CES 2025: An interview with Si Labs’ Daniel Cooley – EE Times

Edge compute is all the rage at CES 2025 with fabless and vertically-integrated companies, as well as OEMs, showcasing their hardware-accelerated solutions based on vendor-specific SoCs and sensors. The applications range from consumer to medical and industrial. 
At the forefront of many of these wireless solutions is WiFi’s newest iteration (WiFi 6), BLE and BLE audio for their already-established place in consumer devices. A chat with Silicon Labs CTO Daniel Cooley illuminated the company’s presence and future in IoT and the intelligent edge. 
The pre-CES big news for Si Labs was the release of the Si917W SoC—a WiFi 6 optimized device that boasts half the power consumption than competing solutions on the market. 
“WiFi is coming big for the IoT. It’s traditionally been PC first and then mobile technology for phones and it’s not been a very good technology for embedded applications that need lower power, lower cost, or just features that simply weren’t available in the WiFi standards itself,” Cooley said.
The standard introduces IoT-centric capabilities, such as target wake time, to minimize power consumption by lowering how often devices transmit and receive data (Figure 1). The demos highlighted the low power consumption of WiFi 6 combined with the power-optimized Si917W SoC.
“We built it for low power, there’s a little innovation in RF and hardware/software co-development, plus WiFi 6 is a better standard for low power. This the secret sauce of Si Labs: low-power design down to the transistors.”
Cooley mentioned how the company’s chip has abundant functionality for the embedded engineer’s application with the needed security to ease the design process in a field where new legislation is continually coming down the pipeline to establish new security measures. 
 Si Labs divested their automotive and infrastructure business to Skyworks for $2.75 billion, ultimately leading to a narrower focus on IoT with four main pillars: compute, connect, secure and smart. 
“We started with wireless where our second product was for mobile phones with frequency synthesizers. We brought in low power MCUs in 2003 through an acquisition called Signal. But it wasn’t a dedicated focus until about 2011/2012,” Cooley said. “You know, it’s very rare that you get to be pure play at scale, this is the kind of thing that is going to have 20+ years in front of it, so we optimized everything around it from the board of directors to our supply chain, to our customers and partnerships.” 
He also highlighted the large differentiating factors from new entrants.
“They can’t build what we can build because they don’t have the ammunition to go build it in dollars,” he said. “There are bigger competitors that are just not focused and are selling in lots of end markets and as a result, have less innovation. So, we’re in a middle spot of being big enough to do what we need to do to lead but small enough to hyperfocus.”
He mentioned one example of how during the pandemic chip shortage, Si Labs gained customers because “bigger competitors couldn’t give their customers chips. They had automotive contracts and had to supply the automotive supply chain instead of embedded applications.”
This very same theme seemed to occur during 2018—a year of heavy merger and acquisition activity. 
When asked about the adoption of Matter, Cooley responded, “It’s not shipping in high volume, it’s in a lot of development is the right way to put it.”
He mentioned how it is not unusual for standards adoption to progress slower than desired.
“It’s great to have very large ecosystem drivers behind it, the Googles, Amazons, Apples, Samsungs, those kinds of companies, because they’re the ones that are driving ecosystems together. They’re the ones trying to make it really friendly for an end consumer like you or me to use this stuff.”
He settled on the key differentiating factor for Matter: it being a protocol that was built to be internet-connected as opposed to the typical “pre-cloud” IoT protocol.
“There’s a certain amount of functionality that the cloud expects when it’s talking to an end node device, and there’s a certain amount of functionality that you need for operating systems and embedded devices. So, the Linux side of things is fine, but if you go down from Linux, it’s been a very wild west, fragmented market.”
Matter is seen as the unifying protocol that drives not only internet connectivity in IoT devices, but operating system adoption within them, as well. “You just don’t typically think about your light bulb running and operating system, but they are, and they will, so I’m really excited about Matter adoption.” 
CES 2025 has introduced a sea of Matter solutions, most of which consisted of Matter over WiFi or Matter over Thread.
“There’s also a third one: Ethernet. We don’t want to forget about that,” Cooley said. He continued by highlighting the fundamental differences between the two protocols. “So, thread is a great mesh network built for IoT, and WiFi is working its way there. Honestly, there are still some benefits of mesh that WiFi does not bring: self healing, range, latency.”
WiFi was built with asynchronous communications without “immediacy”—packets are simply resent if they did not get through.
“That’s not the way 15.4 meshes were built, every bit has to get through. And so, there’s some benefits on the mesh networking side that you carry forward with thread,” Cooley said, citing Apple’s decision to implement Thread in their phones and PCs. “I think it’s a complimentary technology, but it’s not one that has to break the bank. You can do it with the same associates, it’s not like bringing an exotic wireless technology into the fold, like UWB or sub-gig. It can be built, the same radio and the fact that it’s unifying across multiple physical layers is a good thing. Matter over multiple physical Ethernet PHYs, or 15.4 radios, or 802.11 radios like WiFi.”  
When asked on Si Labs’ plans for compute, Cooley replied by elaborating on the range of power consumption and processing of embedded devices.
“There’s the absolute lowest cost thing that you’re going to need, like a microphone or a camera. Then there’s the higher performance side of this, not like autonomous vehicle machine learning, more like drones.”
Currently, Si Labs has a matrix vector processor (MVP) for neural network acceleration.
“The MVP one we have in the chips today is more of the hyper-optimized one for very specific use cases—it does great on benchmarks for that, but our customers are needing higher application stuff.”
The company could have future plans to support more hardware acceleration with alternative methods, such as NPUs. 
Aalyia Shaukat, associate editor at EDN, has worked in the design publishing industry for six years. She holds a bachelor’s degree in electrical engineering from Rochester Institute of Technology and has published works in major EE journals as well as trade publications.
You must Register or Login to post a comment.
This site uses Akismet to reduce spam. Learn how your comment data is processed.
Advertisement

source

About The Author

admin

See author's posts