Spend some time on the SMT line, and you’ll know why.

Automated solder paste inspection, a must for most PCB assemblers, appears to use ridiculously wide specification limits. Most SPI program generators read a stencil Gerber file and set the target paste deposit volumes to 100% of the apertures’ theoretical volumes, with tolerances of +/-50%, unless otherwise directed by the programmer. At first blush, +/-50% of nominal seems like a ridiculously enormous tolerance!

An SPI expert at a top EMS firm offers this perspective: Imagine reading a blueprint that calls out a drilled hole with a half-inch diameter with a quarter-inch tolerance. A half-inch hole that can run 0.250" to 0.750" sounds absurd. And it is. But to a seasoned SMT process engineer supporting a healthy print process, the world is a little more complex than a singular mechanical tolerance. We live in a world full of noise, and our job is not necessarily to eliminate the noise, but to figure out how to maintain robust processes despite it.

Our expert suggests we start by considering the basic setup of our PCB printing process. Examining the tolerance stackup on the surface of a PCB can be eye-opening. Solder mask is nominally 0.001", plus one and minus a half. It’s not uncommon to see mask higher than the pads when non-HASL final finishes are used. Silkscreen is often 0.001" or higher, and if the ink is extra sloppy, it can easily ruin a good print process. If a HASL finish is used, there’s another 0.002" or more of potential variation to the PCB topography.

Now consider the board support tooling: It might be pins that leave locally unsupported gaps; it might be the conforming and flexible type that provides “uniform” support, but just can’t compensate for heavy warpage or larger components, or it might even be the gold standard: a custom-tooled, rigid vacuum fixture. But even they aren’t perfectly planar, and they’re susceptible to the buildup of errant solder paste. So, no matter the type of tooling used, there’s going to be a little slop associated with it too.

Add all the possible sources of tolerances: a mil here, a couple there, and compare them to a nominal paste height of only 0.005". Luckily, all the tolerances don’t usually add together in worst-case scenarios, and even RSS calculations seem a little high with respect to the variation that is actually observed. To allow for 0.001" topographical variation in the PCB equates to 20% of a 0.005" print height. This seems to be a reasonable (if not conservative) estimate of the variation contributed by the mechanical setup of the PCB in the printer.

This 20% does not yet reflect the natural variation contributed by the paste, the environment, the stencil, the printer’s alignment and motion, or the influence of the operators. Considering the number of jokers in this deck, an overall starting tolerance of +/-50% doesn’t sound quite so outlandish anymore, does it?

I like this explanation not only because it effectively articulates the basis for the starting tolerances to someone who’s new to stencil printing, but also because it educates them on the many unseen challenges within the process. To those who really understand printing, however, haphazardly applying an automatic +/-50% on everything seems incredibly oversimplified, even borderline irresponsible. Consider BGAs, for example: On large ones, you make sure there’s enough paste; on small ones, you make sure there’s not too much. How can the same spec and tolerance apply to both? And that doesn’t even consider that the small BGA apertures have tight area ratios and normally release less than 100% of the aperture’s volume. Obviously, these “starter tolerances” need some adjustment.

Unfortunately, there’s no handbook on how to set print volume specs, and there is an element of risk in setting them improperly. Too tight and they slow production; too loose and they permit escapes to create yield loss. My quest for the best guidelines on setting tolerances revealed several different approaches, each with its own strengths and weaknesses, as noted in Table 1.

If any of these approaches were completely effective on their own, they could be coded into a software tool that would automatically set inspection parameters, and there would be far less discussion on the subject. But each strategy comes up short in at least one area; hence the need to consider multiple factors that include a package’s typical soldering defect modes, the reduced paste release on area ratios under 0.8, and the pain of rework in order to determine the optimum values for mean volumes and control limits.

Process owners should not let concerns about setting a bad tolerance prevent improvements. When the starting point is a universally applied theoretical aperture volume +/-50% on everything, any parameter that’s got some analytical thought behind it is bound to be better than the catchall setting from the equipment manufacturer. Maybe +/-50% is the best volume tolerance for certain component types; maybe it isn’t. Many tolerance bands remain at 100%, but get shifted left for devices prone to bridging and right for devices prone to coplanarity issues or head-in-pillow. Others bands are tightened for leadless devices that require a high degree of volume consistency around their perimeters. And yet others keep the +/-50% tolerance band, but reduce the target volume to adjust for lower paste transfer on smaller apertures.

There are no hard and fast rules to setting SPI specs and tolerances, just guidance based on a mixture of process knowledge, statistics and a bit of trial and error. Who’s the best-qualified to set the specs? Usually the engineer or technician closest to the process. How did they learn the craft? Experience. On an SMT line, it is always the best teacher.

Chrys Shea is founder of Shea Engineering Services (sheaengineering.com); This e-mail address is being protected from spambots. You need JavaScript enabled to view it . She wrote this article on behalf of Christopher Associates (christopherweb.com).