FROM EDN EUROPE: Embedded flash for FlexRay controllers
By Graham Prophet, EDN Europe -- EDN, April 13, 2006
Now entering production in Philips Semiconductors ' Nijmegen, Netherlands, wafer fab is a flash/EEPROM memory option for the company's 0.18-micron CMOS process, which is fully qualified for Grade-1 automotive applications. The Grade-1 qualification means that the parts will operate over the extended automotive-temperature range, and that parts meet a more stringent quality assessment. Strategic program manager for the project at Philips is Frans List; he explains that the company has used a two-transistor (2T) memory cell as opposed to the more common single-transistor (1T) cell used in high-volume flash memory. In Philips' 2T cell, Fowler-Nordheim tunneling achieves both programming and erasure (storage and removal of charge)—the more usual method of programming is to use hot-electron injection. List says that use of the 2T cell carries a number of benefits, "Programming and erasure both take place in one cycle, [unlike the iterative process taht programs a standard flash cell which takes a variable number of cycles, until the voltage of the cell reaches], so operation is deterministic. This suits it to use in protocols such as FlexRay, and in the ARM-based controllers Philips is developing for that standard." The 2T structure employs threshold voltages above and below zero (positive and negative), reducing voltage stress in the cell—in a regular flash cell negative values of VT are forbidden and overall positive voltage levels must be correspondingly higher. List adds: "the 2T cell does require more silicon area, but against that there is a trade-off: there is a reduced area needed for charge pumps as the programming voltages required are lower." The design needs no separate "select" line.
Philips will employ the flash/EEPROM option as a low-power embedded option on its standard CMOS process—it needs no modifications to the process, and the existing cell library is unaffected. List says that Philips is confident that the flash cell will scale down to 0.14-micron CMOS—one version is already in production at the same fab, for manufacture of chips for consumer products—and onwards to 90-nm processes.
