the print is considered good. If one or more deposit falls outside its control limits, the print is considered bad. One deposit outside the control limits causes the print to fail. Each component type on the PCB has individual control limits that are based on the stencil apertures theoretical volumes. Figure 10. Solder Paste C Coefficient of Variation It should be noted that the testing process used typical print parameters for densely populated assemblies, and not optimized printing parameters for the individual solder pastes on the test vehicle used. As such, performance comparisons of the different solder paste used would be inappropriate. The performance characterization should be limited to the process inputs that were used on the particular individual solder paste. Many of the print failures were due to the nonconformance of a single deposit. Nonconforming deposits can be viewed on the inspection system s monitor at high magnification. If an attributable cause for the nonconformance, such as a residual fiber on the PCB, was visible, the PCB was manually coded as good. Otherwise, the board was considered a failure. The data from the yield tests showed no definitive patterns. Print yield analysis was considered inconclusive. Print yield charts are shown in Figure 11: Figure 11. Print Yield Data

Originally published in the Proceedings of SMTA International, Orlando, FL, October, 2012 RESULTS AND DISCUSSION The test results did not indicate substantial differences among the use of solvents, as well as selection of the solvents. The current hypothesis of why this occurred is that the ideal or laboratory grade conditions were not as strenuous to the printing process as a real world environment. All of the prints for each set of conditions were done without any pauses or interruptions to the process. While this enabled the experiments to be as efficient as possible, they are not representative of what occurs on a typical production line. In a production environment, stoppages and other noise factors are a regular part of the process. Stoppages can cause residual solder paste to dry in the apertures; dried solder paste is far more difficult to remove than wet solder paste. Other noise factors can cause solder paste to accumulate on the bottom surface of the stencil, which also interferes with print quality by creating gasketing problems. Subtle differences in each paste s response to solvents were also noted. Paste E demonstrated negligible to slightly positive improvements with the addition of solvent, regardless of the individual solvent. In contrast, paste C appeared to react negatively towards all solvents. Different solder paste formulations use different ingredients, which require different solvents to dissolve them. If the solvents and/or materials present are immiscible or incompatible, poor interactions result. Informal testing has shown that certain solder pastes are not compatible with certain solvents. When incompatible solder pastes and solvents are combined, the fluxes tend to ball up, minimizing the solder paste s contact area with the solvent. This leads to the inference that just as with reflowed solder, un-reflowed (wet) solder paste must also be matched up with soil and characterized by their differences and similarities in solubility. CONCLUSIONS All 48 print processes analyzed were considered robust, meeting the benchmark of 80% transfer efficiency with a CV of 10% or less. A few processes narrowly exceeded the 10% CV guideline, but by a margin too small to deem them truly unacceptable (<1%), particularly given that optimization runs were not performed. The print volume data do not indicate any significant differences in average volumes based on stencil coating, wipe frequency or solvent type. Slight differences were noted in the CV of the volumes, indicating that more frequent wiping or the use of select chemistries may improve the printing process level of control. Print yield data was inconclusive with respect to the input variables. This series of tests was performed under nearly ideal laboratory conditions: perfectly conditioned paste; new, clean stencils; climate controlled environment; pastes always at working viscosity via kneading, dummy prints, or continuous running; ultra wipes prior to beginning each test run; technical support engineers from the paste companies on-site; robust PCBs ultrasonically cleaned between each run; and experienced engineers executing the tests. Previous tests that involved non-ideal conditions indicated an advantage of using solvent, as have other studies 5 and real-world production situations. The divergence of the data demonstrating marginal benefit under ideal conditions but substantial benefit under nonideal conditions may indicate the best fit for the application of chemical underwiping is to regain control of processes affected by noise or other aberrations induced by the production environment. Observations of paste response to certain solvents indicated that the same principle on which post-solder cleaning chemistry is selected also applies to stencil underwipe chemistry: the solvent must be matched to the soil. The combination of this investigation, a previous internal study, and a recently published study 5 leads to the following 3-point hypothesis: Stable printing processes that are under control show little dependence on stencil coating, underwipe solvent, or frequent wiping. Non-ideal printing processes, such as those following a long pause, those with compromised paste behavior due to environmental exposure, or those with residual paste on the bottom surface of the stencil from the previous print are ideal candidates for solvent-based underwipes. Different solder pastes will require different solvents that are chemically matched to them for optimum process performance. To investigate the second point of this hypothesis, it is recommended that further testing include Response to Pause (RTP) and controlled noise levels to identify areas where both dry and solvent wipes will most improve process performance. RECOMMENDATIONS FOR FUTURE RESEARCH As noted, solvent wipes have been noted to improve performance in non-ideal processes, but show negligible improvement in ideal processes. This research project has tested both ends of the spectrum, and should move forward by characterizing the effects of different solvents and wipe parameters in overcoming typical production noise factors. This may be accomplished by intentionally inducing controlled noise in a laboratory, or by performing longerterm runs on a production line. The production line test would offer less control of the noise factors, but provide a testing environment most reflective of the real world situations.

