The HW1 by Gesture Platforms is a 10-degree-of-freedom (DOF) high-dexterity robotic hand and wrist built around an ESP32-S3 wireless MCU. It’s primarily designed for researchers, educators, and hobbyists; it bridges the gap between basic DIY robotic hands and expensive industrial models.

The device weighs just around 500 grams but can handle a 1kg dynamic load and a 3kg static load. It communicates via USB-C or Bluetooth 5.0 and comes with a companion desktop app as well as Python and C++ SDKs for custom development.

Gesture HW1 specifications:

• Microcontroller – Espressif Systems ESP32-S3 dual-core LX7 microprocessor with WiFi 4 and Bluetooth 5.0 LE connectivity
• Degrees of Freedom (active) – 10 DOF
  • Finger Flexion – 0° – 90° (at each joint)
  • Finger Adduction/Abduction – -10° – 40°
  • Thumb Flexion/Extension – 0° – 120°
  • Thumb Adduction/Abduction – 20° – 90°
  • Thumb Distal Flexion – 0° – 90°
  • Wrist Flexion/Extension – -50° – 50°
  • Wrist Radial/Ulnar Deviation – -10° – 40°
• Joints – 19 joints total
• Payload – 1kg dynamic / 3kg static load
• Accuracy/Repeatability – ±1mm
• Control Speed – 100Hz
• Sensors – Angle (at motor), Current, Temperature
• USB – 1x USB Type-C port for communications
• Misc
  • 4 x user-programmable RGB LEDs
  • Four-way stretch fabric cover prevents debris from entering joints
• Power – Via the XT60 connector on the side
• Dimensions – ~275mm total height (fits men-small / women-medium gloves)
• Weight – ~500 grams
• Mounting – M3x0.5 threads (depth of 4.5mm)

The HW1 uses aluminum gears at key joints like the wrist for better strength and durability, while its skeleton is made from lightweight, strong glass and carbon-fiber-filled polycarbonate, and the outer shell uses ASA plastic to handle heat and rough use. The fingertips are covered with soft silicone (Shore 30A) to improve grip. Gesture Platforms states that it will provide STEP files, so users can 3D print, modify, or replace the outer parts easily. Most probably, they will release the file after the campaign ends.

On the software side, the company provides a plug-and-play desktop application for Windows 10/11. The app also works completely offline and features four control modes, which include Text (importing .csv files), Timeline (keyframe-style editing), Controller (mapping to a standard game controller), and Stream (real-time input). For more advanced users, C++ and Python SDKs will be available to integrate the hand into custom machine learning and control algorithms.

We have previously written about a range of ESP32-based robotic arms, such as the RoArm-M3-Pro and RoArm-M3-S, the Tobor open-source modular robotic arm, and the MyCobot Robotic Arm, but we haven’t covered robotic hands as much. We did cover the Amazing Hand, an 8-DoF 3D-printable DIY open-source robotic hand, and on the other side of the spectrum, we mentioned the high-end Dex5 dexterous hand in our article about the Unitree R1-A5 and R1-A7 humanoid robot. The Gesture HW1 offers something in the middle in terms of price and functionality. 

The Gesture HW1 ESP32-S3 high-dexterity robotic hand is currently crowdfunding on Kickstarter. The “Launch Day Special” reward requires a $699 pledge (limited to 100 units), while the standard “Early Backer” tier is available for $849. A double pack for both the left and right hands is available for $1,599. Shipping is not included and will be charged after the campaign; deliveries are estimated to begin in November 2026.

The post Gesture HW1 is a 10-DOF ESP32-S3 robotic hand with high-dexterity manipulation (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Secluso is a private, open-source, DIY home security camera system built around the Raspberry Pi Zero 2 W, featuring true end-to-end encryption (E2EE) and on-device AI for human, pet, and vehicle detection. It was designed as an alternative to commercial smart home cameras that require sending raw video feeds to a proprietary cloud, a practice that often raises significant privacy concerns.

Developed by Secluso, Inc., co-founded by UC Irvine professor Ardalan Amiri Sani and John Kaczman, the project utilizes Messaging Layer Security (MLS, RFC 9420) to ensure end-to-end encryption between the camera and the user’s smartphone. Because the system uses an untrusted relay (either self-hosted on a VPS or via Secluso’s free beta relay), the server routing the footage only sees encrypted files and cannot decrypt the video or thumbnails.

Secluso hardware requirements:

• SBC – Raspberry Pi Zero 2W
• Camera – Raspberry Pi Camera Module V1 (OV5647) or V2 (IMX219)
• Audio & Sensors –  HAT with a microphone and safety temperature sensor (included in official kits)
• Enclosure – Official Raspberry Pi Zero enclosure, 3D-printable custom housing, or Secluso’s IR-pass acrylic cover

The software focuses on security and privacy. The core camera hub and server software are now written in Rust rather than in OpenSSL-based C code, making them safer and less prone to memory-related bugs. It also uses post-quantum encryption to protect data even from future attacks where hackers might try to decrypt stored data later. All parts of the system, including Secluso OS and the mobile apps, use reproducible builds, so anyone can verify that the software matches the public source code, and firmware updates are only installed if they come from signed and trusted GitHub releases. The mobile app is available on the iOS App Store and Google Play. Android users can also use Obtainium to download the app directly from GitHub’s release page. For notifications, the system doesn’t rely on Google services and uses its own relay for iOS and supports UnifiedPush on Android, so Firebase Cloud Messaging is optional.

Despite all these features, the system is designed to be deployed in just 5 minutes as the v1.0.2 release introduces Secluso OS, a minimal, Yocto-based Linux image. Users simply flash the image using the Secluso Deploy desktop utility available for Windows, macOS, and Linux. The utility automatically handles injecting unique credentials, configuring the relay over SSH, and setting up the camera without requiring the user to open a terminal. You can find the source code, build instructions, and the detailed security whitepaper on their GitHub released under a GPL license, and the “Build Your Own Guide” explains how to build your own camera. Users can fully self-host the open-source relay/server on their own VPS, so no cloud is required.

For those who don’t want to get into ordering different parts from different vendors and setting up their own server, Secluso is opening pre-orders this month (May 2026). They are offering a limited run of 100 DIY Kits for $50 (which include the night vision camera module, custom mic/temp HAT, and housing parts, and 1 year of Secluso Cloud for free) with an estimated delivery of 2–3 months. A fully assembled Plug and Play version is also available for $100, with an estimated delivery of 4–6 months and a free year of Secluso Cloud. After the included free year, Secluso Cloud is available optionally for $5/month or $50/year. Please note that pre-orders will ship to the US only initially.

Via Hackster.io

The post Privacy-focused, open-source Raspberry Pi Zero 2W DIY security camera offers end-to-end encryption, on-device AI appeared first on CNX Software - Embedded Systems News.

VIA Labs VL610/VL610D MST (Multi-Stream Transport) Hub controllers are designed for multi-display expansion, enabling USB-C adapters to drive up to three 4K UHD high-resolution HDMI or DisplayPort displays simultaneously.

The chips succeed the VL605 single-channel USB-C to HDMI 2.1 signal converter. The VL610 supports three video outputs up to 4Kp60, while the VL610D is limited to two video outputs for different docking design needs.

VIA VL610/VL610D key features and specifications:

• Display/USB Input – DisplayPort Receiver (DPRX) supporting HBR3 (High Bit Rate 3) up to 32.4 Gbps, suitable for up to 8Kp60 and D+/D- USB 2.0
• Video and Audio output
  • VL610
    • HDMI 2.1 FRL transmitter
    • Configurable port supporting DP++ or HDMI 2.1 FRL
    • Configurable port supporting DP++ or HDMI TMDS
  • VL610D – Dual display configuration (not clear details provided)
  • Single display up to 8Kp60 or 4Kp240
  • Triple display up to 4Kp60 or QHDp144.
  • Up to 6x independent audio and video streams with a single DP output supporting up to four MST streams.
  • Full support for color formats and audio
• Misc features
  • DSC 1.2a decoder for decompression to HDMI output or direct pass-through to compatible displays.
  • Variable Refresh Rate (VRR)
  • ECDSA-256 asymmetric authentication
  • Logo Bitmap display capability for brand logos, warning messages, or guidance screens
• Package – QFN100 (10x10mm) for VL610

There’s no product page yet, and VIA only published a press release for COMPUTEX 2026, and I also found an earlier presentation with the slide above. As I understand it, the VL610/VL610D chips are designed specifically for multiple-display DP 2.1 HBR3 MST Hubs, and USB data is limited to High Speed (480 Mbps), so they won’t be found in multi-function USB-C docks.

Thanks to TLS for the tip.

The post VIA Labs VL610/VL610D MST Hub controllers to enable USB-C multi-display adapters with up to three 4K display outputs appeared first on CNX Software - Embedded Systems News.

Broadcom has recently introduced two complementary chips for 50G Passive Optical Network (PON) gateways: the BCM68850 gateway SoC and the BCM55050 Optical Network Unit (ONU) SoC. The devices target fiber-to-the-home (FTTH) infrastructure, such as home gateways, ONTs/ONUs, and multi-gigabit CPE, to deliver ultra-fast, low-latency connectivity for AI workloads, UHD streaming, and more.

The BCM68850 is a 50G ITU-PON gateway SoC with a multicore Arm CPU, integrated NPU for edge AI, and Wi-Fi 8 support, for high-speed routing, Ethernet switching, VoIP, and AI-driven functions. It also supports multiple PON standards along with LPDDR4/5 memory, PCIe, USB, and various security features. The BCM55050, on the other hand, is an ONU/ONT SoC with a dual-core processor and 50G network processor, supporting multiple PON standards, high-speed interfaces, hardware acceleration, and QoS. Together, they are designed for end-to-end 50G fiber deployments with high throughput, low latency, and scalable broadband infrastructure.

Broadcom BCM68850 50G ITU-PON Gateway SoC

The BCM68850 is a single-chip gateway SoC designed to handle small packet WAN-to-LAN routing, L3/L4 packet processing, and tri-band concurrent Wi-Fi 8 wireless networking.

Broadcom BCM68850 specifications:

• Processor – Quad-core Brahma53 (B53) Armv8 CPU
• Network Processor – DI-XRDP with flexible classification and filtering, paired with a dual-issue runner packet processor for line-speed offload of wired and wireless data paths
• AI – Broadcom Neural Engine (BNE) supporting convolutional neural networks for anomaly detection, voice recognition, and cybersecurity analysis
• Memory I/F – Integrated LPDDR4, LPDDR5, and LPDDR5x controller and PHY
• Storage I/F – Support for flexible NAND flash and eMMC
• Networking
  • PON – Integrated Fiber WAN interfaces with 50G ITU-PON, ITU XGS-PON, XG-PON, NG-PON2, 50G HSP PON, 25G EPON, 10G EPON, and Active Ethernet up to 50Gbps
  • Wireless – Native processing for tri-band concurrent Wi-Fi 8 wireless networking
  • Ethernet – Integrated 50Gbps, 25Gbps, and 10Gbps Ethernet interfaces (SLAN) with connection support for an external 10GbE PHY
  • Physical Layer (Transceivers) – Glueless interface to standard SFF transceivers, onboard BOSA, SFP, SFP+, XFP, and SFP56
• USB – 1x integrated USB 3.0 controller
• Other I/Os – Serial Port Interface (SPI), I2C, and Pulse Code Modulation (PCM) highway
• Expansion – Multiple multi-lane PCIe ports, including dual PCIe 3.0 interfaces
• Security – Dedicated security processor (SMC) for secure boot and Post-Quantum Cryptography (PQC) support
• Power – Centralized power management architecture
• Package – BGA (speculative; from part number stamped on the chip BCM68850KFSBG)

The SoC is primarily designed for residential FTTH (fiber to the home), multi-gigabit fiber CPE, and 50G PON Wi-Fi 8 Ethernet gateways.

Broadcom BCM55050 50G PON ONU SoC

For the ONU side, Broadcom has the BCM55050 SoC, a 50G PON chip that combines a dual-core processor, a high-speed 50G network engine, and built-in AI support for edge tasks.

Broadcom BCM55050 specifications:

• Processor – Dual-core Brahma53 (B53) application processor running at 1 GHz, featuring 32 KB I/D caches and a 1 MB L2 shared cache.
• Network Processor – XRDP Network Processor (Dual Issue Runner Data Path) for 50 Gbps Layer 2 packet processing, hardware acceleration, real-time load balancing, classification, and dynamic power reduction
• AI Engine – Integrated AI/ML Neural Processing Unit (NPU)
• Memory I/F – 32-bit wide interface supporting 2133 MHz LPDDR4, with documentation also noting LPDDR5 and DDR3 controller compatibility
• Storage I/F – Flexible flash memory support, including NAND and SPI interfaces, alongside One-Time Programmable (OTP) memory
• Networking
  • PON – Integrated Fiber WAN interfaces supporting 50G ITU PON, NGPON2, XGS-PON, 10GEPON, XGPON, GPON, and EPON
  • Ethernet – Integrated LAN ports with Ethernet MAC that support 100Gbps, 25Gbps, 10Gbps, and 2.5Gbps throughputs, with two RGMII interfaces, four HSGMII interfaces, two 10Gbps XFI interfaces, and two 25Gbps SerDes interfaces
  • Physical Layer (Transceivers) – Internal SerDes circuitry enabling a glueless interface directly to optical transceivers
  • SGMII/HSGMII serial interfaces include support for the G999.1 (G.int) standard.
• Other I/O – PCM, I2C, GPIO, MDIO, and 2x UART interfaces
• Expansion – 1x PCIe Gen 2 (single lane)
• Security – Unrivaled QoS management and hardware security, featuring Secure Boot enforced by a dedicated Security Management Controller (SMC)
• Dimensions / Package – 25 mm x 25 mm FCBGA
• Temperature Range – -40°C to 85°C (Industrial ambient temperature, requiring a heat sink)

Because the processors run independently, the chip can handle both normal ONU tasks and high-speed network traffic simultaneously.

As of 2026, the fastest residential fiber plans typically max out at 8-10 Gbps, such as Google Fiber’s 8 Gig tier. No ISP is currently offering 50 Gbps connections to a single standard home. Broadcom’s angle here is about “network headroom,” handling massive multi-gigabit micro-bursts of data instantly to clear the shared fiber channel faster, thereby minimizing latency for latency-critical applications like AI synchronization and ultra-HD telepresence.

While writing, I came across proprietary and highly specific terms that are worth highlighting, which include:

• Brahma53 (B53) – This is Broadcom’s custom-designed ARMv8-based processor core.
• DI-XRDP / XRDP – This stands for Dual Issue Runner Data Path. It’s Broadcom’s proprietary hardware-accelerated network processing unit (NPU) architecture designed specifically for line-speed packet processing without bottlenecking the main CPU.
• BNE (Broadcom Neural Engine) – This isn’t a generic NPU; it’s Broadcom’s specific ML/AI inference engine designed for edge connectivity tasks (like fault localization and anomaly detection).
• G999.1 (G.int) – An ITU-T standard that defines the interface between the MAC and the PHY in high-speed PON systems.
• Glueless Interface – This means the chip’s internal SerDes (Serializer/Deserializer) can connect directly to standard optical transceivers (like SFP56) without needing any intermediate “glue logic” chips on the PCB. This lowers the BOM cost and saves physical board space.

Previously, we have written about Qualcomm’s Dragonwing Wi-Fi 8 networking platforms, such as the FiberPro A8 Elite, which supports up to 10G PON and supports various wireless features, including 5G FWA and high Wi-Fi speeds, as well as the Broadcom BCM67142, BCM67192, and BCM68565 chips for low-cost WiFi 8 10 Gbps fiber access points. In comparison, Broadcom’s BCM68850 and BCM55050 push fiber performance further with 50G PON for higher capacity. Both platforms also compete in Edge AI, using built-in NPUs to handle tasks like traffic routing, optimization, and detection with low latency.

Broadcom BCM68850 and BCM55050 50G PON SoCs are sampling to early access customers and partners. As usual for enterprise networking chips, you will need to contact a local Broadcom sales representative for exact pricing and sample requests. Some additional information can be found on the Broadcom website and in the press release.

The post Broadcom BCM68850 and BCM55050 SoCs target Wi-Fi 8 and 50G PON fiber gateways appeared first on CNX Software - Embedded Systems News.

Modos Flow is a paper-like, 13.3-inch USB Type-C touchscreen monochrome or color monitor that builds upon the Modos Paper devkit introduced last year with an AMD/Xilinx Spartan-6 LX16 FPGA and STMicro STM32H750 Arm Cortex-M7 microcontroller.

The main difference is that the Modos Flow is more like a consumer product with a full enclosure, a touchscreen, and optional stylus support, 4096-color e-paper display, and frontlight.

Modos Flow specifications:

• FPGA – AMD Xilinx Spartan-6 LX16 FPGA running Caster gateware like the earlier devkit
• MCU – STMicro STM32H750 Arm Cortex-M7 microcontroller for USB communication, firmware upgrades, and standalone applications.
• Display
  • 13.3-inch e-paper display with 3200 x 2400 resolution
  • Refresh rate – 60 Hz with additional power, 40 Hz via a single USB-C cable
  • Monochrome or 4096 colors/16 levels of grayscale
  • Touchscreen support
  • Optimized display modes for reading, browsing, watching, and writing
  • Amber-tinted frontlight (color model only)
• Video Input – USB Type-C DisplayPort Alt-Mode with on-board PTN3460 decoder
• USB – 2x USB-C ports, one for power and DisplayPort Alt mode, one for additional power (and higher refresh mode)
• Misc
  • Stylus support (color model only)
  • VESA-compatible mounting support
• Power Supply – Via USB-C port
• Dimensions – 31.5 x 25.4 x 1.6 cm
• Weight – 690 grams or 1.19 kg with cover and stand

Many e-paper screens feel slow because they refresh the entire display, but with its FPGA-based design, the Modos Flow is different and only updates what’s changing, such as the cursor or the text you’re editing. That’s also probably why it’s one of the few e-paper monitors with touchscreen support on the market, since nobody wants to wait for 19 seconds for a full refresh after a tap. It is compatible with Linux, macOS, and Windows computers.

