Erik: Paul, thanks for joining me on the podcast today.

Paul: Yeah, great to be here. Thank you, Erik.

Erik: Yeah, I'm looking forward to this update. I had your colleague, Srini Pattamatta, on in 2021. So for folks listening, that's episode 99. So there, we did a good deep dive breakdown of your business, and that gives us the opportunity today to discuss more what's new or what's coming up in the near future. So really looking forward to that. I think this is a fast-moving industry. Paul, before we kind of get into the updates, I think it will be useful if you can give everybody on the line a quick refresher on what Atmosic does. What's the value proposition, and what's the core of the tech stack?

Paul: Yeah, so Atmosic has been around now for over seven years. We're a fabless semiconductor company that is based in the middle of Silicon Valley in Campbell, California. We focus on very low power wireless technology and wireless solutions. Our founders came out of various tech companies here in the Valley, had some ideas for how we could reduce the power consumption for very low power wireless, embedded wireless devices, by a factor of three, four, up to five in some cases. They took that technology, and they decided to apply it first to Bluetooth as a very, very common wireless platform. And we are actually in our third generation of technology solutions, which we just announced in January of this year. We've added the capability to also do Thread and Matter so that we can address home IoT markets with that as well. So we're very, very far along now since the last time when Srini was on this podcast, and I'm quite excited by the progress that we've been able to make.

Erik: Could you touch quickly on the value proposition? Because if I think ultra low power, I'm thinking a sensor out in a forest, in a farm, on a container going across the sea where you don't have an energy source. But now you're mentioning home IoT where, of course, you have cables everywhere. So I guess there's two different value propositions for those, or there's maybe multiple value propositions for these different scenarios.

Paul: We're still fundamentally talking about devices that are battery operated or that we're trying to get to be able to eliminate batteries. So the value proposition one is, yes, that low power technology. The other part of the value proposition that I haven't discussed yet is the ability to take in harvested sources of energy and use those harvested sources of energy. So we're talking about things like photovoltaic cells to capture light. You can either also capture some RF energy with an antenna. There are other technologies that could capture thermal energy or vibrational or motional energy. So we have the ability. Our power management unit on the chip has been designed to be able to take in those energy sources, use that energy directly when it's there but also to manage it, to store it out to a rechargeable battery or a capacitor, and then very, very intelligently use harvested energy, then stored energy, and then as a last resort, we can have a battery hooked up. It could be a rechargeable battery, or it could just be a regular battery in the case of if we want to do something with, say, extending the battery life of a device. But we also have applications where the customer has completely eliminated the battery entirely. So from a sustainability perspective, now you can have maintenance-free kind of sensor or a switch that just you simply never have to think about replacing a battery.

Erik: Okay. So let's maybe quickly touch on those two cases, the one where there is some kind of battery but it's long life because ultra low power and then the other where the battery is out of the picture. Let me give you a scenario just so this will make it maybe more concrete for the listeners. We have a client. They do kind of, what would this be called? Well, they do building components. And so one of the things that they do is embed sensors into buildings, into structures for predictive maintenance, and they're looking for a solution that can last for 10 years without touching the battery. So it has to be able to determine is there any structural damage, was there an earthquake, anything like this, and send that data over a 10-year period. They're having a lot of trouble finding that or finding a solution that works there. So I guess this would be a scenario. In this type of scenario where you're looking at a 10-year lifespan, you're embedding this inside a building, right? So I guess there's not going to be solar or kind of those energy sources. Would this be primarily an ultra low power solution, or would there be, even in that situation, opportunities for energy harvesting?

Paul: Something embedded like that is probably going to be a challenge for energy harvesting, unless you had maybe if it was attached or close to a thermal source, like a ventilation duct or something like that where there was hot or cold air or something passing through it maybe. But chances are, a customer like that would really just be looking to take advantage of our low power consumption. So what that could mean is that now the size of that battery to make the device last the 10 years could be potentially a factor of three or four smaller than what they had originally planned to use. Or, maybe for the same size battery in that application, they'd be able to send the data more frequently or send more data and still be able to maintain that 10-year battery life. So we very frequently find we're working with customers to be able to see how we might either reduce the size of the battery and lower the size and reduce also the cost of the device. Or in some cases, we say, "Hey, now this can open up new use cases for you. Now you can send back more data, send it back more frequently than you might have done in previous devices."

Erik: Okay. Got you. Then the second scenario is the scenario where we're removing the battery from the picture. In this case, does that mean that every time that you're sending a package of data, you need the device to be simultaneously collecting the energy required to send that? Or is there some kind of maybe some other short-term mechanism for storing a small amount of energy over a small period? I'm just kind of wondering practically how this works, if maybe the energy source might be somewhat variable but then you need to align that with when the data packages are sent.

