LILYGO T-Display K230 is a handheld dev kit based on Canaan’s Kendryte K230 dual-core 64-bit RISC-V processor. The kit features a 4.1-inch AMOLED touchscreen and supports an optional detachable keyboard with battery and optional nRF9151 cellular and GNSS connectivity.

The device also comes with 1GB LPDDR4 RAM, a 2MP camera, HDMI output, Ethernet, WiFi 4, and either an SX1262 or LR2021 LoRa transceiver. These features make this kit suitable for remote monitoring, wireless communication, and other embedded applications.

LILYGO T-Display K230 Kit specifications:

• SoC – Kendryte K230
  • CPU
    • 64-bit RISC-V processor @ 1.6GHz with RISC-V Vector Extension 1.0, FPU
    • 64-bit RISC-V processor @ 800MHz with support for RV64GCB instruction set;
  • AI accelerator
    • KPU with support for INT8 and INT16
    • Compatible with TVM, TensorFlow, PyTorch, ONNX
    • Typical network performance: Resnet 50 ≥ 85fps @ INT8; Mobilenet_v2 ≥ 670fps @ INT8; YoloV5S ≥ 38fps @INT8;
  • GPU – 2.5D graphics engine
  • DPU – 3D depth processing engine
  • VPU – Hardware H.264/H.265 video encoding and decoding
• Memory – 1GB LPDDR4 RAM
• Storage – MicroSD card slot for firmware and data storage
• Display – 4.1-inch 568 × 123 AMOLED display with RM69A10 driver GT9895 touch controller and MIPI DSI interface
• Video Output – HDMI port supporting up to 1080p30
• Camera – 2MP GC2093 camera supporting 1080p resolution
  • Rotatable camera module for front-facing use
  • Up to three MIPI CSI camera interfaces on the uncased board
• Audio
  • Built-in microphone and speaker
  • MAX98357A I2S audio amplifier
  • 3.5mm headphone jack
• Networking
  • Wireless Connectivity
    • WiFi 4 – RTL8189, 802.11b/g/n, 2.4GHz
    • Bluetooth LE – nRF52840 on the supported expansion board
    • LoRa – Semtech transceiver on the main unit (depending on configuration)
      • Semtech SX1262 – 868/915/923 MHz
      • Semtech LR2021 – 830–940 MHz
    • Cellular – LTE-M and NB-IoT via optional nRF9151-equipped keyboard
    • GNSS – Via optional nRF9151-equipped keyboard
  • 2x MMCX antenna connectors
  • 10/100Mbps Ethernet through RJ45 connector; RTL8152B controller
• USB – 3x USB Type-C ports for power and USB OTG, and nRF52840 debugging
• Expansion
  • 40-pin GPIO header on the main board
  • nRF52840 GPIO expansion header on the cased version
  • Qwiic I2C interface
  • Optional keyboard expansion through a 24-pin docking connector
• Misc
  • AHT20 temperature/humidity sensor on the cased expansion
  • Boot, reset, and power controls
  • Optional backlit QWERTY keyboard
• Power Supply
  • 5V input through USB Type-C
  • Optional keyboard base with two 21700 battery holders
  • BQ25896 battery charging IC and BQ27220 battery fuel gauge
• Dimensions
  • Main board – 104 × 51 × 15.5mm
  • Cased version – Approximately 110 × 58 × 30mm

The optional keyboard expansion adds a physical QWERTY keypad with function keys, arrow keys, and backlighting. It uses a TCA8418 keyboard controller and an XL9555 16-bit GPIO expander. The device also includes two 21700 battery holders and battery management circuitry. It is available in two versions: a standard keyboard and another with a Nordic nRF9151 SiP for LTE-M, NB-IoT, and GNSS connectivity, along with a SIM card slot.

Software support includes the Kendryte K230 Linux SDK, CanMV/RT-Smart firmware, and an LVGL-based launcher with apps for camera, video, audio, LoRa, WiFi, and battery monitoring. The newer Linux firmware also offers experimental Meshtastic and LoRaWAN support. LILYGO provides source code, firmware, examples, and build instructions on GitHub, with additional examples in the CanMV/RT-Smart repository. More details can be found in the wiki.

The design reminds us somewhat of the LILYGO T-Display-P4, a similar smartphone-like development kit built around the ESP32-P4 and ESP32-C6 with an AMOLED display, camera, and LoRa connectivity. However, unlike the T-Display-P4, the T-Display K230 uses Linux-capable dual-core RISC-V processor with a built-in AI accelerator, and support extra features through the optional keyboard. Another related board is the CanMV-K230, launched in 2023, which uses the same Kendryte K230 processor but comes in a conventional development board form factor rather than a handheld design.

The LILYGO T-Display K230 Kit is sold on AliExpress for about $180 (SX1262 and LR2021 variant), and the keyboard adds $133, assuming it is with the nRF9151 at that price. The kit is also sold for about $150 on the company’s official store, to which the optional keyboard adds $70.94 without the nRF9151 module or $113.25 with the module. The bare T-Display K230 board is sold on a different page for $98.79.

The post LILYGO T-Display K230 RISC-V AI dev kit features AMOLED, LoRa, and keyboard with optional LTE-M/NB-IoT cellular connectivity appeared first on CNX Software - Embedded Systems News.

Picotronix is a modular oscilloscope, logic analyzer, and frequency meter from Sydney-based Thort Werx Pty Ltd, based on two Raspberry Pi Pico 2 units, more exactly a Pico 2 used as a signal processor and a Pico 2 W used to drive the display and for wireless connectivity.

The solution is comprised of the Picotronix Main Module with an integrated LCD and three expansion ports, used to connect picoPods plug-in expansion modules to add high-speed ADCs, mixed-logic family logic state analysis, a waveform generator, analog conditioning, and more.

Picotronix key features and specifications:

• Main Module
  • Two-channel digital storage oscilloscope
  • 8-channel logic analyzer, expandable up to 14 channels with a picoPod; up to 150 MHz sample rate
  • Frequency meter
  • Display – 320 × 240 color LCD from Pimoroni
  • Three expansion ports, two 10-pin analog ports, and one 26-pin multifunction port
  • Analog channel
    • Accepts 3 Vpp or 16.6 Vpp at 1 MΩ, with extra BNC AC and DC inputs
    • Captures with the RP2350’s 12-bit ADC, with timebases from 5 ms to 10 µs per division.
  • Auxiliary input doubles as a logic probe or voltmeter and works without either Pico
  • Power – USB, 9–12 V plug pack, or four AAA cells.
• Optional plug-in picoPods
  • Single-ended analog: 500 mV, 1 V, 2 V and 5 V windows, movable anywhere within ±5 V, plus a x10 BNC input to ±50 V.
  • Differential: 2 Vpp, and a ±25 V common-mode pod with software-switched x1/x10 ranges.
  • Wave-05: a 12-bit R2R DAC, 10 mVpp to 3.3 Vpp with a ±5 V programmable offset, plus two precision voltage references.
  • 12-bit Logic State Analyser: sampled at 150 MSPS, in three 4-bit ports that can each serve a different logic family (1.8, 2.5, 3.3 or 5 V).
  • MSO-12: a 100 MSPS 8-bit ADC, 8 logic inputs, and a four-channel 12-bit serial ADC that also detects each logic port’s family voltage.
  • Electronix Lab: a teaching picoPod that connects by cable to all three ports. Configure logic on the pod, generate a real signal from it, and see the result on the display.
  • ADC-100 picoPod: dual 100 MSPS ADCs, capturing up to 75 MSPS on one channel or 50 MSPS on both on a stock 150 MHz RP2350.

The Picotronix is programmable in MicroPython with direct access to the RP2350’s PIO state machines for optimal performance. A PC emulator will also be available to develop MicroPython scipts on your computer.

While we were told that “everything is open”, it’s unclear which part of the project is open-source, since there’s no mention of hardware or firmware release, but the Kickstarter page does indicate part of the code will be open-sourced:

The system’s modular design anticipates diverse software-based functions to be implemented by both our team and the open-source community over time.

….

The open-source nature ensures community development can supplement our efforts, providing multiple development pathways for enhanced functionality.

Documentation is currently limited, and you won’t find a user manual as of now, but more technical details are available in blog posts and YouTube videos, notably a  21-minute comparison to a Tektronix TBS 2000 oscilloscope.

The project has just launched on Kickstarter with a $10,000 AUD (about $7,000 US) funding target. I’ll use the USD equivalent for rewards’ prices. The Picotronix Main Module reward requires a $134 pledge for a full unit and a cable set, and optional picoPods are available at an extra cost. Other rewards are bundles with specific picoPods. Shipping adds $20 to $40 AUD to Australia, and $21 to $42 USD to the rest of the world. Deliveries are expected to start by March 2027.

The post Picotronix modular oscilloscope and logic analyzer features two Raspberry Pi Pico 2 boards (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Synergy Logic SLM-RP4 industrial PLC is built around a Raspberry Pi 4 SBC and designed for industrial automation, machine control, data acquisition, and Industrial IoT (IIoT) applications.

The DIN-rail mountable controller supports 24V DC input and an SPI-based backplane that connects up to eight Synergy Logic I/O modules, providing up to 128 I/O points. It also features Gigabit Ethernet, WiFi 5, Bluetooth 5.0, USB ports, two micro-HDMI outputs, and a fanless aluminum enclosure. The system runs Raspberry Pi OS with OpenPLC pre-installed for standard IEC 61131-3 logic programming.

Synergy Logic SLM-RP4 specifications:

• SBC – Raspberry Pi 4 Model B (2GB RAM)
• Storage – 16GB microSD card
• Video – 2x micro-HDMI ports
• Audio – 3.5mm audio jack
• Networking
  • Gigabit Ethernet port
  • Dual-band WiFi 5 (802.11ac)
  • Bluetooth 5.0/BLE
• USB
  • 2x USB 3.0 Type-A ports
  • 2x USB 2.0 Type-A ports
  • 1x USB-C port for auxiliary power
• I/O Expansion – SPI-based backplane supporting up to eight stackable SLM modules (18 compatible module models, up to 128 I/O points)
  • Discrete inputs – Up to 16 channels of 12–24V DC or 24V AC/DC inputs, or eight channels of 100–240V AC inputs
  • Input simulator – Eight-channel toggle switch module
  • Discrete outputs – Up to 15 channels of sourcing or sinking DC outputs
  • Relay outputs – 8-channel or 16-channel relay modules with SPST or SPDT contacts
  • Combined discrete I/O – Mixed digital input/output and relay modules
  • Analog inputs – Eight channels of 0–10V or 0–20mA inputs with 13-bit resolution
  • Analog outputs – Eight channels of 0–10V or 4–20mA outputs with 12-bit resolution
  • Combined analog I/O – Four analog inputs (13-bit) and two analog outputs (12-bit)
  • Temperature inputs – Four-channel thermocouples supporting J, E, K, R, S, T, B, N, and C types, along with millivolt inputs
• Misc
  • PWR and BASE status LEDs
  • Programmable RGB status LED
  • Toggle switch for PLC run/stop or custom functions
  • Hardware watchdog
• Power
  • 24V DC Class 2 input (±25%) via terminal block
  • 5V/3A via USB-C (controller only, not the I/O backplane)
• Dimensions – Not specified
• Weight – 289 grams
• Temperature Range – 0°C to 50°C, fanless
• Enclosure – Aluminum, open equipment, 35mm DIN-rail mounting
• Certifications – UL 61010 and UL 61010-2-201 listed; CE marked to EN 61131-2 and EN 61010 standards

The SLM-RP4 uses an SPI-based backplane to connect expansion modules through edge connectors, so no extra cables are needed between modules. Modules connect to the right side of the controller, and their positions determine their slot addresses. Currently, 18 compatible modules (not hot-swappable) are available, including digital I/O, relays, analog voltage and current I/O, and thermocouple inputs.

On the software side, the SLM-RP4 comes preinstalled with Raspberry Pi OS, OpenPLC Runtime, and Synergy Logic’s Python and C++ libraries. It supports OpenPLC Editor 4.3.1 or newer, so you can program the controller using IEC 61131-3 languages such as Ladder Diagram (LD), Structured Text (ST), Function Block Diagram (FBD), and Instruction List (IL), as well as Python and C++ function blocks. The editor also includes an SLM-RP4 hardware support package for configuring I/O modules, assigning IEC addresses, and adjusting SPI backplane settings.

The SPI backplane supports adjustable I/O scan intervals from 1ms to 1000ms, with 10ms set by default. It also detects communication errors and can stop PLC execution and reset outputs to zero if errors exceed a set limit. You can also use the provided Python and C++ libraries to control the backplane directly without OpenPLC, but both methods cannot be used at the same time. A hardware watchdog is also included to help recover from application failures.

Synergy Logic also provides documentation for integrating FUXA, an open-source, web-based HMI/SCADA platform that communicates with OpenPLC or custom applications over Modbus TCP. This allows you to create local dashboards with gauges, indicators, sliders, and trend charts. The documentation also includes examples for analog signal monitoring, thermocouple logging, output control, and Modbus TCP communication.

Other Raspberry Pi-based industrial controllers include Industrial Shield’s Raspberry Pi PLC Industrial Controller, also based on the Pi 4, the EDATEC ED-PLC2010 (CM4) with up to 32 expansion modules and CODESYS Runtime, and the Sfera Labs Strato Pi Max (CM4). Synergy Logic previously introduced the SL-RP4, a Raspberry Pi 4-based industrial controller that relies on external I/O connectivity. The newer SLM-RP4 adds an integrated backplane for connecting SLM modules directly to the controller.

The Synergy Logic SLM-RP4 is available for $249, with expansion I/O modules sold separately. The earlier SL-RP4 sells for $149. More information, including setup guides, software examples, and module datasheets, can be found on the product page and the documentation website.

The post Synergy Logic SLM-RP4 – A Raspberry Pi 4 PLC with modular industrial I/Os appeared first on CNX Software - Embedded Systems News.

MYIR MYC-JR3576 is a SMARC 2.2 system-on-module (SoM) powered by a Rockchip RK3576/RK3576J octa-core Cortex-A72/A53 SoC with a 6 TOPS NPU designed for Edge AI applications.

It features up to  8GB LPDDR5, up to 64GB eMMC flash, an optional WiFI 6E and Bluetooth 5.4 module, and exposes all I/Os through a 314-pin MXM connector as defined in the SMARC 2.2 standard. The company says it’s the socketed counterpart to its LGA-packaged MYC-LR3576 (43 x 45 mm) module introduced in 2024, so the same processor is now available in both solder-down and standard form factors from MYiR.

MYC-JR3576 – Rockchip RK3576 SMARC 2.2 module

MYIR MYC-JR3576 specifications:

• SoC – Rockchip RK3576 (commercial) or RK3576J (industrial)
  • CPU
    • 4x Cortex-A72 cores @ 2.2GHz, four Cortex-A53 cores @ 1.8GHz
    • Arm Cortex-M0 MCU at 400MHz
  • GPU – ARM Mali-G52 MC3 GPU with support for OpenGL ES 1.1, 2.0 and 3.2, OpenCL up to 2.0 and Vulkan 1.1
  • NPU – 6 TOPS (INT8) AI accelerator with support for INT4, INT8, INT16, BF16, and TF32 mixed operations.
  • VPU
    • Video Decoder: H.264, H.265, VP9, AV1, and AVS2 up to 8K @ 30fps or 4K @ 120fps
    • Video Encoder: H.264 and H.265 up to 4K @ 60fps, (M)JPEG encoder/decoder up to 4K @ 60fps
• System Memory – 4GB or 8GB LPDDR5
• Storage
  • 32GB or 64GB eMMC flash
  • 32Kbit EEPROM
• Display – Lontium LT8912B MIPI DSI to LVDS converter
• Networking
  • 2x Motorcomm YT8531SH GbE controllers
  • Optional Wi-Fi 6E + Bluetooth 5.4 module (NXP IW623-based AzureWave AW-XM729)
• USB – USB 3.0 hub
• 314-pin MXM 3.0 edge connector
  • Storage – SATA III or PCIe 2.1
  • Video Output
    • LVDS
    • MIPI DSI
    • HDMI
    • DisplayPort
  • Video Input – 3x MIPI CSI (2x 4-lane, 1x 2-lane)
  • Audio – 2x SAI
  • Networking – 2x Gigabit Ethernet interfaces
  • USB
    • 1x USB 2.0 OTG
    • 2x USB 2.0 host
    • 2x USB 3.0 host
  • PCIe – PCIe 2.1 (multiplexed with SATA III)
  • Analog – 6x SARADC, 12-bit resolution, up to 1MS/s (optional)
  • Other I/Os
    • 2x CAN FD
    • 4x UART, 2x SPI, 4x I2C, 1x SDIO, 3x PWM, 14x GPIO
• Security – TPM 2.0
• Power Management
  • 5V/2A input
  • RK806S-5 PMIC
• Dimensions – 82 x 50mm (SMARC 2.2 form factor, 12-layer PCB)
• Temperature Range
  • Industrial grade: -40~85°C
  • Commercial grade: 0~70°C

The company provides a Linux 6.1 BSP (Debian/Yocto) and Android 15 support for the system-on-module and the development board described below. As usual for MYiR, the software is not available publicly, and only provided to customers. They only lists software features like Linux 6.1, U-boot 2017.09, and the list of open-source drivers for the board.

MYD-JR3576 development board

The company also provides a SMARC development platform for the module.

MYD-JR3576 development board specifications:

• Storage
  • MicroSD card slot
  • M.2 (PCIe) socket for NVMe SSD
  • SATA SSD connector
• Video output
  • HDMI port
  • DisplayPort connector
  • MIPI-DSI/LVDS display interface
• Camera – 2x MIPI-CSI Camera Interfaces
• Audio – 3.5mm audio input/output jack
• Networking
  • 2x Gigabit Ethernet RJ45 ports
  • WiFi/Bluetooth on SoM
  • Cellular M.2 socket for USB based 4G/5G LTE Module + 2x SIM card slots
  • 2x external antenna connectors (one for WiFi and one 4G/5G)
• USB
  • 2x USB 3.0 host ports
  • 2x USB 2.0 host ports
  • USB 2.0 OTG port
• Serial ports via dedicated connector
  • Isolated RS232 interface
  • Isolated RS485 interface
  • 2x isolated CAN FD interfaces
• Debugging – USB Type-C debug port
• Misc
  • Power switch
  • 3x Buttons: Reset, User, ON/OFF
  • 3x LEDs: Power, System status, User
  • Fan connector
• Power Supply – +12V/2A via power jack
• Dimensions – 206 x 126mm (6-layer PCB)
• Temperature Range
  • Industrial grade: -40~85°C
  • Commercial grade: 0~70°C
  • Note: WiFi Module operates in the -20~70°C range

Note that the “N/A” port on the carrier board are not supported by the Rockchip RK3576 system-on-module, and are available on the SMARC development board for other compatible modules.

We’ve written about several Rockchip RK3576 system-on-modules over the last two years, including Geniatech SOM-3576 (with 314-pin MXM 3.0 connector, but not advertised as SMARC compatible), the Raspberry Pi CM5-inspired Radxa CM4 and Geniatech XPI-3576-CM5 modules, Boardcon LGA3576 LGA system-on-module, and others. However, the MYiR MYC-JR3576 is the first Rockchip RK3576 SMARC system-on-module we’ve covered here, and a quick web search reveals some other options such as the Advantech AOM-5841 and the Vantron VT-SBC-SMARC-3576.

MYiR Tech is one of the few companies that quotes pricing publicly for industrial-grade hardware:

• MYC-JR3576-64E8D-220-C (RK3576, 8GB/64GB, 0 to +70°C) — $209 US
• MYC-JR3576J-32E4D-160-I (RK3576J, 4GB/32GB, -40 to +85°C) — $192 US
• MYD-JR3576 development kit — $293 or $279 depending on the selected module above.

The modules and development kits are shipping now, and sample pricing applies to the first two units. You’ll find more details, including product briefs, 3D step files, and purchase links, on the product page.

The post Rockchip RK3576-powered SMARC 2.2 system-on-module and development board target Edge AI applications appeared first on CNX Software - Embedded Systems News.

Adafruit has launched the Metro ESP32-P4 Hydra, an Arduino UNO-shaped development board with ESP32-P4, 16MB PSRAM, and 16MB NOR flash, and an ESP32-C6 for WiFi 6, Bluetooth 5.3 LE, Thread, and Zigbee connectivity.

Other features include Raspberry Pi-compatible MIPI DSI and CSI interfaces, a USB 2.0 High-Speed host port, a microSD card slot, and Arduino UNO expansion headers. The board also includes a STEMMA QT connector, LiPo battery charging, and multiple power input options, making it suitable for HMI, camera, and IoT applications.

Adafruit Metro ESP32-P4 Hydra specifications:

• Microcontroller – Espressif Systems ESP32-P4 (v1 chipset)
  • CPU
    • Dual-core 32-bit RISC-V HP (High-performance) CPU @ up to 360 MHz with AI instructions extension and single-precision FPU
    • Single-RISC-V LP (Low-power) MCU core @ up to 40 MHz with 8KB of zero-wait TCM RAM
  • Memory
    • 768 KB HP L2MEM (for dual-core CPU), 32 KB LP SRAM, 8 KB TCM (for LP MCU core)
    • 16MB PSRAM
  • GPU – 2D Pixel Processing Accelerator (PPA)
  • VPU – H.264 and JPEG codec support
• Storage
  • 16MB QSPI NOR flash
  • MicroSD card slot with 4-bit SDIO interface
• Display/Camera
  • 2-lane MIPI DSI connector
  • 2-lane MIPI CSI connector
  • Both are Raspberry Pi-compatible 22-pin, 0.5mm-pitch connectors
• Wireless – ESP32-C6 SDIO module
  • 2.4 GHz WiFi 6 (802.11ax)
  • Bluetooth LE 5.3
  • IEEE 802.15.4 radio supporting Thread and Zigbee (additional firmware integration may be required)
• USB
  • USB Type-C port for power, serial debugging, JTAG, and TinyUF2 drag-and-drop firmware updates
  • USB 2.0 Type-A High-Speed (480 Mbps) OTG host port with software-controlled power and a dedicated 5V boost supply
• Expansion
  • Arduino UNO headers with D0–D13 digital pins and six analog inputs
  • Separate 6-pin SPI header
  • I2C/I3C, UART, SPI, PWM, ADC, TWAI/CAN, and other interfaces
  • STEMMA QT connector for I2C/I3C peripherals
• Debugging – Dedicated hardware/ROM UART debugging connection
• Misc
  • RGB NeoPixel and D13 user LED
  • Reset and Boot buttons
  • On/Off switch
  • Dedicated UART debug connection
  • Battery charging and status indicators
• Power
  • 7–12V DC via barrel jack
  • USB Type-C for power
  • Single-cell LiPo battery connector with onboard charging and voltage monitoring
• Dimensions – 72.3 x 53.3 x 14.7 mm

The MIPI DSI and CSI connectors use the Raspberry Pi 22-pin format, with two data lanes each. Adafruit says the board supports the 5-inch and 7-inch Raspberry Pi Touch Display 2, both with 720 x 1280 resolution and capacitive touch. For cameras, it supports Raspberry Pi Zero-style modules with the OV5647 sensor, but not the newer IMX708-based Camera Module 3.