Modos Flow is open hardware with firmware source code, Caster Gateware code, schematics, and an issue tracker available on GitHub. They still point to the same documentation as for the “Glider” devkit with support for Rust, Python, and C programming language support, so it’s pretty clear that the Flow is based on the same design, although there are some differences, as shown in the comparison table below.

The closest competitor to the Modos Flow is probably the Dasung Paperlike 13K, with a refresh rate of up to 37 Hz, using the same 13.3-inch 3200×2400 E Ink panel, and a similar price. It does lack stylus support, and it’s not an open-source hardware, programmable monitor.

Talking about price, the Modos Flow has just launched on Crowd Supply for $619 (Monochrome) or $719 (Color) with a $150,000 US funding target. Both kits come with a USB Type-C cable and a folding screen cover that doubles as a monitor stand, and the olor version also adds a stylus. Worldwide shipping is included in the price, and deliveries are scheduled to start by mid-December 2026. You can check the videos below (especially the second one) to have a better understanding of the user experience with such displays.

The post Modos Flow – An FPGA-based 13.3-inch USB-C touchscreen e-paper monitor (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Designed by MuseLab, the nanoCH32H417 is a development board for the WCH CH32H417 dual-core RISC-V MCU, which we covered earlier this year for its USB 3.0 (5 Gbps), UHS, and Fast Ethernet support. At that time, only the official CH32H417 development board was available, but this board adds a third-party option.

The board exposes various features of the MCU, including one USB 3.0 port, two USB Type-C ports, a 100Mbps Ethernet interface, and a MicroSD card slot. It also simplifies the prototyping workflow by integrating a WCHLink-E debugger directly onto the board, meaning you won’t need to wire up an external programmer to flash the MCU or view serial output.

MuseLab nanoCH32H417 specifications:

• MCU – WCH CH32H417QEU6
  • MCU
    • QingKe RISC-V5F up to 400 MHz
    • QinKe RISC-V3F up to 144 MHz
  • GPU – Graphics Processing Hardware Accelerator GPHA
  • Memory – 896KB SRAM
  • Storage – 960KB Flash
• Storage – MicroSD card slot
• Display – 12-pin FPC connector with support for common SPI LCDs (e.g., ILI9341, ST7789)
• Networking – 100Mbps Fast Ethernet
• USB
  • 1x USB 3.0 Type-A port (SuperSpeed, labeled USB-3.0-SS)
  • 1x USB Type-C Full-Speed OTG port
  • 1x USB Type-C port for debugging
• Debugging – Onboard WCHLink-E supporting SWD debug and serial port console via USB Type-C port
• Expansion – 4x 13-pin headers exposing GPIO, 5V, 3.3V, and GND.
• Misc
  • Reset, Mode, and IAP buttons (based on standard MuseLab designs)
  • Power LED
• Dimensions – TBD

WCH MCUs can generally be programmed with the MounRiver Studio IDE (MRS), and this is no different. MuseLab has provided a GitHub repository that includes the PDF schematics, English and Chinese documentation, and a basic GPIO toggle project to help verify your toolchain setup before moving on to testing the USB 3.0 or Ethernet stacks. If you prefer an open-source GCC toolchain, the ch32fun project‘s code has a few references to ch32h41, so it might work, although it’s not been added to the list of supported WCH MCUs yet.

The MuseLab CH32H417 development board is available on AliExpress, with prices ranging from $21.49 to $25.50 depending on the configuration, as well as on the company’s Tindie store starting at $17.00. The base variant includes the board and the 13-pin headers. They also offer several bundle options, including a 1.54-inch (240×240) screen for $20, with additional USB 3.0 Type-A and Type-C cables available for a few extra dollars.

The post MuseLab nanoCH32H417 – A $17 WCH CH32H417 RISC-V MCU development board with USB 3.0, Fast Ethernet appeared first on CNX Software - Embedded Systems News.

Dutch hardware designer LichtBit has launched a fully open-source ESP32-based Art-Net/sACN NeoPixels LED strip controller designed for large-scale lighting installations and custom stage design. Built around an ESP32, the hardware routes lighting data over wired Ethernet or Wi-Fi to manage up to 16 universes of addressable LEDs across 4 dedicated outputs.

We have previously written about various NeoPixel LED controllers, such as the xcrhom WLED Type-C, the Adafruit Sparkle Motion Stick, the full-featured Adafruit Sparkle Motion, and others, which are designed for portable lighting setups and standalone animations. But LichtBit’s board targets the professional event industry by bridging the gap between commercial live-performance software (such as Resolume Arena, MadMapper, or xLights) and low-cost digital LED strips.

LichtBit’s Art-Net and sACN LED controller specifications:

• Development Board – Generic ESP32 based developmt board with ESP32-WROOM-32 MCU
• Display – 128×32 I2C OLED display for real-time device configurations, status, and IP mapping
• Network
  • 10/100Mbps Ethernet via WIZnet W5500 module (highly recommended for high frame rates)
  • 2.4 GHz Wi-Fi (802.11 b/g/n)
  • Auto-launched AP mode after 30 seconds of network timeout for standalone debugging/configuration
• Output – 4 physical screw-terminal (4 universes/up to 680x RGB LEDs per output pin, or 2720 in total)
• LED Compatibility
  • 3-channel (RGB), 4-channel (RGBW), and 5-channel (RGBAW)
  • 3-wire synchronous data-only strips like  WS2811, WS2812(B), WS2813, WS2814, WS2815, SK6812
  • 4-wire SPI clock-based strips like APA102
• Misc
  • Onboard addressable WS2812 RGB status indicator LED
  • Physical boot/mode button for configurations (DHCP/Static IP, standalone RGB test cycles, and 10 built-in color presets)
  • Art-Net, sACN (E1.31), and WLED sync support
• Power
  • Wide 5V to 24V DC input voltage range
  • Individual ~8A automotive resettable fuse protection per LED output channel
  • 5V level-shifters for clean 5V data signals across long cable runs
  • 5V power switch to bypass the internal 12V-to-5V step-down regulator for native 5V setups
  • Reverse polarity protection circuits
• Dimensions – TBD

The hardware relies on the I2SClocklessLedDriver library to handle pixel rendering, freeing the ESP32’s processor cores to process network packets from complex software controllers such as Resolume Arena, MadMapper, or xLights. While the main firmware works as an Art-Net/sACN node, the LichtBit board also supports the WLED framework, so if you want standalone effects, you can flash the standard Wi-Fi version of WLED.

Note: For those who want Ethernet capability combined with WLED, an unmerged experimental fork of WLED, WLED-W500, must be used, as the stock WLED branch does not natively support WIZnet W5500 chips via SPI without manual compilation edits.

The developer also provides a custom wled_cfg.json file on the project’s GitHub repo that automatically remaps WLED’s internal configuration to map pins 12, 14, 27, and 26 safely to the onboard level shifters and fuses. The repository also contains 3D design files for an enclosure, binaries, and additional information about the project, while the schematics, PCB layout, BoM, and Gerber files can be found on the EasyEDA OSHWLab project page.

The W5500 Ethernet chip does not support Auto-MDI/MDIX, so there is a small limitation. If you connect the controller directly to a PC, you need to use a crossover Ethernet cable. An easier option is to connect both the PC and the controller through a network switch or router, which works without any special cable.

Traditional stage lighting fixtures use standard DMX512 cables (3-pin or 5-pin XLR). However, a single DMX cable only handles 512 channels (one “Universe”). If you have an addressable LED strip where each LED requires 3 channels (Red, Green, Blue), a single 512-channel DMX universe can drive only about 170 LEDs before running out of bandwidth. To handle large data flows, multiple DMX universes are sent over standard Ethernet networks rather than individual DMX cables. This is done using protocols such as Art-Net and sACN (E1.31), which enable the efficient transmission of thousands of universes from lighting software to controllers.

This board is designed to sit between professional lighting software and raw addressable LED strips, where the software maps a video or complex light show across multiple universes and streams that data over an Ethernet cable. This controller reads those network packets, unpacks the Art-Net/sACN data, and instantly converts it into the high-speed serial data protocols that NeoPixels/WS2812B strips understand.

The fully completed, pre-assembled node is currently available for purchase on Tindie for $82.00 (excluding shipping).

The post LichtBit’s open-source ESP32 Art-Net/sACN NeoPixels controller can drive up to 2,720 RGB LEDs appeared first on CNX Software - Embedded Systems News.

Hardkernel has sent me a kit with the ODROID-H5 10GbE SBC for review. In addition to the Intel Core i3-300 board itself, the kit also comes with an ODROID-H5 Type-1 case, an M.2 card for a second 10 Gbps Ethernet port, and other accessories.

I’ll start the review with an unboxing, my experience assembling the kit, and first boot using an M.2 NVMe SSD with Ubuntu 24.04 and Windows 11 in a dual-boot configuration. In a few weeks, I’ll continue the review after upgrading to Ubuntu 26.04 and perform thorough testing of all features and performance.

ODROID-H5 kit unboxing

An advantage of the ODROID family is that it’s highly flexible, and you can build your own customized system. For this review, Hardkernel sent me a package with the following items:

• External LED power button
• H5 Case Type 1 enclosure
• ODROID-H5
• 16GB DDR5 SO-DIMM (Crucial)
• 15V/4A PSU with US plug adapter
• M.2 10GbE card for dual-port 10GbE implementation

Let’s have a quick look at the board itself:

It’s fanless and cooled by a large and thick heatsink, and most ports can be found on the rear panel: 12V DC jack, two USB 2.0 ports, a 10GbE RJ45, two USB 3.0 ports, one HDMI port, and two DisplayPort connectors.

A look at the top shows one of the M.2 PCIe Gen3 x2 sockets (SSD2), an eMMC module slot, Power and Reset buttons, an RTC battery connector, a 4-pin 12V fan connector, and a 24-pin GPIO header.

The bottom side features a single SO-DIMM connector and three more M.2 sockets. Note that SSD1 and SSD3 are PCIe Gen3 x2 sockets, but SSD4 is only connected to one PCIe Gen3 lane, so that’s something to keep in mind when connecting modules.

The H5 Case Type1 is made of six PCB panels (FR4), and comes with plastic standoffs, rubber feet, and screws.

The 10GbE M.2 card appears to be a standard one for installation in computer cases with RJ45 ports, short and long cables, short and long brackets, standoffs, and screws.

I removed the thin heatsink to check whether it was an RTL8127 module, and indeed it’s based on RTL8127AT. I’m glad this new low-power 10Gbps Ethernet chip exists, as it will help further “democratize” 10GbE networking at home.

The 15V/4A power adapter features a 5.5/2.1mm DC jack and takes 100-240 AV input, so it’s supported worldwide. This type of adapter often comes with three to four plug adapters (US, UK, EU, AU), but to cut costs, Hardkernel only provides one.

ODROID-H5 Case Type1 assembly

Hardkernel didn’t send any storage devices. I have recently upgraded my laptop from a 512GB NVMe SSD to a 2TB NVMe SSD since free storage was tight, and I decided to use the now spare 512GB module, especially since it already comes with Windows 11 and Ubuntu 24.04.

Let’s start the assembly by inserting the 16GB DDR5 module and the NVMe SSD into the SSD1 socket (PCIe Gen3 x2).

I did the same for the 10GbE M.2 module and also connected the RJ45 board through the short cable provided with the kit.

The next step was to add standoffs and rubber feet using the Type_2_D panel as shown in the photo below.

I then attached the RJ45 to the Type_A panel with two standoffs and screws.

I followed the instructions for the ODROID-H4 to connect the Power LED button, since it’s not available yet for the ODROID-H5. Basically, I connected the black wire (GND) to pin 1, the red wire (5V) to pin 4, the Green wire to pin 17, and the blue wire to pin 19.

I almost forgot to add the RTC battery…

… before securing the case with four more screws. The result looks pretty neat.

I now have a dual 10GbE mini PC that could be used for a range of networking applications. The HDMI/DP connectors also enable up to three independent 4K displays, and a few USB ports allow for expansion. I’d just wish there was a full-function USB Type-C port with DisplayPort Alt. mode and USB PD, as I found these to be really convenient now that I have a few USB-C displays.

First boot to Ubuntu 24.04 and Windows 11

Time to quickly test the board. I added an HDMI display (14-inch Crowview portable monitor), a USB RF dongle for a keyboard and mouse combo, and connected the power. The ODROID-H5 quickly booted into Ubuntu 24.04, and I could confirm the Intel Core i3-N300 octa-core CPU, 16GB RAM, and 512GB NVMe SSD in the System->About window.

I’ll have to update that to Ubuntu 26.04, but in the meantime, I rebooted and selected Windows 11 in Grub.

It started fine, but since Windows must have detected hardware changes (from my laptop to the ODROID-H5), it asked me to reset my PIN, and I can’t do that without Internet… I can only connect over WiFi when outdoors, so I’ll skip it and will probably not review Windows on the ODROID-H5 anyway. Instead, I’ll focus on Ubuntu 26.04 in the second part of the review, and test 10GbE networking, the three display outputs, USB ports, and overall performance.

I’d like to thank Hardkernel for sending an ODROID-H5 SBC and accessories for review. Here’s the price breakdown of the kit I received:

• ODROID-H5 SBC – $250
• H5 Case Type 1 – $11
• LED Power Button – $5.99
• 15V/4A PSU (US plug) – $11
• M.2 10GbE card – $76
• 16GB DDR5 SO-DIMM – $220

The total is $573.99 before shipping and taxes. I should be used to RAM pricing by now, but I’m still shocked by it. The first three items must be purchased from Hardkernel or distributors,  but you can shop around for the 10GbE M.2 card and 16GB RAM. The ODROID-H5 price appears to depend on the destination country. While I’m shown $250 to Thailand, if I use a “slow” US VPN, $230 first appears, and then it switches to $260. I haven’t found out the reason behind this yet, I was just curious since some US blogs list the price as $260.

The full price doesn’t include the NVMe SSD, so the total should be quite over $600. A close competitor is the iKOOLCORE R2 Max with 2x 10GbE, 2x 2.5GbE,  Core i3-N305 SoC, 8GB RAM, and a 128GB SSD that I reviewed in 2024 and sold for $449 at the time. However, since the price of RAM and storage has increased, they don’t bother anymore and only sell the barebone model for $469, which means the ODROID-H5 pricing still looks competitive. The ODROID-H5 should also consume much less when using 10GbE since it features low-power RTL8127 chips instead of legacy “Marvell AQC113C-B1-C” network cards. But that’s something I’ll test in the second part of the review….

The post ODROID-H5 SBC Review – Part 1: Unboxing, Type1 case assembly, and first boot appeared first on CNX Software - Embedded Systems News.

PolyCast5 is a portable, hackable ESP32-C5-based multi-tool remote to control devices through five different core wireless technologies: WiFi 6, Bluetooth LE, ESP-NOW, LoRa, and infrared Tx/Rx.

The all-in-one controller can be used for cybersecurity work, a standard IR learning remote control, a voice-enabled password manager, a robotic arm controller, an AI keyboard using the built-in microphone and Bluetooth connectivity, a long-range LoRa remote control, DIY electronics projects through a 4-pin GPIO header, and more.

PolyCast5 specifications:

• Wireless module – ESP32-C5-WROOM-1
  • SoC – ESP32-C5
    • CPU
      • Single-core 32-bit RISC-V processor @ up to 240 MHz
      • Low-power RISC-V core @ 40 MHz acting as the main processor for power-sensitive applications
    • Memory – 384 KB SRAM on-chip, 8MB PSRAM
    • Storage – 320 KB ROM
    • Wireless Connectivity
      • Dual-band (2.4GHz/5 GHz) 802.11ax WiFi 6, with 802.11b/g/n WiFi 4 standard fallback
      • Bluetooth 5.0 Low Energy (LE)
      • 802.15.4 radio for Zigbee 3.0 and Thread 1.3 (note: support not advertised for the Polycast5)
  • Storage – 8MB or 16MB SPI flash (unclear)
  • Antenna – PCB antenna
• Display – Small screen with two colors (adjustable)
• Audio – Built-in microphone
• Wireless
  • Wi-Fi 6 (ESP32-C5) – Dual-band 2.4/5 GHz 802.11ax for OTA updates, network scanning, and AI packet analysis.
  • Bluetooth 5.0 / BLE (ESP32-C5) – The Polycact5 acts as a Bluetooth keyboard controller, including HID payloads, password manager, and AI keyboard mode using the built-in mic (e.g., voice command to open Chrome on the Bluetooth host). Multi-pairing support.
  • ESP-NOW (ESP32-C5) – Low-power, Arduino/ESP32-compatible protocol for custom DIY devices such as door openers, robotic arms, lights, etc…
  • LoRa (SX1262) – 868/915 MHz for long-range (1+ km) control of IoT devices, including the company’s PolyPlug switch with 15A relay. Supports schedules, away mode, and MQTT.
  • Infrared (IR) – 38 kHz universal support with learning (Rx) and replay (Tx) capabilities.
• USB – USB Type-C port for power and programming
• Expansion – 4-pin I2C header for expansion
• Misc
  • 7x buttons
  • RGB LED
  • Vibration motor for haptic feedback
• Power Supply
  • 5V via USB-C port
  • 1400 mAh rechargeable battery good for up to about 7 days in sleep mode or around 19 hours in continuous use.
• Dimensions – 62 x 45.6 mm

There are also offline “tools”, such as a Tetris game, a dice roller, a pomodoro timer, a Bitcoin QR app, a coin flipper, a random number generator, and more. The firmware will also offer customizable homescreen animations, settings, and add-ons.

