TINY System Ultra-Low Power Sensor Hub for Always-on Context Features

Similar documents
Low-Power Processor Solutions for Always-on Devices

HEXIWEAR COMPLETE IOT DEVELOPMENT SOLUTION

THE INDUSTRY CASE FOR DISTRIBUTED HETEROGENEOUS PROCESSING

Building Ultra-Low Power Wearable SoCs

Overcoming Hurdles in Wearable Device Designs. John Logan Atmel

COL862 - Low Power Computing

Use of ISP1880 Accelero-Magnetometer, Temperature and Barometer Sensor

New Technologies for UAV/UGV

Hello, and welcome to this presentation of the STM32L4 power controller. The STM32L4 s power management functions and all power modes will also be

Renesas Synergy MCUs Build a Foundation for Groundbreaking Integrated Embedded Platform Development

IoT Sensor Connectivity and Processing with Ultra-Low Power, Small Form-Factor FPGAs

A Scalable Speech Recognizer with Deep-Neural-Network Acoustic Models

PBLN52832 DataSheet V Copyright c 2017 Prochild.

New STM32WB Series MCU with Built-in BLE 5 and IEEE

Typical Applications: GHz Bluetooth low energy systems - Proprietary 2.4 GHz systems - Sports and leisure equipment - Mobile phone accessories,

Join the forward thinkers who rely on Toshiba for wireless connectivity ICs.

Ultigesture UG Wristband User Manual

STM32F429 Overview. Steve Miller STMicroelectronics, MMS Applications Team October 26 th 2015

Bike Dash. Vincent Altavilla Aziz Elouali Jose Davila. Sponsor: *Duke Energy *pending approval

New STM32WB Series MCU with built-in Bluetooth 5 and IEEE

CEVA-X1 Lightweight Multi-Purpose Processor for IoT

System-on-Chip Architecture for Mobile Applications. Sabyasachi Dey

Product specification

Designing for Low Power with Programmable System Solutions Dr. Yankin Tanurhan, Vice President, System Solutions and Advanced Applications

EMBEDDED SYSTEMS WITH ROBOTICS AND SENSORS USING ERLANG

Last Time. Making correct concurrent programs. Maintaining invariants Avoiding deadlocks

Wireless-Tag WT51822-S1

From Boolean Algebra to Smart Glass

Wearable is so HOT! Intelligence. Value-added service STM32

Advantages of MIPI Interfaces in IoT Applications

STM32L4 System operating modes

THE GROWING USE OF PROGRAMMABLE LOGIC DEVICES IN MOBILE HANDSETS

BLE to Wi-Fi Gateway

Ultra Low Power Microcontroller - Design Criteria - June 2017

Hexiwear. Contents. Getting started. Wake up. Hexiwear. From MikroElektonika Documentation

MediaTek CorePilot 2.0. Delivering extreme compute performance with maximum power efficiency

The Opportunities and Challenges of IOT Market. YuChuan Yang MediaTek Inc.

PPP (PRICE, POWER AND PACKAGE) OPPORTUNITIES FOR INNOVATION IN MOBILE COMPUTING AND IOT

BLE MODULE SPECIFICATIONS

ARDUINO BOARD LINE UP

Cypress PSoC 6 Microcontrollers

Sensing our world PRODUCT OVERVIEW

ez430-chronos Wireless Watch Development Tool: Teardown & Getting Started

Memory Systems IRAM. Principle of IRAM

Ultra-Low-Power Circuits for Wearables

Low Power System Design Using Atmel ARM Cortex -based Products Copyright Atmel Corporation

Sensing our world PRODUCT OVERVIEW

Smartwatches (April 12, 2017) Samsung Gear Live, 2014 Samsung S 3G, 2014 Samsung S3 LTE, November 2016

SAM A5 ARM Cortex - A5 MPUs

ESP-01 WiFi Module Version1.0

Low-Power Technology for Image-Processing LSIs

Let s first take a look at power consumption and its relationship to voltage and frequency. The equation for power consumption of the MCU as it

Title: Using low-power dual-port for inter processor communication in next generation mobile handsets

