The Crossover Processor – Innovation Using Knowledge
• Built on i.MX RT102F crossover MCU. • Optimized light-weight inference engine. •. Face detect, anti-spoofing, face alignment, face recognition. •. MobileNet ...
EUF-SMH-T3850 Build Your Next Industrial IoT Product Using NXP's Powerful, Power-Efficient, Cost-Effective and DesignCompatible NXP Microcontrollers NXP Techday Milan, 3 October 2019 NXP.COM/MCU What computing meant to our parents (last century ) Mainframe Mini-Computer PCs 2 Mobile, the "new" computing revolution (circa 2007!) Mainframe Mini-Computer PCs Mobile 3 Tomorrow arrived. AI/IoT 4 "The 4th Tectonic Shift in Computing" Units Shipped (Millions) 1950 1960 1970 1980 1990 2000 2010 2020E - Mainframe Mini-Computer PCs 10,000 1000 100 10 1 0.1 0.01 Cell Phones AI / IoT 5 1. Source: Jefferies "The 4th Tectonic Shift in Computing" market requirements (to enable these embedded systems) Technology Accessible · Scalable hardware platforms · Power, performance, integration · Safe & secure (root of trust) solutions to sense, process, connect, control 6 · Global reach · Robust S/W enabled solutions · Time to market · Support Accessible Technology · established MCU/MPU industrial leadership · scalable compute platforms · software, solutions and enablement · global distribution & support model · commitment to quality, longevity, supply 7 Uniquely positioned to enable the AI/IoT era, bringing technology to where it matters most, you. 8 Machines are now beating Humans at Inference 26 Deep Neural Nets Introduced ImageNet Large-scale Image Identification contest Top-5 Error (%) 16.4 11.7 2011 (XRCE) 2012 (AlexNet) 2013 (ZF) 7.3 6.7 Machines beat humans 5 3.6 3.1 2014 (VGG) 2014 (GoogLeNet) Human Level 2015 (ResNet) 2016 (Inception v4) 9 Source: https://www.embedded-vision.com/industry-analysis/blog/deep-learning-five-and-half-minutes Machine Learning Is Will Be Everywhere Voice Processing Smart Sense & Control Gesture Control Active Object Recognition Personal / Property Home Environment Multicamera Observation Augmented Reality 10 What can machine learning do Regression (Calculation) · Predict continuous values Classification (Choice) · Recognition, object detection Anomaly detection (Judgement) · Detect abnormal conditions Clustering · Discover patterns / partitions Learn strategies · Reinforcement Learning 11 X=a, y=? It is a ( ) A: Dog B: Cat C: Cow D: Neither Heart is going to malfunction? Y/N Find crowds No need labels How to play the game? What Does a Convoluted Neural Net Actually Do? Take This Image Take This (4k image) Extract unique values that can identify it Infer What It Is (city) Turn it Into This (28x28) 12 ML Deployment A to Z Data Mgmt. & Training Models Machine Learning Enablement and Optimizations Developer Environment (MCUXpresso and Yocto) Target-Specific Enhancements Edge Compute Pre-trained Models ML Software Development Environment Inference Engine Deployment Inference & Edge Training 13 COMPANY CONFIDENTIAL 13 eIQ Components Phased Development Approach Enable Optimize Advanced 1. Feasibility · Open Source Components · Benchmarking · Potential to Optimize/Advance 2. Port · Apps processors and/or MCUs · Build recipes/makefiles · Test examples 3. Deliver (MCUXpresso, Yocto) 1. Optimize Open Source Components · Support community process · NXP proprietary mechanisms 2. Model optimization (off-line) · Quantization, pruning, compression 3. eIQ Sample Applications 4. Deliver (MCUXpresso, Yocto) 1. Optimize ML Functions/ Kernels · GPU · DSP · MLa 2. Heterogenous Compute 3. Advanced