Did you receive this newsletter as a forward? Subscribe here
|
|
Span is tackling our outdated grid one breaker box at a time |
|
By Stacey Higginbotham |
As the urgency of the climate crisis becomes increasingly clear, technology firms are pushing harder toward solutions designed to boost electrification and the use of renewable energy. Solar panels have become more common, for example, as have smart thermostats and a focus on energy-saving services, in both homes and offices. But there are generations of aging infrastructure and regulations standing in the way.
That has not stopped companies building smart home products from trying to nudge customers to more efficient habits through connected devices, however. Last month, Google announced a way to use renewable energy to power a home through Nest, and Tesla is continuing to build electric vehicles that people want to drive. I've also talked to three or four smart home players in the last few months about energy-saving routines they could create for consumers. But the most interesting startup out there right now that's trying to smarten the energy grid is San Francisco-based Span. |
|
— The Span panel and the Span Drive vehicle charger. Together they will cost $4,000. Image courtesy of Span. |
|
|
The startup was founded in 2018. Its CEO, Arch Rao, appeared on the podcast back in January. And this week, it announced new software features and two new products designed to help consumers make their homes' energy systems smarter.
Span makes a breaker box that analyzes and manages the flow of electricity around the home based on where it's needed. I'm excited by the decision to make consumer hardware for the entire home because it allows individual consumers to upgrade an integral part of the electrical grid without the help of a utility, all of which are notoriously slow-moving. And while that may not sound like much, it's actually pretty revolutionary.
In most homes, each electrical load — such as a series of lights and outlets or a large appliance — gets one or two dedicated circuits with a set amperage. If the amperage is 20 amps per circuit and the device plugged into that outlet only needs 10 amps, there's slack in the system. Building codes are designed around this.
But if someone wants to install a device that unlike an outlet or appliance doesn't run all the time, such an electric car charger, they would need a specific number of circuits and amps to be allocated for that device. Span's panel lets its customers allocate extra amps to devices that need them, which means that as people electrify their homes, they won't have to upgrade their existing incoming electricity (which does require a call to the utility company).
I have already run into this problem in my house after installing an induction range and a heat pump in an effort to eliminate fossil fuel use. Add in the electric car charger for my car and I'm pretty much out of circuits for anything new. And my home's electrical system was actually upgraded to 200 amps a few years back.
The Span panel is not for mainstream users, though. It costs $3,500, and the company has only recently adjusted its sales efforts to be able to return calls from people eager to buy it. According to Rao, the sales process is highly personalized, both because a customer has to use an electrician that's certified to install the Span panel and because Span wants a lot of details about the home and the electricity provider. His hope is to sell tens of thousands, not millions, of units in the coming year.
In the meantime, despite its early stage, Span is launching two software features for existing panel customers and two new hardware products as it seeks to meet the demand for a more modern grid.
The first hardware product is Span Drive, a physical car-charging box for electric vehicles that will work with the Span breaker box to optimize charging times. The idea is that electric vehicles can't charge at their full capacity on home electrical networks because they can't pull enough amperage from the circuits without overwhelming them. Span Drive ensures that a car can charge quickly and pull as much as it needs by making sure the car charges only when the full amperage is available, which is especially useful when only a quick charge is needed. Span Drive will cost $500 and requires a Span panel. It will be available in April 2022.
The second product is a second Span panel with an integrated meter box. Rao told me that 40% of homes, typically older ones, in the U.S. have these integrated boxes, which expands the number of potential customers for Span's technology.
On the software side, Span will now let users sign up for a demand response program with participating utilities that lets those users decide how to reduce their home's electrical load. Most demand response programs today are tied to HVAC systems in residential homes, which can leave people surprised when their air conditioning stops cooling as much as anticipated during the hottest part of the day.
Span customers will be able to select appliances and lights to participate in demand response, although given how much electricity HVAC systems use, focusing on them is probably the most efficient way to reduce demand quickly. The other software feature will track the health of a home's wiring as well as any appliances pulling electricity from the grid. Based on fluctuations in demand over time or analysis of a swath of users, Span plans to tell customers if their HVAC compressor is working too hard or their fridge is becoming less efficient over time.
Both of these features are free, although Rao said it's possible Span will charge for other software features over time.
