GreenWaves GAP9 IoT Application Processor Enables AI on Coin-cell Powered Devices

GAPuino Board

GreenWaves Technologies GAP8 multi-core RISC-V microcontroller was introduced last year for artificial intelligence (AI) at the edge at ultra-low power consumption. GAP8 AIoT SoC was shown to perform inference at 3.7mA against 60mA for the same workload on STM32F7 Arm Cortex-M7 MCU. The company has now expanded its GAP IoT application processor family with GAP9 that delivers five times lower power consumption compared to GAP8 microcontroller while enabling inference on neural networks 10 times larger. Greenwaves GAP9 will bring machine learning and signal processing capabilities to (coin cell) battery operated or energy harvesting devices such as IoT sensors in consumer and industrial markets, wearables, smart building, smart farming and so on. GAP9 is said to combine architectural enhancements with Global Foundries 22nm FDX process to achieve a peak cluster memory bandwidth of 41.6 GB/sec and up to 50 GOPS compute power while consuming only 50mW. The increased memory bandwidth (20x over GAP8) allows for greatly improved detection accuracy while analyzing …

Support CNX Software – Donate via PayPal or become a Patron on Patreon

HackEEG Arduino Shield Reads Signals from Your Brain (EEG), Muscles (EMG), and Heart (EKG)

Biosignals Measurements (EEG) Arduino + Raspberry Pi 4

Biosignals are signals from living beings that can be continually measured & monitored, and some common methods to measure those biosignals include electroencephalogram (EEG) to monitor the electrical activity in your brain, electromyography (EMG) for recording the electrical activity produced by skeletal muscles, and electrocardiogram (EKG or ECG) to measure electrical activity of your heartbeat. Those can be used for brain interfaces which according to a recent Ericsson’s report may become commonplace by 2030 with users just thinking about commands, prosthetic arms, health and disease monitoring, and so on. Starcat has designed the HackEEG shield to experiment with all those three methods using an Arduino board and electrodes. HackEEG features and specifications: TI ADS1299 8-Channel, 24-Bit ADC for biopotential measurements SPI EEPROM for storing configuration data 8x analog-digital conversion (ADC) channels, each with a 24x programmable-gain amplifier (PGA). Up to 4x shields can be stacked on one Arduino Due for a total of 32x EEG channels DMA drivers can read …

Support CNX Software – Donate via PayPal or become a Patron on Patreon

$38 Compute Module PoE Board Works with Raspberry Pi CM3/CM3+ Modules

Compute Module PoE Board-Raspberry-Pi CM3 Compute Module

Raspberry Pi 3 Compute Module was first introduced in 2017 with CM3 and CM3L systems-on-module with or without 4GB eMMC flash for $25 and up before the company launched an update earlier this year with CM3+ modules equipped with a slightly faster processor and up to 32GB eMMC flash for $25 to $40. If you wanted to evaluate the solution for your project you’d have to spend well over $100 for either the official development kit or third-party solutions such as balenaFin which may not support all features of the modules out of the box. We previously wrote about Waveshare Compute Module IO Board Plus similar to the official devkit and going for $50 without the actual module. The company has now designed a more compact board that also supports Ethernet and PoE out of the box selling for just $37.99 plus shipping on Seeed Studio or Waveshare not including Raspberry Pi CM3/3+ module. Compute Module PoE board key features …

Support CNX Software – Donate via PayPal or become a Patron on Patreon

iCommSemi SV6166F is a Low-Power WiFi SoC for IoT and Embedded Systems

S274 SV6166F WiFi SoC Development Board

The other day I wrote about Sonoff Micro WiFi USB Smart adapter featuring a CKW04 WiFi chip, but one person suspected it may be based on iCommSemi SV6166F chip with very similar features and the same 6x6mm form factor. It’s still unclear whether that’s the chip used in Sonoff Micro, but AFAICR I never heard about iCommSemi previously, so let’s have a look at SV6166F in particular. The SV6166F is a low-power single-chip device designed for integration into the Internet of Thing and embedded systems with support for 802.11b/g/n WiFi at 2.4 GHz. SV6166F highlights: MCU – Andes Technology N10 processor with ILM/DLM and I-cache. Memory & Storage 128K ROM and 192 KB SRAM for Instruction and data SRAM in total 8K retention SRAM 2MB (16 Mbit) SPI flash in package Flash and PSRAM controller up to 16MB/64Mb with XIP WiFi Connectivity 802.11b, 11g, and 802.11n 1T1R AP/STA mode – Soft-AP WFA features – WEP/TKIP/WPA/WPA2; WMM/WMM PS Advanced 1×1 802.11n …