model optimization 4. Deliver (MCUXpresso, Yocto) Receive Customer feedback, Improve 14 Receive Customer feedback, Improve Receive Customer feedback, Improve NXP Processor and Microcontroller Leadership Embedded MCU & MPU low power, low latency (optional transmit) high transport cost, power & latency execute complex tasks locally ... rethink compute strategy Devices 15 Edge Cloud Why Now? · Low Cost of Technology o Unimaginable device/edge performance capability & efficiency budgets o Sophisticated and accurate sensors o Cloud compute & data storage o Global bandwidth ubiquity · Neuroscience advancements · ML/AI/GP embedded accelerators (voice/pattern/face/gesture recognition, object detection/segmentation, natural language learning, math/cryptography, etc.) 16 NXP Processor and Microcontroller Offering APPLICATIONS PROCESSORS · Full open-source OS platforms · ARM Cortex-A class (+ opt. Cortex-M cores & accel.) · Memory Management Unit · Software abstraction · Rich HMI MICROCONTROLLER · Low-level SW drivers · ARM Cortex-M class (+ opt. accelerators) · Memory Protection Unit · Real-time tasks · Easy to Use 17 Industrial & IoT Microcontrollers Broad Portfolio Expansive Ecosystem Technology Insights High-performance crossover processors to Low-power secure and connected MCUs Common software platform combined with large partner network Learning from our 26,000+ end customers across thousands of diverse applications COMPANY PUBLIC PU1B8LIC Microcontrollers serving numerous end market segments Industrial Energy Secure Transactions Appliance Home, Appliance and Tools Personal Devices Building Control Gaming 19 The Spectrum of Edge Processing MCU 8/16/32bit core On-chip SRAM Wired Connectivity (USB 1.0/2.0, UART, SDIO) Embedded flash Mixed signal analog Crossover Processors 32/64-bit core(s) On-chip SRAM E-NVM Security High performance Mixed Signal Analog Wired Connectivity Wireless Connectivity Graphics Codecs DSP Applications Processors 64-bit cores On-chip RAM Graphics Acceleration 32-bit cores Security & Crypto Accelerators HPMS Machine learning High-speed connectivity Display Codecs Embedded DSP Networking Acceleration · Billions of devices · Balanced power, integration · Many low standby, multi-year operation · Simplicity with ease of use · 100s of millions of devices · Power/performance balance · Hours to weeks of operation · Need for energy efficiency · 100s of millions of devices · High performance is key · Hours to days of operation 20 · For representation only; Blocks not drawn to scale NXP's Scalable Processing Continuum Software compatibility & ease-of-use Performance i.MX RT Amplify market deployment with comprehensive software, solutions and product enablement Functional Integration 21 Look Ahead NXP's Arm® Cortex® -M Portfolio Evolution World's Highest Performing MCU Performance & Integration i.MX RT · Up to 6,500 CoreMark for massive compute potential · Unlimited memory extensibility with added 128KB to 5MB SRAM · Advanced integration for real-time control and HMI systems First 40NVM MCU Platform Power Conscious IIoT Sensor Nodes 32-bit replacement for 8-bit market; Housekeeping Mainstream Efficiency LPC5500 · 400 1200 CoreMark with superior dynamic energy efficiency · Accelerators to offload CPU intensive compute · Enhanced security with TrustZone, Secure Boot and SRAM PUF · Design simplicity Ultra-low Power K32 L · 100 400 CoreMark with low leakage optimizations · Autonomous, power optimized peripherals · Path to secure element grade system protection Value LPC800 · 15 75 CoreMark with low