The board can be programmed using Espressif’s ESP-IDF framework and supports TinyUF2 for drag-and-drop firmware updates over USB. Support for the Adafruit Metro ESP32-P4 was also added to CircuitPython 11.0.0-alpha.2, although the ESP32-P4 port is still in alpha. The company also said it is working on Arduino support. We couldn’t find details documentation for the board, so it must be work-in-progress.

We previously wrote about the Adafruit Metro ESP32-S3, a similar Arduino UNO-style development board but built around ESP32-S3 paired with 8MB PSRAM and 16MB flash. Another similar board is the Adafruit Metro RP2350, which uses Raspberry Pi’s RP2350 MCU and offers an HSTX connector for DVI video output.

The Adafruit Metro ESP32-P4 Hydra is available for $34.95 on the Adafruit store. Note that the board currently ships with the original ESP32-P4 v1 chipset rather than the newer ESP32-P4X v3 revision.

The post Adafruit Metro ESP32-P4 Hydra – An Arduino UNO-shaped ESP32-P4 + ESP32-C6 board with WiFi 6, MIPI interfaces appeared first on CNX Software - Embedded Systems News.

Renesas has introduced the RX74M and RX74N, two new 32-bit RX family MCUs built on a 22nm process with embedded MRAM. Both MCUs integrate a 600MHz RXv3e core and are designed for industrial automation, motor control, robotics, and energy management applications.

The MCUs feature up to 1MB of code MRAM, 1.5MB SRAM, and 256KB of zero-wait-state local RAM for real-time processing. The RX74M adds an integrated EtherCAT slave controller, dual Gigabit Ethernet MACs with TSN support, and an Ethernet switch for industrial networking, while the RX74N offers a single Gigabit Ethernet interface for general industrial applications. Other features include CAN FD, USB 2.0, I3C, high-resolution PWM timers, and hardware-based security with Renesas Protect Zone.

Renesas RX74M and RX74N specifications:

• MCU – 32-bit RXv3e core clocked at 300/480/600MHz
  • Up to 3,604 CoreMark at 600MHz
  • Single- and double-precision floating-point support
  • Trigonometric function accelerator
  • Register bank save function for faster interrupt response
  • 32KB instruction cache and 32KB data cache with ECC
• Memory
  • Up to 1MB code MRAM with dual-bank support
  • 48KB data MRAM (100,000 programming cycles)
  • 128KB user extra MRAM (configurable as MTJ-OTP)
  • 1.5MB SRAM with ECC
  • 256KB local RAM (2x 128KB) with ECC and zero-wait-state access
  • External SDRAM support, up to 128MB
• Storage
  • SD/MMC host interface supporting SD memory and SDIO
  • Dual xSPI interfaces for external Octal Flash and RAM with execute-in-place (XiP) and on-the-fly decryption
• Camera – Parallel camera capture interface
• Audio
  • 2x serial sound interfaces (I2S)
  • PDM microphone interface
• Networking
  • RX74M
    • 2x Gigabit Ethernet MACs
    • 2-port Ethernet switch with TSN support
    • 2-port EtherCAT slave controller on selected variants
  • RX74N
    • 1x Gigabit Ethernet MAC
    • No integrated EtherCAT slave controller
• USB
  • 1x USB 2.0 High-Speed (480Mbps) Host/Device/OTG
  • 1x USB 2.0 Full-Speed (12Mbps) Host/Device/OTG
• Serial – Up to 13x serial communication interfaces (SCI/UART)
• Other I/O
  • Up to 197x GPIOs
  • 3x SPI controllers, 2x I2C controllers, 1x I3C controller
  • Up to 16x 32-bit general-purpose PWM timers
  • High-resolution PWM with up to 200MHz control clock
  • Up to 3x 16-bit ADC units with 36 input channels
  • Up to 2x 12-bit DACs
  • 2x phase-counting interfaces
  • Up to 6x delta-sigma modulator interface channels (RX74M)
  • Up to 3x CAN FD controllers
• Security
  • Renesas RSIP-E50D security engine
  • AES, RSA, ECC, ChaCha20-Poly1305, SHA-512, SHA3-512, TRNG
  • Protect Zone for secure and non-secure execution environments
  • Secure boot, device lifecycle management, tamper detection, and key zeroization
  • On-the-fly decryption for external memory
• Debugging – JTAG, SWD, FINE
• Misc
  • Temperature sensor
  • RTC and watchdog timers
  • 8x DMA controllers, with additional transfer engines
• Power supply – 1.62V to 3.63V with low power mode support
• Packages
  • 100/144/176-pin LFQFP (480MHz, 10/100Mbps Ethernet; also lacks high-speed USB controller, parallel camera capture interface, and external SDRAM controller)
  • 176/224/320-pin LFBGA (600MHz, up to Gigabit Ethernet)
• Temperature range – -40°C to +125°C (junction temperature)
• Manufacturing process – 22nm

From the specifications, it’s clear that the main difference between the two MCUs is in the networking stack, where the RX74M features dual Gigabit Ethernet MACs, a TSN-capable Ethernet switch, and an EtherCAT slave controller for industrial automation. The RX74N, on the other hand, features a single Gigabit Ethernet interface without EtherCAT or TSN support for general industrial applications. Both MCUs share the same RXv3e core, memory architecture, and security features.

The MCUs also include Renesas’ RSIP-E50D engine with hardware encryption, secure boot, and a Protect Zone function to separate secure and non-secure operations. Renesas says these features help developers to meet security requirements such as the EU Cyber Resilience Act (CRA) and support post-quantum cryptography (PQC).

Software support includes the Renesas Flexible Software Package (FSP), e² studio IDE, C/C++ compilers, and tools from partners such as IAR, Codewerk, and RT-Labs. Renesas also lists Zephyr RTOS and eForce as ecosystem partners. For industrial networking, Codewerk provides PROFINET integration services, while RT-Labs offers its P-Net PROFINET device stack for the RX74M. Renesas also provides sample projects for EtherCAT, EtherNet/IP, Modbus, secure boot, and motor control.

Renesas has also introduced the EK-RX74M and EK-RX74N evaluation kits, along with MCK-RX74M/MCK-RX74N motor control kits and MCB-RX74M/MCB-RX74N motor control CPU boards. The EK-RX74M includes a 600MHz RX74M MCU, two Ethernet interfaces, CAN FD, USB, RS-485, external Octal Flash and RAM, a microSD card slot, and Arduino Uno R3, mikroBUS, and Pmod expansion headers and connectors.

What makes the MCUs interesting is the use of MRAM (which makes programming and updates faster) instead of traditional flash memory for program storage. But this is not the first MCU with MRAM, as in the past the company released RA8T2 Cortex-M85 MCUs, which feature a 1GHz CPU, up to 1MB MRAM, and EtherCAT support for industrial motor control. We have also seen Renesas introduce the RA8D2 and RA8M2 with MRAM and additional graphics and processing features. While those RA series MCUs use Arm Cortex-M85 cores, the RX74M and RX74N are based on Renesas’ own RXv3e architecture.

Renesas has also provided “Winning Combinations,” which are just pre-tested system designs including a Maglev Conveyor System for contactless material handling with real-time motor control and industrial networking, and a Smart HVAC Control with Integrated Slave Actuator for controlling fans, compressors, and other HVAC components.

The Renesas RX74M and RX74N are available now in 300MHz, 480MHz, and 600MHz variants, with the 300MHz versions offering more than 20% lower power consumption than the higher-performance models, according to Renesas. Renesas lists the 600MHz RX74M for $10.74 and $7.62 for the 480MHz RX74N in 1,000-unit quantities. The EK-RX74M and EK-RX74N evaluation kits are also available from Mouser and Future Electronics, with prices starting at $114.84 for the EK-RX74N and $206.88 for the EK-RX74M at Future Electronics. More details about the RX74M and RX74N can be found on the Renesas website and in the press release.

The post Renesas RX74M and RX74N – 600MHz industrial MCUs with RXv3e core, embedded MRAM, 22nm process appeared first on CNX Software - Embedded Systems News.

Deki Engine is an open-source modular C++ engine with a Unity-style visual editor for embedded devices, currently optimized for ESP32 platforms such as LilyGO T-Deck Plus, T-Display S3 AMOLED, and soon Elecrow CrowPanel Advanced 7″ ESP32-P4.

The editor is available for Windows and Linux with a compiler for the 2D graphics engine, and everything is released under an Apache 2.0 license on GitHub. The project also works on desktop through SDL3, which may be convenient during development.

Some of the highlights of the Deki Engine project:

• Workflow
  • Build scenes visually (sprites, text, particles, tilemaps, state machines, tweens)
  • Pick a board in the editor’s Build panel
  • Flash it.
  • The project also runs on desktop through SDL3, or headless with no renderer
• Performance – 60 FPS on ESP32-S3 (See demo video below)
• Package-based solution
  • Hardware packages
    • LilyGO T-Deck Plus with support for display, touch, keyboard, trackball, SD card, GPS and LoRa (SX1262)
    • T-Display S3 AMOLED (and apparently T-Display S3 Touch based on the demo)
    • Upcoming support for CrowPanel Advanced 7″ ESP32-P4
  • Other packages – Software 3D render pass, IMU, RTC, audio over I2S, Wi-Fi, BLE and a Tiled map importer.
• Games can run on Tactility OS for ESP32
• Inspector fields are generated with C++26 static reflection.
• The editor, docs and website are currently available in English, Simplified Chinese, Japanese and Korean languages

More details can be found in the documentation website, but I find it hard to navigate and understand, as everything feels all over the place. Deki Engine was developed by Jairo Baldán (Jai), a game programmer from Spain, as a side project.

Looking at GitHub repositories, it looks like all/most commits were co-authored by Claude, so the project relies heavily on AI assistance, but it looks like a working project, and a AI analysis reports that the code is of high quality is high for a solo/hobby project. We are also shown working demos, first for the editor.

A demo (I supposed that’s the relevant repo) is also shown running on the T-Display S3 Touch at 60 FPS.

It might be an interesting for 2D game development on ESP32-based hardware, and the Deki Engine also supports software-rasterized 3D graphics, but I’m not sure how smooth that would be on ESP32 targets.

The post Deki Engine – ESP32 gets an open source Unity-style editor and C++ 2D game engine appeared first on CNX Software - Embedded Systems News.

ZOTAC ZBOX PRO CM5-PICO is a compact fanless industrial mini PC built around the Raspberry Pi Compute Module 5 (CM5) with up to 16 GB LPDDR4x memory, optional WiFi 5 and Bluetooth 5.0 wireless connectivity.

The mini PC measures 114.8 x 76 x 31.3 mm and features two 4K HDMI outputs, Gigabit Ethernet, and M.2 expansion for NVMe storage or AI accelerators, along with an internal 40-pin GPIO header, camera interfaces, and a wide 5V to 20V DC input for industrial automation, digital signage, and IoT gateway applications. All I/O ports are placed on a single side to simplify cable management and stacking.

ZOTAC ZBOX PRO CM5-PICO specifications:

• SoM – Raspberry Pi Compute Module 5 (CM5)
  • SoC – Broadcom BCM2712
    • CPU – Quad-core Arm Cortex-A76 @ 2.4 GHz
    • GPU – VideoCore VII @ 800 MHz with H.265 hardware video decoding
  • Memory – 2GB, 4GB, 8GB, or 16GB LPDDR4x RAM
  • Storage –  0GB (Lite), 16GB, 32GB, or 64GB eMMC flash
  • Wireless – Optional Wi-Fi 5 and Bluetooth 5.0
• Storage
  • microSD card slot for CM5 Lite variants
  • M.2 2280 M-Key PCIe Gen2 x1 socket for NVMe SSD or AI accelerator
• Display
  • 2x HDMI 2.1 ports supporting up to 4K @ 60Hz each
  • 2x internal 22-pin, 0.5mm-pitch, four-lane MIPI CSI-2/DSI combo connectors
• Camera – Same MIPI CSI-2/DSI combo connectors can also be used for the camera
• Audio – 8-channel digital audio via HDMI
• Networking
  • Gigabit Ethernet RJ45 port with PTP (Precision Time Protocol) support
  • Optional Wi-Fi 5 and Bluetooth 5.0 on Compute Module
  • Optional Quectel EG916Q-GL LTE Cat 1 bis Modem via M.2 3024 B-Key slot
• USB – 2x USB 3.2 Type-A ports, 1x USB Type-C port for firmware updates
• Expansion
  • Internal 40-pin Raspberry Pi HAT/GPIO header
  • Internal RS-232 header (RX, TX, GND)
  • UART0 debug header
• Misc
  • Watchdog timer built into the Broadcom SoC
  • RTC with CR2032 backup battery
  • TVS and ESD diode protection on external I/O ports
  • Internal fan header
  • CM5 boot/EEPROM override header and nRPIBOOT recovery button
• Power – 5V to 20V DC input via DC IN jack
• Dimensions – 114.8 × 76.0 × 31.3 mm (0.27 liters)
• Weight – Approximately 250 grams
• Temperature – -20°C to +55°C (Operating)
• Enclosure – Passive cooling with aluminum enclosure

The ZBOX PRO CM5-PICO ships with Raspberry Pi OS based on Debian 13, and also supports Ubuntu for Raspberry Pi, Twister OS, Emteria Android, LibreELEC, and Home Assistant OS. You can also use the Raspberry Pi software ecosystem, GPIO libraries, camera interfaces, and standard Linux development tools.

While it’s the first Raspberry Pi CM5-based industrial computer from ZOTAC, the ZBOX PRO CM5-PICO already has plenty of competition, including the Waveshare IPCBOX-CM5 with dual Ethernet, RS485/RS232, CAN Bus, and optional cellular connectivity, the EDATEC ED-IPC3100 series with similar industrial interfaces, and many more. Compared to those systems, the ZOTAC system features dual 4K HDMI, a single-sided external I/O layout, and a palm-sized enclosure for digital signage, HMI, and embedded computing applications.

ZOTAC showcased the ZBOX PRO CM5-PICO at ISE 2026 earlier this year. Japanese distributor Ask Corporation lists the device, but needs to be contacted for pricing and availability. More details are available on the product page. ZOTAC also specifies a product supply lifecycle extending through January 2036. The device ships with the main unit, a universal AC adapter, dual Wi-Fi tri-band antennas (on wireless models), and an L-shaped mounting kit with screws.

The post ZOTAC ZBOX PRO CM5-PICO – A compact Raspberry Pi CM5-based fanless industrial mini PC  appeared first on CNX Software - Embedded Systems News.

LILYGO T-Halow P4 is a compact M.2 development board that combines an ESP32-P4 MCU, an ESP32-C6 wireless co-processor, and a Taixin TX-AH WiFi HaLow module for long-range wireless connectivity.

The board measures just 80 x 22mm (M.2 2280) and uses an M.2 B-key connector to expose GPIO pins and other interfaces for integration into custom carrier boards. It also supports MIPI CSI cameras and MIPI DSI displays, along with hardware H.264 video encoding at up to 1080p30, making it suitable for compact IoT gateways, remote cameras, and industrial monitoring applications.

LILYGO T-Halow P4 specifications:

• SoC – Espressif Systems ESP32-P4NRW32X
  •  CPU
    • Dual-core RISC-V @ 400 MHz with AI instruction extensions and single-precision FPU
    • Single-core RISC-V LP (low-power) MCU @ up to 40 MHz
  • GPU – 2D Pixel Processing Accelerator (PPA)
  • VPU – H.264 and JPEG codec support
  • Memory – 768 KB HP L2MEM, 32 KB LP SRAM, 8 KB TCM, 32MB PSRAM (according to GitHub and wiki; product page says 8MB)
  • Storage – 128 KB HP ROM, 16 KB LP ROM
• Storage – 16MB QSPI NOR flash
• Display interface – 2-lane MIPI DSI connector for external displays, with touch support
• Camera interface – 2-lane MIPI CSI connector with OV2710 camera support
• Wireless
  • ESP32-C6-MINI-1U with 2.4 GHz Wi-Fi 6, Bluetooth 5, and 802.15.4 (Zigbee / Thread / Matter); connected via  SDIO to the ESP32-P4; External IPEX antenna connector
  • TaiXin Semiconductor TX-AH module
    • 802.11ah HaLow compliant
    • Frequency Bands
      • 859-894MHz (875M SAW) OR
      • 902-928MHz (915M SAW)
    • Channel width – 1/2/4/8MHz
    • Throughput – 150 Kbps to 32.5Mbps
    • Range – Up to 1.2 km (Line of Sight)
    • IPEX antenna connector
• USB – 1x USB Type-C port for power and programming
• Carrier board/Expansion interface – M.2 B-key edge connector with GPIO, I2C, UART, ADC, SD, and other ESP32-P4 interfaces
• Misc – BOOT, Reset, Failsafe, and MCLR buttons
• Power Supply  – 5V DC input via USB-C port
• Dimensions – 80 x 22 mm (M.2 2280 form factor)

On the software side, LILYGO provides examples for WiFi HaLow communication, camera and display operation, and other board functions on GitHub. The examples use ESP-IDF 5.4.1 or newer, with the HaLow module controlled through SPI for data transfer and UART for AT commands. A browser-based configuration tool is also available for configuring the HaLow module through Web Serial.

We previously reported on the original LILYGO T-HaLow, which used an ESP32-S3 alongside the same Taixin WiFi HaLow module. It also supports an OV2640/OV5640 camera, and features an RJ45 Ethernet port, a microSD card slot, and an 18650 battery holder. In contrast, the T-Halow P4 uses a more powerful ESP32-P4 with MIPI camera and display interfaces, while adding WiFi 6 connectivity through the ESP32-C6.

The LILYGO T-Halow P4 is available on AliExpress for $36.98, and on the company’s official store for $26.57 from a China warehouse, or $31.07 for a 915 MHz version (H777) shipping directly from a US warehouse.

The post LILYGO T-Halow P4 – An M.2 2280 board with ESP32-P4, WiFi HaLow, and ESP32-C6 appeared first on CNX Software - Embedded Systems News.

Shanghai-based RoboParty has introduced the RP1 (ROBOTO 01), an open-source humanoid robot designed for robotics research, embodied AI development, and motion-control experiments. The robot is available in two configurations: the standard RP1 with 24 active degrees of freedom (DoF) and the RP1 Pro with 30 DoF.

The robot stands 1.35 meters (~4.43 feet) tall and is built around a Rockchip RK3588 processor, with the Pro model adding an Allwinner A733 processor and an NVIDIA Jetson Orin NX development computing module. Other features include in-house ROMOMO joint actuators with up to 160 N·m peak torque, interchangeable head modules, optional dexterous hands, and a 10Ah battery rated for up to three hours of operation. A separate 480MHz MCU handles real-time motion control, while a dual-voltage battery system supports hot-swapping without shutting down the robot.

RoboParty RP1 specifications:

• Computing
  • RP1 – Rockchip RK3588 processor with a 6 TOPS NPU, 8GB LPDDR4x RAM, and 32GB eMMC storage
  • RP1 Pro – Rockchip RK3588 + Allwinner A733 processor, NVIDIA Jetson Orin NX (16GB) development module
  • Real-time control – 480MHz MCU for motion scheduling, power sequencing, and emergency-stop functions
  • Optional development computing module for the standard RP1
• Display
  • RP1 – Features a motion-control head that uses a simple LED eye display
  • RP1 Pro – Features a dedicated interaction screen
• Camera
  • RP1 – No camera included as standard
  • RP1 Pro – Monocular RGB camera
  • Optional stereo RGB camera with the data-collection head
  • Optional Intel RealSense D435i depth camera and 3D LiDAR with the perception head
• Audio
  • RP1 – Speaker
  • RP1 Pro – Speaker + microphone and interaction display
• Degrees of freedom
  • RP1 – 24 DoF (6 per leg, 5 per arm, and 2 for the waist)
  • RP1 Pro – 30 DoF (6 per leg, 7 per arm, 2 for the waist, and 2 for the head)
  • Dexterous-hand degrees of freedom are counted separately
• Joint actuators – 4x in-house ROMOMO joint module designs using two-stage planetary gear reductions and inner-rotor motors; up to 160 N·m peak torque for the 100mm-diameter module
• Motion
  • Maximum running speed – Up to 4 m/s
  • Slope climbing – Up to 30°
  • Maximum step height – 50mm
  • Minimum obstacle size for avoidance – 50mm (manufacturer specification)
• Head and hands
  • Four interchangeable head configurations: locomotion, interaction, perception, and data collection
  • RP1 – Dummy hands as standard
  • RP1 Pro – 6-DoF dexterous hands as standard
  • Additional grippers and 22-DoF dexterous-hand options
• Sensors – 6-axis IMU
• Networking
  • 2x RJ45 Ethernet jacks for wired networking, debugging, and communication
  • Wireless – WiFi 6, Bluetooth 5.2, 4G cellular
• USB – 2x USB Type-A, 2x USB Type-C
• Expansion – Hardware interfaces and mounting points for additional sensors and peripherals
• Safety
  • Three-channel emergency stop protection and enclosed, pinch-resistant joints
  • Hardwired or 433MHz emergency-stop signaling to the real-time MCU
• Power Supply
  • Internal 24V wiring and adapter board for end effectors
  • 10Ah dual-battery system with hot-swapping
  • 48V main lithium battery and 30V auxiliary battery for maintaining critical control power during battery replacement
  • Up to 3 hours of operating time
  • 140W USB-C fast charging and magnetic charging
  • Approximately 1.5 hours charging time
  • Supports operation while charging
• Dimensions – 1350 x 450 x 200mm; Limb lengths – 250mm thigh, 300mm lower leg, and 550mm single-arm length
• Construction
  • AL6061 aluminum alloy shoulder, arm, wrist, thigh, and lower-leg components
  • AL7075-T6 aluminum alloy foot plates
  • Polycarbonate (PC) chest enclosure
  • Resin head enclosure and selected arm, thigh, wrist, and shin covers
  • Polyurethane (PU) hip-joint bumpers, protective covers, thigh pads, handgrips, and foot surfaces
  • Sandblasted, polished, and spray-painted finishes, depending on the component
• Weight – Approximately 39kg (RP1) / 41kg (RP1 Pro), including battery and end effectors
• Operating temperature – -10°C to +40°C

The standard RP1 features 5-DoF arms and a basic locomotion head, while the Pro version offers 7-DoF arms with additional wrist movements, a two-axis neck, and an interaction head. Both models support interchangeable end effectors, including RoboParty’s RP Hand, a five-finger dexterous hand with six actively controlled axes and CAN-FD communication. The RP Hand has 11 mechanical degrees of freedom, including five passive or coupled DoF, and uses a worm-gear transmission. It weighs approximately 379 grams and supports optional tactile sensing.

