An evolution, if not a revolution, is quietly taking place in the little-watched area of the automotive semiconductor industry. The changes will be as profound as the growing trend of semiconductors replacing electromechanical components in applications stretching from headlights to tailpipe.
Automotive innovation today is almost entirely driven by semiconductor actuation, control, or monitoring. It is estimated that today’s well-equipped upscale automobile generally relies on more than 50 electronic control units. As a result, in-vehicle networking has created a quiet evolution in automotive technology, resulting in, among other things, elimination of unwieldy wiring harnesses once used for control circuits. It also enhances vehicle safety, reliability, performance, as well as reduces combustion emission.
Networking has made possible the most publicized “bells and whistles” for the 2010-11 model including onboard communications systems, the operation of control systems such as lane -departure warning systems and (slippery) road-condition advisories.
Rising fuel prices and environmental concerns are also accelerating electronic innovations in the automobile, with worldwide automotive microcontroller units (MCUs) expected to reach $5.5 billion in 2014—50% of which will be responsible for “green” optimization.
Up until a few years ago, most of the semiconductors targeted for automotive applications, both under-the-hood and in the passenger cabin, came from a handful of integrated device manufacturers (IDMs) that developed, manufactured, and marketed their own semiconductor devices. However, due to economic constraints and the high cost associated with fab construction and fab loading, many IDMs transitioned to a fab-lite or fabless model for nonautomotive products while keeping their automotive products manufactured internally. As their automotive product roadmap moved to more advanced processes, IDMs needed to find a qualified automotive foundry for their production. As a result, foundries have become the foundation for the automotive IC industry.
In addition to this trend, the popularity of Bluetooth connections, GPS, networking, and a host of infotainment applications opened up opportunities for less traditional fabless semiconductor companies to enter the automotive market. These fabless companies faced the same challenges as their IDM counterparts—the need to find a qualified automotive foundry.
Regardless of whether the company is an IDM or fabless, its manufacturing fabs must demonstrate the ability to consistently meet the automotive industry’s tough zero defect standards and wide operating temperature requirements to sell into Tier 1 automotive module suppliers and OEMs.
Developing automotive process qualification standards
The semiconductor industry is no stranger to “process” qualification standards. Almost since its inception, it has operated through a set of commonly accepted “process qualification specifications” for a variety of consumer electronic, communication, and industrial products.
As TSMC began to investigate the requirements for automotive semiconductors, it realized these specifications were not adequate for automotive products and, as a result, searched for commonly accepted process qualification specifications for the automotive industry. The obvious place to check was the Automotive Electronics Council (AEC), whose documentation is widely used by automotive semiconductor companies as well as Tier 1 module suppliers.
AEC-Q100, “Stress Test Qualification for Integrated Circuits,” defined “product” operating temperature classification: Grade 0 for -40°C to +150°C (-40°F to +302°F) ambient temperature operating range, Grade 1 for -40°C to +125°C (-40°F to +257°F) ambient temperature operating range, etc.
It also had detailed “product” qualification requirements that define stress items, sample size, stress conditions, and pass/fail criteria. However, it did not provide detailed “process” qualification requirements for wear-out reliability testing. It only defined wear-out parameters such as electromigration, time-dependent dielectric breakdown, and hot carrier injection, and it defined neither stress conditions nor pass/fail criteria.
To remedy this situation, TSMC, in 2008, presented a paper titled “Automotive Process Qualification Specifications” at the Automotive Electronics Council Annual Reliability Workshop. This process qualification specification addresses the right-hand side of the classic bathtub curve. Today, that specification has been widely promulgated not only by automotive semiconductor companies but also by Tier 1 module suppliers and OEM car makers.
Automotive service package
Developing rigorous process qualification specifications to address the right-hand side of the bathtub curve is one thing, but what about the left side and middle of the curve? Traditionally, semiconductor companies address these issues by applying stringent test methodology such as production burn-in and what is called “high/cold temperature final tests.”
All these tests and procedures, however, occur after wafer process completion. To reduce the process variations and outliers during wafer manufacturing, TSMC created the “Automotive Service Package” as an option for customers to complement their own test methodologies.
Nonvolatile memory IPs
The biggest segment of the automotive semiconductor market is comprised of microcontrollers (MCUs) that require embedded Flash, a form of nonvolatile memory (NVM) technology.
Many of the technologies used in “green” vehicles, such as hybrids, are managed through MCUs. The increase in sales for energy-efficient vehicles will fuel the MCU market.
MCUs play an instrumental role in accelerating electronic innovations in cars by making the vehicle lighter and more efficient and drivers more informed. Increasing complexity in automotive electronics is amplifying the need for higher performance 32-bit MCUs with more embedded nonvolatile memory. Other MCU-enabled applications, such as electronic engine control and the aforementioned networking, reduce overall vehicle weight and improve engine performance and gas mileage by eliminating mechanical systems and dedicated wiring.
AEC-Q100 has very stringent NVM qualification requirements that are very time-consuming and costly for product companies to qualify a product, especially with embedded Flash since it requires 10,000 read/write endurance cycles and 10-year data retention. It is much safer for an MCU to adopt an embedded Flash IP that has been qualified per AEC-Q100 already.
TSMC met this challenge by offering two generations of embedded Flash IP that pass the AEC-Q100 NVM qualification requirement and are available to the general market.
Looking toward the future
TSMC has won the trust of several major IDM companies to jointly develop customized 65- and 55-nm embedded Flash projects for underhood MCU applications. Tier 1 module suppliers and car manufactures recognize that the company is becoming a major automotive foundry as it provides the foundation for IC innovation while working with the automotive IC industry to meet stringent automotive standards and requirements.
Cheng-Ming Lin, Director of Embedded Flash Business Development, and Kuotung Cheng, Automotive Program Consultant, of TSMC, wrote this article for Automotive Engingeering.