#466 – An Interview with Ryan Cousins

Ryan Cousins: This episode is brought to you by the Amp Hour audience. After nine years of the Amp Hour, we're humbled that people still listen to the show. We also want to send a special thank you to all of the Patreon supporters who help us send mics to our guests and support our ongoing server costs. Supporting the show could take many forms, and we appreciate those who share the show with the world. We're doing our 2019 survey. Tell us what you like or don't, guest suggestions, and a bit about yourselves. We'll have a drawing for giveaways, but only survey respondents will be eligible. The survey closes on November 22nd, so get your info in before then. Thanks again for being a listener, and we hope to continue bringing you electronics discussions for a long time. This is the Amp Hour podcast. Released November 10th, 2019. Episode 466. An interview with Ryan Cousins.

Chris Gammell: Welcome to the Amp Hour. I'm Chris Gammell of Contextual Electronics. I'm Ryan Cousins, founder and CEO of Critical. Welcome, Ryan. How are you doing? I'm doing well, Chris. How about yourself?

Ryan Cousins: Good. It's been a while since we've seen each other. We used to hang out in San Francisco at the hardware didactic galactic, or hardware developer didactic galactics, I suppose.

Dave Jones: I know, I know. It's always going to be somewhat concerning when you start seeing some of the familiar faces disappear for a while. Because, I mean, you go to one hardware event around here, and you see everyone at all the same events. So it's, you know, that we can reconnect.

Ryan Cousins: It's interesting, like, the community. And it was kind of like that. You know, there weren't, like, a ton of hardware events, surprisingly, given, you know, Bay Area and stuff like that. But it would be a lot of familiar faces. And, you know, is it a smaller community than I expect? Or is it just a smaller social community than I expect, given that it's, like, in San Francisco proper?

Dave Jones: I mean, San Francisco itself is relatively active in hardware, I would say. I'm not necessarily surprised at the, you know, volume of hardware entrepreneurs, startups, communities, just given it's hardware and it's hard. So most people just aren't really interested in doing it.

Ryan Cousins: Yeah, right. Right. Given, compared to the payout and the, you know, the potential upside of a software career in the Bay Area, I'm sure it's a little bit better, right?

Dave Jones: Exactly. Yeah, yeah. You have the magic dust of software and instantly you're, like, an overnight billionaire. So that's not necessarily how it works with hardware.

Ryan Cousins: Right. Yeah, that's right. In that vein, I mean, you chose to do the hardware route in the Bay Area. So, I guess, first off, what is Critical? And then how did you get to the path of choosing to do hardware in the Bay Area?

Dave Jones: Yeah. Well, a quick background on Critical. I started the company a few years back with two co-founders. We primarily focus on safety-critical and mission-critical applications, but have also gotten a fair amount of exposure in recent years to a lot of video processing and encoding kind of stuff. We're primarily an embedded systems company. So we have our core platform. It's called Snickerdoodle. So it's a little less intimidating. But we focus on providing complete hardware, software, IP, manufacturing, all those kinds of solutions to companies and customers who are looking to develop new products in all sorts of different areas, from robotics to consumer retail applications. I mean, our actual commercial products are kind of all over the map. But, you know, we use our core platform, Snickerdoodle, to build custom systems around it for people developing these commercial products and try and help them get from prototype to production as quickly and as cost-effectively as possible. So, you know, San Francisco is a challenging place, I would say, to run a hardware business. I mean, for as much talk as there is about hardware funding in Silicon Valley, that has mostly eluded us to date for a variety of reasons. Right. You guys are in front of South.

Ryan Cousins: That's the problem.

Dave Jones: Yeah, exactly. Apparently. Yeah. Yeah. Yeah. I mean, it's been good. I mean, it's helpful to have at least some semblance of a hardware community around. You know, I don't – I'm not sure how much of our success or lack of success could be really attributed to our physical location. I mean, other than everything just generally being more expensive here, of course.

Ryan Cousins: Yeah, the DigiKey orders are the same price, but the place where you order them from, the price goes up a little bit. Exactly. That's right. Yeah. So let's talk about the Snickerdoodle real quick because that's going to be a core focus of what we talk about here. You'd mentioned video. You'd mentioned, you know, embedded software IP and stuff like that. And I think once you explain what this thing is, people understand how it could be all these things in one package.

Dave Jones: Yeah, sure. So we built this product kind of out of necessity, you could say. So, you know, my co-founders and I, we were working in primarily the medical field before this, doing a lot of system development, primarily in the area of like mechatronic systems. So real-time motor control, sensor integration, things like that. And what we found is we were doing a lot of the same stuff over and over again for different customers. So everyone would start with a clean slate. You have to architect kind of the main control and integration system for each individual project and customer. And as time went on, we just saw the same patterns over and over. So we kind of figured, you know, we saw this kind of hole in the market between the super low-end cheap stuff that you're, well, depending on the application, most likely not going to use in a commercial product like, you know, Raspberry Pis and Arduinos and stuff like that. And then on the other end of the spectrum are these pretty hardcore systems that are a lot of the time incompatible from either an actual usability standpoint or final, you know, cost of goods or whatever it is. It might be like thousands of dollars for a single development platform. And then after you're done with that, you have to figure out how you're going to turn it into a real product. So we basically set out to have this fill in that gap where we have a very versatile hardware platform that can scale very easily and that can basically get you from an idea to the market pretty as quickly as possible without having to deal with all the headaches of this super hardcore integration piece like down at the Silicon level. So we settled on an architecture that involves, well, an FPGA SOC. So it's a field programmable gate array. I can explain that a little bit, but effectively it's a software reconfigurable piece of silicon with a couple of ARM processors, ARM cores built into the system on chip. And that built around that, we have Wi-Fi connectivity, you know, memory, flash storage, some supervisory microcontroller stuff for power management and other onboard peripherals. And what we wanted to do is give people as blank of a slate as possible to build on top of while still not making them start from complete scratch, both from a hardware and software standpoint. So kind of give them a baseline saying, hey, here's all the support that integration that you don't have to spend a bunch of time on. Now you can really focus on what your value add is. So your specific application. So having a whole bunch of different inputs and outputs that are totally software customizable or reconfigurable allows you to do, you know, that kind of is what enables all these different applications from robotics, you know, sensor fusion, remote sensing, those types of applications, all the way to networking, high bandwidth applications like video processing and encoding computer vision. And it sort of runs the gamut. You're never really, we always saw this limitation with a lot of the kind of more traditional, you know, microprocessor, microcontroller architectures where you would run into this issue where, you know, you would get so far along in your system development, your product development. And then, you know, you might have either some feature creep or something that comes up that you just have to be able to accommodate that you didn't plan for. And then you might have to go back and select a completely different part or architecture and start from scratch or you throw out a lot of the work that you did. So we're trying to, you know, eliminate that burden as well.

Ryan Cousins: Yeah. So you guys pretty much are like, yeah, it seems like the MO is kind of like just over spec it from the beginning, right? You know, you're in robotics, you know, in your video, you know, in your networking, you're going to need a lot of juice. It's here's all the juice you can get right now and in a pluggable form factor. And, you know, it's flexible, it's configurable, it's that kind of thing.

Dave Jones: Exactly. And I can't think of a single project in the past where we've, we've like overestimated the amount of stuff we're going to have to connect to, to a processor, microcontroller. Every time it's like, oh yeah, we should have actually added like a reset signal here, some enable signal there.

Ryan Cousins: This client is perfectly happy from day one. What else could they possibly need? They are so content. Yeah. Yeah. I know. I know those feelings. Yeah. I mean, you're right. And, and I think the, even the, the reduced risk side, I think there's a lot of risk here. And this is something I want to talk about. There's risk in the fact that it is over spec and it is more complicated. I think you guys work against, you know, you work towards making that less so, but like the risk factor of like, hey, you throw this thing in pretty much any pin can be almost anything. Now, of course, that's never true a hundred percent, but like being able to like, so from my personal perspective, oh crap, I hooked this pin up to the, I hooked TX and RX up wrong. In FPGA, who cares? TX is RX. RX is TX. Who cares? It's no big deal. You know, that is, uh, you know, that saves you what, even, even on a fast forward spin, what three days, four days, whatever saves you thousands of dollars, saves you a ton of, of time. Uh, and that can really be valuable from a development perspective.

