Embedded - 294: Ludicrous Numbers of LEDs (Repeat)

Episode Date: August 25, 2022

Mike Harrison challenged us to a PIC fight on twitter. Surprisingly, no blood was shed and we mostly talked about LEDs and art installations. Mike’s YouTube Channel and his website electricstuff.co.uk. He's on twitter as @mikelectricstuf.  Here's a link to what prompted the show: PIC fight on Twitter. His professional hire-him-to-work-on-your-neat-stuff site is whitewing.co.uk For driving LEDs, Mike likes the TI TLC5971: 12-Channel, 16-Bit ES-PWM RGB LED Driver with 3.3V Linear Regulator.

Welcome to Embedded. I'm Elysia White here with Christopher White. A few weeks ago, we mentioned a party to celebrate the 300th episode. How did that happen? It will be the afternoon of September 7th in Aptos, California. That's a Saturday, September 7th. There will be the afternoon of September 7th in Aptos, California. That's a Saturday, September 7th. There will be more details coming up. Most of it will be through the show, so don't expect to necessarily hear about it if you're just reading the show notes. In that same show where we announced the party, we had a listener email that asked about pics, and I thought we gave a perfectly balanced answer. But then on Twitter, we got a message. Hey, at Embedded FM, I know you're fond

of hating on pics. I've used them forever and would be happy to come on the show sometime to defend their corner. See, I think that's inaccurate. I think we're fond of kind of genially disliking on pics, not hating so much. I guess I should have made him read his tweet because our guest this week is Mike Harrison of Mike's Electric Stuff. It's a good excuse to chat with somebody

I wanted to chat with anyway. Hi, Mike. Thanks for joining us. Hi. Could you tell us about yourself? I'm a self-taught electronics engineer. I've always liked taking things apart. And at some point I thought it might be interesting to do a YouTube channel recording me doing what I'd probably be doing anyway, taking things apart. My day job has started off as a micro control consultancy and i've sort of drifted into a weird little niche of

uh lighting architecture um lighting design for basically usually things involving ludicrous numbers of leads incidentally uh there may be some controversy i tend to say lead rather than led it may cause some confusion but it saves time that's my argument and i'm sticking to it okay uh so we want to do lightning round and uh while you gave us lightning round questions we're not going to use them fine worst thing you've ever torn down? Worst in what respect? Just the most depressing assembly or electronics job. Can I swivel that around to the most dangerous thing? Sure.

It was a small fuel cell in a completely ridiculous product, which basically was a fuel cell that cost way more than a similar sized usb battery pack but when you open it contained this very finely ground metal which basically bursts into flames as soon as it reaches air that was an interesting surprise all right uh what is your favorite flux for putting things together um my favorite flux i'll call it here, it's actually, this is Future 315. It's like a very liquidy, it's like water liquidy type thing

I use to refill flux pens. Is having one of your lighting installations lifted by a helicopter the best thing ever? It's up there, definitely. Definitely. Basically, we figured out the original plan was to attach lights to the helicopter but that involved all sorts of approvals and things and but if we just sling it underneath it just counts as cargo so there's very little paperwork or some other?

Never, ever a NeoPixel. They're just not reliable enough for use in big permanent installations. They don't like being soldered. They don't have enough grayscales. They're inefficient. They're fairly horrible. What is a tip you think everyone should know? When you're designing a board layout, placement is everything. fairly horrible. What is a tip you think everyone should know?

When you're designing a board layout, placement is everything. Spend more time on placement, and then the routing just does itself. Okay. Do you want to continue, Christopher? No? Whatever you want. I actually want to ask Mike the questions he wanted to be asked. Okay, let's do that.

But then I'm going to ask why, because it's not not lightning right anymore what's your favorite lead driver i see uh texas instruments tlc 5971 why um because basically it's a 12 channel driver with a spy like interface which has 12 bit pwm on the outputs plus an additional seven bits of current control so it means you can get super smooth grayscales you can stack literally hundreds of them in a chain um and they're just you know they they work really nicely they're easy to use they're easy to drive and they're yeah they could be a bit cheaper but apart from that um they are i use them a lot i've used thousands of them often in applications where if i could buy a reliable neo pixel type device from someone like ti or osram i would use them but you know that we just can't can't really do that for various reasons so i've i've built far many boards than i wished i had with literally

just loads of those tlc5971s with either rg with either RGB or RGBW LEDs on the other side of the PCB. And you said SPY-like. SPY requires one line to go out to everything, but then you said thousands. So is it more like a daisy chain? Yes, it's a daisy chain. It's similar to the APA102 type LEDs where you've got clock and data going in

and then coming out again. And the chip takes all the channels it wants and then forwards the rest down the chain. Cool. That was the chip I was trying to remember. Yeah. Okay, so favorite LED color. I assume it's black. No. That's the easiest color. Yeah, I have created the occasional dark emitting diode in my time

but uh no uh white um for a number of reasons one is it's very very efficient you get a lot of light for very little current um there's also it loses any temptation to do horrible disco mode colors um they're also extremely cheap you know you can buy your osram brand name white leads for the u.s equivalent would be about two two cents if you buy them by the real so they're good quality leads and they're just very very cheap and i've there's one particular one the juris e3 which i haven't actually counted but i must be getting close to have designed in a million in various installations. One installation alone, the Heathrow Airport one, that was 350,000. That's a lot of soldering. Sorry. Yeah, somebody else does that.

I do have my own pick-and-place machine for smaller jobs, but anything that scale gets subcontracted. What color of white? Oh, boy. Wow. About, I would say generally about four to five thousand k that's pretty cool right yes yes it's like not not bluey cool but on the cooler edge of you know warm but although sometimes it's sometimes the case where projects happen on us on a time scale such that we're limited on color choice based on what people actually have in stock.

That seems fair. That is often how I choose color temperature for our home light bulbs. No, no, you don't. Well, yes, that's why I'm not allowed to anymore. What's your favorite voltage? 24. That's a lot higher than I would have expected. Well, no, for large-scale stuff, you need a reasonably high voltage to avoid voltage drops.

