 Ieb o edrychi eich dod i'w gweld i hynny'n을eg yn ystod. Mae'r ffordd yn ymgyrch yn symud ogymru eich ysgrif iawn amgyrchwil wedi'i ar-raddiaeth yn ei wneud bod y roedd gan eich Dodrychi Eich Rhannau ar-raddiaeth. Rydych chi'n gweld i'r rhan cyfle iawn, gyda ei fod yn gymryd. Fyrdd y teimlo fydde i ysgrif iawn gyda'r broseni yn ddeiferach mewn, neu i'n gweithio'r cwmpunio ddiferiannau, os yn mynd i wneud i'r môl i'r olwch, a'n dechrau i'w bobl i'w ddisgwylio ar y lleol sydd wedi'i adresio'r gwybodaeth eraill o drafynnol, ac yn Talkiton i gyfaniaol, trwy'r peresio y cyfanyddiaeth wedi mynd yn cyn dim ei ffnwysgwr ar gyfer y cyflwyn lleol, byw'r reif innovative a'r cyflwyn lleol wedi bod yn gyflwyno cymddiadau. Dyna'r gallu ei gwasanaeth yn ôl wedi'i weithio, ac y cyfrif y grondiadel ei wedi ddim yn siaraddill o gyflwyn lleol ar y cyflwyn lleol, mae ydych yn gweithio'r bwysig o'r ddeinig iawn i'r tyf arall a chi'w gwahanol. Y dyfodol gweithio ar y cyfrifiad cadwydd yn gweithio, dyfodol gweithio ar y dyfodol gweithio, ond mae eu gwahau gweithio'r bwysig iawn i'r ddweud sy'n gweithio'r bwysig, mae'n cael ei wneud yn iawn mewn gwirioneddau i'r gyffredig, ac mae'r cwmhwysig iawn i'r gyffredig, mae'n cyffredig i'r gwirionedd a'r bwysig iawn i'r gweithio. oherwydd, mae'r cyfrifnegol yw llawer yn oed i gynlluniaeth er mwyn Tunbydd. Mae'n ydych chi ffaint rydyn, fel yna ddefnyddio'r pethau, ac eyrna'ch gobeidio. Mae gennym, dyna'r newydd Cb a rydyn nhw yna'r newydd dezgo math i ni. Felly, rwy'n golygu i gael yr ysgu oedd yn amser yn fwyaf gyda'r grannu a'i dda chi. Rydyn ni'r grannu i gael yr ysgu? Mae'r ddweud i'r ddweud. Do you know what's about to come? Do you know how long this is going to take you? It could take you three or four weeks to actually get something back from the manufacturers? Is your board designed for a manufacturer? Did you way back at the start and talk to the manufacturer of your product and go, what process have you got? How do I need to design this board so you can make it and make it reliably? A lot of people don't think about that at the beginning discipline. Then go round theharad, to design for manufacturer. You will need to purchase all your components in advance and do what comes from a free issue, which is that you give your components to the manufacturer and hope that they don't break. If at this point you want to really move on with your project all very good and you have everything together and do 95% of the ideas that started out on this process never make it this ac efallai yn y rhaid i'r gweithio, oes 50% yn cymryd ar y cyflawn. Felly mae'n gwneud o'r rhan o'r rhan o'r rhan o'r rhan o'r brosedau. Ac mae'n mynd i gael ei wneud yn gweithio, ac mae'n mynd i'n gwneud yn y gweithio. Felly, rydyn ni'n gofio ar y ddweud, ond rydyn ni'n gofio? Rydyn ni'n gofio ar y ddweud. Rydyn ni'n gofio ar y ddweud. You've got some pre-made PCBs, which your PCB house has finally made for you and tested and they think it'll work. You've got all your components which you've gone out and bought and spent vast sums of money on, and you've got all of your design data which is which components go where, which way around, how often, and so on. And you're going to take this to your manufacturing house and they're going to, well, charge you even more money before they'll start. First up, they need to take your board and make the components stick to it. To do this, we use solder. Most of you have probably seen solder, it comes in reels, get it hot with the soldering iron, it melts, stick components together. Soldering is joining one metal device to another, in this case, components to the board. To make this easier for the manufacturing process, we use solder paste, which is a little sort of pasty goo, which can be worked with in a very narrow temperature band around 25 degrees Celsius. I'm afraid I've not converted that to Fahrenheit for you, LeMont. The paste is made from an alloy of tin, silver, copper, and other trace metals. It used to be made of tin and lead, but we're not allowed to do that anymore. And these are all held together in an organic binding matrix, lovely term for glue. The paste looks a bit like that. Comes in tubs, that little tub there, about yay high, weighs half a kilo, and yay. Costs about $100 as well. It's this sort of goopy paste, a little bit like the consistency of toothpaste, but for God's sake, don't brush with it. 30 micron balls of toxic metal is not a good thing to be rubbing into your mouth. Paste then has to be got onto a board. And to do that, we use something called a solder mask. Somewhere we have a solder mask. And basically it's exactly the same process as how your t-shirts have been made. There's a screen which has holes in it, which paste goes through onto a board. Again, another friendly company will make that for you. They're lovely people, and they'll charge you between $400 and $500 for one of those. Every time you need one. If you change your board, you're going to need more than one. The mask gets placed into a large metal frame, tensioned, and put into a machine ready for use. Which means, of course, that we need a machine to paste the board with. You can do it by hand, but actually sitting there with a toothpick, putting solder paste onto surface mount pads is really tedious. So this is a picture of our pasting machine on our prototyping line. It will take the board in on its swanky little moving table, lift it up to the solder paste mask, tension it, squeegee the solder paste all over it, and then bring it back out very carefully for you. This whole process takes about 45 seconds. So there's an