Embedded - 309: Nature's Engineers

Episode Date: November 8, 2019

Pete Staples of Blue Clover Devices (bcdevices.com, @theiotodm) spoke with us about tools for manufacturing hardware.  Some posts and products from Blue Clover Devices: PLT Spec Sheet and ICT System How to write PLT Board Test Plans (using YAML) PCB Design Tips and More PCB Design Tips Flying Probe Testers in a Nutshell Zephyr - an OS for IoT BC Devices github Behind the scenes at factories: Shenzhen Tena RT3188 HDMI Stick Factory Tour Inside a PCB Soldering Factory - in China Thank you to our Embedded Patreon supporters, particularly to our corporate patron, InterWorking Labs (iwl.com).

Welcome to Embedded. I am Alicia White. I am here with Christopher White. Have you ever wondered how your code or your electronics get from one unit on your desk to 10,000 tested and shippable boards? Pete Staples from Blue Clover Devices is going to tell us how this works. Hi Pete, thanks for joining us today. Hello, thank you. Could you tell us about yourself? Sure, my name is Pete Staples and I'm the CEO and co-founder of Blue Clover Devices, operating in San Francisco, Hong Kong, and Shenzhen. We've been in business about 16 years. And before starting this company, I was a payload designer at

Boeing Satellite Systems down in El Segundo, California. But you've been doing Blue Clover for quite a while now. Yes, I have. And we're an ODM, and we provide design and manufacturing services to a variety of companies, some very large like Tesla, Ford, and Volkswagen, and also a lot of hardware startups and help them get into production as quickly as we can and fill in the gaps along the way to doing so. Okay. And then there's a tool you make, but I think I'm going to save those questions until after lightning rounds. Are you familiar with lightning round? I am. Yes. Okay. Let's just get to it then. Do you like to complete one project or start a dozen? I like to complete things.

Favorite flux? Kester, 186. How many layers is the optimal number of layers for a board? Six. Flex circuits, good or bad from a manufacturing point of view they're not as easy but we tolerate them favorite acronym plt least favorite component to put on a board in manufacturing? Transformers. Why is that? They don't fit in our machines. They're finicky. The documentation isn't very good on them. And I can go on and on.

Okay. So, so let's go back to, uh, you're an ODM. What does that acronym mean? It's a horrible acronym, so maybe my least favorite. It's original design manufacturer, but it doesn't really answer the question. Traditionally, OEMs are the brand holders and ODMs are contract manufacturers. But rather than be called purely a CM, you would use the ODM term if you're also providing design services. And it kind of grew up out of specialized factories that did one type of product and kind of pre-designed it, like maybe specialized in laptops so that an HP or Dell could show up and say, all right, we want that one, but put this color band on it and put our logo on it and then we'll take it. But we do a variety of products. So we're in a little bit different type of ODM role.

But one of the products you do is your own product that is for other people who do manufacturing. Is that right? So last year, we were using this tool that we fell in love with. We saw a lot of potential for it, and we decided to productize it and brand it. And that's the Production Line Tool, or PLT. And we sell this on our website and have built an ecosystem around this tool. And what it's used for is firmware programming, the act of loading the firmware, and also a lot of electrical and RF tests. Okay, I use my JTAG unit or my ICD or SEGGER unit or whatever to program my boards.

How is what you do different? Those tools are aimed at developers, and so they're suitable for that purpose, but they aren't really built to work in a production line environment where you need to do it many, many times, and you need to have good records of which units got which firmware. And so that's what this is built for, is streamlining that process. And you could use it in the development phase, but it's $2,000, so you wouldn't

normally do that if that was your sole focus. Unless you were very lazy and had a lot of money left over. If you're like, I've got too much space on my desk, I want something bigger, then you might buy this. But the impetus for it is resolving the handoff challenges between development teams and factories. And so we run into a lot of people who travel back and forth between California and Asia, and they often do so because of ambiguities on how to actually do that programming and testing. So we built something that it's more expensive than a J-Link, but it's much less expensive than what you normally find on a production line. And so you can

afford to put it in both places. And that way you can see the same results and get to the bottom of what's going on much more quickly with less travel. What I'm used to is companies making their own, you know, sending some engineers over to China and making their own bespoke programming harness that maybe has, like Alicia said, 10 J-links or something hooked up to a laptop, or if it's serial programming, some big octopus serial thing that goes in. Yeah. You know, they have their own software.

