SXSW: Imagine a Practical, Low-Cost Circuit Board Assembly System (Video)

Video no longer available.

SXSW Create is one of a handful of sub-shows at SXSW which don't require an expensive badge — it's maker-oriented and small, and a few blocks from the slicker parts of the convention. (The local ATX Hackerspace was there showing off robots and giving out soldering lessons and blinkies, without a single corporate pitch.) Under the same tent, I met with Jeff McAlvay, co-creator of Board Forge, which Jeff hopes will make small-run circuit board creation as easy and accessible as small-scale 3-D printing has become in the last few years. ("Think MakerBot for electronics.") The prototype hardware McAlvay had on hand looks -- in fact, is a 3-D printer, albeit one lower-slung than the ones that make plastic doo-dads. That's because the Board Forge's specialized task of assembling circuit boards requires only limited vertical movement. It's using the open-source OpenCV computer vision software and a tiny camera mounted on a movable head to accomplish the specialized task of selecting and placing components onto the boards. The tiny electronic components are lined up in strips on one side of the device, where that smart head can grab them for placement. The brains of the operation include an Arduino-family processor for basic controls, and a Raspberry Pi for the higher-level functions like computer vision. The projected cost for one of these machines — about $2000 — should put instant-gratification machine-aided circuit creation in reach of schools and serious hobbyists, but there's plenty of work before it's set for sale to the public; look for a Kickstarter project in the next few months.

Jeff: My name is Jeff McAlvay. I am working on a project, Board Forge. We aim at being electronics from art to part. So we want to be in a situation where folks can load a design of a board, a blank board and components, and come out with a functional circuit board.

Tim: Okay. Now on the table here, we’ve got a couple of different things, you got what looks like a CNC machine, and you’ve got an electronic brain, and you've got a laptop sending instructions. Can you talk through what each one does? Can you tell us what each of these, the things on this table, do?

Jeff: Certainly. Yes. So this is the machine itself. So it has components that are loaded in. It has a head that is able to move up and down and pick up the components. And it has the boards that are going to be populated with parts. The head has a vacuum tube and a camera on it. Over here, is the brain for the machine. So it is Raspberry Pi doing three things, it is sending instructions to the motor controller, it’s also doing the computer vision, and it is also serving as the interface for the software.

Tim: The software you are using, is this homegrown?

Jeff: Yeah. So we have printouts written in full stack Javascript. So the server is an OJS and the software it is serving is HTML5 and JavaScript.

Tim: Okay. Now what sort of boards can people make with this?

Jeff: So our aim is to use to write different boards, anything from IO equipment that would allow folks to be able to have an accelerometer on their head and be able to move their head and hear from where the sound source is, to also do motor controllers like the one that is built and it’s out of here.

Tim: Now what will this cost?

Jeff: Come again.

Tim: What will this cost?

Jeff: We are aiming at the $2000 price point.

Tim: Okay. Aiming for means when? What’s your timeline right now?

Jeff: We are hoping to kickstart it this summer.

Tim: Okay. And that kickstart is going to raise money to do what, what things do you need that money to accomplish?

Jeff: So the main purpose is to be able to build machines. There are costs associated with making physical things.

Tim: Okay, sure. Now the device itself over here, how did you construct this?

Jeff: Instrumental in the process has been a hackerspace in Chicago called Pumping Station One. Right now I am working with three folks from that space. As well as some folks who are writing software that I met at a hackathon. So those guys have been just crucial in being able to help get to where we are.

Tim: So how big a team is it right now in total?

Jeff: Right now, there are six folks. Three primarily doing the hardware, and then three working on the software.

Tim: And are you all in the Chicago area?

Jeff: Three Chicago, one Miami, and then two California.

Tim: Okay. How do you coordinate that?

Jeff: So far, GitHub has been a big friend of ours, as has been the phone.

Tim: Okay. Now the things this can do, compared to say, MakerBot can do a lot of fine movement, but it can’t for instance place components?

Jeff: Right. Can you say that again?

Tim: Yeah. So MakerBot can do a lot of movement, and it can make things...

Jeff: Right.

Tim: But it can’t pick up the electronic parts, and it doesn’t have a vision system.

Jeff: Right. Yeah. I think in many ways that is one of the big differences between this and more of the traditional 3D printers, while we're concerned with being able to put down material and build structural things. In many ways, this is more concerned with building electronics.

Tim: Okay. Can you characterize the precision with which the board is intended to place components?

