中文 English

Failure Prediction Is Vital for Packaging Technology

Continuous improvement and process optimization are essential to making electronic devices work as expected.

popularity

A while back, I wrote a post on what I call the “10 Commandments of Packaging” – my list of the key things to bear in mind with respect to developing new packaging solutions. Today, packaging engineers must contend with more variables than ever, from new substrate materials to more types of packages with a greater range of complexity. With that complexity comes new challenges, and I felt that some of these commandments bear a bit deeper dive, starting with, “Predict failures.”

Failure prediction requires knowing what to look for and putting a methodology in place to help you anticipate failures so that you can prevent them from occurring. If you wake up one morning and see some water on your kitchen floor, it’s clear that you have a leaky pipe. You have to fix the problem as soon as you see the earlier signs, or it will become catastrophic.

Having a mechanism in place to truly predict failures in the packaging process can make a huge difference in enabling us to stop them before they happen. A key example is optimizing wire bond strength. Wire bonding is an automated process, but the bonds must be tested to ensure pull strength. Bond material can accumulate on the tool, or the tool can experience wear and tear accumulating residue from the bonding process. Visually inspecting wire bonds is not a practical way to ascertain any problems that may result in bond failure.

One way to mitigate this problem is to perform pull testing and employ statistical process control (SPC) techniques to create predictive models. By recording wire bond pull strength on every product (i.e., the amount of force that can be applied before a bond fails), we can create a database of thousands of pulls that the SPC software can analyze in order to pinpoint anomalies. Using SPC is fairly standard. Some companies just plot out their process and post the printout on the factory wall so they can see how they’re doing.

The far better use is to review and analyze the data in order to figure out what it’s telling you over time. For example, a plot may show that the average pull strength was 14g in June, 13.7g in July, 13g in August, and 12.9g in September. This downward trend in pull strength tells us that something is not quite right. SPC plots the points and creates an analysis of this trend. You can then look back at the wire bonding process, and you may find, for example, that you need to change the capillary more often.

Fixing problems in the short term is called containment action – basically, a kind of Band-Aid. Unless you get rid of the root cause, the problem will persist. Techniques like SPC – also known as statistical quality control (SQC) – are vital to achieving this. W. Edwards Deming knew this. Thought of by many as the “father of quality,” Deming was a statistician and business consultant whose focus on statistical methods, continuous improvement and company-wide quality form the basis of what we know today as total quality management, or TQM. He went to Japan after World War II as an advisor to their census process, and his methods helped hasten Japan’s recovery after the war and beyond. He helped Japan and, ultimately, the U.S. and the rest of the world, to understand how to build processes that can be replicated uniformly and consistently, so that you don’t waste time building rejects.

Today, artificial intelligence (AI) and machine learning are being integrated with SPC to speed learning and enable consistency in increasingly complex processes. The notion of lean manufacturing is an extension of SPC; being able to understand and assess variations and why they occur helps us perform more consistently and deliver what our customers need in a more timely manner.

Reliability test techniques
Before mass production, electrical packaging has to pass reliability testing. Thermal cycling test (TCT) is one of the standard reliability tests that has been commonly used in the electrical packaging industry. Ensuring new products pass TCT is a critical issue in the electronic packaging industry. Finite element method (FEM) based design-on-simulation technology can be used as a feasible development methodology for reliability assessment and reliability prediction of electronic packages. 

Another method to detect failure and ensure reliability is to simply operate the device for an extended period to detect problems. High temperature reverse biased (HTRB) burn-in testing is a simple, low-cost, accelerated-lifetime test that enables defective parts to be sorted out of the population. The root cause of failed units may be dielectric failure, conductor failure, metallization failure, etc. These faults are dormant and randomly manifest into device failures during the device life cycle. We use HTRB burn-in testing to stress the device, accelerating these dormant faults to manifest as failures so that they can be screened out during the infant mortality stage, thereby preventing failures from occurring.

When something goes wrong in our business, customers ask for corrective action. We are able to generate a report that examines the “5Ms and E” — man, material, method, machine, measurement and environment. Anything that goes wrong is caused by one of these factors. Knowing which one created a problem enables us to determine how we should solve the problem. By evaluating these six factors at the beginning of a project, before a failure occurs, we can go in with a historical perspective that allows us to know where we’ll need to take action to prevent a problem from happening again. This focus on continuous improvement and process optimization is vital to the future of advanced packaging.



Leave a Reply


(Note: This name will be displayed publicly)