The firmware will be open-sourced on RoboticWorx’s GitHub account after the crowdfunding campaign is over. We are just shown a screenshot of the “Polycast5” repo, which must be set to private. In the meantime, you can find short video demos for the various features of the wireless multi-tool on the project’s website.

The LoRa radio appears to be mostly used for Smart Home applications, notably to send an on/off command to the PolyPlug switch with a built-in 15A relay. The firmware implements schedules, an away mode that turns on/off the attached device randomly to simulate presence, and MQTT support for control outside of LoRa range. We previously saw the MOKO LW005-MP LoRaWAN plug with similar features (plus power meter function), and the PolyPlug adds another option.

RoboticWorx (Justin Atkins) has just launched the PolyCast5 ESP32-C5 multi-tool remote on Kickstarter with a $5,000 funding target. Rewards start at $129 for an Early Bird pledge, plus $15 to $45 shipping depending on the destination country. Deliveries are scheduled to start by September 2026.

The post PolyCast5 – An ESP32-C5 multi-tool remote with dual-band WiFi 6, BLE, ESP-NOW, LoRa, and Infrared Tx/Rx (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Last November, we wrote about the WCH BLE Analyzer Pro, an inexpensive (~$20) USB Bluetooth LE sniffer and analyzer, which looked useful and good value for reverse engineering and debugging.

One downside is that the WCH BLE Analyzer software was only made for Windows 7 to 11, but Xecaz decided to look into it and reverse-engineered the USB protocol to write Linux software using libusb that outputs a standard pcap compatible with popular tools such as Wireshark, or as he puts it: “WinChipHead forgot to ship a Linux driver. We forgot to ask permission.”

As a reminder, the BLE Analyzer Pro features three CH582F Bluetooth LE RISC-V microcontrollers and a CH334 USB hub, supports Bluetooth Low Energy (BLE) 4.0/4.2/5.0, and connects to the host through its USB-C port.

The Linux “driver” for the BLE Analyzer Pro tool can be found on GitHub. Building and installing the software is quick and straightforward:

git clone https://github.com/xecaz/BLE-Analyzer-pro-linux-capture
cd BLE-Analyzer-pro-linux-capture/
sudo apt install libusb-1.0-0-dev 
make
sudo make install
sudo udevadm control --reload-rules && sudo udevadm trigger

You can check the option by running the program without parameters:

jaufranc@CNX-LAPTOP-5:~/edev/sandbox/BLE-Analyzer-pro-linux-capture$ ./wch_capture 
Nothing to do – use -v and/or -w FILE.pcap
WCH BLE Analyzer PRO Linux Capture tool by Xecaz 2026!

Usage: ./wch_capture [OPTIONS]

Options:
  -v            Print packets to stdout
  -w FILE.pcap  Write PCAP (DLT 256, BLE LL + phdr)
  -p PHY        PHY: 1=1M (default), 2=2M, 3=CodedS8, 4=CodedS2
  -i ADDR       Initiator MAC filter  (AA:BB:CC:DD:EE:FF)
  -a ADDR       Advertiser MAC filter (AA:BB:CC:DD:EE:FF)
  -k KEY        LTK, 32 hex chars
  -K PASSKEY    BLE passkey (6-digit decimal)
  -2            Custom 2.4G mode (default: BLE monitor)
  -c CHAN       Channel 0-39: BLE adv 37/38/39 or 0=all (auto per MCU); 2.4G raw
  -A AADDR      2.4G access addr (hex, e.g. 8E89BED6)
  -C CRCINIT    2.4G CRC init (6 hex chars, e.g. 555555)
  -W WHITEN     2.4G whitening init (hex byte)
  -h            Show this help

Capture stops on SIGINT (Ctrl+C) or SIGTERM.

I don’t have a unit, but it can be used as follows:

sudo ./wch_capture -v -w capture.pcap

Then you can open the capture.cap file in Wireshark for further analysis. All three MCUs capture simultaneously on channels 37/38/39, and Wireshark decodes ADV_IND, ADV_NONCONN_IND, ADV_SCAN_IND, SCAN_REQ, SCAN_RSP, CONNECT_IND, OUI lookups, etc. If you want to learn more about the reverse-engineering process, I recommend reading the “RE_PROCESS.md” file on GitHub. An Android app is also being worked on, and Xecaz welcomes testers before releasing the app publicly.

The post WCH BLE Analyzer Pro USB Bluetooth LE sniffer gains Linux software with Wireshark (pcap) support appeared first on CNX Software - Embedded Systems News.

ESPHome 2026.5.0 has just been released with the beta version of the new ESPHome Device Builder web app that replaces the legacy in-tree dashboard with a real configuration editor, a firmware job queue, multi-select bulk actions, labels and areas, out-of-sync detection, cross-config search, distributed builds, and a proper settings UI.

The firmware itself gains optimizations of the main loop, scheduler, and task watchdog to lower CPU and power usage on supported platforms, and a range of other memory/performance optimizations across the API, audio, and helper hot paths. The audio decoder pipeline has been improved and features new microMP3, microWAV, and microFLAC streaming libraries. OTA has also been enhanced with partition-table and bootloader updates, web-server OTA, and soft-brick recovery, and ESP32 MCUs are now handled by up to the ESP-IDF v6.0.1 framework natively, while Zigbee support has been expanded to ESP32 H2 and ESP32-C6, among other features.

Key features of the new ESPHome Device Builder missing in the older/legacy version:

• Visual component and automation builder alongside Monaco YAML, with a left-sidebar device navigator.
• Component catalog with dependency resolution and a per-board pin info viewer that maps GPIO capabilities and shows which component is using each pin.
• Firmware job queue with progress, history, and cancel for compile/install/clean.
• Remote builder. One Device Builder instance can offload compile/install jobs to another over a peer-paired link (mDNS discovery, SHA-256 fingerprint confirmation, identity rotation, per-peer auto-route).
• Labels (colored, searchable, filterable), areas as a first-class field, friendly name as a separate editable field, device cloning, and multi-select bulk actions (update/delete/archive on an arbitrary subset of devices).
• Out-of-sync detection with per-device badges for version, config-hash, and encryption-state mismatches.
• YAML diff view, cross-config YAML search with surrounding context, and a command palette (⌘K / Ctrl-K).
• Card and table views with configurable columns and faceted filters (platform/status/area/labels).
• Settings UI with light/dark/system theme, English, French, and Dutch localization, editor layout, and remote-builder controls. The legacy dashboard exposed almost nothing in-UI.
• First-run Wi-Fi onboarding and USB-plug detection with a “set this up” prompt when a board is connected.

The related code is available in two repositories: device-builder Python backend and device-builder-frontend web UI. Note that the legacy dashboard remains the default in 2026.5.0, and Home Assistant users can try the new Device Builder today by installing the ESPHome (beta) app, where it is enabled by default.

A quick summary of other changes includes:

• Main loop and watchdog architecture overhaul for power savings
• A range of micro-optimizations for performance, notably for audio, BLE, API calls…
• Memory footprint reductions across a range of common components
• Native ESP-IDF toolchain support, alongside the existing PlatformIO build path
• Audio stack modernization – New micro decoders, new HTTP media source, advanced codec configuration…
• OTA platform enhancements
• Faster configuration validation and CLI startup
• Sendspin synchronized multi-room audio
• Radio frequency entity type for representing RF transceivers in Home Assistant
• Various LVGL improvements
• Zigbee expanded to ESP32-H2 and ESP32-C6
• nRF52 and Zephyr platform improvements for deep sleep, Zigbee, OTA, native builds, and more
• New hardware and display support
  • ESP32-P4 USB High-Speed interface gains support for 512-byte USB transfers (was 64-byte, Full Speed)
  • Configurable ESP32 watchdog timeout
  • esp32_ble PSRAM allocation to free about 40 kB of internal RAM on PSRAM-equipped ESP32 boards.
  • I2S-based SPDIF speaker output sends digital audio to optical receivers via any GPIO pin.
  • New modbus_server split with flash savings of roughly 60% over the old wedged-in server mode (1.8 KB vs 4.5 KB) and 40% off the client-mode modbus_controller (3.9 KB vs 6.4 KB).
  • New display variants: epaper SSD1683 + Goodisplay GDEY042T81, Waveshare 3.97” e-paper, Waveshare ESP32-C6 LCD 1.47, Sunton ESP32-2424S012, Sunton 5”/7” mipi_rgb displays, Seeed reTerminal D1001 DSI display
  • WiFi phy_mode for ESP8266 to pin the radio to 11B, 11G, or 11N from YAML.
• BLE Reliability Fix for Bluetooth Proxies

And the list goes on. If you want to check out the full list or want to find out more details about one of the changes above, read the long changelog on the ESPHome website.

Thanks to Hedda for the tip.

The post ESPHome 2026.5.0 released with new ESPHome Device Builder (beta), performance/memory optimizations appeared first on CNX Software - Embedded Systems News.

RAKwireless WisMesh Pi HAT RAK6421 is a modular Meshtastic gateway expansion board for Raspberry Pi 4/5 that adds support for the company’s WisBlock ecosystem. Designed for users running meshtasticd (the Linux-native Meshtastic service), it enables scalable, always-on Meshtastic base stations, MQTT gateways, and backbone relay nodes.

Unlike other LoRa solutions based on Raspberry Pi, the RAK6421 takes a modular approach. Instead of a built-in radio, it offers two WisBlock IO slots and four sensor slots, allowing users to easily plug in or swap modules. This means you can choose between standard LoRa modules like the RAK13300 or higher-power 1W versions like the RAK13302, without dealing with messy USB adapters or breadboard wiring.

RAK6421 WisMesh Pi HAT RAK6421 specifications:

• Compatibility
  • Raspberry Pi 4, Raspberry Pi 5, and Raspberry Pi Zero 2W SBCs
  • Potentially other SBC with standard 40-pin Raspberry Pi GPIO header (HAT+ compliant)
• Expansion
  • 2x WisBlock IO slots for LoRa radio modules and other IO modules
    • SPI, UART, I2C, and GPIO signals
    • 4x analog inputs (0…3.3V) via SGM58031XMS10G/TR (ADS1115 compatible) 16-bit, low-power, precision sigma-delta (ΔΣ) analog-to-digital converter
  • 4x WisBlock Sensor slots (Slot A, B, C, D) for I2C modules like GNSS (Quectel L76K), environmental sensors (BME680, SHTC3), and RTCs
  • 4-pin J8 header exposing AIN1, GPIO4, GPIO5, and VBAT
  • 4-pin J9 header exposing 3.3V, GND, I2C1_SCL, and I2C1_SDA
•  Misc –  Onboard CAT24C32-compatible EEPROM connected to ID_SD/ID_SC for hardware auto-discovery; read-only by default (read-write mode accessible by shorting J10 pins)
• Power Supply
  • 5V input (4.9V to 5.1V) from the Raspberry Pi 40-pin header
  • Downstream power delivery of 3.3V and 4.17V (VBAT) to connected WisBlock modules
• Dimensions – Standard Raspberry Pi HAT form factor
• Temperature Range – -40°C to +85°C (operating)

ESP32-based nodes like the LR2021 LoRa Plus, the T-Display S3 Pro LR1121, the LILYGO T-LoRa Pager, and other similar nodes work well as portable, battery-powered devices. However, fixed gateways that need to handle heavy MQTT traffic, Node-RED, or Grafana dashboards quickly hit performance limits on a microcontroller. The WisMesh Pi HAT RAK6421 solves this by letting you run Meshtastic on a Raspberry Pi. It provides stable SPI, I²C, and UART connections, making it much more reliable for long-term telemetry and gateway use.

On the software side, RAKwireless provides a ready-to-use Raspberry Pi image with meshtasticd already installed and configured, which can be downloaded from their quick start guide. They also offer open-source scripts that automatically install Mosquitto MQTT, InfluxDB, Node-RED, and a pre-configured Grafana dashboard. You can check the product overview page for more details. You can also manually install Meshtastic on standard Raspberry Pi OS or build a custom image. More information on that can be found in the datasheet. The RAK6421 is also officially supported by Meshtastic.

The HAT natively supports a wide array of existing WisBlock modules, including the RAK1906 (Bosch BME680 environment sensor), RAK1901 (Sensirion SHTC3 temp/humidity), RAK12019 (UV sensor), and RAK12002 (RTC module).

The WisMesh Pi RAK6421 HAT is available on AliExpress starting at $12.50 for the standalone board, with bundled options (including sensors like RS485, environmental, and CO₂ modules) priced up to $95.90. Alternatively, you can also find bundles with a Raspberry Pi 4 2GB or Raspberry Pi 5 4GB and a microSD card preloaded with the software for $135.99 and $205.99, respectively. Finally, you’ll also find all these on the RAKwireless store. You’ll also need to add a RAK13300 or RAK13302 (1W) LoRa module if you don’t already own one.

The post RAKwireless WisMesh Pi HAT RAK6421 turns your Raspberry Pi 4/5 into a modular Meshtastic gateway appeared first on CNX Software - Embedded Systems News.

CardputerZero is a pocket-sized computer based on the Raspberry Pi Compute Module Zero (CM0) and designed for makers with a 46-key matrix keyboard, a 1.9-inch LCD, HDMI video output, Fast Ethernet, three USB ports, a microphone and a speaker for voice interaction, a 14-pin GPIO header, a Grove interface, and an IR transceiver (Rx/Tx).

The credit card-sized device comes in two models: CardputerZero Lite and CardputerZero. The latter also adds an 8MP camera, a 6-axis IMU, and a 32GB microSD card preloaded with the software. Both models are powered by a rechargeable 1,500 mAh LiPo battery or directly via a USB-C port.

Cardputer Zero specifications:

• SoM – Raspberry Pi CM0 Lite
  • SoC – Broadcom BCM2710A1
    • CPU – Quad-core Cortex-A53 processor @ 1.0 GHz
    • GPU – VideoCore IV GPU supporting OpenGL ES 1.1, 2.0 graphics
    • VPU – H.264/MPEG-4 1080p30 video decoding, H.264 1080p30 video encoding
  • System Memory – 512MB LPDDR2 RAM
  • Storage – 0GB (Raspberry Pi CM0 Lite)
  • Wireless – 802.11 b/g/n WiFi 4, Bluetooth 4.2 LE with IPEX antenna connector
• Storage – MicroSD card slot
• Display – 1.9-inch LCD (ST7789v3) with 320×170 resolution
• Video output – HDMI port up to 1080p30
• Camera – 8MP Sony IMX219 (3280 × 2464) sensor (4-lane CSI interface)
• Audio
  • ES8389 audio codec
  • MEMS microphone
  • 1W speaker
  • 3.5 mm audio output jack
• Networking
  • 10/100Mbps Ethernet RJ45 port
  • 2.4 GHz Wi-Fi 802.11 b/g/n and Bluetooth 4.2 / BLE on Raspberry Pi CM0 Lite module
• USB
  • 2x USB Type-C host/device ports
  • 1x USB Type-A host/device port
• User Input – 46-key matrix keyboard
• Sensor – BMI270 IMU (gyroscope + accelerometer)
• Expansion
  • 4-pin I2C/UART grove connector
  • 14-pin 2.54mm pitch header with SPI, UART, I2C, USB, GPIO, 5V, GND
• Misc
  • Power switch
  • Boot button
  • RX8130CE RTC
  • Infrared transmitter (Tx) and receiver (Rx)
• Power Supply
  • 3.7 V / 1500 mAh Li-Po battery
  • BQ27220 fuel gauge
• Dimensions – 85 x 54 x ?? mm

M5Stack has yet to say which exact OS runs on the CardputerZero, but I’d assume a customized version of Raspberry Pi OS, since we’re promised a “Full Linux Lab in Your Pocket” to “connect to servers via SSH, run Python code, edit files with CLI tools like Git or Vim”. I can see two related repositories on GitHub. CardputerZero-AppBuilder repo features a utility to develop LVGL apps suitable for the 320×170 display directly on your Linux or macOS machine (and maybe Windows WSL 2), while the CardputerZeroRepository contains the Debian (.deb) repository for pre-built applications.

The company also explains it can run OpenClaw-like tools (possibly PicoClaw), and be used as a retro-gaming console, a portable media player, and so on. The CardputerZero is also extensible thanks to Grove, M5Units, or even custom hardware add-ons for cybersecurity tools (CC1101, NFC…), off-grid communication (LoRa/Meshtastic), and various DIY electronics projects.

In some ways, the M5Stack CardputerZero can be seen as a low-cost alternative to the upcoming Flipper One portable Linux computer, which offers more processing power, more memory and storage, and better connectivity options (dual GbE, WiFi 6, optional 4G/5G cellular), but is also likely to cost $300+.