24 is quite a nice compromise. There's gazillions of 24-volt power supplies out there because it's very common in the industry. Obviously, you can go higher, but as soon as you get higher, things like circuit protection become a little more tricky because obviously you've got like one big power supply that's often being divided down into like smaller units and smaller fixtures so you want to run thin cables so you have to have fusing to protect those cables my typical use or poly fuses and once you get much over 24 volts then you know those devices get thinner on the ground or you have to go to through hole. And I've occasionally done stuff with 48 volts, but you have to bear in mind that a large 48-volt power supply

is not much different from an arc welder. 48 volts likes arcing a lot more than 24 volts. All right. Yes, I can understand that. Okay, now, Pix, we might as well just get this over with did we really slag on picks that much i seem to recall we heard we said something like yeah electrical engineers like them and we don't i mean we were meanish but in a bit of a recurring theme it's not just like the last episode i seem to remember it's come up on multiple occasions. Yeah, and the last one I think was like the nicest we've ever been.

Even we're getting tired of hearing ourselves. Okay, so being an electrical engineer, you completely confirmed our biases because we said EEs like them. But, okay, yeah, why? Why are they better than other microcontrollers? Well, obviously, there's no such thing as the best micro. Everything is, you know, with anything engineering, it's more or less applicable. But, you know, often there's a focus on, oh, it's a weird architecture and the instruction set's weird.

But, I mean, the core CPU just almost never matters. You're writing stuff in C. It doesn't matter if that gets compiled to PIC code or MIPS code or ARM code most of the time. It's all about peripherals. And it's also all about things which are nothing to do with the Archon. For example, most of the PICs I tend to use are the very low-end, things like the 6-pin SOC23s, the 14-pin PIC-8s, and then the mid-range 32s. But across nearly all their range, almost every device,

you can get that same device in DIP, SO, SSO, and QFN packages. So you've got a huge range of choice. The other thing which is maybe less relevant now that they bought AppMail as well is that with their programming service, you can get them to pre-program the chips for you very, very cheaply. For example, in a SOC 23 device, it's something like four or five cents per device for them to program it and put a colored ink dot on it. So if you're doing something, I've done installations that have had thousands of very low-end picks in them. It means that we can just get them already programmed,

even if that's only got a bootloader in it and you put your firmware in later. Imagine even if you're on board programming, a thousand devices take a lot of time. So that's a really major thing. And also, historically, you've always been able to buy them. Microchip have always been very good at, A, almost never obsoleting, but also keeping stuff in stock. It's very rare that you can't get a particular device,

and even if you can't, there's a good chance there's a pin-compatible, like Nexron or like Windows Series, that you can. So a lot of it's about those, not really about the technical aspects of the device, it's about the logistics of using them, getting them. The other thing is, I tend to use the two opposite ends of the scale. And the nice thing is that from a 6-pin SOC23 up to like a 144-pin 32-bit device, you use the same programmer, same development environment a lot of the peripherals are either identical or very similar so you know in a lot of cases the best micro for

a job is the one that you know so you know it means that you've got access to a very wide rate of capability and prices without having to do much into you know you don't have to relearn the dev tools you don't have to relearn the peripheral so a lot of it's nothing really to do with the yeah certainly the core architecture it's yeah it's the actual practicalities of using the part and that makes a lot of sense um especially especially when you're using really tiny ones uh yeah and they are basically plcs to some extent i, they're being pre-programmed, you're using them to do something pretty small, and you want a lot of them. Yeah, and I think that's where

we talk past each other sometimes because most of the things we're doing are pretty processor intensive, and so the architecture does become very important. Like, oh, I do need a floating point unit that is reasonably fast for a particular application. So don't you know i'm not looking using eight bits for anything very often um so that that kind of eliminates a whole class of things

uh and as software engineers we don't care about logistics somebody else's problem somebody probably should yeah just to give an example of like a typical like large-scale lighting installation how what that might look like it could be let's say you've got like say a few hundred fancy looking pendant fixtures with you know maybe 20 or 30 leds in them what you would um typically have is that fixture would have like a little maybe a 14 pin or a 8 pin pick which is taking let's say a 50k board ttl level ur signal and either say using the pig's own pwm to to um control the the leads or driving let's say the texas 5971 drivers that is then fed from like a little box that's got probably a mid-range 32 pin sorry 32 bit pick which is taking maybe a four megabit rs485 signal and then

splitting that out to maybe 16 12 or 16 of those 50k board uart um signals one to each fixture so you've got like a 12 let's say 12 connectors on this board that provides data to one fixture and fused protected power to that fixture and the code in both those you know i very rarely write anything that's more than a few k of code right if that it's just basically doing a small amount of stuff to a lot of data off you know at varying speed from like fairly quickly to really slowly uh depending on the application so it's very different from the thing with the huge user interface and internet of Things and connectivity or whatever. So, you know, it's important to remember that if you're working microcontrollers, you sometimes end up focusing on the sort of things that you do, whereas there's a whole world of other applications.

So it seems, yeah, pretty much whenever I've heard you guys talking about it, it's always been in the context of, you know, big applications that take months and months to develop yeah a lot of the time you know the code will quite often just take a couple of days and that's it finished uh including a bootloader so that i can upload the firmware for both that splitter and the fixture over the wire um because it's very very low i don't like writing software i much prefer building hardware so the less software i have to write the better yeah and i think i think that point is uh a lot of people complain about MPLAB and the tools, especially since they're not free. There's certain limits of size in the things. But if you're working on tiny, tiny projects and you're doing focused applications, but very application-specific things with them, then that doesn't really matter either. Right? I mean, if it's going to fit, it's going to fit.

People, I think, make a bit too much out of the non-free thing. In practice, the 32-bit ones, the free tools work absolutely fine. No problem at all. There's not a huge difference. When you get down to the 8-bit ones, the free tools aren't very good at code optimization. And we're only talking about the C compiler here. We're not talking about MPLAB.

