News

Dynamics in Pb-Free Wave Soldering

Published: 01 February 2006
 by R.A. Szymanowski, D. Casati, E. Saglia, P. Lotosky, K. Howell and G. Hueste
Print

An in-depth study of how three Pb-free alloys performed. Also evaluated: flux selection and convection preheat.

For Pb-free soldering processes, attention focused initially on the SMT primary attach. The number of published works for wave solder, mini-pot, hand solder and hot gas was limited. Even during the development period, solder material price and the annual amount of solder used always played a role in material selection. SMT processes use solder paste, and it is the technology of paste manufacture that significantly drives the cost of the product. The value of the alloy is of less consequence. While still in a state of some change, SMT processes migrated to the SnAgCu alloys and remain there today. There have been some optimization studies in reflow, but there may be a perception that the SMT processes are superior to Pb-free wave solder, where advancement is often considered slow and not progressing as well as SMT primary attach.

Early work in wave solder was affected by the same cost mentality, but in contrast to paste, wave solder bar cost is significantly affected by base metal price. Wave solder will use much more alloy per assembly than reflow, and alloy cost has become a critical selection criteria, even above performance. As an example, we might allow SAC305 or SAC405 solder pastes to cost about $100 per pound, and very little of that paste is discarded. Almost every gram could be used to create a solder joint. For wave soldering, a pound of SAC 305 bar might cost $8 (Ed.: prices are for comparison only), and SAC405 will cost almost $10. This 25% cost increase is attributed to the higher percentage of silver. Next, the wave-solder process will create solder dross, regardless of how many solder joints are made. That dross cost is variable, but a cost estimate is $30/hr., which could add $150,000 per year to the cost of making solder joints. Early work in Pb-free wave solder therefore primarily focused on a SnCu0.7 alloy, with some testing of SnAg3.5. The search for a world-standard Pb-free wave alloy continues and solder bar cost has a strong influence.

As we consider the rapid improvements in SMT and the perception that wave solder has not shown the same rate of improvement, note the following three considerations:

Logical and accurate studies of wave soldering are now being pursued. Wave soldering has always been affected by numerous variables. When solutions address the complete process, rather than a single point, the corrective action can be most effective.

The primary goal of this work was to bring the Pb-free wave solder process to a higher level of performance, to meet demands of boards that are bigger, thicker, heavier, with more thermal challenge, more density and demanding a no-clean process.

Our work focused on wave solder challenges anticipated in heavier boards, looking at the process attributes of flux, preheat and alloy. If the study focused on thinner product, existing materials and processes were seen as adequate.

The strategy was to begin with the confirmation of a flux. Because preheat was identified as a significant factor in soldering thick, thermally challenging boards, a test of preheat performance was planned. Alloy characteristics would then be studied, primarily by using designed experiment methods, followed by small confirmation runs. Current knowledge would focus efforts and optimize the design of experiments.

The approach came from considering the following known conditions:

In the context of soldering thermally challenging and thick board assemblies:

Experiment Descriptions

Flux testing. Flux testing was required as a mandatory building block for the following work and had five major parts. Phase 1 consisted of testing all flux candidates in a laboratory wetting capability test, thereby reducing the number of candidates to test more extensively in machine systems.

Phase 2 consisted of machine testing Alloy A (SAC405), Machine A:

Phase 3 consisted of machine testing Alloy B (SAC305), Machine B; flux application rate testing as before and at a lower rate; and varying contact time and temperature as above.

Phase 4 consisted of machine testing Alloy B (SAC305), configured to Machine B equivalent, with the same conditions as Phase 3.

Phase 5 consisted of visual and production scope inspection, and x-ray inspection to verify hole-fill as required.

Phase 1. A solder alloy pellet was put onto a copper coupon, covered with 10 micro-liters of flux, then reflowed in N2 atmosphere. The melting process is constantly monitored by a video system that permits observation of the alloy profile. The "contact angle" after 20 sec. from beginning of melting is considered. Initial and final situations are reported. Figure 1 shows the final wetting angle of one sample.

Click here to see all tables and figures (1.4MB PDF file).