Price-wise, the CardputerZero will be more affordable, starting at $59 for the Lite version and $89 for the full version for backers reserving a Super Early Bird on Kickstarter. The crowdfunding campaign will start very soon on May 26 at 9 am EDT. Over 5,000 have also subscribed to get notified of the launch of the project, and you can reserve your unit at the aforementioned price by depositing $1 through the M5stack website. Regular pricing will eventually be $99 (Lite) and $149, although I’d expect something in between while the Kickstarter campaign is still running. Talking Sasquach got a early prototype, and you can check out what it can do so far.

The post CardputerZero – A Raspberry Pi CM0 pocket computer for makers (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Earlier this month, I received the Khadas Mind 2 (Intel Core Ultra 7 155H) mini PC with two accessories connecting through the Mind Link (PCIe x8) connector: The Mind xPlay Display and Keyboard combo, and the Mind Graphics 2 dock, adding a range of interfaces and an NVIDIA GeForce RTX 5060 Ti GPU with 16GB of VRAM.

After going through the specifications, an unboxing, and a partial teardown in the first part of the review, I tested the Mind xPlay using the Mind and Mind 2 mini PCs running Windows 11 and Ubuntu 24.04. I’ve now had time to test the Khadas Mind Graphics 2 dock with the Mind 2 mini PC running Windows 11, so I’ll report my experience with the NVIDIA GPU (3D graphics and AI), and test all features, including 2.5GbE networking, the built-in speakers, microphone array, and so on.

A few gremlins and system info

When I first connected the Mind Graphics 2 dock to the Mind 2 mini PC, the NVIDIA GeForce RTX 5060 Ti graphics card was detected, but was disabled with a message reading “the device has been stopped because it has reported problems”. HDMI output from the dock didn’t work at all. To fix that, I connected the Mind 2 to the display through the USB-C DP port, and I went to the Mind app on the mini PC, upgraded the firmware of the dock to version 1.9, and more importantly, downloaded and installed  khadas-nv-43-fix.zip from https://dl.khadas.com/products/mind-graphics-2/drivers/.

That’s what the Device Manager window looks like without (left) and with (right) the Mind Graphics 2 dock.

The microphone and speakers from the “Khadas Mind Graphics Speaker” show up, as well as the NVIDIA GeForce RTX 5060 Ti, the 2.5Gbps Ethernet port, and a few extra USB controllers, including a Realtek USB 3.0 card reader.

GPU benchmarks using Khadas Mind 2 without and with the Mind Graphics 2 dock

Since the NVIDIA GeForce RTX 5060 Ti is the star of the show, I’ll get right into it with GPU benchmarks using the Mind 2 mini PC in standalone mode. mostly testing the 2.25 GHz 8x Xe core Intel Arc graphics inside the Intel Core Ultra 7 155H SoC, and with the dock leveraging the NVIDIA graphics card.

The following benchmarks will be used to evaluate both 3D graphics and AI performance:

• 3DMark – Fire Strike and Nomad Steel (higher-end test)
• Unigine – Heaven and Superposition (newer test, still 2017)
• Final Fantasy XV benchmarks to evaluate the performance of an actual (2024) game
• GeekBench AI 1.7.0
• MLPerf Client 1.6.1

I’ll first list the raw results with screenshots, before adding a comparison table to show both configurations side-by-side. Note that all tests were done in a room with an ambient temperature of 30-31°C.

Here are the results for the Mind 2 in standalone mode.

7,444 points in the 3DMark Fire Strike benchmark, with the software showing “Good” performance.

651 points in the more demanding Steel Nomad benchmark.

Unigine Heaven 4.0 was run at the standard 1920×1080 resolution, and the system could render the scene at 71.8 FPS on average, and a score of 1,809 points.

In the more recent Unigine Superposition benchmarks, the system achieved 3,817 points.

Final Fantasy XV benchmark was run at Standard Quality (1920×1080) resolution. The score was 3,487 points, and the Performance marked as “Standard”.

Time to switch to AI benchmarks, starting with Geekbench 1.7.0 using the OpenVino framework with GPU AI backed, and Intel Arc Graphics AI device.

The scores were 8,087 points for single precision, 12,987 points for half-precision, and 19,698 points for quantized. I used the 18 TOPS GPU, since the NPU (Intel AI Boost) was not an option.

The final test was MLPerf Client 1.6.1.

The NPU test all failed, but we still got the OpenVino GPU test for Llama 3.1 8B Instruct and Phi 3.5 Mini Instruct, which I’ll use for comparison.

Let’s now insert the Mind 2 on top of the Mind Graphics 2 dock and run all these tests again.

30,659 points on Fire Strike with “Excellent” performance.

3,487 points on Steel Nomad with “Good” performance.

The Unigine Heaven Benchmark 4.0 is a bit old for a powerful NVIDIA GPU, and the system delivered 365.8 FPS on average with a score of 9,214 points.

The kit achieved 24,895 points in the Unigine Superposition benchmark.

In the Final Fantasy XV benchmark, the score was 15,542 points, and the performance was rated as “Extremely High” in Standard Quality.

I also maxed out the benchmark with a custom configuration enabling all options to high/highest to stress the system a bit more. The score was 6,028 points, and the performance was considered “High”.

Geekbench AI was configured with the ONNX framework, DirectML backend, and NVIDIA GeForce RTX 5060 Ti device.

The scores were 20,587 points for single precision, 36,368 points for half-precision, and 16,329 points (lower than on the iGPU) for quantized.

All six Base tests from MLPerf Client could be completed. I’ll go with the CUDA results for comparison.

Time for the comparison table. A ratio of 1x means the performance is identical, over 1x the NVIDIA GPU is faster, and under 1x the Intel Arc iGPU is faster.

For 3D graphics, the Mind Graphics 2 delivers 4x to 6x+ higher performance compared to the Intel Core Ultra 7 155H SoC with Intel Arc graphics.  AI models like Llama 3.1 8B Instruct and Phi 3.1 Mini Instruct, tested with MLPerf, are also much faster thanks to the NVIDIA GPU, which yielded about 5x higher performance in terms of tokens per second. GeekBench AI is an outlier with around 2.5x to 3x higher performance for single and half precision, but somehow the quantized score (INT8) is lower than running on the Intel GPU with OpenVino.  It looks like the culprit may be DirectML’s NVIDIA INT8 support rather than anything related to the hardware.

The good news is that I have no stability problem running tests with the Mind 2 connected to the Mind Graphics 2 dock, once the initial driver issue was resolved.

Khadas Graphics 2 features – 2.5GbE, Audio, Video output….

Test all features

• HDMI 1 – 4K 60Hz video OK, audio OK; note I had to manually enable the NVIDIA High Definition Audio driver in Device Manager for this to work (only the built-in speakers would show otherwise)
• HDMI 2 – 4K 60Hz video OK, audio OK
• DisplayPort – 4K 60Hz video OK, audio OK
• Quad display – OK. 2x 4K monitors connected to HDMI 2.1 and DP + a Full HD transflective LCD connected to HDMI 2.1 on the Khadas Mind Graphics 2, and one Full HD portable monitor to the USB-C DP port on the Mind 2 mini PC
• Microphone array – OK
• Speakers – OK (stereo), very good quality, still works when the display(s) is/are turned off
• 3.5mm audio jack – OK, tested with USB-powered speakers
• Card reader – OK. Tested with a Raspberry Pi microSD card (via full SD card adapter)
• 2.5GbE port
  • OK for unidirectional tests. iperf 3.19: 2.35 Gbps Rx; 2.35 Gbps Tx
  • Failed for full-duplex/bidirectional. iperf 3.19: 2.36 Gbps Tx/196 Mbps Rx (potentially not ideal for a BitTorrent node…)
• USB
  • Front USB-C (10Gbps) port – OK, tested with HWiNFO 64 and CrystalDiskMark (1039 MB/s sequential read)
  • Rear USB-C (40 Gbps) port – Neither an ORICO NVMe enclosure nor a Beelink Expand M USB-C dock is recognized. This may be a limitation of the Mind Graphics 2 dock when connected through the Mind Link. The USB-C port doesn’t support 40 Gbps/Thunderbolt, and should fallback to 10Gbbps USB, but it doesn’t seem to work in that case. This port should be most useful when connected to a laptop or other mini PC as a dock/eGPU.
  • USB 3.2 Gen2 (10 Gbps) port #1 (left) – OK, tested with HWiNFO 64 and CrystalDiskMark (1039 MB/s)
  • USB 3.2 Gen2 (10 Gbps) port #2 (right) – OK, tested with HWiNFO 64 and CrystalDiskMark (1038 MB/s)
• Mind unlock button – OK. This button is required to remove the Mind 2 from the dock. A short press should bring a pop-up window, but once it didn’t work, and I had to press it for 8 seconds to force an unlock (video output is turned off when triggered).
  
• Volume buttons – OK
• Fingerprint reader – OK (automatic sign-in). The Mute function also works fine.

Conclusion

If you can afford it, I find the Khadas Mind ecosystem to be pretty neat. The Khadas Mind 2 mini PC itself is already powerful, but you can convert it into a laptop with the Mind xPlay (13-inch display and keyboard) or a full gaming platform or AI workstation with the Mind Graphics 2 eGPU dock with a (non-replaceable) NVIDIA GeForce RTX 5060 Ti 16GB graphics card.

I had some frustrating times, notably with the BIOS update and manually installing the patch for the NVIDIA drivers, but once I got everything set up, I was very happy with the performance of the GPU, which is about 4 to 6 times faster for 3D graphics, and 2.5 to 5x faster for AI workloads than the integrated Intel Arc graphics in the Core Ultra 5 155H Meteor Lake SoC. The combo basically provides the performance of a high-end tower PC in a much smaller form factor.

Other features mostly worked great, too. This includes the built-in stereo speakers (good quality too), the dual-mic array, the 3.5mm audio jack, USB 3.2 Gen2 (10 Gbps) ports, and the fingerprint scanner. The 2.5GbE port also works fine in normal conditions, except for saturated bi-directional transfers, where the performance collapses in one direction using iperf 3.19. The USB-C (40 Gbps) port on the rear panel was not usable for me when testing it with two Thunderbolt mass storage devices. That’s because the port is limited to 10 Gbps USB 3.2 when it’s attached to the Mind 2 through the Mind Link. USB fallback should have worked in theory, but it didn’t here. This USB-C appears to be especially useful when using the Mind Graphics 2 as an eGPU dock with another machine, but we don’t own any hardware with a Thunderbolt port right now, so I had to skip that part.

I’m not quite done yet, as the next step will be to test the Sermoon S1 3D scanner with the Mind 2 + Mind Graphics 2 dock, since the Intel Raptor Lake laptop I used last time around was clearly not powerful enough for the task. I’m not sure I’ll test the Mind Graphics 2 with Ubuntu 26.04, as we received more review samples than anticipated, and I’m probably busy until mid July, but let me know if you’d like a Linux test anyway.

I’d like to thank Khadas for sending the Mind 2, Mind xPlay, and Mind Graphics 2 dock for review. You’ll find the mini PC for about $1,099.00 on AliExpress and the Khadas store in its Intel Core Ultra 7 155H/32GB/1TB configuration, while the Mind xPlay kit goes for $399 on AliExpress/Khadas store, and the Mind Graphics 2 dock sells for $1349 (AliExpress or Khadas Store). This means a complete system, as reviewed above (without the xPlay), costs $2448 plus shipping and taxes.

The post Khadas Mind Graphics 2 review – A powerful NVIDIA Geforce RTX 5060 Ti eGPU dock for the Mind 2 mini PC appeared first on CNX Software - Embedded Systems News.

Yesterday, I wrote about the OpenTrafficMap ESP32-C5 C-ITS receiver board to monitor and potentially optimize traffic using 802.11p / ITS-G5 V2X communication over 5.9 GHz WiFi 6, and map all detected nodes on the OpenTrafficMap website.

Peter Holzhauser  (Pit711) forked the ESP32-C5 C-ITS receiver firmware to port it to the Waveshare ESP32-C5-WIFI6-KIT development board and added BLE streaming. He also designed the V2X2MAP open-source Android app to interface with the board (since 5GHz WiFi on phones can’t usually handle 802.11p), and a Windows installer to flash the firmware.

While the ESP32-C5 C-ITS receiver board integrates an ESP32-C5-WROOM-1 module, a GPS module, and an Ethernet port with PoE, the V2X2MAP project leverages any recent Android smartphone with a USB OTG port and GPS, so you can simply connect most ESP32-C5 boards to monitor live traffic at proximity (a few hundred meters to several kilometers) on your phone including traffic lights, public transports, and other vehicles with 802.11p V2X communication enabled.

More specifically, it captures the signals from vehicles’ on-board units (OBUs) and roadside units (RSUs) broadcast on the dedicated 5.9 GHz V2X band:

• CAM (Cooperative Awareness) — GPS coordinates and speed
• DENM (Decentralised Environmental Notification) — For example, “hazard ahead!”
• SPATEM (Signal Phase + Timing) — Traffic-light countdowns
• MAPEM (Map Extended Message) – Intersection geometry.

This allows the app to update the map without any connection to the cloud once the offline map is downloaded. It can also optionally update the data through MQTT to cits1.opentrafficmap.org or your own server, and record data to open it in Wireshark.

The source code for the firmware, V2X2MAP Android app, and a Python bridge + local dashboard can be found on GitHub, along with documentation instructions to build everything from source. Note that Claude AI was used for some of the code. Everything is released under an MIT license.

While the firmware was ported to the Waveshare ESP32-C5-WIFI6-KIT ($10 to $20 on AliExpress, Amazon, and the Waveshare shop), it should be possible to port it to other ESP32-C5 boards relatively easily if you are familiar with the ESP-IDF framework. Since dual-band WiFi is required, and more specifically, the 5.9 GHz band, this won’t work on other ESP32 boards.

There’s also a legal disclaimer at the end of the README: “Receiving and forwarding ITS-G5 radio data may be subject to national telecommunications law and data-protection law. The Android app shows a disclaimer on first launch. Use at your own risk”. The main issue here appears to be GDPR related when contributing data to the public map.

Another important note is that ITS-G5 is primarily a European technology for 802.11p V2X (Vehicle-to-Everything) communication part of the C-ITS (Cooperative Intelligent Transport Systems) framework, and most other countries rely on C-V2X (Cellular V2X). So, while you can probably use V2X2MAP (legally) in any country, you’ll probably feel lonely doing so outside of Europe.

More details can be found on the project’s website.

Thanks to Ready4Sushi for the tip.

The post Monitor live traffic from V2X signals with V2X2MAP open-source Android app and an ESP32-C5 development board appeared first on CNX Software - Embedded Systems News.

The ESP32-C5 C-ITS receiver project is an open-source hardware board that gathers data over 802.11p V2X communication from nearby traffic lights, public transportation (buses, trams…), trucks, cars, motorcycles, and even pedestrians to display the results on online maps.

It works using the ITS-G5 protocol built on top of 802.11p V2X (Vehicle-to-Everything), which requires a 5.9 GHz WiFi radio, and makes the ESP32-C5 an ideal candidate. The standard requires a C-ITS Station (ITS-S) found in vehicles (on-board units – OBU) or roadside units (RSU) handling both transmission and reception, and a receiver to handle incoming wireless signals, decode C-ITS (Cooperative Intelligent Transport Systems) messages, and feed the data into online maps like OpenTrafficMap (See screenshot below).

ESP32-C5 C-ITX receiver specifications:

• Wireless module – ESP32-C5-WROOM-1 (ESP32-C5-WROOM-1-N16R8 or ESP32-C5-WROOM-1-N8R8)
  • SoC – ESP32-C5
    • CPU
      • Single-core 32-bit RISC-V processor @ up to 240 MHz
      • Low-power RISC-V core @ 40 MHz acting as the main processor for power-sensitive applications
    • Memory – 384 KB SRAM on-chip, 8MB PSRAM
    • Storage – 320 KB ROM
    • Connectivity
      • Dual-band (2.4GHz/5 GHz) 802.11ax WiFi 6, with 802.11b/g/n WiFi 4 standard fallback
        • 20MHz bandwidth for the 802.11ax mode
        • 20/40MHz bandwidth for the 802.11b/g/n mode
        • OFDMA (Orthogonal Frequency Division Multiple Access) mechanism for both uplink and downlink
        • MU-MIMO capability for downlink
        • Target Wake Time (TWT) support
      • Bluetooth 5.0 Low Energy (LE)
      • 802.15.4 radio for Zigbee 3.0 and Thread 1.3
  • Storage – 8MB or 16MB SPI flash
  • Antenna – PCB antenna
• Storage – MicroSD card slot
• Ethernet – 10/100Mbps Ethernet RJ45 port via KSZ8851SNL SPI to Ethernet chip
• USB – 2x USB-C ports for JTAG and UART
• Sensor – LM75BDP temperature sensor
• Power Supply
  • 7 V – 58 V input on active/passive PoE
  • TPS2378DDAR for 802.3af/at compliant active PoE
• Dimensions – 93 x 50 mm

The KiCAD 10 hardware design files (schematics, PCB layout, BoM…), the source code for the firmware, and a Node.js script that bridges raw 802.11 packets on NATS (Neural Autonomic Transport System) into tshark (Wireshark command line), and publishes the decoded JSON back to NATS are available on OpenTraffic’s Codeberg account.

The developer gave a talk at Graz Linux Days 2026 (German) about a month ago, and uploaded the presentation slides explaining the project (AI-translated version in English).  At the time, they had already deployed 20 receivers with ranges of several hundred meters in urban areas, and over 10 km with line-of-sight. The ESP32-C5 can be attached to a 4G LTE modem for internet connectivity, and an OpenSCAD enclosure is also provided.