The robot uses RoboParty’s RBE-MAIN-RK3588S main controller, that features the RK3588 application processor with a separate  (unnamed) real-time MCU. The RK3588 handles AI inference, vision, speech, and interaction, while the MCU manages motion scheduling and emergency-stop functions. Each ROMOMO joint module also has a built-in driver that handles current, velocity, and position control loops locally, with the main controller sending commands and receiving feedback over CAN-FD.

On the hardware side, RoboParty has published its design resources in the RP1_hardware GitHub repository. These include editable SolidWorks assemblies and parts, Parasolid models, mechanical drawings, PCB fabrication files, Gerbers, component placement data, wiring harnesses, and bills of materials for the different robot configurations. The repository also provides manufacturing files for 14 custom circuit boards, including Gerbers, BOMs, component placement coordinates, and 3D models. The company’s original hardware designs are licensed under CERN-OHL-W-2.0.

The software is spread across different GitHub repositories, where the RP1_firmware repo hosts the Linux BSP, USB-to-CAN-FD bridge firmware, and software packaging tools. The roboparty_deploy repository provides a ROS 2-based deployment framework with motor and IMU drivers, Intel RealSense camera integration, ONNX-based policy inference, robot configuration files, and a Python SDK.

For simulation and training, the RP1_description repo provides URDF and MJCF robot models, meshes, and various assets, while the roboparty_train repo includes NVIDIA Isaac Lab-based reinforcement learning workflows, motion imitation, RSL-RL training, and MuJoCo Sim2Sim testing. The 24-axis robot configuration already supports a training-to-deployment workflow. However, the 28-axis body configuration used by the Pro model does not have dedicated training tasks or released default policy files at this time.

The company also provides various learning projects, including AMP, BeyondMimic, MimicLite, and UFO for motion imitation and humanoid control. These projects are part of the PartyOS ecosystem, which connects motion control, simulation, policy training, and deployment on physical robots. Some parts of the software are still under development, and the dedicated RP1 deployment guide in the deployment repository is marked as coming soon. The UFO repository also notes that its RP1 TeCH policy has not yet been tested on the physical robot.

Although RoboParty describes RP1 as a fully open-source platform, not all components are available as source code. The ROMOMO repository provides joint-module specifications, drawings, CAN-FD protocol documentation, and test reports, while host-side motor drivers are available in roboparty_motors. However, the embedded firmware source for the ROMOMO actuators could not be found in the published repositories. The UFO learning framework is also distributed under a CC BY-NC 4.0 non-commercial license.

We have previously reported on similar humanoid robots, including the Unitree R1, a 1.21-meter-tall robot with 26 DoF, and the Raspberry Pi-powered SamuRoid with 22 DoF. We also looked at the Mini π bipedal robot, which was also tagged as open-source but was limited to motion algorithms. RoboParty takes a different approach by open-sourcing mechanical CAD, PCB design files, firmware components, and software for simulation, training, and deployment.

At the time of writing, RoboParty does not confirm pricing or a shipping schedule for the RP1 and RP1 Pro. More details are available on the product page, the documentation website, and the main GitHub repository.

Thanks to TLS for the tip.

The post RoboParty RP1 (ROBOTO 01) – A full-stack open-source humanoid robot with up to 30 degrees of freedom appeared first on CNX Software - Embedded Systems News.

GL.iNet Mango 2 (GL-MG1300) mini dual-band WiFi 5 travel router is built around the MediaTek MT7621A dual-core processor. It is designed to share hotel WiFi, wired Ethernet, or mobile internet connections with multiple devices.

The router measures just 89 x 63 x 15mm and weighs 100 grams. It supports WiFi speeds of up to 400 Mbps on the 2.4 GHz band and 866 Mbps on the 5 GHz band, along with WireGuard VPN speeds of up to 184 Mbps and OpenVPN speeds of up to 95 Mbps with DCO. Other features include dual  Gigabit Ethernet ports, USB 3.0, USB-C power, and others.

GL.iNet Mango 2 (GL-MG1300) specifications:

• Processor – Mediatek MT7621A dual core MIPS 1004Kc processor @ 880Mhz
• Memory – 128MB DDR3L RAM
• Storage – 128MB SPI NAND flash + 2MB SPI NOR flash
• Networking
  • 2x Gigabit Ethernet ports (WAN/LAN)
  • Dual-band 802.11a/b/g/n/ac WiFi 5
    • 2.4 GHz – Up to 400 Mbps (2×2 MIMO)
    • 5 GHz – Up to 866 Mbps (2×2 MIMO)
    • 2x internal dual-band antennas
  • VPN – Up to 184 Mbps with WireGuard and 95 Mbps with OpenVPN DCO
  • Optional cellular connectivity through a compatible USB modem or smartphone tethering
• USB
  • 1x USB 3.0 Type-A port
  • 1x USB Type-C port for power
• Misc – Reset button, configurable toggle switch, status LED
• Power Supply – 5V/2A via USB-C
• Consumption – Less than 5W
• Dimensions – 89 x 63 x 15mm
• Weight – 100 grams
• Operating Temperature – 0°C to 40°C

The router also supports captive portals, so you only need to sign in once to share public WiFi with multiple devices. You can manage the router through GL.iNet’s web interface or mobile app.

The Mango 2 runs GL.iNet firmware 4.10 based on OpenWrt 22.03.4 and supports both WireGuard and OpenVPN client/server modes, with compatibility for more than 30 VPN service providers. It also supports Tailscale for remote access, multi-tunnel VPN configurations, and policies that allow users to select which devices or websites use the VPN connection.

We have previously written about other GL.iNet travel routers, including the Slate (GL-AR750S-Ext), Beryl AX (GL-MT3000), and Slate 7 (GL-BE3600). The Mango 2 is a more compact, entry-level model with dual-band WiFi 5 instead of the WiFi 6 on the Beryl AX or WiFi 7 on the Slate 7. It retains Gigabit Ethernet, USB 3.0, and VPN support, making it suitable for basic travel networking without the higher wireless speeds of the newer models.

The GL.iNet Mango 2 (GL-MG1300) is listed on AliExpress for $ 62.99 and on the company’s official store for $49.99. However, the company doesn’t sell it in the US due to FCC-related restrictions, which will also impact all its future routers. The company offers a 2-year warranty and a 30-day money-back guarantee.

The post GL.iNet Mango 2 mini travel router features AC1300 WiFi 5 and dual Gigabit Ethernet appeared first on CNX Software - Embedded Systems News.

Following our coverage of a 4.3-inch ESP32-S31 display development board, a Facebook reader pointed out the GUITION JC3636F118C-I-YR1 rotary encoder with a built-in 1.8-inch touchscreen display.

It appears to be an update to an earlier ESP32-S3 model, now equipped with an ESP32-S31NRV16 WiFi 6, Bluetooth 5.4, and 802.15.4 SoC with 16MB of PSRAM. The rotary knob also includes 16MB SPI flash, a speaker, a microphone, and an optional rechargeable battery.

GUITION JC3636F118C-I-YR1 specifications:

• SoC – ESP32-S31NRV16
  • CPU
    • RISC-V HP (High-performance) RV32IMAFCP CPU @ 320 MHz with FPU, SIMD, etc.
    • RISC-V LP (Low-power) MCU core
  • Memory – 512 KB SRAM, 32 KB low-power SRAM, 16MB PSRAM
  • Wireless
    • 2.4 GHz WiFi 6 (802.11ax)
    • Bluetooth 5.4 with support for LE Audio, direction finding, Bluetooth Mesh 1.1, and Classic (BR/EDR)
    • 802.15.4 radio for Zigbee, Thread, and Matter
• Storage
  • 16 MB flash
  • MicroSD card slot (for WiFi configuration, music or MJPEG video playback, etc.)
• Display
  • 1.8-inch round 262K color TFT display with 360×360 resolution
  • ST77916 SPI driver
  • CST816 capacitive touch controller
  • Adjustable backlight
• Audio
  • Built-in speaker (PCM5100APWR)
  • Built-in MEMS microphone (MSM261D4030H1CPM)
• Control – CNC knob with haptic feedback (DRV2605LDGSR haptic driver IC)
• USB – USB Type-C port (orientation-sensitive)
  • Reverse insertion (Type-C notch facing up) for file transfer
  • Normal insertion (Type-C notch facing down) for firmware programming.
• Power Supply
  • 5V via USB-C port
  • Optional rechargeable battery (capacity not listed)
• Power Consumption – About 1.3 Watts
• Dimensions – 66mm Ø x 22 mm
• Weight – About 160 grams

GUITION JC3636F118C-I-YR1 comes preloaded with firmware with a clock, a countdown timer, a music player, an MJPEG and AVI video player, a photo frame function, a spectrum analyzer when playing audio, Bluetooth audio, USB audio, and USB HID (knob controls the volume on your computer).

The company also mentions “secondary development in environments such as Arduino IDE, ESP-IDF, MicroPython, and GUITION” program, but the provided zip file only contains ESP-IDF 6.1.0 samples (demo firmware, LVGL-based music & MJPEG video player, Xiaozhi voice assistant) and a currently empty Arduino folder, along with basic instructions, specifications, datasheets, schematics (PDF), and tools.

The main benefits of upgrading from ESP32-S3 to ESP32-S31 in this type of hardware are the faster cores, multimedia capabilities (2D graphics engine, MJPEG decoder), a wider range of wireless options, and a 480 Mbps USB 2.0 interface. The downsides are mainly that the ESP32-S31 is rather new, so firmware is not quite as mature, and it’s slightly more expensive.

GUITION sells the new JC3636F118C-I-YR1 ESP32-S31-based rotary encoder for $40.09 on AliExpress including shipping (but not taxes) with the battery, and $38.49 without the battery. For reference, the earlier ESP32-S3-based “JC3636K518C” model goes for $39.13 on AliExpress with a battery (and I’m shown a $33.39 “SuperDeals”).

Thanks to Rob for the tip.

The post Rotary encoder with 1.8-inch touch display gets upgrade to ESP32-S31 SoC appeared first on CNX Software - Embedded Systems News.

Espressif Systems ESP-Mosaico is an ESP32-S31 development platform with a 480 × 480 OLED touchscreen, haptic feedback, a built-in battery, USB Type-C connectivity, and expansion I/O.

It’s specifically designed for AI <-> hardware integration and targets magnetic expansion toys, motion sensing, square touch display HMI, and edge voice/multimedia applications.

ESP-Mosaico specifications:

• CoreBoard
  • SoC – ESP32-S31NRV16
    • CPU
      • RISC-V HP (High-performance) RV32IMAFCP CPU @ 320 MHz with FPU, SIMD, etc.
      • RISC-V LP (Low-power) MCU core
    • Memory – 512 KB SRAM, 32 KB low-power SRAM, 16MB PSRAM
    • Wireless
      • 2.4 GHz WiFi 6 (802.11ax)
      • Bluetooth 5.4 with support for LE Audio, direction finding, Bluetooth Mesh 1.1, and Classic (BR/EDR)
      • 802.15.4 radio for Zigbee, Thread, and Matter
  • Storage
    • 1 Gbit (128 MB) SPI NAND flash for asset storage.
    • 16 MB (128 Mbit) SPI NOR flash (GD25Q128EWIGR) for program and data storage.
  • Display – 2.16-inch QSPI touch display with 480 x 480 resolution; CO5300 driver + CST9220 touch
  • Audio
    • NS4150B 3W mono Class-D amplifier
    • LMA2718B331 microphone connected to ES8311 low-power mono audio codec
  • USB – USB 2.0 Type-C port for power/charging and programming
  • Sensors
    • 2x 3-axis magnetometers (BMM150)
    • 6-axis accelerometer + gyroscope (BMI270)
  • Misc
    • Power, Function, and BOOT buttons
    • Charge LED
    • Vibration motor
    • 2.4 GHz chip ceramic antenna
  • Power Management
    • 5V input via USB-C port
    • TLV62569 buck converter (3.3V)
    • SY7088 boost converter
    • TP4057 Li-ion battery charger
    • BQ27220 fuel gauge
    • HE9073A33 LDO for audio codec
• BaseBoard connected over a 60-pin B2B connector to the CoreBoard
  • Audio – Onboard speaker driven by the NS4150B amplifier.
  • Expansion
    • 2x 20-pin 2.54 mm pitch headers
    • Back expansion pads with GND, BOOT, RST, RX, TX, 5V, and GND
  • Power management
    • 3.7 V / 65 mAh Li-ion battery
    • TLV62569 buck converter (3.3V)
• Dimensions – TBD

The ESP-Mosaico is advertised as a “modular” development platform, and for now, the company has designed two expansion modules:

• The Camera Module supports a 3.2 MP OV3640 sensor (PCB V1.2) or a 1MB SC101IOT sensor (PCB V1.4); both used at 1024 × 768 resolution
• The Multi-function Interaction Module features a GL5506 photoresistor, an AS312 PIR sensor, an IR LED, six WS2812 RGB LEDs, and two pads for mechanical keys or capacitive-touch channels.

Espressif explains that the Mosaico firmware enables two development approaches:

• Chat Coding through ESP-Claw – It enables users to rely on natural language to quickly build working hardware prototypes without setting up a local development environment.
• Vibe Coding through ESP-ZeroCode AI – In this mode, AI Agents handle most of the engineering workflow, including implementation, deployment, testing on physical hardware, and continuous iteration.

For some reason, there are two places for documentation. The Espressif documentation website provides more technical details about the hardware, while the ESP-Mosaico Guide includes instructions to get started with the Chat and Vibe coding mode described above.

The ESP-Mosaico has been on Taobao and third-party AliExpress stores (at inflated prices) for a couple of months, but you can now purchase the complete devkit with camera and interaction module for $59 on the Mosaico store. You’ll also find an optional $15 battery module there.

The post ESP-Mosaico – An ESP32-S31-based modular development platform for AI-assisted hardware creation appeared first on CNX Software - Embedded Systems News.

The Raytron SE5 1280 is a high-resolution uncooled infrared module designed for integration into OEM thermal imaging systems. Based on a 1280 × 1024 uncooled VOx infrared detector with a 12 μm pixel pitch, the module delivers more than 1.3 million infrared pixels in a compact form factor measuring no more than 29 × 29 × 15.7 mm and weighing less than 26 g.

1280 × 1024 VOx Detector for High-Resolution Thermal Imaging

At the core of the SE5 1280 is a 1280 × 1024 uncooled VOx infrared detector with a 12 μm pixel pitch, operating in the 8–14 μm long-wave infrared (LWIR) band.

With more than 1.3 million infrared pixels, the detector provides a substantially higher spatial resolution than conventional 640 × 512 thermal imaging configurations. This can help thermal imaging systems capture finer temperature patterns and spatial details, particularly when identifying relatively small thermal targets or distinguishing objects that are close together in the image.

The detector also offers a NETD of ≤40 mK, indicating its ability to detect relatively small differences in thermal radiation under specified test conditions. Combined with the high pixel count, the detector provides the underlying sensing capability for applications where both thermal sensitivity and image detail are important.

The SE5 1280 supports thermal imaging at up to 60 Hz, while temperature measurement can be performed at up to 30 Hz, making the module suitable for applications that require responsive thermal video as well as temperature data.

Dedicated ISP for Thermal Image Processing

The SE5 1280 uses Raytron’s LY300 dedicated infrared image processing ISP chip, which is specifically developed for Raytron’s uncooled infrared detectors.

While the detector captures infrared radiation and converts it into electrical signals, the LY300 processes the raw infrared data and converts it into a usable thermal image. The chip integrates a range of image-processing, video-processing, and temperature-analysis functions to support the requirements of uncooled thermal imaging systems.

LY300 supports resolutions of up to 1920 × 1080 at 30 fps and provides key image-processing functions including non-uniformity correction (NUC), dead-pixel correction, horizontal and vertical stripe removal, spatial and temporal noise reduction, wide dynamic range (WDR) processing, detail enhancement, edge enhancement, and brightness adjustment.

For thermal analysis, the processor supports temperature measurement and isotherm analysis. It also provides video-processing functions such as image cropping, image flipping, up to 8× continuous digital zoom, pseudo-color mapping, and OSD.

The LY300 supports multiple video and control interfaces, including DVP, MIPI, USB 2.0, VOSPI, UART, I²C, and SPI, providing flexibility for integration into different OEM thermal imaging systems.

Key LY300 Specifications

Shutterless Non-Uniformity Correction

Pixel-to-pixel response differences are an inherent characteristic of uncooled infrared detectors. Even when individual pixels receive the same infrared radiation, their electrical outputs can differ due to variations in detector materials, microstructure, and readout circuits.

Thermal detectors can also experience response changes as their operating temperature and surrounding conditions change. Non-uniformity correction is therefore required to maintain image quality during operation.

The SE5 1280 supports shutterless NUC, allowing the module to perform non-uniformity correction without relying on a mechanical shutter as the reference for periodic correction.

By reducing dependence on a mechanical shutter, this approach can be useful for applications that require continuous thermal imaging as well as compact mechanical designs with fewer moving components.

Shutterless operation does not mean that NUC is unnecessary. Instead, the correction is handled algorithmically rather than relying on a mechanical shutter event.

Compact Form Factor for OEM Integration

The SE5 1280 can be configured with 19 mm, 35 mm, 55 mm, and 75mm lens options, allowing manufacturers to select an appropriate field of view and working distance for different applications.

Commercial Drone and Aerial Imaging

The module’s compact size, low weight, and low power consumption make it suitable for UAV payloads where size, weight, and power consumption are important design constraints. Its 1280 × 1024 resolution provides detailed thermal imagery for aerial inspection, observation, and mapping applications.

Automotive and ADAS

The SE5 1280 can be integrated into automotive thermal imaging systems and ADAS applications, providing infrared sensing alongside conventional visible-light cameras. Its high-resolution thermal imaging can provide additional information in low-light and reduced-visibility conditions, complementing visible-light sensors for automotive perception.

Security and Monitoring

For security and monitoring applications, the combination of 1280 × 1024 resolution and ≤40 mK NETD enables the module to capture detailed thermal imagery and detect subtle temperature differences. This makes it suitable for fixed or mobile thermal imaging systems operating in low-light or challenging visibility conditions.

Industrial Inspection and Temperature Measurement

The module supports temperature measurement, isotherm analysis, and a range of image-processing functions, making it suitable for industrial thermal inspection and temperature monitoring. Its compact form factor also allows thermal imaging capabilities to be incorporated into space-constrained industrial equipment.

Handheld and Outdoor Imaging

The SE5 1280 combines a compact form factor, a weight of less than 26 g, and typical power consumption below 1.3 W, making it suitable for handheld thermal imaging products and portable outdoor equipment. Its multiple video interfaces also provide flexibility for integrating the module into different portable imaging platforms.

Looking Ahead in Infrared Imaging

The SE5 1280 brings together high-resolution infrared sensing and dedicated image processing in a compact OEM module, providing a practical platform for manufacturers developing thermal imaging products.

The module also reflects Raytron’s broader development of infrared imaging technologies across sensing and processing. Alongside thermal imaging modules such as the SE5 1280, Raytron is developing dedicated processing platforms including Falcon500, its third-generation infrared AI image processor, for the next generation of intelligent infrared imaging applications.

The post Raytron SE5 1280: 1280×1024 Uncooled Thermal Imaging Module for OEM Applications (Sponsored) appeared first on CNX Software - Embedded Systems News.

Earlier this month, we saw how the humble ESP32-S3 could be turned into an SDR transmitter and receiver; then we saw developer h0m3us3r release the w1700k-sdr project, which turns a Gemtek W1700K Wi-Fi 7 router into a wideband SDR receiver.

The LR2021-SDR open-source project by Samuel Reynolds (SAMS0N1TE) follows the same principle and turns the Semtech LR2021 LoRa transceiver into a software-defined radio (SDR) capable of receiving signals from 100 MHz to 2.74 GHz. Using an undocumented command in the chip’s ROM, this project exposes raw 16-bit I/Q samples, allowing software to decode P25 Phase 1 digital voice, POCSAG pager messages, and analog FM signals without an external SDR dongle.

LR2021-SDR firmware features:

• Supported hardware – Semtech LR2021 running ROM 1.24; not compatible with the SX1262 variant for raw I/Q access.
• Receive frequency range – 100 MHz to 2.74 GHz (experimentally tested), extending beyond the datasheet’s specified 150–960 MHz and 1.5–2.5 GHz bands.
• Transmission frequency range – 100 MHz to 2.74 GHz (experimentally tested), extending beyond the datasheet’s specified 150–960 MHz and 1.5–2.5 GHz bands.
• I/Q capture – 16-bit I and Q samples over SPI, with clean short captures at up to 250,000 complex samples per second
• Continuous reception – 125,000 samples/s for up to 391 seconds without reported FIFO loss or dropped sample blocks
• Signal decoding – P25 Phase 1 digital voice, POCSAG pager messages, and analog FM through LakeShark.
• Bluetooth LE – Reception of BLE 1M advertising packets on channels 37, 38, and 39 using the chip’s built-in BLE packet engine
• Recording – Raw I/Q captures for offline signal analysis, decoding, and replay, including triggered recordings with pre-trigger samples

Normally, the LR2021 handles LoRa, FSK, FLRC, and other modulation and provides decoded packets to the host processor. The LR2021-SDR project bypasses this normal operation using an undocumented ROM command that exposes the raw baseband signal. The processor can then process the I/Q samples in software, allowing it to receive signals that the chip’s standard packet interface does not directly support.

The developer tested the receiver across the 100 MHz to 2.74 GHz range using the board’s clock harmonics and separate signals generated by a HackRF One inside a shielded enclosure. The tests confirmed reception at 100 MHz, 1.3 GHz, 2.6 GHz, and 2.74 GHz. However, the measurements were performed on a single board, and the results do not establish calibrated receiver sensitivity or guarantee the same frequency coverage on other hardware.

The developer also tested continuous I/Q capture at 125,000 samples per second, with three sessions running for up to 391 seconds without reported FIFO loss or queue rejection. Short captures at up to 250,000 samples per second also worked, but higher rates of 500,000 and 1,000,000 samples per second caused FIFO overflows because the 16 MHz SPI interface could not transfer the data fast enough.

LR2021-SDR also supports experimental signal transmission, including P25 voice test signals, temperature sensor data, and OOK/FSK remote-control signals. In one test, a separate receiver successfully decoded 72 P25 voice frames and recovered the original test tone. Other tests included sending temperature sensor data received by rtl_433 and replaying remote-control signals decoded by a Flipper Zero. For now, transmission is limited to short bursts, and continuous transmission and live voice support are still under development.

LakeShark (used above) is an open-source firmware project by the same developer that turns the LilyGO T-Display P4 into a handheld SDR scanner using an external RTL-SDR dongle. With LR2021-SDR, the developer has now enabled raw I/Q capture directly from the onboard Semtech LR2021 transceiver, allowing LakeShark to receive and decode P25 voice, POCSAG pager messages, and analog FM signals without an external SDR dongle.

The project’s GitHub repository includes the GPL-3.0-licensed C library, standalone ESP32-P4 capture firmware, Python tools, build instructions, and detailed test results. Users can install LakeShark for live decoding and handheld operation or use the standalone firmware to capture I/Q samples over USB for analysis on a computer. The project remains experimental, and an LR2021-equipped board with an appropriate antenna is required.