To approach a "standard surface," coupons were treated as follows:

Surface roughness and morfology can heavily affect the results, thus the same type of copper-clad laminate (CCL) was used for all fluxes.The treated copper surface of the CCL used in this work was evaluated by AFM.

The wetting capability of flux is reported with a number, scale 0-100, defined here as "wetting index" (Figure 2). It is simply the cosine of the "contact angle" (multiplied by 100).

e.g.: Contact angle = 44 ° -> "Wetting Index": cos 44 x 100
Cos 44° = 0.72
0.72 x 100 = 72 wetting index

The higher the wetting index the better the wetting. A wetting index of 100 equals ideal wetting. It is possible that only the wetting angle by itself could be recorded, but another advantage in using the cosine of the wetting angle is that it is used in formulas to extact the "Work of Adhesion," or "Wa."

Surface tension or interfacial tension (IFT). IFT of liquid flux can play an important role in the soldering process. The IFT of alcohol-based flux is mainly dictated by alcohol; the IFT of IPA is around 22 mN/m. The IFT of water is significantly higher: 72 mN/m. This high IFT would not permit flux to work properly. Surfactants are usually added to the VOC-free flux formulation to lower the IFT.

The lower the flux IFT the higher its capability to spread through holes. Conversely, low IFT could favor the dilution of activator along the board surface. The IFT of VOC fluxes has been measured by Pendand drop.

Phases 2-4. In machine testing of fluxes, the experiments were similar in that they are searching for both flux performance and flux survival. They differed in machine description and alloy. The general customer base may chose either a SAC305 or a SAC405 alloy. Within the company, two different machines comprise the defined base for Pb-free work.

Some indications of the difference in performance due to the alloy are addressed in the next part of the experiment, which looks specifically at solder alloys.

In Phases 2, 3 and 4, the machine setup was always optimized and confirmed as good. In all cases, it was a double wave process. Any parameter outside the DoE was made as uniform and controlled as possible. These phases of testing are designed to take the limited population of flux candidates and rank their performance.

Table 1 shows the specific structure of the Phase 2 DoE. Phase 2 testing was conducted with a SAC305 alloy and a standard Pb-free solder module configuration.

Phase 3 testing continued to look at flux performance and survival, but the amount of flux applied was significantly reduced to further stress the material. Table 2 shows the specific structure of the Phase 3 DoE. Phase 3 testing was conducted with a SAC305 alloy and an enhanced Pb-free solder module configuration.

Phase 4 testing continued to look at flux performance and survival. The amount of flux applied was significantly reduced, to further stress the material. The wave forms were changed to duplicate the machine in Phase 2. Table 3 shows the specific structure of the Phase 4 DoE. Phase 4 testing was conducted with a SAC305 alloy and the standard Pb-free solder module configuration.

Phase 5. In the given project, questions of alloy acceptability or voiding were not a subject of investigation or discussion. Consequently, microstructure analysis has minimal value in this work.

The majority of analysis is visual inspection and inspection under production scopes. Hole fill, solder balls, solder bridges, excess solder, cosmetics and joint shape per IPC-A-610D are applied. Cosmetics are a secondary issue, compared to first-pass yield, given approximately the same visual appearance.

Solder Machine Preheat Testing

For the purpose of Pb-free wave soldering, equipment features that could serve to reduce insufficient hole-fill defects were identified for evaluation.

The value of improved preheat is one of those features. Specifically, in wave soldering, less-effective board preheat creates a greater thermal challenge for the molten metal moving up a PTH to make a top fillet. In Pb-free wave soldering, the process superheat is about 50% that of the SnPb process. All the chemical and thermal requirements to make a full solder joint in a limited amount of time cannot be met without more effective preheat of the sample board.

This study reviewed IR and convection heat units, in top and bottom locations, for contribution to improved machine yield. This study was extended to include cumulative impact of heaters in mixed configurations.

The equipment was: wave solder machine (manufactured in May 2005); solder module configuration was an "A" wave form; traversing head spray unit flux module; variable speed finger conveyor; Table 4 shows the preheater type and configuration.

