#325 – An Interview with David Kronstein (Tesla500)

David Kronstein Tesla500: Hey, everyone. A couple community announcements here. First off, I am now in Chicago, and I've started doing meetups again. I used to do those in Cleveland. The one here is called 3H, or Hardware Happy Hour. The first one is coming up next Tuesday, December 6th at 6.30 p.m. at the Haymarket. If you're in Chicago or you know people who are, we'd love for you to spread the word. There will be a link in the description. Secondly, we are having a contest based on a book that was donated by David Bengston. It is a signed copy of The Art of Electronics, which is very awesome. The way you can enter is to share the episode you're about to listen to. So if it's on Twitter, share a link to this episode, tag The Amp Hour, or send an email to contestattheamphour.com. You can also share via Facebook if you want to just screenshot that and send it to contestattheamphour.com. And finally, my favorite one, if you want to tell a coworker about this awesome episode you're about to listen to, send an email to contestattheamphour.com. Include us somehow. You can BCC your coworkers if you don't want us to see their email address. You can do whatever you want. You need to prove somehow that you did share it either Facebook, Twitter, or email. The more creative, the better. So enjoy the upcoming episode. It was one of my favorites in a long time, and we've done a lot of episodes. Here we go. This is the Amp Hour Podcast. Recorded November 30th, 2016. Episode 325. An interview with David Kronstein.

Dave Jones: Welcome to the Amp Hour. I'm Dave Jones from the EAV blog.

Chris Gammell: And I'm Chris Gammell of Contextual Electronics. And I'm David Kronstein from the YouTube channel Tesla 500 and Kron Technologies, Inc.

Dave Jones: Hey, David. Thanks for joining us.

Chris Gammell: Hey, thanks for having me.

Dave Jones: And it is David, not Dave. Right.

David Kronstein Tesla500: We went over that earlier.

Dave Jones: We'll avoid confusion. Yes.

David Kronstein Tesla500: Make sure you don't confuse each other. Right. And people probably know, have been seeing a lot of stuff, including from our own Dave, reviewing your upcoming, well, your, I guess, running Kickstarter project. Running Kickstarter, yeah. Well, that's great. Good timing. Good timing to be talking to you now. How much? So we're recording on Wednesday the 30th. This will go out Thursday the 1st of December. When is that running till?

Chris Gammell: That is running till end of the day on December 22nd. Okay. So get your Christmas list ready.

Dave Jones: And for those who don't know what it is, tell us all about your Kickstarter.

Chris Gammell: It is a handheld high-speed camera capable of recording a 1280x1024 video at up to 1,000 frames per second and lower resolutions up to 21,600 frames per second.

David Kronstein Tesla500: Is that all? Oh, okay. Yep, that's all.

Dave Jones: And you're basically a one-man band doing this too, right?

Chris Gammell: Pretty much, yeah. Pretty much. A few help from a few friends, but it's pretty much been all me, yeah. It's been an incredible learning experience.

Dave Jones: And you've been doing this for how many years? Like working on high-speed cameras, thinking that you want to make your own?

Speaker ?: 1.4, Dave.

Chris Gammell: 1.4. So at least .4. Yep. Yeah, I've been working on this since 2006, I would say, is when I sort of got into it. So a decade? Yeah, about a decade. There's a whole big story of how I got into it, if you want to hear. Oh, yeah.

David Kronstein Tesla500: Of course. Definitely.

Chris Gammell: That's why you're here, man. Yep. Let's see. I was always interested in high-speed. I started out seeing all the cool things Mythbusters were doing. I got... I was doing sort of things with compressed air guns and things like that way back when. So I wanted to see this stuff in high-speed like they were doing on the show. But cameras were just completely out of my price range.

Dave Jones: What cameras were they using back then, Mythbusters, right at the start?

Chris Gammell: I think they were using some old Photron brand cameras out of Japan. They had like this camera head and these massive two huge like 100-pin cables that went over to this big processor unit.

Dave Jones: Right.

Chris Gammell: That buffered all the data and started to RAM and such.

Dave Jones: And they were low res, right? They were...

Chris Gammell: I think they did something like 1280 by 1024. Okay. Fancy. They weren't too bad. They did HD well before they were even broadcasting HD. But they were only like 500 frames per second. And the quality wasn't that good.

David Kronstein Tesla500: But did they... Was there ever film versions of a high-speed camera? I actually... I'm not a film person at all.

Dave Jones: Yes. Yes, there were.

Chris Gammell: Yes. They had... There's a couple of common ones I've seen. There's one... They're both made by... Ooh, who had made... I forget the manufacturer. But they... There's one like intermittent motion. Like it's... You know how a film camera normally grabs the film, moves it, exposes it, then grabs it and moves it repeatedly. They did that up to 500 frames per second. And that was incredible. The intermittent motion spins at 30,000 RPM. And I can't imagine how they do this without it just tearing the film apart. But somehow they managed this. And they also made another one. This was 16mm. They made another 16mm one that uses... It moves the film continuously across a plane. And this rotating prism scans the focused image onto it. And that can run up to, I think, 11,000 frames per second. So you could do this with film back in the day. Yep. But you're eating like... I think it's about $200 worth of film for a 10-second shot. Yep.

Dave Jones: But that doesn't matter if you're shooting like... You know... Shooting like an explosion. You know, back in the 60s when they were doing the nuclear weapons of 50s. They were doing the nuclear weapons. They were filming those with high-speed video and, you know, stuff like that. So... But they had virtually limitless budgets. Cost didn't matter. Yeah. Right. Yeah. Exactly. Film? Well, we'll just run it for an hour. You know?

David Kronstein Tesla500: Like, no worries. It's like a Gatling gun of, like, film and camera. Right? That's crazy. Yeah.

Chris Gammell: Wow. So after... Yeah. So I was looking around on eBay for high-speed cameras all the time. I always wanted to get one. And this... I think it was an Olympus iSpeed 2 came up. Yeah, the iSpeed. Yep. Yeah, it was 800 by 600, 1,000 frames a second. I said, it would be perfect. I want to do all these YouTube shots and things. Started at $150. And, of course, it got bid up and it eventually sold for $3,800. And this was, like, just the camera. There were no accessories. I'm pretty sure this thing was stolen. Right. But it would have been so cool to get that. And I was so ticked off at not getting that. And I said, screw this. I'm going to build my own. And so I started this journey.

Dave Jones: I bid on one of those once. I bid on an Olympus iView. And it was, like, $1,500. Buy it now or something. And I went, yeah, I'm going to buy this. You know, this was a couple of years back. But the person had no feedback. Right. So I went, I know this is, like, a scam. But I'm going to do it anyway. Because I know, like, the PayPal buyer protection. I'm going to be protected. Right. Who knows? I might get lucky. Right. And actually get it. So, yeah. And then they, I got an email. And then they started, oh, I'm having trouble setting up my PayPal account. Can you PayPal me directly at this other one? And then I knew, yep, here comes the scam.

Chris Gammell: Oh, I have a story. Yep. Sort of like that where it actually turned out well. Oh. There were these two really 3.5 gigahertz Agilent differential probes on eBay. It was like, they went for, I think it was, ended up going for like $400. He had problems setting up his PayPal. But I was talking to him and he had two more of them. And these were, these are like, those are current products that sold for about $3,500 each. Yeah, yeah.

Dave Jones: They're not cheap.

Chris Gammell: So I said, oh, how much for all, how much for all four of them? And then he was, he said, oh, I'll give you all four for $900. I'm like, oh, this is a big risk. This is a bit sketchy, but I'm going to take it. This is too good a deal. And it turned out to be, turned out to be real. Excellent. I've been using those for all the debug on, on the camera and all sorts of other projects.

David Kronstein Tesla500: That's right, kids. So go ahead and send your money to the wind. And sometimes you'll get some nice hardware back.

Dave Jones: Because we'll get into that. So you've used them for all the high speed serial debug?

Chris Gammell: Yes.

Dave Jones: Yep. Yeah. It's difficult to do that without proper details. Yep. Yeah. Yep. Yeah, definitely. So probably most people know you from your Tesla 500 channel. How long have you been doing that now? How many subscribers? I've been doing that. Well, that's all jazz.

Chris Gammell: I think just past, actually, let's check. Yeah, just past 25,000 subs. Cool. Nice. I've been going up really fast lately because of all the attention from the Kickstarter. I think just a month ago, I was at like 18,000. Nice.

Dave Jones: And you were going to set up like a set. Did you end up setting up the second high speed lawnmower destruction channel? Yes.

Chris Gammell: Yeah. If you go check out Will It Mow, I basically set up an upside down lawnmower. And you drop things onto it. In this case, I had back when I, before I'd sent out all the units for promotion, I set up four of them at different angles. So I shot about 10 shots of that and I'm going to be editing them and uploading them.

David Kronstein Tesla500: Why a lawnmower? Is it just because it's not a blender? Is that kind of the idea? Yeah.

Chris Gammell: Different. The whole thing came about when I got a Phantom V4, which is an old high speed camera from the early 2000s, did a teardown, was able to get it working. And I thought, I need something to demo this. So I set it up in the garage with this old lawnmower and dropped a few things onto it. And that video took off massively. So I thought, oh, I'll separate this out into another channel just because the type of people who would often subscribe to that are not the type of people who would appreciate teardowns. And I want to keep Tesla 500 the way it is. Gotcha. That's a good thing.

Dave Jones: Because it's the momentum, the physical weight in the blades, the momentum, like those things will not stop.

David Kronstein Tesla500: Yeah. You know, it's like the destruction. Teenage Chris would like to disagree with you. I have broken a couple of lawnmowers in my day.

Speaker ?: Right.

Chris Gammell: I'm surprised. I have not broken this lawnmower yet. Yeah. It's incredible.

Speaker ?: It's incredible.

Chris Gammell: It's off to, I think it's Black and Decker. Try a boulder. That was, that'll do it. Curbs also. Curbs, boulders, anything concrete. The motor is fine. It's the, this mower was thrown out on the side of the road because the rectifier burnt up.

David Kronstein Tesla500: Oh, really? Ooh. Well, you're, you're, uh, their loss is your gain, right? Yep.

Dave Jones: I'm surprised you've only got 2,500 subscribers for a channel where you destroy stuff. What's going on there?

Chris Gammell: That's just very, very un-YouTube like.

Dave Jones: Yes.

Chris Gammell: Yeah.

Dave Jones: Usually destruction is guaranteed. Yeah. And they're really good videos. And they're really good videos. Yeah. Oh yeah. The hydro.

David Kronstein Tesla500: So Tesla 500 though is mostly, mostly teardown. So if people are interested, Tesla 500 is mostly teardowns.

Chris Gammell: Teardowns. Teardowns, projects. Yeah. It looks like some machining stuff. Machining stuff. Sort of a general like tech channel.

David Kronstein Tesla500: Yeah.

Chris Gammell: That's great. Cool.

David Kronstein Tesla500: That's our kind of stuff. I think, I think everyone should definitely subscribe. Yes. One of my faves.

Dave Jones: Although they probably already are. Probably, yeah. Well, most of our audience are probably. We've only got 10,000 listeners. So 25,000. Yeah.

David Kronstein Tesla500: That, that number works out pretty much. So.

Dave Jones: Yeah.

David Kronstein Tesla500: Anyway. David, what is your, so what is your background?

Chris Gammell: Um, I did, uh, electronics engineering in school. Uh-huh. Uh, yeah. Sort of started, I was always into electronics. My dad showed me projects. I, we were building counters and things using old 7400 series chips back when I was probably 10. Sweet. That sort of got me into it. Yeah. Went to building all sorts of projects. Went to school for it. And then just sort of, I was initially working in battery chargers. And then I sort of, this high speed camera stuff got me a few other jobs in sort of digital design.

Dave Jones: Oh, so your online high speed stuff got you jobs? Uh, yes.

Chris Gammell: Like I.

Dave Jones: People actually saw your channel and.

Chris Gammell: They saw my channel. Um, I think one person on The Amp Hour actually, uh, contacted me through The Amp Hour. Yeah. Networking.

Dave Jones: Yay.

Chris Gammell: Works really well. Hey, that's great, man.

Dave Jones: And they said, hey, we like your videos. Come and design some stuff for us.

Chris Gammell: Is that right? I dropped, hey, here's what I'm working on. Here's my prototype. And they absolutely love it. Yeah. Going in with, to an interview with things to show is the best way to get a job by far.

Dave Jones: Even better than that is to already have your name established via a blogger, a YouTube channel or something like that. Yes. So they can check you out. They know you before you even walk in the door. Yeah.

David Kronstein Tesla500: Oh, definitely. Yeah. Nothing's better than, I've actually, I've actually asked questions of like, not questions of the person, but like, you know, interviews goes well, resume looks good. And you're like, I just can't tell. You can't argue with a YouTube channel. I'm sorry. No, no, exactly. You can actually see what they're doing.

Chris Gammell: It's not, it's really hard to fake that. It's kind of obvious. Right.

David Kronstein Tesla500: Same with hardware. Like you said, I mean, that's, that's true. You bring that into the meeting. There's really, I mean, I guess you could have someone else's hardware, but that's pretty obvious.

Chris Gammell: Yeah. And you can answer questions.

David Kronstein Tesla500: Yeah, exactly.

Dave Jones: That's good. So who was that, by the way? Is that a secret or is it? No, it was. Is it on your LinkedIn?

Chris Gammell: Who was that? I don't remember. The guy who did a lot of radio stuff. I think he's been on a few times. Greg? Shavott? Greg, yes. Yes. Greg Shavott.

Dave Jones: Oh, Greg Shavott. Okay.

David Kronstein Tesla500: I haven't talked to Greg in a little bit. We might have to have him back on. Yeah. Maybe we'll get back on. I guess what he's doing now.

Chris Gammell: Yeah.

David Kronstein Tesla500: Oh, he's doing that crazy medical thingy, the medical startup thingy.

Chris Gammell: Oh, yeah. No, LinkedIn. I don't see it there. He's doing other stuff now. I'm curious.

David Kronstein Tesla500: Yeah, he got pulled off to other things. So we'll try and catch up with him soon, I'm sure.

Dave Jones: Very busy guy.

David Kronstein Tesla500: Yeah. Cool. But he's in Connecticut and you are up in Vancouver, right? Yes. So you were just doing remote stuff for that or what?

Chris Gammell: I wasn't. No, I ended up not accepting it. It wasn't really ready at the time to move across the country.

David Kronstein Tesla500: Right. I got you. Okay. Cool. So I was going to ask, what's the scene like in Vancouver? I've sadly never been. I've heard the city's beautiful, but what's the electronic scene like?

Chris Gammell: It's decent. Nowhere near Silicon Valley, but you can get decent jobs. My one complaint is the pay. In Silicon Valley, the pay is reasonably, it compares well to the cost of living or much better than Vancouver. In Vancouver, you probably make about half as much as Silicon Valley, but the housing is just as expensive, which is really annoying. Oh, okay.

Dave Jones: And are most of the jobs in the center of Vancouver? Is there like a business hub where it all happens?

Chris Gammell: Most of the electronics jobs, at least the ones I've been at, have been more in the suburbs rather than in the downtown core.

David Kronstein Tesla500: Yeah, we've talked about that kind of stuff before, Dave, right? I mean, just like that's where the factories can be built. That's where you can get more space cheaply. Exactly. Exactly. What kind of industries are up there, I mean, in general?

Chris Gammell: Let's see. There's a lot of software stuff. EA is here. Oh. Oh, right. Yes. I think it's Microsoft, Amazon, eBay offices. Oh, so it kind of bleeds over from the Seattle. It sort of bleeds over, yeah.

David Kronstein Tesla500: Okay. Interesting.

Dave Jones: In terms of electronics. So those companies need to have a physical presence in Canada, so they just hop over the border to Vancouver. Is that the city of choice? Yeah.