Obviously this is an expensive upgrade for homeowners, albeit a cheaper option than bringing a new electrical line to the home. And it won't solve the larger problems facing our aged electrical infrastructure. But it's a start, and a good model for how to think about using electricity more wisely. So I'm hopeful that Span's customer service problems associated with its backlogged sales are behind it and that more homes will get something like this installed. |
|
|
|
Waterfall vs Agile for IoT SPONSORED |
|
|
While the Agile methodology has been widely embraced in the world of web development, many IoT products today are still built using the Waterfall approach. Getting the development process right can be the difference between a product’s success and failure — so which is better? In this white paper, you’ll discover the benefits of both Agile and Waterfall development for IoT and learn how to decide what’s right for your project.
Download whitepaper
|
|
|
|
Qualcomm is researching machine learning at the edge |
|
Regular newsletter readers know that I am beyond excited about machine learning (ML) at the edge. Running algorithms on gateways — or even on sensors — instead of sending data to the cloud to be analyzed can save time, bandwidth costs, and energy, and can protect people's privacy.
So far, ML at the edge has only involved inference, the process of running incoming data against an existing model to see if it matches. Training the algorithm still takes place in the cloud. But Qualcomm has been researching ways to make the training of ML algorithms at the edge less energy-intensive, which means it could happen at the edge.
Bringing ML to edge devices means user data stays on the device, which boosts privacy; it also reduces the energy and costs associated with moving data around. It can also lead to highly personalized services. These are all good things. So what has Qualcomm discovered? |
|
— Personalization, privacy, broadening data sets, and improvements in federated learning (FL) are all reasons Qualcomm is investing in training on the edge. Image courtesy of Qualcomm. |
|
|
In an interview with me, Qualcomm's Joseph Soriaga, senior director of technology, broke down the company's research into four different categories. But first, let's talk about what it takes to train an ML model.
Training usually happens in the cloud because it requires a computer to analyze a lot of data and hold much of that data in memory while performing probabilities to assess if the data matches whatever goal the algorithm is trying to meet. So to train a model to identify cats, you have to give it a lot of pictures of cats; the computer then tries to figure out what makes a cat. As it refines its understanding, it will produce calculations that a data scientist can assess and refine further by weighting different elements of the assessment more heavily in favor of elements that make something look like a cat.
It requires a lot of computational heft, memory, and bandwidth to build a good model. The edge doesn't historically have a lot of computing power or memory available, which is why edge devices perform inference and don't learn while in operation. Soriaga and his team have come up with methods that can enable personalization and adaptation of existing models at the edge, which is a step in the right direction.
One method is called few-shot learning, which is designed for situations where a researcher wants to tweak an algorithm to better meet the needs of outliers. Soriaga offered up an example involving wake word detection. For customers who have an accent or a hard time saying a wake word, using this method to improve accuracy can boost detection rates by 30%. Because there is a limited and clear data set, and labels, it's possible to train existing models without consuming much power or computing resources.
Another method for training at the edge is continuous learning with unlabeled data. Here, an existing model gets updated with new data coming into the edge device over time. But because the data is unlabeled — and the edge data may be over-personalized — a data scientist has to be aware of those limits when trying to adapt the model.
My favorite research topic is federated device learning, where you might use the prior two methods to tweak algorithms locally and then send the tweaked models back to the cloud or share them with other edge devices. Qualcomm, for example, has explored how to identify people based on biometrics. Recognizing someone based on their face, fingerprint, or voice could involve sending all of those data points to the cloud, but it would be far more secure to have an algorithm that can be trained locally for each user.
So the trained algorithm built in the cloud might recognize how to differentiate a face, but locally, it would have to match with an individual face. That individual face data would stay private but the features that make it a face would get sent back to help adjust the initial algorithm. Then that tweaked version of the algorithm would get sent back to the edge devices where some noise would get added to the face data to ensure privacy, but also to ensure that over time the cloud-based algorithm gets better without sharing that person's data.
This approach provides large sets of face or voice data without having to scrape it from social media or photo sites without permission. Federating the learning over many devices also means data scientists can get a lot of inputs but that the raw data doesn't ever leave the device.
Finally, we also need ways to reduce the computational complexity associated with building algorithms from scratch. I'm not going to get too into depth here, because there's a lot of math, but here's where you can find more information. Broadly speaking, the solution to traditional training in the cloud is to make training on less compute-heavy devices easier.
Qualcomm researchers have decided that one way to do that is to avoid using backpropagation to figure out how to weigh certain elements when building a model. Instead, data scientists can use quantized training to reduce the complexity associated with backpropagation and use more efficient models. Qualcomm's researchers came up with something called "in-hindsight range estimation," to efficiently adapt models for edge devices. If you are keen on understanding this, then click through to the research paper. But the money statement is that using this method was as accurate as traditional training methods and resulted in a 79% reduction in memory transfer. That reduction computes to needing less memory and compute power.
