This paper gives an overview of the Blackghost 1.0 system-on-chip (SoC) from Qualcomm Research, which was our first test chip that paved the way toward the commercialization of Qualcomm's most recent ultra-low-power Blackghost SoC family. Specifically designed for battery powered Internet of Things, sensor fusion, wearables, and e-medical applications, this highly integrated SoC delivers high-power efficiency through low-power innovations in architecture and circuit domains. It integrates a small footprint sensor control processor based on ARM Cortex-M0, a vision classifier processor, a streaming DSP hardware accelerator, an ultra-low-power analog-front-end, and an on-die power management unit with direct Li-Ion battery attach capability. The die size of this prototype chip is 3x 3 mm² in a 28LP CMOS process technology. To date, this SoC family has successfully progressed to the mass production of near-threshold computing. The logic computation operates at near-threshold voltages (<0.6 V) at frequencies up to 50 MHz and draws less than 9μ A/MHz from the directly attached battery.
16 April 2021 in IEEE Journal of Solid-State Circuits
This article presents on-chip power delivery hardware comprised of nine event-driven (ED) digital low-dropout voltage regulators (LDOs) for a large digital load. The goal is to address the performance degradations in an LDO's accuracy and dynamic load regulation in the presence of parasitics in the power grid of a load. In particular, we investigate the effects of power grid resistance ( R G ), which becomes worse with the size of a digital load and technology scaling. Two critical problems that we address are: 1) the IR drop and 2) the dynamic voltage droop problems. Employing multiple LDOs across the power grid improves the IR drop for mainly better voltage sensing. To tackle the voltage droop problem, we distribute LDOs with ED control such that the LDO closest to a localized droop can instantly correct it. To further improve the feedback control latency, we also enhance each LDO with a novel domino sampling and regulation technique. We prototype the on-chip power delivery system consisting of 3 ×3 digital LDOs in a 65-nm CMOS. We also devise the framework to analyze the stability of the multi-LDO system across R G values. Measurements show that at 0.5-V (1 V) input, the single LDO exhibits 49.8-mV (94.1 mV) voltage droop for a load current change of 4.04 mA/0.1 ns (13.8 mA/0.2 ns) with a 0.1-nF integrated output capacitor. Also, the nine-LDO system achieves a current density of 248.8 mA/mm 2 (1.118.6 A/mm 2 ) at 0.5-V (1 V) input voltage.
Many of us in the field of ultra-low-Vdd processors experience difficulty in assessing the sub/near threshold circuit techniques proposed by earlier papers. This paper investigates five major pitfalls which are often not appreciated by researchers when claiming that their circuits outperform others by working at a lower Vdd with a higher energy-efficiency. These pitfalls include: i) overlook the impacts of different technologies and different Vth definitions, ii) only emphasize energy reduction but ignore severe throughput degradation, or expect impractical pipelining depth and parallelism degree to compensate this throughput degradation, iii) unrealistically assume that memory's Vdd and energy could scale as well as standard cells, iv) use the highest temperature as the worst timing corner as in the super-threshold, but in fact negative temperature becomes much more detrimental in the sub/near threshold regime, v) pursue just-in-need Vdd to compensate effects of PVT, but without considering the high energy loss on DC-DC converters. Therefore, the actual energy benefit from using a sub/near threshold Vdd can be greatly overestimated. This work provides some design guidelines and silicon evidence to ultra-low-Vdd systems. The outlined pitfalls also shed light on future directions in this field.
1 May 2014 in IEEE Transactions on Electron Devices
Nowadays, electronics face a challenge regarding the power consumption of integrated circuits (ICs). There is a need for new devices that can provide improved switching capabilities. The downscaled electrostatically actuated ohmic switch, as a (re)emerging device, is a promising candidate to meet this need. To bring such a device seamlessly into IC design, it must be accompanied by an accurate, fast and robust analytical compact model. The development and the main characteristics of such a model are described within this paper. Extensive numerical simulations and measurements have been used to validate the model.