Malcolm-Jamal Wilson has launched the WLV-01, a fully hackable DIY digital camera built around a Raspberry Pi 5. Designed for hardware tinkerers and photography enthusiasts, the camera features swappable full-spectrum color and true monochrome sensors ranging from 1/1.2-inch to Micro Four Thirds (M4/3), and offers a simple manual shooting experience.

The camera uses a manual Sony E-mount and supports various lens options through adapters, including C-mount, M-mount, M42, Nikon F, Pentax K, Canon FD, and Leica M. For the IMX585 model, the mounting plate supports both C- and E-mount lenses, while Fuji X and Micro Four Thirds mounts are available as optional add-ons. Other hardware features include a 4-inch DSI display, standard 18650 battery cells, a 3D-printed body, and support for focal reducers, helicoids, and speed boosters. The camera also supports full-spectrum photography with suitable IR-pass or hot-mirror filters, plus remote flash triggering through a 2.5mm mono jack.

WLV-01 specifications:

• SBC – Raspberry Pi 5 with 2GB RAM (optimized for thermals and cost)
• Storage – 128 GB internal microSD, expandable via USB stick
• Display – Waveshare 4-inch DSI touchscreen display (800 x 480 resolution)
• Camera system – Interchangeable sensor cartridges and an adaptable manual Sony E-mount, with various sensor options:
  • Raspberry Pi HQ (IMX477) – 12.3 MP, 1/2.3″ Color (visible-light only)
  • Sony IMX585 (Color / Mono) – 8.3 MP (up to 20 MP Burst SR), 1/1.2″ Full Spectrum
  • Sony IMX294 (Color) – 11 MP binned / 47 MP unbinned, Micro 4/3 Full Spectrum
  • Sony IMX492 (Mono) – 11 MP binned / 47 MP unbinned, Micro 4/3 Full Spectrum
• Networking – Built-in Wi-Fi for wireless file transfer and browser-based web interface (enables remote live preview, tethering, and LUT management)
• USB
  • 1x USB Type-C port
  • 1x USB Type-A port
• Misc
  • 2.5 mm mono jack for remote flash triggering
  • Standard 1/4-20 tripod mount
• Power
  • Geekworm X1200 battery HAT supports 18650 cells (battery not included)
  • USB Type-C charging
  • 2 to 3 hours of battery life per charge
• Dimensions – 113 x 84 x 74 mm (body only)
• Body – 3D-printed body built with Proto-pasta HTPLA

The camera supports simultaneous JPEG and RAW DNG still capture, as well as 1080p and 4K H.264 video recording at 24 fps. Its ISO range spans 100 to 6400, while manual focusing is assisted by focus peaking and a rangefinder mode. Additional features include 2x/4x tap-to-zoom, built-in film simulations, custom .cube LUT support, Brenizer panorama stitching, and in-camera double exposure.

On the software side, theWLV-01 runs a custom Python camera app on the Raspberry Pi 5 using Picamera2 and libcamera. It comes with preloaded software, but if you purchase the digital file, you will receive a flashable Raspberry Pi image and the app files. Since the software is Python-based, users can also modify it and add their own features. The camera’s built-in Wi-Fi also lets you access a web interface for live preview, image review, and file management. It also includes a gallery and supports loading standard .cube LUT files for live film simulations. The custom LUT editor specifically supports the conversion of Lumix V-Log LUT files for use directly on the camera.

The WLV-01 is marketed as a hackable, Python-based camera with no locked firmware, but it is not currently open source in the usual sense. There is no public GitHub repository, published license, or free download of its software or STL files. These files are provided with the camera, DIY kit, or a paid digital-files pack. You can modify the Python app and reprint the 3D-printed body, while the maker provides updates and support. So for now, it is better to describe the WLV-01 as a hackable camera, not a fully open-source project.

The WLV-01 is currently available on Kickstarter and has already exceeded its funding goal. Rewards start at CA$85 (~$62 US) for the Digital Files tier, while hardware kits start at CA$360 (~$259 US) for the IMX585 sensor cartridge. Fully assembled cameras range from CA$1,025 (~$737 US) for the IMX585 Color model to CA$1,795 (~$1,291 US) for the IMX492 Mono version. Digital files are expected in October 2026, with physical hardware deliveries planned for December 2026 and January 2027. More information about the WLV-01, including its setup and operation, can be found on the creator’s official website and in the user guide.

The post WLV-01 hackable DIY digital camera features Raspberry Pi 5, interchangeable sensors (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

BE-IIS (Brechel Electronic-Industrial Interface Systems) has released a dual GMSL2 Camera HAT for the Raspberry Pi 5. The board is based on the MAX96716A GMSL2 deserializer and connects two remote GMSL2 cameras through the Pi’s CSI-2 interfaces.

The solution has been validated in a dual-camera setup that operates at 2304 × 1296, RAW10, with two CSI-2 lanes per camera using the same Sony IMX708 12MP camera sensors found in the Raspberry Pi Camera Module 3. Besides low latency, one advantage of GMSL2 is long-range operation, and the HAT has been tested with two cameras connected over 10-meter coaxial cables.

BE-IIS BE-IIS-2CAM (aka BE-IIS-GMSL2-2CAM) HAT+ specifications:

• Dual MIPI CSI camera output compatible with Raspberry Pi CSI interfaces
• Dual GMSL2 camera input
  • 2x Rosenberger FAKRA connectors
  • Up to 6 Gbps each
  • PoC filter and protection per connector
• Analog Devices MAX96716A GMSL2 deserializer
• Misc
  • Single and Dual HAT Operation
  • 40-pin Raspberry Pi GPIO header
• Power Supply
  • Two options:
    • 5V via GPIO header
    • 5V via USB-C port
  • Multiple camera power supply voltages selectable by jumper: 5V, 9V, 12V, 16V, or 24V
• Dimensions – Raspberry Pi HAT+ form factor

On the other side, you’ll also need a MAX96717-based camera-side serializer board (BE-IIS-GMSL2-SER) that connects a standard CSI-2 camera module to the GMSL2 coaxial link and supports I2C control, camera reset/enable signals and Power over Coax. The platform is intended for GMSL2 evaluation, industrial vision, robotics, automotive camera prototyping, stereo vision, and remote camera applications.

Philipp Brechel told CNX Software that Linux integration is an important part of the project. The system runs on the Linux kernel from Raspberry Pi OS and doesn’t require rebuilding the full kernel or installing a custom kernel image. The required drivers are installed through a bash script. Once everything is set up, you can run standard tools such as rpicam-hello to use the cameras.

It’s not the first time the Raspberry Pi 5 has gotten a dual-channel GMSL camera adapter board, since Waveshare released the MAX9296-GMSL-DESER-MODULE board earlier this year. It basically offers the same functionality, but it’s based on the older MAX9296A deserializer, lacking features like tunnel mode, and the form factor is not quite as compact.  If one channel is enough, we also covered a GMSL2 extender solution for the Raspberry Pi SBCs based on two boards for compatibility with Raspberry Pi Camera Modules.

More information about the dual GMSL2 Camera HAT, including a PDF schematic, 3D files, and BoM, can be found on the product page. The BE-IIS-GMSL2-2CAM board is sold with accessories for $229 on Tindie or $300 on Lectronz (maybe with some EU taxes included). You may also need two BE-IIS GMSL2 serializer boards, sold for $79 each on Tindie or $100 on Lectronz.

The post Raspberry Pi 5 gets dual GMSL2 Camera HAT based on MAX96716A deserializer appeared first on CNX Software - Embedded Systems News.

ENIAC (operating as ENILINX) has launched the TICKEY, a 3.7-inch color e-Paper smart badge and desk display built around an ESP32-S3 MCU. The device features Wi-Fi, Bluetooth, and NFC connectivity, allowing users to update the screen from a smartphone without requiring continuous power to hold a static image.

The display uses Modos’ color e-paper technology to deliver 48+ native colors, with refresh times as fast as 3 seconds and sunlight readability. Measuring just 3.8mm thick and weighing 45 grams, TICKEY features an aluminum-alloy body and interchangeable faceplates, and supports NFC and QR code-based web profile sharing. The standard version includes 8MB of storage, whereas the Developer Edition offers 128MB for more advanced applications. These features make it suitable for desk reminders, event identification, digital business cards, and interactive maker projects.

TICKEY specifications:

• Wireless MCU –  ESP32-S3
  • Dual-core Tensilica LX7 microcontroller @ up to 240 MHz
  • Wireless – 2.4 GHz 802.11n WiFi 4 and Bluetooth 5.0 LE connectivity
• Storage – 8MB standard (128MB on the Developer Edition)
• Display – 3.7-inch full-color e-paper screen
  • 720 x 480 resolution
  • 48-color native E-Ink technology (driven by Modos)
  • Fast refresh in 3 seconds; Normal refresh in 13 seconds
• Connectivity
  • Wi-Fi and Bluetooth (BLE) via ESP32
  • NFC (ISO 15693) for sharing pages, profiles, or portfolios
• USB – USB Type-C port for power, programming, and recharging the battery
• Misc
  • 3x physical side buttons (Key Up, Key Down, Key Back)
  • Built-in clip ring hole
  • Compatible with magnets and desk stands
•  Power
  • 5V via USB Type-C port
  • 360 mAh battery for over 1,000 refreshes per charge (estimated at over two months of use as a daily charm)
• Dimensions – 111 x 62 x 3.8 mm
• Weight – 45 grams
• Enclosure – CNC aluminum case

The company mentions that the device is “open-source”, so you can easily write your own applications. ENILINX is also working on a platform where users can find, download, and install third-party firmware. They can switch back to the original firmware whenever they want. But at the time of writing, there is no Git repo or any other information, and like many other KickStarter products, they will be released when the campaign ends.

The device is controlled through the TICKEY companion app, with beta versions available for iOS and Android. The official TICKEY app page provides a direct download for the Android version, while the iOS version redirects users to Apple’s TestFlight for installation. Through the app, you can send photos, pixel art, weather information, or to-do lists to the badge. Built-in NFC lets you tap the badge with a smartphone to open web profiles, portfolios, or event pages.

ENILINX has also partnered with Modos, an open-source e-paper company and the maker of the Flow monitor, to offer a special Developer Edition of TICKEY. It uses the same hardware as the standard version but increases the storage from 8MB to 128MB for more complex applications and comes in a different color.

We have previously covered several e-paper badges and programmable displays on CNX Software, including the BeagleBadge, the Pimoroni Badger 2040, and M5Stack PaperColor, which use ESP32 and e-paper displays that can be easily turned into a badge. Compared to these, the TICKEY stands out with its slim 3.8mm design, 48-color display, and CNC aluminum body with swappable magnetic faceplates. In simple terms, it is a slim, colorful e-paper display with interchangeable faceplates.

TICKEY recently launched on Kickstarter and has already raised well over $600,000 US. The $30 Super Early Bird rewards are sold out, but the KS Special tier is available for HK$391 (about $50). It includes one TICKEY, a Kickstarter-exclusive faceplate and lanyard, a second selectable faceplate, and a quick-release keyring. Duo and Trio packs are also available. Shipping costs extra depending on the destination, such as $10 to India, Singapore, and Japan, or $15 to the US and Canada. Because of pressure on chip production, the company expects the first 10,000 units to ship in November and December 2026, though the schedule may change.

The post ENILINX TICKEY – An ESP32-S3-based color e-Paper badge with open firmware (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

AAEON MIX-PTLWV1 is an industrial mini-ITX motherboard based on Intel Core Ultra Series 3 “Panther Lake” processors with up to two 10GbE and two 2.5GbE RJ45 jacks, and four 4K-capable DisplayPort video outputs.

The motherboard also supports up to 128GB DDR5 C/SO-DIMM memory, M.2 NVMe SSD storage, optional WiFi and Bluetooth, and/or 4G LTE/5G cellular via M.2 modules, a PCIe Gen4 x8 slot for expansion, six serial interfaces, 12V-24V wide DC input, and more.

AAEON MIX-PTLWV1 specifications:

• SoC – Intel Panther Lake processor (PTL-H404/PTL-H484 families) with Intel Xe LPG graphics; TDP: 15 to 25W
• System Memory – Up to 128GB via 2x 262-pin DDR5 C/SO-DIMM sockets up to 6400MHz (H404 SKU) or 7200MHz (H484 SKU)
• Storage – M.2 2280 M-Key (PCIe Gen4 x4) socket for NVMe SSD
• Display Interfaces
  • 4x DisplayPort 1.4 up to 4096 x 2160 @ 60Hz
  • 40-pin LVDS/eDP co-layout connector for
    • Default: 18/24 bit dual-channel LVDS up to 1920 x 1080 @ 60Hz
    • 4-channel eDP1.4 up to 4096 x 2160 @ 120Hz (by BOM)
  • eDP/LVDS T-con IC voltage selection
  • eDP/LVDS panel backlight voltage select (3.3V/5V/12V)
  • LVDS Inverter
  • Quad independent display support
• Audio
  • Realtek ALC897 HD audio codec
  • Line-out jack
  • 4-pin header for 2W/4Ω speakers
• Networking
  • 2.5GbE RJ45 port via Intel I226-V 2.5Gbps Ethernet controller
  • 2.5GbE RJ45 port via Intel I226-LM 2.5Gbps Ethernet controller
  • SKU 4L only – 2x 10GbE via Intel E610-XAT2 10Gbps Ethernet controllers
  • Optional WiFi and Bluetooth via M.2 Key-E socket
  • Optional 4G LTE/5G cellular via M.2 Key-B socket and SIM card slot
• USB
  • 4x USB 3.2 Gen1 (5Gbps) Type-A ports
  • 2.0mm pitch 20-pin header with 2x USB 3.2 Gen 1 interfaces
• Serial
  • 1x RS-232/422/485 box header (5V/12V/RI optional)
  • 5x RS-232 box headers
• Expansion
  • PCIe Gen 5 x8 open edge slot (BoM optional on H484)
  • M.2 2230 E-Key socket (PCIe 4.0 x1/USB 2.0/CNVi)
  • M.2 3042/3052 B-Key socket (PCIe x1/USB 3.2 Gen 1) + Nano SIM Socket; default 3052 by BOM (optional)
  • 8-bit programmable DIO
• Security – Infineon SLB9672 TPM2.0 chip
• Misc
  • Watchdog Timer
  • AMI UEFI/BIOS on 256Mbit SPI Flash ROM
  • H/W Monitor – Temperature monitor on CPU/System, voltage monitor on Vcore/5V/3.3V/12V, fan monitor on chassis
  • I/O Chipset –Nuvoton NCT6126D
  • 4-pin CPU fan connector, 4-pin SYS fan connector
  • Wake-on-LAN (WoL) and PXE  support
  • Power State S4, S5
  • Lithium battery for BIOS
  • CMOS jumper
  • Front panel header
  • ME disable
• Power Supply – 12 to 24V DC input via 4-pin connector; ATX/AT mode
• Dimensions – 170 x 170 mm (mini-ITX form factor)
• Weight – TBC
• Temperature Range – Operating: 0 to 60°C; storage: -40 to 85°C
• Operating Humidity – 60°C @ 90% relative humidity, non-condensing
• Certifications – CE & FCC (Class A)

AAEON supports Windows 11 64-bit and Ubuntu 24.04 with Linux kernel 6.8 for the board. Note that there are two SKUs: MIX-PTLWV1-A10-4L and MIX-PTLWV1-A10-2L, with the only difference being the two 10GbE RJ45 ports on the 4L model.

While the MIX-PTLWV1 is AAEON’s first industrial motherboard powered by the Intel Core Ultra Series 3 “Panther Lake” processor family, Avalue previously introduced the EMX-PTLP thin mini-ITX motherboard with up to an Intel Core Ultra 7 358H SoC, and BCM Advanced Research is working on the MX-PTL motherboard with similar features.

The MIX-PTLWV1 motherboard sells for $951 on AAEON’s eShop. That’s specifically for the MIX-PTLWV1-A10-325-2L variant with an Intel Core Ultra 5 325 CPU and two 2.5GbE ports. The product page has more details, including download links to the datasheet, BIOS, and drivers.

Update: The article was first published on May 28, 2026, and updated following availability on the company’s eShop.

The post AAEON MIX-PTLWV1 Panther Lake industrial mini-ITX motherboard offers quad DisplayPort, dual 10GbE, and more appeared first on CNX Software - Embedded Systems News.

OpenMANET is an open-source project for building long-range wireless mesh Mobile Ad-hoc Networks (MANETs) using Raspberry Pi SBCs and Wi-Fi HaLow (802.11ah) chips from Morse Micro.

This came up when we wrote about Morse Micro USB WiFi HaLow adapters, and an X user commented that “OpenMANET is going to explode with these available, no more adapter boards or M.2 cards”. OpenMANET targets search and rescue, disaster response, airsoft events, and other off-grid or infrastructure-failed situations. So it looks like an alternative to LoRa-based Meshtastic and Meshcore, with higher bandwidth, at the cost of more expensive hardware and higher power consumption.

OpenMANET supports field communications, including push-to-talk, sharing GPS positions, cameras/video streams, and ATAK (Android Team Awareness Kit) integration. The mesh network can access the Internet through a single “mesh gate” over Starlink or cellular connectivity.

The project supports a range of hardware platforms including Gateworks Venice, Raspberry Pi  Zero 2 W, Raspberry Pi 3 Model B, HaLowLink 2, and Heltec HT-HD01 V2, but for the best experience, a Raspberry Pi 4 kit is recommended since it supports the core network, off-grid communication, the camera, BLOS (Beyond Line of Sight) to securely merge two MANETs in different locations over the Internet, and the new user interface.

While Meshtastic and OpenMANET both target off-grid comms, search-and-rescue, field operations, and events, Meshtastic shines for text messages, GPS location sharing, and lightweight sensor data at ultra-low power. OpenMANET delivers higher bandwidth, 1 to 15 Mbps depending on the country of operation, offers full IP network support, built-in PTT voice, RTSP streaming from a camera, and ATAK integration. The main downside is that OpenMANET requires more powerful, power-hungry, and expensive hardware than Meshtastic or MeshCore.

The OpenMANET firmware is based on OpenWrt. The OpenMANETd daemon, written in Go, provides mesh management, an embedded React web interface, and a gRPC/HTTP API. Under the hood, OpenMANET relies on the 802.11s Wi-Fi mesh standard and the batman-adv (Better Approach To Mobile Adhoc Networking – advanced) routing protocol sitting on top of 802.11s.

You’ll find more details on the project’s website, although it mainly points to the project’s GitHub account, the documentation website, and a library of enclosures (also on GitHub).