Chris Gammell: I think they also do it for something to do to get foreign workers in is what I've heard.

David Kronstein Tesla500: Oh, yeah. Oh, okay. Right.

Chris Gammell: They get them to Canada, then they get them to the U.S.

David Kronstein Tesla500: That makes sense. Yeah. Immigration to... I know a couple of people who kind of did that route. Like, I talked to people that are from other countries in California. They say they came in that way, too, as well, which is great. Yeah.

Dave Jones: Hmm. Cool. But there'll be a wall soon, won't there? Oh, please. Please, Dave. Don't stoke my angst. I have to push the buttons. Oh, God. Anyway, so, yeah, so you've worked as an FPGA engineer, engineering physicist. Ooh. That was, yeah, that's one of their interesting titles.

Chris Gammell: What does it mean? I'm actually not sure what it meant, but it was just...

Dave Jones: It sounded impressive.

David Kronstein Tesla500: When your project doesn't work, you drop it, and then you measure how long it takes to hit the floor, and you're like, yep. Yep. How badly it breaks. Parity is measured at 9.9 meters per second, you know, some margin of error. We're fine. Yeah.

Dave Jones: Engineering physics for the win, yes. That's right. Yeah.

Chris Gammell: Yeah, so I started in doing power supply design at Delta Q, which is a golf cart battery charger manufacturer. Oh. So working like one kilowatt range power supplies. Nice. Oh, right. Serious business, yep. Yeah. Yep. Then I moved on, actually, from all the stuff I'd done with the camera to this company called Verisant. They were doing this skin cancer detection system, which was sort of a laser RAM and spectrometer that basically shone a light on your skin. Right. Looked at the spectrum of the reflection. I was doing the control electronics for that. That sounds cool. Got it.

Dave Jones: Was there a lot of red tape because that was medical? That's not medical, though. That's just like skin detection. So it's not really that medical. It's not like an implant or anything.

Chris Gammell: No, because it wasn't. We didn't have FDA approval. That was a big thing. We never really got done before I left. But apparently it's an order of magnitude more difficult to get approval in the US than it is most other countries. Right. We had European, Canada, Canadian. I think we got Chinese approval. Yep. FDA is just so hard.

David Kronstein Tesla500: That's a terrible process. I think Alicia and Chris have talked about that before on their show, about just the paperwork and all that crap you have to do.

Dave Jones: It's not necessarily a bad thing, though, is it? But does it hinder startups and stuff like that?

Chris Gammell: It probably does. Especially this is not really... This doesn't endanger anyone's life. No, that's right. It's just a skin detection. Yeah.

David Kronstein Tesla500: That's right. I had a friend that worked at a sanitizing thing where it was just like a steam sanitizer. And he talked about how... It was for medical tools and stuff like that. But he said the interesting thing is that you just can't change the code, right? For something that is so made to be changed.

Chris Gammell: Yeah. Yeah. That's one of the big concepts. Yeah. Like whole new approval for every code change. Yeah. Right. Just ridiculous.

Dave Jones: Yep. Have you been following the Theranos, if that's how you pronounce it, Theranos debacle? I don't believe so. Oh, Theranos. Anyway, we won't go into that. Google it, people. And yeah. They're going to make a movie about it, apparently. Or they did. They were going to make a movie about how this medical tech startup was valued at billions of dollars. And then it just completely closed. Oh, that one. Yes, I heard of that. Because it was a fraud. It was a fraud. Yeah. Yeah.

Chris Gammell: And it's like they got it so far as a fraud.

Dave Jones: I know. It's brilliant. Yes. A round of applause for Theranos. That's the...

David Kronstein Tesla500: Yeah.

Dave Jones: Faking it until you don't make it. Yeah.

David Kronstein Tesla500: So in these things, though, like, so you were doing... So almost all this is FPGA. Well, you said the first one was power, but...

Chris Gammell: The first one was power. Yeah. Then I moved on to... Yeah. It was sort of designing like a control board. It was like a touchscreen with an FPGA running in an embedded processor. And it talked to a bunch of things like a laser and a camera and an embedded PC and everything.

David Kronstein Tesla500: Okay.

Chris Gammell: That was interesting.

David Kronstein Tesla500: So all self-taught on that side of things? Like the FPGA side of things or what?

Chris Gammell: Pretty much, yeah. That's all the stuff I learned, but I said, screw it. I'm building my camera.

Dave Jones: Right. Okay.

Chris Gammell: That's right. Yeah.

Dave Jones: That's always the way. You have a project and that forces you to learn stuff. Yes. You know, you have a goal. Yeah.

Chris Gammell: It works every time. Not just, oh, you'll bang this information into your head in school without telling you what it's used for. No, do something real. Yeah, exactly.

Dave Jones: A real app.

Chris Gammell: Yeah.

Dave Jones: That's right. Now, you stopped working in July. So are you now full-time at Chrono Technology?

Chris Gammell: Pretty much doing this full-time. Yeah. I've sort of slowed down on some contract work I was doing. It sort of lines up with this. Okay.

Dave Jones: It was contract, was it?

Chris Gammell: Yeah.

Dave Jones: Right. Yeah.

Chris Gammell: Hmm.

Dave Jones: So that, would you have been able to do this Kickstarter if you had a, like a full-time day job, nine to five day job?

Chris Gammell: It probably would have been delayed into the new year and I would have been pulling my hair out the whole time.

Speaker ?: Yeah. Right.

Chris Gammell: Yeah. It was, back when I was actually doing more full-time work, it was really, really stressful. I was, I was not in the best situation. I really had to reduce some load somewhere.

David Kronstein Tesla500: Yeah. Yeah.

Dave Jones: Got it. Yeah. It's tough too.

David Kronstein Tesla500: I mean, just like the mental load of, you know, like if you're working all day, working through problems and then, it's not like this is small stuff when you're getting home either, right? Yeah. This is tough.

Chris Gammell: It's not fun anymore. And that's when, that's when you sort of just want to quit. Yeah. Yeah.

Dave Jones: Yeah.

Chris Gammell: That's too bad.

Dave Jones: Well, let's talk about the camera and the Kickstarter and everything else, shall we? You're currently up to 359,000 Canadian, those weird Canadian dollars.

David Kronstein Tesla500: Weird Canadian pesos.

Dave Jones: I love Canadian dollars.

David Kronstein Tesla500: That's how I used to go gamble when I was 19.

Speaker ?: Right.

David Kronstein Tesla500: 1.6 Canadian dollars to the US dollar when I was 19. Yep.

Dave Jones: And your goal was 65,000. Now, was that an artificially low goal so that you could, you know, so that it would help the psychology behind the whole hitting the target kind of thing with people?

Chris Gammell: I guess a little bit. I mean, I can sort of build these in any quantity. I can just build a batch of whatever size. I figured I didn't really want to do it if it was below about 20, though, just because it did one time.

Speaker ?: Oh, got it.

Chris Gammell: Okay.

Dave Jones: So you priced at that. Right.

Chris Gammell: Yeah.

Dave Jones: Got it. And how many have you built so far?

Chris Gammell: I've built 12 of these current prototypes for various, for testing and for the send out for promotion.

David Kronstein Tesla500: Okay.

Dave Jones: And these aren't just prototypes either. These are basically pre-production. Really?

Chris Gammell: These are basically the final production version.

Dave Jones: Yeah.

Chris Gammell: Nice. Wow. That's great.

Dave Jones: And that is so unlike most Kickstarters. You know, it's like, I've got a 3D render and a breadboard and...

Chris Gammell: Yeah. Like most Kickstarters launch, like on the Kickstarter page there's a picture of the prototypes and there's a 3D printed case version. Most Kickstarters would probably start with that. Yeah.

David Kronstein Tesla500: Step one, get injection molding made.

Chris Gammell: No. Wrong. If the quantities keep going up, I'm going to have to start looking at die casting rather than machining.

Dave Jones: That is what I want to talk about. At what point, at what quantity of volume do you hit the valley of death or whatever you want to call it, where it's uneconomical or too hard to make them yourself in using your current techniques? For money. And not enough money to yet commit to, you know, having Foxconn in the US make them for you. Foxconn in China make them for you.

Chris Gammell: That's sort of a tough one. In terms of making them in-house, one of the most difficult things is the cases because my CNC machine is pushing it to the limit.

Dave Jones: You still made them yourself.

Chris Gammell: I made those 12 myself, but I probably wouldn't do it again because my CNC machine is old. It took a long time.

Dave Jones: Right. But you could just farm that out. You have the files, right? Yes. Yeah. I'm just going to farm that out. Right.

David Kronstein Tesla500: So why die cast versus CNC then?

Chris Gammell: That is probably going to be a decision based on the quantity. I'm thinking just based on some quotes I'd gotten, something around 300 units might be the breaking point for getting a die cast mold made.

Dave Jones: And you currently got 121 backers.

Chris Gammell: Yeah. I think actually in terms of actual cameras, it's like 75 or 80. The rest are just lower level rewards.

Dave Jones: Oh, right. Of course. Right. Okay. So you're about a third of the way there. Maybe a quarter of the way there. And what does a die cast mold cost roughly? What order?

Chris Gammell: In China, ones I was looking at were something around, for this size would probably be something like 30 to 50 grand.

Dave Jones: Oh, okay. So serious money.

Chris Gammell: Once you've got it, it's like $5 or $10 per part. So it's dirt cheap.

Dave Jones: Right.

Chris Gammell: In terms of CNC, I was quoted for the, for this, the entire case set is something on the order of $100 per set.

Dave Jones: Oh, okay. So 20 times less for a die cast.

Chris Gammell: Something like that.

David Kronstein Tesla500: Yeah. Yeah. But you got to amortize all that then. Oh yeah.

Dave Jones: No, of course. For your 50,000.

David Kronstein Tesla500: Could you, I don't actually know, I should probably know this, but what is, what is die cast process? I don't even know what that looks like. Is that like poured in or what?

Chris Gammell: It's basically just like injection molding, but they inject metal at high pressure instead of plastic. Oh, wow.

Dave Jones: Is it a high temperature too, or is it just high pressure?

Chris Gammell: I think it's, well, some maybe a hundred or two, 200 degrees Celsius above the melting point, but it's injected into pressure so it can get into all the little crevices before it silnifies.

Dave Jones: Got it.

David Kronstein Tesla500: That's crazy.

Dave Jones: But they can churn them out for, you know, cheap as cheap.

David Kronstein Tesla500: So like those, those Hammond project boxes, are those like die cast?

Chris Gammell: Those, the metal ones are die cast. Yes. Okay. Okay. So that's like a good example. They're extruded, but most are, like the bowl shaped ones are die cast. Yeah. Right.

David Kronstein Tesla500: The ones that have like the screw on flat top type things. Yeah. Okay. Well, that's a, that's a good reference then.

Dave Jones: Is, is, is, is the finish better with a die cast one?

Chris Gammell: I would say no. It's the, it's hard to beat machining as a finish. Oh, okay. Powder coat it then, right? I mean, you can always do that.

Dave Jones: Oh, right. You can powder coat it and then you'd have the nice smooth finish.

Chris Gammell: Although die cast offers several advantages and you could like have recesses for nice rubber grips and things. And you can do much more advanced shapes that would be, have a much better feel in your hand.

David Kronstein Tesla500: Oh, it's like overhangs and stuff like that. You could, you can do more.

Chris Gammell: Things that are very expensive to do on machining are much more. Oh, curves. Yeah. Yeah. I guess you could do that too. Compound curves. Yeah.

Dave Jones: Got it. Now, if you were. You get real fancy. Let's assume for a second that you were one of these, you know, idiot Kickstarters who have their 3D render and, you know, not much else and not much of a clue. You know, you haven't been working on it for 10 years. If you then got your million bucks on your Kickstarter, how long would it take to actually design this sort of camera? Like if you had to like farm it out to a, you know, as, as most of these people do, they just go, oh yeah, I'll just farm it out to China and they'll design it for me. How would it, because it, you know, it's, it's complex business. All this high speed capture stuff and storage and processing and everything else.

Chris Gammell: Yeah. If they went into it with no, nothing, no base or experience, it would probably take two or three years.

David Kronstein Tesla500: So that would be, is that if they're hiring you, it would take two or three years knowing

Chris Gammell: what you know now? I'd probably do it a little bit quicker, but if you hire just a typical team who had not done it before, it's probably around that, maybe even a bit longer.

Dave Jones: Okay. Yep. Wow.

Chris Gammell: Yeah, that's nuts.

Dave Jones: So that's where it pays to. The experience really pays. Yeah. The experience pays and having a real, not just a prototype, but a, basically a production ready prototype, pretty much. Yes. That, you know, if all things fail, you could go to production with exactly what you got now and most backers would be, you know, entirely happy. Yes.

Chris Gammell: Yeah. Are there any tweaks you're going to do? I'm going to do a board spin. There's a few issues with the battery that I didn't expect. The battery, some, some of these batteries I'm using. Pain in the ass battery. Yep.

Dave Jones: All this high tech, high speed capture stuff. Not, not a problem.

Chris Gammell: Batteries. Yeah. Some of the most annoying things, the hardest things I've dealt with. Um, one is the, one of them is the battery, which some batteries don't report voltage properly. So like I expected them, the one I'd reverse engineered to, uh, the battery product, BMS protocol on reported all that. And then the new ones I got didn't. So I did, that's why the prototypes currently don't have a battery monitoring working. Yep.

Dave Jones: I noticed that in my prototype. Yeah.

David Kronstein Tesla500: You know, I, I hear you can get, you can get really a low cost, uh, Samsung note seven batteries these days. Oh yeah. The problem is in five years, can I still buy that? Oh, well I meant they'll, they'll explode.

Dave Jones: So it was smart to use an off the shelf, uh, you know, uh, existing battery pack. That's pretty much always going to be available. Right. I agree. That was the shape and size.

Chris Gammell: Modern, modern DSLR is like, I use it like a 20 megapixels. That's going to be available for a long time.

Dave Jones: Yep.

Chris Gammell: What else was difficult to the, the, the other thing to do with the batteries, the contacts for the battery is the metal. Yeah.

Dave Jones: How did you get those?

Chris Gammell: The, the ones in the prototypes are just a certain spade terminal. Oh, okay. Right.

Dave Jones: It just happens to fit. Okay. Yeah.

Chris Gammell: And that's, that's what I'm, uh, we'll have to use for now, but if the quantities are high enough, I'm going to actually spring and have a custom connector made because there's just no off the shelf connector that fits these batteries.

Dave Jones: Right.

Chris Gammell: It's just like a blade contacts, 2.5 millimeter pitch. And there was, there was one on Ali, Alibaba that was close enough, but I emailed them and it's like, no, sorry, we don't have that anymore.

Dave Jones: Oh, damn it. But the current, because you could probably get a bulk buy, you know, if you found one, you'd go, right, I'll buy a hundred thousand of those please. You know?

Chris Gammell: Yeah. But yeah, the, the, the terminals work just fine. And I just have to make up a jig to position them properly when, uh, building them.

David Kronstein Tesla500: Got it. So I didn't, uh, I didn't actually get a chance to watch Dave's teardown yet. Um, could you kind of.

Dave Jones: I haven't done a teardown. It's a review.

David Kronstein Tesla500: It's a review. Oh, I'm sorry. Um, so David, could you, could you just kind of walk us through like how, how many boards, what's the assembly like? I mean, is it like stacked boards or is it single board?

Chris Gammell: It's essentially one main board with, uh, either direct connectors or FFC cables going to sub boards as one for some connectors, one that has the power switch and microphone, uh, one that, one that has the image sensor and one that has an encoder and speaker on it. Yeah. I designed it specifically to be really easy to assemble. I hate crimping wires. I hate buying expensive crimpers. Yep. So I designed all of those out. That's great. That's really smart.

David Kronstein Tesla500: And so, and it also has the monitoring screen in the back. So that's why it needs the speaker. That's why I need some of these, the flat flex, like you mentioned, like these are consumer level type things, right? I mean, these are.