Dave Jones: I mean, yeah, it's, it's crazy. It's, I mean, especially when you're talking about hardware, like you're saying when there's not only, I mean, there's so many costs both that are tangible and, you know, real dollars. Uh, but also to a certain extent, intangible, the amount of time loss with new spins and whatever else lead times for parts. Uh, I mean, literally like two days ago, we worked on a new project, um, that's, you know, based on snickerdoodle and, you know, we got the new hardware in, had this whole hardware definition file put together. Uh, that basically configures all the different IO, uh, coming out from, uh, from our module to this baseboard, this custom baseboard we're working on and realized, uh, the, the wires quote unquote were hooked up wrong. Um, but I mean, within an hour it was all fixed and it all worked. So, I mean, it's, it's, uh, you know, you're never, unless you're going to get the Dremel tool out and cut like 30 traces on a board and have a, you know, a pound of magnet wire. Strapped all over the board. Then, I mean, it's just, there's just no way you're going to be able to address it.

Ryan Cousins: We're looking at you, Greg Davil. We know where you are. People who follow me on Twitter know what we're talking about. Uh, yeah. Yeah. Uh, yeah. And, and that makes sense. And, and, and then that's one, right. I think that's the other thing is like from a scaling perspective, right. So now you're, so I've been in the place where I've, okay, I messed up a board, right. I bodge it. And I've got, you know, a bunch of magnet wire going all over the place. No big deal. Right. That's one. I, I bought five. I need to make five for proofs of concept so I can get them to work together or whatever like that. Making one is, is something I don't do anymore. I make, I make a bunch. And if I mess up one, then I have to, I have to fix that one. If I mess up five, I have to fix all five. And then your time scales linearly. Right. And it's like, you have to go and bodge five and then you have a bunch of variability between them and, oh, this one's wire broke, but this one's fine. And I got to track that. It's just, yeah.

Dave Jones: Oh yeah. There's just a ton of risk. Like, you know, absolutely. And I mean, I think, and obviously when you're doing embedded system development, I mean, every software engineer needs a piece of hardware. And yeah, if, if not all those pieces of hardware have the same rework performed or, you know, you cut a trace and like, now you have, you know, some signal line going to ground and you can't see it, you know, even under a microscope because it's some internal copper layers mashed against another. I mean, there's all sorts of, there's so many different problems with that. And not to mention, you know, a lot of times if, even if it's like, say your second prototype run, um, you do want like at least some certain high level of confidence that if you're planning to go to production with your very next harbor spin, that all these issues are actually addressed. And if you're looking at a video bus that now you have to completely reroute and then like you double check it a bunch of times and like really hope that you didn't miss anything. Um, well, you could have a bunch of dead boards, you know, next time they come off the line. So being able to add that extra layer validations, uh, you know, a big time and money saver as well.

Ryan Cousins: Sure. Okay. So let's, let's pull it all forward because, so I, I've been accused or I've accused myself at least of, of, you know, kind of waiting, waiting to the end to, to put the price tag on it. Let's say right up front, how much is a snickerdoodle board, a snickerdoodle one and a snickerdoodle black?

Dave Jones: Yes. So we do have two, uh, kind of two low end and a high end product. Um, so the snickerdoodle one, uh, you can get on Mauser right now for 115 bucks. Um, the snickerdoodle black, the higher end version upgraded FPGA radio memory, all that good stuff. Uh, it's $245. Um, and then we have a couple other, uh, baseboards, uh, that the module can plug into this kind of two different use cases, sort of the lens, the bench top standalone snickerdoodle, which you can like a snickerdoodle one, for example, you can buy with just, you just plug some wires into it and start hacking some stuff together. Um, we also have what we call the pie smasher baseboard, uh, you know, all our raspberry pie that has, uh, you know, HDMI input and output, uh, a couple of gigabit ethernet ports, a bunch of USB FPGA IO. Um, this is 245 bucks as well. So that's kind of a cool, uh, an audio input and output and, and headset. So it's kind of a cool audio video networking, uh, kind of development platform. Um, this is basically extends, uh, the snickerdoodle and then there's a separate like breakout board for just, you know, kind of the same things, um, breaking out a bunch of bios. Yeah.

Ryan Cousins: And so I wanted to do this because, so I think, so right now someone just threw up their hands like, Oh my God, a raspberry pie is only 35 bucks. How is it so much? But I think because we're talking about the risk factor and the integration factors and everything else, that's actually what I wanted to talk about here. Because like, what is the real cost of implementing like Linux systems or highly flexible systems and, and then going to production with them? Right. So like, I, I imagine, and you can correct me if I'm wrong here. I imagine that your core customer is someone doing like 10 to a hundred of something, at least at the beginning. Is that, is that about right?

Dave Jones: So, I mean, so, so in terms of commercial applications, um, as from a prototyping standpoint, yes, I mean, we're pretty much, we, we start almost every single build building, you know, somewhere, but depending on the board size and a bunch of other factors between 10 and 20 units, um, we might like run a hundred PCBs and stuff, 10 or 20 of them kind of thing. Uh, the typical volumes for the applications we're working with range from say like the hundreds a year to, um, give or take like 10,000 units a year. Um, beyond that, which is like a pretty large subset of applications. Um, I mean, it's not necessarily some random commercial, uh, consumer product that you want to sell 10 million of, uh, in which case, you know, you're using a zinc or do whatever crazy custom thing you have going on. But, uh, but you know, it just given the amount of time and, you know, engineering effort it takes to actually get all of this stuff working. Uh, I mean, we can just say from experience that it's not just some, you know, you have somebody come in and say, yeah, I know, I know Linux, I'll figure out this zinc thing, whatever it's got arm on it. It's like, you just give me the, I'll design those hardware from scratch and we'll figure it out.

Ryan Cousins: But I mean, what's that, an A9? A9? Yeah, I've done those. Okay, cool. Exactly. Yeah.

Dave Jones: I mean, it's not, uh, it's not exactly that easy. I wouldn't say, um, you know, and plus there's the hardware validation piece, supply chain. It's, you have to manage, um, there's a few different vendors you're going to introduce. I mean, it's, it's, uh, there's also some cost savings you can get from going from a modular solution. It's like very dense, you know, 14 layer board and maybe your application board only has, you know, six or eight layers, um, that could like cut your piece, cut your PCB costs in half. So there's all these different kind of considerations. Uh, but I mean, I think ultimately not obviously this thing, while it can do a lot, it, I mean, I'm never going to, I'll just be lying to people if I said, this is like the right solution for every application. That doesn't exist in the first place.

Ryan Cousins: You can do this on the next iPhone guys. Come on. Exactly. Yeah. Yeah. Plugable, plugable board into your phone. Come on. It's just all you need to battery and a screen.

Dave Jones: That's right. Well, and on the, obviously on the flip side, I mean, yeah, if you just need to like flick a couple of lights on and off and hook a button and it's, uh, you know, thermistor or something up to something, then, you know, just buy a Raspberry Pi. I'm not going to sit here and tell you. I was going to say less than that, but yeah.

Chris Gammell: Okay.

Dave Jones: Well, yeah, but I mean, people just buy a Raspberry Pi because they have like five sitting around their desk or whatever. But I mean, this is, you know, if you're doing something that has, you know, some higher level of motor control, uh, you know, a lot of more sensor integration, you know, video processing, like I mentioned is a big one, um, networking capabilities. There's always, there's a very, you know, a lot of it's like a kind of industrial IOT, I guess you could call it type stuff. Um, but it's, it's kind of more like robust solution and it is designed to scale quickly from prototype to production.

Ryan Cousins: Yeah. Well, and I think, so another thing I wanted to bring up, you mentioned supply chain stuff and I think you were saying it from the, you know, sourcing. If, if I want to go out and source a bunch of zinc chips myself, you know, I can go on DigiQ or whatever, but like, that's, that's a lot of support circuitry and stuff like that. That's kind of already encapsulated here. Um, another piece though, and something I've dealt with is like, you're not supposed to buy raspberry pies like this. I mean, yeah, sure people do. And I, yeah, I know they have the module for like the, um, the compute module and stuff like that, but it's just not really what it's meant for. And, and it's a supply chain risk in its own right now. I think there's replacement boards that have been out there that are, you know, basically are matching the interface, like the 40 pin header, but, but yeah, they're not meant to be like that. And, uh, and this is meant as a plugin module. And from that perspective, it's, I think there's, there's some power in that.

Dave Jones: Yeah. No, absolutely. I mean, we, we even have like on our, on our website, uh, SOM lifecycle agreement, you know, signed by, you know, yours truly that, uh, you know, guarantees, that does show people

Ryan Cousins: love those things, man.