The post OpenMANET – A higher-bandwidth Meshtastic alternative leveraging 802.11ah WiFi HaLow appeared first on CNX Software - Embedded Systems News.

The AIMORELOGY Ovis is an open-source AI vision camera module built around the CVITEK CV1842H-P SoC, with full-color 1080p night vision and 1.5 TOPS edge AI inference in a compact modular design. It is designed for drones, robotics, security systems, smart cameras, and custom embedded vision products.

The camera features a compact stacked design, with a 20 × 20 mm Core board that includes the CVITEK CV1842H-P SoC, 2 Gbit NAND flash, USB, and UART debug pads. A Sensor board with the SC235HAI image sensor connects on top and also adds Ethernet and UART interfaces. An optional CVBS board goes between the Sensor board and the Ovis Core board.

AIMORELOGY Ovis specifications:

• Ovis Core Board
  • SoC – CVITEK CV1842H-P
    • CPU – 1x Arm Cortex-A53 core @ 1.1 GHz, 1x RISC-V C906 core @ 800 MHz
    • NPU – 1.5 TOPS @ INT8 with BF16 support
    • ISP – AI-ISP with real-time 1080p @ 30 fps full-color night vision support (AI-BNR)
    • VPU – H.265, H.264, and MJPEG encoding up to 8 MP @ 25 fps
    • Memory – 2 Gb (256 MB) DDR3 System-in-Package (SiP)
  • Storage – 2 Gb (256 MB) SPI NAND flash
  • USB – 1x USB 2.0 via 4-pin SH0.8 connector (supports UVC video and USB NCM virtual Ethernet)
  • Debugging – Dedicated UART debug test pads
  • Daughterboard interface – 2x 50-pin board-to-board connectors
  • Dimensions – 20 x 20 x 5.7 mm
• Sensor Board
  • Image Sensor – SmartSens SC235HAI CMOS sensor with integrated lens mount
  • Networking – 10/100Mbps Ethernet via 4-pin SH1.0 connector
  • Serial – 1x UART interface connector
  • Interconnect – 2x 50-pin board-to-board connectors
  • Dimensions –  28 x 28 x 31.7 mm (Core + Sensor board dimensions with lens)
• CVBS Output Board (Optional)
  • Video Converter – MacroSilicon MS7024 video encoder chip
  • Video Output – Analog Composite Video (CVBS) connector
  • Serial – 1x UART interface connector
  • Interconnect – 2x 50-pin board-to-board connectors (top and bottom pass-through for sandwich stacking)
  • Dimensions – 20 x 20 x TBD mm (thickness of the board is not specified)

On the software side, the Ovis runs a dual-OS environment with Linux 5.10 on the Cortex-A53 core and RT-Thread on the RISC-V C906 core, booting from 256 MB of SPI NAND flash. Video output is mutually exclusive: you can choose either UVC mode, which works like a USB webcam, or RTSP streaming over USB NCM, but not both at the same time. The company also offers OVIS Web, an open-source browser interface that connects to REST APIs via Chromium-based browsers (supporting WebUSB) to tune the AI-ISP pipeline (including AI-BNR), configure OSD elements, and toggle inference pipelines. In simpler terms, you can easily switch between AI-ISP night vision mode and edge AI inference mode, adjust video output settings, and configure recognition workflows without writing a single line of code.

The Ovis supports built-in AI functions such as human and vehicle detection, face detection, pose recognition, and single-object tracking (SOT). It also supports custom AI models compiled into SOPHGO/CVITEK BModel files, with documentation currently available for YOLOv8n. For software development, the vendor provides the CVITEK-based aimorelogy-ovis-sdk on GitHub, which includes U-Boot 2021.10, cvi_mpi, hardware image operators (ive), the libsophon TPU runtime, and a Docker build environment (aimorelogy/ovis-build-docker). Additional software-related information can be found in AIMORELOGY’s documentation website.

One of the most interesting parts of AIMORELOGY’s Ovis is the SOPHGO/CVITEK CV1842H-P SoC. It features an Arm Cortex-A53 core at up to 1.1 GHz for Linux and applications, and a T-Head C906 RISC-V core at up to 800 MHz for lighter control work, with 2 Gbit (256 MB) DDR3 stacked in-package, so the module doesn’t need external DRAM.

The SoC also integrates a 1.5 TOPS INT8 TPU with BF16 support and an ISP V4.0 that handles up to 4K image processing, including 3D noise reduction, dynamic range control, purple fringe removal, dehaze, and lens distortion correction. Its video engine supports H.265, H.264, and MJPEG encoding at up to 8 MP @ 25fps, along with two 720p streams. Interfaces include up to four-lane MIPI CSI, MIPI DSI, 10/100 Ethernet, USB 2.0, SD 3.0, I2C, SPI, UART, and GPIO.

The CV1842H-P uses the same 10×10 mm BGA pinout as CV1812H-class chips. If you already designed a custom camera board for those parts, you can swap in the CV1842H-P without redrawing the chip footprint, and you stay on the same SOPHGO/CVITEK SDK. But that does not apply to Milk-V Duo or Duo 256M. Those boards use different chips and packages (CV1800B and SG2002), so you cannot drop a CV1842H-P onto a Duo PCB. You can reuse the Duo software/SDK, but not the board layout.

The AIMORELOGY Ovis is live on Kickstarter. The Standard Kit, which includes the Ovis Core, Sensor Board, and a 4-pin USB cable, requires a pledge of HK$460 (approximately $59). For those needing analog video output, the CVBS Kit adds the MS7024-based output board for about $69 US. Shipping is expected to begin in December 2026. More information about the AIMORELOGY Ovis camera module is available on the product page.

The post CVITEK CV1842H-P-based edge AI camera module offers night vision and AI-ISP support (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Alif Semiconductor’s Ensemble E1C (SK-E1C) and Balletto B1 (SK-B1) StartKits are low-cost development boards built around the company’s Cortex-M55 + Ethos-U55 microcontrollers. The Balletto B1 adds BLE 5.3 and 802.15.4 wireless connectivity, while the Ensemble E1C does not have built-in wireless connectivity.

Both StartKits feature a microcontroller with a 160MHz Cortex-M55 CPU, an Ethos-U55 NPU, 2MB SRAM, and 1.8/1.9MB NVM. They also include an Arducam camera header, a mikroBUS Click Board expansion socket, and two PDM microphones for camera, audio, and sensor-based projects. The Balletto B1 additionally provides the complete RF path to a chip antenna for its wireless subsystem, while both boards feature an on-board Segger J-Link debugger for development and testing.

Alif Semiconductor StartKit (SK-B1 and SK-E1C) specifications:

• MCU
  • SK-E1C – Alif Semiconductor Ensemble E1C 
  • SK-B1 – Alif Semiconductor Balletto B1 
  • CPU – Arm Cortex-M55 core with Helium-M Vector Extension @ up to 160 MHz
  • NPU – Arm Ethos-U55 neural network processor core (128 MACs per cycle)
  • Memory & Storage: 2MB of SRAM and 1.8MB of NVM  (1.9 MB MRAM on E1C)
• Camera Interface – 20-pin Arducam header
• Audio – 2x PDM microphones
• Wireless – BLE 5.3 and 802.15.4 radio subsystem with chip antenna (Balletto B1 StartKit only)
• USB
  • 1x USB Type-C port connected to the SoC
  • 1x USB Type-C port for programming and debugging
• Debugging
  • On-board Segger J-Link ((JL-OB E1) debugger with an Alif E-Series BGA194 MCU
  • Selectable UART routing to the J-Link debugger using JP5 jumpers (SEUART or UART2).
• Expansion
  • mikroBUS Click Board expansion socket
  • GPIO expansion headers for SPI, I2C, UART, PWM, and standard I/O
• Misc
  • User switch and a dedicated hardware Reset switch
  • Green Power LED and Green User LED
• Power
  • 5V DC input via USB Type-C ports
  • On-board TLV75201 3.3V LDO regulator
  • Flex I/O voltage selection jumper (JP3) to configure logic levels for either 1.8V or 3.3V
  • Current measurement jumpers for monitoring power consumption on VDD_MAIN and VDDIO_1V8.
• Dimensions – 86.37 mm x 71.09 mm

Both boards support power consumption testing. According to the schematics, you can physically cut trace shunts (like S18 or S19) on the PCB to insert ammeters to measure current across different power domains, helping evaluate power usage during AI and I/O workloads.

The company mentions that before starting software development, you need to connect the board to a computer through the PRG USB port and use the Alif Security Toolkit (SETOOLS) v1.112.00 or later. SETOOLS communicates with the Secure Enclave through SEUART, programs the on-chip MRAM, and configures the correct target device. Since the tools default to the Ensemble E8 from version 1.107.00 onward, users of the SK-E1C or SK-B1 must select E1C or B1 as the target device.

For the Ensemble E1C StartKit (SK-E1C), the company recommends Visual Studio Code with the Ensemble CMSIS Device Family Pack. The company also provides a VS Code template in its GitHub repository that integrates smoothly with Segger’s J-Link GDB Server.

For the Balletto B1 StartKit (SK-B1), developers working with BLE can use the Zephyr RTOS and Alif Zephyr SDK. Alif’s Balletto App for iOS and Android connects to the factory BLE Blinky demo and other sample profiles. Because the B1 shares the E1C’s digital MCU architecture, developers can also use the same CMSIS, VS Code, and bare-metal development flow for applications that don’t require wireless functionality.

Previously, we have written about Alif Semi’s higher-end Ensemble E4, E6, and E8 SoCs, which feature Ethos-U85 NPUs for Small Language Models (SLMs), as well as the Ensemble E3 dual-core, dual-NPU MCU used in OpenMV AI camera boards. The E1C and B1 chips feature smaller, more affordable options, with packages as small as 3.9 x 3.9 mm and dedicated neural processors for edge AI applications.

Both the Balletto B1 (SK-B1) and Ensemble E1C (SK-E1C) StartKits are priced at around $50, with Mouser listing each at $49.99, and DigiKey offering the SK-B1 for $51.28. Stock is currently unavailable at major distributors, but Mouser expects its next batch to arrive on October 22, 2026.  More information, including StartKit’s specifications, purchase links, user guides, software tools, BLE demo, schematics, PCB layout guidelines, Altium files, BOM, and datasheet, is available on their respective product pages.

The post Alif Semi Balletto B1 and Ensemble E1C StartKits feature Cortex-M55 + Ethos-U55 MCU for IoT and Edge AI appeared first on CNX Software - Embedded Systems News.

Finished your PCB layout in KiCad? You don’t necessarily need to leave the PCB Editor, export a separate set of manufacturing files, and manually upload them to a PCB manufacturer. AIVON’s KiCad Plugin creates a direct path from your PCB design to an online PCB quote.

AIVON has integrated an official KiCad plugin that connects a finished PCB layout to the company’s online quotation page. After DRC, users can click the plugin in the PCB Editor, automatically prepare manufacturing data, and land on a pre-filled quote form instead of exporting, zipping, and uploading files by hand.

Rather than introducing a new plugin framework, AIVON’s integration leverages KiCad’s existing Plugin and Content Manager ecosystem. For developers who already use KiCad for schematic capture and PCB layout, the integration provides another way to connect the design stage with PCB fabrication without manually repeating every manufacturing-file preparation step.

Connecting KiCad PCB Design with Manufacturing

For most hardware teams, the design is not finished when the last trace is routed. A typical prototype cycle still looks like this:

• Export Gerbers for copper, solder mask, silkscreen, and outline layers
• Generate Excellon or NC drill files
• Export a BOM and pick-and-place file if assembly is required
• Zip the package, upload it to a manufacturer portal
• Re-enter board size, layer count, thickness, copper weight, surface finish, and quantity

Those steps are easy to get wrong. A missing layer, an outdated Gerber set from a previous revision, or a drill file that does not match the current board outline can delay engineering review or produce a board that does not match the latest layout.

AIVON’s KiCad integration is designed to shorten this hand-off. The plugin connects the KiCad PCB Editor with AIVON’s online PCB quotation workflow. After the design has been checked, users can use the plugin to prepare the manufacturing data and continue to AIVON’s quotation interface.

The workflow can be summarized as:

KiCad PCB Design → DRC → AIVON Plugin → Manufacturing Data → Online Quotation

The goal is not to eliminate the need for engineering review. DRC, board-outline verification, and design-for-manufacturing checks remain important before a design is submitted for production.

A Shorter Path from Prototype Design to PCB Order

The integration can be particularly useful for hardware developers who frequently build and revise PCB prototypes.

A small change to a board layout can trigger another cycle of manufacturing-file generation and uploading. When these steps are repeated across multiple revisions, even relatively simple manual tasks can add friction to the development process.

By connecting the PCB Editor with the manufacturing quotation workflow, AIVON aims to reduce the number of context switches required after a design is ready.

For engineers interested in trying the workflow, AIVON provides a step-by-step guide explaining how to install and use the KiCad plugin, including installation through KiCad’s Plugin and Content Manager, preparing a board for quotation, and reviewing manufacturing options.

Supporting PCB Fabrication and Assembly Workflows

The integration is not limited to bare PCB fabrication. PCB assembly projects introduce additional manufacturing data, including component lists and placement information. Keeping these files connected to the same board revision can help reduce inconsistencies between the PCB design and assembly process.

AIVON provides both PCB fabrication and PCBA services, allowing the workflow to extend from prototype PCB production toward assembled boards when a project requires it.

For users placing their first order with AIVON, the Welcome Credits program provides up to $60 in new-user support, including a $30 manufacturing credit and up to $30 in shipping support. The new customer offer is designed to make it easier for engineers and makers to move from a completed PCB design to their first physical prototype.

For makers and engineering teams working on IoT devices, embedded systems, industrial electronics, or other hardware projects, this creates a more continuous path from an open-source EDA environment to physical hardware.

Getting Started with the AIVON KiCad Plugin

Engineers who want to test the workflow can install the AIVON plugin through the supported KiCad plugin process, complete their PCB design and DRC checks, and then use the plugin to prepare the manufacturing data and access the quotation workflow. The plugin is free to install, while PCB fabrication and assembly services are ordered separately through AIVON.

For users already working with KiCad, the integration provides a practical way to connect an open-source PCB design workflow with a commercial manufacturing service without requiring changes to the KiCad design process itself.

The post Design in KiCad, Quote with AIVON: A Faster Path to PCB Prototyping (Sponsored) appeared first on CNX Software - Embedded Systems News.

Last year, we saw Morse Micro introduce the MM8108 Wi-Fi HaLow (802.11ah) SoC with up to 43.33 Mbps throughput, along with an MM8108-RD09 USB dongle reference design, for which we had virtually no details at the time.

Morse Micro has now released USB-A and USB-C Wi-Fi HaLow adapters based on that reference design. The adapters let PCs and routers connect to 802.11ah networks through USB without a dedicated expansion card and feature an external SMA antenna delivering up to 26 dBm transmit power.

Wi-Fi HaLow USB network adapter specifications:

• SoC – Morse Micro MM8108 32-bit RISC-V Host Applications Processor (HAP)
  • Standard – IEEE802.11ah Wi-Fi HaLow
  • Frequency band – 850-950 MHz (Worldwide sub-1 GHz license-exempt range)
  • Operating Modes – Access Point (AP) and Station (STA).
  • Channel Width – 1/2/4/8 MHz
  • Modulation – BPSK, QPSK, 16QAM, 64QAM, 256QAM.
  • Data rate – Up to 43.3 Mbps using 256-QAM modulation at an 8 MHz bandwidth
  • Transmit power – Up to 26 dBm
• Antenna – SMA connector on the back of the dongle (includes an external Wi-Fi HaLow antenna in the box)
• USB – USB-A or USB-C port
• Security
  • AES
  • SHA-256 / SHA-384 / SHA-512
  • WPA3
  • OWE (Opportunistic Wireless Encryption)
• Power – 5V via USB Type-A or Type-C port
• Dimensions – TBD
• Operating Temperature –  -40°C to +85°C (silicon/chip — not the enclosed stick)
• Supported Regions – USA, Canada, Australia, UK, Europe, and Japan

Morse Micro also released the hardware design and manufacturing files for the MM8108-RD09 reference design. The company says the USB adapters support Windows, Linux, and macOS. However, I was unable to find the “pre-written driver” for Windows. Linux users need to compile the driver against a patched kernel with S1G (802.11ah) support or use Morse Micro’s OpenWrt fork. The morse_driver and firmware are available on GitHub, along with documentation and resources. More information, including software releases, datasheets, design guides, application notes, and user guides, can be found on the Morse Micro documentation portal.

The MM8108 chipset has already been used in modules such as the Morse Micro MM8108-M20 and Quectel FGH200M. This year, the company also released the HaLowLink 2 Wi-Fi HaLow access point based on that same MM8108 HaLow chipset.

The Morse Micro Wi-Fi HaLow USB network adapters are currently sold on Mouser for $58.00 in both the USB Type-A and Type-C configurations, with AU variants also available. Additional information is available on the product page.

Via Hackster.io

The post Morse Micro Wi-Fi HaLow USB adapters add 802.11ah connectivity to PCs and routers appeared first on CNX Software - Embedded Systems News.

JetKVM Mini is a smaller, lower-cost version of the popular JetKVM KVM over IP solution. The new model features a small information display, offers Ethernet or WiFi connectivity, and supports 1080p video capture.

The main under-the-hood change is that they ditched Linux on a Rockchip RV1106G3 SoC and now rely on an ESP32-P4X (ESP32-P4 version 3.x and greater) to handle video capture, H.264 encoding, USB, and other features. The largest cost saving is removing the RAM and eMMC flash required by Linux.

