02 December 2010
What does it take for a foundry to succeed in the automotive market?
Automotive presents a challenge, but also an opportunity, to the semiconductor foundry industry. The challenge is that automotive has historically been the turf of the IDMs, where the systematic feedback process from designer to technologist to reliability physicist to yield analyst and back to designer creates the wealth of 'how-to' data that accelerates the drive to 0 dppm, which is a key requirement for this market. Since the foundry is physically separate from the IDM in most cases, this communication feedback loop is more difficult – not impossible, but more difficult.
The opportunities that automotive presents for the foundry industry are several. First, it offers a growing business since car electronics is a fast-growing market. In addition, it offers a business that is relatively stable over a long period of time. Thirdly, the highly-disciplined procedures required in automotive can raise the performance level of everything else the foundry does.
So, what are some of the key focus areas for a foundry to be successful in automotive? Frankly, most of the things needed for automotive are needed in other markets as well; they just need to be re-evaluated with a critical eye on reliability and dppm. These fall into 4 high-level categories:
1. Technology accuracy: Automotive technologies can be varied – logic, analog, power. In the analog and power space, designers need a variety of component types. These must be well-characterized for 2nd order effects such as matching, drift, voltage or temperature coefficients, as appropriate. Component Cpk's must be 1.67 but must also be tight enough to permit a cost-effective product design. Automotive can be a harsh environment, so guard-ringing strategies must be developed and guidelines available.
2. Simulation accuracy: The automotive PDK must be robust – i.e., all the detailed component characterization data must be included in the PDK. SPICE models must be taken out to 175C since the product must operate for some time there. While ESD is always difficult, automotive takes that to the next level where 8kV is a common requirement, so special attention must be paid to unique pin applications.
3. Manufacturing accuracy: Key words here are long-term, change control, and defects. Many automotive products run for 10-15 years or more, often in small volumes, which requires a high-quality mindset. While the goal is no process changes, when needed they must be done with rigorous change-control policies in the fab and collaboration with the customer. A relentless focus on defects and yield, using detailed electrical FA is required.
4. Reliability accuracy: Automotive products often have large-sized components on-board, so one must understand the DD levels on components, big components, not just on unit processes. This means the foundry must look for these problems beforehand, not wait for an issue to occur – FMEA! Another special reliability challenge is on-board non-volatile memory (NVM), which must be characterized for DD (large structures) and reliability.
So, these are the technical things that a foundry can do to become automotive-capable. But equally as important as the technical stuff is the non-technical stuff: efficient collaboration. Physical separation is no excuse for not communicating. This is what separates a regular foundry from a truly great foundry.
Christophe Bianchi, general manager, 3MM & European representative, Dongbu HiTek