This research is very exciting because training at the edge has long been the dream, but a dream that has been so hard to turn into reality. As regulations promote more privacy and security for the IoT, all while demanding reduced energy consumption, edge-based training is moving from a wish-we-had-it option to a need-to-have it option. I'm hoping R&D keeps up. |
|
|
|
Build the future with Very SPONSORED |
|
|
Develop new smart technologies or expand the capabilities of your existing products with a battle-tested and trusted IoT partner. Talk to our team today to discover how Very can amplify your current product offerings.
Get started.
|
|
|
|
Episode 344: Energy harvesting sensors are finally real |
|
This week's show kicks off with news from many of the big smart home players offering their plans for the Matter smart home protocol. First, we discuss Google's plans, before focusing on Samsung's latest announcements and then a surprise update from Eero, which is owned by Amazon. Sticking with Amazon, we also cover the news that Alexa is now employed in hospitals and senior living facilities. We cover industrial IoT sensor provider Augury's $180 million round of funding, and a new report from Palo Alto Networks on how remote working and IoT devices have compromised enterprise security before heading into some news from Amazon, Aqara, Inmarsat, and two retailers removing Chinese cameras from their shelves. |
|
— Alexa is heading to senior living facilities and hospitals. Image courtesy of Amazon. |
|
|
Our guest this week is Steve Statler, the senior vice president of marketing at Wiliot, a company that had been making Bluetooth beacons that don't require batteries. Now the company offers sensing as a service and licenses its chip technology. Statler explains the shift and discusses how Wiliot had to build up a web of relationships to make the sensing-as-a-service option possible. We also discuss how smart Bluetooth tags can create what Statler calls the demand chain to track products on an individual level and ensure supply meets demand based on reality instead of estimates. Statler also talks about how to make the tags recyclable, and what he still needs to make that happen. It's a fun interview for people who have high hopes for smart labels, and who want a glimpse of the future where items in your fridge or closet may communicate with you after you've purchased them. |
|
This week on the IoT Podcast Hotline, we answer a listener question about a switch for LIFX lighting without a neutral wire.
The hotline is brought to you by Blues Wireless. Blues Wireless provides the Notecard, a cellular and GPS-enabled device-to-cloud data system-on-a-module. Each Notecard comes packaged with 500MB of data usable over 10 years for $69 or less. Connectivity is globally available in 136+ countries. Learn more. |
|
|
|
|
LEGO inspired Volvo to make a smart, modular wheel loader |
|
|
— This is a fun story. Volvo Construction Equipment has built a 5-ton electric wheel loader that was inspired by a LEGO Technic model. Volvo and LEGO created a partnership back in 2018 to help design LEGO Technic kits. But the engineers at both companies saw the LEGOs as an inspiration and built a real wheel loader with modular parts that can be removed or added to tweak the design to manufacture it in different sizes. This isn't a vehicle that Volvo is making just yet, but it's an inspiring story about how unlikely partnerships can create innovative ideas. As for the IoT angle, the wheel-loader is a sensor-equipped robot that can be programmed to do a job and can also react to the environment in real time. Image courtesy of Volvo CE. |
|
|
|
|
|
|
News of the Week |
|
Check out the data your smart speaker collects: PCMag has created a nice resource showing the types of data that various digital assistants collect about their owners. Amazon's Alexa and Samsung's Bixby seem to know the most, with Alexa being more of a concern since it's in more homes. But I also think it's worth noting that much of the data these devices collect is designed to make certain processes easier, such as using a smart display to call a friend or paying for something using your voice. It's also worth noting that in some cases you'll be asked to opt in to sharing certain information, such as your contacts. But given how many people are unaware of what these devices know and store, these charts are a public service. (PCMag)
Qualcomm's smart city plans get a customer: Three weeks ago, I had Sanjeet Pandit, the head of Qualcomm's smart city, on the podcast, where he spoke about the need to build better broadband before we can have smart cities. He alluded to success in signing up customers for broadband networks that could then be used for smart city deployments. This week, Qualcomm said it would work with CircleGX to deploy a private LTE network using CBRS spectrum in Dallas County, Tex. The network will provide broadband to underserved users as well as the infrastructure for smart lighting and sensors that Qualcomm and its partner Zyter will provide. Eventually, Dallas County could use that infrastructure for other smart city deployments, such as traffic management or smart classrooms. One of the big questions I had for Pandit when I spoke with him was how cities pay for such infrastructure. He spoke vaguely about revenue-sharing opportunities and said we'd find out more as cities announced the networks. However, in this case, I'm still not sure. The CEO of CircleGX said that residents can use federal broadband credits to pay for the connectivity provided by the CircleGX network, and then said the Chamber of Commerce was working on discount vouchers for people who shop at local businesses. At best, the plan didn't sound fully thought out, and at worst it represents another effort to prey on poor people who need broadband but might have to give up their privacy and data in order to get it. (Qualcomm)
