SMT is highly automatable, but must be monitored and controlled.

Better Manufacturing In process optimization, a question that often comes up is where to invest in AOI: post-reflow, post-placement or post-print? The answer is yes. (Perhaps the kind of answer you’d expect from a consultant, right?) Any SMT assembly process will benefit from in-process inspection at any of these points. But where should the initial investment in AOI go? I’ll answer that question with a question. Of the three key automated processes – printing, placement and reflow – which generates the most defects? Various studies have pointed to printing (although others have critiqued the methodology; see “The Printer Effect,” December 2007, pg. 20). Our evaluations of clients’ data confirm these findings – even in processes running 3-Sigma or better.

Essentially, post-print inspection systems come in two flavors: integrated into the printer, or as dedicated, standalone equipment. The latter can be in- or offline. Post-print inspection integrated with the printer has the distinct advantage of having access to the stencil and can quantify board pad coverage by solder paste, bridge detection (between pads) and stencil aperture paste buildup. Recent advances in have synchronized post-print inspection with other post-print functions, including stencil wipe and inspection performed on the offload. Regardless of whether the post-print routine is dedicated or performed synchronously with other functions, it does take process time. Inspection should not create a bottleneck for the process, and decisions should be made whether to use such systems to inspect the entire board or a (key) area.

On-board post-print inspection is 2-D and can analyze alignment and registration of paste to pads, cold slump presence (bridges) and pad coverage. Inline post-print AOI can do this as well, plus solder paste height analysis. (Hence, they are considered 3-D.) Inline post-print AOI systems are positioned after the printer and can perform without necessarily slowing print cycle time and impeding process flow. Of course, these AOIs do not have access to the stencil, which limits inspection to board pad coverage and bridge detection. In some manufacturing schemes, post-print AOI is placed offline and shared by multiple lines, with inspection performed on an acceptable quality level (AQL) basis. This usually includes first article, first article after a line pause of greater than 15 minutes (including shift change), one board every x hours, or one board after x boards.

With regard to methodology, the known good board (KGB), if used, and successive inspected board images are acquired either by an area or line scan camera. This image is processed so the areas where solder is present can be identified and evaluated. Various analysis techniques are used to evaluate the presence of solder depending on the purpose (pad coverage, bridges or stencil aperture blockage). The measurements are compared to user-defined process limits. Historical data can and should be generated for trend monitoring.

The paste detection process used is via an image processing method in which solder is separated from non-solder features of the PCBA resulting in a paste-only image that can be used to evaluate print quality. Solder paste detection methods include direct application of single or multiple thresholds to the captured image, texture-based substitution and image subtraction techniques.

The most common techniques are based on single or multiple threshold methods that use pixel brightness to isolate solder paste regions. A single brightness level (threshold) is chosen to create a binary image depicting paste and non-paste regions. Multiple threshold methods work similarly, but include a scalable transition region of brightness levels to more accurately account for edge pixels.

The subtraction method creates a different image by subtracting one image from another. Subtraction methods typically compare differences between a reference (KGB) image and a newly captured image to detect paste. The images must be registered precisely to each other to minimize error.

The third method for detecting solder paste uses texture-based recognition. This method performs a combination of mathematical and morphological operations on a digitized image to analyze and separate areas of solder paste texture from other dissimilarly textured non-paste areas on a PCB or stencil.

There are, of course, solder defects that do not become evident until after reflow and, hence, cannot be detected and prevented with post-print inspection. Statistically, these defects, along with component misplacements, are usually in the minority. Post-print inspection, on-board or inline, is refined and in my opinion, more repeatable and reliable than post-reflow options.

While SMT is a series of highly automatable processes and not too difficult to control, these processes must be monitored and controlled. Look at it as “pay me now or pay me a helluva lot more later.” Remember, we’re all in this together.

Phil Zarrow is president and SMT process consultant with ITM Consulting (; He still bears the scars, physical and mental, of reflowing convection/IR ovens. His column appears bimonthly.

Submit to FacebookSubmit to Google PlusSubmit to TwitterSubmit to LinkedInPrint Article