Originally published in the Proceedings of SMTA International, Orlando, FL, October, 2012 Further laboratory work should include solubility studies, which use various methods to determine the physical and chemical properties required of a solvent to dissolve a material, on unreflowed solder paste samples. These tests may show that there is not one universal solvent for all unreflowed solder paste and provide scientific indicators for solvent selection based on flux composition. ACKNOWLEDGEMENTS The authors wish to thank two people who have helped provide their expertise on this project: Bruce Barton of Cookson Electronics and Brook Sandy of the Indium Corporation, who supported the studies and actively assisted in the execution of the experiments. Both of these people went above and beyond to help make this testing a reality, we are sincerely thankful. REFERENCES [1] Abbet, K.F., and Jones, M. D., Report on Field Tests of a Developmental Fabric Technology for Cleaning Fine- Pitch Stencils, Proceedings, Pan Pacific Electronics Symposium, February, 2005 [2] Garcia, O., et al, Nano Technology Improve Critical Printing Process, Proceedings, APEX International Conference, February, 2011 [3]Report Number KYZ-2011-1, Shea Engineering Test Report, Print Underwipe Test Report January, 2012. [4] Shea, C., and Whittier, R., Evaluation of Stencil Foil Materials, Suppliers and Coatings, Proceedings, SMTA International, October, 2011 [5] Seeling, K., et al., A Study to Determine the Impact of Solder Powder Mesh Size and Stencil Technology Advancement on Deposition Volume when Printing Solder Paste, Proceedings, IMAPS New England Microelectronics Symposium, May, 2012

Understencil Wiping Does it Benefit your Process? David Lober / Mike Bixenman Kyzen Corporation Chrys Shea Shea Engineering Services Ed Nauss Speedline Technologies 1

Outline/Agenda Introduction Underwipe considerationss Methodology Results Discussion Conclusions & Recommendations Acknowledgements Questions & Answers 2

Introduction 3

Why Wipe the SMT Stencil? Paste PCB Pad Stencil PCB After solder paste flows into the apertures, it sets up and sticks to both the stencil walls and the pads. At separation, the forces holding the deposit to the pad must overcome the forces holding the deposit to the stencil walls. The paste near the wall stretches and snaps. Depending on the pad:wall area ratio, a portion of the paste will release to the PCB, while some will stay in the aperture Some paste may also stick to bottom of stencil due to stringing, bad gasketing or pump out 2012 Kyzen Corporation 4

Understencil Wiping Criticality increasing due to Component miniaturization Density / Placement of components Need for improved yields Wipe options Dry/vacuum Solvent IPA Specially engineered materials Historically, not a lot of science was applied to the process 5

Opinions on Solvent Wiping Good: Removes sticky flux from bottom of stencil instead of smearing it Removes dried paste from inside aperture to improve transfer Vacuum wiping without solvent will actually dry paste inside small apertures, making print quality worse Bad: Affects paste properties (negative interaction) Dries aperture wall which then requires relubrication from next print cycle Very little quantitative data to support any opinions 6

Purpose of this Research Examine effects of Solventless (dry) wiping Solvent wiping IPA and specialty solvents Solder paste interaction Wipe frequency Quantify relationships Based on print yields and volume repeatability Follow up on previous tests that indicated strong benefits of solvents on print process 7

Under Wiping Considerations 8

Automatic Under Wipe Process Process Parameter Min Max Typical Wipe Speed Wet (in/sec) 0.5 5 2 Wipe Speed Dry (in/sec) 0.5 5 2 Wipe Speed Vacuum (in/sec) 0.5 5 1 Vacuum pressure Off On On Wipe Frequency (ppw) 1 ~20 Varies* 10/15/2012 * DOE used typical parameters and 1 & 3 ppw 9

Key Solvent Characteristics Chemical compatibility with the solder paste Dissolve the flux & free up the powder spheres No bad reaction with paste that affects viscosity Process compatibility Dry evenly and at a controlled rate No splitting or tails Leave no residue Equipment compatibility Does not attack valves, seals, flowmeters, reservoirs Safety & health Non-toxic, non-flamable, low- (or no-) odor Always match the solvent to the application 10