JetKVM Mini specifications:

• SoC – Espressif Systems ESP32-P4X
  • CPU
    • Dual-core 32-bit RISC-V HP (High-performance) CPU @ up to 400 MHz with AI instruction extension and single-precision FPU
    • Single-RISC-V LP (Low-power) MCU core @ up to 40 MHz
  • Memory
    • 768 KB HP L2MEM (for dual-core CPU), 32 KB LP SRAM, 8 KB TCM (for LP MCU core)
    • 32 MB PSRAM
  • Storage – 128 KB HP ROM, 16 KB LP ROM
  • GPU – 2D Pixel Processing Accelerator (PPA)
  • VPU – H.264 and JPEG codec support with Pixel Processing Accelerator (PPA) and 2D DMA controller
• Storage
  • 16 MB flash
  • MicroSD card slot for virtual media
• Display – IPS display with 240 x 240 resolution; non-touch, size not mentioned, but possibly 1.54-inch
• Video
  • HDMI input port up to 720p60 or 1080p30; 4K is also supported with JetKVM OS Services
  • H. 264-encoded video streamed over WebRTC to the host
• Networking
  • Mini variant – 100Mbps Ethernet RJ45 jack
  • Mini W variant – Dual-band WiFi 6, Bluetooth (for onboarding), and 802.15.4 for Zigbee and Thread (unused?) via ESP32-C5 module
• USB
  • USB 2.0 Type-C port (480 Mbps) for mouse, keyboard, and virtual media emulation, as well as power
  • USB 2.0 Type-C port (12 Mbps) for secondary power and future expansion (USB-C version of the current RJ11-based ATX, DC power, and serial extensions)
• Power Supply – 5V/500mA USB Type-C port
• Dimensions – 42 × 42 × 23 mm (Aluminum enclosure)

Switching from Linux to the FreeRTOS-based ESP-IDF framework obviously required a complete rewrite of the firmware for the ESP32-P4X. But the user experience will be the same as the original JetKVM device, using the same protocols to integrate with the same web interface and cloud. The firmware will be open-sourced and released on the company’s GitHub account once the JetKVM Mini is closer to release.

On compatible hosts (Windows, macOS, and Linux), the JetKVM OS Services will enable capture of the target computer’s screen at up to 4K, and add a shared clipboard, a guest terminal, and file transfer support. 4K resolution is possible since video encoding happens on the host rather than the JetKVM Mini.

It’s not the first time we’ve seen an ESP32-P4 IP KVM, and ESP-KVM and P4 KVM projects did that with off-the-shelf hardware, but with ESP32-P4 instead of the newer ESP32-P4X. These are limited to MJPEG or really slow H.264 since an additional color space conversion step is required, and is not needed in the new ESP32-P4X.

You can’t buy the JetKVM Mini today, but you can register your interest, and pricing has already been announced:

• JetKVM Mini – $39 for one, $33 per unit in a bundle of three
• JetKVM Mini W – $42 for one, $36 per unit in a bundle of three

The new matchbox-sized KVM over IP device will become available on October 26, 2026.

The post JetKVM Mini – A low-cost, ESP32-P4X-based KVM with Ethernet or WiFi connectivity appeared first on CNX Software - Embedded Systems News.

After launching the SX1262-based M5Stamp C6LoRa tiny module, M5Stack has launched two SX1262-based LoRa add-ons for the 868–923 MHz band: the tiny Stamp LoRa-1262 SMD module for custom PCBs, and the stackable Module13.2 LoRa-1262 module for the M5Stack Core series.

The Stamp LoRa-1262 supports LoRa, FSK, GFSK, MSK, GMSK, and OOK modulation, with up to +22 dBm transmit power and -147 dBm receive sensitivity. The Module13.2 adds an RP-SMA antenna interface, an M5IOE1 IO expansion chip for reset and power management, and DIP switches for flexible pin routing and configurable I2C addresses, enabling multi-module stacking.

Stamp LoRa-1262 SMD module

M5Stack Stamp LoRa-1262 specifications:

• LoRa Transceiver – Semtech SX1262
• Wireless Connectivity (LoRa)
  • Frequency Band – 868 to 923 MHz
  • Modulation Modes – LoRa, FSK, GFSK, MSK, GMSK, OOK
  • Bitrate – Up to 300 kbps (programmable)
  • Transmit power (Tx) – Up to +22 dBm
  • Receive sensitivity (Rx) – Down to -147 dBm (in LoRa low-rate mode)
• Antenna Options:
  • Stamp LoRa-1262 (S014) – Onboard RF pad (antenna trace connected by default)
  • Stamp LoRa-1262 I (S014-I) – IPEX-4 connector (onboard RF pad in NC state)
  • Stamp LoRa-1262 IF (S014-IF) – IPEX-4 connector
  • Note: “I” and “IF” models include a 108mm 868 MHz RP-SMA rubber duck antenna (3dBi gain) and a 40mm RP-SMA to IPEX-4 extension cable
• Host interface
  • SPI (Serial Peripheral Interface)
  • 12-pin 0.5mm pitch FPC connector for I/O breakout (IF only)
• Power – 3.3V @ 3.03 mA (working power consumption)
• Dimensions and Weight
  • Stamp LoRa-1262 and 1262 I – 14.0 x 10.0 x 2.4 mm | 0.7 grams
  • Stamp LoRa-1262 IF – 14.0 x 10.0 x 3.6 mm | 0.8 grams

From the specification above, you can see the Stamp LoRa-1262 is available in three variants. The standard model uses an onboard RF pad; the Stamp LoRa-1262 I adds an IPEX-4 antenna connector; and the Stamp LoRa-1262 IF includes both the IPEX-4 connector and a 12-pin 0.5mm FPC connector on the bottom to expose the module’s communication pins.

Module13.2 LoRa-1262 module

For those who want to quickly evaluate the new LoRa-1262 module, the Module13.2 LoRa-1262 integrates the Stamp LoRa-1262 into a stackable expansion board for M5Stack controllers such as the M5Stack CoreS3, Core2, and Core MP135.

Module13.2 LoRa-1262 specifications:

• Wireless module – M5Stack Stamp LoRa-1262 as described above
• Antenna – RP-SMA female connector; includes 108mm, 3dBi 868 MHz RP-SMA rubber antenna
• Expansion
  • SPI via M5-Bus
  • M5IOE1 IO expansion chip via I2C (address 0x71 to 0x74 configurable via DIP switch)
    • Handles LoRa module reset, bypass control, and 3.3V power management
• Misc – Configurable pins (BUSY, NSS, IRQ) via DIP switches for multi-module setups
• Power
  • Supply
    • 9V to 24V DC input via 5.5 x 2.1mm barrel jack (center positive)
    • 3.7V battery input via 1.25mm 2-pin connector (routes to M5-Bus BAT pin for host boosting)
  • Consumption
    • Transmit (Max) – 277.45 mA @ 4.2V | 137.10 mA @ 5V | 71.38 mA @ 12V
    • Receive – 174.52 mA @ 4.2V | 9.94 mA @ 5V | 4.5 mA @ 12V
    • Standby/Sleep – 79.4 µA @ 4.2V | 693.03 µA @ 5V | 1.06 mA @ 12V
    • Power Off – 2.76 µA @ 5V
• Dimensions – 59.1 x 54.0 x 20.2 mm
• Weight – 23.5 grams (excluding antenna)

The company notes: Before using the LoRa module, connect a matched external antenna first. Do not transmit RF signals without an antenna connected; otherwise, it will cause permanent hardware damage.

Both modules communicate with a host controller over SPI and support direct control of the SX1262 transceiver. M5Stack provides documentation for both the Stamp LoRa-1262 and Module13.2 LoRa-1262 modules, including details on configuration and packet transmission and reception. The Module13.2 additionally uses the M5IOE1 chip for reset, RF bypass, and power management over I2C, while the Stamp is a standalone transceiver module for integration with a host MCU.

The M5Stamp LoRa-1262 is available for $5.50 to $7.95 on AliExpress and the M5Stack store, while the stackable Module13.2 LoRa-1262 retails for $22 on AliExpress and the M5Stack store.

The post M5Stack Module13.2 LoRa-1262 expansion board integrates 14×10 mm Stamp LoRa-1262 module appeared first on CNX Software - Embedded Systems News.

CamThink NeoEyes NE302 is a tiny WiFi 6 Edge AI camera based on an STM32N6 Arm Cortex-M55 MCU with Neural-ART NPU which the company says is “especially suitable for developers and makers”.

I initially thought about it as a smaller version of the CamThink NeoEyes NE301 based on the same STM32N6 MCU and 4MP OS04C10 camera sensor. But it’s quite a different device. First, it doesn’t have a battery, and the USB-C is only used for power. Memory and storage capacities have been reduced to 32 MB PSRAM and 64 MP SPI flash, and a range of built-in and optional features have been removed, including 4G LTE module (global or US), audio wafers, PIR motion connector, and a 16-pin GPIO header, as well as support for PoE power.

CamThink NeoEyes NE302 specifications:

• Vision MCU
  • STMicro STM32N6
  • MCU Core – Arm 32-bit Cortex-M55 CPU @ up to 800MHz with Arm Helium and Arm MVE
  • GPU – Neo-Chrom 2.5D GPU, Chrom-ART Accelerator (DMA2D)
  • NPU – ST Neural-ART accelerator @ 1 GHz, up to 600 GOPS; 3 TOPS/W enabling fanless operation
  • BPU – Hardware-accelerated H.264 and JPEG encoders up to 1080p @ 30fps
  • ISP – ISP with demosaicing, AWB
  • Memory – 4.2 MB SRAM
• Memory – 32 MB PSRAM
• Storage
  • 64 MB SPI Flash
  • MicroSD card slot on the interface board
• Camera
  • MIPI CSI-2 connector fitted with a 4MP OS04C10 camera sensor with 88° (wide-angle), or 137° (ultra-wide angle) field of view
  • Lens mount – Standard M12
  • White LED fill light
• Wireless
  • 2.4 GHz WiFi 6 and Bluetooth 5.3 LE
  • External SMA antenna with 3 to 4 dBi gain
• Sensor – SHT31-DIS temperature and humidity sensor on the interface board
• Debugging – Separate N6-STLINK, U0-STLINK and U6-UART paths on the interface board
• Misc
  • Trigger and Reset buttons
  • Dual-color indicators
  • 4-pin Alarm I/O interface
• Power Supply
  • 5V via USB-C port
  • STM32U073K8U6 MCU for power, wake, and control paths
• Dimensions
  • Main PCB – 38 x 38 mm
  • Device – 42 x 42 x 20 mm
• Temperature Range – -20°C ~ 50°C

While many hardware features have been removed, the NeoEyes N302 relies on the same firmware base as the NeoEyes N301. It supports NPU-accelerated object detection & pose estimation, MIPI CSI-2 → ISP → H.264 / MJPEG hardware-encoding video pipeline, a Web UI for real-time preview and running models, multiple network protocols (HTTP REST API, WebSocket, MQTT, RTSP, RTMP), OTA updates, secure boot, and the STM32U073 for power management.

You’ll find more technical details and quick start, user, software, and hardware guides on the documentation website. The company has also released PDF schematics and layouts, a BSP, the main application, middleware, drivers, the STM32U0 “WakeCore” firmware, and more on GitHub.

The CamThink NeoEyes N301 sells for $99 at launch (MRSP: $109) with a choice of 88° (wide-angle), or 137° (ultra-wide-angle) cameras. The 38 x 38 mm main board can also be used for OEM integration into your own products as a smaller alternative to the OpenNV N6, but the company has not provided pricing information for the board. A Wi-Fi HaLow variant appears to be in the plan too. You can also find additional information on the product page.

The post NeoEyes NE302 – A tiny USB-C-powered WiFi 6 Edge AI Vision camera based on STM32N6 MCU appeared first on CNX Software - Embedded Systems News.

Renesas has expanded its RZ/G Series with the RZ/G3L and RZ/G3SE, two new 64-bit MPUs for HMI and IoT edge systems. Both devices integrate up to four Arm Cortex-A55 application cores clocked at up to 1.2 GHz and a 200 MHz Arm Cortex-M33 real-time coprocessor.

The two chips are pin-to-pin compatible but target different applications. The RZ/G3L integrates an Arm Mali-G31 3D GPU and H.264 video codec for HMI applications like retail terminals and medical devices. The RZ/G3SE removes the GPU and video codec and is designed for IoT devices with basic display requirements, including EV chargers and industrial gateway devices. Both can reduce standby power consumption to around 1 mW and resume operation quickly from deep standby, which is useful for devices that need to remain connected while consuming very little power.

Renesas RZ/G3L and RZ/G3SE specifications:

• MPU core
  •  CPU
    • Quad-core or Dual-core Arm Cortex-A55 @ up to 1.2 GHz
    • Arm Cortex-M33 system management processor @ 200 MHz
  • GPU (RZ/G3L only)
    • Arm Mali-G31 @ 600 MHz
    • Supports Vulkan 1.2, OpenGL ES 1.1/2.0/3.2, and OpenCL 2.0
  • VPU (RZ/G3L only)
    • H.264 codec module for encoding and decoding
    • Maximum resolution up to 1920×1080 @ 30 fps
  • Image processing – Image Scaling Unit 2 (ISU2) for hardware scaling and color format conversion (up to 2048×2048)
• Memory
  • 512 KB on-chip system RAM with ECC
  • 1x 16-bit External DDR interface supporting DDR4-2133 (up to 4 GB) or LPDDR4-2133 (up to 2 GB) with ECC
  • 1x External Bus State Controller (EBSC) with 16-bit SRAM support
• Storage
  • 1x xSPI controller (1, 4, or 8 pins)
  • 3x SD/MMC host interfaces (supports SDHI and eMMC 5.1)
• Display
  • 1x MIPI DSI (1, 2, or 4 lanes) up to 1920×1200 @ 60 fps
  • 1x LVDS interface up to 1280×800 @ 60 fps
  • 1x Parallel output interface up to 1280×800 @ 60 fps
  • Only one display interface can be used at a time
• Camera – 1x MIPI CSI-2 (1 or 2 lanes) up to 1.5 Gbps per lane
• Audio
  • 4x Serial sound interfaces (SSIF)
  • 1x SPDIF input/output
  • 3x Pulse-density modulation (PDM) inputs
• Networking – 2x Gigabit Ethernet (10/100/1000 Mbps) with TSN support
• USB – 2x USB 2.0 Host/Function interfaces
• Security
  • Renesas Secure IP, Secure Boot, Secure Update, Secure Debug
  • Hardware cryptographic IP, True Random Number Generator (TRNG)
  • Arm TrustZone, Tamper detection
• Expansion – 1x PCI Express 2.0 (Gen2, 1 lane, Root Complex only)
• Other  I/O
  • 113x configurable I/O pins managed by the Pin Function Controller (PFC)
  • 3x CAN FD interfaces
  • 1x I3C, 4x I2C (RIIC), 3x SPI (RSPI), 4x RSCI, 6x SCIF
• Misc
  • General-purpose PWM timer (8 ch: 32-bit)
  • Multi-Function Timer Pulse Unit (8 ch: 16-bit, 1 ch: 32-bit)
  • Watchdog timer
  • RTC
• Power
  • Core logic – 1.0 V
  • Memory – 1.1 V (LPDDR4) or 1.2 V (DDR4)
  • I/O – 1.8 V, 2.5 V, or 3.3 V
  • Battery Backup Pin (VBATT) – 1.5 V to 1.95 V
• Packages
  • 368-pin LFBGA (17 x 17 mm, 0.65 mm pitch)
  • 400-pin LFBGA (14 x 14 mm, 0.5 mm pitch)
• Temperature Range – -40°C to +125°C

On the software side, RZ/G3L and RZ/G3SE are supported by Renesas’ Yocto-based Linux stack. That includes Linux BSP Plus for current LTS kernels and the Verified Linux Package (VLP). The Cortex-A55 cores run Linux and the Cortex-M33 core is for real-time work, with Zephyr RTOS named as a launch partner. Atmark Techno’s Armadillo Base OS is also available for these MPUs.

For HMI development, the MPUs are supported by the RZ/G HMI SDK and GUI frameworks, including LVGL, Embedded Wizard, Slint, Candera, Spyrosoft, and Qt. Both processors support deep standby with Linux retained in memory for quick wake-up.

Renesas offers the RZ/G3L Evaluation Board Kit for both the RZ/G3L and RZ/G3SE MPUs. The kit includes an RZ/G3L SMARC 2.1.1 module board, a common carrier board, and the required adapter boards and cables. A pre-built Linux image with the HMI SDK is also available, allowing you to boot Linux and start testing the platform with minimal setup.

The company also works with partners like Advanet, which offers the AdiMB7103 compact industrial motherboard and AdiCS7102 slim industrial computer, both based on the RZ/G3L and supporting Ethernet, USB, RS232C, RS485, CAN-FD, display outputs, M.2, MIPI-CSI, and GPIO interfaces. Advanet also offers the Adsmc7101, an 82 × 50 mm SMARC SoM based on the sameRZ/G3L.

Renesas also works with partners like Virtium/Embedded Artists, which offers SoMs based on the RZ/G3 family. The RZ/G3L SoM measures 82 × 30 mm and provides access to the processor’s dual Gigabit Ethernet, PCIe Gen2, dual USB 2.0, and MIPI display and camera interfaces.

The new RZ/G3L and RZ/G3E microprocessors add more members to the RZ/G family, which is already composed of HMI and IoT SoCs such as the RZ/G2L for HMI applications, the RZ/G3S for low-power IoT and gateway devices, and the higher-end RZ/G3E for more demanding HMI applications. The new RZ/G3L sits somewhere in between the RZ/G3S and RZ/G2L, while the RZ/G3SE adds a mid-range IoT option without 3D graphics and VPU.

The RZ/G3L and RZ/G3SE each come in four variants, and Renesas says both families are available for purchase, but distributor stock is still limited a day after launch. DigiKey currently lists four of the eight part numbers, all with no stock and a 14-week factory lead time, with the first incoming stock expected on December 18, 2026. DigiKey prices start at $34.48 for the RZ/G3L and $31.65 for the RZ/G3SE, while Renesas lists prices of $15.92–$21.65 for the RZ/G3L and $14.28–$20.03 for the RZ/G3SE at 1,000-unit quantities. More details can be found on the respective product pages and in the press release.

The post Renesas RZ/G3L & RZ/G3SE – Ultra-low-power Cortex-A55 MPUs for industrial HMI and IoT applications appeared first on CNX Software - Embedded Systems News.

Amlogic has unveiled the A123X quad-core and C305X2 dual-core Arm Cortex-A320 SoCs for industrial and battery-powered Edge AI and IoT applications such as robots, dashcams, IP cameras, video conferencing equipment, and so on.

The Arm Cortex-A320 low-power Armv9 CPU core was introduced in February 2025, and Amlogic is the first silicon vendor to announce Cortex-A320 SoCs. Details are sparse, with no full specifications or block diagrams and limited software information, but let’s see what we know so far.