Via Hackaday

The post Semtech LR2021 LoRa chip gets the SDR treatment, supports 100 MHz to 2.74 GHz appeared first on CNX Software - Embedded Systems News.

Linus Torvalds announced the release of Linux 7.3 on LKML last Sunday:

It’s Sunday afternoon, and this week I somehow kept my wits together long enough to keep track of the day of the week, so the release isn’t randomly delayed. Will wonders never cease?

Anyway, nothing very special seems to have happened this last week, so as expected the final 7.3 is out, and thus tomorrow the merge window for 7.4 opens. Except I’m on my annual family vacation, so instead of merging anything first thing in the morning, I’ll spend most of the day on a boat and under water. I’ll get back to merging in the afternoon, but I expect my schedule to be a bit unusual for a while.

I hope I can still complete the merge window in the usual two weeks, but it may be that I’ll end up with an extra week of merging just to catch up.

Note that that extra week doesn’t mean that I’ll accept pull requests that come in past the usual timing – the rules are still that what gets merged should be ready *before* the merge window opens, not some kind of late development *during* the merge window. So don’t try any shenanigans.

Below is the shortlog for the last week of 7.3 development. It looks fairly spread out, no real pattern to it: random fixes all over.

Linus

Released about two months ago, Linux 7.2 brought us cache-aware task scheduling for faster performance and fewer cache misses, a fair(er) GPU scheduler, USB4STREAM for streaming data with USB4, and phase IV of the swap table implementation, among other changes. Now, let’s look at Linux 7.3 with a list of key changes and a deep dive into Arm, RISC-V, and MIPS architecture changes.

Some Linux 7.3 highlights

Some notable changes to Linux 7.3 include:

• Improved VRAM management for low-end GPUs – When your GPU does not have enough VRAM (Video RAM), different applications can end up competing for VRAM. This release allows using the dmem cgroup to alleviate the situation where two applications compete for RAM. Instead of backing off immediately, the system will try evicting unprotected buffers from the domain to make space for the protected one. This ensures that applications can actually use all the memory protection awarded to them by the system, without being prone to ping-ponging. More details on pixelcluster.dev.
• Btrfs and Ext4 performance improvements
  • Btrfs regains the Direct I/O performance it lost in previous releases. For non-SSD disks, the removal of an unnecessary 1 jiffy delay during logging workloads leads to decreased latency and 5x increased throughput. fsync with the no-holes feature enabled (default since btrfs-progs 5.15) is also 5x faster.
  • Ext4 improves performance of direct I/O parallel writes and fallocate() performance, and a buffered-write regression has been fixed.
• Better FUSE io_uring performance – Linux 7.3 allows FUSE servers to pass in a buffer pool that the kernel will use for servicing requests, in a way that reduces the memory usage requirements. It also adds zero-copy to FUSE io-uring to further improve performance. Check out the official documentation for details.
• Support for cgroup sub-schedulers (sched_ext) – The implementation of sched_ext is now complete. It allows a cgroup have its processes be managed by its own BPF-based task scheduler. Different cgroups can have different BPF sub-schedulers and take different scheduling decisions. Read this LWN article for more information.

Arm updates in Linux 7.3

• Allwinner
  • Clk driver  – Some cleanups to the Allwinner clock driver library and support for the Allwinner A733 RTC clocks added.
  • Allwinner A733 – Add MMC, RTC, NMI controller support
  • Device tree changes for Linux 7.3
    • A10 – PMU interrupt number fixed
    • A100/A133 – Gains LRADC
    • Enabled overlay support for sun8i-h2-plus pi boards
    • A523 – Gains consumer IR receiver
    • H728-based X96QPRO+ TV box gets S/PDIF audio output
    • PinePhone’s mpu6050 mount orientation fixed
    • Various style fixes
  • Defconfig Arm64 – Enable the LRADC
  • New devices
    • Baijie Helper A133
    • NetCube Systems OpenNMC (dobermann)
• Rockchip
  • Clock driver
    • Clock driver for Rockchip RV1106
    • Fix for Rockchip rk3576 spi source mux
    • Better representing fractional PLL coefficients in Rockchip clk driver
    • Fix for the Rockchip dclk of the 3rd video-port to not affect other components when its rate gets changed
  • PHY – Updated Rockchip clock lane phase tuning and 2500 Mbps support and TMDS rate handling
  • Pinctrl – Added support for Rockchip RK3308B and RV1106 SoCs
  • IOMMU – Rework Rockchip state handling and fix clock, probe, and stale-fault handling
  • CAN-DF controller – Add support on RK3588
  • Watchdog – Added for RV1106
  • New devices
    • EmbedFire LubanCat 4 board (RK3588S)
    • Anbernic RG Vita-Pro game console (RK3576)
    • Vicharak Vaaman2 and Axon boards (RK3588)
    • LCKFB Taishan Pi 3M board (RK3576)
    • Orange Pi 5 Pro board (RK3588S)
    • Graperain G3568 v2 board (RK3568)
    • HINLINK/LinkStar H28K router (RK3528)
    • Youyeetoo YY3588 SBC (RK3588)
    • Alientek DLRV1126 carrier board and CLRV1126F core module (Rockchip RV1126)
    • LincStation E1
• Amlogic
  • Clock driver
    • Added Amlogic A9’s AO and peripheral clks
    • Fix the incorrect parent number of the 32k clock on Amlogic GXBB
  • Meson PCIe controller driver – Correct the PERST# GPIO state so it remains asserted until power and REFCLK become stable to fix enumeration failure
  • SoC driver – Amlogic SoC support for A1 and T7 is added
  • Amlogic drivers
    • Clock measure debugfs fixup
    • Clock measure support for Amlogic A1 & T7
  • Arm device tree – N/A
  • Arm64 device tree for Linux 7.3
    • Amlogic AXG fixups for PCIe and NAND
    • Overall indentation and whitespace fixups
    • Amlogic A9 Pinctrl & GPIO IRQ nodes
    • Clock measure nodes for Amlogic T7 & A1
  • New device – N/A
• Samsung
  • Added Samsung Exynos 1580 high-end mobile phone SoC (from 2024)
  • PHY – Samsung hdptx driver improvements for modernizing the register access and code cleanup
  • Pin control – Added Samsung Exynos 8855 SoC
  • Input
    • Updates to Samsung keypad driver to keep interrupts disabled while the device is closed, along with wakeup logic cleanups and use of pm_runtime_active guards
    • Update to Samsung SUR40 contact count based on PixelSense specification
  • Clock driver
    • Exynos990: Add few missing clocks and correct the gate clock parents in the PERIS clock controller.
    • Cleanup – Use kzalloc_flex for __counted_by checks in Samsung clk driver.
  • SoC drivers
    • Fix very old leak of OF node in PMU driver.
    • Few cleanups and new bindings for Exynos Auto v920 HSI0 syscon.
  • DTS Arm changes – Add Bluetooth to Exynos5250-based Google Manta (Nexus 10) board.
  • Samsung DTS Arm64 changes for Linux 7.3
    • New SoC Exynos8855 (market naming: Exynos 1580) – basic support with its SMDK development board.
    • ExynosAutov920: Add HSI0 syscon node.
    • Several non-functional (white-space and style) changes to match with DTS coding style as reported by recent dt-check-style tool. Goal is to get to no dt-check-style --mode strict warnings over time.
  • Defconfig changes – N/A
  • New Devices – Samsung Exynos8855 SMDK
• Qualcomm
  • New SoC – “Mysterious” Qualcomm’s Cortex-A78C-based “Shikra”
  • ASoC – Qualcomm LPASS fixes for playback constraints and WSA macro volume
  • Backlight – Qualcomm WLED: Remove redundant dev_err() calls
  • MFD
    • Qualcomm SPMI PMIC: Document PMG1110 and haptics device
    • Qualcomm TCSR: Add compatible for Hawi, Maili, and Shikra SoCs, document the IPQ9650 TCSR block
  • Thermal
    • Add Qualcomm Master BandGap thermal monitor support
    • Add Maili Temperature bindings compatible
  • Clock driver
    • Qualcomm Maili GCC, TCSR, RPMh, and video clks
    • Qualcomm Shikra GCC, RPM, GPU, display, and audio clks
    • Qualcomm Nord display and graphics clks
    • Qualcomm Glymur camera and EVA clks
    • Qualcomm Hawi video clks
  • PHY
    • Qualcomm SM8475 QMP USB PHY and PCIe PHY, IPQ9650 QMP PCIe PHY, QUSB2 PHY for Shikra SoC, Hawi support for QMP PCIe PHY and UFS PHY.
    • Glymur QMP PCIe Multi-PHY driver and multiple link-mode support, IPQ5210 PCIe PHY support
    • Qualcomm drop duplicate v8 DP headers, improved runtime handling for QMP drivers
  • Soundwire – Qualcomm SCP address paging, bus mclk_freq support. Increase of data ports to 17 and driver improvements
  • Pin control
    • Added Qualcomm Maili TLMM SoC, PMG1110 PMIC, and Eliza LPASS LPI TLMM SoC
    • Qualcomm TLMM GPIOs are unconditionally marked as wakeup capable
    • Qualcomm improvements on top of the Qualcomm PDC (power domain controller) irqchip improvements from tglx:s tree
    • Qualcomm IPQ806x and IPQ9650 non-urgent pin group fixups
  • PCIe controller driver
    • Add IPQ9650 compatible with global interrupt
    • Add IPQ5210 compatible with IPQ9574 fallback
    • Add DT binding and driver support for Hawi SoC
    • Skip PERST# GPIOs provided by downstream PCIe devices, which should be handled by drivers of those devices
    • Stop advertising Attention Button Present (no Qcom SoCs support Attention Buttons) so pciehp can use Presence Detect Changed events
  • WiFi – Qualcomm (ath12k)
    • Much kernel infrastructure integration work
    • AHB platform MultiPD support
  • IOMMU
    • ARM SMMUv2
      •  Device-tree binding updates for Qualcomm Eliza, Maili, Shikra and IPQ9650 SoCs
      • Add support for Qualcomm SM8450
    • Fix Qualcomm runtime-PM, probe unwind, fault reporting and page table initialization races
  • ASoC
    • Updates in Qualcomm driver stack
    • New platforms: Qualcomm LPI MI2S, SM8475, WSA855X
  • I2C – Added support for Qualcomm Nord SA8797P, SA8255p
  • Driver updates
    • Add Maili platform bindings for SCM, IMEM, AOSS, and PMIC GLINK. Add Shikra IMEM, RPM SMD, LLCC, and UBWC support, including handling for firmware-configured LLCC ECC interrupts.
    • Add the generic Peripheral Authentication Service with SCM and OP-TEE backends, and migrate the MSM DRM and IPA drivers to it.
    • Add SCM SMC-call tracepoints and configurable minidump delivery through always-on SRAM. Correct SCM download-mode pointer ordering and improve download-mode probe diagnostics.
    • Rework the UBWC configuration database and add Milos and Shikra configuration. Add protection-domain mappings for SA8775P and QCS8300, support newer ICE versions, Hawi subsystem statistics, and SDM850 identification.
    • Simplify Qualcomm SoC Kconfig selection and architecture dependencies.
    • Use managed resources in EBI2 and RPMh RSC probe paths to correct cleanup on failures.
    • Add SMEM parsing for DDR configuration data and use its highest bank address bit to select the appropriate UBWC configuration.
    • Enable generic PAS trusted-zone APIs for the Iris and Venus media drivers.
    • Fix SCM probe retry state, reserved-memory cleanup, and an early IRQ-handler NULL dereference. Enable QSEECOM EFI variable access on the Asus Zenbook A16.
    • Correct GENI firmware-size validation using the hardware CFG RAM depth and correct the PMIC GLINK Thunderbolt extradata layout.
    • Document the Nord AOSS side channel and the IMEM minidump SRAM property.
    • Clean up Qualcomm statistics macros and the WCNSS binding schema.
  • Arm64 device tree updates
    • Introduce DeviceTree support for the Shikra SoC, its GCC and RPM clock controllers, CQM, CQS, and IQS SoM platforms, and their evaluation boards. Document the Hawi and Maili SoCs.
    • Add the Eliza CQS SoM and EVK platforms, the QCS8550 RB5 Gen 2, Vicharak Axon Mini, HONOR MagicBook Art 14, Microsoft Surface Pro 12, Xiaomi 12 Lite, Motorola Edge 30, Sony Xperia M2, and Motorola Moto G2 device trees.
    • Extend Eliza support with USB, SD card, touchscreen, PMIC, interconnect, thermal, CPU and LLCC bandwidth-monitor, QUPv3, and ADSP GPR descriptions.
    • Expand Glymur support with GPU, camera and EVA clock controllers, LPASS audio, CoreSight, PCIe, USB, IMEM and PIL memory regions, power domains, thermal cooling, and CRD peripherals. Add SoCCP, PMIC regulator, TRNG, and CPU-capacity descriptions.
    • Add display, IPA, camera, USB, audio, crypto, TRNG, CoreSight, and thermal support to the Kaanapali, Milos, Hamoa, and related platforms. Add embedded controller support for Hamoa and Glymur boards.
    • Improve networking-platform descriptions with IPQ PCIe port, PHY, clock, PWM, regulator, and interrupt updates. Update CTCU, display, Iris, audio, camera, PCIe, USB-C, power-domain, memory-region, and thermal descriptions across established Qualcomm platforms.
    • Correct bindings and DeviceTree validation for Qualcomm compatible strings, legacy fallback compatibles, node naming and formatting, address ranges, clock specifiers, reserved memory, regulators, GPIOs, and peripheral wiring.
    • Add IPQ9574 PCIe bridge descriptions and enable PCIe0 and PCIe2 for QCN92xx Wi-Fi cards on RDP454. Align IPQ5018, IPQ6018, and IPQ8074 PCIe PHY and reset descriptions with the port-node binding.
    • Extend Shikra platform enablement with CQM and IQS SoM Wi-Fi and Bluetooth descriptions. Enable CDSP, LPAICP, and MPSS remote processors on CQM, CQS, and IQS EVKs, and add SMP2P, QUPv3, DDR bandwidth monitor, CPU frequency
      hardware, EPSS L3 interconnect, OPP tables, thermal zones, GPIO reservations, and display and GPU clock controllers.
    • Enable Eliza CQS EVK audio and add its LPASS macro and SoundWire descriptions. Add PMIC ADC support for Hamoa, Monaco, and Lemans, and ADC channel macros for the third-generation ADC block. Add minidump SRAM configurations across affected Qualcomm platforms.
    • Fix QMP PHY vdda-phy and vdda-pll supply assignments across Qualcomm platforms. Correct PCIe IOMMU mappings for QCS6490, Lemans, Monaco, Talos, SM, SC, and related boards. Fix the Glymur/Mahua PDC interrupt mapping, USB HS properties, and wakeup configuration; correct board identifiers, pin configuration, GPIO supplies, default bus frequency, and
      DT validation issues.
    • Enable the X1E80100 deepest idle state, configure SPMI ADC thermal monitoring on SM8550 and SM8650, add the Milos GPU clock controller, and assign SM8650 CCI clock rates.
  • Arm32 device tree updates
    • Add initial device tree support for the MSM8926-based Sony Xperia M2 and the MSM8226-based Motorola Moto G2 (2014).
    • Add touchscreen, Bluetooth pin configuration, NFC, and multicolor status LED support to the MSM8974-based HTC One M8 device tree.
    • Add the coreriver,tc360-touchkey compatible touchkey node to the MSM8960 Samsung ExpressATT device tree and label PM8921 XOADC channels.
    • Consolidate shared IPA properties in the SDX55 device tree.
  • Arm64 defconfig updates
    • Enable the Qualcomm reference device embedded-controller driver for Hamoa and Glymur
    • Enable the DMIC sound driver for Monaco and Lemans platforms
    • Enable the BAM-DMUX WWAN driver for Shikra and other platforms using A2 BAM hardware.
  • New Devices
    • Asus Zenbook A16 (UX3607OA)
    • Eliza CQS SoM platform and Eliza CQS EVK board
    • Vicharak Axon Mini SBC (QCS6490)
    • Motorola Moto G2 (2014 phone)
    • Sony Xperia M2
    • Microsoft Surface Pro 12in
    • Shikra IQ2390S/CQ2390M system-on-modules and EVKs
    • Xiaomi 12 Lite 5G (taoyao)
    • Motorola Edge 30 (dubai)
• MediaTek
  • MFD driver
    • MediaTek MT6397: Add mt6323 AUXADC and EFUSE support
    • MediaTek MT8195 SCP: Add support for MT8189 SoC
  • Clock driver – Add MediaTek MT8173 MFG_TOP
  • PCIe controller driver – Add support for PCIe controller in EcoNet EN7528 and EN751221 SoCs
  • PCIe Gen3 controller driver – Add mediatek-gen3 ‘memory-region’ for restricted DMA buffer
  • WiFi (mt76):
    • MT7928 support
    • MT7925 NAN support
    • MT7996 AP powersave improvements
  • SoC driver updates
    • Fix a missing error check in mtk-socinfo, and also a missing MODULE_DEVICE_TABLE() in the mtk-dvfsrc driver.
    • Rework how the MMSYS tables are specified with the macro, separating the component’s hardware instance from the actual type, resolving a prerequisite for the restructuring of mediatek-drm.
  • Arm32 updates
    • Fixes a minor indentation issue in the MT8127 device tree
    • Adds support for the Auxiliary ADC found in the MT6323 PM
  • Arm64 device tree updates for Linux 7.3
    • Indentation/whitespace fixes for some device trees
    • Documentation improvements for MediaTek Mute-X
    • dtbs-check warning fixes
      • MT8390 Tungsten boards: removal of unnecessary cells
      • MT6795 SoC: removal of deprecated property in UART DMA
    • MT6359 PMIC’s RTC now uses the correct compatible string
    • MT8183/86/88/92/95 SoCs gain support for the UART DMA IPs
    • MT8186 gets a new CCI OPP table to resolve lockups
    • MT8188/83 Chromebooks now got the correct SPI NOR supply
    • MT8192 Asurada drops support for the (unused) MMC1 port
    • MediaTek Genio 510/700/1200 EVKs and Radxa NIO-12L now get their PMIC power input tree completed with assignment of
      the correct supplies
  • Defconfig updates – N/A
  • New devices – N/A
• Other new Arm hardware platforms and SoCs
  • Altera – Cortex-A720-based Agilex72 SoC FPGA.
  • Apple – Gains support for M3 Pro, Max, and Ultra SoC (based on basic M3 support from Linux 7.2),
  • NVIDIA – Lenovo ThinkEdge SE70 desktop box (Jetson Xavier NX)
  • NXP
    • Variscite DART-MX8M PLUS module, Sonata board
    • Variscite VAR-SOM-MX8 QuadMax module, Symphony board
  • Texas Instruments
    • BeagleBadge (AM62L3)
    • Variscite VAR-SOM-AM62 and Symphony board
• Raspberry Pi-specific changes
  • V3D – Reduce PM runtime autosuspend delayThe 100ms autosuspend delay was only ever a workaround: shorter delays caused more frequent runtime suspend/resume cycles on the BCM2711 (Raspberry Pi 4), which exposed the cache and MMU coherency bugs as random GPU hangs.With those hangs resolved, the inflated delay is no longer necessary. Reduce it from 100ms to 50ms so the GPU power domain can be released sooner once the GPU goes idle.
  • VC4
    • Fix binner slot allocation failing on an idle GPU (Raspberry Pi 3 bug)
    • Use managed KMS polling to fix UAF on unbind
  • Add support for 7-inch Raspberry Pi Touch Display 2 on Toradex Versin board

RISC-V updates in Linux 7.3

There are also a healthy number of changes to the RISC-V architecture:

• Add initial definitions and discovery for the Smcsrind, Sscsrind, Smcntrpmf, Ssccfg, Smcdeleg, Zicclsm, Ziccamoa, Ziccif, Ziccrse, Za64rs, and Ssqosid RISC-V ISA extensions
• Improve the RISC-V update_mmu_cache_range() implementation by using flush-by-ASID, enabling performance improvements on
  microarchitectures that support related optimizations; and by taking advantage of the Svinval RISC-V ISA extension on microarchitectures
  that support it
• Shrink CFI shadow stack allocation further (to 512MB) to save virtual address space (and physical memory on systems with strict overcommit policies)
• Add initial CPU context switch support for QoS tagging (Ssqosid)
• Change our vector misaligned access speed test code to be synchronous, avoiding a bunch of bugs related to unnecessary asynchronicity
• Enable ARCH_HAS_ACPI_TABLE_UPGRADE for RISC-V to help with ACPI table debugging (similar to ARM64 and x86)
• Extract vDSO section offsets at build time, rather than run time, to avoid the boot time overhead
• Use assembler directives to control the use of instructions from the RISC-V ISA extensions Zacas and Zabha, rather than compiler -march
  flags that could affect the whole kernel (similar to ARM64)
• Add a kselftest for kprobes support for the c.jal instruction on RISC-V
• When UEFI runtime services are available, use them to restart and power off
• Fix ARCH_HAS_ACPI_TABLE_UPGRADE for RISC-V by adding a missing include file to the ACPI code
• Move contiguous DMA memory reservation later in the boot, to ensure that the correct global and per-node pools are allocated when
  CONFIG_CMA_SIZE_PERNUMA is enabled
• Disallow probes on breakpoint handlers to avoid recursing indefinitely into do_trap_break()
• Patch the compat vDSO during runtime alternatives processing, not only the standard vDSO
• Remove some leftover XIP support elements missed by commit 9b3a2be84803 (“riscv: Remove support for XIP kernel”)
• Fix broadcast timer switching in ACPI LPI power states by implementing a RISC-V version of arch_get_idle_state_flags()
• Miscellaneous fixes, including: vDSO makefile simplification; marking our default_power_off() as __noreturn; error path fixes in the PMU SBI perf driver; RISC-V extension capitalization consistency; the use of BIT() macros in one of our include files; and some documentation fixes
• Alibaba T-Head (TH1520)
  • DRM – th1520-dw-hdmi
    • Fix error check on dw_hdmi_probe() return value
    • Fix remove() callback
  • PWM – Use vertical import style
  • Device tree – Set of changes for the remaining IO expansion chips on the TH1520-based LicheePi 4a board.
• Canaan/Kendryte
  • Added K230 RISC-V-based 64-bit AI SoC, based on the earlier K210 chip
  • Initial support for CanMV-K230 and K230-EVB
• Microchip – Hwmon – PolarFire SoC temp/voltage sensor
• Sophgo
  • SG2000 is a mixed Arm/RISC-V chip that was already supported using the Arm core but is now also added for RISC-V along with
    several other variants of the cv18xx series
  • Device tree
    • Initial support for Milk-V Duo S (SG2000) and Duo 256M (SG2002) boards
    • Allow DMA multiplexer setting channel number for DMA controller
•  Spacemit
  • Thermal control –  Add a devm action to clean hardware interrupts, sampling, and control registers on Spacemit K1
  • PHY – Spacemit USB3/PCIe comb PHY driver
  • PCIe controller driver – Add missing MODULE_DEVICE_TABLE() to generate module alias info for OF-based module autoloading
  • Clock driver
    • Add clock for SpacemiT UFS controller
    • Add SpacemiT I2S clock and fixes
  • Device tree changes
    • K3 SoC
      • Enable SD card support
      • Enable QSPI
      • Fix reset delay of ETH PHY
      • Fix PHY ID of ETH
      • Symbolic PDAM ID
      •  Add SPI controller support
      • Add USB and PHY support
      • Add I2S nodes
    • K1 SoC
      • Symbolic PDAM ID
      • Enable cpufreq support
      • Fix max CPU core voltage
    • K3 boards
      • Pico-ITX
        • Enable ALDPS on RTL8211F PHY
        • Disable 125MHz clock on RTL8211F PHY
        • Enable NOR flash
        • Add EEPROM support
      • CoM260 module –  Add EEPROM support
    • K1 boards
      • Milk-V Jupiter – Enable CPU DVFS
      • Initial support for Banana Pi BPI-CM6 and IO board
      • Orange Pi R2S – Enable PCIe
•  StarFive
  • PCIe controller driver
    • Fix resource leaks on error paths in host_init()
    • Fix runtime PM handling and teardown ordering to avoid register access while power or clocks are disabled
    • Check for runtime PM resume failure to avoid register access while power or clocks are disabled
  • SPI – Add support for StarFive JHB100 SFC (SPI Flash Controller)
  • Device tree – Add bindings for StarFive JHB100 PLIC

MIPS changes

As usual, there are just a few MIPS updates:

• Switch GPIO code to use swnodes
• Rework of TXX9 GPIO code
• Enable multi-VPE for econet
• Cleanups and fixes

Full list:

• mips: dts: econet: Describe dual-VPE 34Kc processor
• mips: econet: add multi-vpe capability to EN751221
• MIPS: ptrace: Fix syscall skipping via PTRACE_SYSCALL
• mips: remove dead select
• MIPS: BCM47XX: Convert buttons to software nodes
• ssb: gpio: Add and register software node for GPIO controller
• bcma: gpio: Add and register software node for GPIO controller
• MIPS: ip22-gio: Drop #include of <linux/mod_devicetable.h>
• MIPS: TXX9: Clean up txx9_iocled_init()
• MIPS: TXX9: Convert gpio_txx9 to dynamic GPIO base allocation
• MIPS: TXX9: Drop GPIOLIB_LEGACY select
• MIPS: TXX9: Use GPIO lookup table for iocled LEDs
• MIPS: TXX9: Reduce TXX9_IOCLED_MAXLEDS to 3
• MIPS: TXX9: rbtx4927: Use GPIO lookup table for TXx9 LEDs
• MIPS: TXX9: rbtx4927: Use GPIO lookup table for SIO DTR
• MIPS: TXX9: Remove txx9_7segled_*() forward declarations
• MIPS: TXX9: Remove tx4938_spi_init() and txx9_spi_init()
• MIPS: kernel: proc: Use two seq_putc() calls in show_cpuinfo()

There were 18,765 non-merge commits in total, of which 1,920 had the “Assisted-by” tag, meaning about 10% of commits to the Linux kernel were assisted by AI agents, or at least AI usage was voluntary reported for these commits. The “Assisted-by” tag requirement will go away in the next releases.