Jeff: So, I don’t want to make any promises yet as to where we will be, but the target we are aiming for is 0402 which is 0.5 mm x 1 mm.

Tim: Okay. And this device you said is going to cost probably about $2000.

Jeff: That’s what we are targeting.

Tim: Okay. That makes it in the range of makerspace and schools?

Jeff: Yeah, we would love to have makerspaces, schools, small university labs, small businesses that do electronic design, and even I think down the road hope to be in industrial spaces where they are doing development of products.

Tim: Okay. How long has this been in the works from your side?

Jeff: We started in September.

Tim: Okay. And if somebody buys one of these when they are available, what kind of software will they need to actually make a circuit?

Jeff: So it comes with a software that we’ve written.

Tim: Okay. So is it GUI-based?

Jeff: Yeah. It is. I think one of our targets is try and make it really accessible for folks to be able to use. I think with the industrial machines, it takes a decent amount of training to get really good facility with them, so we are trying to keep the interface simple.

Tim: Is the software itself open source?

Jeff: Our plan is that once the machine is done, both the hardware and the software will be open source.

Tim: When you say the hardware will be open source, what do you mean by that?

Jeff: The designs for the machine.

Tim: And is everything here makeable with parts from your average makerspace, that if you could, if you’ve got your rail there, and things like that?

Jeff: We haven’t decided exactly the implementation of the final design but we aim at one that is going to be makeable by folks in makerspaces.

SXSW: Imagine...

[Anonymous Coward]

Imagine a News for Nerds site that doesn't have a new SXSW story every other hour....

Re:

[gl4ss]

sxsw has ballooned ridiculously.

however this looks cool.

I'm a bit skeptical about the projected cost of the machine though. and we really need some beefier control electronics than avr's(nobody seems to be producing them en masse now other than with arduinos/avr's) - if you're wondering "why".. well, it has shit for program space and crappier than shit for memory! it's really holding back fw for home 3d printers at the moment.

(raspi isn't practical right now.. though I guess it's not totally out of question

Re:SXSW: Imagine...

[fuzzyfuzzyfungus]

Unless you have some mysterious reason for insisting that the control electronics duplicate, rather than supplement, the ridiculously powerful, RAM-heavy, and massively-mass-storaged computer that you can buy for $200 and use for all kinds of neat stuff, is there a problem with AVRs?

If you are doing a circuit design(or even just downloading one from somebody who did) you presumably own a computer massively more powerful than any microcontroller or embedded system(not counting 'embedded' systems that are server gear with extended temperature ratings put in the same box as the device being controlled) ever made. That PC won't have many PWM outputs, and any DACs and ADCs it has will probably be horribly tweaked in favor of pleasing sound, since they'll be on the sound card; but it will otherwise have ridiculous power to spare.

Microcontrollers make excellent complements, since they have pitiful computational and RAM specs; but tend to be well supplied with PWMs and ADCs. Why reinvent the PC as part of the machine?

Re:SXSW: Imagine...

[gl4ss]

Unless you have some mysterious reason for insisting that the control electronics duplicate, rather than supplement, the ridiculously powerful, RAM-heavy, and massively-mass-storaged computer that you can buy for $200 and use for all kinds of neat stuff, is there a problem with AVRs?

If you are doing a circuit design(or even just downloading one from somebody who did) you presumably own a computer massively more powerful than any microcontroller or embedded system(not counting 'embedded' systems that are server gear with extended temperature ratings put in the same box as the device being controlled) ever made. That PC won't have many PWM outputs, and any DACs and ADCs it has will probably be horribly tweaked in favor of pleasing sound, since they'll be on the sound card; but it will otherwise have ridiculous power to spare.

Microcontrollers make excellent complements, since they have pitiful computational and RAM specs; but tend to be well supplied with PWMs and ADCs. Why reinvent the PC as part of the machine?

realtime reasons? controlling more devices in sync? minimizing delays? I mean those are among the reasons usually cited as reasons when asking why not just have all the devices hooked up to pc separately.

with cnc machines it's common that they're just hooked up to parallel though(though usb is coming more common in home cnc as parallel is going exint). but with repraps, makerbots it's generally preferable to print from sd card on the machine as while feeding the movement codes over usb works ok 99% of the time(that's how I usually do with because of being a lazy ass).. it's that 1%, that one extra pause due to it being usb, that can cause a blip on your print.

on cnc routers that's not a problem though. it's not like the routing bit is going to leak.

so barring all that yeah, I would prefer just a solution where I could just run endless amount of steppers in sync from my pc, of course. but in the meantime I'd rather have lots of more cpu time (and by extension pins) available for io on the avr board that I need to use to run them now.

point being you can buy servos now with optical encoders and arm chips embedded, which makes avr+steppers look quite stone age.(those servos with continous rotation cost a hundred bucks a pop though).