Test and analysis materials and equipment included an ECD Super Mole Gold; 30 AWG, type K thermocouples (T/C); 24 AWG, type K T/C; aluminum, Kapton and high-temperature solder for T/C attachment.

Test Vehicle A (TVA) was a 0.130" thick unpopulated large panel. Five T/C were spaced uniformly across the diagonal. The T/C tips are trapped under torque-set flat washers/screw/nut hardware. The design goal was to measure area uniformity and heat transfer. It was modeled after a BTU oven calibration control sample.

Test Vehicle B (TVB) was a 0.062" thick sample product, populated and previously soldered. It was used to demonstrate reaction to shadowing and physical size differences. T/C attachment method was high-temperature solder on legs and pads.

Test Vehicle C (TVC) was a 0.062" thick sample product, populated and previously soldered. It was used to demonstrate reaction to shadowing and physical size differences. T/C attachment method was high-temperature solder on legs and pads.

Test outline. The evaluation outline, with a description of test and abbreviated results:

1. Opposing element thermal control limits.

  1. While the temperature set point of one of the paired heaters was not varied, the temperature set point of the opposite heater was changed in 20° increments.

  2. Temperature set points were run, with a conveyor speed of 3 fpm, for a minimum of 20 min., as a stabilization allowance.

  3. Temperatures were recorded after 20 min., or until the actual temperature of either heating element exceeded the factory set tolerance range.

  4. After running the tests with the bottom unit as the constant, the top unit was set as the constant. The bottom unit became the variable, and the above outline repeated.

  5. In summary, the IR pair operated with a 500°C max. limit, and the convection units had a 200°C limit.

2. Single pair of heaters and thermal performance on TVA.

  1. Conveyor speed was constant at 3 fpm.

  2. Three combinations of temperatures were used in the paired heaters.

  3. The convection heater combinations were:

    • Bottom at 120°C vs. top at 120°C.

    • Bottom at 120°C vs. top at 140°C.

    • Bottom at 140°C vs. top at 120°C.

  4. The IR heater combinations were:

    • Bottom at 250°C vs. top at 250°C.

    • Bottom at 250°C vs. top at 350°C.

    • Bottom at 350°C vs. top at 250°C.

  5. The temperature seen on the board, the rising slope of temperature and the temperature difference between the five T/C are the points of evaluation.

3. Combined thermal performance on TVA of paired convection and paired IR.

  1. The same temperatures as noted above were used.

4. Paired convection heater performance on TVB.

  1. This test applied the same temperatures as noted, but the test vehicle presented variations in color, size, thermal mass and air flow to the test panel.

5. Paired convection heater performance on TVC.

  1. Same as above.

  2. A second "production similar" test board was used to confirm the first set of results and search for performance inconsistencies.

6. Paired IR heater performance on TVB.

  1. This test applied the same temperatures as noted, but the test vehicle presented variations in color, size, thermal mass and air flow to the test panel.

7. Paired IR heater performance on TVC.

  1. This test applied the same temperatures as noted, but the test vehicle presented variations in color, size, thermal mass and air flow to the test panel.

  2. This is a second "production similar" test board, and it was used to confirm the first set of results and search for performance inconsistencies.

8. Combined convection and IR performance on TVB.

  1. This test applied the same temperatures as previously identified but in the combined configuration.

9. Combined convection and IR performance on TVC.

  1. Same as above.

  2. This is a second "production similar" test board, and it was used to confirm the first set of results and search for performance inconsistencies.

10. Thermal recovery and capacity in a five-panel test.

  1. The purpose of the test was to show that when a production load is put on the machine, there is enough capacity and recovery in the heaters to give the same profile to each panel along the length of the conveyor (that is, a loaded machine is not prone to "cold spots" because it is filled with product).

    • Basically, a series of boards were profiled as a group. A significant profile difference between the individual boards is cause for review.

    • Each panel had three T/C sites.

    • For this test, long T/C wires are required.

  2. There were a minimum of three T/C per board.

    • Lower left corner, 1" along diagonal from the corner.

    • Center of diagonal line drawn across panel.

    • Top right corner, 1" along diagonal from the corner.