instant limitation on the speed with which you can manufacture. And doing it by hand is much, much, much slower. So we've now got some boards, and they're sticky with solder paste, and ready to go. So in order to do this, we need lots of components, and lots of components generally come on something called reels. Reels look like that, and they need feeders to feed them to the machines, and feeders look like that. Or in a little rack, they can look a bit like that. Each feeder will cost between $250 and $300, and it can cost up to $600 for a feeder which is qualified for componentry, which is 0402, which is the size two below the components that you have stuck to your paper. Those are the ones we use. Each pick-and-place machine has between 50 and 75 feeders on average, but we'll see. Some components obviously can't come on reels. They come on trays, which look a bit like that. And trays, generally you'll get CPUs, memory chips, flash, that kind of thing will come in trays. There's some strong-arm CPUs. You must love these. No. Unfortunately, with minimal water quantities, the number of CPUs that you buy tends to be a little bigger than you want. For $14 CPU, $15 CPU, might come in a minimum order of $200, meaning that you've got to spend $3,000 on CPUs, even if you want to make one board. Then some components come in tubes. We have a tube. I'm sure of it. There we go. Because they're awkward. And tubes need a special kind of feeder called a vibrator. And vibrators, unfortunately, cannot be used afterwards. Well, they can in the machines. Then each pick-and-place line will have several machines. Our line has three. And each machine has a number of pickup heads. These pick up the components and put them down on the board. A multi-threaded machine will pick up multiple components at once in what we call a cord. And this reduces the amount of movement that the head has to make. Some machines are obviously used for picking up the bigger components, which might have lots of balls on them, if they're what we call a ball grid array. And so we have a system which is called an upward visioning system. Because unsurprisingly enough, it's a camera that points up and looks at things. And that counts the balls, measures the component exactly, works out any variation in how the component was picked up, and allows you to precision-place a component on the board. The ball grid array that you can see on your sheets, 17x17 grid at 0.8mm between the balls, that's centre to centre, needs to be placed with a 0.1mm accuracy. And that's a big balled BGA. We have some BGAs that have to be placed with an accuracy of 10 times that. And therefore you need expensive visioning. The cheaper components, the smaller things, can just be looked at with a camera, or lasered, more lasers. Before they're picked up by these little pickup heads, and moved over, and put down on the board. The heads use a Bernouy vacuum effect to lift the components out of the feeders, carry them over, and drop them down onto the board. And this is known as the blow-off, yeah? Yeah, it's called puff-off, yeah? Okay, the puff-off. Obviously, doing one of these is easy. Doing lots and lots and lots of them is less easy. And you have a system called a pick-and-place programme. It's a really grand term for move that one there, then move that one there, then move that one there. And if you're really lucky, pick up 10 of those and put them in a line. Pick-and-place programmes are very specific to the line that they run on, the components that you're using, the boards, probably the time of day, the phase of the moon and stuff. And manufacturers will either write the programme for you, and charge you lots of money for it, or tell you what to do so that you can write the programme for them and charge you lots of money for it. The pick-and-place programme can be replicated, so if you have a biscuit, you write one pick-and-place programme for that, and you say there are three in a row. And if you're lucky, your machine will do it. Here you can see one of our pick-and-place machines picking up components from the feeders, which have their reels, and putting it down on a board just here. There are eight boards in that biscuit, and it's making a USB device of some kind, I believe. Now you've got a board with all of its surface mount components put on it. Well, we think they're all on it, so we need to check it. And that needs an optical inspection station. In this instance, what we're doing is we are looking for missing components, components which the machine might have fumbled and dropped on the floor or something. We are looking for the wrong component in the right place, the right component in the wrong place, the right component in the right place, the wrong way round, and various other games which the machines like to play. One of their particular favourites is tiddlywinks. They like to take components, put them down on the board, and then ping them somewhere else. This is because components are as small as a quarter of a millimetre tall. So they've not got much mass. You blow them down into the solder paste, and if the solder paste doesn't quite grab them, they might ping off somewhere else on your board. This is all great until one of them pings under your £20, that's $40 BGA, meaning that it won't solder to your board properly. And you can't see it because it's underneath the BGA, and the BGA is down, and it looks like it might be in the right place, so you've got to try. The machinery to do this inspection is a sophisticated non-contact scanner capable of scanning an entire board and then a lot of software