Silicon Space Technology has developed a commercial bulk CMOS process technology, HardSIL™, which allows optimization of performance, power, and lifetime at high temperatures. A method for preventing latchup, originally developed for use in the space radiation environment, is presently applied to terrestrial high-temperature environments. With the possibility of latchup eliminated in scaled CMOS technology nodes, further designs specific for high-temperature environments have proceeded well. This novel technology has been applied to our 18Mb synchronous burst SBRAM and our ARM® Cortex® M0 microcontroller, and in two CMOS processes at the 130nm technology node (Texas Instruments and GLOBALFOUNDRIES). Extensive temperature testing on these parts demonstrates that bulk silicon CMOS technology has a practical temperature limit of 250°C or higher.
Both the microcontroller and the SBRAM have been tested with clock rates up to 70MHz and at temperatures up to 260°C. Both parts have performed without error and without latchup under these conditions, and with low operating current and low leakage current. For example, the 130 million-transistor 18Mb SBRAM has average core leakage current of 580mA at 250°C and core voltage of 1.5V with test lots and simulations showing further reduction in leakage in the next, terrestrial version of this part. In addition, the 18Mb SBRAM is undergoing an endurance test at 250°C, presently at the 2500 hour milestone. Operation at temperatures beyond the present limit of the testing equipment (260°C) appears possible from extrapolation of current data.
Integration levels of greater than 8 million gates on a bulk CMOS device would allow multi-core processors with large on-chip secondary caches. Additional DSP engines or other compute engines can be accommodated for processing high resolution three dimensional images in real time. This would provide substantial distributed processing in drilling or jet engine control. These system-on-chip (SOC) integration levels can substantially reduce mechanical failures in a subsystem by reducing the number of wire bonds from greater than 1000 connections to less than 100 connections. Integration of mixed-signal A/Ds and D/As as well as on-chip power management provides a path to further reduction in mechanical connections in a sub-system.