Open Source Hardware for IoT with Jason Kridner at OSS Seattle 2024

This is Textron tv. Everybody here, uh, Mitch Ashley here. We are at Open Source Summit in Seattle, Washington.

I'm joined by Jason Kridner, who is with, uh, BeagleBoard. Welcome. That's, that's right.

Yeah. Um, sole talk about yourself and the Beagle board. Yeah.

Yeah. So, um, I've been in the embedded systems industry for, you know, 30 plus years, um, making computers not look like computers, right. Look like something else.

Something that just kind of disappears with your, you know, your environment does things that you want it to do, rather than what somebody else wants it to do. And so we are making open source hardware, computers designed to run open source software really well so that we can get out people's way and allow them to really take their, their ideas and go as far with them as they want. Excellent.

I mean, I think we're all familiar with the right raspberry fry, right? And it's a kind of a Mm-Hmm. Um, a, a, a platform to work on.

Talk about kind of the different direction you've taken with your hardware. Um, well, the first one is that we're open source hardware, right? That really, uh, kind of stands out, right?

You can get, um, the design, um, all the source layouts that you can actually modify it and, and customize it. Um, all the parts are available through distributors and quantity one, so you can, it's really, it is possible to go on and make derivatives. I, a lot of people will kind of sit back and say, well, but yeah, how likely am I to actually go and change the hardware?

And, um, I couldn't answer that for somebody. Some people will mm-Hmm. You might not be one of those people.

Um, but if you make something really, really cool, right? You don't want to end up with that place where you kind of hit the wall. You're locked in, you're locked, locked in.

I I have a different radio I wanna put on this, or whatever it might be. And with, with BeagleBoard, you can do that, right? We have, we, we, we network really well with the open source software experts to help you, um, do the software work for the customization.

Um, but we also have a great network of people that could take our open source hardware design and help get that customized for you, help get you get into manufacturing or really help you if you're able to go, kind of go that, that, that full vertical direction yourself, help you customize it and make something that's your own rather than somebody else's Mm-Hmm. And that, that makes us very different than, than folks you mentioned. Okay.

Very good. Um, so, and we're all familiar with the IOT devices Mm-Hmm. Um, you know, the sensors, things like that.

There's proliferation of devices on the edge. Was there new, any new trends where people are starting to apply? Well, if we're, if we're looking for, you know, buzzwords, right?

org beagle y ai, and, um, the AI is sort in there because we have a neural net accelerator, right? Right. So if we're trillion operations per second accelerator, um, to do, to, to run neural net algorithms, right?

So, um, so bringing intelligence to the sensors, a lot of that's in, in vision. Um, but it could be for all types of, of sensors, right? Mm-hmm.

So time series data, um, you know, if you're monitoring motors, you're, you're, you're, you know, any, any type of things where you want to put some more intelligence and low lower latency at the edge that you can't necessarily do in the cloud. Um, but we do support a lot of connectivity in order to get some of that data into the cloud. But, but I, I think there's definitely a trend to take all of what we're learning with the, the, the, with, with big computing systems and put more and more of that down into the, the much smaller, much more affordable, right?

This, this is, uh, about a $70 computer, um, with four gigabytes of, of ram, um, and, and a lot of computing power, right? And, and I think there's a lot of trying to kind of take that smarts from the cloud and put more and more of that at the edge. Okay.

And, and we make, we make that easier to do with open source software. Um, and, and yeah, that's, that's what we do. What kind of architectural decisions have you made around like CPU?

Are you more arm risk or, um, kind of multiple paths, you know, a a little bit of all of the above? Mm-Hmm. Um, so BeagleBoard does have a line of risk five, um, computers, that's our Beagle five brand.

Um, this happens to be an arm, uh, platform. We see a lot of people like, you know, they're trying to do things today that they're not quite ready to do with risk five. So we're enabling to kind of do some of that stuff tomorrow with risk five, but today, um, helping them get their, um, with arm, but that's also not just the, the high performance application processors on arm.

This also has, um, arm cortex r fives for doing that low latency realtime control. So you've got that combination, and BeagleBoard is really known for that throughout the industry of having that combination of high throughput, um, computing at the edge, um, but also that ultra low latency Mm-Hmm. Um, and we've continued that with Beagle YAI.