cost optimizations · Rich integration inc. DMA, 12-bit ADC, programmable logic, EEPROM, and more · Package scalability with options through 64KB flash memory Metrology & 5V/Touch Vertical Focus KE & KM · Industrial safety with IEC 60730 Class B · Certified, Water-tolerant touch; IEC 6100-4-6 level 3 EMC/robustness (KE MCUs) · AFE & LCD driver for metering (KM MCUs) · Certified by VDE/UL Early LPC4300/1800 LPC54000 KL Series LPC1100 - Product Families LPC4000/1700 K /KV Series 22 i.MX RT CROSSOVER PROCESSORS · Arm Cortex-M cores (v7m/v8m) · Up to GHz performance · DSP & Co-processor Acceleration · High integration · Deterministic instructions · Short latency · Easy to use tools · RTOS support 23 NXP i.Mx RT MCU Series ROADMAP Integrated SRAM Generalpurpose MCUs 35 MB 2MB 1MB 512KB 256KB 24 Performance Arm CM33 Arm CM7 i.MX RT1060 600MHz Cortex-M7, 32K/32K L1 1MB SRAM, 8/16-bit EMI (SDRAM/SRAM) LCD / CSI / 2D acceleration Standard Security i.MX RT1050 600MHz Cortex-M7, 32K/32K L1 8/16-bit EMI (SDRAM/SRAM) LCD / CSI / 2D acceleration BGA196 Standard Security i.MX RT1020 500MHz Cortex-M7, 16K/16K L1 LQFP144 LQFP100 256KB SRAM, 16-bit EMI (SDRAM/SRAM) Standard Security 2017 i.MX RT600 >250 MHz Cortex-M33 with TrustZone Advanced Audio Integration Advanced Security, SRAM PUF, root of trust 4.5MB SRAM for IoT Stacks and Algorithms i.MX RT1170 1GHz Cortex-M7, 32K/32K L1 8/16-bit EMI (SDRAM/SRAM) 2D graphics acceleration, MIPI DSI/CSI Automotive, TSN, Security BGA289 i.MX RT1064 600MHz Cortex-M7, 32K/32K L1 1MB SRAM, 4MB Flash 8/16-bit EMI (SDRAM/SRAM) LCD / CSI / 2D acceleration Standard Security BGA196 i.MX RT1015 i.MX RT1010 500MHz Cortex-M7, 128KB SRAM, QSPI 16K/16K LL1QFP100 Standard Security 2018 500MHz Cortex-M7, 16K/16K 128KB SRAM, QSPI L1 LQFP80 Standard Security 2019 VFBGA176 Product Proposal Planning Execution Production Samples i.MX RT1170 Crossover MCUs Features Performance 1GHz Cortex-M7 with 512KB TCM 400MHz Cortex-M4 with 256KB TCM 6468 total CoreMarks Advanced Security EdgeLockTM 400A Secure Boot High-Performance Crypto On-The-Fly Memory Encryption & Decryption Tamper Detection Rich Feature Set 2MB SRAM 2D GPU and 2D accelerator 2x 1Gbps Ethernet MIPI CSI / DSI Industrial, consumer and auto-grade Low Power 28nm FD-SOI Process Optimized for both active power & leakage power i.MX RT1170 MCU 25 Market Segments Industrial IoT Factory automation Programable Logic controller Unmanned vehicles Smart retail Access control Consumer/Healthcare Smart home Professional audio equipment Patient monitoring equipment Automotive In-vehicle HMI Two-wheel motorcycle/scooter cluster i.MX RT1170 Crossover MCU Family · Specifications - Process: SEC 28FD-SOI - Core Voltage: 1.0V - Package: MAPBGA289, 14x14mm, 0.8mm pitch - Temperature: -40C to 125C (Tj) · Key Features and Advantages - Arm Cortex-M7 processor, 1 GHz, 32KB/32KB L1 Cache, 512KB TCM - Arm Cortex-M4 processor, 400MHz, 16KB/16KB L1 Cache, 256KB TCM - 2MB on-chip SRAM (including TCM for CPU core) - Parallel LCD Display up to WXGA (1280x800) - 8/16-bit Parallel Camera Sensor Interface - 2-lane MIPI CSI and 2-lane MIPI DSI - 2D Graphics Acceleration & OpenVG Acceleration - 8/16/32-bit SDRAM controller up to 200MHz - 8/16-bit Parallel NOR FLASH / NAND FLASH / PSRAM - 2x QSPI NOR FLASH / HyperRAM / HyperFLASH Interface - 2x eMMC 5.0/SD 3.0/SDIO Port - 2x USB 2.0 OTG, HS/FS, Device or Host with PHY - Audio: 4x I2S/SAI, 1x S/PDIF Tx/Rx, ASRC, digital microphone input - 3x ENET: 1Gbps ENET w/ AVB + 10/100 ENET w/ IEEE 1588 + 1Gbps ENET w/ TSN - 3x 12-bit ADC, 