MPLAB's free. It's just the actual C compiler that we're only talking about the c compiler here we're not talking about mplab mplab3 it's just the actual c compiler that we're talking about and um the nice thing about is mplab you just install it run it and it works so this is when i did the um conference badge for the hackaday supercon a couple of years years ago yeah that was a nice thing that i could just give give someone the mplab project they just install mplab load the project press a button and they probe there and they're uploading code straight away you know it's easy it just it just works okay sometimes it's a bit slow on some machines but generally you know everything's in one place you don't have to like find one id from somewhere else and a compiler from somewhere else and mess

about with linkers and the other nice thing the way they handle devices is really simple and literally all you do is you you is you hash include pick.h. And that's it. That's all you need to do in your source code. It handles the default linker settings and everything for you just happens. So if you want to just get a simple job done quickly, it just saves a ton of time. And that makes sense. If you become very good with one thing, like PIC is controlled by microchip from the silicon through the C compiler. where they are trying to compile for everything in the world, and they don't have to, as you mentioned, linkers. They aren't trying to work with different linkers or different machine codes.

It's all PIC, and it's all controlled. And that gives you a lot of ease of use, but it makes it harder to change to other tools. Yeah, if you want to change to other tools. Yeah, if you want to change to other tools. Obviously, yes, I mean, if you're doing something complicated enough that their environment isn't good for you, then you probably are a bit more limited. I mean, let's say the 32-bit.

In fact, the 32-bit tool chain is based on GCC. And in fact, there are ways of making, of overcoming the paid-for restrictions in a way which is probably not quite copyright infringement because GCC is a free tool. There are apparently ways of fairly simple ways of tricking into thinking it's the paid-for version. There's a bit of controversy about how they've actually done that,

which I don't think you know i'm really interested in resolving uh but again on the 32-bit uh compiler you know it's i don't think there's anything i've done that um really needs that last i think it's mostly about code size rather than speed but i've never had issues using the free version of the 32-bit compiler. The 8-bit's a slightly different matter, but the 32-bit, the free version is just fine. I have the 32-bit, and I'm going to disagree with you, although perhaps I need to go find your workaround.

And my problem isn't that it isn't fast enough. It's that I know if I optimised more, I could get to a lower power lower power which is one of the great things about pic is that you can get to low power but i need to optimize more than the free compiler because i know there's stuff there and i can't leave those microamps on the table yeah well that's pretty much just how long you're awake sort of issue yeah do you normally program in c or do you program in assembly when i started using the pic back in like when the original um otp 54 chips came out i did a lot of assembly it took quite a while for me i've always had a very big distrust on of like big bits of software and so on so it took me quite a while to sort of see the light that C was actually the way to do it. And I think C is better.

Compared to Assembler. Having said that, I often write C. I generally treat C as a glorified macro assembler. So, yeah, the sort of code that I like writing the most is the sort of code that you need an oscilloscope to debug and test. You know, squeezing, doing things like, um you know 12 bit bashed uarts that's you know 250k board in software that sort of thing is the sort of thing i like um doing and like doing really weird things with dma controllers and and that sort of thing um but um yeah i i mean i i've never got into i can vaguely follow

the 32-bit Assembler. I've never actually spent the time to look at the MIPS stuff. But certainly on the 8-bit whole cart, sometimes I will look at the Assembler output, the compilers generated, and tweak the source code to make it compile to the Assembler that I think it should be assembling to for speed and so on. Well, we're not arguing. I thought this show was going to be all about arguing with each other i still don't want to use them fine but it's fine if everybody else does that's cool

i mean i there are places to use them i didn't realize they would pre-program them for you that's a huge that's a huge benefit there are times that and it's a very low minimum. It's either 150 parts or $150 and a trivial setup charge. And it's quick. I mean, if it's a stock part, it'll add maybe two or three days to the order from Microchip Direct. So it's actually a very major advantage in some applications. Because although there are third-party programming services, they're generally a lot more expensive because they've got to take the part out, you know, re-reel it and generally mess about. But also, I believe DigiKey offer a programming service,

but they don't offer it outside the USA because they're scared that you might send them some code that's got export control, control degree encryption. Yeah, my weapon's great 8-bit processor. Yeah, exactly. It's ridiculous and it's annoying, but that's their attitude. Yeah, well, that kind of service can make the difference between,

hey, I can do a product on this and I can't. So that's pretty interesting. And setting up a manufacturing line is just so hard. For a small run thing. I mean, on the production side, another neat thing, I'm sure there probably are other tools that do it but again this is nicely integrated into mp lab on the picket three programmer it's got this facility called programmer to go which means you can load your code into the programmer and it then becomes a standalone program you just literally give it

power press a button it'll program your part and then show a red or a green led when it works so that if i'm doing a small production run i'll just program one of those programmers up and send it to my subcontractor and then that becomes part of the production programming test um procedure which again that's you know it saves a hell of a lot of time uh especially you know almost all of my jobs are one off so we don't have time to optimize a test procedure it's a case of okay here's the job build it test it this way and send me the boards. So to be able to just send them a program and say much cheaper than trying to do this with a Cortex M something where you need

a JTAG programmer and a laptop. Or some, you know, multi-gang system with 8 UARTs and yeah, we've done all of those things. I will say I'm a huge hypocrite because the first microcontroller I ever used was a basic stamp. Yeah.

Which I love. Yeah, I mean, the one thing I think Microchip did is they really got people through to the idea of using at that time one-time programmable parts in production, because before that, you had things like 8051s with external EPROMs, you had very expensive

chip packages with an EPROM actually plugged into the back of the top of the package itself to emulate parts that were going to be burned into Mastrom in future. And Microchip were really very much the pioneers of low volume, you know, single chip micros. And now they've eaten Atmel, which was the other one that was good for small micros. Yeah. But Microchip were before Atmel in that respect. Yeah, definitely. for small micros yeah but my microchip were before atmel in that respect certainly yeah definitely yeah okay so it sounds like we're mostly in agreement on all of this much to my shock and horror uh tell me about your lights okay well um maybe i think probably about 12 years ago i sort

of fell into this rather interesting and weird world of basically when people ask me I tend to sort of you know suggest the Venn diagram of if you draw a Venn diagram of artists architects and lighting designers most of my customers fit somewhere in that in that area but generally not completely inside one of them and what I've sort of found the niche that I've fallen into is when people want to do something like a large-scale lighting installation or an art installation and they can't do it with off-the-shelf electronics yeah they call me and it's one of those areas that everyone's you know lots of pretty much everyone in the area knows everyone else for example one of the customers

i worked for originally loads of people left that company started up their own practice or gone to work to other people. So it's all like word of mouth. And I now routinely get like second and third hand referral, like someone will email me and say, hey, XYZ person recommended you. Now I have no idea who that XYZ person is because it's been a second or third hand sort of referral.