Dave Jones: Well, it's, it's nice. Cause like it basically, if, because a lot of these applications are industrial, I mean, even some, you know, defense applications, some of these things where, you know, I would

Ryan Cousins: 10 years. I went 20 years, man. Yeah.

Dave Jones: Yeah, absolutely. Yeah. And, and, and Xilinx knows that. I mean, Xilinx isn't Qualcomm, you know, Qualcomm, you know, throws a chipset in the garbage every year. Yeah, exactly. Uh, you know, they've been building zincs for, you know, decade plus, and they're probably gonna be building for the next 30 years because they're in a bunch of equipment that they need to be able to source for it. Yeah.

Ryan Cousins: It's been a decade already on the zinc.

Dave Jones: Uh, I'd have to look back exactly, but probably about that. Yeah.

Ryan Cousins: Um, well, how long have you guys been around? I guess that that's a good starting place.

Dave Jones: Uh, so we, we started the company about five years ago. Um, really, really dove deep into it about four years ago. So I've been essentially selling the product for four years. Um, so yeah, I mean, it's a very, very mature, uh, hardware platform and it's not, I mean, every time Xilinx tries to phase out, uh, phase out of, you know, some part, they get some massive backlash and then they end up supporting it for the next 10 years. It was kind of, kind of a nature industry kind of thing.

Ryan Cousins: Oh, got it. So like even, uh, even like the older parts, not necessarily the zinc you're saying, but like the Spartan one or Spartan two or something like that.

Dave Jones: Spartan is a perfect example. They just released, I mean, I don't even know how old the Spartan line is, is at this point, but, uh, there's release of Spartan seven last year. Um, so yeah, that part's not going anywhere.

Ryan Cousins: So you've signed this agreement. Uh, Xilinx is pretty good with this stuff. I mean, is the zinc starting to get a little long in the tooth or is it just like, like, I guess that's kind of the interesting thing about industrial stuff too. It's usually it's not super mission critical, but it sounds like this thing can still crank through a lot of video and high demand things. Anyways, it's not like it's running out of room, right?

Dave Jones: Yeah. Yeah. For, for whatever reason we've seen probably the biggest area of interest for us has been in the video space. Um, a lot of video encoding stuff and processing, uh, you know, it, it does at least to the, the part, the specific part we're using. Um, we do start to get kind of pushing up against the edge of what the, the parts capable of technically. Um, but you know, they're, they now have zinc ultra scale, which is, you know, the next, the latest and greatest thing. And of course, silence has already announced the next generation, uh, of their SOCs, which are even more hardcore, of course. Um, so, you know, ultimately, you know, we'll be, we'll be going in the direction of ultra scale. At least that's the tentative plan. Um, once we, you know, have the resources and everything to do that.

Dave Jones: Right. But, uh,

Ryan Cousins: It's not a small undertaking to be like, oh yeah, let's just upend our, uh, our platform. Exactly. We're also supporting all of our customers for 20 years.

Dave Jones: Well, yeah, and like, and for the same reason that, you know, that there is value in having a, a platform like ours, it isn't to your point, not just something you turn around and spin around on a weekend and say, oh yeah, that was a piece of cake, you know? Yeah, exactly. Exactly.

Ryan Cousins: Yeah. I, I, I imagine that if, you know, from a company perspective, if you say you scale from two boards to three, right, you effectively need to, you need to keep supporting the first two. And that means you have to hire someone, right? You have to have someone working on the new stuff and someone to still be working on the old as someone who used to work on old stuff. For sure. Uh, I did, there's, there's a job there, you know, like you got to keep things going.

Dave Jones: Yeah. And I think that's, that's definitely one of the challenges, you know, just in, in general. I mean, we kind of took a slightly, well, I don't know about slightly, but a different approach than a lot of hardware companies do. And that, you know, a lot of our, um, a lot of our kind of strategy was built around this, having a broader community of developers, um, you know, not that are necessarily self-supporting, but, uh, that can help each other to a certain extent, you know, very similar to like BeagleBone, Raspberry Pi, or Duino, all that kind of stuff. Um, but of course, you know, then there's the added layer of having commercial customers using the stuff. So it's, it's, uh, that's been one of the more challenging pieces is keeping the overall community engaged and being able to provide support for both paying and non-paying customers. Uh, I mean, there's, I go on and on about the trials and tribulations of forming an embedded, uh, development platform community. That's a, that's a undertaking in and of itself, but, um, but yeah, support's definitely a, a big, uh, underrated piece. I would say of doing something like this.

Ryan Cousins: Yeah. And I'd actually love to talk about that from the, you know, kind of the trying to get the numbers of a, you know, free community, but then the payment of the, you know, paid community, that kind of thing. Before we do that though, I'm curious, cause you said video is a big, a big piece of what people are targeting your board at. Um, what, uh, what are, what are the options if you're not going with something like the snickerdoodle? Like, like, you know, that you want flexible fabric, you know, you want to write your own encoders or, you know, you know, you're going to be using FPJ. Is like the only real step to like start on a existing platform, like a, uh, digital and board or a Xilinx dev dev board from Avnet or whatever, and then just go straight to a custom board. Is there any other midway between, or is it just you guys?

Dave Jones: Um, well, I mean, so the, like Avnet, you know, has, has modules as well. There are other zinc modules out there, of course, that, that of course, you know, as you can imagine, range, uh, from the relatively affordable to the insane, uh, in terms of pricing and all that. But, um, but I think that, yeah, it's a lot of, a lot of those do take on the form of a single board computer style, um, like the Z board, you know, it's whatever it is, $500 or something like that. Um, for, you know, the same core part, the same SOC we're using on the $250 snickerdoodle black. Um, you know, but it has, they kind of just, they kind of, it's like, as if they integrated, like, uh, more or less like what we did in a modular form. So we took this module, plugged it into what we have as a pie smasher, um, and created this, this like double board computer as opposed to a single board computer kind of thing. Um, whereas they just took one single SBC and put everything in the kitchen sink on it that you could sort of bolt to it. Um, so it's, it's kind of, it's different approaches. Of course, the challenge then going into production is, well, now you have to integrate all that hardware into a totally custom design, which is, you know, right.

Ryan Cousins: Yeah. And that's, and that's really what I'm getting at is that like, do you guys at least have a step where you could say, all right, you know, you don't want to go to a full custom design yet. No big deal. Put our headers on there. And then the, the snickerdoodle can plug into your board. You're sorry. You're the customer's base, uh, daughter, sorry, the baseboard or daughter. Words are hard today. Yeah. Yeah. Base, the baseboard here today. So you have your processing board, which is the snickerdoodle. It plugs down onto a simpler baseboard. Right. And, uh, and you basically get to skip some production steps before going to that 14 layer board, like you mentioned.

Dave Jones: Yeah, exactly. I mean, and like I said, you know, we've, we've seen a lot of customers where it makes sense to just, I mean, use a snickerdoodle in production. Um, you know, granted, like I said, if you're building a hundred thousand or something, maybe not, or, you know, we can, we can talk about that. But, uh, but I mean, you know, realistically, I mean, there's a lot, there's so much development cost in time that you have to really weigh, you know, how, what's the life cycle of your product? Um, what extra margin are you trying to squeeze out by getting rid of this module? Because you still have to buy all the parts that are on there, you know, I mean, for the most part, other than maybe some connectors and a couple of things. But, um, you know, there's, there's obviously a lot of different considerations.

Ryan Cousins: Your, your customers get access to then your higher, you know, higher volume pricing with a Xilinx or someone like that too. That's right. Which, you know, anyone who ever looked at a FPGA on a DigiKey or a Mouser before you're like, oh my God, single piece price is like $150 for a chip. What is going on here? Oh yeah, yeah, yeah. Yeah. It's pretty crazy. Yeah. It's, uh, the, the steepness of that, of the single piece price curve is, it's a little steeper on FPGAs for some reason. Yeah. Probably because just the holding cost and how often people need it is just low. You know, you don't, you don't need, most people aren't buying in a single quantities unless they're really needing it.

