If you’re planning to attend this year’s combined Symposium on VLSI Technology and Circuits, themed “Inflections for a Smart Society,” which takes place June 13-15, 2016 in Honolulu, HI, and you’re primarily interested in advancements in 3D integration and Internet of Things (IoT), here are some not-to-be-missed papers to put on your agenda.
CEA Leti and STMicroelectronics demonstrate for the first time a full 3D VLSI CMOS-over-CMOS integration, CoolCube™, on 300mm wafers, with the top-level CMOS devices fabricated using low temperature (less than 650°C) processes. A functional 3D inverter with either PMOS over NMOS or NMOS over PMOS is demonstrated to achieve compatible performance with state-of-the-art high performance FDSOI devices. Furthermore, the Leti/STM work demonstrates the integration feasibility of CoolCube™ by transferring a high quality Si layer over the 28nm devices with W-M1 and then returning to the front-end-of-line for processing the top CMOS devices. (Paper T17.3)
For the first time, researchers at Stanford and National Nano Device Laboratories have developed a four-layer HfOx-based 3D vertical RRAM, the “tallest” one ever reported, integrated with FinFET selector. The four-layer 3D RRAM is a versatile computing unit for (a) brain-inspired computing and (b) in-memory computing. Uniform memory performance across four layers is obtained (±0.8V switching, 106 endurance, 104s @125°C). The 3D architecture with dense and balanced neuron-synapse connections provides 55% energy delay product (EDP) savings and 74% VDD reduction (enhanced robustness) as compared with conventional 2D architecture. (Paper T18.2)
In the future “Smart Society,” a proliferation of intelligent devices will be wirelessly interconnected in the IoT. Advances in computation, sensing, and wireless connectivity will enable low-power sensor nodes that can be deployed in applications ranging from ambient intelligence to health monitoring. A focus session “Innovative Systems for a Smart Society,” highlights several such systems, with additional IoT-focused papers throughout the Symposium.
Built around a near-threshold voltage microcontroller, Intel demonstrates an energy harvesting sensor node that includes a solar cell, energy harvester, flash memory, and a BLE (Bluetooth Low Energy) radio. Implemented in 14nm tri-gate CMOS, the IA-32 MCU can operate to the peak energy efficiency point of 308mV supply, consuming just 17pJ/cycle. In indoor lighting environments, the sensor node operates continuously in an “always-on-always sensing” state, with continuous reporting of sensor data to the cloud. (Paper C8.1)
Texas Instruments will demonstrate a multi-modal bio-sensing platform that can synchronously capture electrocardiography (ECG) and photoplethysmography (PPG) data with applications for health, fitness, and mobile patient monitoring. The ultra-low-power frontend can be adapted “on-the-fly” with background cancellation and input-signal-aware data-path adaptation while maintaining >80dB SNR. Paired with a low-power MCU and BLE radio, the full node can operate for three days from a coin cell battery. (Paper C8.4)
Imagers represent the sensing interface that most directly correlates to how humans sense the world. Multiple “More-than-Moore” techniques, including novel photodiode structures, 3D integration, and diffraction gratings that achieve superior sensitivity and speed are demonstrated in the highlighted papers below.
For video cameras that do not support mechanical shutters, a global electrical shutter is essential to eliminate the motion blur seen with rolling shutters. Sony demonstrates a CMOS image sensor for next generation immersive user viewing experiences with 4K high-definition video and up to 480 frames per second (fps) for slow-motion instant replays. Low dark random noise of 140µ Vrms is achieved using gain-adaptive column ADCs stacked on top of the image sensor, with the best dynamic range among high-speed image sensors with large optical format. This work shows the potential for CMOS to displace CCD-based imagers in this application. (C21.1)
TowerJazz Panasonic and Panasonic Semiconductor Solutions have implemented a 10µm deep photodiode that improves the quantum efficiency at 850nm by almost 2x compared to a traditional depth photodiode. Through a two-step epitaxial process on an n-type substrate, each stacked photodiode is completely separated from neighboring pixels, maintaining a 40% modulation transfer function and no degradation in dark current. (T22.1)
An invited paper by Rambus presents an overview of lensless smart sensors that rely on phase-modulated diffraction gratings above a conventional imaging array. Compared to a lens, this “More-than-Moore” diffraction grating can be designed for wide wavelength bands and has a lower profile for thinner sensors. Point range finding, eye tracking, and occupancy detection applications are discussed. Because the desired information can be directly extracted from the raw image sensor data without full image reconstruction, privacy can be preserved. (C8.2)
Healthcare and Bio-sensing
While novel sensor platforms show how the IoT intersects with continuous health monitoring, semiconductor technology can further improve existing systems, or create new diagnostic and prosthetic capabilities. These typically use “More than Moore” techniques of integration with non-CMOS sensors, actuators. For example, Delft University of Technology, Erasmus Medical Center, and Oldelft Ultrasound present a front-end ASIC for a 3D transesophageal echocardiography. This ASIC, interfaced to a 1024-element ultrasonic transducer, images the heart from the esophagus. By moving 3×3 beamformers into the front-end ASIC, the number of cables through the gastroscopic tube is reduced by more than 6x, critical for allowing access into the esophagus. Mismatch scrambling eliminates interfering tones from the beamformers’ delay lines. (C4.4)
While still in their infancy, neural implants have the potential to lead to treatments for Parkinson’s disease and other movement disorders. To understand local brain connectivity, high-density electrode arrays are needed. National Chiao Tung University and China Medical University show a neural sensing microsystem that employs vertical integration with dissolvable µ-needles connected through a flexible interposer to ultra-low power recording circuits. AuTSVs replace copper in the interposer for biocompatibility. The vertical integration path not only minimizes dimension, reducing surgical area and promoting success rate, but also reduces loss between the needles and the sensing circuit to improve sensitivity. A 256-channel array at 288µm electrode pitch is demonstrated. (T20.4)
For the full VLSI & CIrcuits Symposium program and to register, visit the website here.