Amlogic A123X

Amlogic A123X specifications:

• CPU – Quad-core Arm Cortex-A320 processor (Armv9.2-A, SVE2)
• GPU – None or not disclosed
• VPU
  • H.264/H.265 encoding at 4K @ 60fps
  • H.264/H.265 decoding at 4K @ 30fps
• AI
  • 4 TOPS ADLA2 NPU for object detection and tracking CNN models
  • 8 TOPS ADLA3 NPU supporting hardware-accelerated Transformer operations for ViT, LLM, etc.
  • Neural network-based hardware SED engine for low-power audio event identification
• ISP – Low-light HDR ISP
  • Supports up to 16M @ 30fps (SDR) and 13M @ 30fps (HDR) resolution
  • Integrated AI-ISP technology up to 4K@30fps to enhance image quality in ultra-low-light environments
• Memory – 32-bit LPDDR4/4X/5 up to 6400Mbps
• Display I/F – MIPI DSI up to 1080p60
• Camera I/F – Dual 4-lane MIPI CSI, BT.656/BT.1120, 8-lane Sub-LVDS
• Networking – Gigabit Ethernet (GEMAC)
• USB – USB 3.0 OTG, USB 2.0 host
• Misc – 2x CAN FD
• Manufacturing process – TSMC 6 nm

The A123X is designed for machine vision industrial applications such as dashcams, sweeping robots, access control, and video conferencing hardware.

Amlogic C305X2

Amlogic C305X2 preliminary specifications:

• CPU – Dual-core Arm Cortex-A320 processor (Armv9.2-A, SVE2)
• GPU – None or not disclosed
• VPU
  • H.264/H.265 encoding at 4K @ 60fps
  • H.264/H.265 decoding at 4K @ 30fps
• AI
  • ADLA2 NPU for AI-ISP algorithms up to 4K @ 30fps
  • 8 TOPS ADLA3 NPU supporting hardware-accelerated Transformer operations for ViT, LLM, etc.
  • Neural network-based hardware SED engine for low-power audio detection
• ISP – Low-light HDR ISP with integrated AI-ISP technology up to 4K @ 30fps to boost low-light image quality
• Memory – 32-bit LPDDR4/4X/5 up to 4266Mbps
• Display I/F – MIPI DSI up to 1080p60
• Camera I/F – Dual 4-lane MIPI CSI, BT.656/BT.1120
• Networking – Gigabit Ethernet (GEMAC)
• USB – USB 3.0 OTG, USB 2.0 host
• Low power
  • Always-on Video (AOV)
  • Pre-roll low-power modes
  • Integrated neural network SED engine (see AI section above)
• Manufacturing process – TSMC 6 nm

Amlogic says the C305X2 is optimized for ultra-low-power battery devices such as solar doorbells, portable hunting cameras, and compact 2K/4K battery IP cameras. It looks to be an upgrade to the C305X dual-core Cortex-A35 SoC.

Cortex-A320 advantages and software support

Amlogic lists some of the advantages of using Cortex-A320 cores:

• 34% SPECint2006 performance increase over the Cortex-A35 Armv8 core
• SVE2 vector extensions in Cortex-A320 deliver 10 times ML processing uplift compared to Cortex-A35.

Paired with the company’s 8 TOPS Transformer-native ADLA3 NPU, the new Cortex-A320 chips can run lightweight computer vision (CV) algorithms and compact LLMs on-device.

On the software side, we’re just told that the A123X and C305X2 processors leverage Cortex-A320’s Linux software compatibility. The company will also offer turnkey reference designs, pre-validated embedded OS stacks, and unified NN SDKs. While the full SDK has not been released publicly, a few C305X2 references show up on the Amlogic-NN GitHub account.

Both Arm Cortex-A320 processors are now in sampling, and mass production is scheduled for Q4 2026.

Via Amlogic on LinkedIn, and thanks to Stane1983 for the tip.

The post Amlogic A123X and C305X2 Arm Cortex-A320 SoCs target industrial and low-power AIoT applications appeared first on CNX Software - Embedded Systems News.

The LattePanda Mu Ultra is an x86 Compute Module powered by an Intel Core Ultra 5 226V or Ultra 7 256V Lunar Lake processor and equipped with 16GB of DDR5 memory.

Offered in the same compact 69.6 × 60mm form factor as the earlier LattePanda Mu Alder Lake-N system-on-module (N100 or Core i3-N305), it delivers much higher performance, including up 115 TOPS of AI performance. Target applications include robotics, industrial automation, portable instruments, vision AI, and a range of Edge AI applications.

LattePanda Mu Ultra

Specifications:

• Lunar Lake SoC (one or the other)
  • Intel Core Ultra 5 Processor 226V
    • CPU – Octa-core processor up to 4.5 GHz
    • GPU – 7x Xe-core Intel Arc 130V GPU up to 1.85 GHz
    • AI – 40 TOPS Intel AI Boost; combined up to 97 TOPS
  • Intel Core Ultra 7 Processor 256V
    • CPU – Octa-core processor up to 4.8 GHz
    • GPU – 8x Xe-core Intel Arc 140V GPU up to 1.95 GHz
    • AI – 47 TOPS Intel AI Boost; combined up to 115 TOPS
• System Memory – 16GB LPDDR5 @ 8533 MT/s (MoP); up to 136.5GB/s bandwidth
• Storage
  • 256Mbit SPI flash for AMI UEFI/BIOS
  • No storage for OS on module itself
• Display Interface – 40-pin eDP 1.4 connector + I2C touch connector
• 260-pin SO-DIMM edge connector
  • Storage – Via PCIe (for NVMe SSD on carrier)
  • Display interfaces
    • 3x HDMI 2.0/DisplayPort 1.4
    • Support for three independent displays
  • WiFi/Bluetooth – CNVio3 interface
  • USB – 2x USB 3.2 Gen2 (10Gbps), 6x USB 2.0
  • PCIe
    • Up to 4x PCIe 4.0 x1 lanes or
    • Up to 4x PCIe 4.0 x2 lanes or
    • Up to 2x PCIe 4.0 x4 lanes
  • 3x UART, 2x I2C
  • Up to 14x GPIO
• Dimensions – 69.6 x 60 mm
• Temperature Range – 0 to 60°C
• Relative Humidity – 0 to 80%

The company provides support for Windows 11 and Ubuntu 24.04/26.04 operating systems.  They also shared some CPU (GeekBench 6.0), GPU (3DMark Time Spy), and AI benchmarks showing how the new Lunar Lake Compute Module compares to the earlier LattePanda Sigma (Intel Core i5-1340P) and/or AMD Ryzen AI 9 365. However, the latter was somehow not included in the AI benchmarks.

LattePanda Mu Mini Carrier Board

System-on-modules typically come with evaluation carrier boards for testing and early software development as a custom carrier board is being developed. Arnon previously reviewed the LattePanda Mu (N100) with the Lite and Full Eval carrier boards, but the new Ultra Compute Module is not fully compatible with the Lite board and is incompatible with the Full Eval board (potential hardware damage!).

So DFRobot designed a new “Mini Carrier Board” that fully works with all four LattePanda Mu/Mu Ultra variants.

DFR1293 Mini Carrier Board specifications:

• Compatible Compute Modules
  • LattePanda Mu N100/Core i3-N305
  • LattePanda Mu Ultra 226V/256V
• Storage – M.2 2230/2280 NVMe SSD socket
• Video Output
  • HDMI 2.0
  • 2x DisplayPort via USB-C ports
• Audio
  • 3.5mm (microphone/headphone) combo jack
  • 2x speaker output connectors
  • Realtek ALC269 audio codec
• Networking
  • 2.5Gbps Ethernet RJ45 port  via Intel I226-V controller
  • Optional WiFi 6/7 and Bluetooth via M.2 E-Key socket
• USB
  • 2x USB 3.2 Gen2 Type-A ports
  • 2x full-featured USB Type-C ports (USB 3.2 Gen2, USB 2.0, USB PD, DP Alt mode)
• Expansion
  • M.2 E-Key slot (PCIe x1, Option CNVio, 2230)
  • M.2 M-Key slot (up to PCIe 4.0 x2, 2230/2280)
  • OCuLink Port (up to PCIe 4.0 x4)
  • 22-pin GPIO header
• Misc
  • Power button with LED
  • CPU FAN connector
  • F_PANEL header
  • GPIO_LEVEL Jumper
  • R_USB_PWR Jumper
  • AUTO_PWR Jumper
• Power Supply
  • 12–20V via USB-C PD or 5.5×2.5mm DC jack, 4-pin connector, or UPS expansion board
  • Minimum power for LattePanda Mu Ultra: 50W
  • PWR MGMT Port (for LattePanda UPS Expansion Board)
• Dimensions – 110 x 94mm

You’ll find more technical details and instructions to get started in the Compute Module and Carrier Board sections of the documentation website.

The LattePanda Mu Ultra starts at $599 (226V) or $699 (256V). However, most people will likely start with the LattePanda Mu Ultra AI Developer Kit going for $643 or $733 on the DFRobot website depending on the variant (226V or 256V) . The kit includes the Compute Module, the Mini Carrier, the Active Cooler, ten Nylon standoffs, five Nylon screws, a button cap, and the product manuals for the SoM and carrier board. That means you’ll need your own NVMe SSD and power supply.

The post LattePanda Mu Ultra Compute Module features Intel Core Ultra 5 226V/7 256V Lunar Lake CPU for AI workloads appeared first on CNX Software - Embedded Systems News.

Arduino UNO Media Carrier board adds three audio jacks, a MIPI DSI display connector, two MIPI CSI camera connectors, and a few RGB LEDs to the Arduino UNO Q and Arduino VENTUNO Q SBCs.

Both Arduino “Q” boards already expose MIPI CSI/DSI and audio interfaces, but only through the JMEDIA and JMISC 60-pin headers, which aren’t very convenient. The Arduino UNO Media Carrier board fixes that with 22-pin MIPI connectors and 3.5 audio jacks.

Arduino UNO Media Carrier (ASX00083) specifications:

• Compatible SBCs – Arduino UNO Q and Arduino VENTUNO Q
• Display I/F – 22-pin MIPI DSI connector compatible with standard MIPI-DSI display modules (5-inch, 8-inch, 10.1-inch Waveshare touch display compatible)
• Camera I/F – 2x 22-pin MIPI-CSI connectors compatible with IMX219 camera modules like the Raspberry Pi Camera Module 2
• Audio
  • 3.5 mm (MIC-IN/Headphones Out) audio jack
  • 3.5mm Line Out audio jack
  • 3.5mm Ear Out audio jack
• Host interfaces
  • 60-pin female JMEDIA connector (passthrough)
  • 60-pin female JMISC connector (passthrough)
  • Available I/Os
  • I2C – CCI I2C (1.8 V, via JMEDIA) and MCU I2C4 (3.3 V, via JMISC)
  • PSSI – 8-bit parallel camera interface (via JMISC)
  • GPIO – SoC GPIO (1.8 V) and MCU GPIO (3.3 V) via JMISC
• Misc – 4x RGB LEDs
• Power Supply
  • Main source – Host board via JMEDIA and JMISC connectors
  • VIN (DC_IN) – 7-24 V DC via J13 4-pin header
  • VCOIN – 3.0 V DC input for host board RTC (max 3.6 V) via J10 4-pin header
• Dimensions – 68.58 x 53.34 mm
• Temperature Range – -10 to 60°C

Arduino provides pinout diagrams, a datasheet, PDF schematics, Altium CAD files, and STEP files for the open-source hardware board on the documentation website.

The documentation also features two tutorials: a generic user manual and a Kiosk Mode tutorial showing how to connect and configure the UNO Q and Media Carrier with a Waveshare touch display.

The UNO Media Carrier board is sold for $19.25 or €19,89 (VAT included) on the Arduino store.

The post Arduino UNO Media Carrier adds MIPI CSI/DSI and audio connectors to UNO Q and VENTUNO Q boards appeared first on CNX Software - Embedded Systems News.

Waveshare ESP32-C5 Pico board brings the ESP32-C5 dual-band WiFi 6, Bluetooth LE, and 802.15.4 (Zigbee/Thread) SoC to the Raspberry Pi Pico form factor.

The board features an ESP32-C5-MINI-1 Series module with PCB or external antenna, two 20-pin GPIO headers, a USB-C port for power and programming, Reset and Boot buttons, and a few LEDs. It also supports Li-Ion battery charging and discharging.

Waveshare ESP32-C5 Pico specifications:

• Wireless module – ESP32-C5-MINI-1 or ESP32-C5-MINI-1U
  • Wireless MCU –  Espressif Systems ESP32-C5HF4
    • CPU
      • Single-core 32-bit RISC-V processor @ up to 240 MHz
      • Low-power RISC-V core @ 48 MHz acting as the main processor for power-sensitive applications
    • Memory – 384 KB SRAM on-chip
    • Storage – 320 KB ROM, 4MB SPI flash
    • Wireless
      • Dual-band (2.4/5 GHz) 802.11ax WiFi 6, 802.11b/g/n WiFi 4 backward compatibility
      • WiFi modes: Station mode, SoftAP mode, SoftAP + Station mode, and promiscuous mode
      • Bluetooth 5.0 Low Energy (LE) with Mesh support, up to 2Mbps data rate
      • 802.15.4 radio for Zigbee 3.0, Thread 1.3, and Matter up to 250 K
  • ESP32-C5-MINI-1 – PCB Antenna
  • ESP32-C5-MINI-1U – IPEX antenna connector (antenna provided with the kit)
• USB – USB Type-C port for power/charging and programming
• Expansion
  • 2x 20-pin Raspberry Pi Pico-compatible headers with up to 26x GPIO, 20x PWM, I2C, SPI, UART, SDIO, JTAG, Serial, 4x ADC, VBUS, VSYS, VBAT, 3.3V, GND
  • Note: 6x GPIO pins through TCA9554PWR GPIO expander
• Misc
  • Boot and Reset buttons
  • Power, Charging, and reverse polarity warning LEDs
  • RGB LED
• Power Supply
  • 5V via USB-C port or VBUS pin
  • 3.7V Lithium battery support via 2-pin battery connector
  • MP1605GTF-Z synchronous buck (step-down) DC-DC converter taking 2.3–5.5V input
  • Eta Solutions (Yutai Semiconductor) ETA6098 switching Li-Ion battery charger
• Dimensions
  • ESP32-C5-MINI-1 variant – 56.85 x 21 mm
  • ESP32-C5-MINI-1U variant – 51 x 21 mm

Four variants of the board are offered:

• ESP32-C5-Pico – PCB antenna, unsoldered headers
• ESP32-C5-Pico-M – PCB antenna, pre-soldered headers
• ESP32-C5-Pico-U – External antenna, unsoldered headers
• ESP32-C5-Pico-UM – External antenna, pre-soldered headers

The company provides support for the Arduino IDE and the ESP-IDF framework. You’ll find additional technical details PDF schematics, datasheets, and instructions to get started on the documentation website.

The company used to share code through tarballs, but Arduino/ESP-IDF code samples and other resources for the ESP32-C5 Pico board can be found on GitHub instead. Code samples are available for board information, the built-in RGB, the I/O expander, dual-band WiFi, UART, and connecting a small external 1.83-inch touchscreen LCD and an SPI CAN transceiver (MCP2515-based).

Waveshare sells the ESP32-C5 Pico for $10.43 to $11.33 on AliExpress, $15.99 to $17.99 on Amazon, and $8.99 to $9.99 on the company’s online store. As usual, the prices don’t include shipping and taxes since these vary by country.

The post ESP32-C5 Pico board follows Raspberry Pi Pico form factor, ships with on-board or external antenna appeared first on CNX Software - Embedded Systems News.

The FREE-WILi 2 (Founder Edition) is an open-hardware, portable hacking multitool and lab designed for hardware hackers, pentesters, and embedded system developers. Designed as an “open electronics multitool,” it combines a range of chips and modules into a handheld device, including two Raspberry Pi RP2350 MCUs, an ESP32-C5 for wireless connectivity, a Lattice ICE40UP5K FPGA, and a Raspberry Pi CM0 for running a full Linux OS.

The device also features a 3.5-inch capacitive touchscreen, a 14-button physical interface, and modular “Orca” headers for hardware expansion. Its wireless features include 2.4/5GHz Wi-Fi, Bluetooth, LoRa, sub-GHz RF, NFC, 125kHz RFID, and IR. These features make it very suitable for wireless testing, embedded hardware debugging, real-time control, electronics development, and retro gaming, with support for Adafruit Fruit Jam, PICO-8, and Doom.

FREE-WILi 2 specifications:

• Compute
  • Main MCUs – 2x Raspberry Pi RP2350B (One for main controls, one dedicated to display)
  • SoM – Raspberry Pi CM0 (to run a full 64-bit Linux OS and run heavy Python scripts)
    • 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 – 8GB or 16GB eMMC flash
    • Wireless – Optional WiFi 4 and Bluetooth 5.0 module
  • FPGA – Lattice UltraPlus iCE40UP5K FPGA with 5.3K LUTs, 1Mbit SPRAM, 120Kbit DPRAM, 8x multipliers with all pins brought out; 8 MB of dedicated SRAM
• Storage – 2x MicroSD card slots (one slot is for device firmware; the other is for CM0)
• Display
  • 3.5-inch 480×320 capacitive touch screen
  • DVI output (Driven by RP2350 HSTX)
• Audio
  • 3.5 mm audio jack (headphone/mic)
  • 4-channel phased-array microphone
  • 0.5W onboard speaker
• Wireless
  • Dual-band (2.4 GHz + 5 GHz) Wi-Fi 6, Bluetooth LE, and 802.15.4 (Zigbee/Thread) via Espressif ESP32-C5
  • Sub-GHz connectivity via CC1101 module
  • LoRa via STM32WLE5JC Arm Cortex-M4 MCU @ 48 MHz with 256 KB flash memory, 64 KB SRAM, SX126x LoRa radio
  • NFC / RFID – 125 kHz RFID (LF) + ST25R3916B NFC (HF) for tag reading, cloning, and emulation
  • Infrared – IR Tx/Rx with dedicated case window
  • SMA connector for LoRa and Sub-GHz
• USB
  • 3x USB host ports (2x 12 Mbps connected to RP2350 display MCU + 1x 480 Mbps port connected to CM0)
  • 1x USB Type-C port (CLI, OneWili, GUI, high-speed SD reader)
• Sensors
  • IMU – Bosch BMI323 (9-DOF with magnetometer) + BMM350 Magnetometer for motion input
  • Environment – Sensirion SHT40 (Temperature & Humidity)
  • Light – OPT4001 ambient light sensor
• Debug
  • Raspberry Pi Debug Probe onboard (for RP2350 + LoRa debugging)
  • ESP32 JTAG via USB
• Expansion
  • 20-pin (FW1) main connector + 10-pin analog connector + mounting posts for “Orca” add-on modules
  • 13x GPIOs (SPI, I2C, UART) with software-selectable I/O voltage
  • 4x analog inputs (0–5V) with op-amp buffering, Programmable Gain Amplifier (PGA), and window comparator
  • 2x analog outputs (0 to ~4.84V @ 25 kHz)
  • CAN FD at full 8 Mbit/s with CAN SIC transceiver
• Misc
  • 5-way D-pad + A/B/X/Y (home/ok/cancel/page) buttons
  • 14 gamer buttons total (emulator-friendly for PICO-8, Doom, Fruit Jam)
  • 5x under-screen context backlit keys
  • Innovative 2-press-per-letter hardware keyboard
• Power
  • 3,000 mAh battery with USB-aware charging
  • Dedicated ultra-low-power (ULP) micro managing 17 independent power zones
  • 1–5.5V / 1.5A programmable power supply with MOSFET crowbar for voltage glitching
  • 60 µA sleep current
• Dimensions – 152.4 × 78.9 × 22.3 mm
• Weight – 290 grams

On the software side, FREE-WILi 2 runs stock firmware that exposes a USB CLI, on-device GUI, and the OneWili API for controlling the hardware from Python, Rust, or C/C++. It also supports rThon for Python-like real-time scripting, WiliBlocks for visual programming, and ZoomIO for sub-microsecond real-time control. You can also load UF2 applications from the microSD card, run WebAssembly programs through WiliWasm, and use the FREE-WILi GUI on a PC without installation.