So does that solve some of these problems too where there's like, okay, we want to do this in parallel? Or is it just... Those are our best customers are the ones that have built the same thing, essentially, and just thought to themselves like, oh my gosh, I can't believe nobody's, you know, it's like my fifth time building this i have to do it a little differently for this project but a lot of aspects

are the same and so when they see it they understand right away what we're trying to do with the plt were you surprised this was an untapped market? It seemed, thinking about it, like, somebody should have done this by now. Why didn't they? I think so. And I think it's a testament to the players in the space being pretty good and strong and just nobody really wanting to take on someone like SEGA or National Instruments

because they've been around and have had their, they have talented sales teams and a lot of history behind them. And so it just seemed like we'd do it, but then we'd be competing with these giants and it's not worth it. But the PLT, it does more than just program the units. I mean, there's the whole testing part.

Are we talking like using my command line to verify all of the hardware or are we talking make sure power and ground aren't shorted together? Definitely the latter is in scope. And we named it production line tool because it's designed to be good enough for testing. And it's not really aimed at debugging your design. And it's not as good at that and so you would still lean on whatever development tools you're comfortable with to refine your your firmware or even aspects of the the design but there is this need for something that the developer can use directly and you know completely configure on their own and know that the handoff is going to be smooth because the hardware is the same, the code that it's executing is the same. And so that's the intent of the product. Testing boards is not something a software engineer usually worries about. Usually by the time I get them, they're working or if they aren't,

it's non-obvious how they aren't working. But that's not a manufacturing thing. That's a development problem. What kind of testing do you do in manufacturing that a production engineer just maybe just doesn't really know about? Well, there are an infinite number of things that can go wrong. And what we try to do on every project is design the testing around the predicted failure modes or people do what's called an FMEA, failure modes and effects analysis, which is a risk assessment on the product, kind of comes out of automotive. So Volkswagen is who initially got us turned on to this, and they did a DFMEA for design risks and PFMEA for production risks. And usually just doing one general one is enough on less high profile projects. And you're just trying to figure out what's going to go wrong and then design tests that can reduce that risk or improve the detectability of that. It could be components that are out

of tolerance. It could be assembly problems, or it could be just noise on circuits, things that you can detect with electrical measurements, but it might be time consuming to do unless you add a fixture expressly for that that's repeatedly doing it and gathering that data. Okay, so the testing fixture. You showed me an ICT100A. Is that the same as the PLT or is that different? It has 78 lines coming out so that we have a lot of I.O. to connect to whatever device we're testing. And we can do that with cables or custom-made adapters. But a common thing when you're testing at the board level is a bed of nails or an in-circuit test fixture. So we built that as well just because that's something people recognize. Nobody really says, I need a production line tool and searches for that and would find that.

But people do have in-circuit test fixtures made all the time. Almost every project has one or more of these made. And the difference is that they're normally totally bespoke, like you mentioned, Chris, where the factory will have that made by a local firm and connect voltmeters, ammeters, some buttons to it, indicators and so on. And rather than make those one by one based on the product, we decided to make that somewhat standardized and then just reconfigure the pins based on the board that we're testing. So the ICT100A is technically the base. And then inside of that base, you can plug in cassettes that are tailor-made to the board that we're testing.

So it's, and it's designed to work with the PLT. I don't know if that clears it up, but. So the ICT was the framework and the bigger part. And then you had a board sitting on it and the board was sitting on a bed of nails and you were testing that board with the pogo pins or whatever it was that the shiny bits were and you said that like half of it half of the cost of it stays the same between boards and you have to

only pay for part of it to be updated when you spin a new board. Am I remembering? I am remembering all that right. Yeah. So our mission is eliminating waste and it just seems like a waste to have these big jigs recreated over and over again. And so the big jig, if you will, is the base. And you can use that as long as the board is no bigger than an A5 dimension, which is about half a sheet of paper. And then you can't get around the need to rearrange the pins. I guess some people do a grid where it is, or you could do flying probe, which we may talk about. But practically speaking, you have to rearrange the pins, the bed of nails.

