12 July 2019 in IEEE Journal of Solid-State Circuits
This paper proposes computational locking (C-lock) in all-digital phase-locked loops (PLLs) to achieve rapid frequency and phase lock acquisition. The proposed approach employs a “lock-accelerator” module that accelerates lock-times ( $T_{\mathrm{ lock}}$ ) in both cold-start (PLL power-up) and re-lock scenarios without impacting steady-state PLL performance. The key premise underlying C-lock is that solving a system of accurate phase-frequency update equations at run-time provides superior lock performance compared to traditional PLLs that rely on linear feedback control. C-lock is readily amenable to runtime adaptation to Process, Voltage, and Temperature (PVT) variation. A 1–2 GHz computationally locked PLL was implemented in a 65-nm standard CMOS process. Measurements over 50 000 iterations of re-lock and cold-start experiments across multiple test-chips, supply-voltage ( $V_{\mathrm{ dd}}$ ), and operating temperatures indicate a mean re-lock time of 12 PLL reference clock cycles (12 $T_{\mathrm{ REFCLK}}$ ), and a “cold-start” $T_{\mathrm{ lock}}$ of 16 $T_{\mathrm{ REFCLK}}$ . Although the silicon demonstration of C-lock is based on a PLL designed for system clocking applications, the proposed approach is applicable to a broader class of PLLs.
11 January 2019 in IEEE Journal of Solid-State Circuits
17 September 2019 in IEEE Journal of Solid-State Circuits
Traditional digital systems employ independent loops to control supply voltage (V dd ) and clock frequency (f clk ). A clock regulation loop, for instance, a phase-locked loop (PLL), locks the system clock to a reference clock (REFCLK). Concurrently, a voltage regulation loop sets V dd to a target value under rapidly varying load current or line-side disturbances. Limitations in voltage regulator loop bandwidth result in significant V dd fluctuation, leading to timing slack degradation in digital voltage domains. Maintaining timing slack in the presence of supply noise or temperature (T) variation incurs energy-wasteful V dd guardbands. In this article, we present a unified clock and power (UniCaP) architecture that instantaneously adapts f clk to V dd and T variations, requiring minimal guardbands. UniCaP absorbs V dd control into the clock regulation loop, using it to adjust f clk and lock system clock to REFCLK. UniCaP enables all-digital construction with aggressive V dd guardband reduction and the capability for on-the-fly dynamic voltage and frequency scaling (DVFS) events. We deployed the UniCaP architecture on a 65-nm buck converter test chip powering a 0.6-1.0-V Cortex-M0 microprocessor with autonomous transition between continuous and discontinuous conduction modes (CCM/DCM) in order to support a wide load current range. The UniCaP test chip demonstrates 82% average Vdd guardband reduction, without any performance loss from adaptive clocking, and temperature margin reductions of 40-55 mV.
20 December 2019 in IEEE Journal of Solid-State Circuits
This article describes an architecture that autonomously sets the operating supply voltage (Vdd) of a digital system for minimum total energy dissipation, inclusive of regulator loss. The technique directly computes the energy per cycle (EPC) delivered by the supply to the system, to guide Vdd adjustment for minimum energy-per-cycle point (MEP; voltage: VMEP and frequency: fMEP) tracking. The proposed approach avoids significant Vdd margins typically required for system operation near-threshold and at sub-threshold Vdd levels by employing a unified clock and power (UniCaP) architecture that regulates target performance across process, voltage, and temperature (PVT) variation with minimal guardband requirements. Measurement results from a 65-nm CMOS test chip demonstrate the VMEP tracking with less than 5 mV of error within the operating voltage range of 0.38-0.58 V. Measurement results also demonstrate the MEP tracking with temperature variation from -30 °C to 105 °C across varied loading conditions over 20 different test chips.