Dave Jones: Yeah. And there's some, some weird considerations that as well, uh, like, you know, that you can, you can buy Xilinx, uh, parts on say like DigiKey and Mouser, but, um, like in, in the case of Xilinx, uh, Avnet is really only like authorized volume distributor for their parts. Um, and part of that has to do with export compliance and some other stuff that makes it just easier to manage if you're going through a single, a single vendor. So, um, so that's like got its own, you know, layer of, I don't know, complexity, if you want to call it that, uh, that, you know, just make it, it's just another, another piece of the supply chain that just, you'd, I mean, I can tell everybody's listening, like, you don't want to deal with any of that. Uh, that's like, I mean, it's, it's like a necessary part of this job, but it's, it's probably the worst, like most, you know, soul sucking part of the entire experience.

Ryan Cousins: Well, that's actually a good segue into, so let's, let's hear about how this whole thing went for critical to company because you guys had to do all this stuff and you had to deal with it and you had to deal with crowdfunding and everything else. So how did it go? It sounds like you were, you're just a proponent of, of starting a FPGA based, uh, uh, embedded

Chris Gammell: board company.

Dave Jones: Oh yeah. Yeah. I recommend it to everyone I come across. Yeah. Yeah. Right. Uh, grandma, hello.

Ryan Cousins: I've got a new idea for you.

Dave Jones: I mean, it's, I mean, it went, you know, I'll put it that way. Um, I mean, the crowdfunding, uh, experience was an interesting one to say the least. Um, I mean, from a, from one standpoint, I mean, it's really obviously gotten us to where we are today for better or for worse.

Ryan Cousins: Um, I mean, we had, when was the crowdfunding and what did you run it on?

Dave Jones: So it was, uh, almost exactly four years ago when we first launched the campaign. Um, we ran it on crowd supply and actually they, it was, it's nice because, you know, crowd supply also can help you transition into distribution pretty well. Um, so yeah, I'd recommend checking them out for sure. If, if you're looking for, uh, if you're interested in doing some sort of, you know, open source or development platform kind of projects, um, they're, they're big in the open hardware community. Um, so, you know, they, they now stock our products. So they're, uh, manufacturer or a distribution, um, you know, center for us now, which has been, you know, obviously helpful. Uh, but the, there's really no getting around, regardless of the platform, there's no real getting around the inherent challenges of crowdfunding.

Ryan Cousins: Um, and you know, we, we haven't, we haven't talked about that at all in the amp hour ever. Yeah, I bet. Yeah.

Dave Jones: I mean, it's, it's a very, uh, it can be a very frustrating and disheartening experience, you know, not, not even necessarily from say like a, you know, you, you go and like, you go out with your, your sign on the street and start, you know, and then you just get annoyed when nobody's just flooding through your door trying to, trying to get your product. I mean, that's, that's just inherent to any business, but, um, but it can be, it's really tough, I think, because, you know, we had a pretty good feel being people who actually would use our product to begin with. We had a pretty good feel for the types of people in the audience that would be interested in something like this. Um, but just the, and probably rightfully so the massive amount of skepticism that people have, uh, for crowdfunding projects, because people have seen so many of these things fail. So many of these money, you know, just thrown down the drain. Um, I mean, honestly, like if I had to be perfectly honest, our project should have failed. Uh, and I mean, I guess, I guess you could say in some cases, it's sort of, I mean, I guess you could say in some cases it kind of did other than the fact that, you know, now we, you know, sell or sell them through, uh, distribution and all that kind of say we have a, you know, functional piece of hardware, a few pieces of hardware and community and some commercial customers and whatever else. So, I mean, kind of like failed our way to success somehow, but, uh, but, but it's, it's interesting, you know, the, I was, I was honestly surprised at how much, you know, negativity there was, um, just in general around the whole concept of crowdfunding campaign. And as a creator, you're really putting yourself out there. When you launch a campaign, you're taking all of the risk. I mean, the incremental risk for somebody to spend 50 bucks or a hundred bucks for something that, you know, may or may not ever show up. Obviously you don't want to go through, you know, life spending your money on a bunch of random stuff like that. I mean, some people do, I guess, but, uh, but I mean, if a lot of disposable income, if like, you know, you have a 20% chance of every product you're buying, she's showing up at your door. Right.

Chris Gammell: Right.

Dave Jones: But in it, but I was, I was just surprised that, uh, you know, that risk is just so much lower than the personal, professional, financial, technical, whatever, all these different risks that you're taking on as a creator, um, you know, to ultimately, obviously, you know, you're getting some hopefully long-term benefit out of that yourself, but providing something new for, for creators like that to use. Um, I don't know. I was, I was like surprised and frustrated. I'm sure you can go back through my various message, message board, you know, postings and different articles and stuff from four years ago and, and probably hear the, uh, the anger being typed, like through the keyboard under the screen. Just like why?

Ryan Cousins: I was just actually, I was basically clicking, curiously searching through the Google right now. I was like, Oh shit, were we, were we, uh, were we negative about this? And then actually, no, I found, I found a positive thing. I actually, when Clint Cole was on the show, uh, back, yeah, this is a long time ago now is at three Oh three, three Oh two rather. He was the, you know, the former president of digital and I actually mentioned the, the, the snickerdoodle to him and, and the notes that I put in the, uh, the show note was that, uh, uh, why care about FPGAs in general? Most people don't understand how they could be useful. Chris mentioned the snickerdoodle enabling high, high end systems. And basically that, you know, we, we were, you know, commiserating that people don't really get it. You don't get FPGAs most of the time, you know, like they, they see like, you can maybe tell them about the applications and like interested in drones and video and all this other stuff. And then you tell them like, Oh yeah. And you can do Verilog and you can, you know, compile systems and you can pay 250 bucks for a circuit board. And people are like, Oh, wait a second. I thought I was going to be able to, uh, make a media center. That's what I was doing with my Raspberry Pi. And, uh, yeah, it's, it's a lot of, it's a lot of frigging work, you know, like these are not simple systems.

Dave Jones: Yeah. And I mean, you, you definitely hit the nail on the head. I mean, the, uh, an FPGA is versatility, uh, is also what makes it so challenging to convey what the hell people do with it. I mean, cause when you, when I just like put, you know, some widget in front of you and tell you this can do anything you want it to do. Like that doesn't, that like makes you just more confused.

Ryan Cousins: And if you've never seen it before, you know, so it's, well, I mean, and engineers love, uh, constraint systems and, and with an unbounded system like that, it becomes difficult to envision, you know, like, okay, well, what do I, yeah, what do I do with it? I don't, I don't get it.

Dave Jones: That's right. Yeah. And honestly, we've, we, like, if we didn't have the Pi Smasher, I don't know if we'd have any customers. It's kind of a weird thing because we, so many people have looked at that and said, like, this isn't really what we're, we're not trying to like replicate this, like what you've built obviously, but this is like, this at least shows us what we can kind of connect to this thing and what we can do with it. So, oh, you have video input and output. Oh, I have this like video encoding box that I want to build, or, you know, you have a gigabit ethernet connections to this thing. Yeah. I want to run like some gigabit ethernet into like some programmable logic, do some like, you know, packet reordering or sniffing or whatever you're doing and create that and put that into some industrial application or whatever. So it's, it's really, it's like once you start to get closer and closer to the end applications, people of course start to get it more and more, but doing that with some, some random development platform isn't necessarily always easy.

Ryan Cousins: You know, I've always wondered, I have no hard data here as well, but like, you know, I mentioned, I mentioned the people doing a media server. I am one of those people. I thought I was going to do a media server with like a Raspberry Pi or something like that. How many are sitting in drawers all over the world? I definitely mentioned this before. You know, it's like, I think that is a lot of what built the pyramid of, of, uh, you know, supporting hardware and supporting interesting, interesting things, you know, that enable other, other things in the future.

Dave Jones: So yeah, I mean, God bless them, you know, a hundred percent. I mean, I mean, the bottom line is like, this stuff is hard. Like, I mean, I don't care if you're using an FPGA or not. I mean, you know, maybe writing a few lines of Python code is something pretty much anyone can get started with in like a couple hours, but I mean, building like commercial, commercially viable products or maybe not even commercially viable, but something that's like relatively sophisticated.

Ryan Cousins: Yeah. Just high complexity, uh, industrial products or high complexity robotics and stuff like that. Yeah. It's, yeah, it's, it's a, it's a big deal. Yeah. So, and I think that that is hard because then you were, you were saying that you were trying, you, you and your, your co-founders understood the struggle and what problem you were solving, right. In the sourcing and the, you know, the getting hardware in front of people, but it's, it is hard to convey that if someone, so if you have a hundred people and none of them have, you know, one of them has ever taken a thing to production and five of them know what this board is. And a hundred, you know, a hundred of them are interested at least marginally in what the application is. It's like, you know, it's hard to say to someone, well, in about three years when you really want to take this to production, it's going to suck. And we're saving you that upfront.