And so that is what we package into a board and a base and the pins themselves and this 78-pin connector. And we call that a PogoPin cassette. So that's $800 that you spend each time you do a new board, but that's a lot less than you would normally spend on a full fixture. How much would I normally spend on a full fixture? A few thousand. And then how much is your whole system? You can say, you know, a range if you want. No, it's all published. So that's unusual in our line of work to just have a shop and publish these things as fixed prices. But that's what we've done. So the PLT is $2,000. The bass is $400.

The cassette is $800. So $3,200, you've got the complete set. And then if you need more PLTs, you can add those on. I should mention by deploying firmware and tests from the cloud backend, which we call PLT Cloud. And that has a subscription component, which is $1,000 a year for a five-person team or $5,000 a year for a 500-person team. And that is so that if I want to do something in my office, I then can forcibly put it onto the device in another country? Almost. So it forcibly makes it available. So if you're updating your computer, it'll say there's an update available and you don't have to do it. And in this scenario, the line also doesn't have to

do it. We didn't really want the line to be suddenly changed because of a keystroke that So we want to give you power, but not quite that much power. And you would have a way to know whether it's actually been updated and is utilized because you have reports in the PLT cloud that show you unit by unit what firmware image is being loaded, what tests have been run, what the result of those tests are, is all viewable. So it keeps people in sync, but you can't quite change the line directly. They have to accept the update and start using it. So we get one of these and there's a board and there's the pins and then there's the 74-pin connector to the PLT. Who writes the code on the PLT? And how do you know what things are supposed to be shorted

and what things aren't? So whoever is designing the product really knows the most about it and is in the best position to design the test. So there's more and more awareness and interest in designing for tests in the beginning. And so whoever's writing the firmware is often the one designing those tests. It doesn't have to be that way. Um, we don't really specify who, who should do it. Um, but what mechanism we're breaking is that you don't have to be an expert in these testers in order to write the tests. So it's a YAML file is the format, and we publish a test plan reference that lists all

the commands that are available and targets, like which microcontrollers you can program using this system. And we think anyone can do it. In practice, it's usually the engineers developing the product that are the ones writing these test plans. And you don't just look for shorts between power and ground. You do get the people who design it usually take their own netlist and use that and put it into your YAML file so they can test the connections they want to. Yeah, that's the idea of it. You can also send UART commands.

You can send CAN bus commands through this tester. So it's quite versatile, but it's intended to capture the things that the in-circuit testers are already doing, which are often electrical measurements, some very basic, some that are more nuanced. So you can talk to it over serial. So if somebody wanted to do a bigger test that involved provisioning Bluetooth information or verifying communications on some other port through a serial command line, they could. Mm-hmm. Okay. I thought it was just the wires and the microcontroller. My mistake for not clarifying that.

This is why we get to talk instead of just me read your website. Okay, so when I release a board and software to my manufacturing team, I usually have to help them figure out how to program the board. That's exactly what always happens. Oh, yeah. And that's what we're trying to say is that, yeah, you could just say, all right, you've got, you know, here's that box and I've released this, I've sent this new release. So run that and then we can look at the test results together. That's the world we want to live in. And when I work with big companies

that are doing consumer products, they always have bed of nail testers. But sometimes the little companies that do like scientific gear, they have flying probe testers. What's the difference? Flying probes are, they're newer and essentially the pins, the bed of nails move, so it's obviously a more sophisticated piece of equipment. And we would love to plug the PLT into a flying probe tester. We're not saying there's anything wrong with them. They are expensive. That's why you don't see them everywhere. We had an intern over the summer research them and write a blog post about them. And from his findings, they were at least $200,000.

