Latest News

High performance IoT solutions for AI at the edge can now be created up to 30 per cent faster, says Sondrel, using the SFA 100 IP reference platform to simplify the creation of battery-powered IoT devices.

The design has an onboard Arm CPU to locally process data gathered from its associated sensors for onward transmission via wire or wireless connection for further analysis. Security is built in using standard secure/encrypted protocols.

The IoT is based on gathering data about the nodes’ environment and the SFA 100 IP platform has been designed specifically for this as a compact, powerful, compute capability to intelligently transform data into knowledge and add intelligence, says Sondrel. Most of the design work has already been done to create this reference design so that all that is needed is the customer’s own IP and any minor customisation needed for the final ASIC. This semi-custom approach, which Sondrel calls Architecting the future, reduces risk, design costs and time to market by up to 30 per cent compared to starting from scratch, says Sondrel.

The SFA 100 provides the ability to integrate a machine learning engine onto a low cost, low power edge device and is based on the Arm Corstone-300 subsystem that ensures a very high level of security. An Arm Cortex-M55 has secure boot with cryptographic algorithm accelerators and supports TrustZone and CryptoCell for additional security for the onward processed data. There is also the Arm Ethos-U55 machine learning (ML) processor that provides a 480x uplift in ML performance, claims Sondrel. This combined processing capability enables the endpoint device to perform ‘smart’ operations such as voice activation, image classification, gesture recognition, filtering, inference or tracking, depending on the application. If more memory is required, it can be scaled up via the DRAM interface

The design has GPIO, I2C, UART and QSPI peripheral interfaces to capture sensor data, such as video, image, sound and both static and dynamic data, and a 12-bit audio DAC for voice commands. The low power requirement of the SFA 100 design enables it to operate on battery power for long periods, together with low power, wireless connectivity using either Bluetooth BT5.1-LE or ZigBee.

Sondrel offers a full turnkey service that turns designs into fully tested, shipping silicon

Founded in 2002, Sondrel is a define-and-design ASIC consultancy, complemented by its turnkey services to transform designs into tested, volume-packaged silicon chips. This single point of contact for the entire supply chain process ensures low risk and faster times to market, says the company.

Headquartered in the UK, Sondrel supports customers around the world via offices in China, India, France, Morocco and North America.

http://www.sondrel.com

Machine learning (ML) inference enables real time analytics in 5G services. Capgemini’s Project Marconi intelligently boosts subscriber quality of experience with real-time predictive analytics. It conforms to Open Radio Access Network (O-RAN) guidelines and is believed to be the industry’s first artificial intelligence / machine learning (AI / ML) -based radio network application for a 5G medium access control (MAC) scheduler.

Capgemini has based Project Marconi on Intel architecture designed to increase the amount of traffic each cell can handle. It allows operators to serve more subscribers and launch new industry 4.0 services such as enhanced mobile broadband (eMBB) and ultra-reliable low latency communications (URLLC) use cases.

Walid Negm, chief research and innovation officer at Capgemini Engineering said: “We gathered and utilised over 1Tbyte of data and conducted countless test runs with NetAnticipate5G to fine-tune the predictive analytics . . . In short, machine learning can be deployed for intelligent decision-making on the RAN without any additional hardware requirement. This makes it cost-efficient in the short run and future proof in the long run as we move into cloud native RAN implementations.”

Cristina Rodriguez, vice president of Intel’s wireless access network division, added: “Our 3rd Gen Intel Xeon Scalable processors with built-in AI acceleration provide high performance for deep learning on the Net Anticipate 5G platform.” The two companies collaborated to deliver fast inference data to enhance the Open-Source ML libraries to create a RAN that can predict and quickly react to subscriber coverage requirements.

Unveiled at MWC21, the AI powered analytics uses NetAnticipate5G and RATIO O-RAN. It forecasts and assigns the appropriate modulation and coding scheme values for signal transmission by forecasting of the user signal quality and mobility patterns accurately. RAN can therefore intelligently schedule MAC resources to achieve up to 40 per cent more accurate MCS prediction and yield to 15 per cent better spectrum efficiency in the case studies and testing. According to Capgemini, it delivers faster data speeds, better and more consistent quality of experience (QoE) to subscribers and robust coverage for use cases that rely on low latency connectivity such as robotics-based manufacturing and vehicle-to-everything (V2X).

http://www.capgemini.com