Keith Bryant columns | September 18, 2006
Does PCB pad finish affect voiding levels in lead-free assemblies
With European environmental legislation and market forces driving the change to lead-free solders for printed circuit board assembly, there will be significant changes in the soldering process.
'Voiding' is one of the known major consequences of these process changes exaggerated by the use of new soldering materials. Excessive voiding can lead to poor quality joints, BGA 'popcorning', open-circuit failures during burn-in or thermal cycling; even field failures can be the result. Therefore voiding needs to be minimised to retain process quality and to reduce rejects and returns in the change to lead-free production. Much has been written about levels of voiding as we move from lead-containing solder pastes and component finishes to a lead-free assembly environment. It is generally acknowledged that this technology shift will exhibit an increased level of voids within the formed joints. Many papers have put forward theories for this and a huge amount of research has been published. From this, certain facts appear irrefutable: * Voiding increases with the use of tin/silver/copper, higher melting point alloys. * The increased surface tension of these materials adds to this phenomenon. * Moisture entrapment within PCBs and components becomes more of an issue with steeper temperature gradients and higher melting point materials. * Many lead-free solder pastes contain more aggressive flux chemistries than lead-containing materials; this often means higher volumes of gas has to vent through the joints. * Reflow profiles and cooling rates have a significant influence on voiding and the position of the voids. There is, however, much conjecture on the safe level of total voiding, the largest safe single void, the safe position of voids within a joint and much else. This article does not enter into this debate, it simply reports on findings and draws comparisons. Background This report brings together results from the SMART Group Lead-Free Experience 2003, where circuit boards with differing pad finishes were assembled over two days using exactly the same materials, components and conditions. Two reflow systems were used here; the results using 'convection reflow' have been included, and those using 'vapour phase' have not. This is simply to allow accurate comparisons. The circuit boards were all made by the same company and stored in the same way; everything possible was done to allow a 'level playing field'. The balled devices on these boards were then subjected to X-ray inspection using a high quality automated system (Figure 1). Figure 1. This screen capture shows the automated void calculation, the first number is ball diameter, the second total percentage of voids, the third gives the largest single void Pad surface finishes There are numerous pad surface finishes in the marketplace, many differentiated only by brand names, so we will work only with materials and methods of application. The most popular types were chosen for evaluation, some more suited to lead-free than others; a few notes have been added on this. The five finishes that follow represent a good cross-section of currently available technologies and will form a solid basis on which to evaluate the findings. OSP (organic surface preservative): This is simply an organic treatment applied over clean copper pads to prevent oxidation; it burns off during reflow to allow soldering to the copper. Due to the poor wetting of lead-free pastes bare copper can be left exposed after assembly. It is also not suitable for repeated assembly cycles. In its favour is that it is the cheapest and if the finish is damaged it is easy to remove, clean the pads and replace the finish. Immersion tin: By removing the oxide from the copper pads and processing the boards using this chemistry, a layer of tin is deposited over the pad areas. As tin is the major part of the solder, the metallurgy is very suited. However, there is the fear of 'tin whiskers'. These are thought to be formed when the tin is under stress and could cause shorts on the board. There are currently many debates on this subject. Immersion silver: This is an electro-less deposit of silver, similar to tin, but heavily promoted at the moment due to the 'tin whisker' issue. Its cost is similar to immersion tin and sits between nickel/gold and HASL. Electro-less nickel/immersion gold (ENIG): The mechanism here is slightly different; the gold provides a shield to stop the nickel from oxidising in air. When it is soldered, the joint is formed with the nickel, not the copper pad. It is the highest cost process, but provides a very flat, easy to solder surface. However, the chemistry has to be well-controlled or defects can appear. Lead-free hot air solder level (HASL): This is the type of board finish that was preferred for many years, then fell out of favour as Quad Flat Packs and other high I/O devices became common. At this time the board manufacturers could not give the assemblers a flat enough platform to mount these devices with this finish. The process is much improved today, giving a much better surface. But the move from leaded HASL