Working with artists seems challenging, both because they tend not to be paid and because their ideas may not be physically possible. Do you find either of those to be true? Not to a huge extent, because most of the people that I, artist-y type people I work with, are already fairly well into the market.

They understand. They're not people that have just got some random ideas um there's probably two of my clients that are fairly big into literally like hang on the wall art type things and both of them have been at it long enough to have at least some sense of what's sensible and they know now know enough to get me in early enough so that you know they don't present this amazing like structure and okay, that's very pretty, but where are you going to put the wiring? You know, things like that. They tend to talk to me fairly early on so I can guide them, you know, into what's feasible, what's practical, cost and so on. So that's not, you know, we're talking sort of mid to high end here, not just, you know, the struggling artist

type people. and so how much input do you get into the design beyond the technical limitations um there's usually some again it varies from client to client there's some where i've literally designed and implemented all of the content side of things because i just hadn't got around to doing it by the time this thing had to be installed and uh um and working and other cases they've got a very firm idea of what they want but there's always a certain amount of negotiation in terms of okay you know that's very nice but if you do it this way it's going to cost you half as much that this sort of thing like you know they say okay right we want

we want 20 000 leds and we want every single one of them addressable separately and i said well actually do you if you can maybe say let's say, have three LEDs per pixel, you know, it's going to slash your power, slash your cost, whatever. And they're often quite amenable to, I might say, okay, well, okay, let's build one prototype that's fully addressable. And then, yeah, play around with your content to see how many pixels you actually need. So I'm actually doing something exactly like that at the moment. It's actually for a lighting rather than art type thing.

But they know they want some sort of movement within the light fixture, but they don't really know what resolution. So we've made some fairly impractical for production demo strips that they can then play around with to test the visual effect of different resolutions on. What kind of visual effects are we talking about? Organic things or or anything pretty much it's either you know organic things or it's things for example the the heathrow thermal

two installation is over a seafood concession so it's meant to represent like shoals of fish swimming around in a fairly abstract way um sometimes it's just pure aesthetic sometimes it's almost playback of video so you know low res video content over maybe a slightly unusual type display um but displaying what is basically video that they generally i'm the the beating point between what i do and what my customers do tends to be the usb port so they generate all the content they write software to do all the graphic stuff and then they just throw pixels at me and I display them. That's the most common sort of scenario.

So there isn't much interaction? There can be. I mean, the problem with interaction is it either has to, and this has sort of become almost Mike's Law in interactive lighting installations, in that unless it is really immediate and obvious like you move wave your arm and something immediately happens it might just as well be random because you see so many installations where you know this the color of this thing depends on the tide in this country or whatever you know and it just yeah most people just don't get it whereas if they can literally work, the thing that people engage with most when you actually see people engaging with things is if they can stand on there and wave their arms and this thing does something cool, they love it.

And it's very obviously interacting. But there's a lot of times when as soon as you break that immediacy, it's very, unless you either get it explained to you or see it on a time lapse, it's, you know, it might just as well be random. LEDs have changed so much over your career. We already mentioned NeoPixels. And things have gotten smaller. Blue is now a viable color. What have been the most interesting changes to you? I mean, there's obvious things like the fact that white LEDs are so efficient

and so cheap, which means that, you know, you can use zillions of them. And certainly the cost of placing them on the board is more than the cost of the device itself. That's probably one of the biggest things. And I tend to prefer doing stuff in like single colors because, you know, as soon as you go multicolor, it's very, you're very risk-tipping over into horrible disco. Like when desktop publishing first came out, people used every single font they could just because it was there. It's the same sort of thing. I mean, certainly I'll tend to, if somebody wants RGB, I say, well, no, you probably actually want RGBW. Because the white will give you the light, the power you want, the optical brightness that you need. But it gives you a lot of subtlety and you can produce a like nice pastel shades and subtle

colors whereas if you have rgb and you want to do white it always looks horrible if you've ever seen an rgb display trying to look white it looks terrible because it's never quite balanced you've got three separate colors where each led has slightly different intensity and like um color it's you know you're never ever going to make that show a nice white whereas with white leads where each LED has slightly different intensity and color, you're never, ever going to make that show a nice white, whereas with white LEDs, it not only appears white and uniform, it also takes maybe about a tenth of the power. Okay, RGBW.

I'm frantically Wikipedian so that I know the difference. I assume it's not the royal gloucestershire berkshire and wiltshire regiment it's red green blue and white wikipedia i like that there was another yeah i know i needed the the disambiguation page so so rgb makes sense we i've used those a bunch and the white comes in because mixing rgb to make white isn't as good as just having a white led but do you use the white other than the rgb i mean you mentioned pastels is that like in order to get a light red you use white and red instead of yeah okay whereas for example let's say if you want like a nice subtle pink if you try to do that with rgb

all you're doing is you're fading the blue and green down very slightly so you have again because most of what you're generating is white that's never going to look very uniform it's going to you know almost always look really horrible whereas if you generate white and then just add a little bit of red to it it it looks a lot more uniform and it takes a lot less power. This is very true. All the times I've played with RGB LEDs, I've had this problem and I didn't even know there was a solution. Thank you. When did these come out? Well, I mean, you've always been able to get an RGB LED and stick a white LED next to it. There aren't that many single parts that do all four. There are some Chinese ones which are sort of okay.