Lab 1 Introduction to Microcontroller

An Introduction to the Stellaris LM4F Family of Microcontrollers

Indoor Air Quality with ESP32 Board # K0068

TI SimpleLink dual-band CC1350 wireless MCU

STM32MP1 Microprocessor Continuing the STM32 Success Story. Press Presentation

EERS: Energy Efficient Responsive Sleeping on Mobile Phones

Feasibility of Gateway-less IoT e-health Applications

Approximately half the power consumption of earlier Renesas Technology products and multiple functions in a 14-pin package

Soitec ultra-thin SOI substrates enabling FD-SOI technology. July, 2015

Open Sesame. Grant Apodaca Jeffrey Bolin Eric Taba Richie Agpaoa Evin Sellin

EFFICIENT TRAINING FOR MACHINE LEARNING MODELS ON EMBEDDED AND/OR LOW-POWER DEVICES

Ultra-low power, Single-chip SRAM FPGA Targets Handheld Consumer Applications

An Ethernet Based Control and Monitoring System Using ARM Processor

Sophon SC1 White Paper

Bluetooth Low Energy Portfolio

Fusing Sensors into Mobile Operating Systems & Innovative Use Cases

Engineer-to-Engineer Note

AN4696 Application note

Processor Applications. The Processor Design Space. World s Cellular Subscribers. Nov. 12, 1997 Bob Brodersen (

ARM Cortex core microcontrollers 12 th Energy efficient operation

QuickLogic TAG-N System User Manual

The new maximum security smartphone No Camera - No GPS - No Recorder

ARM mbed Reference Designs

DesignWare IP for IoT SoC Designs

Sensing our world PRODUCT OVERVIEW

AN Sleep programming for NXP bridge ICs. Document information

ARDUINO PRIMO. Code: A000135

Typical Applications: GHz Bluetooth low energy systems - Proprietary 2.4 GHz systems - Sports and leisure equipment - Mobile phone accessories -

Five Ways to Build Flexibility into Industrial Applications with FPGAs

ESP-07S User Manual REV:

Innovative DSPLL and MultiSynth Clock Architecture Enables High-Density 10/40/100G Line Card Designs

STM8L and STM32 L1 series. Ultra-low-power platform

Product Brief. Model: TLM922S-P01A. Ver.1.0

Towards the Consumerization of Smart Sensors

New STM32 F7 Series. World s 1 st to market, ARM Cortex -M7 based 32-bit MCU

Qualcomm Wi-Fi Connectivity Selector Guide

IoT as Enabling Technology for Smart Cities Panel PANEL IEEE RTSI

Freescale s Next Generation 8-bit LCD Solutions

EE 354 Fall 2015 Lecture 1 Architecture and Introduction

Ubiquitous IoT Perspectives The Power of Connected Sensors and Actuators

Leakage Mitigation Techniques in Smartphone SoCs

Use of ISP1807-LR Evaluation Boards

VIDEO BRIDGING SOLUTION PROMISES NEW LEVEL OF DESIGN FLEXIBILITY AND INNOVATION

2-Oct-13. the world s most energy friendly microcontrollers and radios

Outline Marquette University

STM32G0 MCU Series Efficiency at its Best

MAD Gaze x HKCS. Best Smart Glass App Competition Developer Guidelines VERSION 1.0.0

Transcription:

TINY System Ultra-Low Power Sensor Hub for Always-on Context Features MediaTek White Paper June 2015 MediaTek s sensor hub solution, powered by the TINY Stem low power architecture, supports always-on context to efficiently perform motion/gesture detection, activity recognition, and context-aware computing. The cost-effective TINY System takes advantages of the SoC and advanced semiconductor manufacturing processes to provide high performance processor, large SRAM and more peripherals to satisfy rich context-aware applications with ultra-low power consumption..