And they're also a bit slow. So you don't often see them in a production line because each test, each unit is about 10 times slower than an in-circuit tester. And so they're really good if you kind of have a routine around thoroughly testing new boards over and over that where you don't really have time to build the in-circuit, the bed of nails. You just want to change the program for the new board, program that and test as many points as possible. So they have their purpose, but they are not in most of the factories that we see,

or at least if they are, they're not as heavily utilized as you might think. Which is really sad because they look awesome when they run. They're like little hummingbirds flying around. I can watch the videos for hours and I'm tempted to get one because I love the idea of somebody building a board and then just loading what they want tested and being able to feed that directly into something so that you're not rebuilding anything. You're just essentially on the fly designing or implementing this new test. So someday we may do that. But of course, it is going to be slower, as you mentioned, because think, was a number that

came up as a rule of thumb. And just due to the travel times, like you say. How do Internet of Things devices change manufacturing testing? They make it more complex. The software is a lot more broad and involves a lot more partners when you get into connected devices. And so the reason for us opening the office in San Francisco is really out of awareness that the software is becoming way more important to these devices and that we needed to be closer to where a lot of it was being developed, closer to the semiconductor manufacturers, and just be sure that we had a better understanding of these connected devices. So our tagline now is the IoT ODM because we do embrace that complexity.

How do you deal with all the noise of the devices on the factory floor? And kind of as a corollary to that, because it's happened in my office, how do you test the one you want instead of all of the others that are sitting around? Powered, squawking? Well, often these things are designed to operate in those noisy environments,

and so they can still make themselves heard in a way through their identities. And so a lot of testing doesn't actually require the RF shielding that you might need in certain products if they don't use a publicly identifiable ID. So this is something I've been kind of puzzled by myself is things like Mac IDs that change over time. And we had a guy from IEEE visit our booth at Embedded Systems Conference. And I said, Oh, you I could ask you about this. Like, how do you do you assign blocks of Mac IDs that can change with inside of a product and things like that? And he said that they don't, that you just get individual, you get the blocks of Mac IDs for your assigned to your company that you can then assign to products, but then companies will have IDs that aren't the same format, but aren't actually part of the Mac, part of the IEEE system. I didn't fully understand it actually, but. Is this mostly with Bluetooth devices or with like 802.11 Wi-Fi devices, the Mac address thing?

We do more Bluetooth. So we we see see more of that but um mac ids are assigned for wi-fi as well as far as the noise you have you just design the line around the particular product and so um we do some tile co-branded products where we don't actually need to, even though there are a lot of devices on the line, we don't need to shield them during the test because we can identify which unit through the serial number that we temporarily affix to the unit. And that's how we know which one we're talking to. When we execute a test, we can scan that unit

and then communicate it by name, essentially. Say, hey, I'm talking to you, and are you getting this command? That kind of thing. Yeah, I've done some of that, but there is a point at which a tinfoil lined box comes in pretty handy. Yeah, we think that could be a product extension because that's also something that isn't done all that elegantly in our experience, but it might be nice to just make a standard one. A standard tinfoil-lined cardboard box. Yeah.

Only Reynolds Wrap will do for our Primo RF shielded box. What's inside your PLT? You said it's a computer. Is it like a laptop. Is it like a laptop? It's a Linux box itself. I guess the dimensions, I don't have off the top of my head, but it's like a very sleek toaster oven type size and it includes a single board computer a DAC for data acquisition AC power supply for 100 to 240 volts input a USB hub something we call a carrier board which board, which includes MUXs and relays and FTDI modules. And on that is the large 78-pin connector that we use to engage with the device we're testing and a UX panel on the front with an OLED display and aluminum buttons on it. Do you remember when computers used to be big? Sorry, I'm not talking room size big.

I'm not that old, but still. Let's see. Jakey Poo asked, Jake, can you please? Yeah. Okay. Asked if the Pogo pins are RF compatible or how you do the RF tests? That's a good question about the pins. We never tried to do any RF testing through them. I assume so far that wouldn't work. I'm a bit of a pessimist. We use RF radios inside of the PLT to just transmit and do Bluetooth or BLE testing rather to the device that we're testing. So it's not through the pins. And so if somebody wanted to do some weird RF thing, you'd have to find another module to...

