Hypergrade in power and linear parts with big transistors is nothing new, but for digital chips and more aggressive geometries hypergrade is a challenge. Most "industrial" parts are rated to 125°C, and many automotive parts are being qualified to 150°C. In the landmark 2004 paper "The Changing Automotive Environment", authors Wayne Johnson and John Evans of Auburn University and their industry co-authors cited studies from Delphi, Toyota, and Daimler Chrysler. In those studies, the target for "connected to engine" electronics got progressively higher, from 150 to 175 to nearly 200°C in the Daimler version.
That paper sounded a cautionary pause, however:
"The desire to place engine control units on the engine and transmission … will push the ambient temperature above 125°C. However, extreme cost pressures, increasing reliability demands (10yr/241350km) and the cost of field failures (recalls, liability, customer loyalty) will make the shift to higher temperatures occur incrementally."
That was a system-level view. At a chip-level, there are significant problems moving beyond 150°C. One question Johnson et al raise is "how long will the device see the maximum temperature?" Of course, that depends on where the device is, how it is mounted, what if any airflow or other cooling exists, and what the temperature profile looks like. Not all automotive systems are subjected to on-engine effects, but having lived near Phoenix for 21 years until recently has given me an appreciation for worst-case high-end cabin temperature.
It is clear nobody in automotive circles is willing to compromise on the 10-year reliability mark, or a parameter that goes hand-in-hand for digital electronics: 10-year data retention at maximum temperature. So, it’s not surprising that a bit more than 10 years after that study surfaced and we’re presumably comfortable at 150°C, auto makers are now asking what can be done to get to 175°C or maybe even 190°C near term.
For one item we take for granted – floating gate flash memory – the answer is the data retention figure goes away very quickly at a sustained 175C. The alternative is 1T-OTP, which can be incorporated with enough capacity to handle most control code and data storage requirements. (My factory-installed Bose entertainment system in an Infiniti G35 in Phoenix failed three times over 10 years – it’s a pain, but not a safety problem. We’re talking about vehicle control requirements here.)
Sidense is officially still at 150°C according to their new automotive collateral, but privately they are telling us they are working as we speak on an unnamed foundry qualification for their 1T-OTP at 175°C. It's not a simple procedure. Devices have to be characterized across process voltage temperature (PVT) in all macro sizes. Then, they have to be put through AEC paces: data retention storage life (DRSL), high temperature operating life (HTOL), high temperature storage (HTS), and time-dependent dielectric breakdown (TDDB).
As many of us here including me have been writing about, there are also increasing security concerns in automotive as connectivity increases. 1T-OTP has excellent physical attack resistance as it is impervious to reprogramming and difficult to reverse engineer, side-channel attack resistance with differential mode and bit-line pre-charge, and tamper resistance with lock bits and power and temperature verification among other features. Security of deeply integrated systems with or without connectivity is extremely important given the longevity and the potential exposure to unauthorized forces (such as non-factory repair outlets, not to mention hacking).
All of that adds up to hypergrade for automotive. I'm confident that there are plenty of automotive-qualified mixed signal processes out there that can fab parts, but Sidense 1T-OTP will be critical to achieving qualification under hypergrade reliability, retention, and security requirements for automotive and other markets (consider things like deep-well oil and gas drilling). When Sidense has news to report on this hypergrade front, we'll be here to cover it.