Optimizing Pb-free SMT process parameters: a Taguchi approach reveals the most important process steps and best settings for screen-printing and reflow, as well as the best combination of paste and board finishes for lead-free soldering.

Ed.: For the complete paper, including test methodologies and data results and analyses, and a list of references, please visit circuitsassembly.com/online/0501/0501napoli.shtml.

[TEXT NOT REPRODUCIBLE IN ASCII]

Many investigations have focused on a suitable replacement of the SnPb eutectic solder alloy. As no drop-in replacement for tin-lead solder is capable of maintaining the necessary combination of mechanical properties, cost manufacturability, availability and reliability, a majority of the industries have decided to replace it with SnAgCu alloys. To successfully include the new material into existing processes, investigations were performed to obtain an in-depth understanding of the material's properties.

Several studies and experiments have been conducted on specific SMT processes. Some were aimed at optimizing process parameters for printing or the reflow heating curve, others were aimed at evaluating joint quality when lead-free paste was used in existing processes. The quality of solder joints likely depends on both materials and processes, and it is not easy to set control variables to obtain a desired process. To date, no comprehensive investigation on materials and assembly processes, or transition issues due to lead contamination, has been published.

The objective here is, to a certain extent, to fill this gap, presenting the results of a comprehensive experiment aimed at optimizing control parameters of manufacturing and at improving solder-joint quality. Once the most widely used paste was identified and its physical and rheological properties verified, an experiment based on the Taguchi approach was carried out to establish the optimal printing parameters. A practical technique is proposed to set the control variables of a reflow oven, which ensure the desired heating curve inside the oven. Then, using the obtained printing parameters and the developed setting technique, the relationship between solder-joint quality and materials used was investigated. To this end, a full factorial experiment was conducted on the entire soldering process to determine the best combination of paste/board finishes and to evaluate the effect of nitrogen ambient and lead contamination. The results obtained for the lead-free paste were compared with a control sample consisting in standard tin-lead paste and finishing.

As the physical characteristics of solder paste have a direct impact on the main processes of SMT, tests were performed on tin-lead and lead-free pastes used for process experimentations. The tested solders have spherical particle shape (particle size composition of Type 3 according to IPC norms) and no-clean flux. In particular, the tin-lead paste is Sn63Pb37 with 90.5% metal content, while the lead-free paste is Sn95.5Ag4Cu0.5 with an 89.5% metal content.

[FIGURE 1 OMITTED]

Joint Quality Evaluations

An experiment on the entire soldering process was performed in order to evaluate the quality of the joints. It was conducted using a test coupon produced by Heraeus (Benchmarker II) to highlight the mechanical characteristics of the joints as well as wettability and solder ball formation. The process parameters used in the experiment are those obtained through our earlier experiments [see online version] as the best settings.

[FIGURE 2 OMITTED]

Wettability and solder balls. The test coupon has four areas that are used to evaluate the global wetting index (WIG) of the solder paste. In the first area solder paste is printed with pitches varying from 0.010" to 0.085", and wettability is measured by the number of bridges after reflow (WIA). The other three areas have 14 pairs of pads, circular or rectangular in shape. Paste is printed starting from a 1:1 ratio and then on smaller and smaller areas of the pads, and wettability is measured by identifying the number of pads fully covered by paste after reflow (WIB, WIC, WID). The WIG is a weighted mean, assuming values in the range 0-14.

[FIGURE 3 OMITTED]

As process variables, we considered board finishes and reflow ambient, and kept the temperature constant during the experiment. The results obtained for solder balls were very good both for tin-lead and lead-free finishes. The results show that traditional tin-lead paste has a higher wettability compared to lead-free paste and that better results are obtained in inert ambient.

Mechanical characteristics. Solder joint reliability depends on the solder alloy and the materials used for board and component finishes. In the transition between tin-lead processes and lead-free processes it is foreseeable that leaded and lead-free materials will be used together. To investigate their interaction, we considered as process variables the board and component finishes besides the solder paste and the reflow ambient. For component finishes, solder paste and reflow ambient we chose two possible values, while for the board finishes we chose four possible values. Therefore, in order to test all of the possible combinations we need 32 tests, each one repeated 10 times to reduce the noise influence, leading to a total of 320 boards.

