RAMP

SEMI spoke with Tom Doyle, founder and CEO of Aspinity, about the challenges of packing more localized intelligence into portable Internet of Things (IoT) devices without draining their batteries. Doyle shared his views on Aspinity’s system-level approach – solve the power problems by performing machine learning in analog – ahead of his presentation at the SEMI MEMS Imaging Sensors Technology Showcase, 18 February, as part of the SEMI Technology Unites Global Summit, 15-19 February 2021, online event. Join us to meet experts from Aspinity and other key industry influencers. Registration is open. SEMI: Why is power efficiency so important for IoT devices? Doyle: Hundreds of millions of IoT devices are improving our lives at home and at work. Always on and always sensing the environment for data, these smart devices have traditionally been wall-powered and have relied on the cloud for their data processing needs, but clogged networks, as well as privacy and performance issues, have necessitated the migration to edge processing.Spanning consumer, medical and industrial, these IoT devices are becoming smaller and more portable. And a portion of them is operating remotely in hard-to-access locations. So now we are packing more functionality into the device and we are moving to battery power and the batteries need to last a long time. That is a big challenge before us, and to answer it, we need to find the most power-efficient ways to integrate always-on sensing capability into IoT devices because we cannot afford to have short battery life limit market adoption.SEMI: Why is it so challenging to deliver low-power, always-on solutions and how can sensors suppliers achieve improvements in system power? Doyle: In today’s always-on IoT devices, all sensor data – which are naturally analog – is immediately digitized at high resolution, and then it’s analyzed to determine whether a wake word has been spoken, a specific motion has been made, or some other anomaly has occurred. But since most of the data collected will not contain the information for which the device is waiting, this digitize-first approach wastes significant battery life by continuously running irrelevant data through the ADC and the digital processor.Sensors suppliers have some options to consider for reducing power. If they are satisfied achieving incremental improvements in battery life, both sensors and digital processor suppliers can continue to drive down the power of each individual component in the system. But to achieve revolutionary power savings, we must look at a more holistic system solution.The fundamental problem is that moving data through a system costs power. That is why the most efficient way to save power is to reduce the amount of data down to what’s actually important as early as possible, right at the start of the signal chain, where the physical world becomes data. If we can minimize the amount of data that require downstream processing, then we can maximize battery life.SEMI: Aspinity aims to solve the battery-life problem in IoT devices by introducing a new system architecture. Could you explain how your approach differs from digitize-first?Doyle: Aspinity’s solution, called the Reconfigurable Analog Modular Processor (RAMP), is an analog processing technology that combines analog machine learning (analogML™) and analog compression to enable accurate, ultra-low-power analog event detection and system wake-up. RAMP technology enables a new system architecture, which we call analyze-first, that allows an always-on system to spend just a little bit of analog power up front at the sensor to determine whether sensed data are relevant to the task at hand before waking the digital system for further processing. The analyze-first architecture can extend battery life by months or years over digitize-first architectures because it keeps the higher-power digital components asleep unless important data require digitization and analysis, which in some applications – such as voice-first or acoustic event detection – may occur very rarely. Aspinity RAMP voice activity detection with preroll from Aspinity on Vimeo. SEMI: Can you give us an example?Doyle: Here is a practical example of how this works: For most voice-enabled systems, such as smart speakers, voice-activated TV remotes and hearables, voice is only present 10%-20% of the time – but the digitize-first architecture on which these devices are traditionally based is digitizing 100% of the sound data captured by the microphone, even when most of that data are irrelevant and could not possibly contain a wake word.In contrast, the RAMP-based analyze-first architecture is highly efficient since it uses feature extraction and a neural network to analyze the sound at the microphone, right where it enters the device, to determine if the sound contains voice before waking the digital wake word engine. Additionally, the accuracy of most wake word engines relies not just on waking up and analyzing the wake word, but also on analyzing the 500ms of sound prior to the wake word (preroll). To support wake word engine performance, the RAMP also continuously compresses 500ms of preroll that can be stored in just 2k of memory and delivered to the wake word engine along with the voice data. So, this new analyze-first approach using RAMP technology can extend battery life by 10 times over older digitize-first designs, without sacrificing performance and accuracy.SEMI: What solutions can Aspinity bring to address the current market needs? Doyle: Aspinity offers the only analogML chip for always-on IoT devices that run on battery: the RAMP chip.The RAMP is trainable and programmable to detect many different types of sensor events directly from the raw analog sensor data. One application that benefits from a RAMP chip are devices that are always-listening for voice, for glass break or alarms, or for some other type of sound. Other examples include vibration sensors that monitor industrial equipment for predictive and preventative maintenance, and heartrate sensors that are used to detect anomalies in wearables and other biomedical applications.Aspinity just recently introduced our voice-first evaluation kit – which we will be demonstrating during the Technology Showcase at Technology Unites – to enable our customers to get first-hand experience with our RAMP-based analog voice wake-up solution. With this complete hardware and software kit, customers can experience all of the benefits of analogML and analog data compression – 10x power savings without a reduction in wake word detection accuracy –for their next generation of voice-enabled devices.SEMI: How can technology unite us? What do you expect from your participation at SEMI Technology Unites Global Summit?Doyle: I think this past year has shown us that when time gets tough – and for many of us, the COVID-19 pandemic has been one of the most difficult challenges we have faced – that innovation is critical to solving major problems. The microelectronics industry has played an important role in providing critical components for COVID-19 testing, ventilators, air-purification systems, and other equipment used in healthcare settings. COVID-19 has also accelerated the move to voice as a preferred interface to many devices in an effort to stem the spread of germs on surfaces.The biotech industry is gearing up to provide the vaccines that we hope will restore more normalcy to our daily lives. We can thank the successful collaborations between R D innovators and established companies in many different markets for the new devices and drugs now going into production.With traditional in-person conferences still on hold until the pandemic eases up, attending industry conferences with exceptional speakers presenting interesting content is more important than ever. SEMI Technology Unites Global Summit provides that opportunity, and I’m genuinely looking forward to participating.Tom Doyle, Founder and CEO of Aspinity, brings over 30 years of experience in operational excellence and executive leadership in analog and mixed-signal semiconductor technology to Aspinity. Prior to Aspinity, Tom was group director of Cadence Design Systems’ analog and mixed-signal IC business unit, where he managed the deployment of the company’s technology to the world’s foremost semiconductor companies. Previously, Tom was founder and president of the analog/mixed-signal software firm, Paragon IC solutions, where he was responsible for all operational facets of the company including sales and marketing, global partners/distributors, and engineering teams in the US and Asia. Tom holds a B.S. in Electrical Engineering from West Virginia University and an MBA from California State University, Long Beach.Serena Brischetto is senior manager of Marketing and Communications at SEMI Europe.

