This year’s Embedded World welcomed more than 32,000 visitors, from more than 80 countries. This figure is nearly 20% more than 2023’s 27,000 visitors and the number of exhibitors has grown by 150 compared to last year, totalling “over 1,100” reports the show’s organisers.
This year, the unofficial theme of the show seemed to be edge AI. Among the AI-related announcements, AMD introduced the next generation of Versal AI Edge and Versal Prime adaptive SoCs.
This generation of AI engines triples TOPS per Watt compared to its Gen 1 Versal AI Edge devices and integrates eight Arm Cortex A78AE cores and 10 Arm Cortex-R52 real-time processors. The result, says AMD, will increase scalar compute by a factor of 10 compared to Gen 1 AI Edge and Prime models.
AI Edge devices integrate FPGA programmable logic fabric for real-time preprocessing. They connect to sensors and implement high throughput, low latency data processing pipelines. For AI inference there is an array of vector processors, while post-processing is provided by the Arm CPU cores. The level of integration means that a single chip can be used in place of multiple processors for decision-making and control operations.
The Versal Prime series combines programmable logic for sensor processing with embedded Arm CPUs to accelerate traditional non-AI based embedded systems. There is also a new hard IP for high throughput video processing, including up to 8K multi-channel workflows, making the devices suitable for UHD video streaming and recording, industrial PCs and flight computers.
Early access documentation was announced in Nuremberg and developers can use Gen 1 Versal evaluation kits and design tools (Vivado and Vitis) today. Both the Versal AI Edge and Versal Prime will begin sampling in the first half of next year, followed by evaluation kits and system on modules samples in mid-2025.
The company also announced that car manufacturer Subaru will use the Gen 2 Versal AI Edge series in its new EyeSight ADAS vision system. AI at the edge will enable advanced safety features, including adaptive cruise control, lane-keep assist and pre-collision braking. Subaru uses AMD’s Zynq and PSoC chips.
Cybersecurity
Winbond spin-out Nuvoton announced its NuMicro M2L31 microcontroller series, based on an Arm Cortex-M23 core Armv8-M architecture. Operating at up to 72MHz, the series comprises devices with 64-512kbytes ReRAM and 40-168kbytes SRAM. Operating voltage is 1.71V to 3.6V and the operating temperature range of -40°C to 105°C makes it suitable for use in motor control, PC peripheral and battery management systems. The series also offers high immunity (4kV ESD HBM and 4.4kV EFT), making the microcontrollers suitable for automated industrial applications.
The choice of ReRAM for digital data storage does not require a page erase before writing, which accelerates read/write operation. It also consumes less energy for storing each bit, compared with flash memory. At Normal Run mode it requires 60μA/MHz. Each storage unit can be individually set or reset, which improves endurance compared with flash, says the company. There are three power-down modes: normal, standby and deep power-down. In the normal power-down mode, MIRC, LIRC and LXT can be operated to shorten the stable time of high precision sensors and lower the entire system’s power consumption. An auto-operation mode in the low power domain allows sensor data to proceed through low power serial interfaces without interrupting CPU operation.
There are up to three programmable gain amplifiers, three comparators in conjunction with PWM to ensure cycle-by-cycle over current limiting, one 24-channel, 3.6MSPS, 12-bit SAR ADC, up to two 12-bit, 1MSPS DACs for voltage reference, a 12-channel 16-bit 72MHz PWM and up to 12-channel 16-bit 144MHz EPWM.
The series includes a USB Type-C Rev. 2.1 and USB Power Delivery Rev 3.0 compliant model. The USB Type-C model has pull-high and pull-low resistors, supports dead battery and fast role swap and has flexible power delivery and data transmission over a single cable.
Peripherals across the M2L31 series include RTC with independent VBAT pin, up to 16 channels of PDMA, up to 16 independent capacitive touch key sensing, up to two sets of CAN FD controller, USB 2.0 FS OTG host and device controllers, up to eight sets of UART, up to four sets of I²C and up to four sets of SPI/I2S, one set of QSPI and two sets of universal serial control interfaces.
