The Analysis and Design of Phase-tunable Low-Power Low-Phase-Noise I/Q Signal Sources for Analog Phase Calibrated Transceivers

Due to the demand for low-cost, small-form factor and large-scale integration of system-on-chip wireless transceivers, the image-reject, zero-IF and low-IF receiver architectures have become the main topologies used in mainstream wireless communication systems. Consequently, signal sources with quadrature phase outputs [quadrature oscillators (QOs)] are therefore essential, and their phase noise, driving capability, tuning range, oscillation frequency, and power consumption have a major impact on the overall receiver performance. Additionally, it is required that the QO synthesize precise I/Q waveforms across the signal bandwidth over process, voltage, and temperature variations for adequate image-rejection and signal modulation/demodulation. While the use of symmetrical layout and large inter-digitated devices minimize both systematic and random mismatches, this solution alone may not succeed in achieving the stringent performance requirements dictated by modern wireless standards particularly as the technology scales into the sub-100nm regime, necessitating both phase and gain calibration of the mismatched I/Q channels post-fabrication. Given the necessity for precise RF quadrature signal synthesis, the goal of this work is to investigate low-power low-phase-noise quadrature oscillator (QVCO) topologies with an integrated phase calibration feature.

The first part of this work focuses on the analysis and modeling of cross-coupled LC QVCOs. The analysis focuses on understanding the oscillator basic performance characteristics, design trade-offs, phase-noise performance, effect of including phase shift in the coupling paths, and on examining the quadrature accuracy in presence of process variations. New design parameters and circuit insight are developed and a generalized first order linear model and a one-port model are proposed. Particularly, we introduce the concept of an effective core and coupling transconductances to explain various oscillator properties. Additionally, a new incremental circuit element — the quadrature resistance — is introduced to evaluate the effect of coupling on the open-loop quality factor and hence on the oscillator phase noise performance. Mechanisms affecting the mode selectivity are identified and modeled. A qualitative and quantitative study of the effect of mismatch on the phase imbalance and amplitude error is presented. Particularly, closed-form intuitive expressions of the phase imbalance and amplitude error are derived and verified via circuit simulation.

Based on our understanding of the various mechanisms affecting the quadrature accuracy, the second part of this work introduces a very efficient quadrature phase calibration technique based on the disconnected-source parallel-coupled LC QVCO topology. The phase-tunable LC QVCO (PT-QVCO) achieves an ultra-wide I/Q phase tuning range without affecting the relative amplitude error or consuming additional power or chip area. Additionally, in restoring the phase balance, it is observed that the proposed method restores the phase noise performance to its optimal value which presents a potential advantage over classical calibration techniques. Time domain measurements performed on a 5 GHz prototype show that I/Q signals with phase error up to ~±30°, beyond which the VCO cores are unlocked, can be driven to perfect quadrature phase. The PT-QVCO can be tuned from 3.87-4.45 GHz at the negative mode and 4.4-5.4 GHz at the positive mode, a total of ~1.5 GHz. The fabricated circuit including pad structures occupies an area of 1.1x0.7 mm² and drains 18mW (excluding buffer circuits) from a 1.8 V supply voltage.

The third part of this work introduces a new low-power, low-phase-noise super harmonic injection-coupled LC QVCO (IC-QVCO) topology. Analysis of the waveform accuracy reveals an inverse dependence of the quadrature error on the tank quality factor thus allowing circuit optimization for both low phase noise and precise quadrature synthesis. Additionally, a tunable tail filter (TTF) is incorporated to calibrate the residual quadrature imbalance in presence of a 3-σ variation in the device parameters. An X-band IC-QVCO prototype with a TTF implemented in a 0.18μm RF CMOS process, achieves a measured phase noise figure-of-merit ranging from 177.3 to 182.6 dBc/Hz along the 9.0 to 9.6 GHz frequency tuning range while dissipating only 9mW from the 1.8V supply. The TTF reduces both the 1/f² and 1/f³ phase noise and calibrates the residual phase error within ±11° post-fabrication without affecting the relative amplitude error or the phase noise performance. The circuit performance compares favorably with recently published work.

In the fourth part of this work, we explore the implementation of LC QVCOs as potential I/Q sources at millimeter-wave (MMW) frequencies. Among the several design challenges that emerge as the oscillator frequency is scaled into the MMW band, precise quadrature synthesis and adequate frequency tuning range are among the hardest to achieve. After describing the limitation of using an MOS varactor and a digitally controlled switch capacitor array for frequency tuning, we propose an alternative frequency tuning technique based on the fundamental operation of LC QVCOs. The off-resonance operation, which is defined by the coupling network, suggests varying the coupling current to achieve frequency tuning. In essence, by modifying the bias current of the coupling transistors (G_Mc-tuning), a wide and linear frequency tuning range can be achieved. Extensive simulation results of a 60 GHz prototype, implemented in a 90 nm commercial RF CMOS process, demonstrates a 5 GHz of frequency tuning range (57.5 GHz → 62.5 GHz), a tuning sensitivity of 1GHz/mA, and a 4dB improvement in the phase noise compared to a varactor solution.

Finally, the Appendix includes recent research work on the analysis and design of gm-boosted common-gate low-noise amplifiers (CG-LNAs). While this topic seems to diverge from the main theme of the dissertation, we believe that the comprehensive analysis and the originality of the circuit design introduced in this work are worth acknowledging.