Challenges of Conventional Methods and Features of This Reference Design
Arc faults are one of the primary causes of electrical fires in solar power generation systems, energy storage systems, EV charging infrastructure, and industrial and residential power distribution equipment. Previously, arc fault detection typically involved comparing signals detected by sensors against preset thresholds to determine if an anomaly occurred.
However, in residential and commercial alternating current (AC) circuits, everyday appliances such as vacuum cleaners, power drills, and dimmers can generate arcs at switch contacts or motor brushes, producing signals that closely resemble dangerous arc faults.
Furthermore, in direct current (DC) systems like solar power generation, EV chargers, and energy storage systems, switching transient phenomena3 caused by relay contact bounce, inverter operation, or capacitor inrush currents generate broadband noise in the same frequency band as actual arc faults. This often leads to unnecessary system shutdowns due to false detections.
In contrast, Microchip’s AFD reference design, leveraging machine learning, addresses these challenges. By directly executing edge ML models on the dsPIC33A DSC, which features an integrated DSP engine and advanced analog peripherals, low-latency machine learning inference is achieved. This approach reduces false detections and improves arc fault detection performance compared to conventional methods.
3 Switching Transient Phenomena: Temporary large fluctuations in voltage or current that occur immediately after a circuit switch is turned on or off, before settling into a stable (steady) state.
Role of Current Sensors and Collaboration Between the Two Companies
For high-precision arc fault detection using machine learning, it is crucial for sensors to accurately capture the current signals specific to arc faults.
AKM’s CZ39 and CZ3K are “coreless” current sensors that detect current without using a magnetic core. Their 100ns high-speed response and low-noise performance contribute to the high-precision determination in Microchip’s reference design.
Furthermore, the engineering team at AKM Semiconductor, Inc. (hereinafter “AKMS”), AKM’s U.S. subsidiary, collaborated closely with Microchip on the current detection configuration during the development and validation phases of the reference design. This collaboration has resulted in a new protection solution that combines high-performance current detection with edge processing, supporting the reduction of false detections and enhancing arc fault detection performance.
Comment from Chris Baltar, Vice President of Business Development at AKMS
“We are pleased to have supported Microchip in building and validating this reference design using AKM’s CZ39 and CZ3K series,” said Chris Baltar. “This collaboration demonstrates how combining high-performance current sensing with intelligent edge processing can help designers implement advanced protection features across a wide range of power applications. As AI processing expands, driving needs for higher power density and new high-voltage power systems, we are keen to explore the potential for this reference design in applications such as data centers.”
Application Areas and Future Outlook
This demo application is available through Microchip’s reference design program and can be applied to solar power generation, energy storage systems, EV chargers, smart ignition systems, electronic fuses (e-Fuse), and residential and industrial safety switches.
The Asahi Kasei Group positions its electronics business as a “key growth” area driving overall group profit growth. In electronic components, it provides unique high-performance parts such as ICs for mobile device camera modules and in-car audio solutions. For coreless current sensors, building on their proven track record in EV applications, the company plans to further expand their use into areas such as AI data centers, positioning them as a flagship product for future growth.
For more information on product details and partner companies, please visit the following websites:
