How do you use failure analysis data to improve PCB design and quality?

Leveraging failure analysis data is critical for enhancing PCB design and quality. Analyzing root causes of failures allows for targeted improvements in design specifications, materials, and manufacturing processes. Identifying recurring issues enables the implementation of preventive measures, optimizing overall product reliability. This data-driven approach enhances the design-for-reliability strategy, reducing the likelihood of similar failures in future iterations. Continuous monitoring and incorporation of failure analysis insights into the design process contribute to a proactive quality improvement cycle, ensuring that PCB designs align with industry standards and exhibit superior performance and durability.

One of the most useful approaches to PCB failure analysis is investigating the PCB prototype using multiple visualization techniques. Once the PCB is on the desk and something wrong happens, i.e. the circuit does not behave like it should, we must have the tools to feel what is wrong. A thermal one is one of the most common failure modes for a PCB. One or more components, or the PCB tracks, will heat past their sustainable temperatures and start material degradation and destruction. Hence, pointing a Thermal Camera on the PCB while turning it on for the first time can help debug it, but can also save the board or expensive components by identifying failures before they are unrecoverable while saving time once the damage is done.

Choosing the right components can indeed help avoid failures, but knowledge and analysis of them can also aid in failure analysis. This is especially critical when dealing with Moisture Sensitive Components. Every component has its moisture sensitivity level (MSL), beyond which it can start degrading. This degradation may not produce a thermal footprint, which is therefore a failure difficult to identify even with tools such as thermal cameras. Instead, it would be better to design boards by importing the components datasheets in a searchable and indexed way. By doing so, in front of a problem yielding no thermal footprint, a designer can easily search the schematic for components with low MSL and possibly change them or clean the board.

The best advice to improve the manufacturing process as designers is to improve the design itself. This means identifying companies that are in range to manufacture the PCB we would like to create. After this choice, asking about their technical capabilities, both maximum and in normal production, is the optimal way to ensure the design is done to avoid failure during manufacturing, and post-delivery, and have a better production yield. In general during the design, avoid as much as possible sharp turns of tracks, sharp edges, and anything too thin when there's no reason for such detail to be small.

Reliability testing is as important as the design of the board. The process of testing must be carefully designed to match the expected lifetime of a project. Once this calculation is done, accelerated aging can be performed in thermal chambers, but the data resulting from such experiments must be carefully analyzed. A strong eye should be placed on the bathtub infant mortality graph, showing that statistically, boards and PCB either die very early in the testing or very late when everything starts to fail. An important consideration that can help a company design failure testing campaigns capable of identifying 80% of premature failures in a few hours, but that would require weeks to reach 99% of coverage.

A very important decision to make, especially in failure testing for products that will need to reach the market, is how much to stress a product before releasing it for sale. This is critical as if the product is stressed too much during the reliability testing phase, it will have a lower lifetime once sold or in use. This is why it is important to carefully structure the testing campaign and carefully select components capable of withstanding these tests before going in the market for years of service, or even decades, as we were doing in the LHC at CERN.