The reflow process window for tin-lead solder is relatively wide. The melting point of tin-lead solder is 183 C. The lower temperature limit for reflow is 200 C. The upper limit is generally 235 C, which is the maximum temperature that most components can be exposed to. These high and low temperature limits provide a process window of over 30 C.
The most common lead-free solder alloy in the United States and Europe consists of 95.6 percent tin, 3.7 percent silver and 0.7 percent copper. This alloy has a melting point of 217 C. According to current practice, it needs a minimum reflow temperature of 240 C to ensure good wetting. The maximum reflow temperature is 260 C. This leaves a process window of only 20 C.
Although some components may perform satisfactorily after being exposed to a temperature of 260 C, there are many compelling reasons to limit reflow temperatures. Curtailing reflow temperature minimizes thermal stress on boards and components, reducing the potential for manufacturing defects. High temperatures can put significant stress on plated through-holes and barrels, which can lead to cracking. High first-pass temperatures on double-sided assemblies increase the amount of second-side oxidation, which can cause solderability problems on the second pass. Limiting peak temperatures reduces intermetallic growth, especially on bottom-side solder joints, which are exposed to two reflow passes. It also limits the potential for "popcorning" of components with high moisture content.
The problems related to a narrow process window are exacerbated by complicated assemblies with high component densities. Finding a profile that will reliably reflow a large assembly with high temperature differentials across the board has never been easy. The increase in peak temperatures, combined with components of decreasing size and robustness, means precision tools are required to find profiles that will safely process products using lead-free solders.
The current method of profiling reflow processes is to attach thermocouples to a product, and, using a wireless device or data-logger, run the device and the product through the oven to record the thermal profile. There are several problems with the status quo: Profiling is time-consuming; the software is complicated; and oven setup is a matter of trial and error and experience. With a tight lead-free process window, it will be difficult to find an acceptable profile using this technology.
In developing a thermal profile, the critical factors are the size and weight of the assembly, the density of the components, and the mix of large and small components. Generally, the greater the contrast in component densities, the tougher it is to develop a profile. This is because small components heat up more rapidly and reach higher temperatures than heavier ones. Temperature differentials across the board are a critical factor in determining whether it can be successfully processed with lead-free solder.
With a standard profiling system, it is relatively simple to make the transition to lead-free solder if the board is small and has components of consistent thermal mass. More complex assemblies with a large mix of components offer a more difficult task. The temperature differential across the board may push lighter components past their thermal stress limit, while the densest components may fail to reach an adequate wetting temperature. As a result, complex boards will require new technology to develop a profile that fits the narrow process window.
Reflow Process Monitoring
Automated reflow management systems that combine continuous statistical process control (SPC), line balancing, documentation and production traceability into an integrated software package have recently been introduced. These systems automatically feed real-time process data to engineers and managers, allowing them to make critical decisions affecting production costs and quality. These systems gather real-time thermal process data on every product processed, as opposed to the conventional practice of only periodically checking oven performance. This allows the systems to catch potential defects before they happen, rather than discovering actual defects in inspection. It also eliminates the need for process verification profiles, and it provides real-time feedback and alarms for zero-defect production.
Automated reflow management systems are able to verify the profile of every board through a "virtual profile." A virtual profile is created by running a baseline profile of the product with a real-time profiler, while simultaneously collecting real-time data from thermocouple probes in the oven. The mathematical correlation between the temperatures at the product level and the temperatures on the product itself allows the software to accurately simulate changes in the product profile. Once a virtual profile has been established, the system can simulate, in real time, how the product profile is changing based on probe readings. Process temperature or airflow cannot change without affecting product temperature. Based on the virtual profile, the software's algorithms accurately extrapolate changes in the product profile from changes in process temperature.
Once the virtual profile has been established, the system automatically begins generating SPC data. Every time a board exits the oven, the data set is plotted on frequency histograms. Process data is charted for all critical process specifications, including peak temperature, soak time and time above liquidus. The data is plotted on real-time control charts, and process capability (Cpk) values are calculated for each specification. Any process drift outside of control limits triggers an immediate alarm. Real-time Cpk tracking enables the system to flag an out-of-control process before the oven has produced a single defect.
Testing the New Technology
We decided to test the ability of an automated reflow management system to handle circuit boards assembled with lead-free solder.
Two test boards were used. The first, dubbed "flex board," had a high thermal mass. Used in heavy equipment, it consisted of a flexible substrate attached to an aluminum plate with a high-density connector. This board had components ranging from 0603s to a large quad flat pack (QFP). The second test board consisted of six panelized cell phone boards. This board contained components ranging from 0201s to small QFPs and a micro ball grid array. The two test vehicles were vastly different in thermal mass, component size and component density.
The experiment was divided into two phases. The first phase had two objectives: to compare the automated profiling technique with the conventional trial-and-error method, and to verify that a peak temperature of 242 C could successfully reflow lead-free solder paste. The goal of the second phase was to see if the peak temperature could be reduced to 232 C or less and still produce quality joints. Both boards were used in phase one, but only the cell phone board was used in phase two.
