Inside Electronics

Axelera’s artificial intelligence processing unit is akin to the neural processing units (NPU) and are designed to handle AI inferencing while using very little power. Their quad core Metis AIPUs delivers 214 TOPs at under 4 W. It is available as a chip as well as on boards like their M.2 module. The company's latest Europa delivers up to 629 TOPs and can handle even more streams. Alexis Crowell, General Manager of Americas & G...

In this episode of the Inside Electronics podcast, we explore how modern industrial automation relies on robust, well designed power supplies to keep demanding systems running reliably. Host Alix Paultre, Editor-at-Large of Electronic Design, speaks with Mike Swenson, Marketing Manager for TRACO North America, about why industrial environments are among the toughest places for electronics, from electrical noise and extreme temperat...

While at least a dozen companies offer excellent solutions to deliver the ultra-high currents required to power modern GPUs, Ferric developed a unique capability to integrate thin-film magnetic power inductors within their devices’ packages that eliminates the need for bulky external components. As a result, they can create chip-scale power converters that integrate interface, telemetry, feedback control, and powertrain circuitry.

...

There are all kinds of virtual machines (VM) available and many are being employed in embedded applications. Platforms like the Java virtual machine (JVM) are used on servers and embedded applications with features like garbage collection. Unfortunately, these can have a lot of overhead making them less desirable for small, microcontroller-based systems. Rolf Segger, Founder of Segger Microcontroller Systems, talks about the compan...

IoT empowers intelligent machines to provide advanced functionality to society, but it isn’t easy to get there from here. Key aspects of development include the importance of creating tools that empower others, the role of the market ecosystems in driving adoption, and the challenges in the infrastructure needed to bring connected technologies to the market. We talk to Alistair Fulton, CEO of IoT Labs, about the future of the marke...

Industrial subsystems are distributed so communication between these subsystems needs to preserve this information and deliver it in a timely fashion. This is where time sensitive networking (TSN) comes into play. We talk to Mark Geisler, Marketing Manager at Analog Devices, about the future of TSN in plug-and-play industrial networks as well as the technical and organizational challenges that engineers face when incorporating TSN ...

Nanusens creates nanoscale sensor structures inside the CMOS layers using standard CMOS processes within the same production flow as making the control electronics on the same chip. This approach reduces the size and cost as it benefits from the vast economies of scale of using giant CMOS fabs. The resulting single chip solution has a packaged size of 0.5 mm³. This episode delves into the technology, its future, and some limitation...

RAID, redundant array of independent disks, uses redundancy to provide a more robust disk drive array capable of operation even with the loss of a drive. RAID 1 replicates data with a 50% overhead. RAID 5 and 6 are more common in larger arrays with less overhead as more drives are added to the mix but with a tradeoff in performance.  

Mark Anthony, Product Line Manager at Microchip Technology, talks about the company's SmartRAID 43...

AI workloads and other advanced computing applications continue to expand, pressuring the development community with the dual challenges of performance and manufacturability. Traditional SoCs are approaching their limits in terms of size, yield, and cost, and one solution lies in chiplet-based architectures. In this episode, we talk to Larry Zu, CEO at Sarcina Technology, and how the company is deploying interconnect to minimize si...