Power Management and SRAM for Energy-Autonomous and Low-Power Systems

We demonstrate the two first-known, complete, self-powered millimeter-scale computer systems. These microsystems achieve zero-net-energy operation using solar energy harvesting and ultra-low-power circuits. A medical implant for monitoring intraocular pressure (IOP) is presented as part of a treatment for glaucoma. The 1.5mm3 IOP monitor is easily implantable because of its small size and measures IOP with 0.5mmHg accuracy. It wirelessly transmits data to an external wand while consuming 4.7nJ/bit. This provides rapid feedback about treatment efficacies to decrease physician response time and potentially prevent unnecessary vision loss. A nearly-perpetual temperature sensor is presented that processes data using a 2.1μW near-threshold ARM°R Cortex- M3TM μP that provides a widely-used and trusted programming platform. Energy harvesting and power management techniques for these two microsystems enable energy-autonomous operation. The IOP monitor harvests 80nW of solar power while consuming only 5.3nW, extending lifetime indefinitely. This allows the device to provide medical information for extended periods of time, giving doctors time to converge upon the best glaucoma treatment. The temperature sensor uses on-demand power delivery to improve low-load dc-dc voltage conversion efficiency by 4.75x. It also performs linear regulation to deliver power with low noise, improved load regulation, and tight line regulation. Low-power high-throughput SRAM techniques help millimeter-scale microsystems meet stringent power budgets. VDD scaling in memory decreases energy per access, but also decreases stability margins. These margins can be improved using sizing, VTH selection, and assist circuits, as well as new bitcell designs. Adaptive Crosshairs modulation of SRAM power supplies fixes 70% of parametric failures. Half-differential SRAM design improves stability, reducing VMIN by 72mV. The circuit techniques for energy autonomy presented in this dissertation enable millimeter-scale microsystems for medical implants, such as blood pressure and glucose sensors, as well as non-medical applications, such as supply chain and infrastructure monitoring. These pervasive sensors represent the continuation of Bell’s Law, which accurately traces the evolution of computers as they become smaller, more numerous, and more powerful. The development of millimeter-scale massively-deployed ubiquitous computers ensures the continued expansion and profitability of the semiconductor industry. NanoWatt circuit techniques will allow us to meet this next frontier in IC design.Ph.D.Electrical EngineeringUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/86387/1/grgkchen_1.pd