Re:

[drinkypoo]

You can have AVR (or whatever other similarly-priced microcontroller) and SD, there's no reason why not. If you dump the code to the device and it stores it before attempting to run it then it can keep running even if you are disconnected.

Re:

[gl4ss]

You can have AVR (or whatever other similarly-priced microcontroller) and SD, there's no reason why not. If you dump the code to the device and it stores it before attempting to run it then it can keep running even if you are disconnected.

uploading it on the boards I have takes ages if you do it through the usb serial to the board. besides, there's no free space in the firmware for code to handle the save to sd from the usb serial. basically the fw development is at a dead end because of that. so... whenever I want to print from SD I have to physically move the sd card to the computer, dump the file on the card, move the card back to the machine and start the print(I can start the print from the computer though, or from the control panel on

Re:

[timothy]

"sxsw has ballooned ridiculously.

however this looks cool."

When I was a student at UT, it was all film and music, and (since it's at the university's spring break anyhow) always meant a time for me to get out of Dodge/Austin to avoid crowds / parking problems / traffic jams, etc. It really has ballooned, and the Multimedia (bah! bad word!) part I keep hearing is still the fastest growing.

I still don't care much for loud / dark / drunken nightlife stuff, the crowds, the traffic jams, but it is great to see a

More than pick-and-place

[MetalliQaZ]

I can't see from work but is this anything more than a pick-and-place machine? For home made circuit boards, that is hardly the bottleneck. Etching and drilling is the part that needs machine precision.

Re:More than pick-and-place

[Nadaka]

Ultimately, the machine will etch traces, apply solder paste, place components, cook, and test. Version 1.0 places components.

Re:More than pick-and-place

[Nadaka]

shit damn, I forgot the quotes and the snippy annoying RTFA comment.

Re:

[BlueStrat]

"Ultimately, the machine will etch traces, apply solder paste, place components, cook, and test. Version 1.0 places components."

FTFY on the quotation marks.

My thinking is why replicate functionality that's been available for many years with V.1.0? There are already machines that do that, that have been around for decades.

Here, take your pick: http://www.wotol.com/product-list?&category_title=auto-insertion-machine&category_ids=1091&page=5 [wotol.com]

Call me back when he starts a prototype to actually do the etching, soldering, testing, etc that is mentioned.

Strat

Re:More than pick-and-place

[Alioth]

Those machines are also eyewateringly expensive. This machine is actually pretty damn cheap even if all it does is pick and place.

Re:

[BlueStrat]

Those machines are also eyewateringly expensive. This machine is actually pretty damn cheap even if all it does is pick and place.

Sorry for the late reply. Those machines were just the quickest example that I could link to. The reason for the expense is because the high industrial/production volume/rates they are built for. Those machines are for high-volume, large-scale production operations in large factories.

People have been able to easily build a pick-&-place type machine like one that loads a standard through-hole printed-circuit board using solenoids/servos/stepper-motors and bog-standard industrial/commercial OTS programma

Re:More than pick-and-place

[Alioth]

Actually a pick and place machine is precisely addressing the bottleneck if you need to make small runs. Sure, for a 1-off board it doesn't matter much. But if you want to make a run of 50 boards, you can get the bare PCBs made very cheaply, but assembling the boards is another matter altogether. (I have actually found a company that will assemble 50 boards for a reasonable price, but it would still be a lot better - even if it wasn't any cheaper - to have the assembly done genuinely "in house", not to mention a much faster turnaround time).

Re:

[Grishnakh]

I have actually found a company that will assemble 50 boards for a reasonable price

Which company is that? I've been looking for such a place and haven't found much, only places with ridiculous prices.

Re:

[Anonymous Coward]

I have actually found a company that will assemble 50 boards for a reasonable price

Which company is that? I've been looking for such a place and haven't found much, only places with ridiculous prices.

Ridiculous price of course is relative, some I've found.

http://www.internationalsensor.com/
http://brightcircuits.com
http://ourpcb.com
Many more China outsource companies, just search for pcb assembly

The first one, everything is done in the U.S. (Nebraska actually) and they do a pretty good job. A bit more expensive than the China outsource ones, but it isn't bad.