Introduction Today increasingly more sensors are integrated into mobile and wearable devices to enrich the content and to enable new context-aware features. Sensor-generation raw data are used or computed to be meaningful context, such location, activity, or environment. The context-aware features can be used to create more application services and they are important elements for the development of intelligent devices. Indeed popular context-aware applications can predict the information that a user needs, or know the user location, what the user is doing, or which devices are nearby. A key issue in designing sensor features is power consumption required to support always-on context. When additional power consumption for always-on context is less than 1mW, as seen in the MediaTek embedded sensor hub usage, the additional power consumption is minima similar to that of 3G on standby. Assuming that the power consumption of a smartphone is 3000mAh and 100mA is consumed per hour, the phone can be used for 30 hours. With 1mW (~0.26mA at 3.8V battery side) additional power consumption for always-on context, the phone still can be used for nearly 30 hours. The impact of additional power consumption for always-on context is so minor that users may even not sense it. 2015 MediaTek Inc. 2 of 9

Figure 1. Application Processor without/with Sensor Hub Solutions In traditional chip design architecture, the sensors are connected to the application processor (AP) directly. The application processor must wake up to receive sensor data, as shown in Figure 1(a), which results in high power consumption. The application processor power consumption increases significantly as the number of sensors increases. Indeed the AP will dominate the power consumption of mobile or wearable devices. Already in the market are some solutions for always-on context with ultra low power consumption to avoid waking the application processor to receive sensor data. One solution is the sensor hub, a hardware system isolated from the application processor. Another is the smart sensor, a customized low power sensor. These solutions aim to let the application process wake up only when needed, and to help sensors process data without involving the application processor to save power, as shown in Figure 1(b) above. The smart sensor solution usually has limited capabilities, so only few applications are supported. The sensor hub design is responsible for collecting or computing the sensor data generated by different sensors (e.g., Accelerometer (A), Magnetometer (M), Gyroscope (G), and Proximity (P)). The sensor hub solution usually has two types of designs an external sensor hub and an embedded sensor hub. The external sensor hub is the more costly design. To address these cost and limited application support problems, MediaTek has introduced an embedded sensor hub solution powered on the low power 2015 MediaTek Inc. 3 of 9

architecture of MediaTek s TINY System, a hardware/software integrated sub-system with many lowpower optimization technologies. MediaTek s Embedded Sensor Hub BOM Cost except AP Low Ultra-Low Power Yes, even high loading Peak Performance CM4F (226 MHz) SRAM 512KB Fusion FIFO YES Figure 2. TINY System Features MediaTek s TINY System Sensor Hub Solution MediaTek s cost-effective ultra-low power embedded sensor hub design with always-on context features, shown in Figure 2 above, can reduce the application processor loading and provide outstanding performance. The TINY System combines a hardware/ software platform to mobile or wearable devices to process rich context-aware features and to sense data anytime or anywhere. The TINY System is a very cost-effective solution on the market, as the customer can utilize the same PCB layout. BOM cost without an application processor is one major difference between the external sensor hub, smart sensor, and the TINY System. And the TINY System processor has higher computing capability, thereby allowing for more applications and functions. The system also has larger memory capacity. Indeed the TINY System can work well in both light and heavy workload conditions with very low active power. This excellent performance outshines external sensor hub solutions, which consume much higher computation power for the heavy workload. In addition to the hardware optimization of sensor hub, MediaTek also introduces a sensor hub platform to provide software optimization for always-on context. Therefore, MediaTek aims to provide the best solution for the sensor hub product considering both of the hardware and software designs. 2015 MediaTek Inc. 4 of 9

MediaTek TINY System Sensor Hub Design The TINY System architecture contains four main components, including processor, SRAM, ultra-low power oscillator (OSC) and some peripherals. Key features include: Ultra-Low Power Sensing/Computing High Performance Core Large Internal Low Power Memory Rich Peripherals (for Sensors, and Connectivity) AP Resource Sharing (ex: Memory, Modem system) Figure 3. 2015 TINY Sensor Hub System Hardware Architecture 2015 MediaTek Inc. 5 of 9

Ultra-Low Power Sensing/Computing Three technologies work to achieve the ultra-low power sensing feature. First, the TINY System has independent power domains. This means that the TINY System power switch is separated from other components. The TINY System can operate its own power domain and work independently. Second, TINY System supports some power management methods, such as clock gated, power gated, sleep in the low voltage level, and DVFS (Dynamic Voltage Frequency Scaling). Last, ultra-low power OSC (referred to as ULPOSC) has very low active power and can rapidly be completely set up. High Performance Core The TINY System processor supports both of high performance and high power efficiency. The powerful ARM Cortex-M4F (CM4F) processor is used in the TINY System whose maximum frequency is 226MHz. CM4F can offer DSP acceleration and handle floating point by hardware. It consumes very low dynamic power due to the SoC design and the advanced manufacturing process. Large Internal Low Power Memory The memory cost associated with memory capacity and memory power consumption is an important concern in the sensor hub design. The TINY System provides a large SRAM, 512KB, which can be used to execute more applications or algorithms and let context-computing be more efficient. Figure 4. Fusion FIFO Mechanism In addition to local SRAM, the TINY System proposes a fusion FIFO mechanism to store more sensor data or compute more features with minimal additional power consumption. Typically, the sensor data will 2015 MediaTek Inc. 6 of 9