In phase one, a separate profile was generated for each test board based on a defined process window. Two flex boards and five cell phone boards were built per profile. The starting point of both profiling techniques was a standard tin-lead profile with a peak temperature of approximately 220 C. From there, profiles with peak temperatures of 257 C and 242 C would then be generated. The ramp rate was less than 1.5 C per second, and the time above liquidus (217 C) was limited to 30 to 90 seconds.
Our study found that the software-predicted profiles took fewer iterations and less time to produce than the profiles generated by trial-and-error. The software needed two iterations to go from the standard tin-lead reflow profile to the 257 C profile, and then one iteration to go from the 257 C profile to the 242 C profile. The trial-and-error method took five iterations to go from the standard tin-lead profile to the 257 C profile, and then three iterations to go from the 257 C profile to the 242 C profile. Overall, the software reduced profiling time by more than half compared with the trial-and-error method. Profiles for the cell phone board were generated faster than the ones for the heavier, more diversely populated flex board.
Optical inspection was performed on every board reflowed at 242 C and 257 C. No wetting defects were found. Some were defects found in the assemblies produced with the trial-and-error profiles. These defects were mostly tombstoned 0201s caused by the higher ramp rates near the peak zone.
Cross sections of some components were performed so that scanning electron microscopy (SEM) and energy dispersive spectroscopy (EDS) could be used to evaluate wetting. All the passive components that were evaluated at the peak temperatures 242 C and 257 C showed even and full wetting to both the pad and the component termination. There was no noticeable difference in solder joint quality between these two peak temperatures.
Examinations of the leads on small-outline integrated circuits and thin, small-outline packages showed similar results. As with the passive components, good wetting was achieved on all leads. Full wetting was seen on both the substrate pads and the component leads. The leads were completely wet from the toe of the joint all the way to the component body. SEM analysis was also done to analyze the intermetallic region at both the component terminal and on the board. No real differences were found in the intermetallic regions based on reflow temperature.
How Low Could We Go?
In phase two, we tried to determine the lowest possible peak reflow temperature that would form quality solder joints with the tin-silver-copper alloy. The peak temperature was initially lowered from 242 C to 232 C. It was then lowered in 4 C increments to a final limit of 220 C.
Five cell phone boards were assembled at each peak temperature. Yield and solder joint quality were measured for each change that was made to the profile. Shear tests and cross sections were performed to determine joint quality of various components.
Overall, we found that solder wetting occurred at all six peak temperatures to varying degrees. A noticeable aspect of lead-free solder is that it will not flow out to the edges of the pad. As a result, when small stencil apertures are used, it is unlikely that the solder will fully wet out the entire pad. This phenomenon is independent of the peak reflow temperature. In contrast, standard tin-lead solder wets out the entire length of the pad.
Shear testing was done on a sampling of 0201 components to determine the joint strength of the lead-free solder. Comparison data was taken from a sample of tin-lead components as a benchmark for the lead-free samples.
Our tests showed that the lead-free assemblies had higher average shear strengths than the tin-lead assemblies. Even 0201s assembled with lead-free solder at a peak temperature of 224 C had higher average shear strength than 0201s assembled with tin-lead solder. However, only the lead-free board reflowed at 257 C showed statistically different strength results from the tin-lead board. The average shear strengths at the other different temperatures fell within the standard deviations of each other.
As in phase one, optical inspection was performed to check for defects and evaluate solder joint quality. Sample parts from each peak temperature were then cross-sectioned, and SEM analysis was performed.
No noticeable defects were found on the boards reflowed at a peak temperature of 232 C or 228 C. Some minor wetting defects were found at 224 C. A greater number of defects were found at 220 C. This was not unexpected, as that temperature is only 3 C above the liquidus temperature (217 C) of the tin-silver-copper alloy. Small solder balls were found attached to wetted solder joints, indicating all the paste did not melt.
Cross-section analysis of the components showed that incomplete wetting could be seen on components assembled at temperatures of 228 C and below. At 232 C, component wetting was just as good as it was in phase one. Below 232 C, wetting decreased as the peak temperature decreased. At 228 C, the solder did not completely wet down the pad, although there was full wetting to the component termination. The worst wetting was seen at 220 C.
Although almost complete wetting could be seen at the various peak temperatures tested in phase two, differences could be seen. The most noticeable differences in wetting could be seen in components reflowed at 228 C and below.
From our analysis of boards built in phase two, we believe that good tin-silver-copper joints can be formed outside of the manufacturers' recommended peak temperature range. But, a large reduction of peak temperature can lead to excessive solder balling and incomplete wetting of the solder to both the pads and components. From SEM analysis of the boards, we recommend a peak reflow temperature of 232 C or above to achieve reliable wetting of the pads. Although joints can be formed at lower peak reflow temperatures, even down to 220 C, this may lead to problems. Large parts may not reach a temperature above 217 C, the liquidus point for tin-silver-copper paste, and therefore may be subject to insufficient wetting and reduced reliability.