Paul: Yeah, absolutely. So most people think about batteries as really being a long-term storage device, right? You're talking about months and years of storage. Whereas something that's more transitory like a capacitor in the system, the charging, the energy from a capacitor can bleed off over the course of minutes or hours. But you do need to have some reservoir of energy to be able to meet that burst when you send out that radio signal. Because you're absolutely right. The harvesting of energy can be very intermittent. I'll give one example here of a case of a customer that has created a battery-free wall switch. So it's completely wireless. It's not hooked up to any electricity. It allows you to put the switch wherever you want it. But when you press that switch, you're actually moving a magnet through a coil, and you're creating a burst of electricity as that magnet is forced through the coil when you press the switch. That burst gets captured, if you will, in a capacitor. Then the Atmosic device can then take that energy and then use it to send out the burst of radio signals to say, hey, this button was just pressed. Then when the user presses the button again, same phenomena happens. In between all of that, there's no need for our chip or any of the rest of the system to be powered. You only really care about powering that system when the user presses the button.

Erik: Got you. Okay. Great. So that gives us a good understanding of the different scenarios. Now, coming back to this topic of what's new, you mentioned that you're recently onboarding Thread and Matter as protocols. I saw in the news here that you now have multi-protocol SOCs, which I guess would be — is it Bluetooth plus Thread plus? Do you typically package them together, or is it often going to be segmented?

Paul: Well, in this case, we packaged it together for a number of reasons. One is because the core technology that sits underneath Thread and Matter is this radio protocol called 802.15.4, which is the basis of Zigbee in a lot of commercial industrial applications. It's also the basis of RF for CE in entertainment applications like remote controls. Then also, Thread and Matter is a new standard that's been developed on top of 802.15.4 for home IoT. So in the case of the Thread, Matter standard, it actually uses Bluetooth as part of onboarding. Because Bluetooth is very convenient. You've got it in your phone, so it creates a very good connectivity platform for that. Then Thread and Matter really takes care of the interoperability with all of the other home IoT devices that you have around you.

So what we're interested when it comes to Thread and Matter is again low power, either battery-operated or potentially battery-free devices in the home. So you can think about window and door sensors. You can think about remote temperature sensors for your thermostat. You can also think about things like door locks. In those cases, your window sensor, you might be close to a window. You could put in a small PV cell, and you could use the light energy and some storage to create a completely maintenance-free kind of solution. Now, you wouldn't want to do that in a very dark part of a house that doesn't get any light. But I think in a lot of applications, that kind of small PV cell for energy harvesting can make a lot of sense similarly with like a door lock on your front door. Some folks I know who have smart door locks talk about having to plug in the USB cable to recharge the battery in their door lock every six months or a year, because that battery winds down. Something like energy harvesting can make that a maintenance-free kind of solution.

Erik: Okay. So one significant lever of innovation is on the protocol. So opening this up to more devices. The other which you've just been mentioning is energy harvesting. Over the past few years, what have you seen in terms of developments for either new methods of energy harvesting or innovations? For example, PV cells. I mean, what kind of form factor are we looking at now in order to power a device? What's the trajectory for miniaturizing that? Because I guess that's then a significant factor in terms of what use cases this is viable for.

Paul: Yeah, that's correct. So we've seen a couple of really interesting innovations in the photovoltaic cell space. One is, folks are probably very familiar thinking of photovoltaic cells as a silicon-based, on a piece of glass kind of thing. But now there are technologies that allow you to get away from silicon-based technologies, go to more organic sort of, if you will, a printed kind of thing that can be printed onto flexible materials. So it could be printed onto a piece of plastic or on the inside of a housing that will allow for sort of more creative industrial design, and also to even hide the PV cell. One of the applications that we've been very involved with is remote controls for your TV or your set-top box at home. In this case, people don't like to have a window with the PV cell exposed. So these technologies allow you to be able to literally integrate that PV cell right into the plastic so that it's not so visible.

Then separate from that, these technologies can also be more efficient by a factor of two or three than some of the older silicon-based technologies. So now you're able to again reduce the size of that PV cell to harvest a sufficient amount of energy. So we really see the confluence of our low power and some of the advances in some of the harvesting technologies to be able to enable some of these use cases where you wouldn't want to have to have some huge PV cell on your window sensor, or on your door lock, or on your remote control. Now we can bring the size of that down, and it also helps bring the cost down. It enables a better industrial design too.