I think so. We consider the cassette to be the real estate for customization. So there's really no limit to what we could put on the cassette and do additional testing that's not built into the PLT itself. So that's how we work with clients who can't use the standard products that are on the website but want to use the system for their testing. We just modify the cassette and then it's not the $800 model. It's something more expensive because we customize. How has the reception been to this? Wait a minute, let me go back.

How long have you been doing this product? We launched this year, and the first show we did was in Stuttgart in May, and the reception's been excellent. It's funny how, what a range of adoption there is that some people are extremely skeptical because they've done so many rounds with another way that they just assume it's, you know, they kind of ridicule the device.

And then there are a lot of people who have built something fairly similar multiple times, and they're just foaming at the mouth, really, to get it. And they're just like, I've been looking for you guys. Thank you for finally doing this. And overall, the average response, if you will, is positive. How about sales? Response is one thing. Buying stuff is the other. We sold one yesterday. We're still in the first hundred customers.

So it is a new product and it's something that are at various stages of adoption, but we've sold under 100 so far. Are you using the product to test the products as you manufacture them? Yeah, not yet, but that's coming. So it's really built for smaller firmware images right now, but it's likely going to be extended to being able to handle programming other Linux devices, for example, which this is. Are you getting more excitement from the OEMs, the people doing the design of the electronics and people like Tesla and Volkswagen, you mentioned those, or are you getting more excitement from the contract manufacturers in Asia who see this as a way to simplify their lives.

Because they have multiple lines. If they could have one device for multiple lines, that's kind of handy there too. Right. So far, it's much more warmly received in the U.S. among the brand holders who have, I guess, more at stake send it to the CM that's doing the production. You do have offices, or you spend a built up largely in Shenzhen. So I used to spend 80% of my time there. And in the last five years, it's kind of the reverse. So I'm in the Bay Area a lot more, but I do go about four times a year. Have you, you built this partially because of what you've learned doing your work as an ODM helping companies design and manufacture products.

How many CMs have you worked with that show you things like this? I mean, so there's no competition for this in the U.S. Is there already competition for it in Shenzhen? Not that we've encountered. Definitely products get made, so they have a way to do it. We just don't think that it's not this approach fundamentally. So there is no shortage of ability to make test fixtures in Shenzhen, and we've seen all kinds of ways of doing that. But the folks who are on the ground there don't have that burning desire to make

sure that each of these test results are immediately visible in the cloud. And that desire is more prevalent among the hardware engineers that are designing a product in California, say, and need to make sure that it's really being, being done that way in the factory. And so that just seems to be where the demand is. And that's indeed where we developed it was in our San Francisco office.

And so I'm sure we will see competition eventually, but so far we haven't. And the cloud part, I have to admit the cloud part makes me a little scared because there's so much about security these days. And I would say that the test methodology and the unencrypted firmware would be things that I would not be excited about putting on a cloud unless it was very secure. How are you making sure people feel okay about that? Well, it's way better than what we see all the time anyway. Emailing the SREC over. Here's my top secret firmware. Don't send this to anyone, you know, and we get those emails sometimes.

And we're just like, don't email this to me. Come on. This is the worst security regime you could ever imagine. And yet it happens all the time. And so, we attack it by making sure that you don't have any inclination to do that, simply putting email or putting sensitive test plans or firmware in places that you really have no sense of who actually has access to it or how far it's been forwarded on. So by creating PLT Cloud, you can have 2FA as your company policy for logging into that system. You're logging in as a known individual, putting firmware into this cloud system. And then what we work on is the security between that cloud and the PLT hardware. So there is an authentication chip inside of the PLT that identifies itself and makes sure that that communication is secure.

It is encrypted in transit. To your point, though, it has to be unencrypted at some point before it goes into the device you're programming. And so we don't encrypt it at rest. And indeed, if you took apart the PLT and extracted that firmware you could you could have it so we're looking at ways of adding intrusion detection but right now we essentially make sure that the traceability is very good and that it's very hard to have a man in the middle attack, but you do need to trust somewhat the company that's programming your boards. Which you always did.