That's about the only – I don't think there's any low-power RGBW LEDs from, like, mainstream manufacturers. There's quite a lot of the higher-power ones. And by low-power, I mean, like, sub-50 milliamp. And the higher-power ones are the sort you tend to need to stick to stick on a heat sink um but what i almost always have to do is take an rgb lead from someone like cree and stick an osram white lead next to it on the pcb as close as i can get it oh okay so you you're making the rgbw unit. Yeah. Do you sell those? I mean, not as such.

No, I mean, basically what I tend to be designing is custom LED fixtures or control solutions. So, you know, for example, a job I did about a few months ago, they wanted one meter diameter rings with RGBgbw on the inside and the outside so these are like two pcbs one inside one outside each one was made out of two one meter long strips which had rgb rgb and w leads alternately on one side and drivers on the other side so that was just a custom job where for larger jobs like that i'll do i'll just charge a fee for the design work and the prototyping and then just give them all the information they need to give to a subcontractor to actually manufacture oh i thought you had a store where you could buy stuff um there are one or two things which i've designed which have are now sort of occasional products those are basically high

channel count dimmers for um running rgb single color LED strip, the 12 and 24 volt LED strip. So these are 48 and 96 channel. Basically, they're just PWM outputs to drive constant voltage LED strips. And those have been used in quite a few installations again you know if you want a large number of you know particularly strips occasionally sort of um you know higher power clusters and so on but it's it's generally uh strips and in fact the reason the 48 i think the 48 channel came about there was an installation we were doing at um about five years ago at heathrow airport where we had some large butterflies which had RGB lead tape around the edges so you could light them up and they had like etched patterns in them and there were quite a lot of

these I think we used I think I can't remember the numbers it was something like eight 48 channel dimmers so each 48 channel dimmer would run 12 RGB strips at 24 volts and then the the 96 channel that was actually a company that makes very large architectural models that they want to light up. So, you know, like huge building developments on like a model the size of a table. And again, they had all the lighting they wanted to control. And if you want to control, you know, 96 separate either LEDs or strips of LEDs, it gets really messy doing it with like four or eight channel off the shelf controllers. So that leads me to a question. Some of your installations and things have been, like you said, somebody comes to you and says, I want 20,000. doing it with like four or eight channel off-the-shelf controllers. So that leads me to a question.

Some of your installations and things have been, like you said, somebody comes to you and says, oh, I want 20,000 independently addressable LEDs. Yeah. What is the architecture when you finally arrive at something where, okay, it is a large number beyond 96 in the thousands or hundreds, what does the architecture of your system actually look like to distribute that? It varies. Yeah, I mean, it does vary.

But a very common feature is there is a custom fixture which takes 24 volts and a TTL level data signal. Let's say, for example, it's maybe like a one meter long board that's got LEDs on both sides that sits in an acrylic tube, for example, for like a hanging pendant-type fixture. There's another one I'm doing at the moment where each node is like, I think it's about a two-inch diameter board with about 12 RGBW LEDs on it, and that's going to have like an acrylic diffuser on it. So basically it's a semi-intelligent fixture that takes typically 24 volts and a signal line data signal,

and then that then plugs into a board which takes either Ethernet UDP or RS-485 at fairly high speed and then splits that out into lots of channels at low speed. So in terms of addressing, this splitter unit has a dip switch on it that says which device it is on the RS-485 bus and then which socket you plug your fixture into determines the sub-address within that. And you might have some number of these boxes

from a few to 20 or 30 on perhaps multiple RS-485 buses. I've had installations that have used a 16-port RS-485 to USB converter, throwing data out at, I think, i think four mega board on every single port um for most of the time but my general philosophy is keep it as simple as possible so rs485 is easy because you can just stick on a scope and debug it and do the protocol decodes it's super easy to decode but also you know by splitting into a modular system you know you have the fixture which is basically independent so if that fixture works a thousand fixtures are going to work you have that splitter unit which again if that can drive 20 splitters you can then just multiply those up to however many you need you just need to keep a handle on

overall system bandwidth but also power because obviously power is one of the biggest issues you know you've got one of these splitters might typically have a 24 volt maybe 20 amp power supply connected to it and each output might be have poly fused at perhaps one and a half amps something like that and then within the fixture would either have for example a bunch of let's say six leads in series per pixel running directly off that 24 volt supply or it might have a dc to dc converter in the fixture to convert it down to say five volts if you need to run the LEDs individually so that you end up with a nice thin cable to the fixture and you're not you don't care about voltage drop because that dc to dc converter will deal with that voltage drop along the along your nice thin pretty white decorative cable you make it sound like this might be a little harder than just slapping some LEDs on a controller.

It's easy once you know all the snags. And things like power, voltage drop across the cables, moving your earth ground level, and getting to really, really know your USB to RS-485 converter to figure out how big a packet you can send before the thing stalls. So if you're sending data out to 16 RS-485 ports simultaneously, it's absolutely essential to get the packet size right because it can make a factor of like 30 or 40 difference to your overall throughput. Now that's a software problem. Yeah need to i need to tell the software guys about it because most software guys they're in yeah things like throwing data to an rs485 interface i usb it's

often right at the edge of their capabilities because they're their software guys so they tend to have to be handheld to some extent to explain okay this is exactly what you need to do your packet needs to look like this you need to send split the packets up you need to send like 1k to this port then to that port then to that port before you don't send any more to the first port and all this sort of stuff so i have some standard documents that i tend to send to people if it's like a new custom with a new programmer i sort of send it to them and say you know read this understand this i'll quite happily talk to you if you don't understand it and in what there's one i mean a lot of the programmers working in this area yeah they're into like the graphics and the 3d stuff and all the

that and there was one occasion with a programmer who i shall not name who when i sort of send them documentation showing for example you know they need to send this packet and this bit controls this and that bit controls that he asked me to explain how binary worked ouch ouch yeah don't start don't start from there one of the things i want to go back to was dimmers um which was you mentioned the 96 channel dimmers and i got a little caught up for a second there because looking around my house and LED lights, some of these lights are very bad at dimming. Why are some LEDs so bad at dimming? Because your LED light bulb has a driver inside it that takes your AC mains voltage