How well this works depends on whether infrastructure like traffic lights and public transport implements 802.11p V2X connectivity, and of course, the number of users deploying such C-ITS receivers.

The ESP32-C5 C-ITS receiver board is not available for sale online in the traditional way, but seeing the interest after their talks, the developers ordered 450 pieces of the board, and you can purchase a board with an optional enclosure for about 20 Euros as part of a group buy. The instructions are on the Wiki, and they only ship to several countries within Europe, or you can pick up yours in Graz, Austria.

Thanks to Karl for the tip.

The post OpenTrafficMap ESP32-C5 C-ITS receiver board can help improve traffic efficiency using 802.11p V2X communication appeared first on CNX Software - Embedded Systems News.

SpacemiT has sent me a K3 Pico-ITX Chassis Kit for review. It’s based on the K3 Pico-ITX motherboard with the SpacemiT K3 16-core RISC-V Edge AI processor housed in a compatible chassis.

I’ll start the review with an unboxing, a teardown, and a first boot to the pre-installed Bianbu OS. In the second part of the review, I’ll perform feature testing and run several benchmarks (see early K3 benchmarks for reference) to evaluate the status of the software and performance of the system.

SpacemiT K3 Pico-ITX Chassis Kit unboxing

I received a kit in a retail package reading “RISC-V AI CPI K3 RVA23 Profile Chip” and a UGREEN USB-C dock with a few USB-A ports, HDMI output, and 100W USB PD support. The dock will make perfect sense once we connect the system for our first boot.

I was initially expecting a Pico-ITX SBC, so I was a little surprised to see a complete system/mini PC instead. It ships with a USB debug cable, a screwdriver, and a sheet in Chinese with two QR codes, one of which points to the user guide (available in English).

The front features a power button with LED and a 3.5mm audio jack.

Most ports are on the rear panel: 10GbE SFP+ cage, Gigabit Ethernet RJ45 port, four USB 2.0 ports, a USB OTG Type-C port, and a USB Type-C port with DisplayPort Alt. mode and USB PD for power.

K3 Pico-ITX Chassis Kit teardown

Since the kit comes with a screwdriver, it’s clearly designed to be opened. We can do so by loosening the four screws on the bottom of the enclosure.

The bottom cover comes off easily, and we can access the “SpacemiT MUSE_Pico-ITX” board. As a side note, one of the screws for M.2 slots was roaming freely in the case, so I had to put it back into place…

The main reason for people to open the case is to make use of the M.2 Key-M 2280 (PCIe Gen3 x4) slot for NVMe SSD or other module, or the M.2 B-Key 2242/3052 (PCIe Gen2 x2 + USB 2.0) socket and Nano SIM card for 4G LTE or 5G cellular connectivity.  RTI01 and RTI02 “Real-Time Expansion” connectors on the right might also be useful, but I’ve yet to find an expansion board for these…

Another reason is to connect the serial console for debugging. The Tx, Rx, and GND text above corresponds to the pinout of the 3-pin header. The documentation does not mention anything about the cable itself, so I might have mixed up the color of the wires above. I’ll check this in detail in the second part of the second. There’s also a 2-pin connector next to it for 12V DC power input.

I loosened four more screws and disconnected the System, Audio, and Battery (RTC) cables to take out the board. It didn’t come off that easily, and most people won’t need to do that since there’s no purpose for it, except to check the hardware or move the board to another chassis.

We’ll find all the main chips on this side. The SpacemiT K3 SoC (grey), a Winbond W25064JWSIQ flash (64 Mbit), what should be a 128GB UFS chip (Kingston UFS128-TY7B under the small thermal pad), two 8GB Rayson RS2G32LO5D4FDB-31BT LPDDR4x for 16GB of RAM, RealTek RTL8127ATF 10GbE and RTL8211F GbE controllers, a Slimport ANX7447 USB Type-C crosspoint switch, and an AU4562B step-down regulator.

The cooling solution is a heatsink with a PWM-controlled fan. We’ll also notice the RTC battery and WiFi/Bluetooth antennas placement in the photo below.

First boot to Bianbu OS

Let’s reassemble everything to boot the mini PC.  If you own a monitor with USB PD power output, then a single USB-C cable can carry power and video to the K3 Pico-ITX Chassis Kit. If not, another option is to use a 12V power supply and connect it to the internal 2-pin connector, but this requires opening the chassis.  So instead, I used the provided UGREEN USB-C dock to connect the power (65W USB-C adapter for Khadas Mind 2), and connect an HDMI display (Eazeye Radiant 15-6-inch Transflective LCD). I also added an RF dongle for a keyboard and mouse combo. I was greeted with the Bianbu OS setup wizard in Chinese, and but I could easily change that to English in two clicks.

I went through the setup wizard to set the timezone, keyboard layout, and create a user. Once done, we are asked to reboot the system.

I could login just fine, and quickly checked the system monitor showing the eight application cores, and eight AI cores.

I had no network connection at this stage, so I connected the K3 Pico-ITX to my WiFi 6 router and could browse the web normally using the pre-installed Chromium web browser.

Finally, I connected to the system over SSH to check system info with inxi:

jaufranc@CNXSOFT-spacemitk3picoitx:~$ inxi -Fc0
System:
  Host: CNXSOFT-spacemitk3picoitx Kernel: 6.18.3-generic arch: riscv64
    bits: 64
  Console: pty pts/0 Distro: Bianbu 4.0rc2 (Resolute Raccoon)
Machine:
  Type: RISCV System: SpacemiT K3 Pico ITX details: N/A
    serial: HW3MPK3161280213
CPU:
  Info: 16-core model: Spacemit X100 variant: riscv bits: 64 type: MCP cache:
    L2: 10 MiB
  Speed (MHz): avg: 2200 min/max: 614/2400:2000 cores: 1: 2200 2: 2200
    3: 2200 4: 2200 5: 2200 6: 2200 7: 2200 8: 2200 9: 2200 10: 2200 11: 2200
    12: 2200 13: 2200 14: 2200 15: 2200 16: 2200
Graphics:
  Device-1: saturn-hee driver: spacemit_drm_drv v: N/A
  Device-2: saturn-edp driver: spacemit_drm_drv v: N/A
  Display: unspecified server: Xwayland v: 24.1.8 driver: N/A tty: 80x24
    resolution: 1920x1080
  API: EGL v: 1.4,1.5 drivers: pvr,swrast
    platforms: gbm,wayland,surfaceless,device
  API: OpenGL v: 3.3 vendor: mesa v: 24.0.1 note: console (EGL sourced)
    renderer: softpipe
  Info: Tools: api: eglinfo,glxinfo wl: kanshi,wlr-randr x11: xdriinfo,
    xdpyinfo, xprop, xrandr
Audio:
  Device-1: simple-audio-card driver: asoc_simple_card
  Device-2: simple-audio-card driver: asoc_simple_card
  Device-3: simple-audio-card driver: N/A
  Device-4: simple-audio-card driver: N/A
  API: ALSA v: k6.18.3-generic status: kernel-api
  Server-1: PipeWire v: 1.6.0 status: active
Network:
  Device-1: Realtek RTL8127 10GbE driver: r8127
  IF: enP2p1s0 state: down mac: 50:0a:52:0b:82:18
  Device-2: Realtek RTL8852BE PCIe 802.11ax Wireless Network
    driver: rtw89_8852be
  IF: wlP4p1s0 state: up mac: b0:8c:b3:9a:83:68
  Device-3: k3-gmac driver: dwmac_spacemit_ethqos
  IF: end0 state: down mac: 50:0a:52:0b:e6:64
  Device-4: rfkill-gpio driver: rfkill_gpio
  Device-5: dwmac-5.10a driver: N/A
  IF: end0 state: down mac: 50:0a:52:0b:e6:64
  Device-6: dwmac-5.10a driver: N/A
  IF: enP2p1s0 state: down mac: 50:0a:52:0b:82:18
  Device-7: dwmac-5.10a driver: N/A
  IF: wlP4p1s0 state: up mac: b0:8c:b3:9a:83:68
Bluetooth:
  Device-1: Realtek Bluetooth Radio driver: btusb type: USB
  Report: hciconfig ID: hci0 state: up address: B0:8C:B3:9A:83:69 bt-v: 5.3
Drives:
  Local Storage: total: 119.27 GiB used: 6.27 GiB (5.3%)
  ID-1: /dev/sda model: TY7B-128 size: 119.27 GiB
Partition:
  ID-1: / size: 116.78 GiB used: 6.21 GiB (5.3%) fs: ext4 dev: /dev/sda3
  ID-2: /boot size: 223.7 MiB used: 58.8 MiB (26.3%) fs: ext4 dev: /dev/sda2
Swap:
  Alert: No swap data was found.
Sensors:
  System Temperatures: cpu: 65.0 C mobo: N/A
  Fan Speeds (rpm): cpu: 3329
Info:
  Memory: total: N/A available: 15.63 GiB used: 1.8 GiB (11.5%)
  Processes: 352 Uptime: 1h 0m Init: systemd Shell: Bash inxi: 3.3.40

Everything seems to be properly detected, including the 16 RISC-V cores clocked up to 2.2 GHz, TY7B-128 UFS storage, 16GB of RAM, 10GbE networking, RealTek RTL8852BE  WiFi 6E and Bluetooth 5.3 module, and more. The idle temperature seems a little high at 65°C, so that’s something I’ll have to keep an eye on.

That will be all for today. It’s my first SBC with a Realtek RTL8127, and the second will be the ODROID-H5. Since the K3 Pico-ITX SBC comes with an SFP+ cage, I’ll need to purchase an SFP+ to RJ45 adapter for testing 10 Gbps Ethernet before the second part of the review.  In the second part of the review, I’ll run some benchmarks, the headless ones we’ve all done remotely, plus some others requiring a display, and test all/most features to check out what works and what doesn’t with Bianbu 4.0 OS.

I’d like to thank SpacemiT for sending the K3 Pico-ITX Chassis Kit for review. I have no idea where to buy the exact “K3 Pico-ITX Chassis Kit” I have received for review. SpacemiT relies on a range of distributors such as Banana Pi and Sipeed for the K3 Pico-ITX SBC ($399 for the 16GB/128GB model). The closest option to the “Chassis Kit” is the Milk-V Jupiter 2 with a Pico-ITX chassis using two external antennas. It sells for $575 for the 32GB RAM/256GB UFS version, as the 16GB/128GB model is out of stock.

The post SpacemiT K3 Pico-ITX Chassis Kit Review – Part 1: Unboxing, teardown, and first boot appeared first on CNX Software - Embedded Systems News.

Two ESP32-S31 development boards are currently in the works, and the documentation is already available. The ESP32-S31 was unveiled last March, and it’s the most feature-rich ESP32 wireless microcontroller so far with two RISC-V cores, Gigabit Ethernet, 2.4 GHz WiFi 6, Bluetooth, and 802.15.4 connectivity, LCD and camera interfaces, and more.

The ESP32-S31-Function-CoreBoard-1 development board is based on the ESP32-S31-WROOM-3 wireless module, and offers Gigabit Ethernet, USB 2.0 OTG, and onboard audio peripherals for connected IoT applications.

The ESP32-S31-Korvo-1 is a multimedia-focused development board based on the same ESP32-S31-WROOM-3 module, but featuring a dual-microphone array for speech recognition and near/far-field wake-up, two speaker connectors, a 4.3-inch LCD, a 3MP camera module, and a microSD card slot. It targets low-cost, low-power, connected audio/video and HMI applications.

ESP32-S31-Function-CoreBoard-1 development board

Specifications:

• Wireless module – ESP32-S31-WROOM-3
  • SoC – ESP32-S31NRV16
    • CPU
      • RISC-V HP (High-performance) RV32IMAFCP CPU @ 320 MHz with FPU, SIMD, etc.
      • RISC-V LP (Low-power) MCU core
    • Memory – 512 KB SRAM, 16MB PSRAM
    • Wireless –
      • 2.4 GHz WiFi 6 (802.11ax)
      • Bluetooth 5.4 with support for LE Audio, direction finding, Bluetooth Mesh 1.1, and Classic (BR/EDR)
      • 802.15.4 radio for Zigbee, Thread, and Matter
  • Storage – 8, 16, or 32MB flash (TBC)
  • Antenna – PCB antenna
  • Dimensions – 30 x 22 x 3.5mm
• Audio
  • On-board microphone
  • Speaker output port for up to 4 Ω, 3 W speaker
  • ES8311 low-power mono audio codec
  • NS4150B 3W mono Class D audio power amplifier
• Networking – Gigabit Ethernet RJ45 port via YT8531DC-CA PHY
• USB
  • USB 2.0 Type-A high-speed port, up to 500mA
  • USB 2.0 Type-C full-speed serial/JTAG port for debugging, flashing, and power
  • USB Type-C to UART port for power, flashing, and communication with the ESP32-S31 chip via the on-board USB-to-UART bridge.
• Expansions – 40-pin J2 GPIO header
• Misc – 2-pin J5 header for current measurement
• Misc
  • Reset and Boot buttons
  • 3.3V power LED
  • RGB LED (GPIO60)
• Power Supply
  • 5V via USB-C port
  • 5V to 3.3V DC/DC Converter
  • TPS2051C USB power switch (up to 500 mA output current limit)
• Dimensions – 65 x 55 mm

The ESP32-S31-Function-CoreBoard-1 requires a good USB 2.0 USB-A to Type-C cable and a computer running Windows, Linux, or macOS for development. You’ll find details on the documentation website, where you can also request samples if you have a proper project for it. [Update May 25: It’s now available on AliExpress for $19.70]

ESP32-S31-Korvo-1 multimedia development board

Specifications:

• Wireless module – ESP32-S31-WROOM-3 as described above, except the flash capacity is listed (16MB flash)
• Storage
  • MicroSD card slot (SDIO 3.0 capable)
  • Quad SPI NAND flash footprint, multiplexed with microSD interface
• Display – LCD connector for ESP32-S3-LCD-EV-Board-SUB3 (4.3-inch, 800 x 480 resolution)
• Camera – Camera connector for 3MP OV3660 camera module
• Audio
  • 2x speaker output ports, each with an NS4150B mono class-D amplifier to drive a 4Ω, 3W speaker.
  • 2x analog microphones connected to ES8389 low-power stereo codec with dual ADC/DAC, low-noise preamp, headphone driver, digital effects, analog mixing, and gain control.
• USB
  • USB 2.0 Type-A High-Speed OTG port, up to 500 mA output current
    USB Type-C UART port for power, flashing the firmware, and communication with the ESP32-S31 via the onboard USB-to-UART bridge (up to 3 Mbps).
  • USB Type-C port for power only
• Misc
  • Power switch
  • Reset and Boot buttons
  • Function Buttons:  PLAY, SET, VOL-, and VOL+ for UI control and audio application testing
  • Power LED
  • RGB LED (GPIO8)
• Power Supply
  • 5V via USB-C port
  • Buck DC-DC converter for 3.3V system power.
  • TPS2051C USB power switch with 500 mA current limit.
  • 5V to 3.3V LDO
  • 3.3V to 1.8V LDO for SPI NAND flash (not populated by default)
  • 3.3V to 2.8V and 3.3V to 1.5V LDOs for external camera module.
• Dimensions –  106.5 x 89.20 mm

The ESP32-S31-Korvo-1 board requires one or two speakers, two USB 2.0 Type-A to Type-C cables, and a computer running Windows, Linux, or macOS for development. Both boards are supported by the ESP-IDF framework, but since the chip is so new, not every feature will be supported at this time, and you’ll need to use the master branch. You can find more details and/or request samples for the Korvo devkit on the documentation website. [Update May 25: it’s now available on AliExpress for $58.41]

The post ESP32-S31 development boards unveiled for IoT, Smart Audio, and HMI applications appeared first on CNX Software - Embedded Systems News.

Huion Note E is an Android 15 note-taking device/electronic notebook powered by a MediaTek Helio G99 SoC and equipped with an 8.4-inch color LCD designed to work with a battery-free PenTech 3.0 pressure-sensitive stylus.

The device ships with 6GB LPDDR4 memory, 128GB storage, and a single USB-C port for charging and data. The company says the 1920×1080 display leverages “AG nano-etching technology to create a refined, matte surface that mimics the traditional pen-on-paper experience” and provides an e-Paper-like experience, while maintaining a 60 Hz refresh rate.

Huion Note E (X50) specifications:

• SoC – MediaTek Helio G99
  • CPU – Octa-core processor with 2x Arm Cortex-A76 cores @ 2.2 GHz, 6x Arm Cortex-A55 cores @ 2.0 GHz
  • GPU – ARM Mali-G57 MC2
  • VPU
    • Video Encoding – H.264, H.265/HEVC up to 2Kp30, 1080p60, 720p120
    • Video Playback – H.264, H.265/HEVC, VP-9 up to 2Kp30, 1080p60, 720p120
• System Memory –  6GB LPDDR4
• Storage – 128GB storage
• Display
  • 8.4-inch LED with 1920×1200  resolution, 60Hz  framerate
  • Contrast Ratio – 1000:1
  • Brightness – 300 nits (Typ.)
  • Color Gamut – 99% sRGB
  • Working Area – 180.3 x 112.7 mm
  • DC dimming and a soft-light display
• Audio
  • Built-in speaker
  • Dual microphone
• Camera – Rear-camera
• Wireless – Dual-band WiFi and Bluetooth 5.2
• USB – 1x USB 2.0 Type-C port for charging and data transfer (compatible with USB Type-C headphones)
• Stylus – Battery-free PW510 digital pen
  • 8,192 pressure sensitivity levels
  • 400 pps report rate
  • Tilt recognition
  • Quick eraser
  • PenTech 3.0 technology
• Misc
  • Power button
  • Fingerprint unlock
  • Protective case with magnetic connection for stylus
• Power Supply
  • 9V/2A via USB-C port
  • 4,500 mAh battery good for about 6.5 hours at 50% brightness; 2-hour charge time
• Dimensions – 203 x 143 x 7.4 mm
• Weight – 348 grams

On the software side, the Huion Note E runs Android 15 with “Smart Handwriting tools” to configure pressure sensitivity and type of brush (fountain pen, ball point pen, highlighter pen…), handwriting to text conversion, content search, audio recording and note synchronization, document annotation, and landscape/portrait modes support. The Huion Note app aims to ease note-taking on Android devices, including the Note E.