2Msample/s, up to 24 input channels total - 8x Analog comparator, 2x DAC - Full PMU Integration, DCDC+LDOs - Secure Boot, TRNG, RSA4096, Tamper Detection, Secure Key Storage · Enablement - MCUXpresso, FreeRTOS with SDK - 2A6utosar i.MX RT Series Key Highlights High Performance Real-time Processing · Cortex-M7 up to 1GHz · Cortex-M33 up to 275MHz with Trust Zone (RT6xx/5xx) · Up to 600MHz DSP and other co-processor accelerators (RT6xx/5xx) · Up to 5MB On Chip SRAM Low BOM Cost · Competitive Pricing starting @ $0.99 100K RSL · Fully integrated PMIC with DC-DC · Low cost packages enabling 4 layer (BGA) & 2 layer (LQFP) PCB design · Quad/Octal Serial and SDRAM interface High Level of Integration · High Security enabled by AES-256, HAB and On-the-fly QSPI Flash Decryption · 2D graphics acceleration engine with Parallel & MIPI DSI/CSI · LCD display controller up to WXGA (1366x768) · Digital microphone and I2S interfaces for multichannel, high performance audio and voice 27 · Up to 3x Ethernet and 3x CANFD Easy to Use · MCU customers can leveraging their current toolchain (MCUXpresso, IAR, Keil) · Rapid and easy prototyping and development with NXP FreeRTOS, SDK, ARM mbed and the global ARM ecosystem · Single voltage input simplifies power circuit design · Scalability to LPC, Kinetis & i.MX products Breaking the `GHz' Barrier i.MX RT1100 MCUs 1GHz Cortex-M7 with 512KB TCM Over 6000 Coremark High Performance 28nm FD-SOI Process Optimized for both active power & leakage power Low Power Cortex-M7 Up to 1GHz Cortex-M4 Up to 400MHz Secure Resource Controller 28 Secure Boot, PUF On-the-fly Crypto Tamper Detect Advanced Security Up to 2MB SRAM High-speed ADC, DAC 1Gbps Ethernet MIPI CSI / DSI Rich Feature Set i.MX RT1010 World's Most Cost Effective Crossover MCU Breaks the $1 barrier, enabling replacement of mainstream, low-cost MCUs ½ Price of Competition - $0.99 for volume production 2x Performance over Competition - Up to 500MHz frequency - 128KB SRAM/TCM - HS USB OTG, I2S, FlexPWM 29 i.MX RT600 Crossover MCU Evolution: Power-Efficient Edge Compute · High performance compute: - up to 275 MHz Cortex®-M33 - up to 600 MHz Hi-Fi4 DSP - Up to 5MB of on-chip SRAM · NXP designed DSP accelerator (PowerQuad) - Tightly coupled co-processor to M33 core (v8-M feature) - 5-10x faster for Matrix, FIR, Convolution, Correlation ops - 15x faster running CMSIS-DSP library for FFT / IFFT (while maintaining ecosystem & toolchain compatibility) · Embedded Tensilica HiFi 4 DSP for NLP - Tensilica Instruction Extension (TIE) for Sigmoid transfer function accelerating wake-word Real-time Domain CortexM33 Accelerators Machine Learning Domain DSP Security GPU MIPI 30 NXP's LPC5500 MCU Series Single & Dual-core Arm Cortex-M33 MCU Series from $0.99 at 10Ku · Ultra-efficient 40nm flash technology: - Over 1150 CoreMarks1 and as low as 32uA/MHz2 - 10x improvement for signal processing & cryptography · Numerous mixed-signal interfaces to sense, connect, control · Enhanced safety and security with TrustZone · MCUXpresso developer ecosystem 31 1: 2xCM33 @ 150MHz, 2: 1xCM33 @ 100MHz NXP LPC5500 MCU Series ROADMAP COMMON PLATFORM ARCHITECTURE SCALABILITY, PERFORMANCE & AREA EFFICIENCY Subject to change Low Cost HMI LPC558x/S8x Advanced Efficiency & Integration General Purpose LPC55S6x Balanced 150 MHz Cortex-M33 with TrustZone 640KB Flash, 320KB SRAM Dual-core, PowerQuad, SDIO Advanced Security General Purpose LPC552x/S2x 150 MHz Cortex-M33 w/ TZ Up to 512KB Flash, 256KB SRAM FS & HS USB, SDIO Opt. Advanced Security Low Cost HS USB / CAN-FD Entry LPC551x/S1x 