You can check out a more detailed Linux 7.3 changelog, generated with comments only, using the command git log v7.2..v7.3-rc7 --stat. Alternatively, you can also read the changelog on Kernelnewbies.

The post Linux 7.3 release – Main Changes, Arm, RISC-V, and MIPS architectures appeared first on CNX Software - Embedded Systems News.

BIOSTAR BIWCL-PAT is a 3.5-inch industrial SBC built around the Intel Core Series 3 “Wildcat Lake” SoC family, ranging from the Core 3 304 to the Core 7 360.

The board supports up to 64GB DDR5-6400 memory, three independent displays, and dual 2.5GbE networking. It also features three M.2 slots for NVMe storage, Wi-Fi/Bluetooth, and 4G/5G cellular connectivity, along with USB 3.2 ports, four serial interfaces, and an 8-bit digital I/O header. The compact 148 x 102mm form factor makes it suitable for industrial automation, edge computing, AIoT, human-machine interfaces (HMI), and POS/kiosk applications.

BIOSTAR BIWCL-PAT specifications:

• Wildcat Lake SoC (one or the other)
  • Intel Core 3 304 penta-core CPU with 1x P core @ 1.5 GHz / 4.3 GHz (Turbo) and 4x LPE cores, 1-core Intel Xe3 Graphics @ 2.3 GHz; TDP: 15W
  • Intel Core 3 305 hexa-core CPU with 2x P cores @ 1.5 GHz / 4.3 GHz (Turbo) and 4x LPE cores, 1-core Intel Xe3 Graphics @ 2.3 GHz; TDP: 15W
  • Intel Core 5 315 hexa-core CPU with 2x P cores @ 1.5 GHz / 4.4 GHz (Turbo) and 4x LPE cores, 2-core Intel Xe3 Graphics @ 2.3 GHz; TDP: 15W
  • Intel Core 5 320 hexa-core CPU with 2x P cores @ 1.5 GHz / 4.6 GHz (Turbo) and 4x LPE cores, 2-core Intel Xe3 Graphics @ 2.5 GHz; TDP: 15W
  • Intel Core 5 330 hexa-core CPU with 2x P cores @ 1.5 GHz / 4.6 GHz (Turbo) and 4x LPE cores, 2-core Intel Xe3 Graphics @ 2.5 GHz; TDP: 15W
  • Intel Core 7 350 hexa-core CPU with 2x P cores @ 1.5 GHz / 4.8 GHz (Turbo) and 4x LPE cores, 2-core Intel Xe3 Graphics @ 2.6 GHz; TDP: 15W
  • Intel Core 7 360 hexa-core CPU with 2x P cores @ 1.5 GHz / 4.8 GHz (Turbo) and 4x LPE cores, 2-core Intel Xe3 Graphics @ 2.6 GHz; TDP: 15W
  • NPU – 15–17 TOPS, except Core 3 305 which has none
• Memory – DDR5-6400 SO-DIMM slot supporting up to 64GB RAM
• Storage – M.2 M-key 2242/2280 socket with PCIe 4.0 x2 for NVMe SSD
• Display
  • HDMI 2.1 up to 3840×2160
  • DisplayPort 1.4 up to 4096×2160
  • LVDS via internal header up to 1920×1200 (with 3.3V/5V/12V selection and LVDS inverter header)
  • Triple independent display support
• Audio
  • Realtek ALC897 w/ ALC105 audio codec
  • Support for 4-ohm 2W amplifier
  • Internal headers for Mic-in, Line-out, Line-in, and Speaker-out
  • Onboard buzzer
• Networking
  • Dual 2.5GbE RJ45 ports via Intel i226-LM and Intel i226-V controllers
  • Optional WiFi and Bluetooth via M.2 E-Key module
  • Optional 4G LTE/5G via M.2 B-Key module and Nano SIM card slot
• USB
  • 2x USB 3.2 Gen 2 Type-A ports (rear)
  • 2x USB 2.0 Type-A ports (rear)
  • 2x USB 2.0 interfaces via internal header
• Serial
  • 2x RS232 via internal box headers
  • 2x RS232/422/485 via internal box headers (with RI/5V/12V voltage selection)
• Expansion
  • M.2 B-Key 3042/3052 socket (PCIe 4.0 x1 + USB 3.2) with SIM slot
  • M.2 M-Key 2242/2280 socket (PCIe 4.0 x2)
  • M.2 E-Key 2230 socket (PCIe 4.0 x1 + USB 2.0)
  • 8-bit DIO header (3.3V/5V selectable)
• Security – Onboard TPM 2.0
• Misc
  • Onboard Power LED for APCI power status
  • RTC battery header
  • IT8786E-I Super I/O controller
  • LCD brightness and volume control header
  • CPU PWM fan connector
  • Clear CMOS and front-panel headers
  • I2C and SMBus headers
  • 5-pin header for Volume UP/DOWN & LCD Brightness UP/DOWN
• Power Supply
  • 9V to 24V DC input via 2-pin Phoenix terminal block or 4-pin box header
  • AT/ATX power select header
• Dimensions – 148 x 102 x 1.6 mm (3.5-inch SBC form factor)
• Temperature– Operating: 0°C to 60°C; Humidity: 5% ~ 90%

The SBC is compatible with Windows 11 IoT Enterprise 24H2 or later and Ubuntu 24.04. BIOSTAR provides a user manual and Windows 11 x64 drivers on its product page, including chipset, Intel Management Engine, graphics, LAN, audio, Serial IO, and NPU drivers. Additional downloads include Intel Dynamic Tuning Technology, Innovation Platform Framework (IPF), Smart Sound Technology (SST), and Platform Monitoring Technology (PMT) drivers.

Previously, we wrote about the Advantech MIO-5356, another 3.5-inch Wildcat Lake SBC with up to 64GB DDR5 memory, triple-display support, and M.2 expansion. We also covered the ASRock Industrial NUC 300 series, which offers similar processors, dual 2.5GbE, and DDR5-6400 memory in a smaller form factor, and AAEON recently introduced the even smaller, credit card-sized UP-WCL industrial SBC. The BIOSTAR BIWCL-PAT differentiates itself with four serial interfaces, LVDS display support, and three dedicated M.2 slots for storage and wireless connectivity.

BIOSTAR has not announced pricing or availability details for the BIWCL-PAT 3.5-inch Wildcat Lake SBC. Additional details are available on the product page (note: a PDF is automatically downloaded if you can on that link…) and in the press release.

The post BIOSTAR BIWCL-PAT – A 3.5-inch industrial Wildcat Lake SBC with dual 2.5GbE, triple display support appeared first on CNX Software - Embedded Systems News.

Axera AX8860P is a new 16-core Cortex-A55 SoC with up to 128 TOPS/256 TOPS (FP4) NPU, 256-bit LPDDR5/5X memory, and PCIe 5.0 connectivity, aiming to power desktop AI solutions competing with NVIDIA DGX Spark at a fraction of the cost.

Dr. Qiu Xiaoxin, founder and CEO of Axera, unveiled the new AX8860/AX8860P processor a few days ago at the China International Industrial Fair, along with upcoming products from Sipeed and Radxa.

Axera AX8860P key features and specifications:

• CPU – 16x Arm Cortex-A55 processor
• AI – Up to 256 TOPS
• VPU – 64-channel decode, 16-channel encode
• Memory – Up to 128GB 256-bit LPDDR5x with 273GB/s bandwidth
• Interconnect – PCIe 5.0 + AXLink (C2C) cascading with 80% efficiency for multi-chip cards and systems
• Power Consumption – About 40 Watts under load
• Package – 78 × 78 mm module
• Process – 6nm

The AX8860P SoC is said to support vLLM, SGLang, PyTorch, ONNX, Hugging Face formats, OpenAI-compatible APIs, and Axera’s toolchain. Some documentation is already available showing how to convert a MobileNetV2 ONNX model into an “axmodel” running on the Axera chip.

In terms of hardware, Radxa will introduce a “Pocket Supercomputer” powered by the AX8860P SoC, but has yet to provide details since it’s still a few months away.

Sipeed also has a few AX8860P PCIe cards on the way (with up to four chips with the MaixBlade), and a MaixStation mini AI computer for which we have more details, and you can even reserve one.

MaixStation specifications:

• SoC – Axera AX8860P described above
• System Memory – 32GB or 128GB LPDDR5x
• Host interface
  • USB 3.0 Type-C port connected to a laptop
  • 10GbE SFP+ cage connected to the local network
• Misc
  • Power button
  • Ethernet/USB-C mode switch
• Power Supply – 65W USB PD via USB Type-C port
• Power Consumption – 40W typical
• Dimensions – 110 x 110 x 30 mm

A single 32GB node can run Qwen3.8‑27B, a single 128GB node supports Qwen3.8‑Flash‑Next 125B, a dual node with 256 GB RAM can handle DeepSeek V4‑Flash 284B, and a quad-node (Maixstation Pro) with 512GB manages DeepSeek V4.1‑Flash 552B.

Sipeed also shared Qwen benchmarks comparing the MaixStation 128GB to the DGX Spark 128GB. The numbers show it’s only slightly faster, but they expect further optimization will widen the gap.

The company explains that “256 TOPS” beats the “1,000 TOPS” advertised by NVIDIA because the latter is sparse FP4, which real models can’t use, while Dense is around 500 on paper, but Spark only puts a small slice of it to work, while MaixStation makes better use of its 256 TOPS.

The bandwidth is the same for both systems at 273 GB/s, but Axera AX8860P is said to implement architecture improvements to “keep more of those 273 GB/s busy”. Finally, MTP (speculative decoding) is turned off in the benchmarks above for side-by-side comparison.

The NVIDIA DGX Spark was first unveiled in March 2025 for $3,000 to $4,000, but since then, price increases have kicked in, and it sells now for $6,949.99. While the MaixStation is not for sale now, the company says the 32GB model will go for under $2000 and the 128GB model for under $4000, let’s make that $3,999… Daisy-chaining more than two systems is also much cheaper on the MaixStation since you only need a cable, but a $4000 200G/400G switch for the NVIDIA systems.

Sipeed will launch it on Kickstarter in a few months and invites users to pay a $30 reservation fee to get 100 off once the crowdfunding campaign starts. You’ll find more technical details and the reservation link on the product page.

The post Axera AX8860P 16-core Cortex-A55 SoC takes on NVIDIA DGX Spark at fraction of the cost appeared first on CNX Software - Embedded Systems News.

M5Stack has released the PaperDIY series, an updated version of its 2020 M5Paper ESP32 development kit, featuring an ESP32-S3 MCU and a 4.7-inch e-paper display with 16 levels of gray.

It comes in two versions: the PaperDIY Complete Kit, which ships factory-assembled in a 3D-printed enclosure, and the PaperDIY Assembly Kit, which is designed to be integrated into custom projects. Both boards also include a microSD card slot for extended storage, a built-in 140mAh battery, and a custom M5PM1 low-power management chip.

M5Stack PaperDIY specifications:

• Core module – Espressif Systems ESP32-S3-WROOM-1-N16R8
  • SoC – ESP32-S3
    • CPU – Dual-core LX7 processor with up to 240MHz
    • Memory – 512KB SRAM, 8MB Octal PSRAM
    • Storage – 384KB ROM
    • Wireless – WiFi 4 802.11b/g/n and Bluetooth 5.0 LE
  • Storage – 16MB flash
  • PCB antenna
• Storage – MicroSD card slot
• Display – 4.7-inch (960 × 540) ED047TC1 e-paper display with 16-level grayscale and an 8-bit parallel interface for direct display control, protected by an acrylic screen pressure plate
• USB – USB Type-C port for power and programming
• Expansion – 2x FPC 0.5mm 12-pin interfaces (exposing I2S and SPI signals)
• Misc
  • 1x Power Button
    • Single click (Power on / Reset
    • Double click (Power off)
    • Long press (Enter download mode)
  • 2x User Buttons
• Power
  • 5V via USB Type-C port
  • 5V power pads
  • Built-in 140mAh, 3.7V lithium battery with SH1.0-2P connector
  • M5PM1 power management IC (monitors battery, USB voltage, controls e-paper boost circuit)
  • Consumption – 40.95mA @ 4.2V, with peaks up to 161mA
• Dimensions and weight
  • Assembly Kit – 115.3 × 62.5 × 5.5 mm; 26.4 grams
  • Complete Kit – 120 × 68.2 × 6.3 mm; 43.1 grams; 3D-printed (PLA) enclosure
• Temperature – 0°C to 40°C

On the software side, M5Stack supports Arduino IDE and provides the M5Unified and M5GFX libraries for hardware and display control, along with a separate M5PM1 library for power management. The company also provides factory firmware source code based on ESP-IDF, which also includes examples for battery monitoring, Wi-Fi scanning, microSD card access, and button operation. M5Stack mentions that it currently supports flashing via M5Stack’s WebBurner, but it does not mention UiFlow graphical programming support yet. Additional resources, including schematics, pin mappings, datasheets, and enclosure design files, are available in the documentation.

As mentioned earlier, the original M5Paper was released back in 2020 with a similar 4.7-inch grayscale e-paper display, touchscreen support, sensors, and a larger battery, but it was built around a generic ESP32. The new PaperDIY drops the touchscreen and sensors but upgrades to an ESP32-S3. We have also seen a similar product from LILYGO, the T5 4.7-inch E-Paper S3 Pro, which uses the same ED047TC1 display and ESP32-S3 but features different expansion options, including an optional 18650 battery holder.

Crucial Screw Warning: The documentation includes a “Screw Specification Note” for the Complete Kit’s M2 mounting holes: Screws must have a thread length of less than 3mm. Using longer screws will press against the e-paper display and permanently damage it.

The M5Stack PaperDIY Assembly Kit is available on AliExpress and the company’s official store for $25. The Complete Kit costs $29.90 on the official store, but is currently out of stock.

The post M5Stack PaperDIY Kits pair an ESP32-S3 MCU with a 4.7-inch e-Paper display appeared first on CNX Software - Embedded Systems News.

Seeed Studio has introduced the XIAO IPS touch display series, a family of six compact development boards built around the XIAO ESP32-S3 Plus or XIAO nRF52840 Plus MCU, with 0.96-inch, 1.14-inch, and 1.47-inch IPS color displays.

All six boards include a 6-axis IMU, a PDM digital microphone, user buttons, and LiPo battery support. The larger 1.47-inch version adds capacitive touch and a microSD card slot, whereas the 1.14-inch model includes a Grove I2C connector for connecting external sensors. The ESP32-S3 versions support Wi-Fi and Bluetooth LE connectivity, while the nRF52840 models focus on lower-power Bluetooth LE applications. The boards are designed for portable dashboards, smart home controllers, wearable devices, sensor monitoring, and other small IoT projects.

XIAO IPS Display Series specifications:

• Wireless board (one or the other)
  • XIAO nRF52840 Plus
    • Wireless MCU –  Nordic nRF52840 Arm Cortex-M4F microcontroller @ up to 64 MHz with 1 MB flash, 256 KB SRAM, Bluetooth 5.0, NFC, 802.15.4 connectivity
    • Storage – 2 MB onboard Flash
    • Wireless Connectivity – Bluetooth 5.4 (BLE), 802.15.4, and NFC
    • Antenna – Ceramic antenna
    • USB – USB 2.0 Type-C for power supply and console
  • XIAO ESP32S3 Plus
    • SoC – Espressif ESP32-S3R8
      • CPU – Dual-core Tensilica LX7 microcontroller up to 240 MHz with vector instructions for AI acceleration
      • Memory – 8MB PSRAM
      • Wireless – WiFi 4 and Bluetooth 5.0 LE + Mesh connectivity
    • Storage – 16MB SPI flash, MicroSD card socket
    • Antennas – IPEX1 antenna for Wi-Fi and Bluetooth
    • USB – USB 2.0 Type-C for power supply and console
• Storage – MicroSD card slot (1.47-inch model only)
• Display
  • 1.47-inch – 172 × 320 resolution (JD9853A driver) and capacitive touch support
  • 1.14-inch – 135 × 240 resolution (ST7789 driver); no touch
  • 0.96-inch – 80 × 160 resolution (ST7789 driver); no touch
  • Interface – SPI
• Audio – Onboard PDM digital microphone
• Sensors – Onboard 6-axis IMU (LSM6DS3) for motion sensing, orientation, and wake-on-motion
• Expansion
  • I2C and I2S interfaces through expansion pads (1.14-inch model only)
  • Additional GPIO access through expansion pads
• Debug interface – JTAG (ESP32-S3) or SWD (nRF52840)
• Misc
  • Physical slide power switch (ON/OFF)
  • 2x push buttons (1.47-inch and 0.96-inch variants only)
  • 3x push buttons (1.14-inch variant only)
  • Exposed user-button pads (1.47-inch and 1.14-inch variants only)
  • Touch interrupt and motion detection support
• Power Supply
  • 5V via USB-C port
  • 3.7V LiPo battery through a 2-pin JST connector
  • Battery voltage monitoring on both controller variants
  • Battery status detection on nRF52840 models
• Power Consumption (Active, not deep sleep)
  • 0.96-inch –  ESP32-S3: 159.76mA; nRF52840: 28.69mA
  • 1.14-inch – ESP32-S3: 160.05mA; nRF52840: 39.07mA
  • 1.47-inch –  ESP32-S3: 130.53mA; nRF52840: 83.32mA
• Dimensions
  • 1.47-inch – 26.4 × 51.4 × 12.6 mm
  • 1.14-inch – 26 × 48 × 10.6 mm
  • 0.96-inch – 18.8 × 43.6 × 10.6 mm

So the main difference between the two variants is wireless connectivity and power consumption. The ESP32-S3 Plus variant features Wi-Fi connectivity, making it suitable for connected dashboards, smart home controllers, and IoT applications. The nRF52840 Plus variant supports BLE connectivity and lower power consumption, making it better suited for battery-powered wearables and portable devices.

On the software side, the boards can be programmed using the Arduino IDE. Seeed Studio provides getting-started guides, display examples, and demonstrations for the microphone, IMU, buttons, battery monitoring, and MicroSD storage. The ESP32-S3 version uses the Seeed_GFX2 graphics library, with examples for drawing graphics, touch input, displaying BMP images from a microSD card, recording audio, and motion-based interactions. Additional documentation, including schematics, KiCad design files, 3D models, and factory firmware, is also available through the Wiki. Finally, a free 3D-printable case designed by community member Gokul is also available.

We have previously written about plenty of small display development boards, including Seeed’s own Round Display for XIAO, the XIAO ePaper DIY Kit based on the same nRF52840 Plus, but it drives an e-ink panels, theres also Waveshare’s ESP32-S3-LCD-1.28 and ESP32-S3-Touch-LCD-1.47 boards, Makerfabs’ ESP32-S3 Round SPI TFT 1.28-inch, and similar RP2350/ESP32-C6 1.28–1.47-inch LCD kits. Compared to all of that, the XIAO IPS Display series is a little different, as it combines a pre-soldered XIAO nRF52840 Plus with a rectangular IPS display in 0.96-, 1.14-, and 1.47-inch sizes, plus onboard IMU and battery support.

The Seeed Studio XIAO IPS display series is available for $11.99 to $20.33 on AliExpress, as well as on the Seeed Studio Store for $15.90 to $19.90.

The post Seeed Studio XIAO IPS display series features ESP32-S3 or nRF52840 SoC with 0.96-inch to 1.47-inch LCD options appeared first on CNX Software - Embedded Systems News.

After a successful campaign for Xerxes Pi back in 2025, Rapid Analysis has returned with a Xerxes Home Lab ecosystem that includes the Xerxes Pi 2026 and Xerxes Nano carrier board. The upgrade includes M.2 NVMe storage support on the bottom of the board, an I2S Qwiic connector for digital audio, and added debug and boot jumper pins for easier headless deployment.