The microcontrollers are available in a range of packaging, including WLCSP 25 (2.5×2.5mm), QFN32 (5.0×5.0mm), LQFP48 (7.0×7.0mm), QFN 48 (5.0×5.0mm), WLCSP 49 (3.0×3.0mm), LQFP64 (7.0×7.0mm) and LQFP128 (14x14mm).
The company also supports third parties’ integrated development environments, including those from Keil MDK, IAR EWARM and the NuEclipse IDE with GNU GCC compiler.
Before Embedded World Nuvoton partnered with zeroRISC to offer what it claims to be the first open source, commercial chip based on the OpenTitan secure silicon design.
“Until now, nearly all connected devices have been built without true foundational security and so are open to stealth cyberattacks both in the silicon supply chain and after deployment in the field,” says Dominic Rizzo, founder and CEO of zeroRISC. “By partnering with Nuvoton, Winbond and other coalition partners to produce the first commercial-grade open-source chip in the world, we’re delivering on the promise of the OpenTitan project and democratising access to open, trustworthy foundational security.”
Conventional security starts at the OS layer, yet some of the most damaging cyberattacks target the firmware layer and the hardware on which it runs. OpenTitan secure silicon operates below the OS and is used with secure-by-design firmware. The discrete OpenTitan chip is based on the ‘Earl Grey’ OpenTitan discrete chip design and is described as ensuring the integrity of motherboards and network cards, laptops, phones and IoT platforms with a hardware root-of-trust.
“Open, trustworthy foundational security has become a ‘must-have,’ not a ‘nice-to-have’,” says Erez Naory, vice-president of client and security products at Nuvoton.
Winbond and zeroRISC are actively engaged with the OpenTitan coalition on the joint development of the first OpenTitan-based integrated secure execution environment, which is scheduled for production release later this year.
AI and the IoT
A series of embedded AI-native IoT processors was introduced by Synaptics. The SL-series processors are multi-core Linux or Android SoCs based on Arm Cortex A-series CPUs. They include hardware accelerators for edge inferencing and multimedia processing on audio, video, vision, image, voice and speech. The SL-series processors are part of the company’s Astra platform to introduce AI into products and independently of the data centre to enhance data privacy and latency, says the company. It also announced the Astra Machina Foundation series development kit.
“The SL-Series delivers on the high-performance end of a roadmap of scalable edge AI IoT compute solutions and will soon be complemented by our power-optimised AI-enabling SR-Series of MCUs,” says Vikram Gupta, senior vice-president and general manager IoT processors at Synaptics.
There are three devices in the SL-series. The SL1680 is intended for home and industrial control, smart appliances, home security gateways, digital signage or PoS systems and scanners. It is based on a quad-core Arm Cortex-A73 64-bit CPU, a 7.9 TOPS NPU, a GPU, and a multimedia accelerator pipeline.
The SL1640 is optimised for cost and power, says Synaptics. It is based on an quad-core Arm Cortex-A55 processor, a 1.6+ TOPS NPU and a GE9920 GPU. The third processor is the SL1620, based on a quad-core Arm Cortex-A55 CPU subsystem, a GPU for use in advanced graphics, AI acceleration, audio algorithms and dual displays.
The SL-series processors are available now and the Astra Machina Foundation series development kit, initially based on the company’s SYN43711 and SYN43752 Wi-Fi and Bluetooth combo SoCs, will be available in Q2, 2024.
AI-powered antenna
Spanish antenna manufacturer Ignion introduced its Oxion technology, the next progression in interactive RF design, following the company’s Antenna Intelligence Cloud design guide. Oxion allows IoT device developers to shorten design time using AI/ML to identify obstacles to wireless connectivity. The platform is powered by AWS and provides performance prediction and design guidance tailored for a specific device design using proprietary algorithms. It uses a drag-and-drop function to allow designers to adjust the design online and see immediate performance feedback for optimal antenna component, placement, clearance and PCB size choices.