be only kept in SRAM as shown in Figure 4(a). If SRAM has no free space for storing new data, the AP needs to be awakened to handle the to-be-processed sensor data. When the SRAM size is small, the AP will wake up frequently and more power is generated. MediaTek proposes the fusion FIFO mechanism, a virtual memory concept, which is a large memory space, and consists of SRAM in the sensor hub chip and the low-power external RAM in the AP. The TINY System fusion FIFO mechanism makes good use of memory hierarchy to provide larger memory capacity and good access performance for always-on context applications with a little additional power overhead. The fusion FIFO mechanism can be used to completely support sensor batching function. Rich Peripherals for Sensors and Connectivity As shown in Figure 3 above, the 2015 TINY System adds more peripherals than MediaTek s 2014 product. The TINY System contains low-power peripherals such as I2C and UART, which can connect various sensors (e.g., accelerometer, proximity sensor, ambient light sensor, magnetometer, gyroscope, barometer, and so on), connectivity systems (e.g., GPS, Wi-Fi, BT), and Modem system (e.g., 2G, 3G). Currently, the up-to-date sensors officially announced by Android, including Android KitKat and Android Lollipop, are supported in the 2015 TINY System. When the connectivity system, modem system, and TINY system are manufactured in the same SoC, the connectivity system and MD system data can be directly accessed by the TINY System, thereby saving more power. AP Resource Sharing (Examples: Memory, Modem System) The TINY System can direct access some application processor resources, such as memory (e.g., DRAM) and modem system, without waking up the application processor. When compared with the external sensor hub solution which needs to wake up application processor to access AP resources, the TINY System can save power. The TINY System has very good scalability to access AP resources in the lowpower state. Based on the hardware architecture, the TINY System can support always-on context and has powerful processor and enough memory resources to process many sensor data. In addition, the TINY System provides a platform, which has enormous software algorithms to help to detect some user activity recognitions and other activities (e.g., gesture wake-up, voice wake-up and geo-fencing). The software 2015 MediaTek Inc. 7 of 9

algorithms are tightly integrated within the hardware design to optimize the TINY System functions. The chart below shows power consumption for Android L motion detection activity recognition + pedometer + significant motion detection. Figure 5. Power Consumption for Android L Motion Detection In a nutshell, MediaTek s TINY System can provide the best solutions for always-on context by considering both hardware and software optimizations. A simple experimental result is illustrated in Figure 5 to show the power efficiency of the 2015 TINY System. Figure 5 shows the additional power consumption for the motion detection supported by Android L. The AP w/ sensor buffer (3 second) represents a simple design, in which the sensor has a hardware FIFO to store sensor data, and then the sensor data are processed or computed by AP. The AP w/ sensor buffer (3 second) has the highest additional power consumption. The 2014 TINY System consumes about 1mW. The 2015 TINY System achieves better power efficiency at a half of power consumption. Conclusion MediaTek s sensor hub solution, powered by the TINY Stem low power architecture, supports always-on context to efficiently perform motion/gesture detection, activity recognition, and context-aware 2015 MediaTek Inc. 8 of 9

computing. The cost-effective TINY System takes advantages of the SoC and advanced semiconductor manufacturing processes to provide high performance processor, large SRAM and more peripherals to satisfy rich context-aware applications with ultra-low power consumption. The TINY System is embedded in the application processor chip, and no additional PCB components are required, resulting in very low additional BOM cost. The Fusion FIFO mechanism is adopted to compute complex context data or to store more sensor data. Moreover, the TINY System can provide the best support for the sensor processing of Android L/M, such as the features of sensor batching, and our proprietary sensor features, including significant motion detection and activity recognition. 2015 MediaTek Inc. 9 of 9