Erik: Okay. So you've mentioned a few methods of energy harvesting. PV, I guess perfect where there's exposure to sunlight. Buttons are great for anything where there's kind of a human or some kind of interaction with a device that allows you to capture that physical, mechanical energy. You mentioned also thermal. So I guess if it's in a scenario where there will be a heat source, that's reliable. Aside from those three, what are the other energy sources, you see any new energy sources coming online that would enable use cases and other scenarios where these forms of energy are not available?

Paul: Yes, so the one energy source that we haven't talked about is actually wireless itself. So whether you call it RF harvesting, or some people call it wireless power transfer, you're taking an RF source, an RF transmitter. Our chip actually has a separate antenna connection, and we can attach an antenna. In general, what I like to talk about with folks is, the RF harvesting is a one to five meter or less than five meter type of technology in terms of proximity. You can't just go and stick an antenna up in the air and expect you're going to harvest a lot of energy. You generally have to be somewhat close to some kind of source. And when you do that, now you've got another potential power source that has the same sort of properties, being able to charge a capacitor or recharge a battery. In much the same way that when folks think about they wirelessly recharge their cell phones by putting it on a wireless charging, there you've got very, very close proximity. But you're still doing a wireless transfer of energy. In this case, you're doing it a little bit more at a distance, and you're just not going to get nearly as much energy.

Erik: Got you. Okay. So RF works, but it needs to be within, let's say, five meters of the source.

Paul: Yeah, five meters or less. Now, the other technology that I would really like to see more of — there are definitely some challenges with it — is should we say vibrational or more of a motion type harvesting where you can imagine just the vibration of a motor, or even you can imagine like putting a harvester inside your running shoe. Every time you pound the pavement, that motion of running will be able to harvest energy. That's been a bit more of a challenge. The harvesters there have been a little bit big, a little bit more expensive. But as humans, we're constantly in motion, and we have lots of machines around us in motion all the time. It just seems like there's a lot of energy there to be a great source to harvest. We've talked to some folks about it, and there's definitely some very niche applications for it because of the size and the cost of the harvesters. But it's a very interesting technology. I think a lot of people would love to be able to have automatic tracing, be able to trace their runs and their workouts just by putting their shoes on.

Erik: Yeah, I mean, that case I was mentioning earlier, one of the infrastructure types that they're looking at would be bridges, right? Of course, bridges, you have a lot of vibration. But then I suppose you have this technical challenge of, how to have a device that's capturing vibration energy and that is sufficiently durable to last for 10 years? Right? So that's a significant technical threshold to overcome.

Paul: Yeah, there are companies that have put vibrational harvesters on railway cars and use them for monitoring like the axle and the health of the wheels of the railway car. And so very, very niche application but a great application for something. Whenever that wheel is moving, you want to be able to see how it's doing.

(break)

If you're still listening, then you must take technology seriously. I'd like to introduce you to our case study database. Think of it as a roadmap that can help you understand which use cases and technologies are being deployed in the world today. We have catalogued more than 9,000 case studies and are adding 500 per month with detailed tagging by industry and function. Our goal is to help you make better investment decisions that are backed by data. Check it out at the link below, and please share your thoughts. We'd love to hear how we can improve it. Thank you. And back to the show.

(interview)

Erik: There's another topic. And let me know if this is not particularly relevant. But this is the topic of near space communication. So you see a lot of satellites going up. Then, of course, one of the application areas for these satellites is extending IoT connectivity to areas of the world that previously were unconnected. But of course, satellite connectivity is still expensive, right? So I would imagine that there would be an application there. Are you seeing any significant movement in terms of IoT devices with satellite connectivity? Then what is Atmosic's role in that space?

Paul: It's not a space I pay a lot of attention to. I'll explain why in a minute. But there are quite a few companies that are doing things in that space to be able to have a low power device that would be able to communicate to a satellite. It's a little bit of a challenge for us, because the primary limitation to getting the signal to that kind of greater distance is just the raw energy required in the power amplifier of the radio to just transmit that far. That's something where our technology doesn't play as well. We can improve that some, but I think we'd be waiting a little bit longer to see if there's a dominant network or a dominant application and protocol. Because, in general, they don't use standard protocols like Bluetooth or a Zigbee or something like that with those kind of satellites. They tend to be something very proprietary.

Erik: Okay. Understood.

Paul: We like standards.

Erik: One other question just out of curiosity. I mean, there's been a lot of development, just kind of ongoing development in terms of improved algorithms. How important are compression algorithms to your business? And is there much innovation in terms of that aspect of the tech stack?