Two types of mechanical loading tests were performed on the boards: pull and test. In the pull test, performed on 208-pin BQFPs, three lines of pins of the tested component are cut, leaving just one line of pins. The component is bored in the center and is pulled by means of a hook inserted in the hole, until breaking is reached. The pull test was performed on four of the 10 boards used for each test. The shear test, performed on SOICs and BGAs, consists in applying a shear stress on the face of the component which does not have pins. In particular, as BGA components are square, shear stress has been applied into both possible directions. The shear test was performed on three of the 10 boards used for each test.

In almost all the mechanical tests solder joints held out over the breaking values of the pads.

Tombstoning. Tombstoning mainly affects small-size chips (0402 and 0201), which lift one terminal, thus changing their position from horizontal to vertical. The test coupon, Bench-marker II board, which is designed to study this effect, presents pads with different geometries which are designed to be covered with different solder paste geometries. With reference to pad and solder paste geometries, we observed that the number of tombstones increases with the distance between coupled pads.

In order to study the effect of the process variables regardless of pad and solder paste geometries, using Taguchi's method we performed the ANOVA on the results. The higher values of the F statistic obtained for the component finishes and the reflow ambient implies that the tombstoning effect mainly depends on them. Moreover, a lower number of tombstones are recorded for tin-lead component finishes and air reflow ambient. The better behavior in air is due to the fact that nitrogen enhances the difference of wettability at the ends of the components, while the better behavior of tin-lead paste is due to the higher surface tension of lead-free paste and the larger temperature differences caused by more inhomogeneous heating in the case of lead-free solder.

Voiding. X-ray inspection of all BGA components was performed to detect defects of the solder joints. No defects were noted with the tin-lead paste, while some voiding was observed with the lead-free paste. The impact of voids on quality and reliability of solder interconnection has not been investigated in detail yet, although some earlier studies pointed out that reliability increases for joints that contain voids up to a certain size.

Summary

Taguchi analysis of the results determined the best settings of the printing process and a new procedure for determining the optimal setting of a reflow oven. Differences between physical characteristics of the Sn95.5Ag4Cu0.5 paste and the eutectic SnPb paste imply that optimal settings are different. In particular, lead-free optimal printing parameters are a higher squeeze speed and an increased minimum soldering temperature up to 240[degrees]C, owing to the lead-free paste's higher melting point.

An experiment on the entire assembly process was performed and analyzed using the Taguchi approach, obtaining results on the dependence of solder joint quality on component and board finishes and on the reflow ambient. In particular, we highlighted the following points:

* The higher surface tension of lead-free paste increases the tendency toward tombstoning, which can be kept under control only through correct footprint design.

* Nitrogen ambient increases wettability but, in combination with lead-free paste, increases tombstoning significantly. Chemical tin appears to be the board finish showing the best cost/wettability performance. No evidence of lead contamination affecting joint quality was found in any of the various combinations investigated.

* X-ray inspection of BGA components soldered using lead-free paste exhibits higher levels of voiding. This phenomenon, even if within IPC limits, needs further investigation to gain better insight into the process parameters that may influence it (i.e., reflow temperature profile, flux type, etc.). This investigation will be subject of further research.

TABLE 1: Shelf-life test results.

                      SnPb Paste  Pb-free Paste

After 1 day [kCps]       700           524

After 7 days [kCps]      768           624

After 30 days [kCps]     810           600

TABLE 2: Test Results.

Test Type               SnPb Solder Paste     Pb-free Solder Paste

Wetting balance               >2.2 mN               >2.2 mN

Solder balls                     OK                    OK

Tack time (min/max      3.0-4.4 g/m[m.sup.2]  2.9-4.3 g/m[m.sup.2]
values on 8 hours time

Acknowledgments

The authors would like to acknowledge and thank the many unnamed contributors of IXFIN S.p.A. that made the test program possible. Special thanks to F. d'Ambrosio, G. dello Stritto and R. Gravina for their support during the whole project.

Antonio Buonomo is professor of electronics at the Dipartimento di Ingegneria dell'Informazione, Seconda Universita di Napoli, Naples, Italy, and Alessandro Lo Schiavo is assistant professor of electronics. Fabrizio Rotulo is manager of Ixfin Engineering Division and Campus in Marcianise (Ce) Italy.