A Differential Transmission Gate Design Flow for Minimum Energy Sub-10-pJ/Cycle ARM Cortex-M0 MCUs

Abstract

Ultra-low voltage operation is key to achieving energy-efficient operation for microcontroller (MCU) systems. Variation resiliency, high speed operation, and short design time are the most important challenges for these systems. This paper overcomes these challenges in a new design strategy that enables standard cell design with differential transmission gate logic. The commercial toolchain is extended with in-house developed add-ons and makes use of two custom libraries with different device lengths to allow high speed vs. low leakage trade-offs. The design flow is used to prototype two highly efficient 32-bit ARM Cortex-M0 MCU systems in 40-nm CMOS. The core of the first prototype scales down to 190 mV and 0.8-MHz and reaches 16.07 pJ/cycle at 31.2-MHz and 440 mV. The second prototype benefits from the dual libraries and reduces core energy consumption by 50% at the same speed performance. Minimum energy operation is thus achieved at an even lower voltage (370 mV) with the M0 core consuming only 8.80 pJ/cycle at 13.7-MHz, breaking the sub-10-pJ/cycle barrier for a 6–35-MHz range.