    • All T/C are attached by the aluminum foil method.

    • All T/C are the same AWG.

    • All T/C are 30 AWG, unless otherwise specified.

  3. The test used a minimum of two conveyor speeds and the three temperature range settings noted above.

  4. With the machine stabilized for 20 min. prior to the test, the five panels were loaded onto the conveyor at approximately 12 to 15" spacing.

  5. A single profile run showed the temperature of a given on each panel along the length of the conveyor, satisfying the outline in Table 5.

Alloy Impact on Process

The consideration of different Pb-free alloys for wave soldering raises the question of how the alloy impacts the process. In a similar vein, as production moves from alloy A to alloy B, it would be valuable to know how much process knowledge is transferable between these alloys. These are the points that drove this experiment section.

The goals for this section were to standardize the entire wave solder process, and manipulate only pot temperature and preheat, for three Pb-free alloys. The alloys studied were SAC405, SAC305 and a proprietary low-silver SAC alloy. After measuring the defect rates from the different DoE steps, a limited confirmation study was run. The DoE work and confirmation work data were then compared.

The equipment included a wave solder machine (manufactured in Feb. 2005). The solder module configuration included turbulent and smooth wave forms. The solder module was a "roll-out" assembly permitting for easier alloy changes, and it was the only change that occurred between alloy tests. The solder nozzle assemblies were the same in all cases. They were moved to the alloy pots and reinstalled to exact parameters by use of hard tooling. The flux module was a traversing head spray unit. The conveyor module was a variable speed finger conveyor.

Test and analysis materials and equipment included an ECD Super Mole Gold; 30 AWG, type K T/C; aluminum, Kapton and high-temperature solder for T/C attachment.

Test Vehicle A was 0.062" thick x 6" x 8" with four test components (two DIMM connectors at C3, C2; 0.100" grid at C1, with leg-stick-out under 0.050"; 0.100" grid at C4, with leg-stick-out near 0.125"); three T/C spaced uniformly across diagonal for profiling. TVA had two internal layers of 3 oz. copper, with 0.5 oz. surfaces.

Test outline. The test matrix was a simple, three variable DoE format (Table 6). The correlation of conveyor speed to contact time was: 1.75 fpm = 7.5 sec. of contact time; and 4.5 fpm = 3 sec. of contact time.

The confirmation run was performed at the setting of best yield, based upon field inspection of the collection of DOE samples. Confirmation runs include OSP and Pb-free HASL boards.

In the analysis of samples, visual inspection recorded solder shorts, excess solder, insufficient solder and hole-fill by component type. Solder balls were reported for the entire board. Cosmetics were noted for the assembly. A Nicolet Imaging NXR-1400I system was used for hole-fill confirmation. The reference settings: power at 30; KVA at 80-110; board held at a constant 45° angle.

Results were interpreted with IPC-A-610C and IPC-A-610D; x-ray; main effect plots of DoE and confirmation runs; box plots as above; one way ANOM as above; and interaction plots as above.

Results

Flux evaluation. A VOC no-clean formula with 6% rosin exhibited the fewest solder defects and moderate flux residue. There was no discernable difference when used with either SAC305 or SAC405. (The flux selection work was performed with only two alloys.) Performance was uniform across all temperatures and contact times. The enhanced Pb-free wave solder configuration gave results measurably better than the standard configuration, and flux residues were less obvious.

In the VOC-free, no-clean flux formulas, the resistance to bridges was slightly less than the VOC category leader, but cosmetics were generally better.

Preheater study. This preliminary study showed that paired convection preheaters perform better than an IR-convection pair. Paired (a term used to identify a topside preheater located directly above a similar bottomside heater) convection preheaters operated with good predictability and excellent efficiency. The individual top convection heater was superior to the individual top IR heater. When used prior to a paired set of IR preheaters, the net effect was superior to a single top-side IR preheater.

Table 7 shows the influence of a topside heater vs. the opposite bottomside heater in achieving and holding set-points.

Table 8 shows the influence of top heater vs. bottom heater in achieving and holding set-points.