Designed for home labs, edge AI developers, and small businesses, the Xerxes Pi 2026 supports Raspberry Pi CM4/CM5 and other CM modules, Gigabit Ethernet, optional PoE+, and CSI/HDMI expansion. The larger Xerxes Nano supports NVIDIA Jetson Orin Nano and Turing RK1 modules, adding M.2 PCIe, OCuLink, and SPI/UART connectivity. Both boards support active cooling and compact 1U rack installations.

Xerxes Pi 2026

Xerxes Pi 2026 specifications:

• Supported SoMs – Raspberry Pi CM4, CM5, and compatible alternatives via 2x 100-pin high-density mezzanine connectors
• Storage
  • M.2 M-key PCIe Gen 2 x1 socket for NVMe SSDs (underneath)
  • MicroSD slot for compute modules without eMMC
• Networking – Broadcom BCM54210PE Gigabit Ethernet RJ45 port with optional PoE/PoE+
• USB – USB Type-C for power and compute-module boot/recovery operations
• Expansion
  • SH-series Qwiic connector for I2C and I2S peripherals
  • Mezzanine expansion connector breaking out CSI and HDMI signals
• Debugging – Dedicated SH-series serial debug connector for headless setup
• Misc
  • Boot jumper pin (for easy OS flashing/recovery)
  • PoE-to-USB power selection jumper
  • Standard PWM fan connector
  • Power, activity, and network status LEDs
• Supply – 5V via USB Type-C or PoE
• Dimensions – 120 x 40 mm (fits vertically in 1U rack)

Xerxes Nano carrier board

For more demanding AI workloads, the company has also introduced the bigger Xerxes Nano for NVIDIA Jetson Orin Nano Super (up to 16GB memory) and Turing RK1 (up to 32GB RAM) modules. It also includes an OCuLink connector for connecting external GPUs and other PCIe devices.

Xerxes Nano specifications:

• Supported SoMs – NVIDIA Jetson Orin Nano / Turing RK1
• Storage – M.2 PCIe Gen 3 supports up to three PCIe Gen 3 lanes (under the module)
• Networking
  • Gigabit Ethernet RJ45 port with PoE
  • SFP+ cage
• Expansion
  • OCuLink connector for compatible external PCIe hardware
  • SPI and UART interfaces
  • I²C and I²S interfaces for sensors, displays, and audio peripherals
• Misc
  • PWM fan connector
  • USB-to-PoE power jumper
• Power – USB power or optional PoE, selectable through a jumper
• Dimensions – 120 x 60 mm (half Nano-ITX, fits horizontally in 1U or vertically in 1.5U)

The OCuLink interface can be used to connect compatible external GPUs, AI accelerators, or other PCIe devices, although support depends on the connected compute module, drivers, and power requirements.

Software support, open-source hardware

The boards can run operating systems supported by their respective compute modules, including Raspberry Pi OS, Ubuntu Server, and NVIDIA JetPack. You can also run Docker, Home Assistant, Immich, Open WebUI, and other self-hosted applications. Rapid Analysis has also developed the open-source Xerxes Home Lab management tool, currently available for macOS, with Windows and Linux versions planned. The macOS app can discover devices over SSH, monitor system information such as RAM, disk usage, and fan speed, manage Docker containers, and handle SSH keys. The company also provides KiCad hardware design files for the Xerxes Pi 2026  on GitHub, as well as hardware documentation.

The company also offers 3D-printable rack and enclosure designs including a 1U rack that holds one Xerxes Nano and four Xerxes Pi 2026 boards. Other options include single-board enclosures, 6-slot mini racks, and 12-slot 19-inch racks.

Upcoming Xerxes Compute Module X

The company also talks about the Xerxes Compute Module X, an upcoming Raspberry Pi CM5 alternative with a Rockchip RK3566 quad-core Cortex-A55 processor with an integrated 1 TOPS NPU for lightweight edge AI, and 8GB of RAM. The company calls it a “Lite” because it removes onboard Wi-Fi, Bluetooth, and eMMC storage; instead, it uses Ethernet and external storage such as microSD or NVMe through a compatible carrier board.

When I asked Dr. Rob Newport, founder and chief engineer of Rapid Analysis, for photos and additional details, he replied:

I’d like to keep the CM photos and details confidential at this time. The reason is that it will be using a novel design to mount and rework the memory on the CM to keep costs down. We will be open-sourcing the CM, but we don’t want larger companies scooping up this innovation before we complete our Kickstarter and send out rewards.

Kickstarter campaign and price information

The Xerxes Home Lab campaign is currently live on Kickstarter and has already surpassed its $6,996 USD ($10,000 AUD) funding goal. Rewards start at $49 USD for the Gadgeteer package, which includes one Xerxes Pi 2026 carrier board, PoE accessories, heatsinks, and an enclosure. The Ultimate Home Lab package costs $280 USD and includes six Xerxes Pi boards with a mini rack, while the Agentic AI Studio package costs the same and includes one Xerxes Nano, four Xerxes Pi boards, and a rack. The twelve-board Ultimate Enterprise package costs about $559 US. The 8GB Xerxes Compute Module X is offered as an add-on for about $100 US, with estimated shipping in December 2026.

.

The post Xerxes Pi 2026 and Xerxes Nano carrier boards offer NVMe, PoE, and OCuLink for Raspberry Pi and Jetson modules (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

Just a few days ago, we wrote about the ESP-SDR project, which turned an ESP32 into an SDR receiver (and now SDR transmitter), but the data was streamed over UART, which couldn’t keep up with the sample rate. Now, developer h0m3us3r has released w1700k-sdr, also an open-source project that turns a Gemtek W1700K Wi-Fi 7 router into a wideband SDR receiver.

The project uses the router’s built-in MediaTek MT7996 radio to capture raw 14-bit I/Q samples at up to 320 MSps across the 2.4 GHz, 5 GHz, and 6 GHz bands. The captured data is then streamed over a 10 GbE connection using the router’s Airoha AN7581 SoC to a Linux computer, where you can view the live spectrum and waterfall in a browser. To make this work, the project replaces the router’s stock firmware with a custom Linux system and radio firmware, allowing it to capture raw I/Q data instead of processing normal Wi-Fi traffic.

W1700K SDR features and specifications:

• Supported hardware – Gemtek W1700K (HW 1.x) and W1700K2 (HW 2.1) Wi-Fi 7 routers
  • SoC – Airoha AN7581 with a quad-core Arm Cortex-A53 processor
  • Radio – MediaTek MT7996 Wi-Fi 7 chipset connected over PCIe
    • MT7976 RF front end for 2.4 GHz
    • Two MT7977 RF front ends for 5 GHz and 6 GHz
  • Memory – 2GB DRAM, including a 1 GiB DMA ring buffer for captured samples
• SDR reception
  • Bands – 2.4 GHz, 5 GHz, and 6 GHz
  • Sample resolution – 14-bit I/Q samples
  • Sample rates – 40, 80, 160, and 320 MSps across all three bands
  • Additional 640/2 sampling mode on band 2, with aliasing limitations
  • Tuning – Continuous LO tuning within supported synthesizer ranges, with retuning during streaming
  • RF controls – Automatic gain control (AGC), manual LNA/baseband gain, and 20–320 MHz analog filtering
• Streaming – Zero-copy UDP streaming over 10 GbE with lossless sample packing at the highest rate

The project uses lossless sample packing and buffering to handle continuous high-speed I/Q streaming. At 320 MSps, the radio produces around 1.28 GB/s of data when each complex sample is stored in 32 bits, which exceeds the usable bandwidth of a single 10 GbE connection.

To solve this, the firmware packs half of the data chunks into a smaller 28-bit-per-complex-sample format without losing information. The samples are transferred to a 1 GiB ring buffer in the router’s memory and then sent to the host using UDP. With this approach, the developer mentions that the connection reaches around 9.75 Gbit/s. Unlike ESP-WebSDR, which captures short I/Q snapshots and can miss signals between capture windows, W1700K-SDR supports continuous I/Q streaming without these capture gaps.

The project uses custom firmware for the MT7996, including a ROM patch and separate WM and WA firmware. The developer reverse-engineered MediaTek’s radio initialization and calibration procedures by tracing register operations, then implemented the required functions without loading MediaTek’s proprietary firmware. The receiver supports changing frequency, gain, filter settings, and switching bands while streaming. Changing the sample rate currently requires stopping and restarting the stream.

The host software also includes a “flatten” option that corrects the frequency-response roll-off introduced by the radio’s decimation filters. This helps maintain a more even spectrum display near the edges of the captured bandwidth. You can control the receiver from the browser or use command-line tools to change settings and record I/Q data for further processing.

To get started, you need a compatible W1700K or W1700K2 router, a 3.3V USB-to-serial adapter, an Ethernet connection to the router’s LAN3 port for installation, and a Linux computer with a 10 GbE NIC connected to the router’s WAN port. The Gemtek W1700K is also known as the Quantum Fiber W1700K Wi-Fi 7 router. If you are unsure whether you own the right version, hungvu.tech has a teardown with board images.

The repository provides build scripts, installation tools, and instructions for backing up the original NAND flash. After installation, the spectrum viewer can be accessed through a browser at port 8073. Most of the code is released under the MIT license, while certain kernel, bootloader, and driver components are under GPL and BSD licenses.

The post W1700K SDR turns a Wi-Fi 7 router into a 320 MSps software-defined radio receiver appeared first on CNX Software - Embedded Systems News.

HPMicro has introduced the HPM5100 series, a family of 32-bit RISC-V mixed-signal MCUs, starting with the HPM5151 parts. It features a 400 MHz RISC-V processor, 416 KB SRAM, 1 MB flash, four 16-bit ADCs, four operational amplifiers, and four analog comparators.

It also includes a high-speed USB 2.0 OTG interface with an integrated PHY, CAN FD, multiple UART/SPI/I2C interfaces, and 32 PWM channels for motor control. The chip supports hardware floating-point operations and DSP instructions, while the integrated analog peripherals reduce the need for external ADCs and signal-conditioning circuits.

HPMicro HPM5100 specifications:

• MCU core – Single-core 32-bit RISC-V processor @ 400 MHz (up to 480 MHz in high-performance mode with external power supply)
  • Hardware FPU, SIMD, and DSP instruction support
  • 16 KB L1 instruction cache and 16 KB L1 data cache
  • Up to 3.57 CoreMark/MHz
• Memory
  • 416 KB SRAM
  • 256 KB instruction local memory (ILM)
  • 128 KB data local memory (DLM)
  • 32 KB AHB SRAM
• Storage
  • 1 MB SiP flash
  • 128 KB Boot ROM and 4 Kbit OTP memory
  • XPI interface supporting external serial NOR/NAND flash, HyperFlash, HyperRAM, and PSRAM
• Analog peripherals
  • 4x 16-bit SAR ADCs at up to 2 MSPS, configurable to 12-bit resolution at up to 4 MSPS, with support for differential inputs
  • 4x operational amplifiers configurable as programmable gain amplifiers (PGA)
  • 4x analog comparators with internal 12-bit DAC reference generation
• USB – 1x USB 2.0 high-speed OTG interface with integrated PHY
• CAN Bus – 1x CAN controller supporting CAN FD at up to 8 Mbps
• Other interfaces
  • 5x UART, 4x SPI, 3x I2C
  • 2x I2S interfaces (not available on the QFN32 variant)
  • Up to 79x GPIO
• Security – Secure boot, JTAG protection, 128-bit unique ID, and AES-128 CTR encrypted execution from external NOR flash
• Misc
  • 4x 8-channel PWM modules (32 channels total)
  • 3x general-purpose timer modules, each containing four 32-bit counters
  • 3x watchdog timers
  • Trigger routing and synchronization peripherals
  • 2x 32-channel DMA controllers
• Power
  • 3.0V to 3.6V main supply
  • Selectable 1.8V/3.3V I/O banks
  • Integrated LDO regulators and low-power wait, stop, and shutdown modes
• Package
  • HPM5151IPEX – 14 × 14 mm LQFP100
  • HPM5151IN8X – 8 × 8 mm QFN68
  • HPM5151IN6X – 6 × 6 mm QFN48
  • HPM5151IN5X – 5 × 5 mm QFN32
• Temperature range – −40°C to +105°C ambient temperature, with a maximum junction temperature of +125°C under rated operating conditions

Officially, the HPM5100 is a 400 MHz chip, but it actually features a 480 MHz “extreme” mode. To use it, you need to hook up an external power supply, and you can’t let the room get hotter than 70°C. The company also mentions that you shouldn’t run it that fast 24/7. HPMicro recommends sticking to 400 MHz or less for longer life.

On the software side, the HPM5100 is supported by HPMicro’s open-source HPM SDK, which includes low-level drivers, middleware, and software components such as FreeRTOS and TinyUSB. Support for the HPM5100 and HPM5100EVK was added in SDK version 1.13.0, released in September 2026. Developers can use Segger Embedded Studio for RISC-V, while HPMicro also provides development documentation and example applications through its SDK repository and online documentation.

The company also released the HPM5100EVK evaluation board, which includes an HPM5151 MCU, external QSPI NOR flash, an ES8389 audio codec, a CAN transceiver, and an onboard HPM5301 debugger supporting JTAG and serial communication. The board also features two USB Type-C ports for debugging and USB OTG, a motor control connector, dedicated ADC and OPAMP connectors, and a Raspberry Pi-compatible 40-pin expansion header. For power, an onboard jumper lets you select an internal LDO or an external DC-DC supply. More information about the development board is available in the documentation section.

Previously, we wrote about the HPMicro HPM64G0, a 1 GHz RISC-V microcontroller with a 2D graphics engine and dual Gigabit Ethernet, as well as the HPM6800 series, which offers a 600 MHz RISC-V CPU and optional Vivante 2.5D GPU for HMI and display applications. Compared to the previous MCUs, the HPM5100 series features analog sensing and motor control, with four 16-bit ADCs supporting up to 2 MSPS, up to 48 analog input pins, and four 8-channel PWM controllers.

The HPM5100 series is currently in mass production, with the HPM5151IN61 (QFN48), HPM5151IN81 (QFN68), and HPM5151IPE1 (LQFP100) available from Huaqiu Mall for ¥15.60, ¥16.90, and ¥19.50, respectively ($2.33,$2.52, and $2.91). The HPM5100EVKREVB evaluation board is also available for ¥258 (about $38.50). These prices include domestic Chinese taxes and exclude shipping. Check out the product page for details. [CNXSoft note: the hqchip website blocks some user agents, and doesn’t seem to like Linux hosts. I could enter the website using Chrome on Ubuntu with Surface Pro 10 user agent].

The post HPMicro HPM5100 RISC-V MCU integrates four 16-bit ADCs, four op-amps for analog sensing and motor control appeared first on CNX Software - Embedded Systems News.

Orange Pi O1 is a new single board computer (SBC) based on the somewhat older SOPHGO BM1688 octa-core Cortex-A53 SoC with a built-in 16 TOPS AI accelerator/TPU, and support for decoding up to sixteen 1080p30 video streams, or encoding ten.

The board comes with 8GB or 16GB memory, supports M.2 NVMe, eMMC module, SPI flash, SATA III, or microSD card storage, and offers HDMI and MIPI DSI display interfaces, two MIPI CSI camera connectors, dual GbE, WiFi 6E, Bluetooth 5.3, and optional 4G/5G cellular connectivity,  as well as a few USB ports, audio interfaces, and a 26-pin GPIO header for expansion.

Orange Pi O1 specifications:

• SoC – SOPHGO BM1688
  • CPU
    • 8-core Arm Cortex-A53 processor @ 1.65 GHz
    • RISC-V C906 co-processor @ 1.2 GHz
  • TPU – SOPHGO Neural Network Acceleration Engine (TPU), delivering 16 TOPS @ INT8
  • VPU
    • Video decoding – H.264 & H.265 up to 16 channels at 1080p30
    • Video encoding – H.264 & H.265 up to 10 channels at 1080p30
    • Image Processing (JPEG)
      • Decoding: 480 frames/sec @ 1080p
      • Encoding: 480 frames/sec @ 1080p
• Memory – 8GB or 16GB LPDDR4/LPDDR4X
• Storage
  • Optional eMMC module with 16GB, 32GB, 64GB, or 128GB capacity
  • 128 or 256 Mbit SPI flash (eMMC and SPI flash are mutually exclusive)
  • M.2 M-Key (PCIe 3.0) socket for  NVMe SSD
  • SATA III connector
  • MicroSD card slot
• Display
  • HDMI 2.0 up to 4Kp6
  • 4-lane MIPI DSI interface ( FPC connector)
• Camera – 2x 4-lane MIPI CSI camera interfaces
• Audio
  • 3.5mm headphone jack (audio input/output)
  • 2-pin speaker output connector
  • Built-in electret microphone
• Networking
  • 2x Gigabit Ethernet ports (YT8531C PHYs)
  • Wi-Fi 6E and Bluetooth 5.3LE
  • Optional 4G/5G cellular module via M.2 B-Key socket and SIM card slot
• USB
  • 2× USB 3.0 host ports
  • 2x USB 2.0 host ports
• Expansion – 26-pin expansion header with GPIO, UART, I2C, SPI, PWM…
• Misc
  •  Power and Reset buttons
  • 4-pin DC 12V fan connector
  • 2-pin RTC backup battery connector
• Power Supply – Up to 20V/3.25A via USB Type-C port
• Dimensions – 90 x 70 mm
• Weight – 76 grams

Orange Pi loves to launch hardware before software is ready, and that’s no different for the O1. The product mentions that an Ubuntu image is available, but the Google Drive for it is currently empty, and the documentation is non-existent for now. The company also highlights support for local/private deployment of large Transformer-based models such as Gemma-2B, Qwen1.5-1.8B, Llama2-7B, ChatGLM3-6B, and similar LLMs.

We first noted the BM1688 briefly in 2024 when covering Firefly’s CSB1-N10 series AI cluster servers, but we haven’t noticed another BM1688 platform since then. The BM1688 appears to be in the same family as the SG2300x SoC we tested in the Radxa Airbox Fogwise (again in 2024), just not quite as powerful. It’s unclear to me that Orange Pi would release a BM1688 SBC in 2026, but maybe I’m missing something…

The Orange Pi O1 is priced at $195 (8GB RAM) and $325 (16GB RAM), to which you need to add a $7.9 active cooler. You’ll find both models on AliExpress. A few more details, and eventually the user manual, Ubuntu image, and the Linux source code, can be found on the product page.

The post Orange Pi O1 SBC features 16 TOPS SOPHGO BM1688 AI SoC, supports 4G/5G cellular connectivity appeared first on CNX Software - Embedded Systems News.

Bitmerse Romu is a tiny (15.5 x 12 mm), almost minimal Raspberry Pi RP2354A USB board with a USB 2.0 interface on one side of the PCB, and the dual Cortex-M33/RISC-V MCU on the other side, connected to a few debugging/programming pads, buttons, and an RGB LED.

It doesn’t come with any GPIO, and it’s mostly designed for USB applications such as a security key for passwordless login, a USB MIDI controller, a macro pad, or a custom HID device. You can plug it directly into a spare USB Type-A port of your computer/host without any cables, breadboard, or adapters required.

Romu specifications:

• SoC – Raspberry Pi RP2354A
  • CPU
    • Dual-core Arm Cortex-M33 @ 150 MHz with Arm Trustzone, Secure boot
    • Dual-core 32-bit RISC-V Hazard3 @ 150 MHz (3-stage in-order pipeline; RV32IMAC with Zba, Zbb, Zbs, Zbkb, Zcb, Zcmp, and Zicsr extensions)
    • Up to two cores can be used at a time, either 2x Arm, 2x RISC-V, or 1x Arm + 1x RISC-V
  • Memory – 520 KB on-chip SRAM
  • Storage – 2MB of stacked (in-package) flash memory
• USB – Native USB 2.0 Full-Speed interface (on PCB)
• Programming/Debugging
  • USB DFU or SWD test points
  • UART TX/RX test points
• Misc
  • Capacitive touch button
  • Boot button for firmware flashing
  • RGB LED
  • 3D-printed protective shim
• Power Supply – 5V from USB, regulated to 3.3V onboard
• Dimensions – 15.5 x 12 x 1.65 mm

It comes preloaded with MicroPython, and you can plug straight into any USB Type-A port to start coding in the Thonny IDE (or another IDE). Advanced users can also get an optional programming & debugging jig that connects over USB-C and provides access to a standard 20-pin JTAG connector for SWD debugging, and exposes UART RX/TX pins.

The board is fully open-source hardware, and Bitmerse will release the schematics and PCB design files, bill of materials (BoM), and 3D enclosure STL files on GitHub. Some example MicroPython scripts are already available there, including a dino game, a FIDO2 security key implementation, and a website launcher demo.

We’ve seen many of these “omu” boards over the years, starting in 2018 with the Tomu (Silicon Labs EFM32) and Fomu (Lattice ICE40UP5K FPGA), and followed up with the Somu (STM32L4), Qomu (QuickLogic EOS S3 MCU FPGA), and Comu (CH32V203 RISC-V MCU). Interestingly, they were not made by a single company, but several companies came up with their own “omu” board for whatever reason. Romu is the first “omu” board from Bitmerse, but there’s work on open-source hardware projects such as WiSer serial-to-wireless kit.

Bitmerse has launched the Romu RP2354A board on Crowd Supply with a $5,000 funding target. Rewards are pricier than I would have expected, but it might make sense as a niche product. The board starts at $40, and the optional programming jig adds $65. Bundles are available to somewhat lower the price per unit. Shipping further adds $8 to the US and $18 to the rest of the world. Shipping is expected to start by the end of April 2027.

The post Romu – A Tiny Raspberry Pi RP2354A board for USB application development (Crowdfunding) appeared first on CNX Software - Embedded Systems News.

VIEWE has introduced the PLCM-UES31S043H800V480C-U, a 4.3-inch smart display module that appears to be the first third-party development board powered by the new ESP32-S31 MCU.

The board includes a 4.3-inch display with 2.4 GHz Wi-Fi 6, Bluetooth 5.4 LE and Classic, and IEEE 802.15.4 for Zigbee and Thread. It also features USB-C and USB-A ports, a microSD card slot, RS485 and CAN interfaces, and a 20-pin GPIO expansion header with UART and I2C. The board also includes an ES8389 audio codec, two NS4150B amplifiers, two onboard microphones, four ADC buttons, a WS2812B RGB LED, and BOOT/RESET buttons. It operates on a 5V power supply and supports an operating temperature range of -20°C to +70°C.

