Materials Challenges in Automotive Embedded Non-Volatile Memories
- PDF / 1,575,117 Bytes
- 15 Pages / 612 x 792 pts (letter) Page_size
- 29 Downloads / 177 Views
0997-I02-01
Materials Challenges in Automotive Embedded Non-Volatile Memories Erwin Josef Prinz Technology Solutions Organization, Freescale Semiconductor Inc, 6501 William Cannon Drive West, MD: OE-49, Austin, TX, 78735 ABSTRACT Silicon-based nonvolatile memory modules are widely used in microcontrollers, where they are embedded into a monolithic system on a chip (SoC) which also includes high speed logic transistors, cache SRAM, and peripheral circuits for communicating with the external world. The physical principle most widely exploited for nonvolatile code and data storage is charge storage in floating gates. Charge storage in nitride traps, and more recently, in nanocrystals also has been explored. The most demanding use profiles with respect to temperatures, data retention times, and low failure rates are encountered in automotive engine control applications, where junction temperatures up to 150∞C are common, for 1000ís of hours. Starting with the 130nm technology node, embedded Flash technology has been integrated with copper interconnects, and at the 90nm node, low dielectric constant interlevel dielectrics are also employed to increase circuit performance. To achieve automotive reliability, the materials surrounding the silicon floating gate, nanocrystal, or nitride charge storage area must be evaluated for parasitic charge storage, write/erase stress-induced leakage current, and other parameters important for reliability. Any movement of parasitic charge, potentially over a long period of time, can reduce the sensing window of the Flash EEPROM bitcell. INTRODUCTION Silicon-based nonvolatile memory modules are widely used for code and data storage in embedded systems [1]. They can be integrated with other modules such as a central processing unit (CPU), cache SRAM, and peripherals on monolithic systems-on-a-chip (SoC), which are commonly referred to as a microcontrollers [2]. Compared to volatile memories which loose their information once the chip is powered off, nonvolatile memories retain their data. In most commonly used nonvolatile memories, this is achieved at the cost of much slower write and erase times compared to SRAM memories, and of limited endurance (maximum number of write/erase cycles) of up to 1 million write/erase cycles. For code storage, only a few write/erase cycles are needed in many applications, to enable periodic code updates in the field. Erase can be done on large blocks of bytes which is advantageous for obtaining a small module size. Nonvolatile memory with byte-wide write and block-wide erase is commonly referred to as ìFlash EEPROMî. In a microcontroller, the CPU can read data out of the nonvolatile memory sufficiently fast as to not require cache SRAM.
For data storage, 10,000 to 100,000 write/erase cycles are required. Byte-wide erase is desired, but requires a large bitcell, leading to a large module area. While ìbyte-eraseable EEPROMî modules have sometimes been designed, data can also be stored in Flash EEPROMís ñ in that case, small erase block size is desired. In le
Data Loading...