I previously experienced the company’s pressure-sensitive digital pen technology when I reviewed the Kamvas Pro 16 (2.5K) 15.8-inch drawing tablet with a similar PW517 8192-level stylus. That was quite a niche display targeting artists and graphic designers, and with the Note E (X50), Huion now aimed at the broader consumer market.

The Note E X50 “Smart Digital Notebook” sells for $369 on Amazon and the Huion store and ships with a protective case, a battery-free digital pen, a USB cable, a nib clip, ten replacement nibs, and a Quick Start Guide. You’ll find more details on the product page, including a user manual (download section).

The post Huion Note E – A 8.4-inch Android 15 electronic notebook with battery-free, pressure sensitive stylus appeared first on CNX Software - Embedded Systems News.

M5Stack Core StopWatch is an ESP32-S3-powered WiFi and Bluetooth development board with a 1.75-inch round AMOLED touch display, 16MB flash, and 8MB PSRAM designed for portable IoT devices, electronic badges, and wearables.

The device also features a 1W speaker and a MEMS microphone for voice interaction, two programmable buttons, a vibration motor, a 6-axis IMU sensor, an RTC, and expansion capabilities through a Grove connector and GPIO headers. It’s powered by a 450 mAh battery managed by an “M5PM1 multi-level power management” solution and charged over a USB-C port.

M5Stack Core StopWatch specifications:

• SoC – Espressif ESP32-S3R8
  • CPU – Dual-core Tensilica LX7 microcontroller up to 240 MHz with vector instructions for AI acceleration
  • Memory – 8MB PSRAM
  • Wireless – WiFi 4 and Bluetooth 5.0 LE + Mesh connectivity
• Storage – 16MB flash
• Display – 1.75-inch AMOLED round touch display with 466 x 466 resolution
• Audio
  • 1W speaker
  • MEMS microphone
  • ES8311 audio codec
  • AW8737A amplifier
• USB – 1x USB Type-C port for power and programming
• Sensors – 6-axis IMU BMI270
• Expansion
  • 4-pin Grove connector
  • 7-pin and 6-pin 2.54mm pitch headers for up to 10x GPIO, UART, USB 2.0, VBAT, 3.3V, and GND
• Misc –
  • Power button, 2x programmable buttons
  • Built-in vibration motor
  • RX8130CE RTC chip
  • Lanyard hole
  • Rear magnetic attachment
• Power Supply
  • 5V via USB-C power
  • Built-in 450mAh battery
  • M5PM1 multi-level power management system
• Dimensions – 52.0 Ø x 15.5mm
• Weight – 39 grams

The company provides instructions and code samples to get started with the Arduino IDE using the M5Unified and M5GFX driver libraries, as well as PlatformIO. You’ll find these along with pinout and schematics on the documentation website.

It’s not quite the first ESP32-S3 devkit with a round AMOLED touch display, and competing products include the LILYGO T-Display-S3-AMOLED-1.43 and Waveshare ESP32-S3-Touch-LCD-1.85C, but the M5Stack device integrated the display and audio support in a more compact form factor. The Waveshare model can also handle audio, but it does so via a thicker audio expansion box.

The StopWatch touch AMOLED devkit is sold for $45 on AliExpress and the M5Stack store.

The post M5Stack StopWatch ESP32-S3 devkit offers 1.75-inch touch AMOLED, microphone, speaker, and GPIO expansion appeared first on CNX Software - Embedded Systems News.

Seeed Studio has just launched the reComputer RK3576/RK3588 Edge AI computers designed for developers, embedded AI innovators, robotics, industrial AI, vision AI, local LLMs, and real-world edge deployment.

Rockchip RK3576 and RK3588 computers are pretty common these days, and the Seeed Studio models offer triple video output, dual Ethernet (GbE or 2.5GbE), several USB ports, and M.2 expansion. But the most interesting part is probably the software with Armbian-based Linux OS and support for the reComputer AI Lab platform that enables one-click deployment of AI-accelerated computer vision, audio, and LLM/LVM demos.

reComputer RK3576/RK3588 specifications:

• SoC (one or the other)
  • Rockchip RK3576
    • Octa-core CPU – 4x Cortex-A72 cores at 2.2GHz, 4x Cortex-A53 cores at 1.8GHz
    • GPU – ARM Mali-G52 MC3 GPU
    • NPU – 6 TOPS (INT8) AI accelerator with support for INT4/INT8/INT16/BF16/TF32 mixed operations.
    • VPU
      • Video Decoder: H.264, H.265, VP9, AV1, and AVS2 up to 8K at 30fps or 4K at 120fps.
      • Video Encoder: H.264 and H.265 up to 4K at 60fps, (M)JPEG encoder/decoder up to 4K at 60fps.
  • Rockchip RK3588
    • Octa-core CPU – 4x Cortex‑A76  cores @ up to 2.4 GHz, 4x Cortex‑A55 core @ 1.8 GHz
    • GPU – Arm Mali-G610 MP4 GPU
    • NPU – 6 TOPS (INT8) AI accelerator with support for INT4/INT8/INT16/BF16/TF32 mixed operations.
    • VPU
      • Video decoder – 8Kp60 H.265, VP9, AVS2, 8Kp30 H.264 AVC/MVC, 4Kp60 AV1, 1080p60 MPEG-2/-1, VC-1, VP8
      • Video encoder – 8Kp30 H.265/H.264 video encoder
• System Memory
  • RK3576 – 4GB, 8GB, or 16GB LPDDR5
  • RK3588 – 8GB, 16GB, or 32GB LPDDR5
• Storage
  • MicroSD card slot
  • SPI flash for bootloader
  • Optional NVMe SSD via M.2 socket (see expansion section)
• Video Output
  • 1x (RK3576) or 2x (RK3588) HDMI 2.1 ports
  • DisplayPort 1.4 via USB-C port
  • 4-lane 22-pin MIPI DSI connector
• Video Input (RK3588 only) – HDMI 2.0 port
• Camera – 2x 22-pin 4-lane MIPI CSI connectors
• Audio – 3.5mm audio jack
• Networking
  • RK3576 – 2x Gigabit Ethernet RJ45 ports, one with PoE support (additional module required)
  • RK3588 – 2x 2.5GbE RJ45 ports, one with PoE support (additional module required)
  • On-board WiFi 6 and Bluetooth 5.4 with FPC antenna
  • Optional 4G LTE, LoRaWAN, or WiFi Halow via mini PCIe socket and SIM card slot (for cellular only)
• USB
  • RK3576 – USB 3.0 Type-A port, 3x USB 2.0 Type-A ports, 1x USB Type-C OTG port with DisplayPort Alt mode
  • RK3588 – 4x USB 3.0 Type-A ports, 1x USB Type-C OTG port with DisplayPort Alt mode
• Expansion
  • M.2 Key-M socket (PCIe 2.1 x1) for NVMe SSD or AI accelerator
  • RK3588 only – M.2 Key-M socket (PCIe 3.0 x4) for NVMe SSD or AI accelerator
  • 40-pin GPIO header
• Misc
  • Power, Recovery, and MaskROM buttons
  • Power, Status, and User LEDs
  • 2-pin header for RTC
  • 4-pin FAN header
• Power Supply – 9V-19V DC via power barrel jack
• Dimensions – 123 x 93 x 51 mm; PCB: 115 x 85mm
• Temperature Range
  • RK3576 – 0 to 60°C
  • RK3588 – 0 to 55°C
• Compliance – FCC/CE/TELEC/RoHS

The devices come with Armbian pre-installed in order to provide “long-term maintained system images, security updates, encrypted OS options, and OTA upgrade support for production-ready deployment”. They apparently worked with the Armbiab team on this, but neither Seeed Studio or the reComputer shows up on the website.

The company also offers the reComputer AI Lab platform, which includes optimized edge AI model demos for computer vision, LLM, VLM, speech-to-text (STT), and text-to-speech (TTS). The website also includes interactive tutorials, deployment tools, containerized applications, and community projects. You can also check the VLM + Yolo26 demo at the end of the article.

The reComputer RK3576 starts at $99 on the Seeed Studio’s store with 4GB of RAM, and the RK3588 model at $199 with 8GB of RAM on the same page. Both models may also soon show up on the company’s AliExpress store. Those are pre-orders with shipping scheduled to start on May 30, 2026.

The post reComputer RK3576/RK3588 Edge AI computers are supported by reComputer AI Lab one-click deployment platform appeared first on CNX Software - Embedded Systems News.

Flipper Devices has officially introduced the Flipper One open-source hardware portable Arm Linux platform and networking and Edge AI multi-tool powered by a Rockchip RK3576 octa-core Arm Cortex-A72/A53 SoC, and featuring a Raspberry Pi RP2350 for low-level control.

You may think of it as a successor to the popular STM32-based Flipper Zero hardware and wireless hacking tool, but the company stresses that the Flipper One is NOT a replacement for the Flipper Zero. It is a different product with mainline Linux kernel support, no binary blobs (probably not 100% true), and open-source drivers, and operating at a different level of the networking and wireless stacks with dual Gigabit Ethernet, Wi-Fi 6E, and optional 5G or 4G LTE modem.

Flipper One specifications:

• SoC – Rockchip RK3576
  • CPU
    • 4x Cortex-A72 cores @ 2.2GHz, 4x Cortex-A53 cores @ 1.8GHz
    • Arm Cortex-M0 MCU at 400MHz
  • GPU – ARM Mali-G52 MC3 GPU with support for OpenGL ES 1.1, 2.0, and 3.2, OpenCL 2.1, and Vulkan 1.2
  • NPU – 6 TOPS (INT8) AI accelerator with support for INT4, INT8, INT16, BF16, and TF32 mixed operations.
  • VPU
    • Video Decoder: H.264, H.265, VP9, AV1, and AVS2 up to 8K @ 30fps or 4K @ 120fps
    • Video Encoder: H.264 and H.265 up to 4K @ 60fps, (M)JPEG encoder/decoder up to 4K @ 60fps
• System Memory – 8GB LPDDR5
• Storage
  • 64GB UFS 2.2
  • MicroSD card slot (UHS-I SDR104)
• MCU subsystem
  • SoC – Raspberry Pi RP2350 dual-core Arm Cortex-M33/RISC-V microcontroller @ 150 MHz with 264KB SRAM
  • Storage – 16MB flash
• Display – 256 x 144 pixel display, 64-level grayscale, Gorilla glass; controlled by RP2350 MCU via QSPI
• Video Output
  • HDMI 2.1 port up to 4Kp120 with HDMI CEC support
  • DisplayPort 1.4 via USB-C port up to 4Kp120
• Audio
  • 3.5mm audio (stereo out + microphone) jack
  • Built-in microphone and speaker
  • Nuvoton NAU8822 audio codec
• Networking and wireless
  • 2x Gigabit Ethernet RJ45 ports via Realtek RTL8211F-CG controllers
  • Tri-band Wi-Fi 6 (802.11ax) 2×2 MIMO and Bluetooth 5.2 via WXT2AM2101 module (MediaTek MT7921AUN)
  • Optional 4G LTE or 5G modem via M.2 module and externally accessible Nano SIM (4FF) card slot
• USB
  • USB 3.1 Type-A host port
  • USB 3.1 Type-C host port
  • USB 3.1 Type-C host/device port with DisplayPort Alt. mode, 5V-12V USB PD power input
• Expansion
  • M.2 Key-B socket (PCIe 2.1 x1, USB 2.0, USB 3.1, SATA3, Serial Audio, UART / I2C / SIM card) for 2242/3042/3052 modules
  • 20-pin header for GPIOs from MCU and CPU, USB 2.0 for CPU, 5V, 3.3V, and GND
• Debugging – Debug port for RK3576 SoC and RP2350 MCU
• Misc
  • Directional touchpad, 5-button D-Pad
  • Push-to-talk button, 5x control buttons (Esc, View, Power, Edit, Run), App switcher button, and Back button
  • Power meter LEDs, network LEDs
  • Passive heatsink for cooling
  • Lanyard loop
• Power Supply
  • 5-20V USB PD via USB-C port (3.6V to 24V in PD PPS mode while charging the battery)
  • 7,000 mAh battery (note: mass production capacity not finalized)
  • TI BQ25792 charger IC up to 3.32A
  • TI BQ28Z610 fuel gauge
• Dimensions – 155 x 67 x 40 mm
• Weight – TBD
• Materials – PC/ABS body, TPU bumpers, PC/ABS buttons, anodized aluminum lanyard loop, anodized aluminum bracket

  

  

  

Flipper Devices partnered with Collabora software development company for the upstreaming work for the Rockchip RK3576 to make sure the Flipper One supports “full mainline Linux kernel with no binary blobs, no proprietary drivers, and no vendor-locked board support package”.  Some claims may be hard to meet, notably because the bootloader usually requires a closed-source binary blob for the RAM initialization (but an open-source alternative is being worked on), and more importantly, the firmware for WiFi and cellular is usually closed-source for regulatory (e.g., FCC) compliance.

Development is still in progress, and Flipper Devices welcomes users to contribute through the Developer Portal, containing information about hardware, firmware, Linux (Debian-based Flipper OS), UI using FlipCTL framework, testing, and documentation. It also includes task trackers, architectural discussions, and a live status of mainline Linux support. Finally, you’ll find several Flipper One-related repositories for the Linux build scripts, RP2350 MCU firmware, and more on GitHub.

Overall, the Flipper One will be a highly versatile, modular, open-source hardware and software Linux platform with networking and Edge AI capability. I’m using the future tense because the company is not quite ready to take orders at this time and will announce developer hardware availability, pricing, and consumer release timing in a future update.

The post Flipper One – A Rockchip RK3576-powered portable Arm Linux computer and networking multi-tool appeared first on CNX Software - Embedded Systems News.

Procolored X one is a 3-in-1 machine which combines a UV color printer, a dual-laser engraver (20W diode + 2W IR), and a UV-DTF sticker maker into a single, compact, enclosed unit. It is designed to simplify workflows for small businesses and DIY makers by bringing multiple functions into a single system.

Traditionally, creating full-color engraved products requires multiple machines and a messy pre-coating step before printing. The X one simplifies this by using a special X-axis dual-track design that lets you switch between laser cutting and UV printing without moving the object, making the workflow faster and easier.

Procolored X one specifications

• UV Printing System
  • Printhead -XH6 Dual Printhead (promising up to a 5x efficiency improvement)
  • Maximum Print Width – 33 cm (Supports A3 and A5 platforms)
  • Print Speed – A3 full-size print in ~10+ minutes
  • Ink System – 250ml refillable tanks with smart ink monitoring (notably larger than the 100ml tanks found in typical desktop units)
  • Maintenance – White ink circulation and stirring, automatic ink supply, and an optimized 8ml auto-cleaning cycle
• Laser System
  • Modules – Interchangeable 20W Blue Diode Laser + 2W Red (IR) Laser
  • Engraving Speed – 100–300 mm/s (Blue Diode)
  • Engraving Precision – 0.1 mm
• UV-DTF – Roll-to-roll output for instant peel-and-stick applications (no laminator needed)
• Misc – Vision Positioning with IR Height Sensing for automatic object recognition and zero-error alignment
• Supported Materials  – Wood, metal, acrylic, plastic, leather, stone, cork, silicone, and more; total 400+

One major concern when combining a laser and a UV printer in a single machine is dust. Lasers create dust, and UV printheads are very sensitive to dust, which can easily clog them and cause problems. Procolored solves this with their unique dual-track system. It keeps the laser and UV printing separate: when the laser is working, the UV printheads stay off. This effectively prevents dust from clogging the printheads. The machine is built with a sturdy metal frame, features a stylish orange-lit chamber, and includes a built-in touchscreen for easy control. It also uses a professional G7-certified color system for accurate, high-quality prints.

Speaking about the prints, the company also mentions that the machine comes with integrated UV varnish with instant-curing technology, which enables it to create raised, tactile, embossed “2.5D textures” and high-gloss finishes on materials.

The company does not clearly discuss software support, but a YouTube video shows they are now using a new all-in-one platform to control UV printing, laser engraving, and UV-DTF sticker making. Procolored developed this software in-house specifically for this machine, instead of relying on third-party tools. It supports both Windows and macOS. The software also handles smart features like automatic alignment, height sensing, and design tools, making it easier to use. This is especially useful for Mac users, as many similar machines only support Windows.

Previously, we have reviewed various laser engravers like the Creality Falcon A1 Pro, the ACMER X1, the LaserPecker LP5, and others, but none of these could handle color printing on the material, which makes the Procolored X rather a unique machine.