and produces constant current from the LED. Traditionally, traditional light dimmers that drive tungsten lamps, you know, will mash the waveform up to give the tungsten filament the amount of power it wants, which works fine for tungsten. But when you feed that into a switch mode power supply, it usually doesn't end well. Yeah, there are drivers that are designed for dimming. You'll often find, if you look at even just domestic uh lead house bulbs they'll say dimmable or non-dimmable and that's about the architecture of the internal

driver and even the dimmable ones have some limits you generally they won't generally dim all the way down to zero they'll have a minimum of like 20 or 30 below which it gets really flickery and you can now buy like wall dimmers that you can set a lower limit on to keep it out of that range. Okay. Going back to what has changed over your career, was there anything that you were like, yeah, that's never going to work. We're never going to get that, that we did something or something where you

were like, oh, in a year we'll be able to have hover cars. Not really. I mean, most of what able to have hover cars? Not really. Most of what I do is pretty low level, really. It's controlling LEDs. There's a scale aspect, but that's about just divide and conquer. It's really just the reduction in cost of LEDs themselves. But also, there are some things that I do which are things other than LEDs. I've done a few installations using large liquid crystal displays like say for example a three by three inch single pixel liquid crystal display that goes either transparent or not

transparent and the nice thing about that is the power consumption is very very low um so you can have like huge arrays of those There's actually one we did quite a few years ago in a shopping centre in London that was, I think it must have been about 50 feet high with huge panels on each side with these reflective LCDs, which look great. And the whole thing was powered at 12 volts down a Cat5 cable because these things just used no power because they weren't emitting light. One thing that I would really like to do at some point is a large- scale e-ink e-paper type installation again for the same reason you can get you know large large scale

stuff happening without having to have power cable ridiculous power cables all over the place we need a higher voltage for those but you don't need to power them at all sometimes yeah exactly the fact you can get large area i I think at the Electronica show last year, I did see one company making very low-resolution e-paper displays for architectural applications. So I sent information to a few of my customers that got quite excited about that. Nothing's come of that yet, but I think it probably will at some point

because obviously everyone's looking for something that's different and new that other people haven't done. I'm surprised that I'm not aware of any large-scale ink stuff that's been done um yet and it's obviously when the ink first came out you could only buy them if you were making kindles but obviously now it's easier to buy lower volume stuff but most of the products are still aimed at like text and graphics display whereas what i want is like something that's maybe a4 size with perhaps 16 pixels, something like that. You could have a building that some days it was black, some days it was white, depending on what the building wanted.

I want a display that you can only make out from space. That was actually a brief on a project I talked about a while ago, but it didn't happen. 300 DPI, but you have to be 200 miles up. Yeah. Okay, so it sounds like we can talk a lot more about lights and ink, all of these things. But there is something else you do. You have a YouTube channel where you take things apart and you explain things together.

Yeah, you put things together and then occasionally you have a little minor rant. Why do you do this? Well, I mean, there's really, I suppose there's two strands on the YouTube thing. One is just the teardowns. It's taking stuff apart. As a kid, from when I was a kid, i've been into electronics since i was about five i've always loved taking stuff apart i got quite good at putting it back together again at a fairly early age but i'm always you know i can hardly ever buy any bit of equipment without taking the lid off to see what's

inside it and that side of the youtube is basically making content for people like me who are too scared to take the lid off things or who don't happen to be able to get hold of an x-ray machine or a pill camera or some other weird bit of labic or like a DNA analyzer that you can buy on eBay for not that much money. So that was very much, you know, doing stuff that I'd probably be doing anyway, but also, you know, exposing it to a bigger audience. The other side is I do like electronics, not so much tutorials, but one of the nice things about what I do for a job now is that all my customers care about is this thing is there and working. So it's like design it, build it, and they don't want to hold any of the IP or anything like that. So I can talk about the detailed aspects of the design of these things,

and I've done quite a number of videos on that, and those have proved very popular, just explaining why things are done the way they are and going through the design procedure. I quite like doing videos like that because I think it benefits quite a lot of people just seeing the design process from the inside. And there's just random other things like reviews. I get companies wanting to send me stuff for review quite a lot,

most of which I turn down. The only stuff I tend to take is stuff that I would use anyway, or if it's just something that's such a high price tag that I just can't resist the offer of a freebie. I totally understand. And I understand turning them down because sometimes you're like well if you send that to me and i don't like it am i going to have to say something i will spend time on it now my stand if someone if i get a random key the first thing i say back

is okay you know yes i'm interested in this i'll have a look at it but you won't have any right to review i will say what I think. I will take it apart. And if they're happy with that, then fine. And most people are. I don't think I've ever really slagged anything. If it's something which I clearly think is a piece of rubbish, I won't even bother doing it.

If it's something that's really ridiculous, that's not expensive, I might just buy and just do it for the hell of it. I'd sooner do that than actually get it as a freebie. I'd sooner just have the independence, just buy it and just do it for the hell of it. I'd sooner do that than like actually get it as a freebie. I'd sooner just have the independence, just buy it and just rip, rip, rip, rip the crap out of it.

Um, that, that hasn't happened very often. Um, cause I just don't, I don't have time. The YouTube thing is very much a hobby thing. I don't do it for the money.

Yeah. It took ages to, for me to get around to doing any sort of monetization or Patreon type things. And the only time I do that is if if i have to buy something to do a teardown or if i have to like rent a van to take a bit of kit somewhere um so yeah it just means it just basically doing that has increased the threshold of you know shall i buy this just for the hell of it on ebay some bit of weird weird kit i don't quite know what it is. So things like DNA analyzers and that sort of thing.

And you brought up DNA analyzers, which was my next question. When you do teardowns, you don't do teardowns of, like, things people expect, like the fluke DMM that's sitting on your desk. You did a teardown of a DNA system. Several. Yeah. Weren't there several? I mean, my absolute favorite thing is bits of equipment which were originally ridiculously expensive and now almost completely obsolete.