It also ships with open hardware documentation, an open board support package (WiliBSP), and “Agent.md” files. This allows LLMs like Claude Code or local models running via LM Studio to autonomously write, compile, and debug custom firmware directly on the device using the onboard hardware debugging probes. More information is available on their GitHub account.

The company mentions that they have ported Meshtastic to the onboard STM32WLE5JC LoRa radio, so FREE-WILi 2 can join a LoRa mesh out of the box. That radio shares one external antenna with the TI CC1101 sub-GHz transceiver via dynamic switching. This is a vendor port — FREE-WILi 2 is not listed as an official Meshtastic device, and regional band settings are not documented yet.

We’ve previously written about other portable wireless hacking tools such as the Flipper Zero, MCU-based alternatives like HackBat and the M1 multitool, Linux-based devices like Interrupt and the much more powerful Flipper One; compared to all of those, the FREE-WILi 2 sits in that same pocket-lab space, but packs different types of SoCs and a wide set of radios into one handheld.

The FREE-WILi 2 Founders Edition is currently available for pre-order at $400 directly on the company’s website. Shipping is expected to begin in Q4 2026. More information is available on the project’s website.

The post FREE-WILi 2 portable hacking multitool features two RP2350 MCUs, ESP32-C5, ICE40 FPGA, and Raspberry Pi CM0 appeared first on CNX Software - Embedded Systems News.

Arm has introduced the CSS for Mobile 2 platform as an AI-native compute foundation comprised of the Arm C2-Ultra and C2-Pro CPUs with SME2 (Scalable Matrix Extension 2) units and the Mali G2-Ultra NX GPU with dedicated neural accelerators that enable neural graphics to run natively on the GPU.

Arm C2-Ultra and C2-Pro CPUs

Arm C2-Ultra specifications:

• Architecture – Armv9.3-A
• Extensions
  • Armv9.4-A (Some key features)
  • SME2 Extensions
  • SVE2 Extensions
  • Cryptography Extensions
  • RAS Extensions
• ISA support – AArch64
• Microarchitecture
  • Pipeline – Out-of-order
  • Superscalar
  • Neon and SVE2
  • Optional Cryptography Unit
  • Up to 14x CPUs in cluster
  • 40-bit Physical Addressing
• Memory System and External Interfaces
  • 64KB (I-Cache), 128KB (D-Cache)
  • 2MB or 3MB L2 Cache
  • Optional, 256KB to 32MB L3 Cache
  • ECC Support
  • Bus interfaces – AMBA AXI5 or CHI.E
  • Optional ACP
  • Optional Peripheral Port
• Security – TrustZone, Secure-EL2
• Interrupts – GIC interface, GICv4.1
• PMU – PMUv3
• CoreSight – CoreSightv3
• Embedded Trace Extension – ETEv1.0
• Trace Buffer Extension
• Optional Area-Optimised Configuration

Arm says a C2 CPU cluster delivers the following performance improvements over an equivalent C1 CPU cluster:

• Up to 1.7x performance improvement on AI models
• Up to 15 percent higher single-thread performance
• 15 percent faster web browsing
• 12 percent faster application launch
• 12 percent higher multi-thread performance at the cluster level

One practical example is lower LLM/agent latency of a dual-core C2-Ultra cluster compared to a dual-core C1-Ultra cluster, especially when SME2 units are enabled.

I could not find C2-Pro documentation at this stage (it points to C1-Pro docs), but you’ll find more technical details about the new C2-Ultra CPU core on the documentation website.

Arm Mali G2-Ultra NX GPU

Mali G2-Ultra NX key features and specifications:

• API support – OpenGL ES 3.2, Vulkan 1.4, OpenCL 3.0 full profile
• Bus interface – AMBA 4 ACE, ACE-LITE, AXI
• Configurable dual L2 cache supports up to 8 slices of 2MB each
• Scalability – 10 to 24 cores
• Adaptive Scalable Texture Compression (ASTC) with LDR and HDR support
• Arm Frame Buffer Compression (AFBC), version 1.3.2, 4×4 blocks
• Arm Fixed Rate Compression (AFRC)
• Arm Neural Technology
  • Neural Super Sampling (NSS): AI upscaling from 540p to 1080p
  • Neural Frame Rate Upscaling (NFRU): Frame rate doubling from 30 FPS to 60 FPS
  • Neural Super Sampling and Denoising (NSSD): A combination of NSS and ray reconstruction
• Ray Tracing –  RTUv3 – improved BW efficiency, higher performance, lower area, and HW-based OMM
• Variable Rate Shading – Up to 4×4 shading rate supported

Compared to the earlier Arm Mali G1-Ultra GPU, architectural improvements in the Mali G2-Ultra deliver up to 24 percent higher benchmark
performance, 14 percent higher non-AI gaming performance, and a reduction of 70% for ray tracing workloads. With neural acceleration through NFRU, the system supports longer mobile gaming sessions at up to 120 FPS, and DRAM traffic is also greatly reduced by up to 70%.

You’ll find additional information for the new GPU on the Arm website. You can also visit the launch page for more details and links about the  Arm CSS for Mobile 2 platform as a whole.

The post Arm CSS for Mobile 2 “AI-Native” platform: Arm C2-Ultra and C2-Pro CPU cores, Mali G2-Ultra NX GPU appeared first on CNX Software - Embedded Systems News.

YuzukiHD has made another open-source hardware Linux SBC. YuzukiNeko is a Raspberry Pi Pico-sized single board computer powered by a new Allwinner F101 RISC-V SoC.

The Allwinner F101 processor features a single 1.008 GHz C907 RISC-V core, 8MB or 16 MB PSRAM, and support for various display interfaces such as MIPI DSI, 24-bit RGB, and LVDS. The YuzukiNeko board itself features the F101 SoC, an SPI NOR flash, a microSD card for storage, a USB-C port, and two 20-pin GPIO headers.

YuzukiNeko specifications:

• SoC – Allwinner F101
  • CPU – T-head C907 RISC-V core up to 1.008 GHz
  • Memory – 8MB or 16MB PSRAM @ 252 MHz
  • Package – QFN-68
• Storage
  • 128 Mbit NOR flash (PY25Q128HA-SUH-IT)
  • MicroSD card slot
• USB – USB Type-C port for data, power, and programming
• Expansion
  • 2x 20-pin expansion headers with GPIOs, ADC, 5V, 3.3V, and GND
  • 5x pads with HP out (analog speaker output) and four GPIOs
• Misc – FEL button, power LED
• Power Supply – 5V via USB-C port
• Dimensions – About 51 x 21 mm

You’ll find the PDF schematics, EasyEDA PCB layout, Gerber files, 3D STEP file, BoM, and other hardware resources on GitHub. The Linux source code based on mainline (Linux 7.2) and Buildroot are in separate repositories. In addition to Linux, the board and the Allwinner F101 processor also support Zephyr RTOS.

A product brief for the Allwinner F101 “Smart Control and Display Processor” can be found on the Whycan website, so we have more details about the F101, including a block diagram and specs.

Allwinner F101 specifications:

• CPU – T-Head XuanTie C907 RISC-V CPU @ 1.008 GHz with  32 KB I-cache + 16 KB D-cache
• Display Engine
  • Allwinner Awonder1.1 Lite post-processing
  • De-interlace (DI) up to 720p @ 60fps
  • G2D hardware accelerator supporting rotate and mixer functions, among others
• Video engine
  • Image decoding
    • JPEG up to 16384×16384
    • PNG up to 2048×2048
  • Video encoding
    • JPEG/MJPEG up to 720p @ 30fps
    • Maximum resolution: 8192×8192
• Memory
  • F101S2-XXX – 8MB 16-bit PSRAM (SiP)
  • F101S3-XXX – 16MB 16-bit PSRAM (SiP)
• Storage – SPI flash interface
• Video Output
  • RGB888 up to 1280×720 @ 60fps
  • Single-link LVDS  up to 1366×768 @ 60fps
  • 4-lane MIPI DSI  up to 1280×800 @ 60fps
• Audio
  • DAC
  • Analog audio – HPOUT
  • Digital audio – I2S/PCM, OWA OUT
• USB – USB 2.0 DRD
• Other peripherals
  • SDIO2.0/3.0
  • 2x SPI, 6x UART, 3x TWI  (I2C)
  • 4-ch PWM, 2x PWM_BL, IR_RX
  • 1-ch GPADC, 4-ch TPADC
• Security – Integrated 512 bits eFuse storage space
• Package – QFN68, 7 x 7mm

The YuzukiNeko is not for sale yet, but may be soon. Other open-source hardware from YuzukiHD, like the Avaota A1 and Yuzuki Chameleon, ended up being sold online. One member on a Whycan forum thread mentions that the Allwinner F101 sells for 18 yuan, or about $ 2.70 US. This could mean the YuzukiNeko board or similar designs could sell for under $10.

Thanks to Freire for the tip.

The post YuzukiNeko – A Linux-capable Allwinner F101 RISC-V SBC with Raspberry Pi Pico form factor appeared first on CNX Software - Embedded Systems News.

When the Part Exists, but the Data Doesn’t

Reverse engineering often starts with a physical part but no reliable digital data. In restoration and customization, manual measurement can provide basic dimensions, but it becomes limiting when complex curves, mounting points, and tight clearances need to be captured accurately.

MetroY Ultra captures the actual three-dimensional geometry of a part, providing a reliable digital reference for reverse engineering, custom design, and manufacturing.

Designing a Custom Motorcycle Seat Tray from Accurate Scan Data

In a custom 1984 Honda CB400 build, the builder needed to create a seat tray and seat pan for a custom seat while integrating the rear fender into the modified rear subframe. The challenge was not simply designing a new part, but making that part accurately fit a motorcycle whose geometry had already been modified. Traditionally, this kind of work is often done with welded sheet metal or fiberglass layup. Both methods can produce strong parts, but they also require cutting, trimming, sanding, fitting, and repeated manual adjustment. When the geometry of the existing structure is not fully known, these adjustments are often part of the design process itself.

Instead of starting with metalwork or fiberglass, the builder moved the project into a digital workflow. The motorcycle had already been 3D scanned, and MetroY Ultra was then used to capture the rear fender from multiple angles, including both the top surface and underside. Markers were placed on and around the fender to support stable tracking, and enough overlap was captured so the scan data could be merged into a complete model. Capturing both sides of the fender was important because the part’s overall shape, mounting features, and relationship to the surrounding frame all needed to be represented in the digital model.

The scan was captured and initially processed in Revo Metro, Revopoint’s software for MetroY Ultra. From there, the data could be reviewed and prepared before moving into further cleanup. One practical challenge was that some marker-related holes were similar in size to the real mounting holes on the fender, so the model had to be cleaned carefully: unwanted holes and scan artifacts were removed, while actual bolt holes were preserved. This distinction was important for the downstream design: preserving the actual mounting geometry while removing scan artifacts ensured that the digital model remained a useful representation of the physical part.

Once cleaned, the rear fender scan was combined with the full motorcycle scan in CAD software. This gave the builder an accurate digital reference of how the fender, frame hoop, and seat area related to one another. Instead of working from a series of measurements and trying to reconstruct these relationships manually, the builder could now design against the actual geometry of the modified motorcycle. Using the scanned frame geometry as a guide, he designed a 3D printable seat tray with cutouts matching the actual frame, along with cable routing holes and mounting points for brass inserts.

For this type of custom-fit design, MetroY Ultra provides accurate scan data that gives the early design stage a more reliable foundation. The value is not simply having a 3D model of the fender, but having its actual geometry available as a design reference. Instead of starting with rough measurements or repeated trial fitting, the builder could test and adjust the design digitally before cutting metal, laying fiberglass, or moving toward a full-size 3D print. For a one-off custom build, this helps move fitment from a physical trial-and-error process into a digital design process.

Accurate Digitization of a Classic Car Part

A radiator grille may look like a simple exterior component, but in classic vehicle restoration it can be a fitment-critical part. In this project, MetroY Ultra was used to scan a radiator grille for a VW Golf MK1 Convertible, creating a high-accuracy digital reference for a part that may be difficult to source, repair, or reproduce.

For older vehicles, original CAD data is often unavailable, and replacement parts can vary in condition, shape, or fit. For a grille, accuracy matters beyond overall dimensions: edge profiles, mounting points, openings, surface transitions, and connections with surrounding bodywork all affect fitment.

MetroY Ultra’s high-accuracy scanning captures these details more completely than manual measurement alone. Instead of relying on visual approximation or limited dimensions, restorers and designers can work from the actual geometry of the part. The resulting digital model provides a reliable reference for repair planning, reproduction, modification, or long-term digital archiving.

By accurately digitizing the grille, MetroY Ultra helps make restoration and redesign more controlled, especially for parts that are no longer easy to replace through normal supply channels.

Scanning a Black Part for Custom Fitment

Another representative application came from a modified S197 setup equipped with a 132 mm Whipple throttle body. The project required a custom thermal spacer kit, where accurate understanding of component interfaces was critical.

In performance modification projects, custom parts often need to fit a specific setup rather than a factory configuration. Important geometry includes mounting areas, bolt holes, edge profiles, clearances, and surrounding structures. Without accurate data, repeated adjustments may be needed to achieve proper fitment.

The component had a black surface, which can make 3D scanning more challenging, especially for curved areas, recessed sections, and fitment-critical edges. Yet MetroY Ultra was able to capture usable geometry across these challenging areas, providing a digital reference for spacer design and alignment.

Using scan data from the actual component helps reduce guesswork before modeling or manufacturing. It allows designers to better understand fitment requirements, clearance needs, and key features required for proper installation.

A More Reliable Starting Point for Reverse Engineering

Together, these examples demonstrate three capabilities that matter in real-world reverse engineering: accurate geometry for highly customized fitment, high-accuracy digitization of existing parts, and reliable scanning of challenging black surfaces. Revo Metro helps connect the scanning stage with downstream work by supporting scan capture, review, and initial data preparation. The result is a workflow that starts from the actual geometry of the part, rather than rough measurements or assumptions.

For designers, makers, restorers, and engineers working with custom parts, legacy components, or small-batch projects, MetroY Ultra offers a practical way to bring real-world parts into a digital workflow. Whether the goal is to build around an existing structure, reproduce a hard-to-source component, or capture a challenging surface, the process begins with reliable geometry. MetroY Ultra helps turn that geometry into data that can support the next stage of design and manufacturing.

CERTIFIED METROLOGY

Revopoint MetroY Ultra

Built for demanding reverse engineering, inspection, and design validation workflows, MetroY Ultra captures complex parts with high accuracy using multiple blue-light scanning modes. It is ideal for digitizing mechanical components, automotive parts, and hard-to-scan surfaces for CAD reconstruction and product development.

Price: from ~$1,899
Discount code: REVOYCSCNX

View at Revopoint.

The post Reverse Engineering Without CAD Data: How MetroY Ultra Turns Physical Parts into Digital Designs (Sponsored) appeared first on CNX Software - Embedded Systems News.

Seeed Studio has expanded its XIAO Plus Series introduced in 2025, with the XIAO SAMD21 Plus and XIAO RP2040 Plus USB-C boards adding several GPIOs and integrated Li-ion battery management to the XIAO (SAMD21) and XIAO RP2040, launched in 2020 and 2021, respectively, while keeping the same form factor.

XIAO SAMD21 Plus

Specifications:

• MCU – Microchip SAMD21G18 Arm Cortex-M0+ microcontroller at up to 48MHz with 256KB Flash,32KB SRAM
• USB – 1x USB Type-C port for power and programming
• Expansion I/Os
  • 2x 7-pin “legacy” headers with 11x analog inputs, 11x digital I/Os, 1x DAC, SPI, UART, and I2C
  • Castellated holes and pads with 16x digital I/Os, I2C,and I2S
  • 3.3V I/O voltage (not 5V tolerant)
• Misc
  • Reset and Boot buttons
  • RGB LED
  • Charge LED
• Power Supply
  • 5V via USB-C port
  • Pads for battery
  • On-board PMIC
• Dimensions – 23.5×17.5 mm
• Temperature Range – -40°C ~ 85°C

The upgraded variant of the XIAO SAMD21 features 16 additional SMD GPIOs and integrated Li-ion battery management. The board supports Arduino, PlatformIO, MicroPython, CircuitPython, Zephyr, and more. Check out the wiki for more details.

The XIAO SAMD21 Plus is sold for $5.90 on Seeed Studio, and should soon be offered as an option on the company’s AliExpress and Amazon stores.

XIAO RP2040 Plus

Specifications:

• MCU – Raspberry Pi RP2040 dual-core Cortex-M0+ up to 133 MHz with 264 kB SRAM
• Storage – 2MB SPI flash
• USB – 1x USB Type-C port for power and programming
• Expansion I/Os
  • 2x 7-pin headers with 4x analog inputs, 11x digital I/Os/PWM, SPI, UART, and I2C; 2.54mm pitch
  • Castellated holes and pads with 15x digital I/Os/PWM, 1x I2C 
  • 3.3V I/O voltage (not 5V tolerant)
• Misc
  • Reset and Boot Buttons
  • Charge and User LEDs
  • RGB LED
• Power Supply
  • 5V via USB-C port
  • Pads for battery
  • On-board PMIC
• Dimensions – 21 x 17.8 mm
• Temperature Range – -20°C ~ 70°C

The upgraded variant of XIAO RP2040 features 15 additional SMD GPIOs and integrated Li-ion battery management. The board supports Arduino, PlatformIO, MicroPython, CircuitPython, TinyGo, Rust, Zephyr, and more. You’ll find more details and instructions to get started on the wiki.