VIEWE PLCM-UES31S043H800V480C-U Specifications:

• Wireless module – ESP32-S31-WROOM-3-N16R16V
  • SoC – ESP32-S31NRV16
    • CPU
      • RISC-V HP (High-performance) RV32IMAFCP CPU @ 320 MHz with FPU, SIMD, etc.
      • RISC-V LP (Low-power) MCU core
    • Memory – 512 KB SRAM, 32 KB low-power SRAM, 16MB PSRAM
    • Wireless –
      • 2.4 GHz WiFi 6 (802.11ax)
      • Bluetooth 5.4 with support for LE Audio, direction finding, Bluetooth Mesh 1.1, and Classic (BR/EDR)
      • 802.15.4 radio for Zigbee, Thread, and Matter
  • Storage – 16 MB flash
  • Antenna – PCB antenna
  • Dimensions – 30 x 22 x 3.5mm
• Storage – MicroSD card slot supporting 4-bit SDMMC / SDIO 3.0
• Display
  • 4.3-inch IPS TFT LCD with 800 × 480 resolution
  • RGB Vertical Stripe pixel arrangement, 16.2M colors, 400 cd/m² brightness
  • Connected over a 40-pin 24-bit RGB interface via ST7262E43-G4 driver IC.
  • Capacitive touch panel (CTP) with GT911 IC
  • Active display area of 95.04 mm × 53.86 mm
• Audio
  • Everest ES8389 stereo codec
  • 2x NS4150B Class-D speaker amplifiers
  • 2x on-board analog microphones (MIC1, MIC2)
  • L/R speaker connectors (2-pin 1.25 mm)
• USB
  • USB Type-C port for 5V power, programming, and serial debug via CH340C (UART0)
  • USB 2.0 Type-A High-Speed port connected to the OTG PHY
• Expansion
  • 6-pin 5.0mm terminal block with RS485 and CAN
  • 2×10-pin 2.54 mm external GPIO header for UART0/1, I2C, RS485/CAN GPIOs, 3.3V/5V/GND
  • 4-pin 2.5 mm UART helper connector
• Misc
  • 4x ADC keys (SW3–SW6), plus dedicated RESET and BOOT buttons
  • WS2812B RGB LED
  • 3 kHz buzzer
• Power Supply – 5.0V typical via USB Type-C, or DCIN pins
• Power Consumption – 320 mA typical / 500 mA maximum (max backlight); 100 mA typical (backlight off).
• Dimensions – 105 × 67 mm
• Temperature – Operating: −20 ~ 70°C (LCD limit); Storage: −30 ~ 80°C

Since the ESP32-S31 is still a preview device, you can program it only with the ESP-IDF framework. You also need to use the ESP-IDF master branch with the –preview flag. The company provides examples, including LCD, touch panel, ES8389 audio playback, SD card access, RS485/CAN interfaces, and a complete LVGL factory UI on GitHub to help users get started. More hardware details and display timing information are available in the Wiki. Linux support might eventually be a possibility, since Espressif released an ESP32-S31 Linux BSP preview last August.

VIEWE 4.3-inch ESP32-S31 display development board is sold for $38.99 on the VIEWE store, and we’ve been told it should become available on the company’s AliExpress Store in the next few days.

The post VIEWE launches 4.3-inch ESP32-S31 display development board appeared first on CNX Software - Embedded Systems News.

Microchip Technology has introduced the MTCH3380P and MTCH3240P, two new capacitive touch controllers that handle touch detection internally, making it easier to build display-free HMI devices with touch-sensitive buttons, sliders, and wheels.

The main difference between the two controllers is the number of supported touch sensors. The MTCH3380P supports up to 38 self-capacitance or 64 mutual-capacitance keys, while the smaller MTCH3240P supports up to 24 self-capacitance or 32 mutual-capacitance keys. Both controllers use parallel sensing with 12 on-chip ADCs to improve touch response times and noise immunity. They also support touch detection through thick cover materials and when users are wearing gloves, making them suitable for home appliances, industrial control panels, EV chargers, and outdoor kiosks. Both also support IEC/UL 60730 Class B functional safety requirements.

Microchip MTCH3380P and MTCH3240P specifications:

• Controller engine – Microchip patented charge-transfer acquisition engine with pipelined analog sensing and digital processing.
• Touch sensing
  • Mutual capacitance – Up to 32 keys (MTCH3240P) or up to 64 keys (MTCH3380P)
  • Self capacitance – Up to 24 keys (MTCH3240P) or up to 38 keys (MTCH3380P)
  • Sliders/Wheels – Up to 6 groups (MTCH3240P) or up to 8 groups (MTCH3380P)
  • Sensing Schemes – Single-ended Mutual, P2P Mutual, Single-ended Self, and P2P Self capacitance
  • Panel Materials – Glass up to 6 mm, Plastic up to 3 mm
• Noise immunity
  • Up to 240 VPP between 1 Hz and 1 kHz (no touches)
  • Noise suppression for electrical and power-line interference
  • IEC 61000-4-6 Class A conducted noise immunity at 10 Vrms
• Safety and Security
  • Functional safety – UL/IEC 60730 Class B ready with power-on and periodic self-diagnostics.
  • Encryption – AES 128 encryption for configuration parameters and object protocol messages.
  • Authentication – Firmware authentication mechanism to detect tampering.
• Host interface
  • I2C client interface supporting Standard (100 kHz), Fast (400 kHz), and Fast-mode Plus (1 MHz)
  • Dedicated CHG interrupt line
• Misc
  • Adjacent Key Suppression (AKS) technology
  • 64-byte non-checksummed custom data store
  • PWM output for display backlight, audible buzzer, or haptic feedback (MTCH3380P)
• Power
  • Digital supply (VDD) – 3.3V nominal
  • I/O voltage (VDDIO) – 1.8V to 3.3V nominal
  •  Additional 3.3V analog supply and internal voltage doubler for a 6.6V sensor drive (MTCH3380P only)
  • Configurable Active/Idle modes with automatic wake-up on touch detection
• Package
  • MTCH3380P – 56-pin XQFN (6 × 6 × 0.4 mm)
  • MTCH3240P – 38-pin XQFN (4 × 4 × 0.35 mm)
  • Both feature a 0.35 mm pin pitch
• Operating temperature – -40°C to +105°C

The touch controllers can be configured, tuned, and tested through the mTouch Studio utility, and Microchip also provides host drivers for Linux and Zephyr. Apart from that, software information is limited, which may not be that surprising since those are not microcontrollers…

The company also introduced the EV-MTCH3380P-DEV-PCB and EV-MTCH3240P-DEV-PCB evaluation kit control boards. Both boards support mutual-capacitance or self-capacitance sensing.

We previously wrote about Microchip’s maXTouch M1 automotive touchscreen controllers designed for curved displays and physical knob-on-display technology. We have also seen more flexible, low-power capacitive touch solutions like the Renesas RA4L1 ultra-low-power Cortex-M33 MCU family.

Both the MTCH3380P and MTCH3240P are “available” through Microchip Direct for $1.1 and $0.84 for order of 5000 pieces in tray and tape-and-reel packaging. The tray versions (MTCH3380P-MAU001 and MTCH3240P-MAU001) have a minimum order quantity of 245 units, while the tape-and-reel versions (MTCH3380P-MAUR001 and MTCH3240P-MAUR001) must be ordered in multiples of 1,000 and 6,000 units, respectively. Additional information is available on the respective product pages and in the press release.

Thanks to TLS for the tip.

The post Microchip MTCH3380P and MTCH3240P capacitive touch controllers support up to 64 buttons with parallel sensing appeared first on CNX Software - Embedded Systems News.

Unveiled at Network X 2026, the Qualcomm Dragonwing FWA Gen 6 Elite Platform combines 5G Advanced (3GPP Rel-19), NR-NTN (New Radio – Non-Terrestrial Networks) satellite communication, and tri-band Wi-Fi 8 with a 5×5 radio configuration.

The new Fixed Wireless Access (FWA) solution features a Qualcomm X105 Modem-RF with 5G NR over satellite support, quad-frequency GNSS (L1, L2, L5, and L6), Qualcomm Turbo DSDA for 60% higher throughput compared to Gen 5, a quad-core CPU, an integrated 10 TOPS NPU, LPDDR4x and LPDDR5/5x, and quantum security (CNSA 2.0 compliant). It offers 500 MHz of aggregated bandwidth, 6- and 8-layer MIMO, peak download and upload speeds of 14.8 Gbps and 4.2 Gbps, respectively.

Dragonwing FWA Gen 6 Elite Platform (QMX105) specifications:

• CPU – Quad-core CPU
• AI accelerator – Up to 10 TOPS, 5th-generation Qualcomm 5G AI processor handling
  • OTT traffic detection for apps (chatbots, social media) 
  • 5G gaming traffic prioritization
  • Smooth 5G-to-Wi-Fi transitions 
  • 8Rx/6Rx/4Rx antenna management for mobile/FWA
  • Advanced sensing software that can predict and adapt to real RF conditions such as cell congestion detection and mitigation and mobility conditions
• Memory – LPDDR4x and LPDDR5/5x
• Cellular Modem-RF
  • Modem – Qualcomm X105 5G Modem-RF System
  • Peak Download Speed – 14.8 Gbps (FR1 + FR2), 13.1 Gbps (FR1 with 500MHz BD and 1024 QAM)
  • Peak Upload Speed – 4.2 Gbps (4Tx UL CA)
  • Cellular Technology
    • 3GPP Rel-19
    • FR1 + FR2 DC
    • Sub-6 carrier aggregation (FDD-FDD), sub-6 carrier aggregation (FDD-TDD)
    • 1024-QAM sub-6
    • Dynamic Spectrum Sharing (DSS)
    • LTE, WCDMA, LAA, GSM/EDGE, CBRS, mmWave, 5G SA (standalone), 5G NSA (non-standalone)
  • Satellite communications integrated with NR-NTN, NB-NTN, and RFFE modules and filters
• Wi-Fi
  • Generation – Wi-Fi 8, Wi-Fi 7, Wi-Fi 6E, Wi-Fi 6, Wi-Fi 5, Wi-Fi 4
  • Standards – 802.11bn, 802.11be, 802.11ax, 802.11ac, 802.11n, 802.11g, 802.11b, 802.11a
  • Spectral Bands – 2.4 GHz, 5 GHz, 6 GHz (tri-band) with 5×5 radio architecture
  • Spatial Streams – Up to 12; 8L MIMO on DL and 4L MIMO on UL; up to 4Tx on UL
  • Modulation – Up to 4K QAM
  • Encryption – AES-CCMP, AES-GCMP
• Ethernet – 2x 10GbE
• Process – 6 nm node 

Qualcomm lists support for OpenWrt and prplOS operating systems, the RDK-B SDK, and the Qualcomm AI Stack for artificial intelligence workloads.

Details are rather light at this stage. The company also introduced other (wireless) network infrastructure solutions at Network X 2026, but none stood out more than the FWA6. A little bit more information about the 6th-gen DragonWing FWA platform is available on the product page and in the press release.

The post Dragonwing FWA Gen 6 Elite Platform combines 5G Advanced, NR-NTN, and tri-band Wi-Fi 8 appeared first on CNX Software - Embedded Systems News.

adapter through a case

We received two unusual little devices for review this month: the PlugMate and the PlugClaw, both built by TrustKernel, a security company better known for the trusted-execution tech inside other people’s phones than for consumer gadgets. Each is a complete, self-contained computer the size of a USB stick (its own processor, memory, and encrypted storage) that runs a private, hardened Android system called PlugOS. You plug one into your phone, tablet, or computer, and that host lends it a screen, a keyboard, and a network connection. Nothing of yours crosses over.

The easiest way to picture it: it’s a little computer (or more accurately, the hardware of a small phone) that borrows your screen, mouse, keyboard, and other peripherals to interact with you. IT folks will be tempted to compare it to a virtual machine, and the analogy isn’t bad, except that nothing here is virtual. It’s real, separate silicon; it just shares your host’s display and plumbing the way a VM shares resources. Unplug it, and TrustKernel says it leaves nothing behind.

The PlugMate (US$199 on sale, US$299 RRP) is pitched as the serious one: a hardware vault for your private apps, data and identity. The PlugClaw (US$149 on sale, from US$239 RRP) is the same basic platform with an AI agent bolted on top (OpenClaw running in an Ubuntu VM alongside PlugOS) that can operate apps, answer messages, and run workflows for you. Same shell, same size, different pitch, lower price. Which immediately raises the question: why does the AI one cost less? We’ll come back to that.

First things first: update before you do anything

Here’s the bit that matters if you’ve just bought one: run the system update before you set anything up. On our unit, the latest OTA went through cleanly and left our test files alone, but updates replace the system image, TrustKernel warns beta builds “may cause data loss,” and any reset on this platform rotates the encryption keys, meaning gone really means gone. So treat it like a fresh phone, not a folder: plug it in, skip the setup tour, go to PlugOS Settings → System Update, let it do its thing, and only then start installing your life on it. And if you’re updating a device you’re already using, back up first: PlugOS has a “Local Backup” that writes to the host device, though it insists on a storage location being set in the host’s PlugOS client first (the dialog literally points at the PC client’s settings), and forum users are still asking for a fuller backup path. Treat anything on the stick as precious until you’ve tested a restore.

What’s in the box

Both arrived in the same generic blue “PlugOS — USB-C Extended Computing Device” box with a “Made for iPhone | iPad” badge and a sticker telling the two units apart: ours are a red PlugClaw (6 GB/128 GB) and a black PlugMate (128 GB). Interestingly, both labels say Model PS01, which tallies with TrustKernel’s claim that PlugClaw is built on PlugMate’s foundation.

In the PlugClaw box: the stick itself, a U-shaped USB-C adapter that folds the device flat against the back of your phone, two mesh protective sleeves, a setup guide, and the product-key/warranty cards. The adapter earns a mention: it’s a sturdy little thing, and on our phone it connected happily through a chunky case, solving the classic “fat plug won’t fit the case’s port cutout” problem before it starts. PlugMate’s pack is much the same, swapping the mesh sleeves for a ribbed case, which exists, as it turns out, for a reason (it’s a heatsink, more on that below). The box also carries a warning worth repeating: always shut the unit down from inside PlugOS before yanking it out. It’s an eMMC-based computer, not a flash drive, and TrustKernel warns that hot-unplugging can corrupt the file system.

The hardware

Fifty millimetres long, nineteen wide, eight thick, and a claimed 14.3 grams: our kitchen scale says 15–16 grams bare, for both sticks, which is within sneezing distance of spec. Either way, these things are smaller than most USB drives you’ve got lying around.

Inside, per TrustKernel’s spec sheet, sits a MediaTek Helio G80 octa-core chip (two Cortex-A75, six A55), 4 GB of RAM (6 GB in our PlugClaw; a 6 GB/256 GB PlugMate exists on paper but is permanently out of stock) and 128 GB of fully encrypted eMMC storage. One USB-C male connector does everything: power, data, authentication, display out. There’s no battery, no modem, no SIM slot; more on what that means later.

Worth a footnote here: TrustKernel’s own support forum has the processor listed as a Helio G70, while the spec pages and an independent teardown (which found an MT6769V SoC on the board) point to the G80. We asked our unit directly: CPU-Z inside PlugOS reports a Helio G80 at 2 GHz, so the spec sheet wins that argument. It’s running Android 14 underneath; how fresh that really is gets an entry of its own below. It’s the sort of detail that matters less for performance than for the principle: on a security device, you’d like the paperwork to be exact, and a forum typo is a small thing, but exact is exact.

Setup and first boot

Setup is the same on both: install the PlugOS client on your host (Android 8+, iOS 16.6+ on iPhone 15 and up, macOS 14.5+ on Apple Silicon, Windows 8+ on x86, or the new Linux beta client on Ubuntu/Debian/Fedora/RHEL x64), plug the stick in, scan the product-key QR code, set a PIN, wait.

And a quietly lovely detail: the Android client asks for no account, no registration, no e-mail: plug, pair, go. In 2026 an app that doesn’t want to know who you are is practically a manifesto, and it happens to match the “zero data collection” pitch nicely. Boot time, plug-in to usable desktop: 24 seconds on the PlugMate (Android host), 20 seconds on the PlugClaw (Windows, firmware 1.1.8). First boot after a reset is slower (58 seconds on ours, generating keys and running first-run init), but warm boots are consistently quick. For context, PCMag reported multi-minute boots on their unit; either they were unlucky, or a year of firmware updates has genuinely paid off. Our money’s on the latter.

Cross-platform reality check

TrustKernel claims plug-and-play across Android, iOS, Windows, macOS and Linux. We threw our menagerie at it: a Windows machine, a Fedora box, an Android phone and a MacBook Pro:

Host	Client	Result
Android phone	PlugOS Android app	Works — full session on the handset, U-adapter fits through a case
Windows PC	PlugOS PC client	Works — screen capture of the session is blocked until you switch off PlugOS's capture protection (a user toggle, not a hard lock)
Fedora Linux	PlugOS client 1.5.7 Beta (Flatpak)	Works — official installer handles Fedora; full Settings UI present
Intel MacBook Pro (2017)	macOS client	Refused — the client requires Apple Silicon + macOS 14.5; our Intel machine wouldn't accept the app

Every host we could try worked, and the failures were predictable ones. The interesting discovery was on Windows: the client ran fine, but screen recordings came back with the PlugOS window missing entirely; the capture protection is host-side, and it turns out it’s just a toggle in PlugOS settings. Disable it and the session records normally. The Fedora beta client earned its keep too: the official installer handles dnf properly, drops a sane udev rule, and the result behaves like the Windows client.

Worth knowing: TrustKernel says iPhone 15 and up only; Lightning won’t work even through adapters, and it’s reported running happily on iPads and iPhones (we had no iOS device to verify with). macOS support is Apple-Silicon-only per the download page, and our Intel MacBook confirmed it the hard way: the client won’t install at all. We don’t have an iPhone or an Apple-Silicon Mac in the lab to check the iOS/macOS side (readers feeling generous: the review bench happily accepts donations, ideally with ports). Windows on Arm isn’t supported yet. And the Linux client is still beta: early builds lacked file transfer, backup, and clipboard sync, though one forum user reports it running happily under Qubes OS, which feels very on-brand for this product.

Living with PlugOS

Living with it is unremarkable in the good way: it’s a phone. Apps open at phone speed, scrolling is phone scrolling, and YouTube ran at 1080p without a stutter in our testing. Nothing about the daily experience reminds you that the computer doing the work is plugged into a USB-C port, which is quietly the whole point.

Boot it, and you get a phone-shaped Android 14 desktop, portrait on phones, a portrait window on PCs. The usual suite is here: settings, file manager, browser (a privacy-focused one called Veil), the open-source HeliBoard keyboard (nice choice), a Play Store-capable app store, and PlugOS Settings, which is where the product earns its keep:

Virtual Peripherals is the party trick. Apps inside PlugOS never see real hardware; they get a virtual view you control. GPS can be set to host mode or anywhere you drop a pin; we parked ours in Alaska, and Google Maps dutifully agreed we were freezing to death in Anchorage. Worth noting: it only spoofs the GPS coordinates. Your traffic still rides the host’s network, so IP-based geolocation still outs you, as TrustKernel’s support staff confirmed on their forum, suggesting a VPN inside PlugOS for the full effect.

The same panel lets you simulate a SIM card (IMSI, IMEI, ICCID, carrier, even a phone number) plus Wi-Fi MAC/BSSID, camera (black frames), microphone (silence), and sensors. To apps inside PlugOS, you look like a phone that isn’t yours. Let’s be precise about what that is and isn’t: there’s no modem and no eSIM here. You can’t make calls or get a number that receives texts; you’re faking the identifiers apps read, not real telephony.

The firewall is a proper system-level one: per-app allow/block lists in blacklist mode, plus Net Logs and a DNS/NTP toggle. It’s off by default, which matters if you bought this for the privacy defaults. We verified the controls; one forum user did report flaky checkbox behaviour, so watch it rather than trust it blindly.

GMS is optional. Google Mobile Services are off by default; flick the switch, and the Play Store appears for whatever needs it. On Play Integrity, our unit passes BASIC and DEVICE but fails STRONG; a forum screenshot showed STRONG passing on a different firmware, so your mileage may vary by release, and STRONG-requiring apps (Krungthai NEXT again) are the ones that refuse.

Duress PIN and self-destruct. PlugOS Security gives you two related weapons: a duress code (a second unlock you can enter under pressure that silently wipes the device) and a configurable brute-force limit: pick 5 to 15 wrong attempts and the unit auto-resets on the nth failure. TrustKernel’s docs say the wipe rotates the encryption key and can present a fake “hardware malfunction” screen as a decoy. We verified both features exist in the settings (the reset-on-Nth-attempt menu is right there, options 5 through 15); deliberately setting our own unit off is a test we’re saving for last, since a device full of your data is a one-way trip.

Containment, verified. This is the core promise: apps inside PlugOS can’t see the host’s data, IMEI, or identifiers, and the host can’t see inside PlugOS. TrustKernel’s whitepaper claims all peripherals are virtualised, and the USB link is an encrypted, access-controlled channel. Our checks back the mechanism: About phone inside PlugOS reports a PlugOS-assigned IMEI rather than the host’s, the device identifies itself as PlugOS hardware rather than the phone it is plugged into, and on the network side apps see a private 10.x NAT address, which PlugOS’s own settings call an “internal network address, inaccessible from the external network”. Inbound, nothing can knock. Outbound is where the catch is: traffic still exits via the host’s real public IP, so IP-based geolocation and the host network’s visibility of your traffic are unaffected; “can’t be knocked on” is not “can’t be watched”. Independent testing went further than we dared: dumping the stick’s partitions with mtkclient and reflashing a modified boot image: the userdata partition wouldn’t read (encrypted), the modified boot wouldn’t boot, and the unit then disabled itself entirely, product key invalidated, dead. Someone bricked their own device trying to crack it; call that an endorsement, depending on your sense of humour.

Performance — a 2020 phone, and that’s fine

Let’s not pretend: a Helio G80 is a midrange chip from 2020, and no amount of aluminium makes it a Snapdragon. Geekbench 7 inside PlugOS on our PlugMate returned 306 single-core, 782 multi-core. In practice, everyday apps open promptly, scrolling is smooth enough, and YouTube played 1080p flawlessly; the short version is “phone apps, phone speed”. It browses, messages, streams, and runs bank apps fine; it is not a device you buy to feel speed.

The catches — because there are a few

• It runs hot. The aluminium body is the heatsink, and that ribbed sleeve we assumed was a protective case? It’s a passive heatsink, and after a solid session its fins are genuinely hot to the touch. PCMag measured 122.9 °F (about 50.5 °C) on their unit during downloads and called it uncomfortable to hold; ours gets properly warm, so keep it in the sleeve under load and away from curious fingers; the warning TrustKernel gives is for a reason.
• Screenshot lockdown — aimed at the host, not you. Host devices can’t capture the PlugOS session: on Android the recording comes out black, and on Windows the client window simply doesn’t appear in screen recordings at all (Windows lets apps exclude themselves from capture, the same trick banking and DRM apps use). Inside PlugOS itself, its own screenshot and screen-record tools work fine; your private system keeps its own camera, so to speak. For our purposes, capture = PlugOS’s own tools or a second camera pointed at the screen.
• It’s a phone UI, always. On a PC you get a portrait window, and your mouse imitates a finger.
• Localisation wobbles. The system timezone may snap back to China time after setup. Trivial until your banking app’s 2FA clock disagrees.
• Old bones. PlugOS is Android 14 on a 4.14 kernel. Our unit shows the August 5, 2025 security patch even after the latest update. TrustKernel says the base is deeply customised and updates will keep coming on 14. Your mileage may vary.
• The camera is virtualised too. Apps inside PlugOS receive compressed images from the host camera; usable, but visibly softer than native (TrustKernel admits this on their forum).
• Power sensitivity. Some Android phones can’t supply the wattage; there’s a Compatibility Mode that throttles the stick to compensate. Our phone ran both sticks at full tilt without needing it.