100 MHz Cortex-M33 w/ TZ Up to 256KB Flash, 96KB SRAM FS & HS USB, CAN-FD Opt. Advanced Security 150 MHz Cortex-M33 w/ TZ Up to 1.5MB Flash, 512KB SRAM Advanced integration, high speed interfaces Motor Control / Appliances / Metering LPC557x/S7x 150 MHz Cortex-M33 w/ TZ Up to 1.5MB Flash, 512KB SRAM Advanced integration, high speed interfaces Motor Control / Appliances / Metering LPC553x/S3x 150 MHz Cortex-M33 w/ TZ Up to 256KB Flash, 96KB SRAM Advanced Analog/Timers, CAN-FD Opt. Advanced Security Lowest Cost Cortex-M33 MCU LPC550x/S0x 100 MHz Cortex-M33 w/ TZ Streamlined Integration Opt. Advanced Security 32 2019 2020 2021 Arm CM7 Arm CM33 Arm CM4 Arm CM0+ NXP LPC5500 MCU SERIES NXP'S SECURE EXECUTION ENVIRONMENT Protecting end products from unexpected threats over their entire lifecycle Root of Trust SECURE BOOT MANAGER HARDWARE PROTECTED KEY STORE & DEVICE ID SECURE DEBUG ACCESS CONTROL SECURE RUNTIME & PHYSICAL PROTECTION TRUSTZONE PROTECTED RESOURCE ISOLATION SECURE DATA ENCRYPTED COMMUNICATION PROTECTED DATA ENCRYPTED FIRMWARE Expanding NXP's MCUXpresso Software & Tools · MCUXpresso Configuration Tools, now including TrustZone Configurator · Flash programming command line tool (vcom/USB HID) including SRAM PUF key provisioning (Blhost for Linux/Mac/Windows) · Host-side command line/GUI tool to create and sign a secure flash image (elf2sb) · Secure debug authentication & debug credential certificate generator tool Third-Party Ecosystem to Reinforce Product Accessibility · From Secure Product Development/Mastering/Provisioning to Programming/Deployment/Management 33 Introducing K32 L MCU Series Secure, Low-power edge/end nodes based on Arm® Cortex®-M combining Low Power Optimizations · Exceptional Sleep Currents · Fast Wake-up High Performance Mixed-Signal · Power Optimized Analog Integration Advanced Security & Protection · Physical Tamper · Authenticate Boot · Crypto Acceleration 34 Integrated SRAM Power Optimized NXP K32 L MCU Series ROADMAP 384 KB 256 KB K32 L3 A60 72 MHz Cortex-M4 Dual Core with 72 MHz Cortex-M0+ 1.25MB Flash, 384KB SRAM Enhanced Security 96 - 128 KB 16 32 KB K32 L2 A31/A41 96 MHz Cortex-M0+ 256-512KB Flash, 128KB SRAM Baseline Security K32 L2 A40 Dual 96 MHz Cortex-M0+ K32 L2 B31 48 MHz Cortex-M0+ 256KB Flash, 32KB SRAM Standard Protection K32 L2 B21 128KB Flash, 32KB SRAM K32 L2 B11 64KB Flash, 32KB SRAM 2019 2020 35 Arm CM7 Arm CM33 Arm CM4 Arm CM0+ Product Proposal Planning Execution Production Samples Introducing K32 L MCU Series Ambition K & KL MCUs 2014 - 2017 · Broad scalable MCUs · Flexible power modes · Rich analog integration · Crypto acceleration, RNG K32 L2/L3 MCUs 2019 K32 L4 MCUs 2020+ · Up to 50% run current reduction · Improved dynamic efficiency & leakage optimization · Autonomous peripherals and flexible power modes · Higher density memories with optimized pricing · Lower power peripherals, analog subsystem · Extended peripheral and analog integration · Addition of tamper protection and secure boot · Further security enhancements 36 CONFIDENTIAL 37 TrustZone Brings Added Security with Resource Isolation · Reduce Security Attack Surface for hackers · Reduce amount of security relevant code - Decreases amount of code that requires security thinking (thinking like a hacker) - Decrease the amount of time needed for negative "hack" testing - Decrease the amount of people that "need to know" security secrets - Allow Non secure code to be developed by less trusted resources - Reduces the scope of security certifications Application Assets 38 Developing code