Dave Jones: Well, no, I mean, it's, it is true. It's, it's funny because, you know, you mentioned some of this, uh, some of these issues around FPGAs, which have been present ever since FPGAs were invented, which is they're, they're, they're, you know, extremely open-ended and pretty complex to use. And of course, one of our big visions with Snickerdoodle was eliminating a lot of that, uh, a lot of that, you know, no, nobody wants to sit around. I mean, well, I shouldn't say nobody. I don't want to like, I mean, we love FPGA engineers, so a lot of them just love writing VHDL and Barilog, like, you know, eight hours a day, which is great. Uh, but the vast majority of software developers don't know what VHDL stands for, you know, and they don't care to really learn what it stands for. So it's, uh, let alone learning a completely new paradigm and, and, you know, language and all the rest of that. So I think, you know, part of our vision was to eliminate a lot of, a huge part of that barrier to entry, um, and make the actual integration of these programmable logic systems much easier, like much more kind of, uh, um, friendly to your average software systems developer. Um, you know, of course, well-played, well-laid plans kind of thing. Uh, you know, we are, our, our grand, uh, our ideas are grants or basically proven some traction with a crowdfunding campaign and going and raising a bunch of money so we could build out some big platform for making everyone's lives easier. Uh, you know, didn't exactly go according to plan, but you know, do what you can.

Ryan Cousins: Well, but now you don't have to deal with VCs all the time either. That's true. Yeah. That's what I'm saying. Yeah. Right. I mean, so obviously there's a value, we kind of talked to the value add on the supply side of things. You know, obviously there's a module that someone could buy off the shelf, design a baseboard, plug it into it, get a lot of flexible processing in there. That sounds, that sounds great. That that's definitely useful. You've mentioned, you know, platform side of things though. And so what does that look like? I mean, is it, are people using the Xilinx tools and just kind of building the system like anyone would with a normal dev board? Or is there some other thing that critical and with a snickerdoodle is, is like, there's like a software layer that's, that you said you're trying to make easier.

Dave Jones: Yeah. So right, right. We, out of the gates, I mean, and this is another one of those things where I can't say that we would do it the same way a second time. But one of the, one of the major things that we did from day one was be able to support Ubuntu Linux out of the box. So, you know, you get a, you know, fully supported FPGA platform that's running Ubuntu. And then you can, of course, build on top of that using the Xilinx tools and all the rest of it. You know, what we're, what we're working on is getting the actual IP integration side, much more streamlined and simplified for people.

Ryan Cousins: Um, and, and in addition to that, like a app store for digital, uh, or rather for, for hardware layer type things.

Dave Jones: Effectively. Yeah. I mean, it's, it's, uh, it's definitely, that's one of the biggest pain points obviously you see is even if there's commercial IP available for a specific thing you want to do, um, getting that, well, a buying that is usually extremely expensive. And like, Oh yeah.

Ryan Cousins: I mean, you can, you're talking to lawyers and you're signing very big contracts. Oh, it's totally insane. Yeah.

Dave Jones: I'd say, I mean, you're, you're, I mean, minimum, I mean, there's a few pieces of IP for a few hundred bucks that you can buy, but like generally speaking, you're like five to six figures, you know, and not counting any of that legal overhead or anything. Um, and then you actually have to get whatever it is to work on your hardware, which is like, that's right. It might cost you at the same amount or double or quadruple what you paid for the IP in the first place, you know? Um, let's talk about building a custom Yocto systems. Oh yeah.

Chris Gammell: It's, it's ridiculous.

Dave Jones: Yeah, exactly. Cause then you have the whole like operating system integration side of that thing. I mean, the whole, I mean, that's ultimately like, you know, we've, we've kind of chipped away at that already just in terms of making, getting something usable out of the box. You know, you have this supported thing. It's like a closed system, you know, you kind of know what you're getting as opposed to just building everything from scratch, which is just a nightmare. Um, right.

Ryan Cousins: But you know, yeah, that is the thing that I always think about is like, okay, so I've great. I've, I have written, I have written some awesome Verilog. I've got it tested. I've, you know, simulated, I got it installed. It's hooked into my bus and it's talking to my processor and everything's great. It's like, Oh, I gotta go write a driver now. Like Linux doesn't know what that is. You know, like it's, you gotta go like write some code. Now you've got a processor. It's got to talk to the hardware. Like that's the thing.

Dave Jones: Yeah. No, no, no. Linux has been like, I mean, it's, it's, you kind of, of course, like, uh, as with everything, there's pros and cons, but the Linux has been probably the biggest drain on our resources of anything we've ever done. I mean, it's, it's like, it was like basically getting Linux to run properly on this thing, like is, was more effort than like probably designing the hardware was. Um, I mean, it's just, I mean, like in hindsight, of course we, we wanted to provide, uh, some, like a platform that would be more familiar to your kind of average software developer, which is how we eventually settled on Ubuntu. Um, you know, there's all sorts of, you know, lot Yocto or Petal Linux and all these other things that are probably more readily supported by Xilinx, but like nobody cares about any of those really. I mean, I mean, to just be perfectly honest about it, like. Yeah.

Ryan Cousins: I mean, if you're, if you're trying to target software people and you want to like apt, get install something, it's like, yeah, you're not doing that in Yocto. You're going to, you know, fetch the things and build all your software yourself. Exactly. You got to go. Yeah. Yeah.

Dave Jones: Which is, you know, it's good and bad. I mean, now we have this super familiar platform, but now you're maintaining Ubuntu on a, on an embedded platform. That's right. It's just a nightmare. Every six months. Yeah, exactly. Yeah.

Ryan Cousins: April and October. You're, you're doing more work. Oh yeah. For sure. Not before that, you know? Yeah. So the Linux side of things, what, I mean, what, what is the hard part of that? And, and, and what are other systems? I wanted to compare and contrast again for like what you guys are offering versus what other, other things that are out there. Right. So like you'd mentioned the Z board and things like that is Z, is the Z board running Linux or is it more bare metal or like what's, what are available to people if they're like, Oh, I need flexible fabric. Is it the go build Yocto?

Dave Jones: Um, I, generally speaking, that's the impression I get, uh, you there. So the, the zinc book, that's literally the name of it. Um, it was originally written for, uh, one of these admit boards, but it's of course applicable to any zinc platform. Um, is a really good resource. Uh, you know, we've, we've, we've interacted quite a bit with the, some of the professors at university of Strathclyde originally wrote it. Um, and you know, so that's kind of a, uh, uh, an interesting place. They have a lot of good resources for doing like bare metal stuff. Um, which is like, you want something to actually work and you know, you're not going to sit and, you know, the rebuilding device trees and kernel modules and like drivers and like this 600 other things that you need to integrate with Linux. Um, you know, this isn't supported anymore. Now you have a different version of this. It doesn't work with that, you know?

Chris Gammell: Yeah.

Dave Jones: Uh, so, I mean, it's just so, so that, uh, that's a really, you know, that's an interesting approach. Um, of course that doesn't necessarily work for everything. Uh, you know, some there, I don't want to like just totally poo-poo Linux and make it sound like it has zero value because once you get into your wireless and networking and stuff like that.

Ryan Cousins: Um, yeah, just displays. I mean, like displays are crazy. If you're, it's like, oh, you want to go write a, you know, like go good, make a custom, you know, QT based display, you know, free R toss or something like that. Yeah. Which, you know, it happens, but it's not easy.

Dave Jones: That's right. Yeah. And, and it's, it's pretty hard to justify. I, it's much easier for like a company, so to speak, to justify that kind of stuff because there's like a commercial demand for it, or at least ideally. So somebody, IE, somebody's paying for it, you know? Um, it's, you know, what the, part of the, the, you know, idea, at least with Linux is, oh, this is all just, you know, free open source, you know, we're going to get this thing working and in a hackable state.

Ryan Cousins: And in theory, it just works. Now, obviously that is a very big asterisk on the end of that theory, but like, but like, I think from a, from a, you know, how you've talked about your customers, especially, you know, starting from a team of one, if I think about myself, how I would go and design an FPGA board. I, first off, I have no idea where to start, but if I went and grabbed like a snickerdoodle, I could design a baseboard for it, I think. Yeah. And I bet I could, I bet I could go and roll that into a, you know, in a system that has, you know, you know, I need Linux, I need a display. I could put all that stuff. I could hook that into the connector bus. Yeah. And then, and then start to drive it with this device tree that, or to start with the Linux based thing that has a device tree that needs to talk to everything, blah, blah, blah.