to lead-free HASL could put more stress into the bare board, due to the higher temperature. This could lead to board warping or in severe cases, delamination of multilayer boards. Having said all this, it is a low cost option without any metallurgical mismatch issues. Void measurement and calculation The use of a fully digital system with 65 000 levels of greyscale allows voiding to be located and measured accurately. The algorithms give accurate, reproducible numbers for void percentage and largest single void size. The ability to view the balls from oblique angles through 360° allows the position of the voiding to be confirmed (Figures 2 and 3). At the start of this work the void position was not considered to be very significant but findings did change this dramatically. Figure 2. This is an oblique image of a BGA assembled with nickel gold pad finish exhibiting voiding. Figure 3. The angled views above, and single ball images, confirm that the void shown is at the ball-to-device interface Summary The total percentage void results is displayed in graphical form as a mean value of all of the measurements taken for each pad finish. Any results which varied more than two standard deviations from the norm were not included, as these would tend to point towards an issue with the process, component or bare board finish. One of these issues is covered at the end of this report. The largest single void results display the average size of the five largest individual voids for each type of pad surface finish. Average percentage total voiding: This chart indicates that the surface finishes, which exhibit the lowest percentage of voids, are immersion tin and lead-free HASL. This could be due to the affinity of the pad finish to that of the component termination and the solder paste. It could be construed that the two pad finishes with the highest percentage of voids are the ones which are likely to produce the most gas from the action of removing oxide from the surface prior to soldering taking place. It may be possible to draw more conclusions from these results, however there are many influences which can affect the levels of voiding and it may be misleading to look any deeper into these numbers. Current IPC Class 1 allows 30% of the ball area to be devoid of solid material, so all of these results could be considered as good. There is also a theory that voids inside a joint can act as stress relievers and reduce or diffuse cracks in the same way that rubber is used in castings and adhesives. Largest single void: These results show that the voids in OSP joints tend to be much larger and taken with the previous results suggest that this finish produces or traps the most gas in lead-free joints. It is generally accepted that small voids are potentially less of an issue than large voids, but there is conflicting opinion that any voiding is detrimental. It also points to the fact that the electro-less nickel/gold pad finish produced a quite high percentage of voids, which tend to be small in diameter. This could be due to the fact that the gas produced is of different composition, it takes longer to produce and is 'frozen' in the solder before it joins together to form larger bubbles. It is noticeable that these voids are more central or close to the joint face, whereas the OSP voids are often close to the joint open edge. The void profile of tin, silver and HASL is much smaller and therefore it could be concluded that these pad finishes are more suitable for lead-free assembly. Additional findings This work has highlighted variances in voiding which have been seen whilst using digital X-ray systems to inspect BGA assemblies and given some information to help in the choice of PCB pad finish for lead-free assembly. Whilst conducting this research another voiding phenomenon was seen for the first time, which has the potential to become a severe failure mechanism. It is not easy to locate and cannot be seen on many X-ray systems. As seen in Figure 4, it consists of a large number of very small voids formed at the pad to ball interface. Due to its position and the fact that all the voids lie in the same plane it could easily lead to an open circuit or high resistance joint after thermal cycling, burn-in or worse still, failure in service. Potential causes of this are oxidation of the pad finish, issues with the PCB manufacture or some form of intermetallic reaction. Figure 5 illustrates very clearly the problem and its location. It has been found on boards with OSP, nickel/gold and immersion silver pad finishes. It is seen infrequently and without a defined pattern. This would tend to point to a random failure mode, not simply a material mismatch. Needless to say, more evaluation is going on and it is hoped to report more findings soon. Figure 4. This shows a large number of voids all in the smaller area of the balls. This suggests that all the voids are in the pad-to-void interface. Oblique views confirm this to be correct Figure 5. The picture is an excellent example of this defect, not only does it show the voiding in the pad area, it is also visible on the track - clearly seen in the lower left-hand side of the image. By: Keith Bryant, Dage Group, UK