For years, semiconductor manufacturers have leveraged manufacturing data throughout their globally-dispersed supply chains to improve quality and reduce return material authorizations (RMAs). Automotive OEMs and Tier 1 suppliers are now working to meet the similar challenge of reducing defective parts per million (DPPM) and beyond in vehicle production. The ability to share and connect data backwards and forwards throughout the supply chain is now seen as a key capability to address this challenge. How can sharing data throughout the automotive supply chain reduce DPPM?
Getting all the data
Whether manufacturing is housed in a single factory or across multiple factories within a value chain, the ability to quickly access and analyze manufacturing data is a key element in determining the quality of automotive electronics parts or products. The sharing of aggregated manufacturing data across various test stages or throughout subcons located in different geographies can drive significant insights and learnings, including faster time-to-quality for accelerating new product introductions (NPI), more robust detection of outliers that help to reduce RMAs and the ability to create “test more” and “test less” populations that help to reduce overall test costs.
Until recently, semiconductor and electronics companies primarily worked in siloed environments due to legacy data infrastructures that were not conducive to sharing manufacturing data and contractual restrictions that limited data sharing across the value chain. Recent high-profile consumer and automotive electronics recalls are a perfect example of why it is important for manufacturing data to be shared as often and as early as possible among value chain participants. It only takes an extremely small fraction of product failures to cause catastrophic recalls and brand damage.
While these number of devices involved in a recall can make it appear that these failures are stochastic in nature, the application of advanced product analytics to the manufacturing data can identify that the failures are actually the combination of systematic variations in the properties of the devices and the usage patterns of the device. As a result, product quality cannot simply be correlated to a specific supplier or batch of material. In fact, what product analytics can demonstrate is that the quality of every device manufactured is truly unique because “quality” is comprised of many different elements, even when the primary cause is determined to be a “design defect”. This is the reason why in a broad electronics product recall, many of the devices recalled do not exhibit any of the issues that forced the recall in the first place.
And once the product is in the field, imagine if chip suppliers had access to data from electronics systems suppliers on how their chips perform when they are in-use. This can provide tremendous insight for current and future chip development. On the flip side, if electronics board suppliers can receive manufacturing data from all their chip suppliers, electronics board suppliers can more easily diagnose problems they find by linking data from multiple suppliers, avoiding critical product recall situations.
In theory a “supplier quality network” can connect chip manufacturers with electronics manufacturers through an electronic data hub that protects proprietary information of all the participants in a value chain but shares enough bi-directional data to enable engineers to solve problems, even ones that appear to be unsolvable.
The benefits of a supplier quality network are tangible and go beyond information sharing. With complete product genealogy, there is increased knowledge and productivity for the entire value chain. Electronics parts manufacturers can expect lower RMA costs because they can determine if parts are really good before shipping to suppliers and can more quickly pinpoint the root cause of problems, even those that involve multiple factors or appear as “no trouble found” which can account for a third of failures at the device level.
In addition, electronics manufacturers can achieve faster time to quality by reducing unnecessary, redundant and expensive tests. If manufacturers have good information on all the products from their own factories as well as their suppliers, they can improve their test methodologies and be more selective with tests, such as reducing the number of expensive tests like burn-in. With data flowing back and forth between electronics suppliers and chip manufacturers, manufacturers can even determine which components will provide better performance and capability for the overall system.
To put the importance for improved electronics quality through data sharing into perspective, the first autonomous cars are set to come to market within the next decade. At SEMICON Europa 2015, Audi shared that a modern premium vehicle already has about 7000 semiconductor devices. Assuming a defect rate of one part per million and an annual production rate of two million vehicles, that translates into a failure rate of more than one per hour, every day of the year.
With electronic and semiconductor content in cars only increasing, it will become imperative to improve quality another order of magnitude over the next decade. The quest to achieve “defective parts per BILLION” will need to be built upon data sharing across supply chains of both chip and electronics manufacturers.
If supplier testing and performance data can be shared throughout a value chain, it can dramatically reduce the time needed in NPI to reach the necessary quality and yield levels to move into high-volume production. At the same time, advanced analytics such as multi-variate analysis can also be applied to all the components of a system to identify composite issues that can negatively impact near- and long-term product quality.