Embedded Computing Design

AI consumes a lot of power simply because it requires a lot of compute to handle AI calculations. If you could reduce the amount of compute that’s needed to handle those calculations, you could reduce the power that’s required. It sounds simple, but I assure you, it’s anything but. 

However, the developers at Velaura AI feel that they have cracked the code, so to speak, in reducing the power needed to run AI computers. 

To understand how they’ve done it, and what it means for the future of AI, check out this week’s Embedded Executives podcast, where I spoke to Sanjay Gupta, the Co-founder & President of Velaura AI.

Nary does a conversation take place these days in my professional circles without some mention of AI. And for good reason. It’s not just the latest cool technology. It is, without question, a game changer. 

You are likely familiar with the Z-Wave Alliance, a standards development organization that is dedicated to developing and advancing Z-Wave wireless technology. What you may not be aware of is all the work that’s going on behind the scenes at the organization, as well as what’s coming in the near future. 

To help rectify that, for you as well as for me, I spoke to Avi Rosenthal, the Board Chairman for the Alliance, on this week’s Embedded Executives podcast.

AI is front and center in so many conversations these days. However, we tend to focus on the processing element when it comes to AI. While that element is vital to the growth of AI, there’s another area just as important and somewhat under the radar. And that’s power. We need a way to efficiently power these huge processing elements. 

The term software-defined vehicle (SDV) is not new, not even close. The technology has been around for a long time, and you could argue it has been in place and in use for a long time. But the vision goes so far beyond what we have seen to date. And to me, that’s pretty disappointing. 

To gauge whether my opinion was accurate and, if so, why that’s the case, I spoke to Andrew Hunter, a Senior Vice President with Infineon Technologies, on this week’s Embedded Executives podcast.

Many engineers have been in their current positions for a long time. In many instances, that’s a great thing. But it could mean that you’re missing out on a great opportunity. That said, where do you turn if you’re “job shopping,” and if you get an interview, what do you do?

Jeremy Werner, the Senior VP and GM of Micron's Core Data Center Business Unit, gave the Keynote Address at the recent Future of Memory and Storage conference.

The community has been designing embedded systems for a very long time. If we look at one slice of time, say the last 25 years, how have embedded systems changed over the last 25 years? 

The short answer is: immensely. 

When you think about a security camera, Wi-Fi is usually part of the equation, unless your intended location is somewhere remote, like a barn, a construction site, or some other location where there is no Wi-Fi and no practical way to get it there.

One vendor, SEGGER Microcontroller Systems, can come up with a way to simplify the design of embedded applications, a term they’ve coined as EmApps. You can have SEGGER write the apps for you, or you can get an SDK from them and do it yourself. 

I dove way deeper into this issue with Rolf Segger, the Founder of SEGGER, on this week’s Embedded Executives podcast. We get into where TinyML should be deployed and why, and where it shouldn’t.

Embedded devices get updates and patches on a regular basis. Those updates consist of bug fixes, security vulnerability updates, and lots of other stuff. But at some point, you’re going to run out of memory space for these patches, especially if your system has been in the field for years or even decades. 

So what do you do? 

That's not an easy question to answer, but there is an answer, and that’s what I got from Steve Hanna, a Distinguished Engineer from Infineon, on this week’s Embedded Executives podcast.

Do you know the difference between a microcontroller and a microprocessor, aka an MCU and an MPU? 

Many people use the terms interchangeably, but that would be inaccurate. But more importantly, when do you use one versus the other in your design? 

That’s the topic Rich and Vin tackle in this episode of DevTalk with Rich and Vin, using the Renesas devices as examples.

NR+, aka DECT-2020 NR, is a non-cellular 5G wireless technology standard that operates without cell towers or operator subscriptions. It offers private, highly scalable mesh connectivity in the license-exempt 1.9-GHz frequency band for such applications as industrial IoT and industrial automation. 

I often hear people in our embedded community use the terms artificial intelligence and machine learning interchangeably. The two terms are certainly related, but they're not synonymous. And if you're the engineer responsible for designing an embedded system, understanding the difference can have a significant impact on the hardware you select, the software architecture you build, and ultimately, the system you deliver.

The Connectivity Standards Alliance (CSA) has a significant task in its hands. As a global body encompassing hundreds of companies, its goal is to create and promote open, universal standards in the wireless space, including (in a big way) the IoT. 

The goal sounds great, but having most of the major suppliers contributing and providing products for all corners of the globe can cause the process to become unwieldy. 

To understand how this group works, and how it is moving forward—which it is clearly doing—I spoke to Tobin Richardson, the CSA’s President and CEO, on this week’s Embedded Executives podcast.

When you’re a semiconductor vendor, it’s not always easy to show off your technology. In most cases, you have to rely on your customers and/or partners to help you out. NXP decided to eliminate that middleman, so to speak, and built its own smart home on its Austin, Texas campus. Obviously, it’s not a real home, but the simulations were close enough that I could easily understand the power of the technology the company was showcasing in each “room” of its house.

We’re seeing new life coming to the Thread standard. If you remember, Thread is part of the IEEE 802.15.4 wireless protocol and operates at very low power. But it doesn’t offer the bandwidth of Wi-Fi. Wi-Fi continues to get better, so I question whether there’s really a need for Thread. 

When Matter debuted a few years ago, it was hailed as the answer to one of the IoT industry's biggest headaches, namely interoperability. Getting smart home devices from different vendors to work together had become a frustrating experience for both manufacturers and consumers. Matter changed that conversation, at least in theory.

It seems like nobody wants a data center in their backyard. In fact, they are starting to be banned completely by some states. But we need them if we are going to move AI technology forward. So, what do we do? 

Developers know that securing an industrial Ethernet network can be tricky, and the rules continually change. We know that the AES-256 encryption algorithm can serve as a starting point, but it should not be trusted to do the job all by itself. One reason for this is that encryption must be deployed without redesigning an existing network, adding latency, or introducing another appliance that must be managed and maintained.