Paul: We do see it get worked into things, for instance, like audio applications. Today, we don't play that much in audio space. But absolutely, the recent — I guess this goes back a couple of years — announcement of Bluetooth low energy audio. A big aid in that was the improvements that they have in the codec that were required to code and decode an audio stream on the fly. So we do have hardware acceleration and support to be able to support some of these algorithms. They tend to be very specific for a certain application. Some customers come, and they'll want to run some kind of data compression that is specific for what they're trying to do. So we have to make sure that our parts have the capability, the CPU and the memory to be able to handle that kind of process.

Erik: Got it. Okay. When we were chatting earlier, you mentioned one area that's a bit more future looking. This is the topic of TinyML or Edge AI applications. I know we're still quite early there in terms of capabilities. There's a lot of people across the value chain that are working on making this more feasible. But can you give us a little bit of an update on where we are today, and then maybe what's a reasonable timeline for when we'd expect more, let's say, low power AI to be on the edge?

Paul: Well, we are seeing things move to the edge more and more. In order to have all of these great machine learning algorithms running, they require data. So capturing more data, which we're capable of doing because of our low power consumption, adding energy harvesting into of all that. So now that you've got all of that data, what you want to be able to do — even though you've got, let's say, a tiny sensor — you want to be able to send the learnings that have come from that data back to that sensor so it can do inference. We've looked at and are actively involved with TinyML and some of the platforms that allow you to run those inference pieces down on the platform.

So you can imagine an application. I was mentioning the railway car application, where you're harvesting the vibrational energy but you're also wanting to monitor that vibration. And if that vibration gets excessive, that can be a sign that maybe you're having an axle problem or a wheel problem. And with some machine learning, then you could create a model that would maybe allow that sensor to be able to make a decision on the fly about, oh, this is a really serious bearing issue, or maybe this is a small vibrational problem, and it's transitory. It's nothing to be concerned about. Being able to update that model is something that could have a lot of value for safety or efficiency knowing when a piece of machinery is likely to fail.

Erik: If we use that example, what would be the timeline where you think that it's going to be, I'd say, not just technically but more maybe commercially viable to have ML embedded in that type of sensor using ultra low power. Because I guess you could do that today with just a large battery.

Paul: I would say it's even feasible today but it's just going to take time for the industry to latch on to some of these things. One thing I will say we haven't touched on this is our applications. We span across consumer applications, commercial, industrial. Especially on the industrial side, what we've seen is, the timelines there are extended. Because once you deploy something in industrial environment, it can take longer for you to go through the approval process, to get it designed in and deployed. Then once it's deployed, it can be there for a very long time, and it can be very difficult to — people may not necessarily want to replace what's already there. So it's a longer cycle to see that kind of change in these commercial industrial environments. That doesn't mean it's not happening. But we tend to look at a longer, three- to five-year horizon when we look at those kinds of applications. Whereas sometimes on the consumer side, you know that designs are going to be changing every 9 months, every 12 months. So there's a faster cycle with consumer applications.

Erik: Absolutely. And consumer, yeah, it's much easier to convince a segment of consumers to test out a new technology before it's maybe entirely mature. So we've discussed innovation or adoption of new protocols, also developments on energy harvesting. If you look at it from a vertical perspective, what does this mean for use cases? Are there certain use cases over the past couple of years that you've seen start to ramp up rapidly in terms of adoption?

Paul: Yeah, one of the applications and use cases that I'm sure some of your listeners have seen is the electronic shelf label. So going into a big-box store, your supermarket, electronic store and seeing these electronic labels, the price label which actually has a little bit more information about the product. It has a display. That's the same kind of technology that's in your Kindle reader. So it actually doesn't consume any power. It only consumes power when you change the display. Then you've got these shelf labels that are wirelessly connected to a central point which allows a lot of configurability, whether it's making price changes, making specification changes as you move products around the store. Being able to add sensors to those labels so you can even get a sense of how many people are passing by, or having wireless connectivity so that the customer can interact with and learn more about the product from that label as well. So there have been very, very strong rollouts across Europe of electronic shelf labels. There are some retailers. It was just recently, an announcement by Walmart that it plans to bring electronic shelf labels to all of its stores in North America. So that's a very, very exciting vertical, very high volume. It's really being positioned as the retail experience of the future, if you will, with some of the pricing and sensor technology that's going to be deployed in stores.

Erik: Yeah, that's an interesting one. I mean, my first job when I was in high school was at a grocery store, and so I know the logistical effort that goes into actually keeping all these little paper labels updated. To some extent, electronic shelf label has been around for quite a while, right? At least pilot cases. But yeah, it has had quite a long ramp up period. I suppose that's because of the challenges of just having a stable product. So I think that's interesting, that there's this set of use cases which have technically been feasible but they haven't really been economically viable until you get to the point where you can have a stable product in the field at the right price point. That's the type of innovation that we're looking at here. Are there other circumstances like that that you see right now where there's things, use cases that maybe have existed but haven't yet or are just now reaching the ramp up point?