Chris Gammell: They're all super low cost, super easily available.

David Kronstein Tesla500: Yeah. So are you, you're designing for just whatever is the most common you can get or how did you decide? Pretty much.

Chris Gammell: What's, what's common? What makes it easy to assemble? Yeah. Just try to make it, make it easy, make it not, not a pain to get out.

David Kronstein Tesla500: Okay. So, uh, so you're talking about quantity for the boxes. And getting to the, uh, the die cast would be nice. Um, any chance you're getting to quantities on the FPGA or any chips you have on board? I mean.

Chris Gammell: The quantities for decent discounts. Are you getting any kind of discount at all? We're only talking about a couple hundred. Yeah. There's not going to be any huge, any, there's not gonna be any huge discounts on parts. Mm-hmm. In terms of assembly, I have a supplier in China called Our PCB who has been very good so far. I think for these boards, like the main board on this, which is about a thousand parts and about 130 unique bomb line items. It's about a hundred dollars assembly.

Dave Jones: A hundred bucks for the assembly.

Chris Gammell: That was for, um, quantity 12. And it gets down to. Oh, wow. Yeah. I'm sorry. A hundred bucks each for assembly quantity 12.

Dave Jones: Yeah, yeah.

Chris Gammell: And then it gets down to.

Dave Jones: That is cheap. That is cheap.

Chris Gammell: And then it gets down to, I think it was 60 if I ordered a hundred.

Dave Jones: That is, that is still seriously cheap for that sort of low volume. And we're talking serious parts here. I mean, you know, it's, these are loaded on both sides. We've got high density, um.

Chris Gammell: We've got like a 600 pin BGA. 600 pin BGA. A 700 pin BGA. Yep.

Dave Jones: Oh, oh, oh. That's fun. That's serious business. Yeah. So that is, yeah, that's seriously good pricing. Yeah.

Chris Gammell: I've been really impressed by them. Like they, they said, oh, we failed extra inspection on one. We're reworking. We're holding in a few days. They got, I got them all back. There was only one rework required. One solder bridge on one cap. Wow. So I was really impressed.

Dave Jones: Yep. That's awesome. Now, architecture wise, I'm going to, please correct me if I'm wrong here, but image sensor, 16 LVDS lines coming out.

Chris Gammell: Yep.

Dave Jones: That goes into a, uh, EPC5 FPGA.

David Kronstein Tesla500: Uh, yep. That's right. Yeah. ECP5 FPGA. What is that? Sorry. Lattice ECP5. Oh, it's a lattice part. Okay. Yeah. Lattice. Why did you choose that? We'll talk about that in a sec. Oh yeah. Sorry. Yeah. That's, that's good. Keep going. Sorry.

Dave Jones: So people can visualize the architecture. I'm making hand gestures at the moment. It's a signal flow. And, um, that then, uh, so that into the FPGA, the FPGA does all the high speed, uh, decoding into that and storage into the, uh, DRAM, the high speed DRAM.

Chris Gammell: Yep. That's right.

Dave Jones: And then it encodes a video format. And sends the video format to the processor, which is like in a standard video format, which is a, uh, TI DaVinci TMS 320. Yeah.

Chris Gammell: DM8148. DM8148.

Dave Jones: DM8148. They've got 10 million variants. Yeah. So, yep. We'll talk about that. And then that does the conversion to H264 and drives the LCD as well. And everything else, user interface.

Chris Gammell: Yep. Yeah. It runs Linux, drives the UI.

Dave Jones: Yep. Nice. Got it. And that's, and that's pretty much it. It's a two main chip solution.

David Kronstein Tesla500: Yeah.

Chris Gammell: Yeah.

David Kronstein Tesla500: Really simple. Well, three, right? Because the, the sensor. Sensor. Yeah. Yeah. Oh, well the sensor, of course. Yeah. Of course. Okay.

Dave Jones: So let's start with the sensor. How, which sensor are you using? Was it easy to get? Is it like in low volume, all that sort of stuff? How did you acquire? Yeah.

Chris Gammell: It's the Luxima Lux 1310.

Dave Jones: Yep.

Chris Gammell: Yeah. That was surprisingly easy to get. Like there, I really have to, hats off to Luxima. They're really easy to deal with even as a little guy.

Dave Jones: Nice.

Chris Gammell: Like I, I think some of my emails to them were even copied to, I believe the guy who designed the silicon on the sensor. Oh, wow. Yeah. So support was really good. That's great. Wow. Yeah. I wish all semiconductor manufacturers were so easy to deal with.

Dave Jones: That's it. Is that because they're small niche that they're able to do that?

Chris Gammell: It's a pretty small company.

Dave Jones: Right. Okay. They obviously don't fab them themselves. They fab out the sensors to.

Chris Gammell: Yeah. They come from Taiwan. I think it's one of the manufacturers out there.

David Kronstein Tesla500: Hmm.

Chris Gammell: Yeah.

David Kronstein Tesla500: I never, I don't know. I don't know. I don't know anything about camera sensors.

David Kronstein Tesla500: Yeah.

Dave Jones: Who would manufacture the camera sensors? Do you, was it, would it be Sony or would it be like, like you can't just go to one of your regular chip fabs and presumably. And fab.

Chris Gammell: Maybe you can. I'm actually not sure about that.

Dave Jones: Hmm. Interesting.

Chris Gammell: Probably some special process required to make efficient sensors though. I would imagine.

Dave Jones: Totally. Yeah. I'm sure. Um, which I didn't think your regular chip fab could do, but I could be wrong. Chris, you're from a chip fab.

David Kronstein Tesla500: I know I should know this stuff, but I don't. Yeah, I know. I don't. I bet you don't. If anyone knows, maybe leave a comment. Yes.

Dave Jones: Anyway, so you were able to get those parts in onesies, twosies. Did they give you free samples or did you have to pay? I didn't know free samples.

Chris Gammell: I did have to pay for them. But yeah, it was easy to deal with.

Dave Jones: Nice. Yeah. And did you have to sign an NDA or anything stupid like that?

Chris Gammell: Nope. The, like the data sheet's not public, but they just email them and they send it to you. And other people have had the same. Yeah.

Dave Jones: That's great. Win. Oh, we'll have to get them on the show.

Chris Gammell: That's how they run a semiconductor company.

Dave Jones: Yeah. So. That'd be interesting actually to get the chip designer on the show. Get them on.

David Kronstein Tesla500: Yeah.

Dave Jones: Yeah. Talk about CMOS sensor design. Hmm.

David Kronstein Tesla500: David, what about the, so the, so the, the thing I never understand about the sensors is, so it's 1.3 megapixels, which is a, can you explain like just the, the, the measure of what that is versus the resolution versus what the output image can be? Cause I never quite, it seems like it, it doesn't seem to match up with what, whatever is coming out versus like they get oversampled or something.

Chris Gammell: Um, no, well, yeah, the, the imager or image sensor obviously has a pick an array of some size and you, you always, you can either read out that full array and get the full resolution video out. It's just usually sampled from top left pixel to the right and in a normal raster fashion. And in these high speed sensors, you can usually also window it down to smaller resolutions and read the same pixel rates. You get a higher frame rate. So that's how we get the higher frame rate.

Dave Jones: Right. And that's why you get the, you know, the 600 by 96 resolution or whatever it is for your 21,000 frames per second. So it's that long landscape narrow thing. Now I've always wondered why, okay, you window it down, but why can't you have like separate windows and then separate clocks out? Is it just the architecture of how the sensor works or?

Chris Gammell: Yeah. There's basically the, as you read out the chip, there's some, you select which horizontal, uh, row you want to read out that gets put onto each. There is, in this case is the pixel, the imager is a hundred, is a thousand 24 pixels, sorry, 1280 pixels wide. Yeah. There's 1280 parallel lines. And you select, say you select the first row, those 1280 lines get driven with the different voltages for each of those pixels. Yeah. And then those are, there's MUXs that select groups of 16 to go off to the ADCs. Right. I think in this sensor there is, the ADCs run at 90 megahertz and there's 16 of them. Oh my God.

Dave Jones: Why can't they just increase the number of ADCs? Why can't they just have like a thousand ADCs on there and have a thousand separate lines coming out and you'd get a hundred thousand frames per second?

Chris Gammell: Some cameras do something like that. Some other sensors they make have column ADCs, which is just one ADC per every column.

Dave Jones: Per column. That's what I'm talking, that's exactly what I'm talking about.

Chris Gammell: The downside of that is you can't trade horizontal resolution for speed because you're limited by the rate of those ADCs, like one line every clock. Right.

David Kronstein Tesla500: Huh.

Dave Jones: Interesting.

David Kronstein Tesla500: So did you play, I'm looking at the data, I'm looking at the, we'll link in the thing to the Lux 1310, but the, it says frame rate up to 1 million frame per second at 1280 by one pixel.

Chris Gammell: I'm not sure where that number came from because I can't get it to run at that rate. It seems really, what did you even do with that? Like that line, there it is. Yeah. It's going to be useful as a street camera if you just want to see the horizontal motion of things.

Speaker ?: Yeah.

David Kronstein Tesla500: Oh, okay.

Chris Gammell: You can buy special street cameras that are just some width by one.

Dave Jones: Yeah, single pixel. Yeah. You can buy a one, if you can buy a one pixel sensor, 128 by one pixel sensor. They use them on scanners and other applications like that.

Chris Gammell: Yeah. Document scanning, things like that.

Dave Jones: Yeah. Document scanning and stuff like that. That makes sense then too. Yeah.

David Kronstein Tesla500: So if you want to frame, if you want to, if you want to scan 20,000 pages per second, you're in good shape. Yeah. Good luck with your motor drivers.

Dave Jones: All right. So sensor, any thermal issues with the sensor? What is, I think I haven't, haven't got the data sheet. It draws a couple of watts or half a watt or something.

Dave Jones: Yeah. I initially.

David Kronstein Tesla500: Two watts at full resolution.

Dave Jones: Two watts. Okay. At full res. Yeah.

Chris Gammell: When I built the first camera, I was like, yeah, this might get somewhat hot. So I put a hole in the back of the board to connect that directly to the back of the sensor. And that was very useful because yes, it does definitely need heat sinking to keep it at a reasonable temperature. Right. So in the camera, there's this.

Dave Jones: What happens if it gets too hot?

Chris Gammell: I've run it at full clock without heat sinking. It was okay. But on one sensor, when I was first bringing up this new camera, I forgot that there is an internal PLL. Oh. So I had, I accidentally put 270 megahertz into it when it should have been 90. And that sensor now has some dead lines on it.

Speaker ?: Oops.

Chris Gammell: Yeah.

David Kronstein Tesla500: See, now could you capture the smoke being let out of the chip at that super high rate? Because then we could, it's like a recursive thing, you know?

Chris Gammell: It wasn't actually any smoke, but you could.

David Kronstein Tesla500: You never know.

Dave Jones: So if the chip could capture smoke release from its own sensor.

David Kronstein Tesla500: Yeah. Hmm. You could line them all up and you could just destroy like thousands of dollars worth of cameras at once, you know?

Dave Jones: Nice. Now, how does this perform in light wise? Is it particularly good? Is it particularly meh?

Chris Gammell: It's reasonable. It's better than the, the previous version of this camera used another sensor called Lupa 1300-2. That was about, it's twice the pixel size and it was what half the speed, but this new sensor is about the same sensitivity, even with much smaller pixels. So I'm actually pretty impressed with it. Nice. The sensitivity, how do I describe this in terms people will understand? To shoot, let's see, to shoot in 60 frames a second, like 160th of a second exposure takes about 400 lux to fully saturate the, to fully saturate the image sensor.

Dave Jones: For those who want to know, that's 400 lux is average office lighting probably? Yeah, that's good. Yeah. Yep.

Chris Gammell: Okay. Yeah. Yeah. So with, with no gain, you need about something on the order of 10,000, 8 to 10,000 lux to run at full speed. Which is, how many light bulbs? Is it like studio lighting kind of thing? Something pretty good studio lighting. Yeah.

Dave Jones: Yeah.

David Kronstein Tesla500: Or how about, so we, Ben did a demo video too. And I saw, I saw, I saw that one. Sorry, I didn't watch Dave's. But the, I saw that. So he had like a 600 watt bulb or something like that or something. Wasn't it? Yeah.

Chris Gammell: 600 watt incandescent. Yeah. Yeah.

Dave Jones: Yeah.

Chris Gammell: Yeah. Is it the lighting, unless you're shooting really high resolutions, it's not too bad. Like you can just about do in, like in the decent lighting have in the garage with a little bit of gain, you can do decent shooting just in the room lights. Okay. Right. But to get really good images, you need lots of light. So, and what is, what is the physics reason for that? Is that just the, um. It's just the number of photons hitting the sensor. So just charging up the capacitor. Yeah. Charging up the capacitance. There's just a physical limit there that you just can never get around.

Dave Jones: And of course you can increase the gain, but then you get the noisy image. You get that pixelated noise on, on, on your image. That's just.

David Kronstein Tesla500: So it's like a low level DC problem, but just done really, really quickly.

Chris Gammell: Yeah. It's just like trying to capture the most data you can in the least amount of time. The less time you have, less you can capture. Yeah. Huh.

Dave Jones: So it's the photon to noise ratio. Is that.

Chris Gammell: Yeah. It's just. Problem. It's not the correct term. Yeah. Yeah. The size of the pixels, their quantum efficiency, which is what. Percentage of photons actually get turned into electrons. And how fast you're. How long your exposure time is. Those three things sort of define the sensitivity. Huh.

Dave Jones: Now, what is the difference? This is a CMOS sensor, but a lot of people are familiar with CCD sensors. Have you used those in the past? And what's the difference?

Chris Gammell: Yeah. I've used CCDs in the past. They are better in a lot of ways. Uh. The CCD is quite different. The CCD sensor works by having this, you have a silicon substrate with these little, uh, electrodes across it. And you set the voltage on these electrodes to create, like, potential wells. So that when a.

Dave Jones: So they have physical wires. Physical little. Yeah.

Chris Gammell: Little wires going across the chip. Right. And then different doping done in the, in the silicon. So that when a electron, when a photon hits the chip, it knocks electrons free. And those electrons are just, just sit there, held in place by the electric field of these wires. And by how, however, the chip is doped. Huh. And then you move, you change, you, uh, pulse these wires on and off in a sequence to push the electrons down the chip. And you push them off to the end row. And then you, on the end row, you push them off to the side and off the side to an, to, uh, an ADC.

David Kronstein Tesla500: That sounds tough.

Chris Gammell: Yeah. It has issues, but it generally has a pretty good quality, like better, very good signal to noise ratio.

Dave Jones: So they still use CCD for like space, um, and satellite, um, you know, and space probes and stuff, don't they? Yeah.

Chris Gammell: We used them in, um, at Verisont for this camera because it had, we, the image sensor was actually cooled by a Peltier device down to like minus 40 degrees Celsius to reduce the thermal noise. Wow. Yep. And the CCD really works well at reducing noise when you reduce temperature, but CMOS doesn't work that way.

Dave Jones: Yeah.

Chris Gammell: Yeah.

Dave Jones: Space is cold. So yeah, you've automatically, you don't need to cool it down on your telescope. Expose the back of it to space and then boom. That's actually not true, Dave.

David Kronstein Tesla500: That's not true. No. No. Space is terrible at cooling things. Have you not read about space? Oh, of course.

Dave Jones: Duh. Yes, of course.

David Kronstein Tesla500: Because it's a vacuum. Yeah. Yes. But either way, it's probably still something like that. Yeah. I don't know why it would be good in space, but space is not good at cooling things. I've learned this from sci-fi. No.

Dave Jones: Anyway.

David Kronstein Tesla500: So those are harder to manufacture though?