Dave Jones: Yeah, exactly. When you're starting, when you're starting from an already supported platform, you know, in some state, like, I mean, you just, you saved yourself from going from having no idea what to do to, okay, well, I can look it up, you know, some, on some random Linux forums and figure out how to actually get this thing to work. Yeah.

Ryan Cousins: Right. You're not out in the dark, like reading man pages, you know? That's right.

Dave Jones: Yeah. And maybe all you have to do to build like your, your little, like a, your little rover drone or whatever it is that you're, you're doing is just the electrical engineering side of that. It's like, you know, you need some power supply stuff, some, you know, some connectors and some other random things that are going to be bolted to whatever baseboard that you could just like.

Ryan Cousins: Or, or, you know, you call a great, you know, uh, electrical engineering, uh, consultant like myself and, uh, sorry, I just had to go plug in there. But yeah, I mean like a software person or software, you know, focused team, you know, you could do the hardware once effectively and that's, and then the value add for that team is adding software layers, adding more and more features on the software side. Yeah. And so it really enables like software people, like you're saying. For sure. Yeah. Yeah. Yeah.

Dave Jones: Yeah. And I guess to just follow onto that one, one thing that I can't, it's kind of hard to figure out how much interest that, or how much this is really being used. I think it's pretty big in education, but there's a, uh, there's a kind of a, well, relatively new framework, um, called pink. It's like P it's like zinc Z Y N Q, but with a P P O N Q. That's it's like a Python for FPGAs generally speaking. Um, which is kind of interesting. Yeah. Like I said, they use it a lot for education just cause it makes it, it kind of lowers that barrier to entry.

Ryan Cousins: Um, it's not something that we think you have a 12 week course or whatever. You're like, okay, well we're going to start at a transistor level. We need to build a Linux based system on custom hardware. Yeah. Yeah.

Dave Jones: Usually that's spread over like four years of education, you know, rather than like 14 weeks or whatever, but no, I mean, it's, it's a, yeah.

Ryan Cousins: And then someone's going like, and I can't have, I can't have spaces in Python. Is that, is that right? Yeah, exactly.

Dave Jones: Yeah. So it's, so there's some interesting frameworks like that that people are going to come up with, but, um, yeah, I mean, it's none of it's, none of it's easy. You just try and you try and abstract it as much as you can without extracting it too much, because then you like take, if you abstract it too much, then you take away the whole benefit to having this, this reconfigurable hardware in the first place, you know?

Ryan Cousins: Right. Exactly. At that point, why not just, you know, build it on just any other piece of any other arm processor, that kind of thing. So, yeah, well, let's, let's, let's abstract a little bit here, um, from like what you were saying, uh, let's hear some of your customer story, you know, tell us who the customers are or, you know, what their specific things are, but how are they using it in that custom way? Because I think that's another thing, like we're trying to make something, we're trying to talk about, you know, a thing that is pretty abstract on its own, but we'd like to take it back to, uh, I wanted to abstract the customers, but not what actually what they're doing. You know, like, you know what I mean? Like, uh, what, what are people using this for and how are they using the custom stuff specifically?

Dave Jones: Yes. So that's a good question. I mean, one of the, one of the things we wanted to make sure that we did was like, I was kind of saying earlier is give all our customers as blank of a slate as possible without it being totally blank. So, you know, on the module we have, you know, wifi, we have Bluetooth, um, we have a couple of buttons and things like that on there, LEDs, uh, that are tied to this microcontroller for power, power, uh, management, things like that. So we have a few components on the board that allow us to give the end user customer, um, full access to every single FPGA pin you could have access to. So, uh, which is really nice, um, in terms of flexibility, uh, for any custom application. So to give you a kind of an example, one of the early projects we worked on, um, was actually a super simple, uh, baseboard, um, well, relatively simple anyway. Uh, if you're, if you're a hardware electrical engineer. Um, so we, uh, helped this customer design a, uh, a custom baseboard that Snickerdoodle plugged into that was for, uh, an industrial six axis, uh, robotic arm. So, um, there was just some basic power in front ends and sensor feedback and stuff like that. They would, they're all mapped to various FPGA IO. So we were doing real-time control. Uh, and that was sort of an interesting one because there's two arm cores, uh, on board. We're doing some of the more like hardcore signal processing and stuff like that in the FPGA. Um, and that's, that was sent over to, uh, to one core where we were running, um, I gave it a bare metal or an RTOS, uh, but, you know, doing the real-time control loops. And then in the other core, we had Linux running. So you could run say Ross and do robot operating system and then, you know, do, uh, some of the path planning and some of these higher level, um, kind of control things. So you get both the, the real-time, you know, safety critical element of that, uh, in addition to the kind of user level use, uh, UI UX kind of stuff that makes it easier to, um, kind of integrate.

Chris Gammell: Right.

Dave Jones: Um, you know, we've done stuff like that. We've done things with, uh, systems that are, uh, they're, well, defense related. Uh, applications are doing some interesting things with, um, sensing, uh, like remote sensing where usually people, when people think of FPGAs, uh, they think of like high power consumption. But what's interesting is, uh, FPGAs are actually like orders of magnitude, uh, or can be orders of magnitude more computationally efficient than something like a CPU or GPU. Um, yeah, yeah, yeah.

Ryan Cousins: So, and yeah, cause GPUs are just scaling. It's just going parallel, but they're, you got a ton of them and they're still going in a linear fashion, right? Instead of making it a parallel actual algorithm.

Dave Jones: Exactly. And like, you could selectively use certain percentages of the fabric. You turn off pieces, parts of the fabric you're not using. So that that's literally sitting idle while you're doing work in another area. So it's, there's all these different, uh, architectures for that, but because of this microcontroller we have on board, um, we're able to, uh, put this pretty high powered FPGA module and like deep sleep mode. So basically turning everything effectively off, um, with a couple different external triggers or internal timers or whatever, however you want to set that up and have the board stand by at, you know, sub 10 milliwatts kind of thing, um, until an event happens. And then when that happens, it triggers this whole suite of sensors to this hardcore signal processing. Um, while that event's being triggered. And then once the activity stops, put the thing back into a deep sleep. So you could have this thing sitting in the field for weeks at a time or running on a little, uh, little like, uh, solar panels or battery or whatever. Um, so there's some interesting, like kind of these, these high performance, but low power applications, um, you know, that we have a pretty wide input range. Like, so you can use like a single cell lithium, uh, ion battery or lithium polymer battery that you can plug in there. And so there's, there's, uh, there's also stuff that's a little bit more consumer facing, um, working with a company on, um, building a, uh, a pinball accessory. Um, the base, he takes these old pinball machines and, um, brings them into the 21st century. It probably can't go too far into the description of what that exactly is doing, but you probably make some guesses.

Ryan Cousins: But that is a lot of inputs, right? Like a lot of those have like discrete components and, you know, each light is a light and each switch is a switch and you can't really gang them to get, you know, you can use expanders and stuff, but you can also hook them into FPGA fabric and stuff. Yeah.

Dave Jones: A hundred percent. And I mean, you think of like displays, they have a lot of these like dot matrix displays. It might have like, you know, 20, 20 or so signals coming in and doing all sorts of stuff. So, um, yeah. And you know, smart.

Ryan Cousins: It's old enough too. And you can't, and you can't get the display driver anymore either. Now you just drive it yourself. Oh yeah. Yeah. Yeah.

Dave Jones: I mean, none of those guys are going to support you. I mean, they are either all out of business. So they've all been like consolidated into like, you know, two basic pinball companies and they have their own proprietary stuff. But, you know, if you can, if you can like put, you know, your wires in between theirs and get some information on that, uh, you know, then you can do stuff like that as well. So I think, you know, there's, there's like smart retail applications, you know, a bunch of, it involves a bunch of different camera streams being encoded and, um, you know, in addition to some sensors and stuff like that. So it's, I mean, there's a, again, it's, it's sort of a good and bad thing. I mean, because you can just do whatever you want with this stuff. Um, it's very open-ended, uh, so it's, it can be kind of difficult to convey the specific value that a customer, a particular kind of customer could get or what the different use cases are. Um, but you know, it usually we kind of categorize these things into two main areas. It's either highly mechatronic. So you've got a whole bunch of motors, sensors, uh, actuators, whatever it is, uh, UI stuff, um, that you're integrating together, or it's some sort of high bandwidth application, which is usually like audio, video, networking, things like that. Um, and some of the, sometimes those two things come together, but those are. Those are kind of the two typical applications.