IPA Isopropyl Alcohol The de facto standard for under wiping Worked great on rosin-based (tin-lead) fluxes Dries fast Doesn t attack printer HW Low cost Easy choice + - Less effective on more complex (lead-free) fluxes May react with paste and make it runny or gummy Flammable High VOC Not allowed in some facilities or geographical areas 10/15/2012 IPA may no longer be a suitable solvent for modern stencil printing processes 11

Methodology 12

Designed Experiment Tested: Wipe sequence Dry with Vacuum Wet-Vacuum-Dry Wipe Frequency 1 or 3 prints per wipe Wipe Solvent No solvent - Dry IPA (2) Engineered Solvents, K1 & K2 Laser-Cut Fine Grain Stencils With and without nano-coating Lead-Free Solder Pastes (3) No-clean and (1) water clean tested (2) water clean not tested due to time constraints 13

Test Vehicle Densely populated production PCB used on many previous print tests: µbga pads 14

Test Vehicle 14,466 solder deposits Paste volume measured on all pads (6) 0.5mm BGA36 per board; 32 boards per array 6912 BGA pads per print, ~300mil 3 each One print per wipe 10 boards printed Sample size = 69,120 solder deposits Three prints per wipe 12 boards printed Sample size = 82,944 solder deposits 15

DOE Matrix 16

Test Facility Speedline (MPM) Applications Laboratory, Franklin, MA Printer: MPM Momentum Solder Paste Inspection: Koh Young Aspire 17

Output Variables Print Yields All 14,466 pads inspected, result is PASS or FAIL One or more out-of-spec deposits = FAIL Expressed as a percent Print volume repeatability µbgas only 10.8mil circular apertures in a 4 mil foil = Area ratio of 0.68 Average volumes, std deviation as percent of avg 18

Results 19

Print Yields Inconclusive Many of the failures were for a single insufficient deposit (out of 14,466) All failures verified visually Time constraints precluded optimizing each paste for the test vehicle Previous tests with optimized print parameters demonstrated yield advantages of solvent wiping 10/15/2012 20

Print Yields Paste A Nano Paste A No Nano 100 100 80 80 60 40 20 1 ppw 3 ppw 60 40 20 1 ppw 3 ppw 0 Dry IPA K1 K2 0 Dry IPA K1 K2 Paste E Nano Paste E No Nano 100 100 80 80 60 40 20 1 ppw 3 ppw 60 40 20 1 ppw 3 ppw 0 Dry IPA K1 K2 0 Dry IPA K1 K2 10/15/2012 Pastes B & C demonstrated similar inconsistencies 21

Volume Repeatability 0.5mm µbgas only (10.8mil circles) Basic metrics for robust print process Minimum 80% transfer efficiency Deposits average 293mil 3 greater 300 mil 3 = 82% transfer, axes set to show 300 mil 3 Std deviation <10% of mean volume Aka, Coefficient of Variation (CV) All print processes tested met criteria Cannot directly compare solder paste performance because print parameters were not optimized for each formulation 10/15/2012 22

BGA Volume (mils3) Coefficient of Variation, % No-Clean Solder Paste B 400.0 Paste B Coated Stencil Only 14.0% Volumes above 300 CVs below 10% 350.0 300.0 250.0 200.0 150.0 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% 100.0 Dry IPA K1 K2 1ppw 3ppw CV-1ppw CV-3ppw 0.0% 23

BGA Volume (mils3) Coefficient of Variation, % No-Clean Solder Paste C 400.0 Paste C Coated Stencil Only 14.0% 350.0 12.0% Volumes above 300 CVs below 10% 300.0 10.0% 250.0 8.0% 6.0% 200.0 4.0% 150.0 2.0% 100.0 Dry IPA K1 K2 1ppw 3ppw CV-1ppw CV-3ppw 0.0% 24

BGA Volume (mil3) Coeffiient of Variation % No-Clean Solder Paste A min 400.0 350.0 300.0 Paste A, Volume Comparison max 12.0% 10.0% 8.0% Paste A, CV Comparison 250.0 6.0% 200.0 4.0% 150.0 100.0 Dry IPA K1 K2 2.0% 0.0% Dry IPA K1 K2 CV -Uncoated, 1ppw CV - Coated, 1ppw Uncoated, 1ppw Coated, 1ppw Uncoated, 3ppw Coated, 3ppw CV - Uncoated, 3ppw CV - Coated, 3ppw 25