Chris Gammell: Yeah. They take a very different process. I believe CMOS sensors are a much more standard process that can be done on normal semiconductor equipment. So that's why they're sort of preferred. And you can also integrate all your control electronics, control logic, and ADCs right on the chip where you can't do that with CMOS or sorry, CCD.

David Kronstein Tesla500: Oh, interesting.

Dave Jones: Got it.

Chris Gammell: Cool.

Dave Jones: All right. So we've got our sensor. You hook up 3.3 volts, 1.8 volts, analog and digital. And bingo, out pops the magic data from the LVDS lines.

Chris Gammell: Right. Yes.

Dave Jones: So these go into your FPGA?

Chris Gammell: FPGA. Those are these.

Dave Jones: Let's talk about the FPGA. Sure. Why Lattice and why not the other two biggies?

Chris Gammell: Let's see. I'll start out why I chose Lattice initially. And that was because back at the time, which was 2000, no, 2011, I think, or 2010, it was the only vendor that supported full 64-bit laptop sodiums or computer memory in a low-cost FPGA, like below $1,000. Ah, interesting.

Dave Jones: Right.

Chris Gammell: Yeah. Since then, most of the manufacturers now support that. But I was sort of...

Dave Jones: So they've got hardware support in the chip for those SODIMMs. So doing all the refresh and everything else.

Chris Gammell: Not quite. They have hardware accelerated certain functions because DDR3, the minimum clock speed is 300 megahertz, which is really difficult to deal with.

Dave Jones: Oh, okay. I didn't know there was a...

Chris Gammell: That's one of the really difficult things. Okay.

Dave Jones: Minimum. Right. Yeah.

Chris Gammell: DDR4 is even worse. It's about a gigahertz.

Dave Jones: Whoa.

David Kronstein Tesla500: Right. Which is... Which people might be thinking like, oh, well, there's lots of processors that do that, but a lot of FPGAs don't need to go that fast.

Chris Gammell: Not getting IO to run at that speed is a huge challenge. Yeah. Yeah. Right. And power draw and everything else, right? I mean, that's crazy. Oh, yeah. Yeah. Yeah. So they have these special output drivers that buffer two bursts of DDR into four pixels, four, not four pixels, four bits wide to get the data rate down to one quarter. So you can actually deal with it in the FPGA logic.

Dave Jones: Yep. So how much of a hassle was it to get the DDR working on this FPGA?

Chris Gammell: Was it easy? The old one, the ECP3 I was using in the previous version was easy. The ECP5 was much more difficult because it's a new chip. I think I might be the only...

Dave Jones: Does that mean they didn't have example code? They had example code.

Chris Gammell: They had example code that didn't work. Oh. Great. Yay. It took me almost two months to get this working. I might be the only person, even including Lattice, who has this working properly with 64-bit plus. Oh, wow.

David Kronstein Tesla500: You should sell it back to them and, you know, make your money back, you know? I'll give you this. You give me the chips for free. That's what you say, right? Yeah. I should try. Yeah. What is the relative cost on an ECP5?

Chris Gammell: I think for this one, the ECP3 I was using before was about $200. Oh, my God. This is down to, I think, about $70 or $80. Okay. So it's not too, too horrible. It's half the cost of anything else that will do a similar thing. Yeah.

Dave Jones: And how many pins did you need and were you forced to use a high pin count package just because that's all they had?

Chris Gammell: Oh, Dave's touch point. I had to use, yeah. I had to use the highest pin count just because that's what you need to run the 64-bit memory bus. Wow.

Dave Jones: Right.

Chris Gammell: There's a few pins free. Actually, there's about 20 or about 30 pins free. Okay.

Dave Jones: So you did genuinely use all the IO on this thing? Yeah.

Chris Gammell: Like, just because of bank voltages, it is a really strange for IO on this. Oh, interesting. It's pushing this low-end FPGA pretty far, but the benefit of saving like $120 is pretty big.

David Kronstein Tesla500: So what about the, do you have an idea of the, so I only, oops, sorry. I only have a relative measure. I know some of the families in Altera and Xilinx. Do you have like, what does this usually compete against, this FPGA?

Chris Gammell: This would definitely be the lower-end Cyclone, like Cyclone 4, Cyclone 5.

Dave Jones: Okay.

David Kronstein Tesla500: Okay.

Dave Jones: Yep. But the high pin count one, you know.

Chris Gammell: Yeah. I think Lattice was going to release an ECP-4, but they can't, it was going to be higher performance, but they canceled it and then released the ECP-5, which was a lower performance, much cheaper chip, but it seems to fill this market pretty well. Just the price is so much better than anything else.

Dave Jones: Interesting. At $200, the price is so much better. Yeah.

Chris Gammell: Or at $70 or whatever, yeah.

Dave Jones: Did you get this first spin? Because like, you can't just breadboard this thing, right? And like, you know, you've got to commit. Do you have a dip package, please? You've got to commit to like, what, a six or eight layer board or something to fan out all the stuff. And we'll probably talk about that. And like, yeah, did it work first go? Did you have issues?

Chris Gammell: Yep, it worked first go. There were, I made a few mistakes that I was, I was like, wait a minute, that's not right. Oh, no. I swapped those two. I was just going to brick the whole thing. But no, I swapped the two of the bank voltages, 1.5 and 2.5 volts. Yeah. But luckily it still works.

Dave Jones: Okay.

Chris Gammell: All right. But yeah. Rugged process. Next revision, I'm going to correct that.

Speaker ?: Mm-hmm.

Dave Jones: So the camera I've got has got a wrong bank voltage.

Chris Gammell: Yeah. The one of the 1.5 volt IO banks is driven by 2.5 and vice versa.

Dave Jones: That's hilarious.

David Kronstein Tesla500: How many, how many total, because I always hear about the promise for FPGAs of like, they're like, oh, we're reducing our stuff. And then I see designs for these fancy FPGAs and then I'm like, oh no, there's like eight power rails. So how many rails do you have?

Chris Gammell: Let's see. There's 1.2 volt core, 2.5 volt for some analog stuff like PLLs. And beyond that, it's just whatever IO you need. So in this case, I think there's 1.5, 2.5 and 3.3.

Dave Jones: Okay. But you need an additional 1.1 for your processor as well, don't you?

Chris Gammell: The process, actually, no, that's, yeah. The FPGA is now 1.1 volts. That was the old one was 1.2.

Dave Jones: Oh, okay. Yeah. The processor. Okay. So you're able to share that with your processor. processor.

Chris Gammell: Actually, no, the processor needs a variable supply between 1.2 and 1.35, depending on what clock speed you're running at, which is annoying. So we need a whole nother regulator for that. Yeah.

Dave Jones: It needs a variable source.

Chris Gammell: Most of these system on chips, like certainly the ones in your phone, the power supplies are variable voltage so they can get the best power efficiency. Huh. Right. So when they clock down the chip, they reduce the voltage.

Dave Jones: Oh, okay. So dynamically clocks down? In software?

Chris Gammell: Yeah. In this case, I just have it locked. I'm not so, so concerned about power here because everything else is so power hungry. Yeah. So I just leave it running at the max 1 gigahertz and 1.35 volts all the time.

David Kronstein Tesla500: Okay.

Chris Gammell: But in a software that you could potentially implement the proper voltage scaling.

David Kronstein Tesla500: And then do you need another, because DDR requires the half scale voltage. Is that right?

Chris Gammell: Yeah. There's a dedicated regulator that produces the 0.75 volt termination voltage.

David Kronstein Tesla500: Okay. Yeah. That always trips me up too. I'm just like, that's crazy to me.

Chris Gammell: Yeah. DDR is such a pain to deal with. I wish they would just make it LVDS pairs. Like RamBus had Ram that was that. It was beautiful, but no one adopted it.

David Kronstein Tesla500: Because RamBus is a shit company with terrible licensing. Yeah.

Dave Jones: Yeah. So your mid, can you tell us about that bias in for a sec, the mid rail? Biasing?

Chris Gammell: Yeah. So at the end, let's just take an example of the address bus that runs from the FPGA to all the Ram chips. And this is a pretty long, even on those little SODIMs, the total trace length is like six or eight inches.

Dave Jones: Yeah, it's not much. Yeah.

Chris Gammell: And at the end of that, you don't want reflections from your high speed signal. So you put a termination resistor. And to get the minimum power, you connect that termination resistor not to ground or VCC, but to half of VCC. And that minimizes the power dissipation.

Dave Jones: For a power dissipation reason, not necessarily for a reflection reason.

Chris Gammell: Yes. You could terminate them to any voltage. Well, there's some other issues. Like if you terminated it to ground, it would pull the DC level down and then your thresholds would all be all messed up. Yeah. Because it also uses the 0.75 volts as a reference for the input buffer. So they know exactly where half VCC is. So there's minimal variability.

David Kronstein Tesla500: Right.

Dave Jones: Now, did you use resistors for this? Or did you use dedicated end termination chips? Because you can actually get end termination chips for these things. I've used those in the past.

Chris Gammell: The termination is all done on the SODIM. So I actually don't have to worry about that. Oh, OK. Of course.

Dave Jones: Right. Yeah, yeah. Right. Yes. Because you're using the SODIM module. Yeah.

David Kronstein Tesla500: That's like the plug-in. There's a header and you plug in. You can swap out the memories then. Is that right?

Chris Gammell: Yeah. There's standard SODIM sockets.

David Kronstein Tesla500: Oh. That's nice.

Dave Jones: Exactly like the PC you're using, Chris. Yeah. Well, it's like the laptop style. It's not like the wide one, right? No. Well, no. Yeah. Huh. Desktops. Yeah. Well, that's nice too then. So for sourcing. But if you went away from the modules and you used the DDR chips directly on your board, then you would have to roll your own termination solution.

Chris Gammell: Yes. It's not too bad. It's basically just a whole bunch of resistor packs to VTT. Yeah. Yeah.

Dave Jones: Yeah.

David Kronstein Tesla500: Got it. Can we step back to the sensor talking to the FPGA real quick? OK. Sure. So that's LVDS. And I've done one sensor project in the past or image sensor with an FPGA before and it was using CERTA's stuff. Those are different, right? Was it just based on the output of this particular sensor? Like why one of the other advantages or what?

Chris Gammell: In this case, there's not enough CERTA's channels. We just use the input buffers on the FPGA. But you could use CERTA's as well if the FPGA has enough of them. Also, because of the format, the data format, it doesn't really line up with CERTA's. So you have to do some bit of deserialization logic in the fabric to get the proper, to sync up the data. I gotcha.

Dave Jones: For those who don't know, FPGA's typically have a CERTA's, which is a serializer, deserializer hardware embedded on some of the IO pins, not all of them. So if you've got a serial input that you want to convert into parallel, then it can do that for you in its own dedicated hardware block without you having to do it in your FPGA fabric.

Chris Gammell: Yes. And it's also usually much faster.

Dave Jones: Faster. Yes, exactly. But you're saying there weren't enough, so you just didn't use any CERTA's?

Chris Gammell: Yeah, and also the chip that includes CERTA's costs quite a bit more. Yeah.

Dave Jones: Ah, right.

Chris Gammell: Okay.

David Kronstein Tesla500: It's like an upsell they do.

Chris Gammell: If you look at the comparison graph. Yeah. In terms of signal integrity, the LBDS is dead easy because it's a nice terminated differential pair. The difficulty is really in the RAM because that's this multi-drop bus still running at a very high speed. Yeah. Boo. And that's a gig, you said? The DDR runs at 300 megahertz DDR, so 600 megabits per second.

David Kronstein Tesla500: Yikes.

Chris Gammell: Yeah. As long as you follow the guidelines, it seems to be okay.

Dave Jones: And on the FPGA front with all this integrated hardware, for those who don't know, I think we've talked about it before, the dream of FPGA was that it was just a fabric, a sea of gates. And that was it. And you could do whatever you want. You could include the CERTA's in there. You could include a processor. You could throw in any hardware, UART, any hardware you wanted into this thing. And then they realized it doesn't work so well, especially at high speed. So we'll put dedicated hardware blocks of real silicon in there for these CERTA's and other processors. Yeah. And ARM processor, it's not very quick when you implement it in the FPGA fabric. Yeah. Just as an example. You could put in an ARM core.

Chris Gammell: When I had the old version of the camera had a soft core. And that was taking up nearly 40 or 50% of the FPGA. But on this current one, where it uses a separate CPU.

Dave Jones: And ARM soft core? What soft core was it?

Chris Gammell: No, it was just a microblades. I think it was the Xilinx microblades.

Dave Jones: Oh, Xilinx microblades. Yep.

Chris Gammell: And the new one, where it doesn't have the soft core, the whole project is only using up about 15% of that FPGA. Wow. It's so much smaller.

Dave Jones: But you still had to buy that expensive couple hundred dollar FPGA because of the I.O. count. Yes. Or the speed or whatever the features. Yep. Right. Bummer. But you're only using 15% of the space.

Chris Gammell: Yeah. That's great, though. That means you can do field upgrades later. I definitely plan to do some better demosaic algorithms I want to implement. Those are going to take up some. But it's nice to have, it's nice to not be fighting and trying to cram every little bit on. Yep. There was one job I'd interviewed for and they were saying, oh, we have this camera out there and the FPGA is 95% full. I was like, nope.

David Kronstein Tesla500: Got that one walked out of the interview.

Dave Jones: I'm not taking that one. Yeah. I remember I had to roll my own by hand. I had to do my own place and routing by hand to fit everything in to a CPLD. I routed, like I did the auto compiler.

David Kronstein Tesla500: Back in your day was what, 20 gates, right, Dave? You just. Yeah, right. Back in, you actually could do that. A few thousand gates. Yeah. A couple of thousand gates. Yeah.

Dave Jones: You can place a route by hand. And it was, you know, and I fitted the damn thing in. Damn it. So.

David Kronstein Tesla500: Screw you, lattice. David, what kind of stuff is actually inside then? I mean, so you have 16 parallel paths of data coming in. Yep.

Chris Gammell: So those get. Serial. Yeah, serial. And they're converted to parallel. Yeah, those get clocked into a shift register that converts them to parallel. Then they go into some FIFO buffers. And that gets dumped.

Dave Jones: How big are the buffers?

Chris Gammell: They are. They're pretty small. They're. They are 256 bit wide. And they're, I think, 512 words deep.

Dave Jones: And why do you need the FIFO buffers?

Chris Gammell: That's because when the RAM has to be accessed by multiple things, it has to be accessed by the logic. The rest of the data has to be read out for display. It has to be able to refresh. So you always have to be able to pause. And the data coming in has to go somewhere. Oh, wow.

Dave Jones: So if you had. So if you had true dual port memory, you wouldn't need the FIFO.

Chris Gammell: You wouldn't need that. No.

Dave Jones: Yep.

Chris Gammell: That's not. That's not available in DRAM. So that's not. No, that's right.

Dave Jones: It's not really an option.

Chris Gammell: So. Oh, wow.

David Kronstein Tesla500: What else? So if I was visualizing the data from the. It's like a fire hose of data, right? Yep. And then you're trying to put it into a bucket. And then at the same time, though, you have.

Chris Gammell: It's like you're filling up a bucket with a fire hose. Then you have this super powerful vacuum that quickly sucks it all out. And then it fills up again. And it sucks it all out.

David Kronstein Tesla500: Wait. So. Oh, so you have to pull it. OK. So the fire hose is coming in. It's filling up this bucket. But then it has to go back out. And that's going out to where?

Chris Gammell: At some point, it hits a fill level. And that trips logic that starts reading the data out of the FIFO and dumping it into RAM.

David Kronstein Tesla500: Oh, I thought the bucket was RAM. OK. So the bucket is the FIFO. Bucket is the FIFO. Yeah. OK. Right. OK. Interesting.

Chris Gammell: And then. Yeah. Then it gets written to RAM. And then a separate piece of logic reads out the data specified to get a certain address and processes all the display functions. Like it does some gain and offset correction. D-mosaics. It does color matrix. All that image processing stuff.

