Startups that manufacture electronics are faced with a limited set of choices when it comes to getting their hardware assembled. Each of the methods below have an array of possible approaches, some good some less so. I will try to cover what I’ve seen in my quest, hoping this will be a good overview of what’s out there. If I miss details or interesting methods please let me know and I will keep this list updated. This is part one, where I look at options of assembly without getting into a real commercial pick and place.
Doing it manually. Everybody started here, including Adafruit, Sparkfun and others. It’s great and fun at first and you learn a lot about surface mount technology (SMT) and also gain respect for how hard quality assembly really is. And we’ve seen countless ways people do this around the world. Some spread paste with a toothpick or dispense with a DIY syringe, then place parts with tweezers. This works for about 5 minutes before you go crazy, especially if you go smaller than 0603. Others laser cut mylar stencils to speed up the paste deposition. I etch my own stencils from soda can aluminum, this is fast and works great. For placement I’ve seen people make some fancy ball bearing jigs to help them move their hand, a bit better.
This is very cool and I almost did this myself but I think it’s overkill and a bit slow and … more expensive than it needs to be for what it does, but I can appreciate the effort of its creator. One advantage of this fixture is the ability to better hover over larger PCBs while placing. FWIW the maker even went as far as mounting a camera to help placing the parts, a nice touch, but looking at his video of it in progress the angle of the camera looks weird and looking away on a monitor when you work with your hands is what cuts the speed. Here’s another one of the same kind for reference. I do the same as these, only probably faster, with my aquarium pump pick and place (hold your laugh for a moment there) which cost me about $20 in parts, took about 30 minutes to make, and has full stereo vision – my eyes, can’t beat that for the price. I like to brag about this method not because I’m using it, but because it’s probably the best 100% manual method I’ve seen. Why you ask? I can do up to 7-800 CPH (measured!) with it on less complex boards. The only downside is obviously having to sit there all the time, with some neck and back strain, and the part alignment is less than ideal, but paste reflow really helps. I’ve assembled many thousands of PCBs this way, it works nicely and gets stuff done, but it’s not great. I wish I spent all that time developing new projects. So what is better than this?
Outsourcing assembly – Well the next alternative is to have someone else do it for you. You can relax and spend more time developing instead of doing repetitive monkey work. There are now batch services that do this for you, or if you prefer there are many professional places which will cost more but you can expect more peace of mind. Here’s an example, used by Steve at bigmessowires.com to make his Floppy EMUs. If you choose such a service it’s probably smart to look for one around you so you can visit and talk to the owner before you start throwing money at it. I found at least 2 professional assembly houses close to my location from a simple search and asking around. The plus is you pay and get them done. The biggest downside is the cost. If you make a million of something it makes a lot of sense, more so than having an assembly line yourself, but if you’re low volume like me then the cost can be prohibitive in the long run. Whenever I looked into this the cost turned me away, but my designs are simple, complex designs are hard to make by hand and that’s when you might not have a choice but outsource. I think most kickstarters are forced to outsource even with simple designs because there’s too much going on and too little time to make it happen. Ok, what else?
Home made pick and place machines. I’ve seen some admirable efforts from various people who’ve done this. Some start as promising projects but don’t really work, like this one. Here’s one that actually works, looking pretty cool for a DIY garage project. Here’s another one from Daniel Amesberger, it picks from strips of tape which is the drawback (he talks about making/buying feeders), but otherwise probably one of the best I’ve seen.
Another one I spotted lately is this, youtube here, and details here. It’s quite good, has upward vision for alignment, not sure about head vision. Tape advance is done by the head, which is pretty acceptable for DIY, and tape covers are pulled bacl by weigths, not ideal but again acceptable, it makes it almost automatic, too bad the documentation is lacking.
And finally for this section, the best ever DIY pick and place I’ve seen is this one:
It uses real commercial feeders adapted for this build, has top and bottom vision for alignment and hence can be fully automatic. The quality is pretty amazing for a home made machine, more so than some of the commercial pick and place machines I will describe later. Brian even made a vibe feeder that works much like a real one. It’s well documented and I only wish it was a bit faster, looks like everything is done sequentially rather than doing non essential parts while the head is moving. Looking at the quality he put into this robot only makes me wish I had a fraction of his skill:
Commercial pick and place machine kits. I think I’ve seen at least 2 or 3 of these. The only one I think is worth mentioning here is the red frog.
You can buy the kit for $3689 at the moment of writing. I think it’s an admirable effort by its creators. When I first saw this I got excited, I thought maybe this is is the affordable answer to my pick and place problem. However the more time I spent researching it it was obvious it would be a poor choice to burn almost $4K, as much as I wanted to believe in it. There are many reasons, but mainly the kit assembly is less than smooth, many ways to mess it up, forums reveal lots of issues there, potential missing parts and required creativity to get it assembled. This is not your typical IKEA furniture assembly session, it will be long before you go from kit to functional machine. Then it’s made of wood, which is very dimensionally unstable depending on temperature and humidity, big bummer there. I mean this is supposed to be a super precision instrument, the wood dooms it from the get go. Then there’s the features you’d want in a pick and place, or lack thereof. You can watch it work here. There’s some vision but no alignment, and it seems hopelessly slow. I can do at least 2 times faster with my manual pick and place. The UI looks daunting. Hanging weights pulling back the cover tape means it’s not fully automatic, you have to be nearby to raise the weights when they reach the floor. But really nicely engineered machine.
I wanted to mention these DIY-ish methods, so they are here before I move on to commercial options. Unfortunately I see a gap between very manual methods (like my aquarium pump “pick and place”) and an actual automatic machine that doesn’t need spoon feeding. Most of these home made machines (except Brian’s and Daniel’s machines) are pretty hopelessly slow, lack alignment, or have vision alignment which slows them down to about half the speed. But to be fair, I can’t really expect a DIY machine to have advanced features like laser alignment, but there are other methods for on the fly alignment of components. I will touch on this subject later when I talk about what important features make a great pick and place. So if you’re not a mechanical engineer or don’t have the skill to build a machine from scratch, you’re left with buying one. And frankly that’s what ends up happening when you realize that kits and DIY machines will end up just draining your time and won’t work as you dreamed. Remember, I’m only talking about entry level machines here, not high end ones. The sort of thing that Adafruit got into when they got their first PNP.
One last thing to mention is that in general machines that use extruded T-slot type assemblies are not regarded as high precision high-reliability and repeatability machines. The reason is explained in this video, so I’m glad I didn’t say this, although it’s pretty obvious why this is true. So I will opt out of looking into any machine using t-slot segments frames.
In my next post about this subject I will talk about a few entry level commercial options, and try to emphasize the features that you probably want if you’re looking for automation and not baby sitting a machine. Sorry if this series doesn’t come with a lot of visual glamour, actually I wish I had all this info aggregated when I started reading on pick and places. So hopefully this is somewhat informational for those who want to learn more about these expensive and mysterious machines, and I don’t know a whole lot yet. Who knows, maybe in a few years when the maker movement has its own section at Home Depot you just get all your assembly needs there, including a desktop pick and place, just like buying a snow blower.