Improving Robustness of Ultra-Low Power Systems through Power Reduction in Wired Signaling Circuits

The Internet of Things (IoT) will offer new levels of services and rich data sets. Services like health monitoring will be improved through more personalized care. Rich data sets provided by applications like infrastructure management will allow a better understanding of wear and tear on buildings and bridges, creating a safer environment for us to live in. In order to achieve this, IoT systems must be placed everywhere from inside clothing to monitor body functions, to inside concrete bridge structures to monitor vibrations. There will be no dedicated supply of power to nodes in these places, meaning that energy must be harvested from the environment. Due to the nature of energy harvesting circuits, the energy storage component should last through many charge cycles, especially in smaller form-factor applications. Existing commercial technologies last hundreds to few thousands of cycles and often require regular maintenance and periodic full discharge cycles to avoid damage, limiting their use in self-powered, IoT applications. Electrical energy storage in capacitors and super-capacitors provide a more robust solution, having a cycle life greater than 500000 cycles. As a result, the energy efficiency and power consumption of IoT systems must be reduced to the power levels a capacitor or super-capacitor can deliver.

Power is reduced by decreasing supply voltage and optimizing circuits, architectures, and system level knobs. Circuit level optimizations have a greater effect if they were previously contributing to a large portion of the ULP system budget. Circuits within this category are mainly analog, memory, and chip Input and Output (I/O). Analog and memory circuits are heavily researched, and lower power solutions are being developed. On the other hand, ULP communication circuits have not been thoroughly researched, and optimizations can provide great reductions in both circuit and system power. Off-chip power reduction will allow for regular communication between nodes without needing to manage how often data is transmitted. ULP On-package communication will allow integration of different process technologies on the same chip, leveraging the best aspects of each process. On-chip communication can be improved to allow more fine-grained system level tuning to configure a platform for more desired operating modes.

This dissertation presents techniques at both the circuits and systems level to reduce the power consumption and increase the level of integration of a proof of concept battery-less, self-powered, System in Package (SiP). The heart of these improvements are in communication. Techniques are introduced to improve power and energy of off-chip communication, and a transceiver for robust, wired, on-body networks is demonstrated. An ULP on-package cold-boot bus is presented, enabling integration of an ULP non-volatile memory for loading instructions when the system powers up. A flexible on-chip bus is implemented to allow fine-grained power and energy SoC optimization at the logic block level. Also presented are applications and improved test methods to facilitate commercialization of self-powered, battery-less platforms. A proof of concept relative positioning system is implemented to reduce reliance on high-power GPS radios, and provide positioning when a GPS signal cannot be acquired. Lastly a test methodology is presented for subthreshold SoCs to reduce functional test time by testing at a higher optimal voltage and predicting delay and power of the system at the lower operational voltage.

Lower power communication, improved application space, and a faster test process will enable commercialization of systems. Having ubiquitous commercial battery-less platforms that are more robust will provide large sets of data to give insight to new areas of medical, industrial, and environmental research and will improve quality of life. The improved quality of life provided by these devices through an increase in personalized medical care and an abundance of sensor data will progress how our society works and interacts.