David Kronstein Tesla500: So. But the FIFO doesn't ever stop. Oh, I guess it's a FIFO anyways. Right. So it's not really a bucket. It's like a. It's like a. I don't know what it would be. Like a tube. Right. Yeah. It's like a tube.

Chris Gammell: Like a bucket. A bucket is like a stack. It's sort of the opposite of a FIFO. Yeah.

David Kronstein Tesla500: Yeah. OK. Oh. And then it goes into the.

Chris Gammell: Then it gets read out for display. It gets read out of RAM into another FIFO. And then that gets read out much more slowly at the sort of like 60 frame per second pixel rate to be sent to the to the system on chip.

Dave Jones: Now, here's the interesting thing. You actually convert it to a video format.

Chris Gammell: Yes.

Dave Jones: To go between the FPGA and the processor. Now, I can guess at why you did that, because it's normally an extra step you wouldn't do. I'm guessing. Yeah. But because you did it because the FPGA. Sorry. The processor has support for video in.

Chris Gammell: Yes. Is that the reason? It supports standard 24 bit parallel RGB video in. Got it. So as long as it's in that format, it's really easy to deal with all the software they supply just works. Got it. And it's really easy to develop for.

Dave Jones: Yep.

David Kronstein Tesla500: That sounds nice.

Dave Jones: So that was a development time decision. You know. Yes. Why roll your own processor code to decompose video and stuff like that.

Chris Gammell: You need a couple of software engineers for a year to do that.

Dave Jones: A year to do that. Right. Got it.

David Kronstein Tesla500: Well, that's really smart then. And so. Wait. So what does it take to get to that RGB 24 format?

Chris Gammell: It's basically the data from the image sensor. There's one 12 bit. There's a 12 bit word for each pixel. And even on the color sensor, there's only one 12 bit word per pixel. So it has to go through an algorithm that interpolates based on which color filter that pixel is. It interpolates some of the pixels around it to get the full RGB value for that pixel.

David Kronstein Tesla500: Oh, because it's 24 versus 12 bit, you mean?

Chris Gammell: Yeah. Well, it's 12 bits. And then that ultimately gets converted down to 8 to be output.

David Kronstein Tesla500: Oh, 8 per color. Is that the idea? 8 per color. Yeah. Okay. I see.

Dave Jones: Now. Damn. Can we quickly. I don't want to derail our nice sequential design. But he's about to. Discussion here. He's about to. I'm about to. So just jump back to the sensor. The sensor is available in monochrome and color. And you actually offer people, don't you, in the Kickstarter, a monochrome or color version.

Chris Gammell: That's right. Yes.

Dave Jones: What is the. Like, why would you want monochrome?

Chris Gammell: There's a few reasons. And a lot of people actually ask that. Because of the color filters, if you, to understand the color filters, look, Google Bayer color filter. I'm sure we'll have a link there.

Dave Jones: I've done those in one of my videos. Yeah. It's very interesting.

Chris Gammell: Because of that color filter, you lose a little bit of resolution depending on the scene you're looking at and depending on the demosaic algorithm that converts that Bayer pattern into RGB. So a monosensor completely lacks that color filter. And it's just native monochrome pixels that are sensitive to the whole spectral range. So you get about twice the sensitivity. And depending on the scene. Nice. Almost twice the resolution. Oh, wow. That's intense. Okay.

Dave Jones: So the quoted resolution of the sensor, the 1280 by, what is it? 1400. 1280 by 1024, yeah.

Chris Gammell: That doubles? In certain situations, like if you're looking at, for example, a pure blue object, that'll be half on the color sensor.

Dave Jones: Oh, right. Of course. Yes.

Chris Gammell: For a typical scene, it's almost one-to-one. It's not that bad. But for certain edge cases, it can be pretty bad. And that's where you don't have that problem on mono. Right. Wait.

Dave Jones: So if you were shooting an object that was blue. So if you're doing, so if you've got a car that's blue and you're filming the crash test of this car, then you'd get half the resolution, roughly.

David Kronstein Tesla500: About that, yeah. Maybe.

Dave Jones: Because it's blue. Because you're shooting in.

David Kronstein Tesla500: Is it because you're throwing away all the red and green data? Is that what it is?

Chris Gammell: Yes. Because like the only pixels that would light up would be the ones with the blue filters. And that's only one quarter of them have that blue filter.

David Kronstein Tesla500: Oh, because the filters actually exist on the picture? On the die. Yes. Oh, okay. That's interesting. So it's like you're selecting in hardware and then you're basically, you're like downsampling in the physical world.

Speaker ?: Yep.

David Kronstein Tesla500: That's right. Yeah.

Dave Jones: I've got, in one of my teardown videos of like an old Sony camcorder or something, I actually look at the sensor under a microscope and you can see the RGB filter strips in there. You can actually see the strips. It's really, yeah. That's cool. It's really cool. So what, what typically, who would typically use a monochrome version like that? If you're doing. Are you targeting a certain market for that?

Chris Gammell: Usually if you want to run the camera at low resolution to get the highest speed, you usually want to get the best resolution possible. So you may use monochrome there. I had one person who requested it to, they wanted, I think they wanted to film bats and they wanted to eliminate the scene with an IR light. So that would be, in that case you don't need color. Perfect.

Dave Jones: Yep.

David Kronstein Tesla500: And it's sensitive to, to, to outside the visual range. Is that right?

Chris Gammell: Sensitive up to around 900 nanometers, I think.

David Kronstein Tesla500: Oh, wow.

Chris Gammell: Yeah. The color version actually has to have an IR cut filter to remove everything above about 650. Otherwise IR gets mapped into colors weirdly and things look wrong.

Dave Jones: Right. Got it. So if you're doing nighttime stuff, that's IR illuminated, then you choose the black and white. Definitely.

Chris Gammell: Yeah. It's just most people, most people like color just because they want pretty shots. But there's a lot of, once you actually use the monocamera, it's actually very nice in a lot of situations.

Dave Jones: Yep.

David Kronstein Tesla500: So is that why if you take filters off of certain like webcams, you can get like IR sensitivity too? Like it's just innate to the CMOS?

Chris Gammell: Yes. Yeah. The people, that's a hobby. A lot of people have to take out the sense, take out the color filter and you get really interesting. Like I think trees light up like a whitish or pinkish color rather than the normal green. That's cool.

Dave Jones: Very cool. All right. See, that was an interesting tangent. You're right. I'm glad you asked that. Thank you very much. That's good. Yes. All right. Back to our flow. Where were we up to?

David Kronstein Tesla500: I mean, David, could you summarize maybe like, so what's coming out? I can't really see in my mind how much of this is contained within the FPGA, right? So you said it's 15%.

Chris Gammell: So everything you've been talking about is in the FPGA so far.

David Kronstein Tesla500: Oh, really? Okay. Yes.

Dave Jones: Including generating that RGB color space signal.

David Kronstein Tesla500: So you're doing all this stuff in Verilog? Is that right? All in Verilog, yes. Okay. Yeah.

Dave Jones: It's all relatively... Any reason for Verilog? We always ask that.

Chris Gammell: I just like, I like the syntax much more sort of C-like. I'm familiar with C. Mm-hmm. VHDL is, I don't like the verboseness of it. But there are certain advantages in some. It's more capable in some ways, but I'm just, I like nice syntax and not duplicating a lot of things like you do in VHDL. Yeah.

Dave Jones: Sure. Cool. And all the tools for these are free, are they? For your...

Chris Gammell: The FPGA... Development? The dev tools are free for the ECP5. The one thing that's not free is the RAM controller license. Oh, okay. Yeah. Other than that... How much does that cost? That's about $500 a year, I think. Okay. So it's not... It's not too horrible.

Dave Jones: You're right. Yep.

Chris Gammell: That's for like the IP? That's the IP core, yeah. Okay.

David Kronstein Tesla500: Yeah. That's... I mean, if you really look at it, like, that's saving you a lot of time, not having a right

Chris Gammell: So that's like a year's project to do a PDR controller. Yeah.

Dave Jones: So do you get the source for that?

Chris Gammell: No. They give you this... They give you an encrypted source along with a... Some non-encrypted config files and the program that generates all the config files. Yeah. I wish they'd just give out the source.

Dave Jones: You're right. Yep. But then it'd leak and...

Speaker ?: Then it'd leak.

Dave Jones: Yeah. Everyone.

Chris Gammell: It's so tied into the hardware in that chip that you couldn't just copy it. Exactly. I know.

Dave Jones: They're selling hardware. I know. Anyway. So, okay.

David Kronstein Tesla500: So then what is your interface out of this FPGA then?

Chris Gammell: Because... There's a few. There is... To the CPU, there's two of these 24-bit RGB streams. Okay. What else? There's not much else. There's a GPMC bus, which is basically just a 16-bit parallel bus that you use to access registers in the FPGA. So the CPU can control settings in the FPGA.

David Kronstein Tesla500: Oh, that's nice. So that's like when you're using the click wheel and trying to change stuff?

Chris Gammell: Yeah. Like when you change a frame, just write to register. Oh, I want to display this frame. The FPGA goes and does that and starts spitting out video. And there's a few other minor things. There's some trigger I. The trigger I.O. is connected to the FPGA. The ADC for the analog channels is connected there too. So it can buffer the data right onto acquisition RAM.

David Kronstein Tesla500: I'm having a panic attack thinking about how much FPGA knowledge there is in this thing. So much. Yeah.

Dave Jones: This is a massive project. It's really impressive though. It really is. Is this one of the most complex Kickstarters you've seen? It's probably, yeah.

Chris Gammell: Like in terms of hardware and software development? One of the most complex projects I've seen basically a single person do that I know of.

Dave Jones: Same here. Yeah. I thought it was. Yeah.

David Kronstein Tesla500: I mean, it's reminiscent of like, I mean, Jerry did a lot of FPGA stuff too. And I mean, it's reminiscent of that. I think the scale of the number, I don't know the internals of the Cast AR stuff, but like it seems like it's the FPGA stuff in general when it's doing this fancy image processing stuff. It's nuts.

Chris Gammell: Yeah. There's just so much stuff. So many. It took a long time to develop, but it was, yeah, it was a really fun project. Good. But yeah, it's so long. So it's like, I just want to get this done. I have other projects I want to do, but got to get this out.

David Kronstein Tesla500: Oh no, dude, this is the rest of your career. I don't know if you know that. I mean, this is.

Dave Jones: So did you have this end goal in mind that you'd produce a product to sell or. I guess sort of in the back of my mind. At what point did you go? Oh, I think, yeah, I'm going to take this serious now.

Chris Gammell: I think sort of in probably around 2012 when I had the first sort of alpha prototype on on the way to running and saying, yeah, hey, I can actually maybe sell this and this new Kickstarter thing is out. This would be useful.

Dave Jones: Right. Okay. So you've been thinking about it for four years. Four more years. I'll wait. That's what you're thinking.

David Kronstein Tesla500: And now we wait. Now it's finally there.

Dave Jones: Were you at all tempted to do the Kickstarter thing and saying, oh, look, I've already got this prototype that I've been working on for a few years and I'll just go do the 3D renders and everything else. And then I'll get all the money. I'll get a million dollars and then I'll hire a team. Was that ever?

Chris Gammell: I was not happy at the state of development. I asked myself, would I be confident in selling this? No, I wouldn't.

Dave Jones: You're an engineer and not an entrepreneur. That's your problem.

Chris Gammell: A business person would bust through that barrier. I know. Raise a shit lot of money. And there is a competitor camera, the Edgertronic came out in 2013. And it was quite similar, but they did a pretty good job, but they made a few mistakes like not having an onboard LCD. You had to connect it by Ethernet to a computer and you needed to tether it to a power supply and it was a lot of effort to use. But apparently they've sold 700 of those at $5,500 each.

Dave Jones: Oh, okay. Now there's another competitor that was on Kickstarter, wasn't there?

Chris Gammell: Yes. FPS 1000, which was, I was really surprised. That's the project where I found out how fast Flash is.

Dave Jones: Right. Yes. I had no idea. Because people talked about that on the forum. They write directly to Flash. And I thought, oh shit, that's not possible. And then I thought about it for 10 minutes and I thought, oh yeah, they could probably do that.

Chris Gammell: Yeah. Then I looked it up and, oh, these Flash chips, a single chip you can write 200 megabytes a second sustained to.

Dave Jones: Yep.

Chris Gammell: The only issue is, yeah, I looked up the data sheet. It's 60,000 write cycles. So if you leave it recording in a loop for two weeks, it will completely use up the Flash's life.

Dave Jones: Don't they use our Flash where leveling?

Chris Gammell: The problem is when you use the entire Flash. There's no point.

Dave Jones: You use the entire Flash, right. Yes. Of course. Right. So that's their problem. So that's their issue.

Chris Gammell: Yeah. I mean, it's a really ingenious design. Hats off to, I think it's Graham Rowan for coming up with this. It's way cheaper than you could do it with the architecture I'm using. But it's just, there's some issues with it. And it's been delivered? Is that right?

Dave Jones: Yeah. They got 125,000 pounds and it's been delivered.

Chris Gammell: I've heard some haven't. I've heard conflicting stories. I'm not sure about that. But he did run a second Kickstarter just a couple of months ago because he needed more money, apparently.

David Kronstein Tesla500: Oh, yeah. I'm looking at the updates. Okay.

Chris Gammell: Yeah. I think that was FPS 1000 HD. Okay.

Dave Jones: Oh, okay.

David Kronstein Tesla500: Rebranding a new product launch.

Dave Jones: And it's significantly cheaper than yours. Is that because they went for the FPGA solution or is it sensor?

Chris Gammell: What's lower cost sensor? The case isn't as nice? A lot of those. There's, yeah, that's on the first one, FPS 1000. Those are the case. On this new one that he just released, they're actually using the same sensor I am. And he's selling it for, I think it's $2,200 US dollars. It's a little bit cheaper than mine.

Dave Jones: Okay. A little bit cheaper. Okay. Got it.

Chris Gammell: But yeah.

Dave Jones: I don't see that on, oh, no, because it's in pounds. That's right. It's 1,300 pounds. Okay. When you can, no, sorry, 1,900 pounds. Okay.

Chris Gammell: The one issue with that is he doesn't have the system on chip with H.264, so it only saves to raw format, which is huge files that you have to process in your computer and the workflow is difficult.

David Kronstein Tesla500: Gotcha.

Dave Jones: Got it. Okay.

David Kronstein Tesla500: I mean, it definitely is a nice product. I mean, this looks nice. Oh, definitely. Yeah. Yeah. That's great.

Dave Jones: Cool. But it doesn't have the same sort of system capability your one has with the synchronization and the IO and you've even got analog inputs. You can superimpose waveforms over the top, which I think is a really cool idea.

Chris Gammell: Yeah. No, it doesn't do any of those. It's a very, it's a low cost. It serves that low cost market.

Dave Jones: Yeah. Got it. So, but you were shooting for like the professional industrial market. Yeah.

Chris Gammell: I was trying to get into that, but still be at the price point. I felt sort of like the $2,500, $3,000 was the price point where people buy DSLR cameras that are that expensive. Yes. It's not completely ridiculous.

Dave Jones: Yep. True. Okay. So you went into this with a price point in mind.

Chris Gammell: Yes. The old version, alpha version that used the other sensor, that would have sold for probably $5,000. Wow. So I don't think it would have been anywhere near successful. Yep. I'm very glad. I made a sort of, I was feeling, is this feature creep? But about two, a year and a half ago, it was like this new sensor came out, the Luxima one. I was like, should I change this? It's like a fifth the price of the other one and twice as fast. Yeah. I'll redo the whole thing. I think that decision paid off. Yeah.

David Kronstein Tesla500: That was a good decision.