The seemingly simple act of commanding consumer devices by voice is a choice that nearly 118 million Americans now make every day, according to a recent report from eMarketer, the digital marketing research firm.While the voice interface is convenient for users, its implementation comes at the potential loss of individual privacy. The reason? Always-on, always-connected voice-first devices such as Amazon Alexa and Google Home require a wall plug and an internet connection to powerful cloud processors, making it possible for cloud companies — however benignly — to collect data on personal habits, location and conversation that were never intended for sharing. Move processing to the edgeTo address concerns over user privacy, device designers are attempting to do more of the audio processing within the consumer device, rather than sending users’ voices into the cloud. Moving more processing to the edge is a trend across the Internet of Things (IoT) industry, and not just for voice data but for other types of sensitive or proprietary data as well.Yet designers have realized limited success because the conventional approach to always-listening edge processing is notoriously inefficient: It digitizes and processes 100% of incoming sound data even though up to 90% of the data is irrelevant noise. This digitize-first approach wastes vast amounts of system power digitizing and analyzing the audio signal as it searches for a wake word when there isn’t even speech present, making it impractical for use in small, battery-operated devices.Workarounds don’t workTackling this power issue is critical to keeping private data secure. Unfortunately, it’s also exceptionally difficult. Design engineers have tried workarounds to decrease power consumption in an always-listening system, including duty cycling and reducing the power of each individual component in the audio signal chain that handles the data. The reality is that these kinds of approaches don’t address the root cause of the problem: too much data.To truly tackle the problem, we need to change our approach to a system solution, not a component solution. By moving to a more efficient edge architecture that intelligently minimizes the amount of data that moves through the system, we can focus the system’s energy resources on analyzing voice and not on searching for a wake word in irrelevant noise. Analyze, THEN digitize It’s time to move away from the digitize-first approach that has dominated voice wake-up device architecture since the invention of voice-first applications.Inspired by the way the human brain efficiently filters incoming information, differentiating, for example, a dog bark from a baby’s cry, an ultra-low-power analog machine learning technology is changing this paradigm. For the first time, device designers can use low-power analog machine learning to detect which data are important for further processing and analysis prior to data digitization.Leveraging an analyze-first architecture, a new analog neuromorphic semiconductor platform allows the higher-power-processing components in the system to stay asleep until voice has actually been detected, and only then does it wake them to listen for a possible wake word.Delivering a post-microphone audio chain that draws as little as 25µA of current when always-listening and collecting preroll data, this analyze-first architecture allows designers to extend battery lifetime significantly. That’s the difference between smart earbuds that run for weeks instead of hours or a battery-powered smart speaker that runs for months instead of weeks.More importantly, it’s the difference between the current always-listening devices that indiscriminately record and send all sound data to the cloud, and one that has the localized intelligence to select and send only the relevant data, reducing the user’s vulnerability to the loss of private data.Balance convenience with privacyThe trade-off between making our lives easier and keeping our personal information private is a choice that we are asked to make throughout our day in a hundred different ways. Bringing more audio processing capability to the mobile device without draining the battery is the first step toward delivering more secure voice-first solutions. But to succeed in this effort, we must shift to a bio-inspired architecture that determines which data are important and requires further processing at the earliest point in the signal chain. Once we move to the analyze-first approach, only a small fraction of the tens of zettabytes of data collected by the forthcoming generation of always-on IoT devices will require further processing in the device and in the cloud.A better balance between cloud and edge processing is a better balance between convenience and privacy, and that’s a win for everyone.About the AuthorTom Doyle is CEO and founder of Aspinity. He brings over 30 years of experience in operational excellence and executive leadership in analog and mixed-signal semiconductor technology to Aspinity. Prior to Aspinity, Tom was group director of Cadence Design Systems’ analog and mixed-signal IC business unit, where he managed the deployment of the company’s technology to the world’s foremost semiconductor companies. Previously, Tom was founder and president of the analog/mixed-signal software firm, Paragon IC solutions, where he was responsible for all operational facets of the company including sales and marketing, global partners/distributors, and engineering teams in the US and Asia. Tom holds a B.S. in Electrical Engineering from West Virginia University and an MBA from California State University, Long Beach. For more information, please visit https://www.aspinity.com/Technology.Aspinity is a member of MEMS Sensors Industry Group (MSIG), a SEMI technology community, that enables the MEMS and sensor industry to address common challenges, innovate and accelerate business results.