Utilizing digital design techniques and circuits to improve energy and design efficiency of analog and mixed-signal circuits
MetadataShow full item record
Technology scaling has long driven large growth in the electronics market. With each successive technology generation, digital circuits become more power and area efficient. The large performance increases realized for digital circuits due to digital scaling have not translated to similar performance improvements for analog circuits. First, noise-limited analog circuits are not capable of leveraging the reduced parasitics of advanced processes, since capacitor sizes are generally set by noise requirements. Second, analog circuit performance is closely tied to the achievable device intrinsic gain, which degrades as process sizes shrink. Reduced supply voltages further exacerbate this issue, as the achievable gain per stage is limited by the number of devices that can be stacked while maintaining all devices in saturation. Finally, process variation increases with decreased feature sizes, so analog circuits have deal with increased mismatch and wider variations in threshold voltages, increasing the time required to design a circuit that is robust across process, voltage, and temperature (PVT) variation. This work seeks to address the limitations of analog circuits in advanced technologies by leveraging digital techniques and digital-like circuits that offer improved scalability. The first half of this dissertation investigates replacing the traditional closed-loop residue amplifier in a pipeline analog-to-digital converter (ADC) with an open loop dynamic amplifier. Previous works incorporating dynamic amplifiers have struggled to achieve large gains and have suffered from offset mismatch between the comparator and amplifier, which will only get worse in more advanced technologies. We propose the usage of a residue amplifier that combines an integration stage, to ensure low noise operation, with a positive feedback stage, to ensure high gain and high speed operation. By utilizing this topology, the proposed amplifier was the first dynamic amplifier to achieve a high gain of 32. Additionally, the proposed amplifier can reuse existing comparator hardware in the ADC, removing all offset mismatch between comparator and amplifier. Digital calibration techniques were applied to ensure a constant gain across PVT. The next part of this dissertation tries to overcome the scaling challenges for noise-limited ADCs with band-limited input signals. By leveraging digital filtering techniques to generate a prediction of the band-limited signal, the conversion can be limited to a range that is a fraction of the total ADC input range, allowing for significant decreases in reference and comparator power consumption. This work extends previous works by enabling accurate predictions for any band-limited signal characteristic. Previous works only focused on accurate predictions for low-activity signals. Finally, the large compute power enabled by modern technology scaling is leveraged to improve the design efficiency of analog circuits. A new automated circuit sizing tool is proposed that can achieve better performance than manual designs done by experts in a much shorter amount of time. All of these techniques help to alleviate the power and design efficiency limitations caused by technology scaling.