Support CNX Software – Donate via PayPal or become a Patron on Patreon

RISC-V based PolarFire SoC FPGA and Devkit Coming in Q3 2020

Microsemi unveiled PolarFire FPGA + RISC-V SoC about one year ago, but at the time, development was done on a $3,000 platform with SiFive U54 powered HiFive Unleashed board combined with an FPGA add-on board from Microsemi. I’ve now been informed that Microchip has announced its Linux-capable PolarFire FPGA+RISC-V SoC would start shipping in Q3 2020 at the RISC-V summit and that a development kit will be sold for a few hundred dollars. PolarFire SoC FPGA   PolarFire SoC FPGA key features and specifications: Mid-Range FPGA optimized for Low Power High-speed serial connectivity with built-in multi-gigabit/multi-protocol transceivers from 250 Mbps to 12.7 Gbps Up to 461k logic elements consisting of a 4-input Look-Up Table (LUT) with a fracture-able D-type flip-flop Up to 31.6 Mb of RAM Power optimized transceivers Up to 1420 18 × 18 multiply-accumulate blocks with hardened pre-adders Integrated dual PCIe for up to ×4 Gen 2 Endpoint (EP) and Root Port (RP) designs High-Speed I/O (HSIO) supporting …

Support CNX Software – Donate via PayPal or become a Patron on Patreon

Intel Roadmap 2019-2029 – 1.4 nm Processors Expected within 10 Years

ASML Intel Manufacturing Roadmap 2029

A roadmap from a slide by ASML presented at the IEDM conference recently shared on Anandtech shows Intel Roadmap for the next ten years with 7nm manufacturing process expected in 2021, 5nm in 2023, 3nm in 2025, 2nm in 2027, and 1.4nm in 2029. Intel did however contact Anandtech to explain the slide from ASML is a modified version of the Intel slide shown below that does not show actual process nodes, only the dates. I suppose Intel does not want to make any commitments seeing how their 10nm technology suffered delays after delays, and in any case the actual dimensions of the process may vary by that time frame depending on difficulties or new discoveries made. Both slides mention +, ++, and backport opportunities for all new processes. + and ++ are just iterative improvements for the current process, while back- porting is the option to port a process developed on the latest node to the previous ++ process …

Support CNX Software – Donate via PayPal or become a Patron on Patreon

Google ADT-3 is a Developer-Focused TV Box for Android TV on Android 10

Google ADT-3 Android TV 10 Developer-Kit

Back in 2014, Google killed Google TV and announced Android TV, and as a result, introduced ADT-1, the first developer kit specifically designed for Android TV.  Since then ADT-2 was introduced in 2018 in an HDMI stick form factor, and now the company has announced the launch of Android TV on Android 10 together with ADT-3 devkit targetting Android TV app development for Android 10. The “developer kit” just looks like a standard TV box, although It’s really tiny and comes with what looks like a voice remote control. Google did not provide the full specifications, but ADT-3 is equipped with a quad-core Cortex A53 processor coupled with 2GB DDR3 memory, and supports 4Kp60 HDR video playback and output via an HDMI 2.1 output port. That probably means the box is powered by an Amlogic S905X2 processor since few other TV box SoCs combine 4x A53 cores with HDMI 2.1. The device is pre-certified, and will receive updates and security …

Support CNX Software – Donate via PayPal or become a Patron on Patreon

SiFive Learn Inventor is a Wireless RISC-V Development Kit Inspired by BBC Micro:bit

SiFive Learn Inventor is a RISC-V educational board partially inspired by BBC Micro:bit board with the same crocodile clip-friendly edge connector, and an LED matrix. The board is also fully qualified to work with the Amazon FreeRTOS real-time operating system. Shaped in the form of a hand, the board features SiFive FE310 RISC-V processor found in the SiFive HiFive1 board, as well as ESP-WROOM-32 WiFi + Bluetooth module. SiFive Learn Inventor specifications: SoC – SiFive FE310-G003 32-bit RISC-V (RV32IMAFC) processor @ 150 MHz with 64KB of internal SRAM Storage – 512 KB flash “Display” – 6×8 “widescreen” array of RGB LEDs with 262,000 colors each; LEDs can expand off-board onto external arrays via the edge connector Wireless Connectivity – 802.11b/g/n WiFi 4 (2.4GHz) and Bluetooth 4.2 LE via an ESP32 module (ESP-WROOM-32) USB – 1x Micro USB port for power and programming/debugging Expansion A/D Converters (four) accessed via on-board coprocessor BBC Micro:bit compatible edge connector with I2C, SPI, UART, GPIO, …

Support CNX Software – Donate via PayPal or become a Patron on Patreon