Chris Gammell: But yeah, that set the project back at least a year and a half. Wait, wait, you did redo it? Yeah. I basically threw out the whole thing and redid it. Oh, wait. Wow. But that's not the Luxima one. You mean the... No, the one I previously was using was the ON semiconductor Lupa 1300.

David Kronstein Tesla500: Okay. Okay.

Chris Gammell: Yeah. And that was, that sensor's like $1,500.

Dave Jones: Oh.

Chris Gammell: That's just... Yeah. Yeah. And it's only half as fast as the current one.

Dave Jones: Yep. So that's why you went bugger it. I can't make a viable business out of that.

Chris Gammell: No, Edutronic is too much of a competition in that segment. So I said, I've got to change this.

Speaker ?: Huh.

David Kronstein Tesla500: Yep. And how long did that redesign take?

Chris Gammell: And it took you a year and a half. I started it. I first prototyped the sensor on the old alpha version in mid-2015. I started laying out this new version in beginning of this year. Got prototypes in June. And then it's just been a sprint until now to get it running.

Dave Jones: How long did it take you to do the schematic and the PCB layout?

Chris Gammell: Schematic was generally just a copy of the old version. So that wasn't a huge change. That was like a few weeks. Then PCB layout was about, I'd just like to say about three weeks or a month.

Dave Jones: Yep. Yep. That's typical. Yeah. And how many layers?

Chris Gammell: Eight layer.

Dave Jones: Eight layer. Because you need to fan out the 600 pin BGA. Yes.

Chris Gammell: Just no way you can do it on any fewer.

Dave Jones: No, you can't.

Chris Gammell: Oh, and so you ripped up all the FPGA stuff too. Oh, I ripped. I changed the FPGA. I packed two boards down to one. It was a complete redesign. Complete relay out. Yeah. The schematic was generally the same. Yeah. Yeah. Wow. And the form factor stayed the same too, right? Pretty much. It got a few millimeters bigger. Okay. But definitely worth it to go to one board. Yeah. Oh, yeah. Definitely. Because when I had two boards, about a third of the space was taken up by the board-to-board connectors and the traces connecting to them. Right. And they're not cheap connectors usually. No, those are expensive. It's like $40, no, $20 worth of connectors there.

David Kronstein Tesla500: That's how Samtech makes their money. Yeah. And they make their money.

Dave Jones: And they're very nice connectors. Yeah. Yeah. And their sampling service is second to none. Yeah. So what's the major thing driving the cost in this? Or is it just a combination of bloody everything?

Chris Gammell: I'd say the image sensor is almost a third of the cost. The main board is about another third. And then it's all sorts of random stuff like battery and boards and assembly and LCD and all that stuff.

Dave Jones: Housing and...

Chris Gammell: Housing, yeah. Got it. Wow.

Dave Jones: Wow. Are you... So what sort of contingency do you have price-wise if things go wrong in production? Can you handle all your boards coming back? I've failed. Can you...

Chris Gammell: For the early bird batch is about $25,000. And I can handle... If those boards completely don't work, I can easily handle building. It's about $10,000, $12,000 to build a whole other set. So I can do another design spin if needed. And it's not the end of the world.

David Kronstein Tesla500: Try not to do that though. That's what I've learned.

Chris Gammell: I'm not going to build a hundred or something right away without testing it in some way first.

David Kronstein Tesla500: Yeah.

Dave Jones: So did you deliberately price the Kickstarter to factor failure in?

Chris Gammell: I sort of will price the early bird versus non-early bird rewards to have this buffer of small... A smaller quantity as a buffer. And then, yeah, obviously the later rewards are a little bit more expensive just because it gives a bonus to people starting out getting in early. Yeah.

Dave Jones: You're right. Yep. Huh. Interesting. Okay. Because this is almost pre-production. Well, it is basically production. You said you're going to re-spin the board a little bit. So is that the only risk at the moment? Although you have a lot of stuff to write in the firmware. Yes. You're changing the OS, aren't you?

Chris Gammell: Yes. Sorry, not the OS.

Dave Jones: The OS is Linux. You're changing the development tool chain.

Chris Gammell: Tool chain, yes. Because of some limitations in TI's, almost complete dropping of software support for their chips, I have to switch from the TI SDK to this Ridge Run one where they actually have implemented a lot of these features that I need. Right.

Dave Jones: So this is for the TI processor. This is for the DaVinci processor. What does that TI development kit do for you?

Chris Gammell: Basically, the Linux build, the kernel build, builds for Uboot, and the file system that goes on the card that has all the main file system for the Linux system.

Dave Jones: Does it have video drivers and all that sort of jazz?

Chris Gammell: Video drivers, all the drivers for that special input hardware for the H.264 compressor. But it's just basically featured. The Ridge Run one has many more features, some of which I really need to implement some of these record modes, interesting record modes.

Dave Jones: You mentioned that the TI SDK does not support two video streams at once. Yeah. And you need that for your run and gun mode where you're recording video into the memory plus you're saving it and processing it at the same time or something.

Chris Gammell: Yes. So you need one stream to display onto the LCD and one stream to send into the compressor that goes onto the card so you can continue shooting while saving.

Dave Jones: And that just can't handle it.

Chris Gammell: Yeah. I asked on their support forum, sorry, we only support one stream at the moment.

Dave Jones: Right.

David Kronstein Tesla500: Interesting.

Dave Jones: So this Red Lake, is this like a third-party thing?

Chris Gammell: It's a Ridge Run. Yeah. It's a third-party company that does. They seem to cover almost all of TI's chips and they're just, they seem to have much better documented stuff. They'll provide support, but of course you have to pay for it. Yeah. Got it. I think it's like five grand for the license.

David Kronstein Tesla500: Again, but you're basically buying an employee at that point, right? You're basically buying. Sort of, yeah. Yeah. So what is this Linux build then? I mean, is it some flavor of Linux or what is?

Chris Gammell: I think it's called Orego.

David Kronstein Tesla500: Okay.

Chris Gammell: It's some like shrunk down something? Yeah. It's pretty much shrunk down. TI's customized it with all their drivers.

David Kronstein Tesla500: Okay. Wow. That's intense though.

Chris Gammell: Yeah. It's been interesting.

Dave Jones: So is that the risk? Have you actually tried this Ridge Run thing?

Chris Gammell: Not yet. I've been waiting for the Kickstarter, which I don't want to drop five grand on it yet.

Dave Jones: Right. Is there a backup plan if it all goes tits up?

Chris Gammell: Then I'll just have to, I will just have to develop it myself and it will just take longer.

Dave Jones: Right. Yeah.

Chris Gammell: All the stuff to do it is there. Like Ridge Run has done it. It's just the case of timing. Yeah.

Dave Jones: Could you deliver, if that comes to pass, if, you know, look, sorry guys, I need another three months to work on this. Could you actually deliver the hardware to backers and give them the current firmware and go, look, I'm still working on the firmware or how would that work?

Chris Gammell: Yeah. Right. Because the firmware started an SD card just like a Raspberry Pi, so updating it is no problem.

David Kronstein Tesla500: Oh, that's great. So just boots direct, huh? Yep. Oh, that's really nice. Got it. There's no brick in your SD card. No, you can't brick this one.

Dave Jones: So there's little risk in actually getting hardware to backers with some sort of working hardware to backers. Yeah.

Chris Gammell: Like the couple of, I do have to test a couple of things. You test some audio, the audio stuff, you test the analog stuff, but pretty like all the main video stuff is working. And yeah, even if it didn't have some of those, I think most backers would still be very happy with it.

Dave Jones: They'd still be. I'm happy with mine and mine's got no audio and lots of other, no run and gun mode.

Chris Gammell: Almost no other cameras have these features really. So it's a bonus.

Dave Jones: Yep. And the waveform on screen, because you've got some ADC inputs on this, which I think is very novel and you, in your demo, in your Kickstarter video, you demo, oh, you would have like waveforms. You could feed in signals, like test signals coming from sensors. What application do you envisage that being useful for? Is there a particular case or did you just think, hey, I don't know. Oh, a few things. I have no idea, but it might be good.

Chris Gammell: You could visualize audio and signal and see how a speaker moves. You could do like, I'd shown in a demo, like an electric motor, seeing the 60 Hertz waveform and how the motor accelerates and how sparks occur. You could visualize pots.

Dave Jones: Was that a real demo you did or was that like Photoshop-y kind of?

Chris Gammell: Yeah, I don't have it actually working hard, but it'll be something like that. Or it will save to like a binary data file or CSV or whatever.

David Kronstein Tesla500: Oh, so it's just, it's like, this is a logging function? Is that the idea? It's basically a data logger. Yeah. Okay.

Dave Jones: And it's superimposed on the screen in real time as you're shooting, right?

David Kronstein Tesla500: Yeah, or as you play back.

Dave Jones: Or as you play back, sorry, because it'll be too high frequency to see it.

David Kronstein Tesla500: Wow.

Dave Jones: Cool.

David Kronstein Tesla500: So you basically are starting like a science company, you know, like it's like this

Chris Gammell: is going to be like scientific equipment. People were saying like they wanted just to, the camera just not even for high speed, just to develop FPGA image processing.

David Kronstein Tesla500: Yeah. Wow. Like a dev kit. Okay.

Dave Jones: As a platform. Maybe you could just sell the board as a dev board. Possibly, yeah. Wow.

Chris Gammell: Get your quantities up. Depends what people want. Some people, people were asking me, can I buy a sensor, can I buy a monochrome just sensor board and a color camera so I could swap the two? Yeah. I was like, if you're confident, yeah, I'll sell you one.

David Kronstein Tesla500: You know what I would say is that, I don't know if they're listening, but maybe some of their employees are, but I would, if I was Lattice, I'd be like, hey David, I would love to, you know, pay you to rebrand this as one of our dev boards. Or if I was TI, I would listen to this and be like, I'd love to make this our dev board.

Chris Gammell: And then, yeah, exactly that. I'm actually wondering if Luxima may want to sell, but I think this camera is, if I recall, it might be cheaper than their dev system for their sensor. Yeah. Actually, no, I think it's about two grand. Oh, wow. No, yeah, their dev system was about a couple of grand or something. I don't remember exactly. Right. Wow.

David Kronstein Tesla500: Or they could just acquire you outright. They can just, you know, if you, do you want to buy a Canadian, we'll sell you a Canadian. The Amp Hour is selling a Canadian. Luxima, TI, Lattice, give us a call. We'll give you a great price, Canadian dollars.

Chris Gammell: Of course. One thing I would not do, I would not sell out just to have it suppressed. Right. Of course. Right.

Dave Jones: Yes. Yes. Because I want this out there. So where do you see the company? Yeah.

David Kronstein Tesla500: You need 10 years of your life down the drain, you don't want to just make it.

Dave Jones: Where do you see the company going? Because this is your full-time job now, basically. Pretty much, yeah. Do you have plans for taking on people?

Chris Gammell: Yeah, definitely looking for a software engineer, someone who knows the embedded Linux side of things. I'm not as familiar. I could definitely use help there.

Dave Jones: Well, we happen to have an audience listening right now, and we have a great track record.

Chris Gammell: We will sell you some nerds, David. Yeah. We've got nerds. Let's just hope the Kickstarter does really well and I can afford to pay people.

David Kronstein Tesla500: Yeah, right.

Chris Gammell: You might not.

Dave Jones: So are you interviewing people now if somebody's actually some, if someone's- Do we have an email address we could send resumes to? That's what Dave's really getting at.

Chris Gammell: Yes. I will set one up, yep.

David Kronstein Tesla500: Okay.

Chris Gammell: Cool. But yeah, one person who backed the cameras has done a lot of Linux stuff, and he's interested in doing some, helping with some software. Like, he's currently working on fixing some of the, doing a boot splash screen and fixing some of the boot up time issues. Nice.

David Kronstein Tesla500: Yeah. Community stuff. I mean, that's going to be great too, right? Yeah. I mean, it seems like that would be a good crossover. Yes. Yep. And I'm sure there's a lot of software people out there that have a lot of excess cash sitting around that want to film water balloons popping and stuff too. Funzies.

Dave Jones: Have you looked at the potential market worth of this sort of, of this camera at this price point? What sort of volumes do you think you could sell if you were successful?

Chris Gammell: Just judging by how Edutronic sold their camera, 700 of them at twice the price, I can imagine that it's much more than 700s. So it could be into the thousands.

Dave Jones: Over how many years did they do that?

Chris Gammell: That was three, three years.

Dave Jones: Okay. Right. So, and so are they the go-to one at the moment for that sort of price point?

Chris Gammell: Sort of the go-to for like real professional low cost, high speed.

Dave Jones: That, that still seems low to me. I mean, you know, how many universities and research groups and other, you know, companies that want to do this sort of thing? Like, I can imagine just, you know, one car company want to do crash testing and stuff like that. They'd order 50 cameras or something, you know? Something like that, yeah. It's like pittance for them.

Chris Gammell: I think it's just, they're being held back by the lack of slightly poor usability requiring the computer, et cetera.

Dave Jones: Okay. Okay. So it's a system problem. It's probably, yeah.

Chris Gammell: I mean, it's a great product. Right. It did a really good job. It's just not, it didn't quite hit the market perfectly, but it still did very well.

David Kronstein Tesla500: Got it. Can I ask a pessimistic question? Mm-hmm. Are you worried about people ripping off your design? I mean, we, we usually talk about openness and stuff.

Chris Gammell: It's pretty much secured by the FPGA bitstream security.

David Kronstein Tesla500: Oh, it is. Okay. So that, yeah, that is a way to do it.

Chris Gammell: As long as I keep that secure, the design is like the most of the effort, the majority of the effort was making the FPGA firmware.

Dave Jones: Okay.

Chris Gammell: Well, that's.

Dave Jones: I suspect at this sort of technical level, like no one's going to rip it off. Yeah. It's not, yeah. It's not like a low cost product. If you know how to do this, you're going to make your own. It's not a low cost product. That's a good point. Yeah. Yeah. Yeah. Okay.

Chris Gammell: It's not like, like you hear these stories of Kickstarters happening and then the Chinese are selling it before the Kickstarter is even done. And, but those are like very simple.

Dave Jones: That's because it's a $50 product.

Chris Gammell: It's like a mechanical device. Yeah. Yeah.

Dave Jones: Yeah. Yeah. That's it. It's copied. Bam. No, this is.

Chris Gammell: This is very different. Yeah.

Dave Jones: Yeah. Entirely different. So no, I wouldn't be concerned with that.

David Kronstein Tesla500: Not at all. We had some questions on Reddit, which I will kick myself if we don't get to. Sure. Oh, right. Yeah. I forgot about that. I think a lot of them we answered. Mike Harrison asked about the sensor costs. Wegmat asked about the journey and logistics. I think we got all that stuff. Mm-hmm. The one at the top though, Genlock. Is there any support to Genlock 2 cameras? So I'm guessing there's a lot of industry type of interplay type stuff. Is that, is that what this is? I don't quite understand what this means. Yeah.

Chris Gammell: Genlock is being able to synchronize the camera to take frames synchronized to another source. So yeah, it will support that. And then basically you can either have it like a rising edge trigger and image to be captured. You can have the shutter be directly controlled by the level on the trigger. Like it just exposed while high. Yeah. Those are the basic two modes of doing Genlock.

David Kronstein Tesla500: Okay. And then what about the, so then this is also, this is Maurice who was asking this about this, but then also the, do you have recommended lighting rigs? I mean, so we saw Mike doing that 600 watt incandescent. I mean, do you have other high power type stuff?

Chris Gammell: Well, I'm not really selling it, but I built a five kilowatt LED lighting rig and I have a video on YouTube. Yeah, that's cool.

Dave Jones: If you haven't seen it, guys.

Chris Gammell: So why aren't you selling it? It's a whole nother product and it's not, it requires water cooling and expensive power supplies.

Dave Jones: Yeah, no, that's messy. That's messy. Don't go there.

