2014 Theses Doctoral
Power-efficient Circuit Architectures for Receivers Leveraging Nanoscale CMOS
Cellular and mobile communication markets, together with CMOS technology scaling, have made complex systems-on-chip integrated circuits (ICs) ubiquitous. Moving towards the internet of things that aims to extend this further requires ultra-low power and efficient radio communication that continues to take advantage of nanoscale CMOS processes. At the heart of this lie orthogonal challenges in both system and circuit architectures of current day technology.
By enabling transceivers at center frequencies ranging in several tens of GHz, modern CMOS processes support bandwidths of up to several GHz. However, conventional narrowband architectures cannot directly translate or trade-off these speeds to lower power consumption. Pulse-radio UWB (PR-UWB), a fundamentally different system of communication enables this trade-off by bit-level duty-cycling i.e., power-gating and has emerged as an alternative to conventional narrowband systems to achieve better energy efficiency. However, system-level challenges in the implementation of transceiver synchronization and duty-cycling have remained an open challenge to realize the ultra-low power numbers that PR-UWB promises. Orthogonally, as CMOS scaling continues,
approaching 28nm and 14nm in production digital processes, the key transistor characteristics have rapidly changed. Changes in supply voltage, intrinsic gain and switching speeds have rendered conventional analog circuit design techniques obsolete, since they do not scale well with the digital backend engines that dictate scaling. Consequently, circuit architectures that employ time-domain processing and leverage the faster switching speeds have become attractive. However, they are fundamentally limited by their inability to support linear domain-to-domain conversion and hence, have remained un-suited to high-performance applications.
Addressing these requirements in different dimensions, two pulse-radio UWB receiver and a continuous-time filter silicon prototypes are presented in this work. The receiver prototypes focus on system level innovation while the filter serves as a demonstration vehicle for novel circuit architectures developed in this work. The PR-UWB receiver prototypes are implemented in a 65nm LP CMOS technology and are fully integrated solutions. The first receiver prototype is a compact UWB receiver front end operating at 4.85GHz that is aggressively duty-cycled. It occupies an active area of only 0.4 mm², thanks to the use of few inductors and RF G_m-C filters and incorporates an automatic-threshold-recovery-based demodulator for digitization. The prototype achieves a sensitivity of -88dBm at a data rate of 1Mbps (for a BER of 10^-3), while achieving the lowest energy consumption gradient (dP/df_data=450pJ/bit) amongst other receivers operating in the lower UWB band, for the same sensitivity.
However, this prototype is limited by idle-time power consumption (e.g., bias) and lacks synchronization capability. A fully self-duty-cycled and synchronized UWB pulse-radio receiver SoC targeted at low-data-rate communication is
presented as the second prototype. The proposed architecture builds on the automatic-threshold-recovery-based demodulator to achieve synchronization using an all-digital clock and data recovery loop. The SoC synchronizes with the incoming pulse stream from the transmitter and duty-cycles itself. The SoC prototype achieves a -79.5dBm, 1Mbps-normalized sensitivity for a >5X improvement over the state of the art in power consumption (375pJ/bit), thanks to aggressive signal path and bias circuit duty-cycling. The SoC is fully integrated to achieve RF-in to bit-out operation and can interface with off-chip, low speed digital components.
Finally, switched-mode signal processing, a signal processing paradigm that enables the design of highly linear, power-efficient feedback amplifiers is presented. A 0.6V continuous-time filter prototype that demonstrates the advantages of this technique is presented in a 65nm GP CMOS process. The filter draws 26.2mW from the supply while operating at a full-scale that is 73% of the V_dd, a bandwidth of 70MHz and a peak signal-to-noise-and-distortion ratio (SNDR) of 55.8dB. This represents a 2-fold improvement in full-scale and a 10-fold improvement in the bandwidth over state-of-the-art filter implementations, while demonstrating excellent linearity and signal-to-noise ratio. To sum up, innovations spanning both system and circuit architectures that leverage the speeds of nanoscale CMOS processes to enable power-efficient solutions to next-generation wireless receivers are presented in this work.
- Vigraham_columbia_0054D_11891.pdf application/pdf 6.98 MB Download File
More About This Work
- Academic Units
- Electrical Engineering
- Thesis Advisors
- Kinget, Peter R.
- Ph.D., Columbia University
- Published Here
- May 14, 2014