“Can’t my phone do this already?” — the alternatives

Fair question, because software has been quietly copying the idea. Xiaomi’s Second Space (there since MIUI, still in HyperOS) gives you a whole second profile (own apps, accounts, files, separate unlock finger), free, on the phone you already own. Google’s Private Space in Android 15 does a similar trick at AOSP level, complete with the option to hide that the space exists at all. Samsung’s Secure Folder has offered Knox-backed separation for years, and even humble app cloners give you a second WhatsApp.

Here’s the difference, and it runs deeper than a feature list: all of those are second drawers in the same house. They run on the same kernel, the same modem, the same SIM, the same hardware identifiers, so an app inside Xiaomi’s Second Space still sees your real IMEI, your real sensors, your real everything. They can’t fake a location, they can’t firewall an app off the network, they can’t be duress-wiped, and they most definitely can’t be unplugged and carried to your mate’s iPad. PlugOS is more like a separate house that happens to share your driveway; different land title, different letterbox.

Same logic for the other usual suggestions. GrapheneOS hardens your phone beautifully, but it is your phone. Tails on a USB stick borrows the whole computer to boot: great for PCs, useless on a phone, and you’re trusting the host’s hardware and bootloader. A second phone costs more, needs charging, and can’t lend an iPhone an Android. No other hardware product currently does what these two do; the last similar category was Intel’s Compute Stick, and it was neither secure nor long-term for this world. If all you need is hiding Tinder from your flatmate, your Xiaomi or Pixel already does it for free. PlugMate and PlugClaw earn their price when the requirement is physical separation that travels.

PlugClaw: the one with the AI inside

Same stick, same PlugOS, plus an Ubuntu VM running OpenClaw, TrustKernel’s agent framework, with an “Android-Use” layer that lets the agent operate apps the way a person would: tapping, typing, navigating. The marketing line is “ChatGPT talks, PlugClaw acts.”

First, a wrinkle that deserves its own paragraph: remember the lovely “no account, no registration” client? That courtesy ends at the AI’s front door. PlugClaw wants activation: a one-time code sent to your phone number or email (the phone field helpfully defaults to +86), plus a ticked User Agreement and Privacy Policy, before the agent will do anything at all. Perfectly ordinary for a service that dispenses credits, mildly ironic on a stick sold on privacy. Worth flagging without being a dealbreaker: if you bought the Claw precisely to avoid handing anyone your digits, BYOK with a self-managed model is the cleaner path.

The claim, in practice: it works, and it really does operate the phone-OS side of the stick. In our tests it told a joke, generated a logo, translated text into Thai, produced a five-second video clip of a puppy, and (asked to review itself) used its android_use hooks to interrogate its own hardware and write up the results (specs correct, battery level invented: “24%, surviving on charm and optimism”; there is no battery).

Set loose on a real chore (“install a Play Integrity checker from a trusted source and report the result”) it found the right GitHub repo, realised it was source-only, pivoted to a mirror site, met a Cloudflare wall, and kept throwing itself at the door until we intervened. You have to admire the persistence, though it’s a reliable way to burn credits. On pure on-device control, it fared better: asked to use the calculator, it opened the app and tapped through the sum step by step: slowly, but correctly and visibly.

Model choice is broad: the picker lists three free models (Qwen and MiniMax variants) plus a paid ladder from 0.2× to 47× the base credit rate, with names like GPT-6-Astra and a Claude Sonnet that the vendor lists as “Caude-sonnet” (a typo shipped in their own UI). Availability varies by region (the setup list doesn’t include Thailand, so we activated under the USA), and a separate GUI model (Qwen3-VL) is what actually drives the screen-tapping. Video generation is a single option: ByteDance’s Seedance-2.0. You can also bring your own API key, and (the bit the marketing undersells) PlugClaw exposes a local API gateway at 127.0.0.1, so the stick doubles as a portable LLM endpoint for other tools.

Pricing is pay-as-you-go in “credits”: 100 per US dollar, sold in $5–150 packs via Stripe, valid for a year, no refunds. You get 1,500 credits (about $15) free at activation. Ours went 1500 → 1429: a few credits for text and image chores, about 66 for the video and self-review. Media generation is where the meter spins; chat is cheap. There are no subscriptions, which is a welcome change from the usual AI mugging.

The privacy pitch is “Confidential AI”: TrustKernel claims cloud inference runs inside hardware-protected confidential VMs neither they nor the provider can see into, while admitting upfront that frontier models do see your request content. Your workspace stays encrypted on the stick.

You can also text the stick on WhatsApp or Telegram and have the agent work remotely, and there’s a skills marketplace plus third-party ClawHub and MCP support, though the marketplace is in its early days: most community skills show single-digit install counts and lean heavily towards Chinese services (Qzone, ZAKER, Baidu trends). The way we’d put it: PlugClaw is a PlugMate with a junior employee living inside it (tireless, occasionally charming, prone to running face-first into Cloudflare), and the employee works for tokens.

Who is this actually for?

After weeks of research (the reading honestly started before the sticks arrived; we were keen) and our own testing, the genuine fits:

• The work/life splitters. A second phone’s worth of separation without a second phone: work apps, logins, and notifications living on the stick, vanishing when you unplug. Works, and it’s the cleanest use case.
• The second-phone substitute. Reality check first: it can’t replace a second handset for calls and texts. There’s no modem, no eSIM, no number. But for the more common reason people carry two phones (separate apps, separate accounts, separate files), it absolutely substitutes, and it weighs less than your car keys.
• iPhone and Mac people who need Android. If you daily-drive an iPhone but need that one Android-only app, this is far cheaper and less faff than a second handset, and it’s been run happily from iPads and iPhones.
• Identity compartmentalisation. (That’s the longest word in this review, and it earns every letter.) Related to containment but different: apps inside PlugOS don’t just fail to see your data; they see a different device. Fake the IMEI, the SIM identifiers, the MAC address, the GPS coordinates, and your actions inside PlugOS simply don’t attach to the identity your phone leaks everywhere. For anyone who needs an activity that can’t be joined to their real-world identity, for reasons legitimate or merely private, that’s the whole product in one sentence.
• App containment. The dodgy APK you need but don’t trust, the crypto wallet, the bank-token app — install them where they can’t read your photos, contacts, or IMEI, because there is no “your” IMEI to read.
• An AI agent with its own machine. If the idea of handing an agent your phone’s guts makes you itch (it should), PlugClaw gives it a separate, sandboxed machine to live on; your accounts and files stay outside while the agent works inside. The emerging niche of “AI that needs to do things, on hardware you control” is basically what this thing was designed for.
• Travel and the “banking bunker”. A use case close to our hearts: Thailand’s banking rules change often enough to be a hobby. Keep every banking and government app inside PlugOS, and your banking environment survives phone upgrades, losses, and customs-friendly factory resets. Our real-world run: TrueMoney Wallet worked, SCB Easy worked, Krungthai NEXT refused. It wants a genuine mobile-data path, demanded we turn off wifi, and wasn’t fooled even by a full Thai-carrier simulated SIM: it checks the transport, not the identity, and there is no modem to check against. Bonus entertainment: NEXT also detects screen-capture tools itself and refuses to open while you’re recording, which is more than PlugOS itself manages. So: two out of three Thai banking apps on our stick; check yours before you rely on it, because bank apps are a lottery, not a guarantee.
• Privacy-conscious users and activists — with conditions. Hardware isolation, a duress PIN, virtualized identifiers, and a real firewall are a genuinely useful toolkit for people who need compartmented identities, and the tamper-resistance demo above is impressive. But the conditions matter: closed-source OS and client, no published audit, Chinese jurisdiction, and a compromised host can still watch your screen and keystrokes. If your adversary is a thief or a nosy employer, this helps a lot. If it’s a state, talk to someone smarter than a gadget review first.
• The company-phone escape. If your employer hands you a managed phone, your personal life technically lives inside their MDM’s field of view. PlugOS sits outside it: keep your own accounts, photos, and chats on the stick, and when you unplug, the company device (and the company) is none the wiser about your weekend. One thing worth flagging: some workplaces frown on unexplained USB gadgets, so maybe don’t wave it around at the IT desk.
• Tinkerers. People have run Ubuntu chroots, local Llama 3.2 models, and automation macros inside PlugOS. It’s a legitimately fun little sandbox.

And who it’s not for: anyone wanting a phone replacement (no modem), a desktop replacement (portrait phone UI), a gaming device (it’s a Helio G80 in a 14 gram body, be reasonable), or plausible deniability against nation-state forensics (see above).

PlugMate vs PlugClaw: which to buy

They share the same PS01 platform, the same PlugOS, the same security toolbox. The PlugClaw adds the AI stack and undercuts the PlugMate by US$50. TrustKernel says the hardware is “streamlined” to hit that price, though on paper our units are near-identical (we benchmarked the Mate only; the Claw’s extra 2 GB showed no disadvantage in daily use). Availability nudges the decision further: the PlugMate’s only paper advantage (a 6 GB/256 GB config) is permanently out of stock (TrustKernel blames storage-chip supply, no restock timeline), so the Mate you can actually buy is the 4 GB/128 GB one. And the Claw dips lower still: a 4 GB/64 GB SKU drops the entry price beneath the US$149 we quoted, which is plenty if the stick’s job is apps and agents rather than hoarding files. If you want the vault and don’t care about AI, PlugMate makes sense only if you truly value its positioning; for most people, PlugClaw is the better buy. Between you and I, same isolation, plus an eager little helper that does your chores, for less money. And after testing: the agent does earn its keep; on-device it genuinely operates apps, slowly but correctly, as long as your chores stay inside the stick and away from Cloudflare.

Verdict

Both are clever, genuinely unusual devices that mostly deliver the thing they promise: a real, encrypted, second Android that lives on a keyring and borrows whatever screen you feed it. The isolation is real, the privacy toolbox is thoughtfully designed, the duress and tamper behaviors are above anything else at this price. You do pay in other currencies, though: closed source, unaudited, Chinese jurisdiction, host-keyed PIN, a phone UI on desktops, and heat you’ll feel. At US$199 for PlugMate and US$149 for PlugClaw on sale, PlugClaw is the easy recommendation of the two, assuming you have chores it can actually do. We’d like to thank TrustKernel (PlugOS) for providing both review units; the opinions, measurements, and gripes above are our own. You can purchase either on the PlugOS website, and CNX Software readers get a 5% discount with the coupon code CNX.

The post PlugMate and PlugClaw Review – Thumb-sized Android-based private computers with optional Ubuntu/OpenClaw support appeared first on CNX Software - Embedded Systems News.

Meta, Facebook’s parent company, has released the open-source Muse Gadgets SDK, designed to work with ESP32 boards or Linux-based SBCs like a Raspberry Pi.

This requires some explanation… I’ve been on holiday for a while, so I’m not up to date on everything, but apparently Meta released Muse on September 6, yet another personal AI agent… Earlier this month, Nat Friedman, head of product at Meta Superintelligence Labs, shared an X post about the Muse Gadgets “side project” that aims to make ESP32 and Linux devices work with Muse, basically enabling hardware AI monitors and Physical AI with Muse.

The Muse Gadgets SDK has been released on GitHub under an Apache 2.0 license (for most of the code) with two main directories:

• ESP32 Device SDK – It can show images, supports audio input and output, and can interface with sensors, actuators, buttons, etc…
• Linux Device SDK – Works with Raspberry Pi or other Linux SBCs and boxes. One use case is having Muse handle sysadmin chores or your Home Assistant setup.

A third directory contains skills from a range of hardware such as Apple HomePod Mini, printers from Brother and Epson, ESPHome devices, 3D printers, various TVs, Smart Home devices, and so on.

Meta recommends Raspberry Pi 5 for Linux, Waveshare ESP32-S3-Touch-AMOLED-1.75C devkit (1.75″ AMOLED touchscreen, speaker, mic, and battery), or Seeed reTerminal E1002 color e-ink display to get started, but you can use any other similar devices with some work. An HDMI stick will also be supported, but no details yet.

Alternatively, Meta also manufactured 5000 Muse Home Link devices that will be provided for free to Muse subscribers in the US in the next few weeks as long as stock lasts. It’s based on an ESP32-C5 wireless MCU providing dual-band WiFi 6, BLE, and 802.15.4 (Zigbee/Thread) connectivity, features 8MB flash and 8MB PSRAM, a USB-C port for power and programming, an LED, and not much else.

You’ll need a Muse SDK token and review the Gadget SDK Terms before pairing a Muse Gadget with the Muse app on iOS or Android (Settings > Devices). Developer mode must be enabled, and you should find a device called “MuseGadget”. Besides the GitHub repository, you’ll find more details on the project’s website, where you can also claim a free Muse Home Link if you qualify (US-based Muse subscriber).

JJ Englert shared one demo with Muse integrated with a Waveshare ESP32 “Muse Gadget”.

Got my Muse working on my ESP32. Even gave him a voice. Really cool! https://t.co/xaA1REByRm pic.twitter.com/jlxzsaPnz6

— JJ Englert (@JJEnglert) October 3, 2026

The post Facebook releases Muse Gadgets SDK for ESP32 and Raspberry Pi/Linux-based devices appeared first on CNX Software - Embedded Systems News.

Wireless-Tag WT32P4C5-43S (ZX4D30CE405-V1.3)  is a fully integrated 4.3-inch touch display development board that combines an ESP32-P4 MCU with an ESP32-C5 wireless module supporting dual-band WiFi 6 (2.4/5 GHz), Bluetooth LE, and 802.15.4 connectivity. Unlike previous ESP32-P4 + ESP32-C5 solutions, this one comes with a built-in touchscreen, so you don’t need to connect a separate display.

Designed for industrial automation, smart control panels, and edge AI vision terminals, the board also integrates a 15-pin MIPI CSI camera interface, a microSD card slot, two USB Type-C ports (USB 2.0 High-Speed and Debug/JTAG), and a 14-pin GPIO expansion header.

Wireless-Tag WT32P4C5-43S (ZX4D30CE405-V1.3) specifications:

• Core module – Wireless Tag WT0132P4-A1-N16R32
  • SoC – Espressif Systems ESP32-P4
    • CPU
      • Dual-core 32-bit RISC-V HP (High-performance) CPU @ up to 400 MHz with AI instructions extension and single-precision FPU
      • Single-RISC-V LP (Low-power) MCU core @ up to 40 MHz with 8KB of zero-wait TCM RAM
    • Memory
      • 768 KB HP L2MEM (for dual-core CPU), 32 KB LP SRAM, 8 KB TCM (for LP MCU core)
      • 16MB or 32MB PSRAM
    • Storage – 128 KB HP ROM, 16 KB LP ROM
    • GPU – 2D Pixel Processing Accelerator (PPA)
    • VPU – H.264 and JPEG codec support
  • Storage – 16MB NOR flash
  • Dimensions – 25 x 20mm
• Storage – MicroSD card slot
• Display
  • 4.3-inch IPS TFT LCD with 800 × 480 resolution and capacitive touchscreen
  • Driven via an 18-bit parallel RGB interface and I2C capacitive touch interface
  • Transmissive backlight with 500 cd/m² brightness, 16.7M colors, all-around viewing angle
  • Capacitive touchscreen (GFF structure) with Hynitron CST3240 controller via I2C
• Camera – 15-pin, 1.0mm pitch MIPI CSI-2 FPC connector (2-lane data + 1-lane clock, Raspberry Pi CSI camera compatible)
• Wireless – Espressif Systems ESP32-C5 32-bit RISC-V wireless SoC with dual-band (2.4 GHz / 5 GHz) Wi-Fi 6, Bluetooth 5 (LE), and 802.15.4 radio (Zigbee 3.0 / Thread 1.3)
• USB
  • USB 2.0 High-Speed Type-C port for power and programming
  • Debug USB Type-C port (USB 2.0 Full-Speed) for firmware flashing, serial console, and JTAG debugging
• Debugging – 2x 7-pin Molex PicoBlade-compatible connectors (1.25mm pitch)
  • 1x ESP32-P4 debugging
  • 1x ESP32-C5/C61 wireless module debugging
  • Compatible with the ZXACC-ESPDB debugger
• Expansion – 14-pin (2×7, 2.54mm) pin header with 7x GPIOs (GPIO19–23, plus GPIO26/27 multiplexed with USB1 D+/D-), 5V, 3.3V, and GND
• Misc – Reset button
• Power
  • 5V/1A  via either USB Type-C port
  • 5V input/output and 3.3V output via the 14-pin expansion header
• Dimensions – 117.20 × 81.04 × 16.40 mm (overall display dimensions + max thickness)

What’s interesting is that though the ESP32-P4 supports MIPI DSI displays, the onboard 4.3-inch screen uses an 18-bit parallel RGB interface (RGB_D0 to RGB_D17), leaving the MIPI CSI-2 interface for Raspberry Pi-compatible camera modules. Because the board features two separate ESP32 chips, the PCB includes 7-pin MX1.25 debug headers for both the main ESP32-P4 controller and the companion ESP32-C5/C61 wireless module, along with the onboard USB-C debug port.

The board can be programmed with Espressif’s ESP-IDF (v6.x), Arduino IDE, and the LVGL graphics library for building HMI applications. The manufacturer also provides two GitHub repositories: the ZX4D30CE40S_V13 repository is for the main ESP32-P4 firmware and BSP drivers, and ZX4D30CE40S_V13_slave for the ESP32-C5/C61 wireless coprocessor firmware. The company also provides PDF schematics, 2D DWG drawings, and 3D STEP files, with links available in the Chinese user manual.

We have previously covered Wireless-Tag’s WT99P4C5-S1 and WTDKP4C5-S1 development boards, both combining the ESP32-P4 with an ESP32-C5 for dual-band WiFi 6 connectivity. We also looked at the Maker Go ESP32P4C5 Core Board, which features Raspberry Pi-compatible MIPI display and camera connectors. However, all three boards need a separate display. The WT32P4C5-43S includes a 4.3-inch touchscreen, making it a more complete “turnkey” solution for HMI applications. We have also covered the GUITION JC4880P433, which has a similar 4.3-inch touchscreen but uses an ESP32-C6 module limited to 2.4 GHz WiFi 6.

The Wireless-Tag WT32P4C5-43S is available from the company’s online store for $32, while a second option listed as “WT32P4C5-43S with” (no typo or missing word) costs $37. The product page does not clearly explain what the more expensive option includes.

The post Wireless-Tag WT32P4C5-43S – A fully integrated 4.3-inch ESP32-P4 and ESP32-C5 touch display devkit appeared first on CNX Software - Embedded Systems News.

M5Stack has just introduced the ToughC5, a rugged, weatherproof outdoor IoT controller built around the ESP32-C5. The device features a bright 2.0-inch capacitive touch IPS display and RS485 connectivity.

The ToughC5 is an upgrade to the ESP32-base3d M5Stack TOUGH introduced in 2021, with the upgraded model featuring a more modern ESP32-C5 MCU with dual-band (2.4/5 GHz) Wi-Fi 6, Bluetooth 5 LE, and an IEEE 802.15.4 radio for Zigbee and Thread. Other changes include faster Octal PSRAM, an upgraded RTC chip with a dedicated coin cell backup battery, a passive buzzer replacing the previous 1W speaker, and a custom power management system (M5PM1 + M5IOE1) replacing the AXP192 PMU. These features make it suitable for industrial field control, outdoor edge data collection, and smart building applications.

M5Stack ToughC5 specifications:

• Wireless SoC – Espressif Systems ESP32-C5HR8
  • CPU
    • Single-core 32-bit RISC-V processor @ up to 240 MHz
    • Low-power RISC-V core @ 40 MHz acting as the main processor for power-sensitive applications
  • Memory – 384 KB SRAM on-chip, 8MB Octal PSRAM
  • Storage – 320 KB ROM
  • Wireless Connectivity
    • Dual-band (2.4GHz/5 GHz) 802.11ax WiFi 6, with 802.11b/g/n WiFi 4 standard fallback
    • Bluetooth 5.0 Low Energy (LE)
    • 802.15.4 radio for Zigbee 3.0 and Thread 1.3
• Storage
  • 16MB flash
  • MicroSD card slot
• Display – 2.0-inch IPS LCD (ILI9342C) with 320×240 resolution, up to 853 nits brightness; CHSC6540 capacitive touch driver
• Audio – Onboard passive buzzer
• USB
  • USB Type-C port for power, programming, and debugging (Inside the waterproof housing)
  • Weatherproof M12 to USB-A cable (1-meter long)
• Expansion
  • 4x internal HY2.0-4P expansion interfaces (RS485, I2C, GPIO, UART) designed to be routed through the second M12 connector if needed.
  • 4-pin I2C expansion header (Inside the waterproof housing)
• Misc
  • RX8130CE Real-Time Clock (RTC) with CR2032 coin cell battery
  • Physical button (Inside the waterproof housing)
    • Single click – Power on / Reset
    • Double click – Power off
    • Long press – Enter download mode
  • Antenna – Internal FPC antenna (IPEX-2 connector)
  • 2x M12 body cable routing holes for external expansion
• Power Supply
  • 5V DC @ 500mA via USB Type-C
  • 6V to 24V DC @ 1A via RS485 interface input
  • 2-pin battery connector
  • Multi-level power management via M5PM1 and M5IOE1 chips
  • Consumes 320 mA @ 4.2V at full load and 10.41 µA @ 4.2V in shutdown state
• Dimensions – 75.9 x 58.0 x 42.5 mm
• Weight – 119.0 grams
• Enclosure – UV-resistant PC+ABS housing with basic water and dust resistance (no IP rating, not designed for immersion)
• Mounting – Magnetic mounting support

Like other M5Stack devices, the ToughC5 can also be programmed with the Arduino IDE. The company provides a quick start guide, along with M5Unified and M5GFX driver libraries, on its documentation page. The company also talks about the UiFlow2 firmware, which is actively being developed; once finished, it will let you create applications using a web-based, Blockly-style programming interface. You can install the firmware using M5Burner, and the UiFlow2 Web IDE supports programming and debugging over USB or Wi-Fi.

The M5Stack ToughC5 IoT Dev Kit is now available for $49.9 on AliExpress and the company’s official store. The device ships with a 1-meter USB-C cable, a pair of Allen wrenches, terminal cables, and a waterproof O-ring. Since the written documentation and accompanying photos for the ToughC5 IoT controller are rather poor, you may want to watch the short video below to clearly understand how the connectors are arranged.

The post M5Stack ToughC5 weatherproof ESP32-C5 IoT controller offers dual-band Wi-Fi 6, BLE 5, and Zigbee/Thread appeared first on CNX Software - Embedded Systems News.