Instead of relying on a datasheet, Oxion provides an interactive digital twin, explains CEO Jaap Groot. He adds that the installed base of connected devices is expected to reach 70 billion in 2026. A free-to-access platform was created through the “spirit of transparency and democratisation using AI” and with a grant from the EU Innovation Council’s Accelerator program and working with Barcelona-based Basetis, which developed active learning algorithms for the real-time results engine, he says.
Cloud-based system development
An expansion to the Quick Connect Studio by Renesas Electronics brings real-time code customisation, remote debugging and the ability to graphically co-optimise hardware and software for quick validation of prototypes.
Engineers can drag and drop device and design blocks on the cloud and then generate, compile and build the base software automatically. According to Renesas, this is “a major shift towards the no-code paradigm of development”. Compilation is handled by the cloud to automatically generate system software that can be iteratively tested on the hardware.
The development kit supports all the company’s RA MCU boards, as well as multiple Renesas wireless modules and sensors. Devices from partners such as ams Osram, TDK and Arducam have been added.
The development kit provides automatically generated code, which can be customised in real-time in the browser application. Using remote board farms, the code can be tested on target hardware dynamically to verify the operation before buying or building the physical board and setting up the test system. This “dramatically” speeds up building proof of concepts and prototyping systems by concurrently working on both software and hardware components of system design.
Sensor networks
Still with the cloud, NeoCortec demonstrated its NeoMesh Click boards from MikroE, which integrates sensors for wireless mesh networking, interfaced to Avnet’s IoTConnect cloud to process and analyse data in smart buildings, metering or security applications as well as agricultural, transport, medical or logistics use.
The Click boards can be used by engineers to save developing PCB hardware and to conduct proof of concept testing on the NeoMesh network protocol stack in a real applications.
The boards use NeoCortec’s low-power NC1000 and NC2400 network modules. Three versions are available, operating at 868MHz, 915MHz and 2.4GHz. Average power consumption is down to 20µA.
Design tools
Getting back to the roots of Embedded World, embedded design tools were the focus of LDRA’s stand. The first was the introduction of security-first capabilities to its tool suite. The company has included extended taint analysis and vulnerability heat maps to help developers address design issues early in the design cycle. The four new capabilities are security vulnerability reports, security audits, security reviews and taint analysis.
“Security vulnerabilities in mission-critical systems are just waiting to be exploited, but LDRA can help developers identify and remediate them before release,” says Ian Hennell, operations director, LDRA. The new features abstract “much of the noise and complexity of traditional analysis tools” says the company. “Spending less time on ‘analysis paralysis’ enables embedded software teams to focus on early vulnerability remediation and faster compliance,” continues Hennell.
The security reporting, audits and heat maps can identify and pinpoint vulnerabilities in code and the suite also provides deep visibility into component-level vulnerabilities, software complexity and memory allocation impacts. The automated CWE, MISRA, and CERT-C security reviews reduce the time needed for compliance.
The company also introduced domain-specific software productivity packages for the LDRA tool suite. The single, streamlined workflow replaces different traceability, software analysis and compliance reporting point solutions for functional safety and security standards across different industry sectors.
The packages have all the necessary software tools to demonstrate compliance with the highest required assurance levels, says the company, instead of using a selection of verification and validation tools, often from different vendors, to analyse and test software, provide traceability between development artifacts and generate documentation and compliance reports.
Each package is domain-specific (for example, medical, defence) and offers capabilities ranging from static and dynamic analysis with unit and integration testing to complete compliance reporting for specific industry standards and assurance levels.
Addressing functional safety and security, each includes coding standards compliance, security assessments, automatic test case generation, execution, artifact generation and compliance reporting. There are also add-in products for specific industries, applications and required integrations.
“We’ve strategically crafted these packages to support large-scale industry demands and the focused requirements of start-ups and small- to medium-size companies that need tailored solutions fitting growth-scale budget constraints,” Hennell says.
Finally, there was an announcement from industrial computing company Advantech and Qualcomm Technology, announcing their intention to collaborate in an “open and diverse” edge AI ecosystem for AIoT. Qualcomm will provide semiconductors and SoCs for an Advantech portfolio of AI-on-modules, AI function boards and AI edge platforms. Advantech also intends to develop a dedicated edge AI software development kit.