BGA Volume (mil3) Coeffiient of Variation % Water Soluble Solder Paste E min Paste E, Volume Comparison 400.0 350.0 300.0 250.0 200.0 150.0 100.0 Dry IPA K1 K2 Uncoated, 1ppw Coated, 1ppw Uncoated, 3ppw Coated, 3ppw max 12.0% 10.0% 8.0% 6.0% 4.0% 2.0% 0.0% Paste E, CV Comparison Dry IPA K1 K2 CV -Uncoated, 1ppw CV - Coated, 1ppw CV - Uncoated, 3ppw CV - Coated, 3ppw 26

Discussion 27

Stable Printing Processes 48 print processes tested 2 pastes with 4 solvents, 2 wipe freqs, 2 stencils 2 pastes with 4 solvents, 2 wipe freqs, 1 stencil Performed in a lab with minimal process noise Did not demonstrate sensitivity to stencil coating Demonstrated some sensitivity to Under wipe solvent Wipe Frequency 1ppw was better than 3ppw All processes met desired performance criteria, despite not being individually optimized 28

Results of Prior Tests Showed considerable print quality improvements with solvent wipe, despite Compromised paste quality Pauses in printing process Delays in measurement (allowing slump) Achieved 100% print yields at 3ppw using solvent, even with process noise! Test results not published, available upon request While ideal print process may not benefit from solvent under wipe, non-ideal process does 10/15/2012 29

Non-ideal Printing Processes Solder paste printing processes that experience: Pauses in process Extended working life or reuse from prior run Alignment issues Inadequate board support Stencil or PCB dimensional inaccuracies HASL finish or other topographical issues Undesirable rheological behavior (pastes that string or slump) Process noise that creates non-ideal conditions can be addressed with under wiping 30

Factors in Stencil Cleanliness Cleaning Processes Solder Paste Paste Delivery: Squeegee type, angle, stiffness, coating, etc OR Pump properties Original Cleanliness Wipe parameters: Sequence, speed, frequency Printer Setup Wiper type Paper/fabric type Board Support PCB Flatness PCB-Stencil Alignment Print Parameters: speed, pressure, separation Solvent effectiveness Dryness Baseline release characteristics Foil Thickness (Area Ratio) Alloy Aperture size & position Stencil Working time Exposure time Temperature Aperture wall geometry and topography Release coatings in aperture or on surface 2012 Kyzen Corporation Solvent compatibility Powder Size IDEAL STENCIL CONDITION No unwanted paste inside apertures or on bottom surface 31

Match the Solvent to the Soil The same principle behind post-reflow flux cleaning also applies to pre-reflow flux cleaning No one size fits all under wipe solvent Response of paste to flux demonstrated by increases/decreases in print volume variation Poorly matched solvents may: Not clean effectively May leave an unacceptable residue Change paste consistency and behavior Make the process worse instead of better 32

Conclusions & Recommendations 33

Two Tests with Contrasting Results Test with lots of process noise showed very positive improvement in print quality with solvent underwipe Tests with minimal process noise showed little improvement with solvent underwipe Need to identify which noise factors are best addressed by solvent 10/15/2012 34

Three-Part Hypothesis 1. Stable printing processes that are under control show little dependence on stencil coating, underwipe solvent, or wipe frequency 2. Non-ideal printing processes that result in errant paste inside the apertures or on the bottom of the stencil are excellent candidates for solventbased underwipes 3. Different solder pastes will require different solvents that are chemically matched to them for optimum process performance 10/15/2012 35

Continuing Research Test runs with noise Typical noise of production environment Alignment, PCB variation, untimed pauses, varying paste lots, production equipment & operators Intentionally induced noise in lab Misalignment, aperture-pad size mismatch, suboptimal print parameters, shear out from excessive working, long pauses (15, 30, 60min) Solder paste solubility studies in chem lab To identify best solvent for each paste To look for common trends in paste composition 36

Acknowledgements The authors wish to thank the following people for their support: Brook Sandy, Indium Corporation Bruce Barton, Cookson Electronics Tim O Neill, AIM Rob Tyrell, Stentech 37

Thank You! 38

Authors & Contact Information David Lober Process Chemist, Kyzen Corp. david_lober@kyzen.com Mike Bixenman CTO, Kyzen Corp. mikeb@kyzen.com Chrys Shea President, Shea Engineering chrys@sheaengineering.com Ed Nauss Process Engineer, Speedline Technologies (MPM) enauss@speedlinetech.com 39