Paul: Yeah, one thing I would just add to that about the electronic shelf labels is, because of some of the energy harvesting PV, energy harvesting technologies that I was talking about, being able to get a very small PV cell into that shelf label, that's something. So we're starting to see some momentum there. I think that's another one, where in the next few years, you're going to see folks coming to product with a completely — There will still be a rechargeable battery or something inside the shelf label. But think about it. When you're in a store and you could have hundreds or even thousands of shelf labels, you don't want to be going around replacing all those batteries. So there's a big opportunity there that we see.

We're also very excited about positioning and tracking technologies, both for consumers. I mean, everybody sees what's happening with AirTags and those kinds of applications and the technology growing there. But even more so in environments like hospitals and factories and places where you could be trying to track doctors and nurses and patients, or pieces of equipment, or product as it's moving through the factory, and adding sensors to it to be able to tell when something in the manufacturing line is maybe getting too hot or too cold. Or one of the applications that we're involved with right now is nurses and doctors in a hospital wearing a badge. Not only the tracks where they are, but that has a button on it. And so if there's a problem, they can press the button very quickly and get attention, whether it's potentially a safety issue or a health issue. So those kinds of applications are very exciting for us. They represent a great use case for us with, again, both the low power. There are people that are looking at, how can I put one of those PV cells that can be embedded right into the plastic so that, again, you never have to worry about replacing a battery in a device like that? The sustainability of some of these solutions, we love to throw out numbers about how many batteries are being thrown away or being disposed of each day. As the IoT grows and grows, whether it's with consumers or factories or hospitals, there's just going to be more and more battery usage unless we apply some of these technologies to just eliminate that.

Erik: Yeah, right. So on a sustainability perspective, we're not really looking at energy consumption. I guess that's not the critical factor. It's the wastage of batteries and then the pollution caused both in the production and disposal of them. You mentioned a stat earlier, which was the battery size could be reduced by maybe 3x if you use kind of an ultra low power solution. Is that kind of an industry standard? What are you typically looking at when a company tries to assess what kind of sustainability impact this could have on their device?

Paul: We certainly do help them look at battery size. We have a number of partners that we work with that have newer rechargeable battery technologies and form factors that can be either small or be, say, very flat, things like that. Then we can help them assess the environmental impact of, say, eliminating. Maybe it's you eliminate 10 battery changes over the course of the life of a product. And what is the carbon footprint of those batteries versus the carbon footprint of one of those small, sustainable, flexible, organic PV cells and a small rechargeable battery? Again, when you add that up over hundreds of thousands or millions of devices in some of these larger applications, you're talking about it can be on the order of tens to hundreds of tons of carbon pollution, an avoided environmental impact. So a lot of companies have sustainability goals and targets. Some companies we talk to, even from product generation to product generation, they are challenged by their sustainability groups to reduce the carbon footprint of successive generations of the product. And so we can come in and we can show how we can really help with that process.

Erik: Yeah, great. Well, you guys are riding a couple of important trends here. You're certainly well-positioned for serving the future, I'd say the needs of the future. Paul, maybe just to wrap us up, what should people be paying attention to in the next 6 to 12 months in terms of either technology, new solutions coming out of Atmosic?

Paul: I think for us, as I mentioned, the Thread and Matter home IoT space is going to be a very exciting space. One of the things that's really been holding back the IoT in the home are these different platforms, whether you're on the Apple home, or you're on an Amazon, or a Google platform. What Matter was designed to do was to cut through all of that and create a common standard that could work across all those platforms. That's got a lot of people very excited about being able to — the consumer doesn't have to worry when they go in a store. If they see the Matter label, they know they can take it home, and they know it'll work in their system. We think that's going to just open up not only more folks to the home IoT market but also more applications, people who have more sensors in the home. Other devices, they may be powered. They may be plugged in, or they may be battery-operated. It can be something as simple as a window sensor. It could be controlling the blinds in your home based on when it's hot outside, and there's sunlight streaming in and you want to be able to try to keep the interior of the house cool. So you might automatically have the blinds go down. Those kind of interactions are things that are potentially — they're great use cases for home IoT.

Erik: Yeah, fantastic. Well, Paul, thank you so much for taking the time to give us this update today. And yeah, wish you the best in building this business in the home IoT.

Paul: Oh, thank you very much, Erik. I really enjoyed talking with you.