Figure 3, a representative profile of a top convection preheater, shows a rapid and uniform effect on the board, with a minimal temperature difference across the panel. Set-point temperatures were approximately 50% of IR.

Figure 4, a representative profile of a top IR preheater, shows a slower and nonuniform effect on the board, with a wide temperature difference across the panel. Set-point temperatures were approx. double that of the convection system.

Alloy study. In the course of the study, there was a reoccurring situation with large voids, especially in a single drilled hole size. It was explored, and the following is presented as points for future study of bare board quality standards, MSL of bare boards and process parameters.

In summary, a minor imperfection in a PTH, in conjunction with a hydroscopic laminate, and a solder contact time over 3 sec., at a 265°C pot temperature, resulted in large voids. Figures 5, 6 and 7 show the existing standard per IPC-A-610, and a sample void found in the study.

Figures 8 to 24 show a statistical representation of the alloy interaction study.

Conclusions

The flux evaluation methodology was successful for the conditions applied. A large field of candidates was put under evaluation, increasing the opportunity to find the most appropriate candidate; the laboratory test process reduced the field of candidates quickly, with low cost, and was both impartial and repeatable; a minimum number of test vehicles was required for actual machine work; the machine time used for the testing was short; the program can be modified to suit business needs.

For preheat, given the rules of solder joint formation that demand a time-temperature threshhold to be met, the performance improvements seen in the test are desirable. Compared to other preheaters, the set-points on the convection unit are generally lower. From the profiles, three attributes in convection heaters were noted: 1) the assembly reacts to the thermal energy more quickly; 2) in a shorter period of time, the assembly has a lower delta t than the alternative methods; 3) convection heaters are less sensitive to color or thermal mass. Given the study detail and results, top convection heaters are identified as a process advantage.

The alloys tested were: SAC405 (Alloy 1), SAC305 (Alloy 2) and a proprietary, low silver SAC (Alloy 3). The ANOM charts (Figures 25 to 27) of the excess solder data show, except for one point, the study results to generally be within the anticipated range of statistical error. In effect, this supports the statement that there is not a significant difference in the process characteristics of each alloy.

In Figure 25, the response in OSP is the single point that does not follow the trend. For those test conditions, it cannot presently be explained. All samples were the same lot, and the same OSP. An extended study is suggested. Presently, a single, small study cannot be considered a process standard.

While one ANOM chart supports no statistical difference between alloys, the analysis of main effects and interactions for solder balls and hole-fill are of interest.

The large voids found unexpectedly in this study may be associated with bare board specifications that are not adequate for Pb-free wave solder.

As unexpected as the voids were, another unexpected finding was the total lack of defects on C4. C4 was a 0.100" grid pin header. The through-board stick-out was purposely made very long, to 0.175", to encourage solder bridges. There was never a bridge on this connector, while there was often a bridge on the exact same part that had a stick-out of about 0.050". In another study, the leg stick-out was about 0.090", and bridges occurred.

In discussion, there are other ways to interpret this study, and one is to look at all test conditions, and make an alloy selection based on the alloy with the largest number of test runs that have few defects. This logic suggests that Pb-free HASL surface finishes may have a place in Pb-free wave soldering.

Future work will include a study of defects by alloy and surface finish, with larger sample sizes; explanation of the large voids formed after exceeding a minimum contact time and solder pot temperature; process characterization in board thicknesses over 0.135", with mixed component loading and variations in thermal mass; a comparison of selective soldering machines, with Pb-free alloys, to the optimized Pb-free wave solders.

Ed.: This article was first published at IPCWorks in November 2005 and is used here with permission of the authors.

 

Richard A. Szymanowski, senior process developing engineer, D. Casati and E. Saglia are with Celestica Corp. (celestica.com). Paul Lotosky is global director, technology implementation at Cookson Electronics (cooksonelectronics.com); plotosky@cooksonelectronics.com. Keith Howell, product marketing manager, and Gregory Hueste are with Speedline Technologies (speedlinetech.com).

Submit to FacebookSubmit to Google PlusSubmit to TwitterSubmit to LinkedInPrint Article
Don't have an account yet? Register Now!

Sign in to your account