Ask the Expert: What Technologies Enable Deeply Embedded Far Edge IoT?

IoT devices

battery-powered

endpoint AI

always-on

always-listening

voice command



Ambiq’s blog series, ‘Ask the Expert,’ brings you industry insight and technology expertise straight from the pro’s mouth. We reach out to our in-house experts for the big answers and turn them into bite-sized highlights for you to enjoy.

Today’s Topic: What technologies make the Internet of Things (IoT) possible? We brought in Ambiq’s Director of Technical and Solutions Marketing, Marc Miller, to share his expert knowledge.

Marc has a wealth of experience working in Silicon Valley. Four years ago, he shifted his focus to consumer products. Currently, he is focused on defining new products that stretch the limits of ultra-low power, always-on sensing, communication, and computation at the far edge.

Through Ambiq’s ultra-low-power processors, endpoint devices such as smartwatches and fitness trackers can be always connected and always listening to increase the quality of life for their users.

The Internet of Things is quickly becoming the backbone of a fully connected and data-driven world. What technologies have enabled IoT to become so widespread?

The trigger for Smart, Always-On-Voice was the Amazon® Echo. The Echo and its derivatives do a good job of locating you, extracting your voice from the surrounding noise, and communicating with the cloud to get a response. However, the Echo is a complex system with many microphones, adaptive beamforming, and an application processor. It uses a lot of power and needs to be plugged into AC power. In our home, car, work, and at play, we expect portable and mobile devices to be just as capable as those plugged into AC power.

Thankfully, the processors found in today's smart devices are able to provide greater processing capabilities, having evolved into always-on edge processing. Edge processing enables endpoint devices like a smart band or a smartwatch to handle increasingly complex computation directly on the device without sending processes to the cloud. Especially with Ambiq’s proprietary technology, the SPOT™ Platform, these endpoint devices can do more on less power.

Voice Enabled IoT Endpoint Devices

What is the SPOT™ Platform?

SPOT stands for Subthreshold Power Optimized Technology. Essentially, SPOT enables semiconductor products to operate at ultra-low voltage levels throughout the processor chip. Since power scales as voltage squared, this dramatically increases energy efficiency and battery life of the devices.

Subthreshold circuits use the leakage of "off" transistors to compute in both the digital and analog domains

How much more efficient are we talking?

Ambiq’s embedded solutions have lowered the power consumption of endpoints by a factor of ten. We’ve helped companies make IoT possible by enabling their devices to last days, weeks, or even years on end without battery recharging or replacement. And when the ultra-low-power technology can run on 1/10th of the power, all kinds of new applications and use models become possible.

Where can we see Ambiq’s SPOT technology in action?

Just about anywhere. With SPOT, Ambiq has enabled top OEMs to extend battery life and add new features in small form factor, battery-powered products including smart bands, smartwatches, hearables, remote mics, smartcards, and IoT sensors. Technologies that once seemed far-fetched, such as credit cards with dynamic CVV (card verification values) codes or fingerprint authentication are already here or are on the way. Soon, there may be a wireless personal assistant button that is the size of a quarter while having the capabilities of a smart speaker like an Amazon Echo.

With so many devices becoming connected and data flowing between them, should consumers be worried about their privacy and information security?

Believe it or not, the level of privacy and security can actually increase as smart devices become more personalized. The more processing that a device can do locally, the less will be transmitted to the Internet. Currently, endpoint processing is focused on sensing motion, and extracting keywords from noisy backgrounds.

It seems like half of the engineers in Silicon Valley are devoting work to minimizing false recognition of the keyword and undesired transmission to the cloud. The whole determination is based on whether the key word is heard, and in some cases, whether the enrolled person has said it. The process is limited to signal processing to extract the voice from the noise, and neural nets devoted only to matching the sound to the wake word. Now imagine a factor of 10 or 100 increase in AI analysis in your smartwatch or earbud, doing a deeper contextual and insightful analysis on your behalf, so that you and your devices keep personal what you want to keep personal.

What will it take for a true IoT experience to be realized?

For IoT to be truly realized, endpoint devices need to interact with each other without human instruction. To do this, they need to be intelligent—artificially intelligent. Endpoints can’t afford to run machine learning (ML) in the cloud, where there is an unlimited amount of computing power. They also have strict requirements that have prevented their ability to run inference models and mimic human intelligence. Endpoint devices need to be real-time, always-on, portable, mobile, compact, secured, and, most importantly, battery-powered. All these requirements need to be packed into a self-constrained, system solution that includes sensing, data storage, analysis, inference, and communication in a 1mW power budget.

Voice-on-SPOT enables the lowest power & highest quality always-on voice processing

What can we look forward to in IoT devices as endpoint AI technology becomes more sophisticated?

As AI technology becomes more sophisticated, endpoints will be expected to do more and run longer. This may result in an evolution of how we use devices like smart fitness trackers and wearables. If you rarely have to remove your smartwatch for charging, it is a different experience that will further enable the lifestyle where you have small, portable, personal devices embedded in and around you to make your life better and safer.

The overriding challenge for the industry is providing our personal cloud that senses us, hears us, and protects us. The Echo capabilities are the starting point, not the end point for what we want available to us whether it is a watch, earbud, remote control, or smart lock.

Looking for more resources on how ultra-low-power solutions are enabling IoT devices? Chat with Marc on always-on voice command ultra low-power solutions at mmiller@ambiq.com

Marc Miller

Written by Marc Miller

Marc’s marketing and engineering career has spanned 100W FPGAs to 1mW embedded processors. He is currently focused on stretching the limits of ultra-low power, always-on voice communication at the far edge. At Ambiq he takes a system level viewpoint with partners and developers, so hearables, wearables, and smart home devices can provide unique capabilities when every milliwatt counts. Before joining Ambiq, Marc had product management and business development roles at QuickLogic and Intel/Altera.

Subscribe Here!