I mean, before you were emailing them your binary, and then it was going on to a computer, and then it was going through a SEGGER or other programming device. So it's not like you're adding a layer of insecurity here. Thank you. Right. We're not polluting the water or anything. We're just only filtering as much as we can. And we feel good about it. We think this is a much more secure and yet usable system than what we see day in, day out in other places. Are there other things about running a factory that product creators don't really think about? Are there things manufacturing engineers would like to tell design engineers to do or to stop doing? Change is hard.

So as a manufacturer, we absolutely rely on version control and and that's that's something that uh we find that we have to get a little bit preachy about that you can't just send an email saying oh you know we want to try we're going to swap out this part and um not keep track of of changes so in the beginning, when you're prototyping things, yeah, you can keep track of builds and material in Excel and just take parts off the board and put new parts on and keep some, scribble some notes, that's okay. But once you're planning to actually put this in the hands of customers, you really have to have rigorous version control and know which version is which units are on which version and and track that stuff it's funny as a software engineer i'm like yes of course you do but as i've seen in manufacturing with i've seen it where

the manufacturing engineers have tried to explain this, that if we change this and it breaks, you're going to blame us. So we'd rather not change anything. Yeah, I guess, I mean, that's, being a manufacturer is tough because there's this implied responsibility, even though, you know, you sometimes don't even know the details of the product. And yet if it comes from your building, you know, people really hold you accountable. And we, we accept that, but it does make it very challenging because when something goes wrong, it, it could be a decision that you really, really warned against.

But people forget those conversations sometimes. Yes. Especially when they happen at odd times because of time shift. Yeah. Time zones. Are there any other things that design engineers should know or good rules of thumb that every board designer should know? Like only put heavy objects on one side? Yeah, that's a good one.

We are working on a blog post about how to design a board so that it's easy to test. So our recommendations are put tons and tons of test points on the bottom where the bed of nails can engage easily. And if you can just put all the parts on top and don't sweat adding an additional layer, it does make the board a little more expensive, but there are a lot of folks who realize that adding a layer really does give you a better signal integrity and makes people's life easier overall. And there are just some handy dimensional aspects to the board that can make it flow more easily that we'll put in the blog post. I will link to that when we get it. I recently saw the Shenzhen Tena HDMI stick factory tour. Are you familiar with that video? No. I got it from Alan Cowhen's book, The Prototype to Product.

And it is a video in which the CEO of a factory takes you all around the factory and shows you the reflow ovens, the assembly pieces, the testers, the boxers, the people who do the rework, and it was kind of eye-opening to me, the number of people and the amount of time they spend looking at tiny, tiny things. And there's another one from Strange Parts, Inside a PCB Factory, PCB Soldering factory. I've seen that one.

And I just recently came across strange parts and found that guy's pretty good. I like how they take you behind the beaded curtain and you do learn things from that. For us, it's a little bit of the world we already know, but I think those videos are pretty spot on. It shows how different it is from our desks in California. Yeah, I mean, even little things like people developing on Macs here, and then they get to the line and everything's on a PC with maybe Windows 95 on it or something. And so, you're just like, oh, this tool I'm totally leaning on might not even work on this thing. And you can't just say, all right,

I think you guys should all just get Macs now and, you know, everything will be fine. So, they, you know, it's just, it makes things that don't seem like they should be difficult in practice turn out to be pretty difficult. How did you, how did you get started in BCD? You were, you were a systems engineer at Boeing and then you left to found a company. How did that happen? Well, I think I always was meant to start a company. And I had a boss at Boeing that recognized that I had an opportunity to do a business case in the company for a new satellite constellation. And that kind of awakened the entrepreneur within. And he just said, you should probably go get an MBA and start a company. And I respected his advice and followed it. And we didn't really, I didn't intend to go into manufacturing. That was really an accident,

