Semiconductors shift to ensure greater auto reliability
Robert Cappel of KLA Corp. discusses how increasing needs in ADAS and AV technology and Zero Defect reliability are driving semiconductor development for automotive applications.
The future potential of autonomous cars is driving innovation in the semiconductor industry. While fully autonomous vehicles (AVs) are years off, features that improve automotive safety and efficiency and provide in-vehicle entertainment are on the rise. These features are driven by innovation, and a shift in the semiconductor industry is key to overall success.
As consumers and regulators demand more capability from automobiles, semiconductors have become a critical part of the advanced solutions. However, the same chips designed to bring safety and economy to the car’s operation also bring complexity and increased requirements for reliability. Many of the semiconductors are now part of advanced driver assistance systems (ADASs) that are critical to the function and safety of the vehicle, where failures cannot be tolerated.
The rapid push into autonomous driving capability accentuates the need for all chips to work together without incident to protect the safety of both the car’s occupants and others in the surrounding environment. This is the impetus for car manufacturers and Tier 1 suppliers insisting that Zero Defect programs be put in place across all automotive semiconductor suppliers. These programs are having a profound impact on the importance of process control for automotive semiconductors. Across the ecosystem, companies are recognizing the need to make fundamental changes in their historical approach to automotive chip manufacturing to respond to critical new trends and be successful in the automotive market.
One trend is the increase in semiconductors in vehicles. Semiconductor content is growing rapidly, both in volume and as a percentage of the raw parts cost of the vehicle, to serve safety, connectivity, and automation functions in the smart, connected, or autonomous car. This upward trend shows no sign of slowing soon. Up from a few hundred, large-design-rule controllers, MEMS, power regulators, and other components a decade ago, a modern vehicle may now contain up to 8000 semiconductor chips that represent 20-30% or more of its cost.
As cars are tasked with automating more of the functions of driving, all semiconductor device segments are seeing growth, and cars are now likely to contain multiple complex SoCs (systems on chips), dozens of image sensors, and more memory (e.g., flash and DRAM) than a laptop or mobile phone. However, electronics are currently the number one failure item in cars under three years old. Vehicle manufacturers are under pressure to reduce these failures to shrink their warranty costs, and more importantly to preserve their reputation for quality, which is one of the top factors in consumer brand loyalty.
Another trend is the move toward ensuring that the semiconductors in the car are reliable and function over time in the most severe conditions—especially systems that are critical to the safety of a car’s occupants. The semiconductor industry has been discovering that automotive quality is quite different from consumer-grade quality. Reliability expectations for consumer-grade devices are orders of magnitude lower than the automotive market, allowing up to 10% failure rates within the first two years in a relatively tame operating environment.
This is the standard that many non-automotive fabs are accustomed to serving. In contrast, automotive quality requirements demand parts per billion (ppb) failure rates, giving rise to the Zero Defect concept of manufacturing. Zero Defect requires a multi-disciplinary approach to quality, but process control is a predominant factor in its success. The types of defects that cause reliability issues are the same type that cause yield issues. Therefore, total defectivity can serve as an effective proxy for reliability. Reaching ppb automotive quality means pushing further up the yield curve than the industry has become accustomed to for shorter lifetime consumer electronics (See Figure 1).
Latent reliability defects (Figure 2) present obstacles to Zero Defect success. The size or location of these defects may not initially kill the die, or they may lie in an untested area of the die—an increasing problem with complex SoCs. As a result, the at-risk die passes electrical test and “escapes” into the supply chain. The demanding automotive environment of extreme cold, high heat, humidity, and vibration can sometimes “activate” these latent defects and cause a premature failure.
The industry has long relied on electrical testing as the method to cull bad die, but latent defects pass electrical testing, so other methods are required to stop escapes near the source where costs are lower. Industry estimates have pegged the cost of an escape as increasing 10x for every layer of integration it passes (fab, test, board, system, car 0km, warranty, recall), creating a strong push to find and eliminate the at-risk die in the fab.
The best approach for improving semiconductor reliability is to manufacture devices with fewer overall defects by implementing well-proven process-control solutions. Closely monitoring and controlling the processes used to make the chips and employing continuous improvement programs that reduce the random defectivity introduced by the process tools or environment are critical for reducing overall defectivity.
The second approach is to ensure that the process is sampled frequently enough to provide traceability. When the inevitable process excursion happens, Zero Defect fabs know definitively where the problem started and stopped so they can quickly quarantine the affected parts until they can be effectively dispositioned or culled. The combination of these two methods for reducing reliability defects has resulted in the automotive fabs adopting process control tools at a higher rate (75th percentile), as well as typically using defect inspection systems designed with sensitivity for one or two design nodes more advanced than they are manufacturing so they can detect the smaller defects that may affect reliability.
A method that is receiving increasing interest is utilizing inline defect information not only to control the process, but to identify die at risk for reliability problems while they are still in the fab—where the cost of correcting the problem is the lowest. Automotive fabs have long relied on screening (See Figure 3), where a high throughput defect inspection system inspects 100% of the die on all wafers at a handful of final layers late in the process. Die that meet the defined failure criteria (defect size/type/location) are excluded or “inked.”
While effective for large defects, this method is inadequate alone for smaller, latent defects. A new inline technique, called I-PAT (Inline Parts Average Testing), may be the answer. It leverages a 20-year-old automotive industry technique known as Parametric Parts Average Testing (P-PAT.) This original e-test based method identifies any die whose test results lie outside of the normal distribution of the population, even if they are within the operating specifications.
For a small sacrifice of 0.5-2.5% of yield, significant improvements in reliability are gained, with some automotive fabs seeing a 20-30% improvement when these outlier die are culled. I-PAT moves this concept inline, using massive data sets and artificial intelligence to identify die with outlier defect populations across the multiple steps within the manufacturing process. These outlier die are statistically more likely to contain the latent defects that the industry desperately wants to eliminate. I-PAT results could be used to cull these at risk die and improve the overall go/no-go decision for die.
The combined requirements for additional sensitivity to small reliability defects, faster yield learning cycles at advanced design nodes, and increased sampling for traceability all drive the need for increased process control in automotive semiconductor fabs. Additionally, the introduction of I-PAT may help automotive semiconductor suppliers better utilize their process control equipment to identify the latent defects that are a top priority. These approaches can help the automotive industry achieve the reliability goals for all the semiconductors implemented in modern ADAS and future autonomous vehicles.