Ryan Cousins: You're talking, talking about a dancing robot, aren't you? That's right. The ultimate application. Yeah.

Dave Jones: Yeah.

Ryan Cousins: Yeah. No, I think that's, that's, you know, the high bandwidth one is always the one that I point to because like when I try and explain FPJs to be, I used to work on FPJs way back in the day, like Oh four, I go five and my college days and not very successfully either, but I always love them from that perspective. And like the idea of like data streams moving through a system like that, that to me is always the way to try and explain FPJs as a value, uh, a value ad value valuable thing, just because like, you know, instead of a filter going into an ALU and like chunking around and like continuously doing a, you know, a multiply cycle or something like that. Now it's doing that in fact, obviously there's, you know, there's a max that are inside of FPJs too, but like the idea that it's like a streaming, a streaming thing. That's always the applications that I think really target, uh, target FPJs because you can build so many things in line in the data path that really optimize for that kind of thing.

Dave Jones: For sure. Yeah. Yeah.

Ryan Cousins: So, uh, okay. So then, so then of these customers, right, you gave us a smattering of the customers of the overall, and that's, that's fine. I think that was a good, good overview. How many of them are using the Linux side of things, using the Ubuntu side of things specifically? Like, are they all using it? Some of them using it? None of them using it? And specifically on the customer, sorry, the, uh, the commercial customers, the ones that are making money off this kind of stuff.

Dave Jones: Yeah. I would say it's a mix. Um, I would say most of them are using it at least to some extent. Uh, some of them are using it a lot. Uh, and then others are either not using Linux at all or using it very, very sparingly. Um, like maybe just using it. Yeah.

Ryan Cousins: Troubleshooting, displaying, stuff like that. Yeah.

Dave Jones: Displaying or maybe like some network related stuff, uh, running the network stack, like TCP IP, TCP, TCP IP stack on there or whatever. Uh, so it's, it's, that's a mix. It's, it's almost, well, I would say more customers are using, using it than not. Um, but not, I mean, I definitely wouldn't say it's necessarily because that's the best, uh, way to do it.

Ryan Cousins: Um, you know, there's not the main feature that they, they came to the platform for either.

Dave Jones: It sounds like, well, that's, that's true as well. Yeah. I mean, there are a lot of, a lot of these customers who are using Linux that I would, you know, if we had a different solution for that, which we would like to have and are working on, um, there, I would, I would say don't use Linux, use this other thing, especially when you're talking about some of these like safety critical, high reliability systems. Uh, there's a lot of things about Linux that do not lend themselves to, or does not lend itself to safety critical anything.

Ryan Cousins: Um, so even you're saying, even if there's a second processor that's monitoring and doing all of the safety stuff.

Dave Jones: Yeah. I mean, well, that's, that's like as, as out of a good of a, uh, uh, compromise, if you want to call it that, that's going to do that. I mean, you're not going to like be running your real time motor control loops in Linux.

Ryan Cousins: Uh, well, I mean, you could, I'm not going to be running real time, anything in Linux.

Dave Jones: That's right. Yeah. Yeah. Even like real time Linux, like we don't waste our time with, cause that's like, it's not, I mean, it's not really real time, but, but like, but I mean, you know, that's, it's, uh, of course you kind of get back to the whole, you know, who's, what kind of engineers and what kind of developers are going to be working on the thing. I mean, if all the engineers you have on staff or Linux engineers, then, you know, well, you can't just like not, you're not going to not use Linux at that point.

Chris Gammell: Right. So what about Windows CE? Does anyone think of a Windows CE?

Dave Jones: Yeah. Yeah. We've definitely worked on some previous projects. Like we made sure to like select the platform that doesn't support Windows anything. Cause we're, yeah.

Chris Gammell: But that's just not going to fly. Poor Windows. Still getting hate. Still getting hate. Yeah.

Dave Jones: Yeah. They're really struggling over there at Microsoft. They're here.

Ryan Cousins: Yeah. Yeah. But you know, it's tough to roll around in all that cash. Yeah. Yeah. It must be. Yeah. Yeah. Make it out the door. Yeah. Yeah. Well, I can imagine like, even if you did have a Linux, uh, you know, say, so you said there's two processors on the zinc, right? There's a two, two, uh, a nine cores, whatever. Um, and even if one of them is running Linux and the other is running real time OS, and then it's talking to some custom peripheral, I can imagine some safety critical stuff going from custom peripheral in the fabric up to a micro that's doing, you know, some kind of processing on it in bare metal, handing it over to Linux and then acting on it, pushing it back down. Like that's still a decent amount of timing to get all the way up and all the way back down. If you need some kind of higher level intelligence doing its thing on it. So I could, I could, I could see how that would also impact, you know, the use case of, of a Linux or any, really any operating system at that point.

Dave Jones: No, for sure. I mean, cause then of course, then you got to consider like what your buffer size is, like how frequency is Linux serving this servicing this thing? What happens if like that call never comes, you know, you're stuck with all this data you got to dump. Like, I mean, yeah, there, there's, there's a lot of, uh, a lot of different considerations there. I mean, you know, it's, it's one of those things you're never going to really. Get away from, cause that's going to be a factor in any embedded system. What's, what's nice is, uh, you're dealing with a single architecture, you know, it's, it's just like a very tightly integrated SOC. So communication bandwidth is never an issue between the FPGA and the ARM cores cause they're

Ryan Cousins: connected by actually taking it down to a serial, serial link and then, you know, re going back up or whatever. Yeah, exactly.

Dave Jones: I mean, I can, I mean, we've worked on so many, so many projects in the past where, you know, you have like seven microcontrollers and like a couple microprocessors spray it all over a PCB. Now you have to come up with like an insane communication architecture to make sure those all work together and all that.

Chris Gammell: So like, and it's going to be custom.

Dave Jones: Exactly. Oh yeah. A hundred percent. Yeah. And then, I mean, then you get into issues with like validation, verification and validation too. If you're doing medical or something that's regulated like that. I mean, you know, Linux, of course, you know, with their 10 million lines of code or whatever, isn't necessarily a good option when, you know, you could do the same thing with like a few thousand lines and running bare metal on something. So, you know, it's all sorts of considerations for that stuff.

Ryan Cousins: But how many of your customers, so I also wanted to mention, so you guys, critical also does engineering services. You guys will help, help your customers. Someone who might start with the product, then they're like, oh, I want to take this to a, you know, a broader thing. How many of them are doing that? I mean, how many people are coming in and saying, yeah, can you help us with the hardware? Can you help us with the, you know, the fabric or the software?

Dave Jones: Yeah, quite a few. I mean, I would say at any given point in time, we usually have like two or three like hardware development or well, system development projects like that. And then, you know, ranges from anything, anything from, we just needed to basically, you know, build some, a custom baseboard for us and like get it up and running to, we needed to basically design this whole product. You know, Encloser, you know, the software stack, like all the IP integration, like they're just the, every, all the ins and outs get us to basically get us to like a sustainable manufacturing state. Because a lot of times we'll, when we're doing, especially the hardware stuff, we'll, we'll help our customers with the first couple of manufacturing runs and then they can go and hand it off to some volume, volume CM or whatever.

Ryan Cousins: So yeah. So would you say that starting a product like this is a good loss leader for, for a consulting business? Well, would you recommend this to me, Ryan? It is really funny.

Dave Jones: I mean, honestly, it's funny you mentioned that, Chris, because it's the, I will a lot of times tell people that if you want to get a consulting business off the ground, launch a crowdfunding campaign. It's, it's a really, it's a really, I mean, you got to be careful with that because that anyway, I'm not, you know, back to the whole crowdfunding disaster, you know, like nightmare scenario. There's a whole bunch of stuff that goes into that. But, but in general, I mean, we've gotten more consulting business out of the one crowdfunding campaign than pretty much anything, you know, and it's anything else we could have done, you know, basically just sticking a website up and handing out a bunch of business cards. Yeah. I mean, it's a very, very good marketing tool for, for consulting for better or for worse. I mean, if that's what you want to do. Right. Right. Right. Great.

Ryan Cousins: The problem is you have to just deliver at the end of the day. That's, you know, if it wasn't for that, it would be great. Yeah. Yeah. Yeah.