The Procolored X one full-color laser engraver is available on Kickstarter and has already surpassed its $100,000 goal with over $700,000 raised so far. It’s available in three bundles. The Basic bundle ($3,799) includes the machine, a 20W laser, platforms, inks, and accessories; the Standard bundle ($3,999) adds a UV-DTF kit and rotary attachment; and the Professional bundle ($4,199) also includes a 2W IR laser. Extra add-ons like an air purifier and ink bottles are sold separately. Shipping is about $199 to the US and $299 to Europe/UK, with no shipping to South America, Saudi Arabia, or India. Delivery is expected around January 2027.

The post Procolored X one – A 3-in-1 laser engraver with UV and UV-DTF color printing (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Synaptics Coralboard is a development board powered by a Synaptics Astra SL2619 Edge AI SoC with a 1 TOPS Synaptics Torq inference engine implementing a Google Coral NPU, and supporting hardware-accelerated Google Gemma 3 lightweight models.

The board features 2GB of RAM and offers MIPI camera and display interfaces, a microSD card slot, a USB Type-A port, microphone inputs (I2S), mikroBUS and Qwiic- expansion connectors, and optional Wi-Fi and Bluetooth connectivity through an M.2 expansion slot.

Synaptics Coralboard specifications:

• System-on-module – Grinn AstraSOM-261x
  • SoC – Synaptics Astra SL2619 SoC with 2x Cortex-A55 cores @ 2GHz, 1x Cortex-M52 core @ 200MHz, and 1 TOPS Synaptics Torq (Google Coral NPU)
  • System Memory – 2GB DDR4 @ 3200Mbps (optionally 1GB)
  • Storage – 16GB flash, up to 64GB
  • Input Voltage – 3.2V to 5.5V
  • Dimensions – 25 x 25mm, LGA178 package
• Storage – MicroSD card slot for mass storage
• Display – 4-lane MIPI DSI connector
• Camera
  • 22-pin 2-lane MIPI CSI connector
  • Google I/O 2026 edition variant – Arducam OV5647 1080p mini camera module
• Audio interfaces – I2S2 and I2S3 audio connectors (1.8V)
• Connectivity – Optional WiFi and Bluetooth via M.2 module
• USB
  • USB-A host port
  • USB-C port for power (5V only, no power negotiations) and data (device mode)
• Expansion
  • M.2 E-key or E+A key slot (SDIO 3.0, UART/PCM, and GPIOs)
  • I3C connector (1.8V)
  • Qwiic Connect System for I2C
  • mikroBUS connector – Google I/O 2026 edition variant: MikroBUS Sensor HAT board with camera connector, 2x microphones, a piezo buzzer, and a few user LEDs.
  • 20-pin header GPIO header
  • I/O voltage – 3.3V unless otherwise stated
• Debug interfaces
  • UART console for kernel messages
  • JTAG interface, THT holes for use with POGO-pin probe clip
  • Shunt resistors and related TPs (for power measurements of various functions) are accessible by pogo pins from the add-on board.
• Misc
  • Reset and user buttons
  • 3x LEDs (1x power, 2x user)
• Power Supply
  • 5V via USB-C port
  • 2-pin battery connector without charging functionality
• Dimensions – TBD

The board runs a Yocto Linux distribution, and is supported by the Synaptics Astra SDK with necessary bootloaders, kernel drivers, and configuration files. It can run Google’s Gemma 3 270M model out of the box via the open-source, MLIR-based Synaptics Torq toolchain. Learn about to get started with the board on the Google Developers website.

It is being showcased at Google I/O 2026 as part of “Jellectronica” live, AI-powered musical performance combining computer vision, real-time inference, and generative audio. In practice, an NPU-accelerated YOLOv8 object detection model tracks the movement of jellyfish from the Monterey Bay Aquarium via a live stream, and the motion data is translated into control signals that drive a generative music performance powered by Google DeepMind’s Lyria Realtime model. So basically, a live concert by a bunch of jellyfish “performers”. Watch the video demo below.

When I asked a representative (PR company) about pricing and availability, the current board is a limited edition specially made for Google I/O 2026 only. However, general availability and pricing of the Synaptics Coralboard will be announced later this year. A few more details may be found on the product page, and you can register your interest by filling out this form, but note that it reads: “Submission does not guarantee availability or participation“.

The post Coralboard features Synaptics Astra SL2619 Edge AI SoC, supports Google Gemma 3 inference appeared first on CNX Software - Embedded Systems News.

Little Gadgets’ Hacknect is an ESP32-S3-powered wireless USB Type-A hacking cable with Wi-Fi control, HID automation for keyboard/mouse events, payload deployment, and a microSD card slot hidden in one of the Type-A connectors.

The USB cable is designed for makers, developers, automation enthusiasts, and cybersecurity researchers and can be wirelessly controlled from your smartphone or computer. It just looks like a normal cable, so any free USB cable that comes your way may look suspicious in the future.

Hacknect specifications:

• Wireless SoC – ESP32‑S3 with WiFi 4 and Bluetooth 5.x
• Storage – MicroSD card slot (inside one of the USB Type-A ports)
• USB – 2x USB Type-A male ports (Full-speed, 12 Mbps)
• Features:
  • Keystroke injection – Automated keyboard payloads using HID emulation.
  • Mouse injection – Simulated mouse movements and automated actions.
  • Payload slots – Store and manage multiple payloads directly on the device.
  • Wi‑Fi Triggers – Trigger actions wirelessly from smartphones or computers.
  • Self‑Destruct Mode – Quickly erase stored payloads and sensitive data.
• Dimensions – TBC (mostly looks like a standard USB-A to USB-A cable)

The AHacknect cable can be controlled through mobile devices or desktop computers and laptops through a web‑based control panel accessible from a compatible web browser. You can just click on one of the topions to launch payloads instantly. Here are the options:

• Self destruct
• Keystroke injection
• USB keylogger
• Payload drop
• Send file
• Lock system
• Network disconnect
• Packet viewer
• System analyzer
• System reboot
• Remote console
• USB device monitor
• WiFi triggers

Technical details are fairly light for an “open-source” project, but we’re told that the firmware source code, example payloads, documentation, demo projects, getting started guides, and example automation scripts will be released once the cables start shipping.

It’s not the first such USB hacking cable around, as we previously covered the Hackstar cable with RP2040 or ESP32-S3 MCU offering similar features minus the microSD card slot.

The Hacknect has recently launched on Kickstarter and raised close to $14,000 so far.  Like other security-related/hacking devices, there’s a high premium compared to the actual BoM cost of the device, and rewards start at 65 Euros (about $75 US) for a white or red cable for the “Early Bird” pledge. This doesn’t include shipping, which adds about $19 US, and deliveries are expected to start by August 2026.

The post Hacknect – A wireless hacking USB cable with a built-in microSD card slot (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Hardkernel ODROID-H5 is an affordable Intel Core i3-N300 octa-core Alder Lake-N SBC providing a 10GbE RJ45 networking jack, and four M.2 PCIe slots for storage, wired/wireless networking, or AI accelerator expansion.

It relies on the same SoC as in the earlier ODROID-H4 Ultra SBC, but trades a single PCIe Gen3 x4 M.2 slot and four SATA ports for four M.2 slots, and upgrades from dual 2.5GbE  to single 10GbE. It still features four USB ports, but only one USB 3.0 port, and three USB 2.0 ports.

ODROID-H5 specifications:

• SoC-  Intel Core i3-N300
  • CPU – Octa-core “Alder Lake-N” processor @ up to 3.8 GHz (Turbo single-core) or 2.3 GHz (Turbo multi-core) with 6MB cache
  • GPU – 32EU Intel UHD Graphics @ 1.25 GHz
  • TDP: 7W
• System Memory – Up to 64GB DDR5 4800 MHz SO-DIMM memory
• Storage
  • Up to 4x 2280 NVMe SSDs via M.2 slots (See Expansion section)
  • eMMC flash connector
  • Optional SATA expansion card adding 6x SATA ports
• Video and Audio Output
  • HDMI 2.0 ports up to 4Kp60
  • 2x DisplayPort connectors up to 4Kp60
  • Triple independent display support
• Networking
  • 10GbE RJ45 port with WoL support via Realtek RTL8127AT 10 Gbps Ethernet controller; PCIe Gen3 x2; tested up to 9.49 Gbps with iperf3.
  • Optional 10GbE M.2 card(s)
  • Optional WiFi 6E or WiFi 7 via third-party M.2 modules
• USB
  • 1x USB 3.0 Type-A port
  • 3x USB 2.0 Type-A ports
• Expansion
  • 3x M.2 (PCIe Gen3 x2) slots
  • 1x M.2 (PCIe Gen3 x1) slot
  • 24-pin I/O expansion port with 2x I2C, 3x USB 2.0, 1x UART, 1x HDMI-CEC, 1x external power button, 5V, 3.3V, and GND
• Misc
  • Power and Reset buttons
  • System LEDs for Power, Sleep, PMIC, and NVMe (data transfer)
  • Passive heatsink
  • 4-pin PWM + Tacho fan connector for optional 12V fan
  • Dual BIOS with backup battery
• Power Supply – 11-20V via 5.5/2.1mm DC jack (15V/4A DC adapter recommended)
• Power Consumption
  • Headless Idle – 3.3W
  • Desktop GUI Idle – 4.5W
  • CPU + GPU stress test – 25W
  • Power off – 0.4W
  • Suspend – 0.9~1.3W
• Dimensions – 120 x 120 x 44 mm
• Weight – 320 grams with heatsink

ODROID-H4 and ODROID-H5 SBCs are quite expandable, and the Korean company provides multiple expansion cards for their Intel SBC. However, the new ODROID-H5 doesn’t support them all, and unsurprisingly, drops support for M.2 expansion cards (since it already has four) and the 2.5GbE networking card.

Here’s the ODROID-H5 board fitted with a 6x SATA port M.2 module on top…

… and three M.2 modules (NVMe SSD, Hailo AI accelerator…) and a DDR5 SO-DIMM module on the bottom side.

The company tested the board with Ubuntu 24.04/26.04 (host), Windows 11, and Debian 13 (Guest OSes) using hardware virtualization (VT-x), each showing on a separate display to demonstrate operating system support and triple display configurations.

The company also offers a range of cases and accessories, just like for the previous ODROID-H series boards. You’ll find more details in the long forum announcement about the new ODROID-H5 board and in the wiki.

I can’t find a direct competitor to the ODROID-H5, but the closest alternatives are probably 10GbE Alder Lake-N platforms, such as the iKOOLCORE R2 MAX mini PC and MW-N100-NAS mini-ITX motherboard, although they are based on older 10GbE controllers, rather than the new, low-cost, power-efficient RTL8127 10GbE PCIe controller.

The ODROID-H5 SBC sells for just $250, which looks pretty good for an octa-core Intel SBC with 10GbE networking, four M.2 sockets, and triple display support. However, it’s a barebone system at that price, and you still need to add storage, memory, a 15V/4A power supply ($11), and potentially other accessories such as an enclosure, wireless module, and so on.

The post ODROID-H5 – A $250 Intel Core i3-N300 SBC with 10GbE networking, four M.2 PCIe slots appeared first on CNX Software - Embedded Systems News.

After SpacemiT officially launched the K3 Pico-ITX SBC, and the K3 chip entered volume production, an Edge AI mini PC, and a laptop motherboard (for Framework 13) have been released. To add to the list, Firefly has recently launched the CSC2-N48SPK3, a massive 2U rack-mounted server based on multiple SpacemiT K3 SoCs designed to bring RISC-V computing power to enterprise racks.

While consumer devices and modular laptops are great for developers, large-scale server-side AI workloads require more powerful hardware. The CSC2-N48SPK3 addresses this with up to 48 SpacemiT K3 RISC-V compute nodes, each with an octa-core X100 SoC delivering up to 60 TOPS (Sparse) AI performance, up to 32GB LPDDR5 RAM, 128 GB UFS storage, and an optional NVMe SSD. To manage all of these nodes, Firefly relies on a Rockchip RK3588 octa-core Arm processor as the central control node.

Firefly CSC2-N48SPK3 specifications:

• Server Form Factor – 2U rack-mounted 48 Node high-density server
• Architecture
  • 48 distributed computing nodes + 1 control node
  • Total AI Performance – 2880 TOPS (60 TOPS x 48, INT4)
• Compute Nodes (x48)
  • SoC – SpacemiT Key Stone K3 8-core X100M 64-bit RISC-V AI processor; up to 2.4GHz
  • Performance – 130K DMIPS general-purpose computing power per node; single-core SpecInt2006 exceeds 9.0/GHz
  • Compliance – Fully complies with RVA23 Profile
  • Memory – Up to 32GB LPDDR5 per node (available in 8GB, 16GB, or 32GB)
  • Storage – 128GB UFS 2.2 per node
• Control Node / BMC (x1)
  • Based on Rockchip RK3588 octa-core Cortex-A76/A53 processor @ up to 2.4GHz
  • RJ45 Console port (BMC debug serial port, 115200 baud rate)
• Storage – 48x M.2 2280 PCIe NVMe SSD slots (optional)
• Display
  • HDMI out  up to 1080p60 for BMC management
  • Touch display on the front panel for real-time device info (temperature, fan speed, network IP, etc.)
• Networking
  • 4x 10GbE SFP+ cages
  • Gigabit Ethernet RJ45 port (MGMT, used as BMC management network)
• USB – 2x USB 3.0 ports (lower port supports OTG for BMC upgrades via USB flash drive)
• Misc –
  • Reset, Power, and Restart BMC buttons
  • 12x high-speed cooling fans
• Power – 2x AC redundant power supplies (hot-swappable)
• Dimensions – 724.0 x 430.0 x 88.8 mm
• Weight – 23.1 kg (net), 25.3 kg (with packaging)

The server supports the RVV vector extension, native FP8 inference, and multimodal acceleration, delivering over 10 tokens/s on local 30B parameter models. It ships with a pre-installed Linux distribution and supports build systems like Open Build Service and Koji. Firefly claims it is much faster than emulation, with single-node performance up to 7× higher than x86 QEMU. In one example, the RISC-V K3 node compiled the Linux 6.18 kernel in 22 minutes and 28 seconds, compared to 2 hours and 50 minutes on an Intel Xeon Gold 6548Y+ running QEMU.

On the hardware side, the storage options are very extensive. The board includes 48 M.2 PCIe slots, so each compute node can have its own dedicated NVMe SSD. If you use 8TB or 16TB SSDs, that adds up to as much as 48 × 16TB, or 768TB of storage. It also supports security and virtualization, including three-level privilege modes (M/S/U), built-in protection against Specter and Meltdown attacks, and full hardware virtualization using RVH1.0, RV AIA, and RV IOMMU extensions. Previously, we have written about several interesting Arm-based rackmount servers, including Firefly’s own Cluster Server R2 and Huawei  Taishan X6000 V2 Server, both of which are 2U servers.

The Firefly CSC2-N48SPK3 RISC-V server is now available, but as expected for enterprise rackmount hardware, it is quite expensive. Pricing on the official Firefly store starts at $38,829 for a configuration with 48 compute nodes, each with 16GB RAM and 128GB storage, along with an 8GB RK3588 control node. A version with 32GB RAM per node is also available. More details are available on the product page.

The post Firefly CSC2-N48SPK3 – A 2880 TOPS RISC-V AI server with 48 SpacemiT K3 Nodes, 48 NVMe SSDs appeared first on CNX Software - Embedded Systems News.

Canonical has just introduced Ubuntu Core 26, based on the recently-released Ubuntu 26.04 LTS and designed for IoT devices and embedded systems, with precise Linux builds, optimized OTA updates, live kernel patching, and enhanced hardware-backed protection for mission-critical deployments.

Offered with up to 15 years of security maintenance, Ubuntu Core 26 minimal, immutable operating system enables reduced installation times, 90% smaller OTA updates, and precision-led builds via Chisel. Every component is a containerized snap, just like in prior Ubuntu Core releases. Canonical says it can help companies meet requirements for the EU Cyber Resilience Act (CRA) with securely designed and private AI deployments relying on hardware-based trust.

Ubuntu Core 26 highlights:

• Faster snap installation and updates – An improved snap-delta format reduces update sizes between 50% and 90% for most snaps. For instance, updates to the Core base snaps dropped from 16MB to 1.5MB in size. This lowers data usage and increases uptime.
• Precision builds with Chisel – The build system relies on Canonical’s release-specific package slice definitions, enforcing explicit, traceable dependencies. As a result, every file in the filesystem can be attributed to its originating slice and source package, improving the accuracy of integrity checks and vulnerability triage. The company explains this contrasts with approaches like Yocto builds, where provenance and dependency closure are largely implicit in layered recipes and post-processing. The new build system also contributes to a 7% reduction in base image size.
• CRA compliance through hardware-rooted security – Canonical assumes Manufacturer responsibilities under the CRA for the operating system’s release cycle by providing security maintenance for its core modules, continuous CVE monitoring and coordinated disclosure, and compliance with IEC 62443-4-1 (Cybersecurity in Industrial Automation and Control Systems).
• Livepatch rebootless kernel patching now works on both AMD64 and ARM64 targets.
• Developer tools improvements
  • New system snaps that speed up device deployment to new features in Snapcraft
  • Ubuntu Frame display server now supports multiple graphical applications on a single display, with configurable layouts, custom client placement, and a new accessibility launcher.
  • Canonical Observability Stack, an observability solution built on Juju and Kubernetes, enables Ubuntu Core to stream logs and metrics from the device to Cloud-based or private Grafana, Loki, and Prometheus infrastructure.
  • Snapcraft components offer a flexible new way to distribute large or optional resources, such as debug symbols, translations, or optional drivers, alongside their main snap without bloating the base installation.
• New API documentation – The Snapd REST API is now OpenAPI compliant and includes overhauled Swagger-based documentation.