that will do a comparison between a board that you know to be good and the board that you've made. Unfortunately, it needs an extremely powerful machine to do this in any reasonable amount of time, and it's not very good at it. This really is an instance where a human being is quicker and better than a machine. So we have a Gary. Gary is our process engineer, and there's Gary looking at some boards, same kind of boards that the pick-and-place machine was just placing. As you can see, he has a big magnifying glass for a head, and his arm is made out of trays. We then need to cook the components to the board. What this is doing is melting the solder and therefore joining the components to the board. We have a machine that does this. It looks a bit like a pizza oven, but if you put pizza in it, then our hardware engineer will come around and put your head in it. It has several zones inside it maintained using what is called forced air reflow. Each zone must have its temperature very carefully set so that as the board passes through the oven at a particular pace, it will go through a particular temperature curve and therefore melt the solder without blowing components up. The air flow has to be exactly right so that it doesn't blow the components off the board. If you've got your air turned up too high, your BGA might rotate by a fraction of a fraction of a degree and connect all the balls up wrongly. And then it will separate the binding agent from the solder and in the case of the solder the paste that we use, that then basically disintegrates, turns into harmless molecules, gets passed through an activated carbon filter and goes somewhere else. So we have to do this, correctly soldering all the components to the board without losing or damaging any of them. To do this, we have a machine which looks a bit like this and as you can see it has a computer to control it. Here you can see the temperature zones. I think there aren't any boards in it at the moment. The most important thing on any of these machines is the emergency stop button for when you've got your hand in it. When the board comes out of the end of the oven, it's still very hot so you have to let it cool down before you look at it again. So that's another optical inspection station. Then you've got to put all of the components that couldn't be surface mount placed onto your board. There are many different kinds of through-hole components. Typically they are ones which have leads on them so LEDs, dips, that kind of thing. And we are going to overrun, by the way. So if you have to run away, please remember to put your rubbish at the front but it's lunchtime next to you, you don't have to run. And there are connectors, so VGA connectors, parallel connectors, that kind of thing. And on our case study, we've got great big connectors because we've got parallel ports and all sorts of things. So usually use a Gary to do this or a through-hole placement robot but they're really expensive and crap. But just to confuse you, that's a Gavin doing it instead. It's very important that all the components are in the right places the right way round. The legs have to be trimmed because they're about to go through a machine which, if they're not, the board will foul on it and cause all sorts of problems. So to do that, we need some solder. And we have a nice machine called a flow solverer which has within it a standing wave of molten metal. Mmm, yummy. Boards are placed into a tensioned carrier. They get waveguides and cleaning spans and so on put on them and then they get put into a machine. The machine looks a bit like that. This is a baby machine. The carrier gets introduced to the machine at this end and comes out at that end. So first up, we need to make sure the solder will stick to the board. To do that, we use a substance called flux which is an organic solvent. It gets sprayed onto the board. So don't breathe it or your lungs might disintegrate. And what happens is the board gets passed over a fluxer which goes ahead and squirts flux all over the board. The boards then pass into a preheating stage where they're raised to about 100 degrees Celsius and then passed over a standing wave of molten metal. Molten metal, standing wave. There's about 120 kilograms of solder in our soldering machine. The solder is held at 260 degrees Celsius which is 491 Fahrenheit. I converted that one for you, LeMont. And it costs $2,000 to fill that machine with solder. It costs $15 every time you turn the machine on and about $5 an hour to run it. It takes that much energy to keep that much metal melted. The solder has to be checked for purity regularly and interesting chemicals that you really don't want to deal with are added to it at various times to make sure that the bath stays exactly right. Since the boards have to be fluxed, heated and soldered, there is an absolute maximum rate of about one frame of boards a minute and this is a completely human intensive task. You have to monitor it because if you don't bad shit happens which we'll see later. This means that you have another friendly process engineer looking after your flow soldering machine or you have Gary. Again, once that's done you've got a visual inspection to do. You need to make sure that there are no solder bridges. That is solder that has joined pins together that it shouldn't have. You need to clip any of the remaining long legs so that they don't stab people and you need to hand solder any of the components that you couldn't flow solder because while flow soldering is great, it is not a be all and end all, large connectors have too much thermal