It was Alexa all along! This week, Amazon's Alexa Smart Properties business said it would deploy Alexa smart speakers and displays in hospitals and senior living facilities. During the pandemic, Amazon has been testing the use of Alexa in senior living facilities as a way to keep residents connected with their families. Hospitals will use Alexa differently. By letting patients call nurses to make requests using Alexa — as opposed to hitting a call button, which requires that the nurse walk to the room and then take a second trip to get and bring back whatever is needed — Alexa speakers or displays can help boost productivity. Patients can avoid having Alexa in their rooms if they'd like, and Amazon says none of the data will leave the hospitals or be stored on Amazon's cloud. So while I think some patients will shun Alexa for privacy reasons, most will accept it. After all, in a hospital setting people are sick and want to get better; they're not focused on fighting for digital privacy. But what's remarkable to me is that Alexa in this case is a Trojan horse, sneaking into patient rooms as a productivity tool that could easily become a platform for voice interaction with patient medical records. Instead of writing on a patient's chart, doctors or nurses could tell Alexa the details and have that information shared with Epic or whatever patient record system the hospital uses. So while the platform might provide a way for developers to offer skills that can inform or entertain patients, it could easily grow into something far more interesting. I'd keep an eye on this. (Amazon)
Connectivity means more consolidation in the commercial lighting business: GE Current (Daintree) is acquiring the commercial lighting business of Hubbell Inc. for $350 million. While mostly of interest to the lighting world, the consolidation here is driven by the long life of LEDs and the transition to providing lighting as a service that has arisen because connected lights also make a great home for other sensors that can help drive service revenue. As products become connected and product makers turn to services for revenue, consolidation makes a lot more sense. It allows for a larger footprint and economies of scale as well as the option to add new services. (Lighted Mag)
Smart button maker Flic has a new product coming out: If you're a fan of smart buttons (and who isn't?), then get ready for the Flic Twist, a smart button that also acts as a dimmer. The Flic Twist will launch via a Kickstarter on Nov. 2 and will ship in January for an early bird price of €79 ($92.39). (Shortcut Labs)
Get your IoT device ready for Verizon's 5G network: If your IoT device was certified to work on Verizon's network, then get excited because now it will also work on the carrier's 5G network provided it has the appropriate radios. I'm not sure how many IoT products in the wild are ready for 5G given the cost of the modules and the amount of power 5G radios require, but if you're building a latency-sensitive, big bandwidth product this is good news for you. In probably more relevant news for IoT companies is that Verizon has certified a $4, low-power module for its network, which is a pretty good price for a cellular modem to provide connectivity. (ZDNet)
Semtech launches a sensor service to protect users from RF exposure: Semtech, which makes mixed-signal chips including LoRa radios, has launched a new line of people-sensing chips called PerSe designed for smartphones and other areas where a lot of radios might be crammed into a small space. The idea is that the PerSe products will help detect people nearby (PerSe is a portmanteau of person sensing) and then regulate the nearby wireless signals to comply with safety regulations. It's possible that some devices using certain wireless bands cannot transmit data at full power because it would be considered unsafe for nearby people. But if someone isn't standing too close, then the data transmission can continue at higher power and complete the transfer faster. With easily-attenuated spectrum, such as mmWave spectrum used in 5G, safety becomes a valid concern, and optimization based on distance could help improve device performance. This sort of sensor tech might find homes on smartphones, wearables, and other personal devices worn close to the body. (Semtech)
Samsung and Amazon's Eero are betting on Matter: Sure, it's delayed, but companies are starting to promote their plans for the Matter smart home standard, with Eero and Samsung being the latest. For more, check out Kevin's article. (Stacey on IoT)
Thanks for your support. If you'd like to advertise this newsletter and our podcast, please request a media kit. |
|
|
|
|
|
|
|
|