Chris Gammell: You can buy, I would suggest people can just buy the, there's cheap LED film lights that are super plentiful. That's actually what I got in the mail before, right before the show.

David Kronstein Tesla500: Yeah. I don't know what the output is, but yeah.

Dave Jones: Yeah, yeah, no. That, that is a market you don't want to get. Yeah. You're just going to get beat in that market.

Chris Gammell: I'll probably just, I'll open source my like water block design if people want it. Maybe if people, enough people want it, I'll sell those, but I don't want to sell the complete thing. Like getting UL approval on the power supplies, like 20 grand, it's not worth it. Got it.

Speaker ?: Yeah.

Dave Jones: Now, Nick282000 wants to know what the real data rate from the sensor is. The data sheet says 2, 1.5 gigabits. Yeah. Oh, sorry. Bit versus bit was what we were.

Dave Jones: A second. Yeah. Yes.

Chris Gammell: The real data rate is, just a second, let me calculate that.

Dave Jones: Are you using Windows calculator or a real calculator?

Chris Gammell: I'm using Windows XP calculator, not the Windows 7 one. In scientific mode or in just normal mode? In scientific, because in the Windows XP calculator, they made the key combination that does square, square now does square root. Seriously? Yep.

David Kronstein Tesla500: Wait, what was it before? It was a slash?

Dave Jones: Get a real calculator. Get a, like.

Chris Gammell: And by escape, just a minute.

Dave Jones: Your entire reputation is now down the toilet, David, because you're using a Windows calculator.

David Kronstein Tesla500: We're going to just, we're going to, we're going to get remixed, just, just David using Windows calculator over some techno music. Calculator. That'll be after, that'll be after the, after the outro.

Dave Jones: You'll have to insert, you'll have to insert MC Frontalot's song about calculus.

David Kronstein Tesla500: Oh, okay. Okay.

Chris Gammell: Why was that vacation germane to the math? Cause of good data policy in the office and the fast.

David Kronstein Tesla500: MC Frontalot's. I didn't know he had, that's your favorite.

Dave Jones: I was going to actually feature in that video. He actually approached me and said, hey, I'd like to use this voice clip from one of your videos. And I actually remix it into the song.

David Kronstein Tesla500: Oh.

Dave Jones: And it, it didn't make the final cut. Your chance of fame. Sorry, Dave. Yeah, it didn't make the final cut. Or I might be hidden in there somewhere, but I just didn't notice it. Okay.

Chris Gammell: The final data rate is average 3.1 gigabit, gigabytes per second. Gigabytes. Yes. Okay.

Dave Jones: So multiply by eight for bits.

Chris Gammell: Yes. So 24.8 gigabits per second.

Dave Jones: Right.

David Kronstein Tesla500: And that's coming out of the.

Dave Jones: That's coming out of eight lines.

Chris Gammell: 16.

Dave Jones: Sorry, 16 serial lines.

David Kronstein Tesla500: Yes.

Dave Jones: Right.

David Kronstein Tesla500: Yeah. Jeez, that's a lot of data.

Chris Gammell: Yeah, that's pushing.

Dave Jones: So 28 meg divided by 16. So we're still talking about 1.2. 1. 1. 1. 1. 1. 1.

Dave Jones: 1.

Chris Gammell: 1. 1. Yeah. It's really.

Dave Jones: Just over a gig per, per SIRDES. Yeah. So each SIRDES has to handle one gig. Yeah. Right. The main, the limitation is. Would the hardware.

Chris Gammell: Oh, go ahead.

Dave Jones: One handle. Would, would the hardware SIRDES handle that? Or did.

Chris Gammell: It's, no, it's limited. Yes. Oh, easily. Been like three times that. It's just the number, quantity. Yeah. Yeah.

David Kronstein Tesla500: They have like standard numbers for those for some reason. It's like one 5.6 gigabit. Right. Or.

Chris Gammell: It's weird numbers. Yeah. Yeah.

David Kronstein Tesla500: 23, 24 or something like that. Yeah.

Chris Gammell: But the, yeah, the limitation is pretty much RAM. Like I'm pushing the limit. But the theoretical bandwidth on the RAM is about 4.8 gigabytes per second. So we're using more than 50, more than like 60% of that. Oh, wow.

Dave Jones: What, what if it was just as, what if you got through halfway through this design and you went, oh, it's just a smidgen over the DDR3 data rate. I can't do the maximum resolution. What would you have done? Would you have gone to DDR4 or?

Chris Gammell: I'd probably reduce the speed or increase the RAM clock slightly. Either reduce the image of the speed.

Dave Jones: Or you'd overclock. You'd go, oh yeah, I can overclock this baby. Right.

David Kronstein Tesla500: Wow. Wait, so how close to the margin are you right now? I mean, like, or how much margin do you have rather? There is.

Chris Gammell: So that is 3.1. Oots, oots, oots, oots, oots.

Dave Jones: I love where we're doing calculations live on the show.

Chris Gammell: 65% of the theoretical bandwidth. Okay.

Dave Jones: Oh, that's, yeah.

Chris Gammell: I've never seen an image artifact due to like RAM over or under flow. So it's solid.

David Kronstein Tesla500: Okay. That's great.

Dave Jones: Nice.

David Kronstein Tesla500: And then, okay, so then at the, so this is all self-contained at this point, right? So then it gets, you can also push to like a recording medium, like a SD card or some kind of physical medium.

Chris Gammell: Once you have your recording saved in RAM, you can save it to SD card, to USB, to eSATA drives. You could save it over a network to network share. However you want to get rid of it. Yeah.

David Kronstein Tesla500: And that's just standard, the standard output from Linux drivers driving to these, these standard ports. Is that the idea?

Chris Gammell: Yes. Yeah. It's all, that's one nice thing about Linux is all this stuff that would be a massive difficulty to do on your own is all just taken care of.

David Kronstein Tesla500: Right. If you were doing just like an embedded platform from scratch or something like that would be. That's why, that's, that's why the old scopes all struggled even putting stuff on USB drives, right? Yes.

Chris Gammell: They finally got that working like mid 2000s. Like that took a while. Yeah. That is you. Yeah. That's true.

Dave Jones: What was the hardest part of this whole thing? What was the one that really kicked your ass?

Chris Gammell: Probably the most worried I was, was getting the RAM working on the ECP5. I was a few weeks away from my deadline of, if this doesn't work, I'm going back to ECP3. Oh, wow. But I was able to get it working.

Dave Jones: Sweet.

David Kronstein Tesla500: And you said you have to pay.

Dave Jones: Did it just magically work one day or did it just partially work and you were able to figure it out?

Chris Gammell: I was able to partially work. I was able to hack some of the generation scripts and he was able to make it work.

David Kronstein Tesla500: Sweet. And you're paying 500 bucks a year for that DDR controller that doesn't really even work? I should ask for a refund. You totally should. Yeah, totally.

Dave Jones: Yeah. Or sell it.

David Kronstein Tesla500: Just work it into the deal when they ask you to become a dev board. Totally.

Dave Jones: No, that's, they should get on that.

David Kronstein Tesla500: How did you, how did you set the price on this whole thing? I know that kind of switches topics as well.

Dave Jones: I think we answered that. Did we? Yeah, I sure we did.

Chris Gammell: I had a price target of this like $2,500, $3,000, sort of just had to make the hardware fit within that.

David Kronstein Tesla500: Okay.

Chris Gammell: Yeah, you use a lot of commodity parts, put a lot of effort into pulling cost out of the bomb, removing expensive parts, using lower cost ones.

David Kronstein Tesla500: Yeah. Okay. Sorry, Dave, I cut you off.

Dave Jones: How much time did you spend on the software versus the hardware? Because this is just a bigger software project. I mean, I can't imagine writing software for this.

Chris Gammell: Probably four or five times as much time on software as hardware. Wow.

Dave Jones: There you go.

Chris Gammell: Yep. Easily, yeah.

Dave Jones: For you newbies out there who think something like this is, you know. Yeah. I mean, this is one of the highest end hardware projects I've seen on a Kickstarter. And you're saying it took four or five times longer to develop the software. I mean, that's, you know.

Dave Jones: Like, just to give you an example.

Chris Gammell: Wow. This is probably, I was learning, it was a very steep learning curve for learning the embedded Linux stuff. But it took me about a month, month and a half to figure out how to switch the UART from UART 0 to UART 4. Because there's like 10 different places in the kernel source tree that I had to go find and change these register values. Oh, it was horrible. Don't become a current Linux kernel developer, no.

Dave Jones: Oh, this is.

David Kronstein Tesla500: Mama, don't let your boy grow up to be a kernel developer, huh? Wow, that's crazy.

Dave Jones: That's insane.

Chris Gammell: I'm sure it would be faster for me to do it now. It was just that learning curve, learning where this is. I did so much search everything in the whole kernel directory for whatever term. And finally, I sort of know where things are now. It's a little bit easier.

Dave Jones: All right. Wow. 10 years of development. I can't believe it's been that long.

David Kronstein Tesla500: I think that you don't really think this is going to end after the delivery of stuff, do you?

Chris Gammell: No, I want to do more cameras. There's lots of more nice sensors out there.

David Kronstein Tesla500: Okay, so you're not like, I'm going to go build Wi-Fi widgets. You're like, I just want to build different stuff.

Chris Gammell: All the things I want to build are weird, esoteric things that have not been done before.

Dave Jones: Okay.

Chris Gammell: Or do some problem in a much better way. Any hints?

Dave Jones: But does that...

Chris Gammell: Any hints? Any hints? I wanted to work on a raster scanning laser projector, and I have an idea how to do that and get full 1080p resolution.

Dave Jones: But is this taking focus away from the company you're trying to build here?

Chris Gammell: Yes, that's why I'm not doing them right now. Yeah.

Dave Jones: Right. And I... Yeah, like, it's not like, oh, I've delivered all my things. Now I'll go work on my fun little project.

Chris Gammell: No, I've got to... If I want to keep the company going, I've got to keep updating, making new products. It's just a never-ending cycle. Otherwise, someone else is going to come in.

Dave Jones: Does it? Do you need a new product, though? Or do you... I think the systems integration here is more important. Providing solutions for people who want to have multiple cameras all synchronized and tied together and big rigs and...

Chris Gammell: Yes, definitely. I think... Stuff like that. Once, yeah.

Dave Jones: I can see that being your... Where you can add value. Like, not just selling a one-off camera and that's it.

Chris Gammell: Yeah, I think that's definitely something where other people haven't... Other companies haven't really focused too much on.

Dave Jones: So are you going to do that? Because I can see that being your differentiator. Yes, I think... We provide system solutions, you know, like...

Chris Gammell: And once you've got those systems, the whole CPU side is sort of disconnected from the image sensor side. So you can just swap in different sensors, different FPGAs, make a faster camera. Right. All that other stuff is all done. Got it. That's sort of where I want to take this.

David Kronstein Tesla500: You list at the bottom of your page, this thing is going through FCC Part 15, but it's not a big deal.

Chris Gammell: I don't think so. That's probably... I shouldn't say that because I'll have bad luck, but...

Dave Jones: Because you're doing the worst possible thing for a product here. Massively high-speed parallel digital stuff.

Chris Gammell: Yeah, well, I'm going to... As soon as possible, I'm going to book a time in the local test lab. Hopefully, I can try to get some video of that. They're often pretty picky about NDAs with other people's products sitting around, but I'll see what I can do.

Dave Jones: Okay.

Chris Gammell: Getting some video of actually debugging this.

Dave Jones: Is this a showstopper for delivering product to people? Can you still deliver even if you fail? Yeah, that's a good question. Because they're like prototyping.

Chris Gammell: You can deliver a certain number. I forget what it was.

Dave Jones: Yes, because these would not be classified as consumer product. They'd be classified as a perk. I mean, it's not even a product, right? People aren't buying the product. They donated you the money. Yes. Right?

Chris Gammell: So you're giving them a free gift. That's probably okay, but not... I'll be selling them directly.

Dave Jones: Oh, of course. Yeah, that's different.

Chris Gammell: Yeah. But that's why I sort of built in room for at least one PCB respin. Yeah, that's great. Right. Yep.

Dave Jones: As in financial buffer room.

Chris Gammell: Financial room and just time, yeah.

Dave Jones: Time. Yep. When are you scheduled to deliver these?

Chris Gammell: Early birds in March next year and the non-early bird normal ones in May.

David Kronstein Tesla500: March of 2017? Yeah.

Chris Gammell: Good Lord, man.

Dave Jones: Dude, it's the first of December today. Yeah.

David Kronstein Tesla500: I know. That's what I'm saying. That's why I said it's really close. Yeah.

Dave Jones: I think it's reasonable. But considering the one I've got in my hand right now is practically the finished product.

David Kronstein Tesla500: I think a lot of Kickstarters, you see one-year timelines these days.

Chris Gammell: Oh, yeah.

Dave Jones: Yeah, yeah, yeah.

Chris Gammell: So... Yeah. Like, I'm working... Just before this, I was working on those revisions on the final PCB, and I should have that sent out to manufacturing within the next couple of weeks. So, get ahead of schedule.

Dave Jones: Yeah.

Chris Gammell: That's crazy.

Dave Jones: How much money do you think you've spent on this, of your own money? Probably... Because it's all your own money up till now.

Chris Gammell: Probably $35,000 or so.

Dave Jones: Okay. So, you are keeping track?

Chris Gammell: No. It's just desk debt.

Dave Jones: It's an estimate.

Chris Gammell: Yeah. Yeah. It's over four years. It's not that bad. It's not, like, paying a person.

Chris Gammell: Yeah, yeah, yeah. Yeah. But it's still not a... It's still a decent amount of money. Yeah.

Dave Jones: Yep. That's more than most people put into a Kickstarter, let me tell you. Yeah. Oof. I... Like... And I'm... Like, this is how Kickstarters should be. You pony up your own... I know that's not the original intention. The original intention was to use the Kickstarter to get money to then... Because you didn't have any money to produce the product. Like, but it's just being abused. Yeah, but the...

David Kronstein Tesla500: No, the difference would be, like... If you were supporting an artist before, it'd be, like... The difference would be paying someone to go start painting or paying someone to go figure out how to paint. Right? Right.

Chris Gammell: Yeah. That's a big difference. Yeah.

Dave Jones: Right.

Chris Gammell: But, yeah, I like going into things more confident rather than less confident. Yeah.

Dave Jones: Because you're an engineer and not an entrepreneur, in quote marks. Because if you're an entrepreneur, you wouldn't care about this. Or, like, I'll just farm it out to somebody, you know? And, uh, someone in China will magically... The Oompa Loompa factory will make it for you.

David Kronstein Tesla500: Yeah. Yeah. Well, that does beg the question, though. So, the... So, the... Are there any downsides to you not being the... I mean, I don't know how much business experience you have. I mean, are you worried about any of that side of things? Of the logistics or the... You know, like, the moving forward type stuff, right? I assume this is going to work out fine. But it's only 100 cameras. Yeah.

Dave Jones: Like, it's only...

Dave Jones: That's this I'm saying.

Chris Gammell: If the Kickstarter, like, super explodes into 1,000, then I'm starting to have to worry about getting a commercial space. Yep. But if it's just like...

Speaker ?: No, no, no.

Chris Gammell: I built 12 of them. Building 100 is not that much more difficult.

Dave Jones: I mean... At what number do you think that you would have to change the way you do things?

Chris Gammell: Probably in the few hundred, I would say. That's when it starts becoming, like, difficult to find space to store things. Like, we're talking about pallets. Like, pallets of parts.

Dave Jones: So, at the moment, you still plan to do, like, all the assembly yourself and stuff? I mean, obviously, you only get the PCB made in China and assembled in China and all that sort of stuff. But you plan on doing most of it, like... Yeah, but the final assembly...