Seeed Studio sells the XIAO SAMD21 Plus for $4.90.

The post XIAO SAMD21 Plus and RP2040 Plus USB-C boards get additional GPIO pins, integrated Li-ion battery management appeared first on CNX Software - Embedded Systems News.

After checking out the hardware of the ACEMAGIC Kron Mini K5 with an unboxing and a teardown in the first part of the review, we’ve now had time to test the Intel Core 3 304 Wildcat Lake mini PC in more depth with Windows 11 Pro.

So in this article, we’ll report our experience with the penta-core mini PC, testing features, running system and AI benchmarks, playing 4K and 8K YouTube videos on Firefox, measuring network throughput (2.5GbE and WiFi 6), evaluating thermal performance, and measuring fan noise and power consumption.

Software overview and feature testing

The System > About window confirms we have a 1.5 GHz (base frequency) Intel Core 3 304 computer paired with 12GB of RAM and 477GB of storage, running Windows 11 Pro 25H2 build 26200.9168.

HWiNFO 64 provides additional details about the 15W Intel Core 3 304 penta-core (1P+4E) Wildcat Lake-U processor, the motherboard with 12GB of LPDDR5 SDRAM, and Intel Wildcat Lake integrated graphics with 1 Xe core running at 1000MHz (up to 2300 MHz).

We also ran GPU-Z to get more information about the iGPU, as we usually do when testing mini PCs. However, the Wildcat Lake SoCs are very new, and the utility has not been updated, resulting in the Intel GPU being identified as “FD81” and several “Unknown” fields.

The PL1 and PL2 power limits are set to 22W (PBP) and 35W (MTP), respectively, with PL4 at 60W. That’s a bit different from the values reported for the Beelink EQi Wildcat Lake 304: 15W (PBP), 35W (MTP), and 80W (PL4), so multicore performance should be a little higher by default if the cooling system is sized adequately.

On the memory side, HWiNFO64 confirms that the system comes with 12GB of soldered LPDDR5 SDRAM, consisting of four 3GB Samsung chips.

Task Manager shows that Windows can use up to 11.7GB, running at 5,600 MT/s, while 332 MB is reserved for the hardware, notably the GPU and NPU, both of which are also detected in Task Manager.

Let’s go back to HWiNFO64 to check out networking. The 2.5GbE port relies on a Realtek RTL8125 Gaming 2.5GbE Family Ethernet Controller.

A Realtek RTL8852BE Wi-Fi 6 802.11ax PCIe adapter (module) shows a maximum link speed of 573 Mbps, which should be OK, but is quite lower than the link speed on most other mini PCs. For reference, we had a 2999 Mbps link on the Beelink EQi Wildcat Lake.

We went back to Device Manager to check the Bluetooth version. “LMP13.xxx” firmware version looks up to Bluetooth 5.4.

We also took the opportunity to quickly test Bluetooth connectivity using a file transfer from the mini PC to a Redmi Note 14 Pro 5G smartphone. No issues.

The ACEMAGIC Kron Mini K5 (Intel Core 3 304) has six USB 3.2 Type-A ports and one USB 3.2 Type-C port. Some ports are 5 Gbps, others are 10 Gbps, but sadly none of them are properly marked on the mini PC. Nevertheless, we double-checked the actual speeds using an ORICO M234C3-U4 M.2 NVMe SSD enclosure, HWiNFO 64 utility, and CrystalDiskMark benchmark.

For reference, here are the results for the front left USB 3.2 Type-C (10 Gbps) port… (note: advertised as 20 Gbps)

… and the left USB 3.2 Type-A (5 Gbps) port on the rear panel.

Summary for all seven ports from left to right:

• Front panel
  • USB-C – USB 3.2 – USB 3.2 Gen2 (SuperSpeedPlus+ 10 Gbps) – Read speed: 1,062.51 MB/s; Write speed: 1,039.35 MB/s
  • USB-A #1 – USB 3.2 – USB 3.2 Gen2 (SuperSpeedPlus+ 10 Gbps) – Read speed: 1,060.28 MB/s; Write speed: 1,045.51 MB/s
  • USB-A #2 – USB 3.2 – USB 3.2 Gen2 (SuperSpeedPlus+ 10 Gbps) – Read speed: 1,057.36 MB/s; Write speed: 1,040.01 MB/s
• Rear panel
  • USB-A port #1 (top, left)– USB 3.2 – USB 3.2 Gen1 (SuperSpeed) – Read speed: 455.90 MB/s; Write speed: 463.75 MB/s
  • USB-A port #2 (top, right) – USB 3.2 – USB 3.2 Gen1 (SuperSpeed) – Read speed: 455.89 MB/s; Write speed: 463.55 MB/s
  • USB-A port #3 (bottom, left) – USB 3.2 – USB 3.2 Gen1 (SuperSpeed) – Read speed: 455.98 MB/s; Write speed: 464.15 MB/s
  • USB-A port #4 (bottom, right) – USB 3.2 – USB 3.2 Gen1 (SuperSpeed) – Read speed: 455.97 MB/s; Write speed: 464.51 MB/s

Since nothing is marked on the mini PC, you’ll just have to remember that the faster ports are all located on the front panel. Note that the USB-C port is advertised as a “USB 3.2 Gen2×2 port up to 20 Gbps”, but we couldn’t test this with our peripherals.

The Kron Mini K5 supports up to three independent displays via HDMI 2.0, DisplayPort 1.4, and USB-C with DP 1.4 Alt Mode.

We verified this by connecting three displays to the mini PC: a KTC A32Q8 32-inch 4K Smart Monitor via DisplayPort 1.4, a CrowView 14-inch laptop monitor via HDMI 2.0, and a CrowVi 15.6-inch display via USB-C. All three screens worked fine in extended desktop mode.

ACEMAGIC Kron Mini K5 (Intel Core 3 304) benchmarks on Windows 11 Pro

Before running benchmarks on the mini PC, we set the Power Mode to Best Performance. Note that the tests were performed in a room with an ambient temperature of about 28–29°C, so your own results may differ depending on room temperature and system settings.

The Kron Mini K5 mini PC scored 5,309 points in PCMark 10, broken down into Essentials (8,266 points), Productivity (9,758 points), and Digital Content Creation (5,035 points).

3DMark Fire Strike is a CPU+GPU workload. The system scored 1,880 points.

The system achieved an overall score of 2,686.2 points in PassMark PerformanceTest 11.

The Disk Mark score is a little lower than a typical score in a mini PC with an NVMe SSD. CrystalDiskMark was used to confirm the mid-range performance of the included 512GB NVMe SSD, which reached 1,780.49 MB/s for sequential reads and 1,675.72 MB/s for sequential writes.

Cinebench R23 is useful to evaluate both single-core and multi-core CPU performance and get a first idea of the mini PC’s cooling ability.

The Kron K5 mini PC scored 1,729 points in the single-core test and 5,027 points in the multi-core benchmark, with an MP ratio of 2.91x on a heterogeneous (1P+4E) penta-core processor.

Let’s test the GPU with Unigine Heaven Benchmark 4.0. The system averaged 18.8 FPS and scored 473 points at the standard 1920×1080 resolution, confirming that the Intel Xe3 Graphics in the Core 3 304 is better suited to general graphics work and everyday use than demanding 3D gaming.

Next up is 4K and 8K YouTube playback in Firefox, tested at 30 and 60 FPS.

The results showed that the Kron Mini K5 could play YouTube videos from 4K 30 FPS through 8K 60 FPS smoothly, with no obvious stuttering and no (or few) dropped frames. That’s the same result as on the Beelink EQi Wildcat Lake mini PC, confirming the Intel Core Series 3 processors are solid choices for video playback.

The Intel Core 3 304 SoC comes with Intel Xe3 Graphics delivering up to 9 TOPS of AI performance, and an Intel AI Boost NPU delivering up to 15 TOPS. We therefore tested performance with Geekbench AI using OpenVINO on the GPU and NPU.

On the GPU, the Kron K5 mini PC achieved 2,752 points for the single-precision score, 9,722 points for the half-precision score, and 11,787 points for the quantized score. The full results are available on the Geekbench website.

On the NPU, the scores were: 850 points for single precision, 16,657 points for half precision, and 26,212 points for quantized. The full results are available on the Geekbench website. The results show that the NPU has a clear advantage with quantized models.

ACEMAGIC Kron Mini K5 Windows 11 benchmarks comparison against other mini PCs

Now that we have benchmark data, let’s compare the Windows 11 benchmark results of the ACEMAGIC Kron mini K5 against other entry-level to mid-range mini PCs: Beelink EQi Wildcat Lake and GEEKOM A5 Pro 2026 Edition (AMD Ryzen 5 7530U).

Here are the basic specifications for the three systems

Let’s compare the benchmark results

As one would expect, the two Intel Core 3 304 mini PCs have similar scores, but there are still some interesting differences.  The slower 5600 MT/s LPDDR5 memory on the Kron K5 leads to a lower Memory Mark score and appears to somewhat negatively impact 3D graphics and AI scores compared to the Beelink EQi Wildcat Lake. While the NVMe SSD on the ACEMAGIC mini PC is fairly low-end, it’s still faster than the UFS 3.1 storage on the Beelink device.

The GEEKOM A5 Pro 2026 Edition mini PC performs better with its AMD Ryzen 5 7530U 6-core/12-thread processor, especially on multi-core and 3D graphics workloads. The NVMe SSD is also significantly faster. It does cost $100+ more than the Wildcat Lake systems.

Network performance testing (2.5GbE and Wi-Fi 6)

We tested the 2.5GbE port of the Kron Mini K5 (Intel Core 3 304) mini PC using the iperf 3.21 utility and an UP Xtreme i11 Edge mini PC (192.168.31.12) on the other end of the connection.

• Download

PS C:\Users\aey\Downloads\iperf3.21_64\iperf3.21> .\iperf3.exe -t 60 -c 192.168.31.12 -i 10 -R
Connecting to host 192.168.31.12, port 5201
Reverse mode, remote host 192.168.31.12 is sending
[  5] local 192.168.31.116 port 61912 connected to 192.168.31.12 port 5201
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-10.00  sec  2.76 GBytes  2.37 Gbits/sec
[  5]  10.00-20.00  sec  2.76 GBytes  2.37 Gbits/sec
[  5]  20.00-30.01  sec  2.77 GBytes  2.37 Gbits/sec
[  5]  30.01-40.00  sec  2.76 GBytes  2.37 Gbits/sec
[  5]  40.00-50.01  sec  2.76 GBytes  2.37 Gbits/sec
[  5]  50.01-60.01  sec  2.76 GBytes  2.37 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-60.01  sec  16.6 GBytes  2.37 Gbits/sec    0            sender
[  5]   0.00-60.01  sec  16.6 GBytes  2.37 Gbits/sec                  receiver

iperf Done.

• Upload

PS C:\Users\aey\Downloads\iperf3.21_64\iperf3.21> .\iperf3.exe -t 60 -c 192.168.31.12 -i 10
Connecting to host 192.168.31.12, port 5201
[  5] local 192.168.31.116 port 62092 connected to 192.168.31.12 port 5201
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-10.01  sec  2.77 GBytes  2.38 Gbits/sec
[  5]  10.01-20.01  sec  2.76 GBytes  2.37 Gbits/sec
[  5]  20.01-30.00  sec  2.76 GBytes  2.37 Gbits/sec
[  5]  30.00-40.01  sec  2.76 GBytes  2.37 Gbits/sec
[  5]  40.01-50.01  sec  2.76 GBytes  2.37 Gbits/sec
[  5]  50.01-60.00  sec  2.76 GBytes  2.37 Gbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-60.00  sec  16.6 GBytes  2.37 Gbits/sec                  sender
[  5]   0.00-60.02  sec  16.6 GBytes  2.37 Gbits/sec                  receiver

iperf Done.

The 2.5GbE port reached 2.37 Gbps in both upload and download. All good.

We added a Xiaomi Mi Router AX6000 to the mix to test Wi-Fi 6 5 GHz.

• Download

PS C:\Users\aey\Downloads\iperf3.21_64\iperf3.21> .\iperf3.exe -t 60 -c 192.168.31.12 -i 10 -R
Connecting to host 192.168.31.12, port 5201
Reverse mode, remote host 192.168.31.12 is sending
[  5] local 192.168.31.128 port 57583 connected to 192.168.31.12 port 5201
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-10.01  sec   983 MBytes   824 Mbits/sec
[  5]  10.01-20.01  sec  1007 MBytes   845 Mbits/sec
[  5]  20.01-30.01  sec  1018 MBytes   854 Mbits/sec
[  5]  30.01-40.01  sec   973 MBytes   816 Mbits/sec
[  5]  40.01-50.00  sec  1023 MBytes   859 Mbits/sec
[  5]  50.00-60.01  sec   987 MBytes   827 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate         Retr
[  5]   0.00-60.02  sec  5.85 GBytes   838 Mbits/sec   64            sender
[  5]   0.00-60.01  sec  5.85 GBytes   838 Mbits/sec                  receiver

iperf Done.

• Upload

PS C:\Users\aey\Downloads\iperf3.21_64\iperf3.21> .\iperf3.exe -t 60 -c 192.168.31.12 -i 10
Connecting to host 192.168.31.12, port 5201
[  5] local 192.168.31.128 port 57573 connected to 192.168.31.12 port 5201
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-10.01  sec   962 MBytes   807 Mbits/sec
[  5]  10.01-20.00  sec   979 MBytes   821 Mbits/sec
[  5]  20.00-30.01  sec  1.00 GBytes   860 Mbits/sec
[  5]  30.01-40.00  sec   944 MBytes   792 Mbits/sec
[  5]  40.00-50.02  sec  1006 MBytes   843 Mbits/sec
[  5]  50.02-60.01  sec  1.00 GBytes   860 Mbits/sec
- - - - - - - - - - - - - - - - - - - - - - - - -
[ ID] Interval           Transfer     Bitrate
[  5]   0.00-60.01  sec  5.80 GBytes   831 Mbits/sec                  sender
[  5]   0.00-60.03  sec  5.80 GBytes   830 Mbits/sec                  receiver

iperf Done.

Download speed was 838 Mbps and upload speed was 830 Mbps, which is a decent result for Wi-Fi 6 and sufficient for high-resolution video streaming, file transfers, and general network use. The maximum link speed of 573 Mbps reported in HWiNFO 64 appears to be erroneous…

Thermal performance

We ran 3DMark Fire Strike CPU+GPU workload on the mini PC to test it for potential CPU throttling. The maximum CPU temperature was about 90°C, and HWiNFO 64 detected thermal throttling on all Efficient cores, but not on the Performance core. This is different behaviour from the Beelink EQi Wildcat Lake, which didn’t have any thermal throttling, mainly because of the lower PL1 thermal limit (15W vs 22W), and the Beelink mini PC exceeded power limits instead.

After rebooting the system to run Cinebench R23 multi-core, the maximum CPU temperature rose to about 92°C, and thermal throttling was observed on all five cores this time around.

The different thermal strategy doesn’t seem to impact performance that much. However, the 15W configuration should be better for sustained loads. We also ran the test in a room at 28-30°C, so it’s possible that in a cooler room (20°C), the ACEMAGIC Kron Mini K5 wouldn’t have reached thermal limits.

Fan noise

The Kron Mini K5’s fan is not particularly noisy at idle and during light workloads, but under high load the fan becomes noticeably louder. We used a sound level meter placed about 5 cm above the top of the chassis.

• Idle – 37.4 – 38.0 dBA
• 3DMark (Fire Strike) – 38.0 – 41.6 dBA, rising to 41.2–46.0 dBA during heavy scenes
• Cinebench R23 multicore – 45.7 – 47.0 dBA

The test room had a background noise level of about 36.5–37.0 dBA. For reference, the Beelink EQi Wildcat Lake mini PC was measured at just 39.8 – 40.4 dBA when running the Cinebench R23 multicore benchmark. So the Beelink is a better system if you need a relatively quiet device, and it’s one of the quietest actively-cooled mini PCs we’ve ever reviewed.

Kron Mini K5’s power consumption

The Kron Mini K5 mini PC can be powered by either a 19V/3.42A adapter through a DC jack or a USB-C PD adapter.

We used a wall power meter to measure power consumption with the provided 19V power adapter:

• Power off – 0.7 – 0.8 W
• Idle – 3.7 – 4.3 W
• YouTube 8K 60fps in Firefox – 12.5 – 15.9 W
• 3DMark (Fire Strike) – 13.0 – 17.9 W (during combat scenes)
• Cinebench R23 Multi-Core
  • First few seconds: 35.7 – 37.1 W
  • Sustained: 29.7 – 33.0 W

We recently received a Sabrent Voltik 252W 8-port charging station with a built-in per-port power meter function, so we used it to measure the power consumption through USB-C. This is a bit different, since the losses in the charging station are not part of the measurements, and the values are somewhat lower. It’s better if you want to focus on the mini PC’s power consumption per se, but not if the purpose is to evaluate the electricity costs.

USB-C power measurements:

• Powered off – 0.2-0.4 W
• Idle – 2.8 – 3.4 W
• YouTube 8K 60fps in Firefox – 11.6 – 17.5 W
• 3DMark (Fire Strike) – 12.0 – 18.9 W (during combat scenes)
• Cinebench R23 multicore
  • First few seconds: 31.1 – 33.7 W
  • Sustained: 26.7 – 32.7 W

Note that during testing, the mini PC was connected to Wi-Fi 6, a USB-A RF dongle for a wireless keyboard, a USB mouse, and a 32-inch 4K KTC A32Q8 monitor (HDMI 2.0).

Conclusion

The ACEMAGIC Kron Mini K5 mini PC works well with Windows 11 Pro. It delivers strong single-core performance, 2.5GbE and WiFi 6 networking, supports up to three 4K displays through HDMI, DP, and USB-C ports, and plays 4K and 8K YouTube videos smoothly up to 60 FPS. The Intel Core 3 304 is also suitable for light AI workloads, although it doesn’t qualify as a Copilot+ PC with a 9 TOPS GPU and 15 TOPS NPU.

In many ways, it offers similar performance and features as the Beelink EQi Wildcat Lake Core 3 304 mini PC. On the downside, it lacks high-speed interfaces like 10GbE and Thunderbolt 4/USB4, the memory has lower capacity and bandwidth (12GB LPDDR5-5600 vs 16GB LPDDR5-6400), and its fan is not as quiet as the one on the Beelink mini PC. A different power limit configuration (22W vs 15W PL1) means the mini PC hit thermal limits instead of power limits in our stress tests. While the 512 GB NVMe SSD included in the ACEMAGIC Kron Mini K5 can be considered “mid-range”, it’s still faster than the 512 GB UFS found in the Beelink EQi Wildcat Lake. Having said that, both have two NVMe SSD sockets, so upgrades are possible.