Ubuntu Core requires 512MB RAM, 1GB storage, and is compatible with the following architectures: AMD64 (Intel/AMD 64-bit), ARM64 (64-bit Arm), ARMHF (32-bit Arm), and RISCV64 (64-bit RISC-V). Canonical provides Ubuntu Core 26 testing images for Generic x86 systems and Raspberry Pi 4/5 SBCs. Additional details may be found in the documentation.

The post Ubuntu Core 26 targets IoT devices and embedded systems, offers up to 15 years of security maintenance appeared first on CNX Software - Embedded Systems News.

Waveshare’s “Isolated RS232 / RS485 / CAN / CAN FD Expansion Board For Raspberry Pi” kit transforms your Raspberry Pi 4B or 5 into an industrial computer with a DIN rail or wall mount enclosure.

The solution also provides access to isolated interfaces, including two RS485, RS232, CAN FD, and CAN Bus interfaces, as well as 7-36V DC input through screw terminals. The complete solution also benefits from the Raspberry Pi’s Gigabit Ethernet, four USB ports, and full-size HDMI port (via adapter), as well as support for Raspberry Pi HAT and PCIe expansion boards for NVMe SSD, an additional GbE port, 4G LTE/5G cellular, etc…

Specifications:

• Compatible SBCs – Raspberry Pi 4B and Raspberry Pi 5
• Host interface – SPI + UART
• Isolated industrial interfaces via screw terminals
  • 1x CAN FD with 5kbps ~ 8 Mbps baudrate via MCP2518FD + MCP2562FD chips
  • 1x CAN Bus with 5kbps ~ 1 Mbps baudrate via MCP2515 + SN65HVD230 chips
  • 2x RS485 with 300 bps ~ 921600 bps baudrate via SC16IS752 + SP485 chips
  • 1x RS232 with 300 bps ~ 921600 bps baudrate via SP3232EEN chip
• Expansion
  • Optional NVMe, Ethernet, USB, 4G/5G through PCIe interface (Raspberry Pi 5 only)
  • Raspberry Pi HAT compatibility through 40-pin GPIO header (note: partial due to height restrictions and potential interference with SPI/UART)
• Misc – “Eco-friendly” plastic case with support for wall-mount & rail-mount installation
• Power Supply (one or the other)
  • 5V via USB Type-C port on Pi5 Connector Adapter (B)
  • 7-36V DC wide voltage input via screw terminals
• Dimensions – 154.6 × 83.7 × 59 mm
• Weight – 323 grams (kit without Raspberry Pi)

Software-wise, the kit is supported by Mike McCauley’s C library for Broadcom BCM2835, the wiringPi library, and various Python packages. You’ll find the schematics (PDF), RS-485 and CAN bus code samples, and instructions to get started on the wiki.

It’s far from the first Raspberry Pi industrial solution, but most are complete Raspberry Pi CM5-based industrial PC or controllers, although there are a few Raspberry Pi 5 industrial computers. The Waveshare kit enables users to convert an existing Raspberry Pi 4 or 5 board with RS232, RS485, and CAN Bus interfaces, a wide input voltage range, and a DIN rail enclosure at a relatively low cost.

Waveshare sells the “Isolated RS232 / RS485 / CAN / CAN FD Expansion Board For Raspberry Pi” kit for about $60 on AliExpress, $68.99 on Amazon, and $54.99 on the company’s online store.

The post $60 kit transforms the Raspberry Pi 4/5 into a DIN Rail industrial computer with isolated RS232, RS485, and CAN Bus appeared first on CNX Software - Embedded Systems News.

The ESP32-S3 PowerFeather V2 board is an ESP32-S3 WiFi and BLE IoT board with an Adafruit Feather form factor that supports LiFePO4/LFP batteries, as well as Li-Ion or LiPo batteries, and up to 18V DC input for solar panel connection.

As one could have guessed, it’s an update to the ESP32-S3 PowerFeather board introduced in 2024 with support for solar panel input, Li-Ion, and LiPo batteries. The V2 design is virtually identical, except it features an Analog Devices MAX17260 fuel gauge and a TPS631013 buck-boost regulator that keeps 3.3 V stable to add support for LiFePO4 batteries. Lithium Iron Phosphate batteries are said to be safer and longer-lasting than Li-ion or LiPo batteries, albeit at the cost of lower energy density.

ESP32-S3 PowerFeather V2 specifications:

• ESP32-S3-WROOM-1-N8R2
  • SoC – ESP32-S3
    • CPU – Dual-core Tensilica LX7 up to 240 MHz
    • Memory – 512KB SRAM, 16 KB RTC SRAM
    • Wireless – 2.4 GHz WiFi 4 and Bluetooth 5 LE + Mesh; PCB antenna
  • Memory – 2MB QSPI PSRAM
  • Storage – 8MB QSPI flash
• USB – USB Type-C OTG port for power and programming
• Expansion
  • 2x 16-pin 2.54 mm pitch headers with 23x multi-function GPIO:
    • UART, I2C, SPI, I2S, SDIO, PWM, CAN, RMT, Camera, LCD capable
    • Analog – 6x analog input capable
    • 5x touch input capable
    • 12x RTC capable (deep sleep pin hold, wake-up source)
    • Semitec 103AT input on thermistor pinhole
  • 4-pin JST SH STEMMA QT connector with I2C
• Misc
  • User and Reset buttons
  • Charging status LED (red), user LED (green)
• Power Management
  • Power Supply
    • 5V/2A via USB-C port (VUSB)
    • 5V to 18V DC up to 2A via VDC pin
    • Up to 4.2V/2A via 2-pin JST PH battery connector; BQ25628E battery charger
    • Maintained supply voltage (can be used to set MPP voltage)
  • Output
    • 3.3V up to 1A shared between board, 3V3 header pin and VSQT STEMMA QT connector
    • 3.3V to 4.2V up to 3A shared between board and VBAT header pin
    • 5V to 18 V up to 2A shared between board and VS header pin
    • TPS631013 3.3V buck-boost regulator
  • Monitoring
    • Supply – Current and voltage measurements, good supply detection
    • Battery – Voltage, current  (charge/discharge), and temperature measurements; charge estimation; health & cycle count estimation; time-to-empty and time-to-full estimation; low charge, high/low voltage alarm; MAX17260 fuel gauge with support for Li-Ion, LiPo, and LiFePO4 chemistries
  • Battery protection
    • Undervoltage Detect @ 2.2 V, Release @ 2.4 V
    • Overvoltage Detect @ 4.37 V, Release @ 4.28 V
    • Overcurrent protection @ 3A
    • Trickle charging safety timer @ 1 hr
    • Temperature-based charging current reduction based on JEITA, cutoff at 0°C and 50°C.
  • Misc – 3V3 enable/disable; VSQT enable/disable, FeatherWing enable/disable
  • Power States – Ship mode, Shutdown mode, and Power cycle
• Power consumption using BATP (at about 3.7V) vs V1
  • Deep-Sleep, Fuel Gauge Enabled – 24 μA  vs 26 μA (initial) and 18.5 μA (settled)
  • Deep-Sleep, Fuel Gauge Disabled – 19 μA vs 18 μA
  • Ship Mode, Fuel Gauge Disabled – 1 μA vs 1.5 μA
  • Shutdown Mode, Fuel Gauge Disabled – 1 μA vs 1.4 μA
• Dimensions – 65 x 23 x 7 mm (Adafruit Feather form factor, supports FeatherWings); 2x 2.5 mounting holes

Power consumption has changed due to the new regulator and fuel gauge, but it remains in the same range.

Alongside the hardware, the PowerFeather-SDK V2 has also been released for programming the board with the Arduino IDE or the ESP-IDF framework. The main change is adding support for LiFePO4, precisely for charger and fuel gauge configuration. Sample code reports battery and supply information, but doesn’t need to care about the battery type since it’s automatically detected by the SDK.  The documentation for the V1 and V2 boards looks fairly detailed, and includes guides to reduce power usage, use a solar panel, and create an ESPHome device.

While looking for alternative ESP32 boards with LiFePO4 battery support, I didn’t find any, but noticed several people trying to hook up this type of battery to their ESP32 boards, so it could be useful to some. Having said that, we did cover several Raspberry Pi solutions compatible with LiFePo4 battery, including the LiFePO4wered/Pi+ UPS HAT and, more recently, the AQEX qUPS-P-BC-2.0 UPS HAT, which works with a range of battery chemistries, including LFP.

The price of the ESP32-S3 PowerFeather V2 is the same as the V1, and the board goes for $30 on Elecrow. I was curious about the price of LiFePO4 batteries, and prices start at about $4 for a 3.2V/7000 mAh battery. It can go much higher depending on capacity and voltage. I also saw a 48V/300Ah (15kWh) LiFePO4 battery to power a full-house energy storage for close to $1800, and clearly not suitable for this little ESP32-S3 board….

The post ESP32-S3 PowerFeather V2 board gains support for LiFePO4/LFP batteries appeared first on CNX Software - Embedded Systems News.

Semtech FX86E is a 5G RedCap and 4G LTE cellular modem designed for industrial IoT applications such as rugged, long-range security cameras, heavy-duty vehicles, and various solar/battery-powered applications.

Based on the company’s EM8695 M.2 module powered by a Snapdragon X35 5G RedCap modem, the pre-certified modem supports public and private networks, boasts a MIL-STD-810H military-grade design, IP30 protection, E-Mark certification, and low power consumption.

Semtech FX86E specifications:

• SoC/Memory/Storage – TBD
• Cellular
  • 5G RedCap (Sub-6 Ghz) and LTE Cat-4
  • Frequency bands
    • 5G RedCap – 5G Sub-6 Ghz (FR1 1CC): n1, n2, n3, n5, n7, n8, n12, n13, n14, n18, n20, n25, n26, n28, n30, n38, n40, n41, n48, n66, n70, n71, n77, n78, n79
    • 4G LTE – LTE (Cat-4 DL / Cat-3 UL): B1, B2, B3, B4, B5, B7, B8, B12, B13, B14, B17, B18, B19, B20, B25, B26, B28, B30, B34, B38, B39, B40, B41, B42, B43, B48, B66, B70, B71, B106
  • Nano SIM (4FF) slot
  • Antenna – Optional 5G/4G adhesive mount puck antenna
  • Approvals – FCC, PTCRB, IC, GCF, RED, RCM (Planned)
  • Telecom & Carrier – Telstra, DoCoMo, SoftBank, KDDI, Anterix, AT&T, Verizon, T-Mobile
• Ethernet – 10/100Mbps Ethernet RJ45 port
• USB – Micro USB 2.0 port
• Misc
  • Signal LED, user LED
  • Mounting options – Bracket for screw/wall and DIN rail mounting
• Power Input – 4.75V to 32V DC
• Dimensions – 82 x 60 x 32 mm including connectors
• Weight – 158 grams
• Temperature  Range – -30°C to +75°C (MIL-STD-810H, test methods 501.7, 502.7)
• Humidity – 95% relative humidity over a temperature range of +20°C to +60°C  (MIL-STD-810H, test methods 507.7)
• IP Rating – IP30
• Vibration and Shock
  • MIL-STD-810H (test methods 503.7, 514.8, and 516.8)
  • SAEJ1455 (Sec 4.10.4.2 and 4.11.3.4)
• Vehicles certifications – E-Mark UN ECE Regulation No.10 Rev 7, CB IEC62368-1 and UL Listed SAEJ1455 (Sec 4.13.1, 4.13.2)
• ESD – 8KV contact discharge, 15KV air discharge
• Hazardous Locations – Class 1 Div 2 (C1D2); planned

On the software front, the system runs OpenWrt with an “intuitive, browser-based ” web UI interface for device configuration and network management. Semtech also mentions boot protection and firmware over-the-air (FOTA) updates via secure HTTPS. Documentation is limited (non-existent) for the modem itself, but you can find some resources for the EM8695 module itself. including AT command set, firmware, Windows drivers, etc…

Semtech/Sierra Wireless didn’t provide pricing information for the FX86E 5G Red Cap modem, but it should now be available, considering it was sampling during Embedded World 2026, last March. For reference, the EM8695 M.2 module alone is sold for $152.90 on Techsip. Additional information about the FX86E modem can be found on the product page and in the press release.

Thanks to TLS for the tip.

The post Semtech FX86E – A Compact 5G RedCap and 4G LTE modem for industrial IoT applications appeared first on CNX Software - Embedded Systems News.

Back in November 2024, we wrote about Qualcomm’s QCC730M and QCC74xM modules for low-power IoT devices, but at the time, there was no evaluation board EVK for those SoCs. Fast forward to today, and the official development boards for the QCC74x series have finally been released, complete with a surprisingly affordable price tag.

The Qualcomm QCC74xM EVK is a tri-radio evaluation board with Wi-Fi 6, Bluetooth 5.4, and IEEE 802.15.4 (Thread/Zigbee) with a price tag of just over $13. But what is interesting is that, while it carries the Qualcomm brand, documentation from the Zephyr Project clearly shows that the QCC74x series is actually based on and mostly equivalent to the Bouffalo Lab BL61x series, specifically the BL618.

Qualcomm QCC74xM EVK specifications:

• Wireless Module – Qualcomm QCC74xM LGA Module (QCC743M, QCC744M, or QCC748M depending on the board variant)
  • CPU – 32-bit RISC-V processor (RV32IMAFCP) @ up to 325 MHz with FPU and DSP
  • Memory and Storage
    • 484 KB SRAM (with 48 KB Cache) and 128 KB ROM on-chip
    • 8 MB stacked pSRAM (QCC748M only)
    • 16 MB stacked NOR flash
  • VPU – Motion JPEG (MJPEG) video encoding up to 720p (QCC748M only)
• Display – DBI via GPIO (QCC748M only)
• Camera – DVP via GPIO (QCC748M only)
• Audio
  • I2S (Master/Slave) interface
  • 1x DAC for speaker output and 1x ADC for MIC input via GPIO  (QCC748M only)
• Networking and wireless
  • 10/100Mbps Ethernet via GPIO (QCC744M & QCC748M only)
  • 2.4GHz 802.11 b/g/n/ax Wi-Fi 6 1×1 SISO
  • Bluetooth Low Energy 5.4 with coded PHY
  • 802.15.4 Thread and Zigbee
  • Support for Matter over Wi-Fi and Thread
  • PCB Antenna or an RF connector options
• USB
  • USB Type-C port (J1) for UART/serial console and programming
  • USB Type-C port (J9) for Direct USB/Power (QCC748M only)
• Other I/O
  • Dual-row pin headers for GPIOs (Up to 32x GPIOs on QCC748M; up to 19x on QCC743M)
  • 2x UART, 2x I2C, SPI (Master/Slave), XIP QSPI
  • 1x ISO 11898 CAN Bus (Automotive/Industrial)
  • General Purpose 12/14/16-bit ADC and 12-bit DAC
  • 1x PWM (4 channels), IR Controller (Receiver), Timers (RTC, 2x 32-bit, 1x 16-bit), and JTAG (via GPIO)
• Security
  • Integrated crypto acceleration with public key acceleration, TRNG, and QSPI (XIP) on-the-fly AES decryption
  • PSA Certified Level One featuring secure boot and secure debug
• Misc
  • BOOT button (S1)
  • RESET button (S2)
  • Power LED and VDD33 LED
  • 2x 2-pin header jumpers (J3, J4)
•  Power
  • EVK Input Power  – USB or 5V
  • Module Input Voltage – 2.97-3.63V
  • I/O Voltage – 1.8V/3.3V
  • 1.8V power measurement pad (J5)
  • 3.3V power measurement pad (J6)
• Dimension -TBD
• Operating Temperature – -40 to +85°C

After going through the specifications, it’s clear that it’s designed to directly compete with Espressif modules like the ESP32-S3 or ESP32-C6, featuring a 325 MHz RISC-V core with an FPU and DSP, as well as short-range wireless standards. The specific EVK available right now (the QCC748M) represents the higher end of the stack, with up to 8MB of integrated PSRAM and 16MB of Flash.

Also, from the Block diagram, it’s clear that the only clear difference between the three boards is in the GPIOs and on the USB port.

On the software side, the QCC74x series is designed to operate in a hostless mode, running both the protocol stack and IoT applications without an external MCU. The official software SDK (hosted on CodeLinaro) is built on FreeRTOS and is open-sourced on CodeLinaro, alongside a Microsoft Visual Studio Code (VS Code) extension for development. Due to its Bouffalo Lab roots, the hardware is also already supported by the Zephyr RTOS ecosystem. More information is available on Qualcomm’s quick start guide.

The Qualcomm QCC74xM EVKs are available on DigiKey for $13.12. This includes the QCC748M, QCC744M, and QCC743M development boards. The QCC748M EVK comes in two variants – the EVK-QCC748M-2-01-0-AA with a built-in PCB trace antenna, and the EVK-QCC748M-2-01-0-AB with an RF connector for an external antenna. The QCC744M and QCC743M EVKs are also available in the same two antenna configurations (PCB trace and RF connector variants). The modules themselves for under $3 to $4.75.

Via Hackaday

The post $13 Qualcomm QCC74xM EVK offers Wi-Fi 6, Bluetooth 5.4, and 802.15.4 connectivity (Bouffalo Lab BL618 inside) appeared first on CNX Software - Embedded Systems News.