Great. You know, low power is also a big issue, right? It's a, it's a huge issue, man.

You're like, did somebody tell you all about this ahead of time? They gave me the cheat sheet? No, no.

I, I worked in, I worked in the security part of this industry, digital certificates for IO OT devices and when energy sensors and all that kind of thing. Yeah. So, um, this is a, a much lower power option, right?

So it's really optimized computing. Um, we're using, um, cortex a 50 threes in here, but in a format that's, um, like with the kind of the latest of, um, uh, of where latest ecosystem kind of op platform. Um, but we're using the arm cortex a 50 threes, which is lower power than some of the other ones.

Um, more suitable for embedded systems, embedded devices, um, with, and then the neural net accelerators and the R fives really, they're, they're more op, they're very optimized for their, their task. They're very inner energy efficient for the tasks that they do, right? So it, it is very much a, a power optimized platform.

Um, and you mentioned security as well. There's, um, the ability to fuse, um, security keys into the device so that it will only run your chips, only run your software. Yeah.

Yeah. So there, there's, there's that, that provisioning in here as well. Right.

But of course, you know, our first goal is right to, to, to really enable the open source development, but I talk about removing barriers, right, of what you want to do. And sometimes that means locking it down and making sure nobody's gonna mess with it in the field. And so we do provide provisions for that.

Mm-Hmm. It sounds like a, a, a great path to go down when I'm not ready to go to an OEM and have something made for me yet. Um, but, and all the shelf thing may be not what I want and not the right technology or chips or whatever architecture I can take what you have or help kind of customize my own version of it almost back to the breadboard days.

Right. You know, kind of build my own thing. Yeah.

But, but I think, I think people are getting more and more kinda leery of just doing the, the breadboarding. Like, you know, we love the, the breadboarding, um, um, but people are getting a little bit more leery that more, more, more plug and play, um, sort of of things. So there's, yeah.

You know, we, so we're trying to work with the broadest variety of, um, of add-on hardware, right? That doesn't require you to go and make custom hardware. Um, we're doing a lot of things with the, the kernel developers to improve support for things like MIC Robust, where they have about 1500 different sensors and actuators improving the software experience for that.

So you don't even have to write the software Mm-Hmm. Um, for those, for those drivers. Um, and then we're not just doing, so today we have, you know, the single board computer that allows you to use super rapid prototyping and it's open hardware so you can kind of go where you want, but in between we're also building our own line of modules.

Um, so you can just customize the connectors and solder down the hard to do compute portion. Okay. And, and leverage that same software across those platforms.

Great. How about marrying, you know, software os to whatever the hardware that you're using, you know, not everybody's like comfortable like sorting out the driver issues and kind of working about all that stuff. Do you help figure that part out?

We do. Um, now our, our strategy working with Texas Instruments, uh, is primarily this, the, what we call upstream first. Um, and that means to en to enable the Linux kernel itself to run really, really well.

The Lin's, you know, um, version of you try to get as much of a, you know, ubiquitous Linux that'll run on most things, right. So we can, you know, just grab the, the latest kernel and all the new improvements, right. When we want to go and solve some of those, those problems, right.

We just pull the latest tree and it just works. Right. So our goal is the vanilla Linux kernel that's coming from, from Linux Tour Vaults, um, just works on our hardware.

Um, and um, to continue to kind of push in the extra bits, um, as you have a new driver that you want to get support, make sure that mm-hmm. Um, the people that you know, can do that for you. All the consultants are really comfortable with the hardware and they can add it for you quickly or you can come to us.

Very nice. Very nice. Well, it was great talking with you Jason.

And, uh, so Beagle, board Beagle, like the dog, right? That's right. org and be sure and Yep.

And I, I could have a lot of fun with that, so thanks for stopping by. Thank you. See, we haven't talked about this part of the hardware part of Open, open Hardware open source, so it's good to be, it's good to know we can do this things, you can do a lot of experimentation at home and teach kids have a lot of good problems with doing this kind of work.

So thanks for being with us. Thank you very much for having me. Bet.