In the third part of the review, we will install and test Ubuntu 26.04 on the Kron Mini K5 mini PC to see how Linux performs on the Intel Core 3 304 mini PC.

We’d like to thank ACEMAGIC for sending the Kron Mini K5 (Intel Core 3 304) mini PC for review. It is sold for $399.99 on Amazon, a fair price compared to the $469 asked for the Beelink EQi Wildcat Lake (16GB RAM, 10GbE, and Thunderbolt 4) and the $380 price for the BOSGAME ECO with a similar 12GB/512GB configuration, but a different port layout.

CNXSoft: This article is a translation – with some additional insights – of the original review on CNX Software Thailand by Suthinee Kerdkaew.

The post ACEMAGIC Kron Mini K5 Review – Part 2: A $400 Intel Core 3 304 Wildcat Lake mini PC tested with Windows 11 Pro appeared first on CNX Software - Embedded Systems News.

OpenTrailPaper is an open-source, DIY e-paper bike computer project comprised of firmware for the ESP32-S3-based T5 E-Paper S3 Pro devkit and Android and iOS companion apps for maps, routes, and settings.

It shows ride data and offline maps, follows standard GPX (GPS Exchange Format) routes, records FIT (Flexible and Interoperable Data Transfer used in Garmin devices) files, and connects to Bluetooth (and soon ANT/ANT+) heart rate, power, and cadence sensors.

Before looking into the project’s details, let’s check out the LILYGO T5 E-Paper S3 Pro specifications:

• ESP32-S3-WROOM-1 wireless module
  • SoC –  ESP32-S3N16R8 dual-core Tensilica LX7 microcontroller @ up to 240 MHz with
    • 2.4 GHz 802.11n WiFi 4 and Bluetooth 5.0 LE connectivity
    • Memory – 8MB PSRAM
    • Storage – 16MB SPI flash
  • PCB antenna
• Storage – MicroSD card slot
• Display – 4.7-inch e-Paper display with 960×540 resolution, 16 shades of gray, capacitive touch (GT911); ED047TC1 drive IC
• Wireless
  • 2.4 GHz 802.11n WiFi 4 and Bluetooth 5.0 LE connectivity (ESP32-S3)
  • LoRa module (SX1262) supporting 433 MHz, 868 MHz, and 915 MGz bands
  • GNSS – u-blox MIA-M10Q or Quectel L76K module (autodetected by the firmware)
• USB – USB Type-C port for power and programming
• Expansion – 4-pin Qwicc connector
• Misc
  • Power, Reset, Boot, and User (IO48) buttons
  • PCF8563 RTC
• Power Supply
  • 5V via USB Type-C port
  • 3.7V, 1,500mAh battery good for under 8 hours per charge with OpenTrailPaper firmware and front light off; (note: might be optimized)
  • BQ25896 charger and BQ27220 fuel gauge
• Dimensions – 129 x 69 x 11 mm

The LILYGO T5 ships with either the company’s factory firmware or MeshCore for off-grid communication and location sharing. If you’re going to flash OpenTrailPaper on the device, simply purchase the LILYGO firmware version since you’ll save a few dollars that way. You’ll find additional technical details and code samples for low-level hardware features of the T5 E-Paper S3 Pro on GitHub.

Back to the OpenTrailPaper e-paper bike computer project specifically. The project’s GitHub repository features the source code for the ESP32-S3 firmware based on the ESP-IDF framework, the OpenTrailPaper Android and iOS companion apps, as well as tools to create compatible maps from a smartphone using the companion app, a Python script in a terminal, or JavaScript code running in a web browser. Everything has been developed with Anthropic Claude as an assistant, but with human supervision, and it’s a working project, albeit with some caveats for now (more on that below). Further documentation can also be found on the project’s website.

The easiest way to get started is to go to the setup page in a browser supporting the Web Serial API. You’ll go through the instructions to flash the firmware over USB and build offline map tiles stored on a microSD card.

The companion app can plan and transfer routes, build offline maps, copy recorded rides, change settings, and install firmware updates over Bluetooth. The bike computer itself does not need the phone during a ride.

Here’s what runs on the bike computer (during a ride) and the smartphone (before or after a ride).

One of the main reasons for getting an e-paper bike computer is to be able to clearly see data and the map while riding outdoors. I don’t usually need GPS when I ride a bicycle where I live, but when I travel, I typically connect my smartphone running the MAPS.me offline maps app to a bicycle stand to get directions. In most cases, it’s not super convenient to use since it’s unreadable unless I find/make some shadows.

To reproduce the project, you’ll need to purchase the LILYGO T5 E-Paper S3 Pro for $90 on Amazon or $103 on AliExpress, as well as find a way to mount it to your bicycle. A phone bicycle stand should probably do. Bear in mind that there are a few limitations:

• The T5 S3 Pro is not waterproof, so you may have to put it in a ZIP bag to protect it from rain.
• No pressure sensor. Climbing is derived/estimated from elevation data in the map tiles
• GPS could be improved – A multi-band GPS module would lock on faster and hold a fix better in various conditions
• No magnetometer – The map only orients using direction of travel, so it doesn’t work while you’re stopped.
• Impractical, small hardware buttons and the touch panel might not work well with winter gloves or in the rain.

The post OpenTrailPaper transforms LILYGO T5 E-Paper S3 Pro devkit into a DIY e‑paper bike computer appeared first on CNX Software - Embedded Systems News.

Arduino has just announced the release of ArduinoCore-Zephyr version 1.0.0, with new hardware support, including the VENTUNO Q SBC, an updated Zephyr 4.4.1 base, support for the Nicla Vision camera, and various fixes and improvements.

Arduino first announced its plans to switch from Arm Mbed to Zephyr RTOS in July 2024, before releasing the first beta of Arduino Core for Zephyr in December of the same year. ArduinoCore-Zephyr 1.0.0 is not the first stable version, since that was version 0.90.0 released on July 29, 2026. The new release builds on that, so let’s see what’s new.

ArduinoCore-Zephyr 1.0.0 highlights

• Support for Arduino VENTUNO Q SBC, specifically the STM32H5F5 real-time MCU on the board
• Update to Zephyr 4.4.1 with the latest upstream improvements, security fixes, and driver support
• Added Nicla Vision camera support to capture and process image data directly through the Zephyr core
• Various bug fixes and CI improvements
• Currently supported platforms from Arduino
  • UNO Q board
  • VENTUNO Q board
  • Portenta H7 board
  • Portenta C33 board
  • Opta microPLC
  • GIGA R1 WiFi board
  • Nano Matter board
  • Nano RP2040 Connect board
  • Nano 33 BLE board
  • Nicla Sense ME module
  • Nicla Vision module

You’ll find the full list of changes on the project’s GitHub release page.

If you want to check it out, download and install the latest version of the IDE, update the board files if requested, and install or update Arduino Zephyr Boards to 1.0.0. If you want to test the latest release on the Renesas EK-RA8D1, NXP FRDM-MCXN947, or NXP FRDM-RW612 board, install Zephyr Community Boards, also maintained by Arduino.

That’s about all you have to do in the IDE before building the sketch. The Arduino IDE should automatically upload and install the Zephyr loader when uploading the firmware for the first time, but in some cases, you may need to manually install the Zephyr loader onto your board. A manual update appears to be mandatory for the Portenta C33.

The post ArduinoCore-Zephyr 1.0.0 adds support for Arduino VENTUNO Q, updates to Zephyr 4.4.1, and more appeared first on CNX Software - Embedded Systems News.

If you are not into ducks, but still would like to play around with physical AI, the Petoi Quaddle is a mini robot dog equipped with 4 servos controlling its legs and is designed to help people learn robotics, coding, AI, and 3D printing.

The robot is powered by an ESP32-S3 motion core for servo control and wireless connectivity, and an optional ESP32-S3 AI core to enable voice commands and Smart Home control. The robot can walk, strafe, spin, talk back, and react to the user’s touch.

Three models are available depending on the user’s requirements: Builder, Buddy, and Scout.

A Raspberry Pi Zero can also be mounted on the top in place of the ESP32-S3-based AI core. Petoi Quaddle runs OpenCat open-source firmware (Arduino) already used in the earlier Petoi Bittle robot dog, first introduced in 2020. The robot can be programmed in a visual programming IDE with drag-and-drop blocks, as well as Python, MicroPython, Arduino, C++, and ROS. A Raspberry Pi Zero (UART) or a computer (Wi-Fi or Bluetooth) is needed for the latter.

Other highlights include compatibility with LEGO and 3D-printed shells, partnerships with TuyaOpen for Smart Home integration, the ability to program AI Code to use ESPHome or Xiaozhi AI, mobile apps for control, and more. The documentation is already up but under construction, and you can already play around with the Quaddle simulator.

Petoi has launched the Quaddle on Kickstarter with a $50,000 funding target. The three main rewards are as follows:

• Quaddle Builder – $99 ($149 MSRP)
• Quaddle Buddy – $139
• Quaddle Scout – $199

You can order it as a kit or preassembled. Shipping adds $15 to most of the world and $18 to the European Union. The company expects to ship rewards in December 2026.

The post Petoi Quaddle – A mini robot dog for physical AI experimentation (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

CNXSoft: This is a guest post by Pavel Burgr, who recently added Puya PY32 MCU support to TinyGo 0.42, the latest version of the Go compiler for microcontrollers. We first covered TinyGo in 2019, and the project is still going strong with 17.7k stars, 1.1k forks, and 243 contributors. I just made a few edits and added illustrations and a few links to the article.

TinyGo 0.42 adds support for the Puya PY32 family of 32-bit Arm microcontrollers. The new port includes startup code, generated device-register definitions, linker scripts, GPIO and UART support, and targets for 87 individual MCU configurations. It also includes board definitions for two inexpensive EmbedFire development boards available from AliExpress.

The port was contributed by Pavel Burgr and merged through TinyGo pull request #5106. One motivation was to establish a solid, easy-to-use firmware platform for small IoT devices built around the Registry service and the BleRiot low-power radio protocol (more on that at the end of the post). Those projects are separate from TinyGo, but they show the sort of resource-constrained application the new targets are intended to enable.

Broad PY32 target coverage

TinyGo is a Go compiler for microcontrollers, WebAssembly, and other constrained systems. It uses LLVM to produce compact native binaries and provides a hardware abstraction layer through packages such as machine, while retaining familiar Go syntax and tooling. Running Go on a part with only 2 KB of SRAM may sound a little like a dare, but this is precisely the sort of dare TinyGo was made for.

The PY32 port covers both Cortex-M0+ and Cortex-M4 devices. The target table contains 87 concrete memory configurations across 40 device-definition families, ranging from parts with 8 KB of flash and 2 KB of SRAM to devices with 512 KB of flash and 144 KB of SRAM.

Supported series include:

• PY32E407
• PY32F001, F002, F003, F021, F030, F031, F032, F040, F071, F072, F403, and F410
• PY32L020 and L090
• PY32M010, M020, M030, M031, and M070
• PY32MD310, MD320, MD410, MD420, and MD430
• PY32T020, T090, and T092

Each chip target selects the correct flash and SRAM map. Peripheral definitions are generated from Puya’s CMSIS data through the supporting py32-svd repository, giving driver authors typed register names in device/py32 instead of private address tables.

The machine API in TinyGo 0.42 covers clock and timer startup, GPIO, and UART, including a second UART on selected parts. ADC, I2C, SPI, and PWM do not yet have portable machine drivers, but their generated registers are available to low-level code.

Two supported EmbedFire boards

Two small Puya32 development boards have dedicated targets and pin aliases, making them the simplest way to try the port. Both are available for under $2 on AliExpress, and the PY32F030 board goes for $7 and up on Amazon.

The PY32F030 board maps LED1, LED2, and LED3 to PA2, PA3, and PA4, and its two keys to PA5 and PA6. Its default UART uses PA7 for TX and PA8 for RX. The smaller PY32F002B board maps the LEDs to PA1, PA5, and PA4, the keys to PA3 and PA0, and the default UART to PA6 and PA7.

Both boards define machine.LED as an alias for the second onboard LED. As embedded tradition demands, the same first program can perform the ceremonial blink on either target:

package main

import (
	"machine"
	"time"
)

func main() {
	led := machine.LED
	led.Configure(machine.PinConfig{Mode: machine.PinOutput})

	for {
		led.Set(true)
		time.Sleep(500 * time.Millisecond)
		led.Set(false)
		time.Sleep(500 * time.Millisecond)
	}
}

With a supported SWD probe connected and pyOCD configured for the selected Puya device, the program can be built and flashed with one command, with no Makefile archaeology required:

tinygo flash -target=embedfire-py32f030 .

For the smaller board, only the target changes:

tinygo flash -target=embedfire-py32f002b .

The board targets inherit the exact py32f030x8 and py32f002bx5 memory maps and invoke the matching pyOCD target during flashing. Bare-chip targets are also available for custom boards. A small TinyGo PY32 project template provides additional blink examples and a starting Makefile.

From low-cost MCU to Registry-connected IoT node

With the traditional blink out of the way, the practical motivation behind the port is more interesting. Pavel developed the PY32 support while building BleRiot, a simple and reliable request/response protocol for low-power RF nodes.

Each node exposes signed 32-bit registers and understands just two operations: GET and SET. That is the whole vocabulary, which is a feature rather than an omission. The Linux hub polls registers and handles names and unit conversions as well as retries and timeouts. This small, predictable interface keeps firmware compact, while recovery from a dropped packet is handled once in the hub instead of by every application.

One hub communicates with many nodes over a custom 250 kbps GFSK radio link with BLE-compatible packet framing, although it is not Bluetooth LE or GATT. The reference node pairs a PY32F030x8 with a PAN211x 2.4 GHz radio, while the hub uses a simple USB radio dongle.

BleRiot publishes the node values to Registry, a lightweight local service that gives each register a name, current value, optional metadata, and a time-to-live. Stale values expire automatically. Applications can use its HTTP/JSON API directly; the supplied Node-RED nodes combine with standard MQTT nodes for bridging, and a built-in Prometheus endpoint exposes numeric registers as gauges ready for Grafana dashboards. In other words, getting a temperature onto a dashboard does not require teaching a 3 KB MCU about Grafana.

A controller follows the same path in reverse when it requests a change. BleRiot shares device types and register definitions between the TinyGo firmware and Linux hub, and its tooling can generate a node identity, build the firmware with TinyGo, and flash it over SWD.

Registry and BleRiot are not required to use TinyGo on PY32, and the PY32 support is not tied to a particular radio. They are useful examples of why low-cost microcontrollers benefit from a maintained compiler, reproducible chip targets, and a standard Go-facing hardware API.

Availability and next steps

PY32 support is available in TinyGo 0.42.0. The two EmbedFire targets provide the most direct entry point, while the broader chip list gives developers a foundation for custom boards ranging from tiny sensor nodes to larger Cortex-M4 controllers.

TinyGo 0.42 emphasizes runtime stability, correct memory maps, GPIO, and serial communication rather than claiming complete peripheral coverage. Future contributions can add portable ADC, I2C, SPI, PWM, and board definitions on top of the generated device/py32 register layer. For developers already using Go on the host side of an IoT system, the port makes Puya’s inexpensive MCU family a much more approachable endpoint.

CNXSoft note: TinyGo support for AT32 MCUs is apparently next.

The post TinyGo 0.42 adds support for 87 Puya PY32 MCU variants and two EmbedFire boards appeared first on CNX Software - Embedded Systems News.

Modern websites rarely work in isolation. They rely on third-party services to exchange data. API integration into website projects can improve user experience.

From payment gateways to CRM platforms, almost every feature can be integrated. Build your successful API integration solutions today.

What Is API Integration and Why Does Your Website Need It?

An API allows two software applications to communicate with each other. They can exchange information securely to facilitate the production process. It is convenient because you don’t have to develop every feature. You can simply connect your websites with specialized services.

Common API integrations include:

• Payment gateways
• Social media integrations
• Customer relationship management (CRM) systems
• Supplier portals
• Email marketing platforms
• Weather services
• Maps and location tools

For many businesses, API integrations are a way to:

• Save development time
• Utilize advanced applications at more affordable rates
• Provide a richer customer experience

What Should Businesses Consider Before Integrating an API?

API integrations may seem like a no-brainer. Still, you need to consider your technical and operational requirements to launch them effectively. Depending on the situation, you need to deploy different technicians for the integration.

Define your needs in areas like:

• Business objectives
• API documentation quality
• Authentication methods
• Security requirements
• Data privacy compliance
• System compatibility
• Scalability
• Vendor reliability
• Risk management

The blueprint can reduce operational risks. It is crucial for achieving better project outcomes.

How to Integrate an API Into a Website Step by Step

Understanding how to integrate an API into a website becomes straightforward when there is a structured process.

Step 1: Define your requirements

Determine the functionality you expect from the APIs. What should they provide? How should they achieve your objectives? Before you rush into selecting APIs, evaluate your current pain points. Identify weaknesses and the expected improvements.

Step 2: Select the right APIs

There are plenty of API solutions. But not every one will fit your project. Aside from choosing one that matches your niche, it should come with:

• Comprehensive documentation
• Reliable support from developers
• Strong performance with a proven track record
• Specialized features that target your pain points

Step 3: Test and implement

Check the authentication requirements. Integrate the APIs into your website with proper configurations. It is essential to test the integration several times. Testing can verify:

• Data accuracy
• Error handling
• Performance
• Security
• Browser compatibility
• Mobile functionality

Step 4: Monitor and adjust

Check how well employees are adjusting to the implementation. Provide sufficient training if they are struggling. Monitor the system’s performance as well. You can make adjustments as you see fit.

What Are the Common Challenges of API Integration?

API integrations can present challenges.

Common obstacles include:

• Incomplete documentation
• Authentication errors
• Data format inconsistencies
• Legacy system compatibility
• Security vulnerabilities
• Version updates

The best way to reduce the risk is to partner with an experienced firm like Atlantic BT. It has the advanced resources and expertise to address different issues. Proper integration testing can reduce performance flaws after the launch.

How Much Does It Cost to Integrate an API Into a Website?

The price range can be huge. It can go from $1,500 to $10,000+. An API integration into a website project covers multiple aspects. Therefore, you can only get an accurate quote after an initial assessment.