And DNA analysis is one of those fields that's moved so quickly that literally, you know, the first one that I did, I paid, I think, something like 200 pounds for. And the brand new price of that was like half a million dollars or something ridiculous. And it had like a camera inside that was like $150,000. It had this massive, great CCD sensor. But you can't even buy the chemistry to use it. So it's like having an expensive printer they don't make the cartridges for anymore. And even if they did, the cartridges would be so expensive that you could just go buy a cheaper printer that performs better

because that that field has moved moved forward so quickly but the nice thing about that sort of thing is it's niche industrial thing that was just built almost regardless of cost um so you get very nice like beautiful mechanics inside it amazing optics you know i i totally dna yes i totally understand this i i went to hp's uh bioscience division in like 1998 and we did one of the first dna scanners big dna scanners and it was definitely half a million dollars and um and now i'm working with a little company that's making one that's kind of field oriented like you take it out into the field and it's it's it's just an app it's like just point your camera at somebody yeah it's in the kilobuck range not in the millibuck range

and it's just amazing how much that technology has. Well, a lot of the cost of those things goes into the custom. I mean, it's not necessarily the parts, although the CCD cameras usually are pretty pricey in the motors and stuff. But, I mean, I was thinking that I wanted Mike to someday get a hold of the medical laser I worked on years ago at the terror department to see what he thinks. But a lot of the expense of that was just the optics. I mean, we had this gold-coated scanning wheel that cost, you know, not that much to make,

but the NRE was millions and millions and millions and millions of dollars. So, yeah, a lot of the cost is in the, okay, how do we get this all arranged, not in what parts of the... And, of course, design and development. A lot of these things are built in fairly small quantities, so there's a huge amount of development cost.

With the Agilent ones, it were all lovingly hand-built. Yes. Do you have any other favorite teardowns? Not. I'd probably have to look through the list and remind me because I've just done so many. I've not done that many for a while. But, again, it's the thing that i have no clue what's inside i know there's going to be something interesting inside but i don't really know what that's my favorite

thing it's like that that discovery and the surprise of thinking oh there's you know what's all this mechanical stuff or just some or say like in that fuel cell thing oh look it's catching fire and i'm taking it apart oh look it's catching fire well these are fun to watch but but you also i mean some of them have some interesting info like you did the comparison of the the electric car chargers which i thought was really cool yeah because it was like is it really worth it to to spend you know a little extra for one that's not from random company uh but the thing that was terrifying to me seeing that Tehran, I didn't realize they carried, you know, the full load

on traces on the PCB. It just makes me really nervous for some reason, not being, you know, an expert doubly or anything, but it's like, what, 32 amps across a couple of traces? Well, I mean, that just goes back to,

for example, that 48 channel LED dimmer. I mean, that is normally rated as a one amp per channel. So, you know, you can have 48 amps going through that terminal. So that's like a 2-ounce PCB with tracks on both sides, but with very, very careful layout design that you're not moving it very far. And I think on the 96-channel, I do actually solder some wire across to distribute the power, just to reduce the heating a little bit but um yeah

large amount of currents do do get quite fun to uh to push across pcbs but again in the case of those chargers you could see the way they designed it they did you know they designed it to minimize the length that current was going going across the pcb they've done it on both sides and you know pointing a thermal camera at it yeah they were clearly it was clearly okay because it wasn't getting particularly warm. But, yeah, once you get into 10s of amps, interesting things can start happening. Do you have any videos that you worked hard on or that you were especially pleased with that people didn't seem to notice? Again, it's hard to remember. i tend to get fairly consistent view nut view counts

obviously most of my most of it's from subscribers i don't think many people sort of randomly stumbled across it i think things like for example the car charger one that got quite a lot more because obviously there were people looking from the the electric vehicle type forums that i posted it on um but yeah i don't tend to spend too much time thinking, okay, was it worth doing that amount of work? I mean, as you can see from my production values, I don't spend a huge amount of time on that. It's generally like shoot it in the afternoon,

edit it in the evening, and it's on YouTube that night, partly because if I don't do it, then I'll be busy with something else. It will never get done. But also I sort of shoot everything in sequence sequence so i don't have to remember everything between shooting and editing so it's very much just get it out there um as easily as possible i get numerous complaints about that but you know this is just something i do for fun um and you know it's either interesting to you or not yeah whether it's superbly polished or not. It ought not to matter to most of the people that are actually interested in it.

Yeah, well, it's this or nothing, kids. Yeah, absolutely. That's exactly it. Well, Christopher, do you have any other questions? At what point did you realize you could make a career out of this? Was it fast or was it like a couple of years? It happened many years previously. I basically was starting to tinker around with picks anyway. And this was in the time before, for example, big companies like Farnell stocked them.

You had to talk to specialist distributors. So I was talking to my local distributor from a company that I don't know if they even exist anymore, just to get some parts and some programmers and just really just for playing around with them. I think I'd started doing some electronics consultancy there. And they just called me one day saying, oh, we've got this customer we're trying to sell pics to. They're currently using Maskron parts, but we need someone to do the software. Do you want to sort of meet them and you know um see if we can get something happening so i went along with him they clearly had really bad experiences with the previous consultant who you know they're using these mask grom parts

they spend months and months and you know this product was something literally it was a fire alarm sounder so it had to make like one of i think 10 different combinations of beeps and whoops and whatever. So it was a really simple thing. But this was in like a 1K memory or no, 512 word memory pick. So I said, yeah, that should be okay. I'll prototype it. I'll maybe come back in a week and show you working.