mass to be flow soldered. So you have to hand solder them because quite frankly you cannot get them up to temperature otherwise. This is all done by your process engineer or Gary. So very well and good saying all that. How much does that lot cost us? Well, economics one on one. We have a solder paste machine imported out at the beginning. He looks a bit like that. Costs you $30,000. Then you've got to pick and place line. Well, we've got three of those machines. Each of them is $200,000 each. Then you need to add these feeders at $300 a time and they're $50, $60 on a machine. It's $200 for the line usually and you need to double them up because you need spares in case when those run out you need to stop the machine and put a spare on and if you have to reel up in whilst the machine is stopped you're losing money on the production line. So that lot comes to $750,000. Then you've got the optical inspection stations. We've already said they're not terribly good. We do have some and they're about, we have a Gary actually, but they're about $20,000 each and you need a couple of them usually. Reflow oven. Ours is quite a small one. That was the one we saw earlier. He only costs about $30,000. Flow soldering machine. We saw really? Pretty metal. That's another $40,000 and of course a couple of thousand dollars to actually put the solar in the thing. Support machinery because all of this lot needs compressors, vacuum systems, air scrubbers because you're producing toxic gases and all the other building to put it in people, that sort of thing, set you back another $130,000. Your friendly manufacturing engineering is priceless and for everything else there's Gary. I'll try and pick this up a little bit. Our manufacturing case study, we have about 800 components on the board. Yes, I've run out of time. This means that this is extra bonus talk that you're getting for free, which is something you don't get with hardware manufacture. We have about 30 through-hole components including very large connectors and a manufacturing house in terms of prototyping runs will typically charge you about 10 cents to place those little small components but they'll give them to you because they only cost a hundredth of a cent each. They'll cost charge you about 20 cents to place the medium devices such as, I don't know, the RAM chips and probably up to 40 cents or more for the large devices such as your processor and this is quite simply because they take longer to place and you have to have a more expensive machinery to do it. They have to make their million dollars back. All of these prices are obviously prototyping runs as you move into production size runs you can probably knock about 25% off that but that's not a lot. You have to choose your manufacturing process. This is all about whether or not the lines run hot or cold. As we talked about earlier the datasheets give you all sorts of physical information about the components. What the component must be in order to solder properly. Another one they give you is how hot the component will be before it explodes. Often these are inverted because datasheets are really good. So the concept of a cold process is one which runs the bottom of that curve. We hug the bottom of the curve we run a cold process which means that we would rather fail to solder than explode components because when components blow up they might send shattered bits into your oven and then you get the smell of burning chips for ages and I don't know if you've ever done that but your kitchen smells bad for ages. So once you're done you have nice shiny board with all sorts of components on and connectors and power input and video output and hard drives and all sorts of things and it's really good. However the board that comes back has no guarantees attached to it. Basically any faults that the manufacturing process has introduced are your fault. They're your problem and you have to pay for them anyway. If you damage part of their production line because they were stupid it's your fault. So I hope you've got a lot of money or really good insurance. Unsalvageable hardware that is broken beyond repair is yours to absorb the cost of and you bought the chips you bought the boards and you still have to pay for them to be placed. Here's an example of an unsalvageable board. This board took a bath in the solder. So it has solder all over it and you really don't want this to happen because one it ruins your board and two it has a tendency to pull components off the board and into the solder. Which means they have to stop their soldering line clean the machine and you get charged for that as well. You have to test each board fully. I'm not going to over stress this too much but it is very good practice to test the board when it comes back by putting software onto it. It is possible to buy pre-programmed parts that is flash chips with your software already and don't get to go through the full test process. Often if you manufacture like that you don't notice until an entire batch goes out with the chip soldered the wrong way round and your customer comes back and complains a lot. Here is one of our boards under test. Unfortunately we haven't got the audio test plugged in but you can see serial console, USB, there's a whole bunch of serial ports there some ethernet, some power. That power cable is a bit complicated because it's got some testing equipment built into it video output and this little ribbon cable here is JTAG which we mentioned earlier. This board will actually be brought up from raw straight off the manufacturing line all the way through to full functional