Chris Gammell: Final assembly testing yourself? Final assembly is basically just 15 minutes, so it's not too bad. And then some... Right. Testing, some automated test sequences afterwards.

David Kronstein Tesla500: Okay. Yeah, true. You front-loaded all that and you're redesigning to a single board, I'm sure, too, right? Yeah. Yeah.

Speaker ?: Yeah.

Dave Jones: What sort of test automation do you have for your PCB? Is it, like, just final output? Like, feed it, hook it up, put it in the box and hook a lens on? And if you get video out, it's a win?

Chris Gammell: For the prototypes, it was pretty much that. Plus, I checked a bunch of voltage rails, basically just manually doing what an automated test jig would do. But yeah, for the production one, I'm definitely going to come up with an automated test jig that will test all of the... Test all the voltages, make sure the CPU... Sorry, CPU is running.

Dave Jones: Right. So you've built test points into your PCB and everything for that for a... Like a... You know, you put it on a test bed and...

Chris Gammell: Yeah, like I made sure to put voltage monitoring, voltage points. I actually labeled them so you'll know what the voltages should be.

Dave Jones: Hmm. Right.

Chris Gammell: Excellent. And then it'll run through an automated calibration where it takes a bunch of sample images, checks the imager is all working.

Dave Jones: Oh, okay. Is that built into the firmware or you put special? That's built into the firmware, yeah. Cool. Okay, right. Nice.

David Kronstein Tesla500: Do you expect people are going to want to, like, access this thing or, like, upgrade it themselves or hack it or... I mean, is it too low?

Chris Gammell: People ask, can I upgrade the RAM? I think people might try to do that.

David Kronstein Tesla500: Okay.

Chris Gammell: Yeah, they could definitely do that. I'm not doing anything deliberately to make it hard.

David Kronstein Tesla500: Okay.

Dave Jones: Right. Is there a limitation why you're not just putting whacking more RAM in it? I mean, because it only records for 4 point something seconds at the moment.

Chris Gammell: Right now, the... There's two... There's two different sizes of sticks available. It was 8 gig and 16 gig in sodium size. And right now, the RAM controller only supports one stick. I think I have to get it to support two. And I know that in signal integrity-wise, it's fine. The controller doesn't support it. I think I can make it work with two sticks.

Dave Jones: Got it. So you could potentially double...

Chris Gammell: You could potentially go to 32 gigs, yeah.

Dave Jones: And that'll give you what? How many seconds? 16 seconds record time. 16 seconds record time. Yep.

David Kronstein Tesla500: At any of the speeds?

Dave Jones: At any res, yep.

David Kronstein Tesla500: Yeah.

Dave Jones: Oh, sorry. No, it changes with the...

Chris Gammell: Doesn't it? The record time is about the same at any resolution. Just because the pixel rate remains the same.

Dave Jones: Oh, right. Yes, of course.

Chris Gammell: And that makes me think my calculation earlier was wrong. Yeah, it's about 2.1 gigabytes per second. I multiplied by 12 over 8.

Dave Jones: Okay. Oops.

David Kronstein Tesla500: I wonder if people have been screaming at their podcast player for the past 10 minutes. Ha, ha, ha.

Dave Jones: So having met your goal in 24 hours or like 10 hours or something... Five hours. I couldn't get to sleep that night.

David Kronstein Tesla500: I was...

Dave Jones: Yeah, I was watching your tweets or something. Oh, your updates. Yeah. That was hilarious. It's actually fun when you go to sleep and you wake up with some big number. It's actually quite nice.

Dave Jones: Yeah. I've done that. And it's, yeah. Like, holy shit. What happened overnight? You know? But I guess you couldn't help yourself.

Chris Gammell: Nope.

Dave Jones: You had to watch.

Chris Gammell: Yeah. My friend was up. We were talking. We just couldn't stop. Another pledge. Another pledge. Just watch it. It's more boring than watching paint dry, but people still do it.

Dave Jones: Yeah, yeah, yeah. Refresh, refresh, refresh.

David Kronstein Tesla500: Paint doesn't have as much implications, though. No. Yeah.

Dave Jones: So now that you've met the goal, have you immediately started... No, because you've got a board spin, right? So when have you started? Yeah, doing the work and the board spin.

Chris Gammell: Probably going to get that out in a couple of weeks. I started ordering parts.

Dave Jones: Are you ordering any parts? Have you started ordering parts now in anticipation of getting the money?

Chris Gammell: I'm working in ordering some lenses. Yeah, I found a good company that has a bunch of these lenses I need, so I'm getting those ordered in.

Dave Jones: Nice. So you're ponying up your own money again. How very un-entrepreneurial. Because you're guaranteed to get this money coming through. Yes. So you know that, right? Yeah. So you can safely go spend your own money knowing it.

Chris Gammell: It's going to come in, yeah.

Dave Jones: Minus 8%, of course. Yeah. About 8? I think it's about 8.

Chris Gammell: I think it's, yeah, it's 5% Kickstarter plus 3 to 5% payment processor.

Dave Jones: Yep. So how high do you think it can go? Do you think you've exhausted...

Chris Gammell: I still have a bunch of promo going. A bunch of YouTube channels haven't released their videos yet, so...

Dave Jones: Oh, okay.

Chris Gammell: I'm hoping to keep the current momentum up, which would put it... It's about 30 grand a day, so it would put it just about 1 million. That's probably a bit optimistic.

Dave Jones: Oh, okay. Oh, yes. I haven't put it on the kick tracker or whatever. There's a tracker website that shows current growth and it projects out. Have you found that you're getting consistent ones per day? So far, surprisingly.

Chris Gammell: Yeah, I think it's varied from 27,000 to 32,000 a day over the last five days. That's pretty close.

Dave Jones: Yeah. Wow. Okay. Oh, but it's only been five days. It's only been five days.

Chris Gammell: We'll see what happens.

Dave Jones: People are still watching videos. So who has released videos? Myself, Ben?

Chris Gammell: Yeah, you, Ben, Cody's Lab, Telflater Mouse. Those have all released. Hydraulic Press Channel has released.

Dave Jones: Oh, have they? Yeah. I didn't see that.

Chris Gammell: Yep. They released a couple days ago, I think.

Dave Jones: Okay, cool. So was this your plan from day one? I'm a blogger myself. I know how important this shit is. I know how important their audience is. I am going to target.

Chris Gammell: Yeah, like the other ones just targeted the big magazines like TechCrunch and Tested. You know, I'm going to go a different way. And it seems to have paid off really well.

David Kronstein Tesla500: Yes. Well, because eventually the TechCrunches of the world will pick it up anyways, right? Because it'll be like, I keep seeing this thing everywhere. And it's awesome footage, right? Like I said, I've only seen Ben's. But that bullet time stuff was crazy. That was so nice.

Chris Gammell: Yeah, Telflater Mouse released his video when he first got the camera almost a month ago. And then it immediately got picked up on all those big news websites right away.

Dave Jones: Yeah, I saw that. Yeah. But that was a shame because the Kickstarter wasn't ready yet. So you probably missed a lot of opportunity there. Yeah.

Chris Gammell: But it seems most of those have reposted it now that it's running. So it seems to be okay.

Dave Jones: Oh, okay. Yeah. Nice. Yeah. Did anyone refuse your offer?

Chris Gammell: I did. No. Actually, no. I got a few people took a long time getting back to me. Then I offered it to other people. And then those early people got back and say, oh, yeah, I do want one. Oh, right. And then I ended up, I only have two cameras plus the bench dev system left here. And there's one sort of rolling around in Europe right now to a few people.

David Kronstein Tesla500: Nice. Right. That's great. Oh, okay. So what about if, so if our listening audience, I mean, like this is, you know, it's, it's definitely a very affordable, like probably the most affordable option out there if you want to do this kind of stuff. But if someone was thinking, I think this might be me. I, I, what, what do you imagine that an amp hour electronics person would use it for specifically? Anything electronics wise or is it more nature?

Chris Gammell: I think it's more nature stuff. Like you can, you'll get cool electronics, like components blowing up. You could be able to do some things like more, the electronics applications might be something like a high current power distribution test where you want to see if your wires are going to fling apart through the magnetic field or something like that.

David Kronstein Tesla500: Okay. Okay. Or what's his name? Who's the guy? Photonic induction. Yeah. Photonic induction. Right. Yeah. Is he going to do one like something blowing up real fast? Man, that pumpkin video he did was nuts. Yeah.

Dave Jones: So did you approach photonic induction?

Chris Gammell: No, I didn't actually.

David Kronstein Tesla500: Huh.

Chris Gammell: Dave, you should send yours, man. Interesting. Yeah. I sort of had to make a choice of who I thought was going to be the best and I, those are the ones that I had. You chose Dave? Good Lord. Yeah. I have to, you have to be on Dave's channel.

Dave Jones: I've got a lousy 400,000 subscribers. I'm no hydraulic.

David Kronstein Tesla500: I just, I just mean that thing's going to sit on a shelf now until, until you come on next time and Dave's like, oh shit, I was supposed to be using this thing, you know? So I'm just saying, you know.

Chris Gammell: I think your channel, maybe it was just because it was the first, but yours got the most backers of any of the promo that has been done so far.

Dave Jones: Nice. There you go.

Chris Gammell: So definitely worth it. Don't stroke. Don't stroke.

David Kronstein Tesla500: We don't need this on this, on this show. We don't, we don't need to stroke his ego. Okay.

Dave Jones: Yeah.

David Kronstein Tesla500: We don't, no, don't, please don't.

Dave Jones: Excellent. Oh, that's, that's good to hear.

Chris Gammell: Get a slow motion video of Dave looking smug. How can you track that? Can you just based on time when the video was released and when pledges came in? Oh, right. When it was released.

Dave Jones: Because I, I, I put a link. Can you see where links come from? I actually forget. Yes. See where your traffic comes from?

Chris Gammell: The problem was most of the links were from YouTube and YouTube is just one thing. It just says YouTube.

Dave Jones: Oh, one thing. Oh, okay. Right. Yep.

Chris Gammell: I should have, that's where I failed. I should have made different, you can do special referrer links. I should have done that for each channel. Yes. But I didn't do that. That was a mistake.

Dave Jones: Then you know. Yep. Oops. Next time.

David Kronstein Tesla500: Do that on your next million dollar Kickstarter. Yep.

Dave Jones: Where, where are you hoping, do you have a secret? And like, are you looking to crack the seven digits? I mean.

Chris Gammell: I would like to. It's a little bit scary doing that all at once. And in some ways I would like to do it later because then I don't have to pay Kickstarter's cut. Oh yeah. Yes. Yeah. Sell more cameras later.

David Kronstein Tesla500: Yep.

Chris Gammell: Yeah.

David Kronstein Tesla500: And you are going to keep selling these though, right?

Chris Gammell: Oh, definitely. Yes. I'm going to set up a store pre-order on my website and then just sell it directly afterwards.

David Kronstein Tesla500: Well, this has been a solid hour and a half, two hours. Two hours. Oh wow. Two hours. Hour and 53 minutes. What, what are we missing? I mean, there must be something that we're missing.

Chris Gammell: Oh, I have lots of stories. I'm about to come back some other time and tell some of these stories of development.

Dave Jones: Yeah.

David Kronstein Tesla500: You're welcome back anytime.

Dave Jones: Yeah. It's probably, yeah. We need probably another half an hour of stories, right?

David Kronstein Tesla500: Yeah. Yeah. We don't want to step on Vincent's toes and, you know, go over the three hour mark or anything. Yep.

Dave Jones: So, all right. Well, let's, yeah, let's end it. It's been two hours.

Chris Gammell: So, okay. So where can people find you? Yep. They can find me on my website, uh, crontech.ca. They can find me on, uh, YouTube as Tesla 500. Also as, uh, Willit Moe. And of course I'm on, uh, requisite Twitter and, uh, Facebook and everything. Gotcha.

Dave Jones: And I asked you this before the, before we started recording. Why is it Kronos with a C and not Kronos with a K?

Chris Gammell: That was because my significant other suggested the name and then I didn't pay attention to whether it was K or C and I just, oh, I'll just whack that on the sulk screen. I'll send the board out to production. And then, wait, that's supposed to be K. I've already got that. I got the promotional. I've lasered it on the case. I can't change it now.

Speaker ?: Oh, no.

Chris Gammell: And now we're sort of stuck with that. Yeah. Also, the other reason was if you go say, go Google Kronos, they're going to do it with a C because that's the normal way of doing it. I don't, then you don't have to say, oh, Google Kronos with a K. It was a K-R. It was a K-H-R. It's, I think it makes it a bit easier.

David Kronstein Tesla500: Right.

Dave Jones: Right.

David Kronstein Tesla500: But it is by Krontech with a K. Yeah.

Dave Jones: At least I got that right. It is by Krontech. Yeah. With a K. Yes.

David Kronstein Tesla500: Makes sense. Awesome. And then one last thing. We will have an email address that you'll set up if people want to send you a resume specifically, right? Yes. Awesome. That should be.

Dave Jones: So that's for a software Linux-y penguin geek.

Chris Gammell: Yes. That'll be, I'll make that jobs at Krontech.ca.

David Kronstein Tesla500: Okay. And we will link that to it. There it is. Cool. Well, David, you're a monster, man. I don't know how the hell you did this. Like, monster in a good way. Like, yeah, this is, this is so impressive.

Dave Jones: I was just blown away at the end there when you, when you said that you have a significant other. Like, I think it's the time to do this. Yes. It's like, you know.

Chris Gammell: It's a bit stressful trying to manage everything, but it's not too bad. Yeah. Yeah. Yeah. That's crazy.

Dave Jones: And you would not have been able to do this if you weren't full-time at it now, really.

Chris Gammell: No, I would have been delayed very, at least into the new year. And then it probably would have, yeah. Would have just been ripping my hair up the whole time.

Dave Jones: Yep. Yep. Terrible. All right. Well, I hope you cracked the seven digits. Me too. Yeah. There. It's, it's worthy because this is one of the, you know, this is a genuine Kickstarter, not a, you know, bullshit entrepreneurial marketing piece of rubbish.

David Kronstein Tesla500: Awesome. Well, it's great. Looking forward to seeing more of this stuff and seeing lots of, man, I love those videos too. I love seeing this stuff. Awesome.

Dave Jones: Yeah.

Chris Gammell: I want to get some more, I want to, I wish I had more cameras. I would keep shooting more videos. I just set them all up and I gave them all away.

David Kronstein Tesla500: You should set up like a, if you can ask the people that already have them too, to like set up like a tag on YouTube or something, that would be great. So like people could just see that stuff.

Chris Gammell: They wanted like a support forum or a wiki or something. I'll have to set that up as well. Yeah.

David Kronstein Tesla500: Okay. That's good. Well, thanks again, man. Sweet. We'll talk to you soon.

Chris Gammell: Thanks, David. Yep. Thanks for having me.

David Kronstein Tesla500: Catch you next time.

Chris Gammell: See you later.

David Kronstein Tesla500: So if you could tell us what you had for breakfast this morning, that'd be a good test.

Chris Gammell: I had mini wheats.

Dave Jones: What are mini wheats?

Chris Gammell: There are these shredded wheat things. Do you have them in Australia? Maybe they're called something else.

Dave Jones: Probably.

Chris Gammell: Yeah. They're basically just like shredded wheat compacted in these little, I don't know, pellets.

Dave Jones: Oh, yes. Yes. I've seen that right.

Chris Gammell: Probably not a very appetizing way of describing it, but it's correct.

David Kronstein Tesla500: Yeah.

Speaker ?: I'm going to turn it in.

David Kronstein Tesla500: And now I will insert a clip about shredded wheat.

Chris Gammell: I should record a video of shredded wheat dropping into a bowl and you can put it at the end. I have to do this now.

David Kronstein Tesla500: And you can do like another one of like it breaking apart, like slowly breaking apart.

Chris Gammell: Oh, hydraulic press for shredded wheat.