And the guy, his facial expression said that he clearly did not believe that this was possible. So a week later, we went in. It all worked, and I then did work for them, both hardware design and firmware for many, many years. And then, for example, guys from that company left another company and they wanted some stuff done. So for many, most of what I was doing was just general electronics consultancy based around picks, occasionally other things and then um i went to a

presentation on uh by you may i think you may have it in the someplace in the usr dork bot which is basically a presentation or people doing presentations on interesting things that the tagline is doing strange things with electricity and one of the presentations was from the company that i subsequently you know went on to do a lot of work for where they did um it was a a um like a conference table with lighting above it and rfid readers in the table so as they moved various bits of table furniture around the lighting reacted and it was very clear from that presentation that they were at the very limits of their technical abilities to get this thing working so i just went up to the guy and started chatting to him and said look you know i think i can be i think i can help you guys and yeah that that was it i'll be i'm still doing work for them today that was probably uh at least sort of 12 or 13 years ago um and the nice thing about that

sort of work is that it's a case of design it build it ship it next job you don't have all the really boring stuff like you know the production engineering and all the you know documentation all that sort of stuff i'm very much you know the next new job is always more interesting than anything i'm working on at the moment and the actual design and design and problem solving is the thing i love the most so yeah i don't do any production stuff at all these days i just can't maintain enough interest throughout that process to do it that's a great place to be yes definitely sometimes on the other hand you you talked about the hub and spoke architectures and and testing things and being able to test things modularly, which means that the production has a path, it has a plan.

And when you're doing installed electronics and you've done outdoor installations, it's not a cushy environment. No. I think you did one that was near the ocean. And my response to that was like oh my god salt fog was so horrible um i think i don't remember you might be confusing it with something else but generally not much stuff is outdoors and if it is i just go over the you know because the actual cost of the materials isn't a huge factor generally you know anything that's outdoors

goes into a die-cast aluminium ip68 box, not a plastic one. It's got decent cable glands. It's got decent quality waterproof connectors, et cetera. So, I mean, generally, when I go into a customer, the first thing I do is I terrify them by explaining all the ways in which you could go wrong. And then when it doesn't, they then get a nice surprise. But it's generally, you know, there's no point in cheaping out on things like connectors you know using near pixels whatever on certainly your temporary installations are a bit different but for opponent installation you know the cost of it's not worth saving a few

pennies here on there because that's nothing in the scheme of things but the the cost of it going wrong is much higher like for example you know if this power supply go you know say for example you know don't put power supplies in the ceiling so you need a cherry picker to go up and get them if they if they go wrong um you know the cost of going out somewhere which may be in another country to fix something and because it's a one-off you can't just call your local tech to go fix it because they haven't got clear what it is and that's that's actually a fairly long-term persistent problem so you have to just do everything you can to make sure that A, things don't go wrong. B, if they do

then, for example, as easily you just say okay, just swap this for your spare modules always make plenty of spares so if something goes wrong they can just swap it out and just try and think ahead of all the things that might feasibly go wrong and try and do your best to mitigate that.

Is there anything uh as you continue to do this do you keep on top of new developments and things is there anything you're looking forward to to make your life easier or do nifty new things um no i mean what i would like the most is for let's say somebody like osram and ti to get together and do a lead with a built-in driver that I could trust. An RGBW lead. Yeah, an RGBW, even if it costs like a dollar each, I don't care because that would just save the fact that I can design something in with confidence that it's just going to work. Because a lot of the time, at most, we'll maybe do one prototype,

but it's generally a case of we have one chance to get the production run right, and it's got to be right. So we don't have the time to qualify parts over extended periods of time and do extended testings and burn-ins. There isn't time to do that. You mentioned you went to electronics self-taught. Why didn't you go the uni route i started i did start university i've always been terrible at maths because it the way it was taught that they were the other they'll teach you calculus explain but without explaining you why you know why you might want to know the area under a graph so i i didn't really pay much attention so i did like i started

the computing with electronics type course and dropped out of after the second year because i i'm very lazy when it comes to academic things but b you know i just didn't really so again some of the computing side was a bit too abstract for for some of my liking i've always been a practical hands-on person so if i want to learn something i'll just tear into it and play with it and you know figure it out myself and pretty much everything i know i've learned by yeah building things taking stuff apart um yeah and all the other sort of cool stuff and you'll see this was obviously before the internet was a big thing so i used to have like a massive wall full of data books.

But yeah, I've always been the sort of person that I learn by just doing things. It's impressive to me. I learn from books, and when I do things often, I just get frustrated and lost. So I'm always interested that people learn in different ways. Yeah. frustrated and lost so i'm i'm always interested that people learn in different ways yeah just quickly going back to the pics one other just minor thing on why i use 8-bit pics at all is things like supply voltage you know you can run like anything between say two and five volts on the device so and also they've got like one percent on chip rc oscillator so for example my my led fixture that takes youART data in and spits data out to the drivers,

literally, it's just the pick, a couple resistors, and the decoupling cap. The 5V supply goes straight to the pick. There's no crystals or anything, so it just keeps things nice and simple. Yeah, I think one of our listeners mentioned the power friendliness was a big feature.

And they're fairly robust. All you need is a series resistor to protect it against a short from the data line to 24 volts, for example. Didn't I say in that show electrical engineers like them because they require fewer parts? Yeah. That's always a good thing. You're fine.

I'm just making sure. Well, Mike, thank you for being with us. Do you have any thoughts you'd like to leave us with? Never try to catch a falling soldering iron. That is great advice. I have a couple of burns on my hands from last weekend, but they were from the desoldering. And you can never have too many test points. Oh, yes. Please, all electrical engineers,

listen to that one really well. Our guest has been Mike Harrison of Mike's Electric Stuff YouTube channel. He's also the founder of White Wing Logic, an electronics consultancy in Essex in the United Kingdom.

Thanks, Mike. Thank you. You're welcome. It's been fun. And I would like to remind you that we are having a party for the 300th episode. It will be Saturday, September 7th in the afternoon in Aptos, California. You can mark your calendar now, and we will release the details and RSVP information around August 1st. Thank you to Christopher for producing and co-hosting, and of course, thank you for listening. Thank you to our Patreon supporters for our mics, Mike, and you can always contact us at

show at embedded.fm or hit that contact link on embedded.fm. And now a quote to leave you with. Let's see how far I get on this one. Peter Piper picked a peck of pickled peppers. A peck of pickled peppers Peter Piper picked. If Peter Piper picked a peck of pickled peppers, where's the peck of pickled peppers peter piper picked embedded is an independently produced radio show that focuses on the many aspects of engineering it is a production of logical elegance an embedded software consulting company in california

if there are advertisements in the show, we did not put them there and do not receive money from them. At this time, our sponsors are Logical Elegance and listeners like you.