test. I'm told I'm still at zero. I know I'm at zero. Why are you running out of tape? You didn't make negative signs you just weren't thinking ahead. So test every board it's not good enough to test one board within a batch because variance within a batch exists. We have all sorts of things that I could say about that but I'm being told to skip. So you take your boards you package your product and hopefully you can sell to the unsuspecting public and profit. Unfortunately up until now you haven't been able to sell anything. You've dispersed around about $40,000 in terms of PCBs, components paying to have them placed at your own time and in your hands you have up to 10 prototype boards which you can't sell each of which is worth about $4,000. It's worth nothing but it cost you $4,000. You also have assuming you've gone to production a minimum of 200 production boards because anything smaller than that they're not going to do in a production run for you and each of those will have cost you about $130 for our case study and there have been about $5,000 of non-refundable engineering costs at the beginning of that. So can you sell enough of them can you sell them soon enough to make enough money that you get to eat next month? Good luck. I'll try and be quick here. There are some sensible ways to cut the costs in all of this. You can use module hardware such as one of our modules you can also use other modules like the balloon and there are lots of modules on the market lots out there if you look hard. Popular components but not too popular. We had a problem last year where the Nand chips in one of our products we couldn't get hold of them because Apple bought a lot of those out and Samsung who sell them their flash bought every single piece of Nand on the market including Infinion's, Heinex's and everybody else's and then rebadged them as their own. So not too popular. If you designed for a manufacturer from the start as we suggested you don't have to go through the whole prototype phase again and again so it's cheaper. Use the right people in the right ways. What we're saying here is that if you've got a Gary, if you've got a clever hardware engineer who knows his job you can reduce the costs quite rapidly and you don't have as much wastage. It's a good thing. I have more to say but we have it at a time. Open hardware designs are easy, useful and good ideas. We have produced because we think they're good ideas we've produced our first open hardware design you can come and have a look at the schematic in a bit. It's the same module pen out as the modules we've all got at the moment which means obviously a modular design is a lot, so we're suggesting is a lot here, easy to integrate and the connector for that is a couple of dollars so they're really easy to put down on board and play with and do things with what we would like. I get to say things again. Very quickly this is a physical process. It has a physical impact. The environment is fairly boring but it's kind of really important. I am 10 from red cheese. I'm getting nought points now. We produce quite a lot of waving I can't land. Lots of solder waste and lots of metal waste from all of these components that you have to clip legs off. Solder waste looks a bit like that, that's Gary's little pot. Solder waste goes back to our solder suppliers and about 50% of what we give them as waste comes back to us as solder. About 80% of what they have left goes out to other manufacturing people as different things like oxides different substances which they can then use in different chemical processes so it's actually really well recycled. Scrap wire, we keep in a nice cream tub because it tastes nice can be recycled like any other metallic waste like you might recycle your tins from your food and there's lots of board wastage or salvage. We can give them to unsuspecting people at talks but they have to give them back and generally they get built up and returned to the PCB houses I believe for them to recycle and I have no idea what they do with them, maybe they build houses or make robots or something. That is one month of salvage with my shoe next to it so you can see how big the pile is and then we get trays reels tapes all sorts of things which are rubbish. Trays are basically graphite and you can use them in the winter. Reels, plastic everything else, plastic paper simple to recycle. You have to build smarter not harder obviously if you make your design carefully you're going to have less wastage and that's very important because if your PCB is hard to make half the PCBs you get back won't work and that's all waste so if it's hard to manufacture half the stuff you get back might not work and that's all waste so be good at it and you make less waste then there's this fantastic thing called ROHS or Rosh which basically says no you may not use nasty chemicals you must use these nasty chemicals instead because they are worse or something it essentially means to us that we can't use lead in the majority of what we do there are exemptions to this such as medical military and so on where you can because they don't want to pay the extra but lead free processes are more expensive but at least you can sell them in more places so you may as well give up and go home because as Douglas Adams said it makes you wonder why we came down from the trees in the first place so dependent on whether or not you want to stay around we'll take questions the answers are 42 if they're not we can't guarantee to have the answers if we have them they might not be correct and whatever way they'll be very very expensive thank you