Automated Optical Inspection (AOI), Apps, and Machine Learning: Tools for PCB Quality Control

Can you assess a PCB's similarity to its golden sample just by looking at it with your eyes? As components and leads get smaller over time, this becomes increasingly difficult and nearly impossible when considering scalable PCB fabrication. How has visual inspection of PCBs evolved?

Visual inspection of PCBs has long been an important portion of the manufacturing process. Technological advancements over the last several decades, however, have made visual inspections more exact—and in some cases removed humans from the picture altogether.
Here's a high-level look at some examples of these advancements in action.

What Is a Golden Sample?

A "golden sample" is a manufactured sample deemed to be the best embodiment of the design as it is intended. It is a "perfect" or "ideal" example of a device or PCB that serves as a bar against which to measure all other fabricated samples. Using a golden sample to ensure quality for fabricated PCBs requires intensive attention to detail, from the skew of components on the board to the integrity of leads on a QFN package.

A golden sample can benchmark factors including those that can be measured by oscilloscopes and other test and measurement equipment. Visual inspection, however, is also vitally important for pinpointing errors in the fabrication process. Modern visual inspection usually includes the use of microscopes. Apps designed to work with these microscopes are intended to further aid in board assessment in increasing in popularity. Also popular is AOI or automated optical inspection, which uses a camera to do autonomous inspections of PCBs in post-production.

All of these methods involve a form of photogrammetry, the term for measuring objects based off of photographs taken of them. While this term can be applied to measurements of, say, buildings or natural formations based on aerial photography, it also applies to the measurement of PCB features on the scale of microns.

Measurement Apps with High-Powered Microscopes

Most PCB visual inspection requires the use of microscopes by virtue of the fact that surface-mounted components are increasingly crammed onto tiny spaces.

Danish digital microscope maker Tagarno has unveiled a new app that allows users to perform side-by-side comparisons of golden sample images to live PCBs or sample images that will help accelerate the inspection process.

Tagarno says their newly released side-by-side app will allow users to better conduct quality control checks following component placement or reflow operations, and help users catch misplaced components or other errors in a more efficient manner.

The maker has previously offered a number of other apps to more closely analyze measurements, verification lines, and focus stacking, but company spokesperson Sabine Svane said there was tremendous demand from global distributors and end customers for the additional capabilities provided in the new application: “Prior to using the app, the procedure was to place the boards next to each other and perform a purely visual comparison and evaluation of two boards,” according to Svane. “In regard to other situations, there are [none] where you have the precision, alignment features and ease of use in the same way as you do with our app.”

Screengrab used courtesy of Tagarno.

Users connect a mouse and keyboard directly to the microscope in order to utilize the app, which works with the FHD Trend, FHD Prestige and FHD Uno models. Users can perform a range of functions with the comparison app, including save templates, add text, arrow, and circular annotations, and adjust colors.

The auto-alignment feature in the app allows users to more easily match up the reference to the sample, by making the reference sample partially transparent. Prior to officially launching the app, the company said the product went through test use with distributors and end users.
As with any app, one of the qualities Tagarno highlights is the ability to update features and fix bugs remotely. The company issues quarterly updates on the firmware on its applications to make sure any errors or required enhancements are addressed.

Charles King, principal analyst at Pund-IT, said the application is similar to other automated optical inspection solutions used in the quality control process in the PCB market, but it's not clear what the unique point of differentiation is for the Tagarno application. AOI options can range from basic web-cam based inspection technology used for limited production use to more robust commercial grade options for high-volume processes.

Machine Vision in AOI

Assessing 3D and 2D images for automated optical inspection can be performed by machine vision algorithms. This example from Visteon and the Polytechnic Institute of Setúbal presents an algorithm for assessing 3D solder joint inspection.

One method of ensuring equal assessment of each PCB is the use of specific AOI cameras to reduce the effects of how distance from the lens affects focus. Compared to traditional cameras, AOI cameras can reduce distortion by using a telecentric lens, creating an orthographic projection of a complex 3D structure by "flattening" it to two dimensions. This allows machine vision algorithms to more reliably make accurate measurements of a board.

Another method is to capture 3D images instead, allowing algorithms to assess a three-dimensional rendering of a board. Last month, MEK Marantz Electronics announced their latest system for 3D inspection of PCBs.

Taking top-and-bottom AOI. Images courtesy of MEK Marantz Electronics

Nine cameras are used to take a detailed, 3D image of a PCB. This latest version also allows top and bottom views of the boards for additional information.

---

What's your experience with visual inspection of PCBs? Do you have expertise about assessing flattened vs 3D images? Is there an important AOI method or technology missing that you'd like to see covered? Share your thoughts in the comments below..


Featured image courtesy of Tagarno.