probably one that I would not do if I started now. I just think manufacturing is really, really stressful, but it was something that our customers needed. And so we opened the factory because we couldn't find the factory that we needed to take care of our customers. And you got an MBA? I didn't even see that. Okay. How was it to shift from being a hands-on engineer to being a CEO? Uh, yeah, it's just completely different, um, patterns of working. I I've had to become a lot more collaborative and you're, you're just trying to be in people's heads so much more as a manager, you know, really think from other people's perspectives all the time. And as an engineer, you can define the parameters of the problem more clearly than you can as a CEO. Did getting an MBA help with that? Did it show you the path to go from engineer to CEO? Or was it sort of the MBA was useful for other things? It gave me the time. I think it does just take time thinking about what you're going to be doing

later. And so if I just skip that and started, I think I would have struggled even more. I did learn things in business school. So I was at UCLA for that. The instruction was good. The case study method is helpful. I enjoyed that. And I met good people that I might not have run into in your previous career. And it was helpful. Networking. Networking is weirdly helpful. So my uncle is an entrepreneur, and he thought that business school would cure me of wanting to be an entrepreneur. And fortunately, I stuck with that plan, though.

How big is Blue Clover these days? About 80 people total across the three offices. And the Shenzhen site is the largest. And you have a company mascot. We do. We do. That is Ben, the beaver. And our office is pretty close to GitHub

and I have some friends there and I just saw how much fun they have with their Octocat and I panicked. I thought thought what if we get too big and then we can't have a mascot so uh you know apple can't have a mascot now it's it's too late so we we decided to hurry up and have a contest and we picked the beaver because that's nature's engineer and it's some it's a species that works in a team and we just like the vibe of that so we had a had some artists do renderings and the winner is the one that you see on our website i was a little confused to start with i thought it should have been a goat

because of the whole clover thing. But your explanation makes sense. Thank you. Apparently, beavers eat clover, too. Oh. That's what I've heard. Nobody should eat blue clover. No. If you ever find one, it's a rare and wonderful thing.

Let me ask you a question about kind of current events. You're tied into China and Shenzhen. The way things are going with tariffs and whatnot, how is that affecting the landscape? How is that changing your thinking with where people are looking to do the manufacturing? effect, that's for sure. It effectively makes our factory more expensive than factories in any other country. And I guess this is one of the reasons that we're investing more heavily into the tools, which are essentially immune to that because we build the PLTs in San Francisco, and we think that this is something that is more about the underlying software, so it's much more flexible in how we construct it, and we could build it in a variety of places. And people are thinking about where they do their production a lot more now. And we think that's also a place to our strength because if you have to move production and your factory has this $200,000 rig of equipment, it's not very easy to move that. But our testing solutions are essentially portable. We actually set it up at your 300 party a few months ago, and that's how portable it is. You can just pop it up anywhere. And so, we want to embrace that and affect change in the things we can do.

And there's not much we can do about the tariffs or anything. So we're just kind of letting that happen. We're not planning to move or anything. So if that was the underlying question. No, no, not at all. I just was curious how it was shaping what you guys do. Yeah. Yeah, it definitely affects us. And we get a lot of questions about it.

Well, Pete, I am thoroughly confused by the time change, and I don't know whether it's breakfast or lunch, but it's something. My stomach is demanding food. Okay. So I think I need to, to wander off before we do that. Do you have any thoughts you'd like to leave us with? Uh, no, I think I'll, I'll pass on that one. Okay. Our guest has been Pete Staples, CEO and co-founder of Blue Clover Devices.

That's bcdevices.com. You'll see in the show notes. Thanks, Pete. Thank you. Thank you to Christopher for producing and co-hosting despite his cough. And of course, thank you for listening. You can always contact us at show at embedded.fm or hit the contact link on embedded.fm. Now, a quote to leave you with. He's going to leave all that in, isn't he? I am.

Yeah, I thought so. Okay, I'm going to... A quote to leave you with from Prototype to Production by Alan Cohen. While design development culture is all about nonstop creativity, basically contained chaos at times, manufacturing culture is all about process, precision, and repeatability. Embedded is an independently produced radio show that focuses on the many aspects of engineering. It is a production of Logical Elegance, an embedded software consulting company in California. If there are advertisements in the show, we did not put them there and do not receive money from them. At this time, our sponsors are Logical Elegance and listeners like you.