Is the slew of new materials, coatings and processes truly unique, or just the same old hype?
The SMT stencil market is burgeoning with new materials, coatings and cutting technologies, all of which promise to improve the solder paste printing process. Which of these options really help assemblers keep yields up and costs down, and which are nothing more than techno-snake oil born of marketing genius? I was able work on a fantastic stencil study this summer with Vicor’s (vicr.com) Ray Whittier to find out. We were holding the results close to our vest pending initial publication at the SMTA International conference in October. Now that details of the study are available, I’m happy to share my takeaways with Circuits Assembly readers.
The investigation took a comprehensive look at SMT stencils from a production perspective, which is extremely refreshing, since many studies happen in laboratories. It tested actual performance, using a production PCB design in a production environment. No laboratory simulation could possibly produce a better bellwether of real-life performance.
Four different suppliers, three stencil manufacturing methods, five materials and two foil thicknesses were put to the test on a highly miniaturized, densely populated production PCB. A single print contained roughly 8,500 µBGA deposits and 2,000 0201 deposits; all these were packed into a 3 x 7" area – about the size of a checkbook. With only 10 strokes of the squeegee, pretty significant sample sizes were generated. The print volumes were measured by an inline SPI system and subsequently analyzed statistically. Every bit of the test setup and results is detailed in the SMTAI paper, which is available for download online.
Here are some of the more interesting finds:
Nano-coating really does work. When we tested the two-part material, which gets manually applied to the stencil, we found that, in most cases, the transfer efficiency of coated stencils was no different from uncoated ones at 0.77 area ratios, and was actually lower on coated stencils at the 0.65 area ratios.
So if it didn’t improve transfer efficiency, what did it do for the process? It was like a shot of steroids for the print yields. We tested 13 pairs of “identical” stencils: one with coating and one without. Of the 13 pairs, seven of the coated stencils produced 100% yields, while only one of the uncoated produced a 100% yield. It improved yields on almost all stencil pairs, even the poor quality ones. That was enough to immediately implement the coating for production, where it has helped raise print yields an average of 5% across all product lines.
By what mechanism were yields boosted? We assume the coating performed as promoted with respect to encouraging stray solder paste to stick to the PCB instead of the stencil. In this test, we wiped after every print, so we have no real means of proving it. The higher yields combined with the slightly lower transfer efficiency on the µBGAs makes us think that it basically produced crisper print definition. Maybe we’ll go back and study that some more; maybe not. This is a production-oriented test, and our main takeaway is that it bumped print yields considerably; perhaps we’ll leave it to the lab folks to document the finer points of how and why.
Electroformed foils can exhibit unacceptable amounts of thickness variation. We had previously come to accept the notion that the electroforming process can produce a slight variation in foil thickness due to current density variations around the apertures, but what we found astounded us: When measurements were taken roughly 1.0" outside the print area (recall that it is only 3 x 7"), some electroformed stencils showed corner-to-corner thickness variation as high as 50%. Average variation from spec was as high as 25%. And it wasn’t limited to electroformed stencils with apertures! It was also observed on electroformed nickel blanks that were subsequently laser cut, which basically negates the oft-accepted argument about aperture density locally influencing foil thickness.
Not all suppliers submitted sloppy electroformed stencils or foils, which indicated to us that the process is controllable. We don’t pretend to be plating experts, but when one supplier’s process can produce a foil that’s dead nuts on-spec, while the other’s is as much as 50% off, we naturally suspect that the process and quality control methods are very different between the two operations.
The effect of an overly thick stencil on a miniaturized product can be devastating. In the case of the µBGA, a stencil only 15% over its 0.004" spec pushes the area ratio over the brink from the precipitous 0.66 benchmark down to 0.55. Twenty-five percent over drags it down to the 0.45 neighborhood – a place that no one printing type 3 paste wants to find themselves. The printing process is difficult enough to control; we really don’t need our suppliers throwing monkey wrenches into it for us.
Electroformed stencils can exhibit unacceptable amounts of positional variation too. When we compared transfer efficiencies among stencil types, we found some of the electroformed stencils were releasing far greater proportions of paste than stainless steel. We’d heard the argument that electroformed has better release than SS because it has smoother walls, but given our findings on thickness, we weren’t going to buy into this without a little more investigation. What else could possibly cause >100% transfer efficiency on 0.65 area ratios? Bad alignment, perhaps? When we dug into our SPI database, we found average positional offsets of 0.002" in at least one direction, sometimes in both.
Consider 0.010"-0.011" apertures and pads: A 0.002" offset in one or both directions can easily blow the gasket and permit a considerable amount of paste to pump out. We are assuming this is the reason for the unexpectedly high paste volumes, but even if that’s not the real root cause, we frankly don’t care. We just don’t want stencils with apertures 0.002" off target, and we don’t want stencils that supply 120% of the aperture volume when we are expecting 80%, because both are going to create unacceptable solder bridges and balls. Garbage in, garbage out.
What’s the source of the variation? The inferior positional accuracy of electroformed stencils is usually attributed to their manufacturing process. They are formed in a chemical bath on a flat plate, and when mounted and tensioned in their frames experience a little bit of stretch, whereas stainless foils are pre-tensioned in their frames when laser cut, so they don’t experience the same stretch. That’s the story we’ve been sold anyway.
Positional variation wasn’t limited to electroformed stencils. We found an unacceptable level on some nickel stencils that were laser cut, too. A pair of the laser-cut Ni stencils both produced positional offsets of 0.002" or more in the Y direction only. They were nearly perfect in X. The two other pairs of laser-cut Ni we received – one of which was from the same supplier – all averaged offsets of less than 0.001" in both directions. In fact, nine of the other 10 laser-cut (SS) stencils all showed variation of less than 0.001" in both directions, and the single outlier was only slightly over the line in one direction.
While it would appear that laser cutting generally provides better positional accuracy than electroforming, it clearly does not guarantee it. When one of eight pairs of stencils demonstrates such a unique but consistent deviation, the machine that cut them is naturally suspected as the special cause of variation.
Chrys Shea is founder of Shea Engineering Services (sheaengineering.com); firstname.lastname@example.org. She wrote this article on behalf of Christopher Associates (christopherweb.com).