Low-power ADC designs in scaled CMOS process

This thesis presents advanced design techniques for successive approximation register (SAR) analog-to-digital converters (ADCs), continuous-time ∆Σ ADCs, and single-slope (SS) ADCs in nano-scale CMOS technologies. (1) In high-speed SAR ADCs, metastability of the comparator limits the performance, which even results in the sparkle code errors. Proposed background calibration utilizing the comparator decision time detector removes the metastability-induced sparkle code errors by controlling the metastability detection window. At the same time, 1-bit resolution increase is gained from the proposed technique, which results in the fewer comparison cycles. Along with the relaxed requirement on the comparator, this cycle reduction helps to achieve the good power efficiency in high-speed SAR design. A prototype ADC in 40nm CMOS achieves 35.3dB SNDR and consumes 0.81mW while sampling at 700MS/s. (2) In the proposed continuous-time ∆Σ ADCs, conventional power-hungry opamp is replaced by voltage controlled oscillators (VCOs) that perform the data conversion in the phase domain instead of the voltage domain. In contrary to the opamp which is difficult to achieve good performance in the advanced CMOS process, VCOs have many advantages in the phase domain. To solve the nonlinear gain of VCOs, dual VCO-based integrator is used to suppress the dominant second-order distortion. To address the distortion from the DAC, a novel DAC calibration technique that both digitally senses and removes DAC mismatch errors is proposed. It has low hardware complexity by taking advantage of the intrinsic clocked level averaging (CLA) capability of dual-VCO-based integrator. It ensures high linearity regardless of the VCO center frequency. By lowering the VCO center frequency, power consumption is reduced. A prototype ADC designed in 130nm occupies an area of only 0.04mm² . It achieves 71dB SNDR over 1.7MHz bandwidth (BW) while sampling at 250MS/s and consuming only 0.9mW from a 1.2V power supply. The corresponding figure-of-merit (FOM) is 98 fJ/conversion-step. (3) A SS ADC has advantages of high linearity and a simple architecture. Thus, it is well suited for the column-parallel architecture for the CMOS image sensors. However, conversion speed is severely limited in high-bit resolution since more than 2 [superscript N] cycles are required for a N-bit resolution. To tackle this limitation, a two-step approach becomes popular. In this thesis, a two-step SAR/SS architecture is presented. In addition to reducing the conversion time, analog correlated double sampling (CDS) can cancel kT/C noise, which enables capacitor area reduction. A prototype ADC in 180nm CMOS occupies only 9.3µm x 830µm. It achieves 60.5dB SNR after CDS while sampling at 256kHz and consuming 91µWElectrical and Computer Engineerin