with TrustZone for secure IoT applications Composing a system from Secure and Non-secure projects · Partition project place minimal security related code in secure project · Non-secure project cannot access Secure resources · Secure project can access everything · Secure and Non-secure projects may implement independent time scheduling USER PROJECT Non-secure state SECURE PROJECT Secure state User application Function calls Firmware update Communications stack Function calls RTOS Start Crypto keys, certificates Crypto library Secure boot & bootloader 39 Certified Safety for Home Appliances and Industry 4.0 Appliance IEC60730-1, IEC60335 Class A Class B Class C Industrial* IEC61508 SIL 1 SIL 2 SIL 3 NXP Deliverables · IEC60730/B UL&VDE Certified Self Tests for MCU cores and key peripherals · Dedicated on-chip safety features · Integration with MCUXpresso SDK (in 2019) · Appliance specific safety tests · Expert customer support for FMEA (Failure Mode and Effects Analysis) 40 *SIL2 enablement being investigated for 2020 MCU Kinetis Kxx KV1x/3x/5x KE15/16/18 K32 L3 K32 L4 i.MX RT 1020/50/60 i.MX RT 1170 i.MX RT 1010 LPC800 LPC5500 i.MX8M Mini DSC56F8xxx Nevis3 S08 PB16 IEC60730B Coming! 2H 2019 2H 2019 2H 2019 2H 2019 2H 2019 2H 2019 2H 2019 Commitment to Total Quality & Manufacturing Expertise Process Improvement Product Quality Customer Loyalty Methodology Tools Best Practices in HW & SW Development Customer Satisfaction · ISO 9001 Certification · CMMI · Lean Internal Productivity NXP Quality Mgmt is the heart of SW & HW Design Methodology Supply Chain and Manufacturing Excellence · Maintain excellent delivery performance to our global customers NXP Corporate Quality and HW/SW methodology teams drive Best Practices across NXP · Maintain competitive lead times on a broad mix of products · Guarantee longevity for 10 15 years ... consistently exceed our minimum longevity guarantee 41 Unified Development Software and Tools for MCUs Software compatibility across NXP's broad portfolio of microcontrollers and crossover MCUs. Online Community 42 Online Documentation Online SW & Tools * Highlighting just two example many more from NXP! NXP Solutions, Reference Designs and Demos Low cost Sinusoidal & Trapezoidal Motor Control Detect and recognize face in <250ms · Built on KV MCU Series · MCUXpresso SDK drivers & motor control SW · FreeMASTER, GUI based run-time debug monitor and data visualization tool · MCAT (Motor Control Application Tuner), GUI based FreeMASTER plug-in for real-time monitoring, tuning and updating of motor control system parameters. 43 · Built on i.MX RT102F crossover MCU · Optimized light-weight inference engine · Face detect, anti-spoofing, face alignment, face recognition · MobileNet model, SVM face classifier · Supports on-device learning, cloud learning, and mobile app Comprehensive Ecosystem MCUXpresso Software & Tools MCUXpresso SDK Runtime software including peripheral drivers, middleware, RTOS, demos and more MCUXprersso IDE Edit, compile, debug and optimize in an intuitive and powerful IDE MCUXpresso Config Tools Online and desktop tool suite for system configuration and optimization Partner Software and Toolchains Industry leading support and intuitive software configuration tools to accelerate application development. Probe and Programming Tools NXP and partner solutions for hardware debugging solutions. LPC-Link2 44 NXP and the NXP logo are trademarks of NXP B.V. All other product or service names are the property of their respective owners. © 2019 NXP B.V.Adobe PDF Library 19.12.66