Dave Jones: It doesn't look that great in your consulting resume with like the product that you advertise and sold totally fails. A bunch of people paid for it.

Ryan Cousins: You know, but if people, if people would have just paid me directly to do the hourly stuff, I would have just made them a custom product. It would have worked then. That's right. It would have definitely worked then. Yeah.

Chris Gammell: That's right. Oh my God. Yeah.

Ryan Cousins: That's, no, it's interesting. I think it is. It's like broadcasting out of the world. It's like, Hey, look, I've got this thing, you know, and especially if you are crafting these stories around like, Hey, this is a, a really good fit for, I don't know, like autonomous vehicles or something like that. I can imagine that would be a good business for you guys to have like, Hey, we need to do a lot of vision all going into the same thing. And you know, if, if one of these industry spikes, they need help, but they're going to be looking around or look on the web and whatever. And it's, Oh yeah. Well, Ryan, Ryan and his company, they, they, they do this kind of thing. They're, they're good with FPGAs. Maybe you should just ask if they can build more of it for you. Yeah. Oh yeah. Okay. Great. You know, that's how conversations go. It's kind of crazy. No, a hundred percent. Yeah. Well, it's, that's great. I mean, I'm, I think that's, that is the, the, the happy ending from the crowdfunding campaign, which, you know, it sounds like it was painful, but it, it is, you know, it made a ongoing business and ongoing, um, you know, it's like a, it's about the size of a calling card. This is like your calling card.

Dave Jones: No, I mean, it's true. It's, it's, it is funny. I mean, the crowdfunding stuff, I mean, I will mention, I think this, uh, literally this next week, uh, we're expecting the very last components of one line item that we, that were still back ordered from our crowdfunding campaign to come in and ship out. So it's literally like four years, almost to the date, uh, to the day is, you know, we'll have shipped a hundred percent of everything that we promised we would ship. Uh, so it's like, it's one of those things where, I mean, in the end it all works out one way or another. Uh, I mean, most people, or at least most smarter people than us probably would have like quit and just said, sorry, everybody a long time ago. But, uh, you know, for some reason we decided we like really needed to deliver for, for everybody that we, uh, who we promised we delivered to.

Ryan Cousins: So, you know, we, we got all of these, uh, are these all in the, uh, the, the campaign updates on the, the, uh, crowd supply page?

Dave Jones: Yeah. Yeah. We eventually, well, I think there was a somewhat recent update on there, but yeah, they are in there. You can, if you want to follow the timeline of like disasters, you know, for the, for the last, for the three years or whatever it took us to do most of the deliveries, you can, you can find them on the crowd supply page. But, uh, yeah, I mean, it's, it's interesting. I mean, I keep on, you know, telling myself I'm going to, I'm going to, uh, you know, put together some book next.

Ryan Cousins: Yeah.

Dave Jones: Something like that. But of course that, that like has like a, you know, you almost have like PTSD, you know, to go going back to revisiting some of these old experiences of just like, you know, it gets your heart pounding, just thinking about it. But, uh, yeah. One of these days I'll do that.

Ryan Cousins: Waking up from a dream, thinking around a plane to China or something like that. Basically. Yeah. Well, uh, where can people find out more about, I'm, I, I feel bad. Like, so we kind of hit our hour here, but we didn't actually talk about the crowdfunding that much. That's why I'm kind of pointing people towards the, the, uh, the crowdfunding page that that's going to, you know, we would end up just rehashing a lot of that stuff here anyways. Um, but where can, where can people find information about you, the crowdfunding campaign, everything like that and how to buy the product and get in touch, whatever. Sure.

Dave Jones: Sure. Uh, well, yeah, you can definitely always go to our website, obviously critical.com spelled, uh, in the most annoying techie kind of way possible. Uh, no vowels. Yeah. No vowels. And this, the K's, we use K's instead of C's. So it's K R T K L, uh, which makes it really easy to find if you're just doing a Google search because we're the first like 5,000 results for, if you type in K R T K L. Uh, right.

Ryan Cousins: Right. You just have to know that. Exactly.

Dave Jones: That's the other, that's the challenge. Yeah. Um, snickerdoodle.io is a little easier to remember. So you can always check that out as well. Uh, yeah. On crowd supply, um, you can find the snickerdoodle campaign. We can kind of read through our, our, our original campaign. Well, I might've changed a bit over the, during the campaign, but it's all posted up there. You can still buy snickerdoodle on there. Um, it's also all of our stuff's available on Mouser now. Uh, so you can search for it on there. Uh, yeah, we have like newsletters and stuff. I've of course been extremely lacking on the frequency of which I'm sending out newsletters. Oh yeah. I think you blog posts, but that's pretty, pretty standard, you know?

Ryan Cousins: Right. If, if, uh, if crowd, if crowd, uh, funding campaigns are built on people, you know, putting, uh, you know, boards and drawers at the end, uh, MailChimp's entire businesses on people that have, they continue to pay for monthly subscriptions and don't send out any newsletters like myself.

Dave Jones: That's right. Exactly. Pretty much. Yeah. Yeah. Um, but no, it's, it's, uh, it's all good. I mean, you know, we, uh, you know, I try and like, if anyone ever, you know, has, has questions or, you know, wants to hear the horror stories or, or I guess I had, I occasionally have some words of encouragement. Although there's a few and far between. Yeah. Right. Yeah.

Ryan Cousins: I come up for beer folks and that's when you get the good stuff and the bad stuff.

Dave Jones: Exactly. Yeah. Yeah. I mean, people can email me anytime at just my last name cousins at critical.com. Uh, I'm always willing to help. Like, I mean, I like to help other, the hardware startups or, or inventors or whatever, as much as I can, because I know how hard this stuff really is.

Ryan Cousins: So, and do you guys have a getting started page so that people could see what it's like to get started with, uh, with the actual, uh, snickerdoodle as well?

Dave Jones: Yeah, we have a couple. I mean, honestly, that stuff, that's one of the areas I really need to work on improving. Um, we have a pretty, well, relatively active forum. Uh, that's like always, of course, another challenge to kind of keep going. But, um, yeah, if you go to the snickerdoodle page, we do have some, uh, a couple of drop down menus for documentation, getting started, some tutorial stuff. We have like a decent resources page or some downloads. It's kind of, it's kind of all on there. Um, yeah. Yeah. Yeah.

Ryan Cousins: I see a hello snickerdoodle. That's a, that's a good start. I think exactly.

Dave Jones: Yeah.

Ryan Cousins: I'll link all that stuff into the show notes too.

Dave Jones: Yeah. It's always kind of challenging, like, uh, the community stuff's tricky because, you know, we have these forums and these wikis and things that are, you know, our GitHub, obviously you can check out as well. Um, you know, it's all kind of geared towards, uh, community driven content and resources. Um, of course, keeping people engaged enough, you know, once, once people get the answers that they need, having them hang around long enough to help other people. Yeah, exactly. I guess that's kind of tricky, but so we end up answering, you know, 95% of the questions anyway, but, uh, but it's all right. I mean, this just kind of comes with the territory, I guess.

Ryan Cousins: Well, Ryan, thank you so much for telling us about this. I mean, it sounds like from, uh, you know, the ashes of a crowdfunding, not even ashes, you guys had a successful crowdfunding campaign from the pain of a, uh, crowdfunding campaign. You've, you've, uh, made yourself quite a business here and, uh, it sounds like a really good service for people who want to get started with an FPGA and an really complicated system, all kind of package in a small thing. Yeah.

Dave Jones: Yeah. Yeah. No, I appreciate the kind of words and yeah, it's, it's, uh, it's something that'll, you know, only get better with time is really the best way to put it. Um, you know, we're, we're still just kind of, I mean, it's like, uh, we have this kind of internal joke where we're saying, you know, as an overnight, it was a four, it took us four years to have an overnight success kind of thing. Um, you know, I don't know if I necessarily like, uh, across the board qualify the entire thing as a success, but it's, it's something that'll only improve with time as more people are using it and the platform continues to evolve and, you know, just, yeah, hopefully, uh, you know, any little spark to some additional innovation or creation that we can get, uh, you know, is, is makes it all worth it for us. So that's kind of what we're going for.

Ryan Cousins: That's great. All right. Well, we're looking forward to the next, uh, the next rev of things and, uh, we'll talk to you soon.

Dave Jones: Yeah. I appreciate it, Chris.

Dave Jones: Uh, let's go.

Speaker ?: x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x x