A few key deciding factors include:

• Project complexity
• Number of APIs
• Custom development requirements
• Third-party licensing fees
• Security requirements
• Existing website architecture

Simple integrations may take only a few hours to execute. But enterprise-level projects will involve substantial planning and months to release.

More importantly, you should not focus solely on the upfront cost. API integrations can lead to better user experience and operational speed. It is likely to reduce your future production costs as well.

API Integration Best Practices for Businesses

If you are wondering about how to integrate an API into a website, follow these best practices.

• Use secure authentication methods
• Encrypt sensitive data
• Validate all incoming data
• Document every integration
• Monitor API performance regularly
• Plan for API version updates
• Build scalable architecture
• Test thoroughly before deployment

Ongoing maintenance keeps your website competitive. API providers release updates regularly. You need to monitor compatibility and adjust accordingly.

In conclusion, Successful API integrations enable businesses to expand efficiently. You can automate workflows and deliver more satisfactory customer experiences. Speak to an expert to find out suitable API integrations for your company.

The post How to Integrate an API Into a Website: A Step-by-Step Guide for Businesses (Sponsored) appeared first on CNX Software - Embedded Systems News.

Boardcon CM311Y3 system-on-module (SoM) is powered by an Amlogic A311Y3 octa-core Cortex-A78/A55 SoC with an 8 TOPS NPU and targets 4K video surveillance systems, AI edge computing devices, interactive terminals, enterprise thin clients, and autonomous robots.

The module comes with up to 16GB LPDDR5 and up to 256 GB eMMC flash, exposes 210 pins through castellated edges, and a development board is also provided for evaluation and early software development. We don’t often see new SoMs or SBCs based on Amlogic SoCs these days, so let’s have a closer look.

Boardcon CM311Y3 System-on-Module

CM311Y3 specifications:

• SoC – Amlogic A311Y3
  • CPU
    • 2x ARM Cortex-A78 cores @ 2.4GHx
    • 6x ARM Cortex-A55 cores @ 2.0GHz
    • RISC-V core for system control processing
    • RISC-V core for Always-on power management and Sensor Hub
  • GPU – Arm Mali-G625 MC1 with support for OpenGL ES 3.2, Vulkan 1.4, and OpenCL 3.0
  • VPU
    • Video Decoder – H.264 up to 4Kp60 H.265, AV1, VP9, AVS3 up to 4Kp120
    • Video Encoder – H.264/H.265 up to 4Kp60
  • ISP – 16MP resolution
  • NPU – 8 TOPS AI accelerator; supports INT4, INT8, INT16, FP8, FP16, and BF16
  • DSP – HiFi 5
• Memory – 4GB, 8GB, or 16GB LPDDR5/LPDDR5X
• Storage – 32GB, 64GB, 128GB, or 256GB eMMC flash
• Networking – Realtek RTL8211H Gigabit Ethernet PHY
• Castellated edges (210 pins) for carrier board
  • Storage – SD card interface
  • Display Interfaces
    • HDMI Tx
    • MIPI DSI/LVDS(A)
    • eDP/MIPI DSI/LVDS/VX1(B)
  • Camera/Video Input – 2x MIPI CSI, HDMI Rx
  • Audio – S/PDIF output
  • Networking – 2x Ethernet (2.5GbE and GbE from A311Y3 specs, but the carrier board below uses 2x GbE)
  • USB – USB 3.0, USB 3.0 DRD, USB 2.0
  • PCIe – PCIe 3.0 x2
  • Low-speed I/Os – UART, 2x I2C, GPIO, PWM, etc…
  • Misc – SDIO, IR Rx, Debug
• Supply Voltage – 3.4 to 5.5V DC
• Dimensions – 52 x 45 mm (8-layer PCB, 0.9 mm pitch QFN)
• Weight – 10.7 grams

The company provides support for Android 16 with Linux 6.12.69 and U-boot 2025.01, and all necessary drivers.  They also offer a “Ubuntu 22.04 cross-compiler environment”, which probably means a VM image with Ubuntu 22.04, development tools, and the SDK pre-installed.

Idea311Y3 Amlogic A311Y3 development board

Idea311Y3 specifications:

• SoM – Boardcon CM311Y3 module described above soldered on board
• Storage
  • MicroSD card slot
  • NVMe SSD via M.2 M-Key slot
• Display Interfaces
  • HDMI 2.1, up to 4K @ 60Hz
  • 2x MIPI DSI, up to 2560 x 1600 @ 60Hz
• Camera/Video Input
  • HDMI 2.1a input, up to 4K @ 60Hz
  • 2x 4-lane MIPI-CSI
• Audio
  • 3.5mm audio I/O jack
  • 2x speakers via 12-pin pluggable terminal block
  • 8-channel audio via HDMI
• Networking
  • 2x Gigabit Ethernet RJ45 ports via Realtek RTL8211F-CG controllers (one on module, one on carrier board)
  • WiFi 5 (802.11ac/a/b/g/n) and Bluetooth 5.4
  • Optional 4G LTE via mini PCIe socket and nano SIM card slot
• USB – 2x USB 3.0 Type-A ports
• Serial port
  • 3-pin serial port for debug
  • RS232 via 12-pin pluggable terminal block
  • RS485 via 12-pin pluggable terminal block
• Misc
  • Reset, Power_EN, and Power buttons
  • RTC with battery connector
  • 2-pin header for USB Boot
• Power Supply – 12V/3A DC input via power barrel jack
• Dimensions – 155 x 115 mm

The development board is supported by the same SDK as the module, and the company can also provide various accessories, namely an EC20 mini PCIe LTE module, an IMX415 8.29 MP CMOS camera module, an IMX307 2.13MP CMOS camera module, a 12V/3A power adapter (EU/US),  and a 10.1-inch 1280 x 800 display (MIPI DSI or LVDS).

Boardcon hasn’t provided any pricing information for the new Amlogic A311Y3 system-on-module and development board. Additional information may be found on the product page.

The post Amlogic A311Y3 Cortex-A78/A55 Edge AI system-on-module delivers up to 8 TOPS appeared first on CNX Software - Embedded Systems News.

Artery Technology AT32F406 and AT32F408 are two new Arm Cortex-M4F MCUs running at up to 216 MHz, with the F408 adding USB OTG High-Speed and an integrated PHY for industrial automation, IoT, USB accessories, and gaming peripherals.

The AT32F406/F408 series both offer DSP instructions and a floating-point unit (FPU), integrate up to 512 KB Flash and 192 KB SRAM, and a 4 KB OTP memory for critical data and parameters. The microcontrollers also offer a QSPI interface for external Flash or RAM expansion, and a range of digital and analog peripheral interfaces.

Artery AT32F406/AT32F408 specifications:

• MCU – 32-bit Arm Cortex-M4F CPU up to 216 MHz; also features Memory Protection Unit (MPU) and DSP instructions
• Memory/Storage
  • 128 KB to 192 KB SRAM
  • 256 KB to 512 KB Flash memory
  • 26 KB boot memory (configurable as a bootloader or general instruction/data memory)
  • 4 KB OTP (one-time programmable) memory
  • QSPI interface for external SPI Flash or SPI RAM with address mapping support
• Peripherals
  • Up to 57 fast GPIOs (AT32F406), 54 GPIOs (AT32F408); almost all 5V-tolerant
  • USB – USB 2.0 OTG High-Speed (OTGHS) controller with on-chip PHY and a dedicated 4 KB buffer (AT32F408 only).
  • Low-speed I/Os
    • Up to 3x I2C interfaces (SMBus/PMBus support)
    • Up to 8x SPI interfaces (up to 36 Mbit/s) with multiplexed half-duplex I2S
    • Up to 6x USART and 2x UART interfaces (supporting LIN, IrDA, RS485 drive enable, and ISO7816)
  • Analog
    • 3x 12-bit 5.33 MSPS A/D converters with up to 16 external input channels, configurable resolution (12/10/8/6-bit), and hardware oversampling up to 16-bit equivalent
    • Dedicated VREF+ pin (-7R type LQFP64 package only)
    • Internal temperature sensor and internal reference voltage
    • EMUXC – External Multiplexer Controller (Enables fast ADC scanning without CPU intervention)
  • Timers
    • 16-bit 8-channel advanced timer with complementary channels for PWM and dead-time generator
    • Up to 7x 16-bit and 1x 32-bit general-purpose timers.
    • 2x 16-bit basic timers.
    • 2x watchdog timers (general and windowed) and 1x 24-bit SysTick timer
    • Enhanced RTC (ERTC) with auto-wakeup, subsecond accuracy, and hardware calendar
  • Clock – 48 MHz High-speed Internal Clock (HICK)
• Misc
  • Infrared transmitter (IRTMR)
  • 2x 7-channel DMA controllers
  • CRC calculation unit
  • 96-bit unique ID
  • SWD and SWO debug interfaces
  • Power-on reset (POR), low voltage reset (LVR), and power voltage monitor (PVM)
• Power – 2.4 V to 3.6 V
• Packages – LQFP64 (7×7 mm); QFN48 (6×6 mm)
• Temperature Range – -40°C to +105°C

On the software side, the AT32F406/F408 are supported by Artery’s AT32 IDE, VS Code extension, Workbench, AI Studio, and AT-Link tools. They also include sLib, which protects selected Flash memory areas from being read while still allowing code execution, helping protect proprietary software. Artery also provides BSP libraries, documentation, and an AT32F405-to-F408 migration guide.

To help developers get started quickly, Artery has released the AT-START-F406 and AT-START-F408 evaluation boards alongside the new MCUs. They feature Arduino-compatible headers to stack existing shields and prototype quickly. At the time of writing, I could only find documentation for both development boards, but purchasing information is not available.

Artery is not exactly well known, but we did write about them in 2023, as a “Black Pill” clone board featured the company’s AT32F403ACGU7 Cortex-M4F MCU, itself a drop-in replacement (but not a direct clone) of ST’s STM32F103 series. The new AT32F406/AT32F408 parts don’t seem to offer similar compatibility against any STM32 SKUs.

The AT32F406 and AT32F408 MCUs are currently in sampling, with mass production underway. More details and documentation are available on the respective product pages and in the press release.

The post Artery AT32F406 and AT32F408 Cortex-M4F MCUs feature optional high-speed USB OTG PHY appeared first on CNX Software - Embedded Systems News.

Layered Ventures LLC has launched the TickrCast, an expandable HUB75 LED ticker display built around an ESP32-S3. The TickrCast ships as a full consumer product designed to display live esports data, traditional sports scores, social media metrics, and flight tracking across a daisy-chained array of panels.

Unlike a typical LED matrix development board, the system is a complete product with a custom controller board, plug-and-play panels, power connectors, a companion app, and an expandable enclosure. The display starts with two panels and can be expanded by adding additional panels, with the company targeting configurations of 10 or more panels. The system is controlled through Wi-Fi, while Bluetooth LE is used for initial provisioning, and the manufacturer provides more than 30 plugins at launch with no subscription required. The platform also supports signed over-the-air (OTA) firmware updates and hardware flash encryption for improved device security.

TickrCast controller specifications:

• ESP32-S3-WROOM-1U wireless module
  • SoC –  ESP32-S3
    • CPU – Dual-core Tensilica LX7 microcontroller @ up to 240 MHz
    • Memory – 8MB PSRAM
    • Storage – 16MB SPI flash
    • Wireless – 2.4 GHz 802.11n WiFi 4 and Bluetooth 5.0 LE connectivity
  • IPEX antenna connector
• Display – HUB75 based LED matrix panels, 64×32 resolution per panel
• Connectivity – 2.4 GHz Wi-Fi and Bluetooth LE (Via ESP32-S3 module)
• USB – USB Type-C port located on the left edge for power and data/programming
• Expansion – Modular design starting at 2 panels, expandable to 10+ panels (panels bolt together for easy servicing)
• Misc – EN (Reset) and BOOT buttons
• Power – High-power 5V-15A JST box-type connector for power delivery
• Dimensions – TBD

The TickrCast runs custom ESP32-S3 firmware for HUB75 matrices, with a web flasher, web dashboard, and mobile apps (coming soon to the App Store & Google Play) for setup and control, along with signed OTA updates, flash encryption, and a plugin system for live esports, sports, social, music, news, weather, flights, and custom graphics. The company says the platform is already running on a 24/7 soak-test fleet, with Bluetooth provisioning, local content caching, a live plugin marketplace, and 60+ official plugins, while developers can publish their own plugins and use an MCP hook to generate them from plain-English requests.

The creator, Joseph Capehart, mentions that handling Wi-Fi and HUB75 Direct Memory Access (DMA) simultaneously on the ESP32-S3 caused conflicts during development, eventually requiring a platform migration. TickrCast solves this by moving the content composition to a server-side rendering and segment caching system, leaving the ESP32-S3 to handle the display output. This helps keep scrolling smooth and updates responsive, while the cloud infrastructure has been load-tested with up to 22,000 concurrent devices on a single server. Note that the project is not open-source, which may be an issue for a device connected to the cloud. However, the developer promises to release the server and firmware code as open source if the company ever shuts down.

Previously, we have written about similar projects that put live data on an LED display, including Meterbit’s Pixlpal, an open-source ESP32-S3 128×64 RGB LED matrix ticker for news, crypto, and smart-home alerts. We have also written about Pimoroni’s Interstate 75 W (RP2350) and Adafruit’s PioMatte; both are designed to control those same HUB75 matrices.

The TickrCast is live on Kickstarter, where a single-panel foundational unit starts at $99 for early birds, while the standard 2-panel “Enthusiast” kit is priced at $159. For those wanting maximum screen real estate, an 8-panel kit goes for $699, and a 10-panel kit for $799. Deliveries are scheduled to begin in February 2027. Stretch goals include injection-molded enclosures at $20,000, native Home Assistant integration at $175,000, and a free extra panel for every backer at $500,000. The company also mentions that the first 50 units will ship with a 3D-printed enclosure.

Note: The company says that, because a grounding issue was discovered in testing, the 10-panel hardware was respun. Backers of this tier will receive 8 panels initially, with the final 2 shipping later once the revised board passes validation.

The post TickrCast – An ESP32-S3 HUB75 LED ticker display with server-side rendering and OTA (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Harald Kreuzer has built an ESP32-P4 weather station and air quality monitor that features a 10.1-inch dashboard display and a built-in Sensirion SEN66 sensor for particulate matter, CO2, VOC, and NOX, and connects to wireless LoRa and ESP-NOW sensors.

The project is entirely made of off-the-shelf boards, so no custom hardware is used, and it should be relatively easy to reproduce. It doesn’t require any cloud or user account, although it can connect to a choice of weather APIs (Open-Meteo, OpenWeatherMap or Visual Crossing) to display weather forecasts. The build is fully documented in a long blog post, so we’ll highlight the key features of the project in this article.

Hardware specifications for the main unit:

• Based on ESP32-P4-Module-DEV-KIT-C
  • Credit card-sized ESP32-P4-Module-DEV-KIT SBC based on an ESP32-P4NRW32 module with ESP32-P4 with 32MB PSRAM and ESP32-C6 for WiFi and BLE
  • 10.1-inch IPS panel with 1280 x 800 resolution, 800:1 contrast ratio, and 400 cd/m² brightness
  • 8Ω 2W speaker (not used in this project)
  • RPi Camera (B) and 15-pin FFC cable for the camera (not used in this project)
• LoRa connectivity – Heltec V4 (board only)
• Sensors
  • Sensirion SEN66 multisensor for PM1.0, PM2.5, PM4.0, PM10, nitrogen oxides (NOx), and volatile organic compounds (VOCs).
  • Brightness control through
    • BH1750 light sensor
    • DFRobot C4001 24GHz mmWave sensor for human presence detection (12-meter range variant)

Harald is also working on a version based on the 10-inch version of the GUITION ESP32-P4 touchscreen displays, which should probably be thinner since the ESP32-P4 is directly integrated into the display board.

While you could use the main unit as is, you may want to add a few extra LoRa and ESP-NOW modules:

• Sensor-Receiver – LoRa + ESP-NOW receiver based on Heltec WiFi LoRa 32 (ESP32-S3)
• Sensor-LORA-BME280 – Temperature, humidity, and pressure LoRa node with Heltec WiFi LoRa 32 V3/V4
• Sensor-LORA-BME280-HTCC-AB01 – Temperature, humidity, and pressure LoRa node with Heltec CubeCell HTCC-AB01 V2 (cheaper, no display)
• Sensor-LORA-SHT45 – Temperature and humidity LoRa node with Heltec WiFi LoRa 32 (ESP32-S3)
• Sensor-LORA-RadiationD – LoRa Geiger counter based on RadiationD-v1.1 tube and Heltec WiFi LoRa 32 V4 (ESP32-S3)
• Sensor-LORA-RadiationD-HTCC-AB01 – Same as above, but with Heltec CubeCell HTCC-AB01 V2
• Sensor-ESP-NOW-BME280 – Temperature, humidity, and pressure ESP-Node node using a generic ESP32 board (oldest sub-project)

That’s about it for the hardware. The source code for the main unit, the receiver module, and sensor modules can be found on GitHub. It mostly relies on the ESP-IDF framework, except for the CubeCell-based modules whose firmware is based on Arduino/PlatformIO. The main unit renders the user interface with LVGL and was designed with EEZ Studio open-source GUI editor.

You’ll also need to print the enclosures for the main unit and sensors for your 3D printer. The only issue is that the STL files don’t seem to have been released just yet, as I could only find files for an older version of the project, and they won’t be suitable for the new variant. The STEP and SolidWorks files can now be found in the SolidWorks folder of the aforelinked firmware repository.

There are also plenty of DIY weather station and air quality monitor projects such as MoreSense MS-07, Project Aura, or Spakfun ESP32 Arduino IoT Weather Station, but this project is quite advanced and the first I’ve seen based on ESP32-P4.

While you may have fun and gain knowledge and skills while building your own ESP32-P4 weather station & air quality monitor, you will almost certainly not save money compared to cost-optimized commercial products. The costs are estimated as follows for the main unit:

• Waveshare ESP32-P4-Module-DEV-KIT-C –  $77.99 on Waveshare or $98.55 on Amazon
• Heltec V4 board –  $29.99 on Amazon
• Sensirion SEN66 – $62.03 on the Sensirion website
• Light sensor BH1750 – $1.39 on AliExpress
• C4001 24GHz (12 meters) – $12.90 on DFRobot

That’s at least $185. You’ll need to add taxes and shipping, as well as a few nuts and bolts, the filament for 3D printing, and extra electronics components if you want to add external modules.

The post DIY ESP32-P4 weather station & air quality monitor connects to LoRa and ESP-NOW sensors, various weather APIs appeared first on CNX Software - Embedded Systems News.