Just pick and place?

[PPH]

Or will this have interchangeable heads that do mask printing, PCB milling, drilling, etc?

Close...

[gabereiser]

But show me one that makes the board, etches, and then assembles and you'll have my money in a heartbeat!

Re:Close...

[wierd_w]

So, inkjet a wax and asphaultum solution through a heated nozzel, touch gently with a few passes from a hot air blower, then dunk the whole thing in the etching bath.

Wait a preprogrammed amount of time, fish it out, then plunk it in hot soapy water, agitate, then hang up to dry.

(News for nerds: beeswax and asphaultum have been used as a deep etching mask for centuries, and is used to mask iron cutlery blades for acid etched artwork. Filtered mixtures of the stuff would lend themselves very well to existing 3d print systems, as it is both cheap, and reusable, with a low melting point. Copper etches much faster than iron, and the depth of etching is far shallower. The most expensive product involved would be the acid etchant itself, and let's face it, a strong solution of CLR will work just fine here, as would dollarstore knockoff HCl based toilet cleansing gel, and those are both pretty damned cheap.)

Re:

[gabereiser]

yeah i know you can etch your own boards (I do) but to have a little desktop "autobot" that can not only cut the copper to size, etch the circuit on it, then assemble the pieces would be a HUGE breakthrough for us hobbyist tinkerer's.

Re:

[wierd_w]

That's what I was getting at. A carousel tool cassette with fixed position head, an X,Y,Z table with 2 bath tubs on another x axis slide table, 2 reservoirs, one with solder paste and another with wax+asphaultum mask, and a simple means to drop and raise the table in and out of the tubs, and you have a winner.

Whole process:

you still manually saw/cut your prototype board to size, then lock it to the table.

The first NC program is loaded and started. The robot uses the camera to find the edges of the prototype

Re:

[gabereiser]

You sir, have my money! Please... I don't have the time, nor the patience to build this but this is pretty much exactly what I (and other) want. MakerBots are cool, but what about a PrinterBot? You could probably add the reflow step to this machine if it was enclosed somehow, and then its up to the user to hand-solder the large through-hole components.

Pick-place and solder paste are the issues

[dbc]

A $2K device that does solder paste and Pick-and-Place is what we need. You can have circuit boards made easily and cheaply from a number of places. It's been a loooong time since I thought it was worth the time and hassle of playing in the soup myself. I don't see the point of trying to make PCBs at home any more. Toaster oven or hot plate soldering works great for suface mount. The two killers are 1) applying solder paste, and 2) pick and place. So, a cheap reliable stencil is one option for older. A friend of mine has a Mikini 1610L CNC mill, and we did a hack to add a manual solder paste syringe (one of the compressed-air driven hand-held units) as a tool head. Our first attempt got some nearly usable boards, but it would require tuning and another rev to get the right amount of paste and make it all work. Other people have done hobbyist grade Pick-n-place. Combining the two operations, adding the webcam for precise part orientation, and hitting $2K would be a game changer.

Re:

[egcagrac0]

Instinct tells me that CAM-daubing solder paste is only barely efficient for a one-off. At about a quantity of 4 pieces, it should be far easier to cut a stencil, align it with the bare board, and squeegee solder paste onto it.

If the $2000 device can make the stencils, now you're onto something.

There is a slight advantage to making the bare boards yourself - time. You're not waiting days or weeks for the board shop to turn your job, you can load up a blank board and have an assembled prototype tomorrow (w

Re:

[dbc]

There are plenty of board shops where you can get 2 to 8 or so boards turned at reasonable (hobbyist-friendly) prices and have boards in 4 or 5 days (no mask and silk screen). With 6/6 design rules and plated-through holes -- if you can do 6/6 and through-plating in your garage with reasonably priced equipment and good reliability, and repeatability, and not spend more than a couple hours at it, then wow, show me how. Until then, I'd rather spend my time designing and debugging than messing around with a

Re:

[drinkypoo]

A quick to make and cheap disposable stencil would be a winner, but I haven't found a way to make that yet.

How about punching the holes in a sheet of some kind of plastic? Using the nearest hackerspace's laser cutter seems like a reasonable solution to me, however.

Re:

[Grishnakh]

There's a couple of places that make laser-cut Kapton stencils for pretty cheap prices: ohararp.com and polulu.com. But for a one-off board, it is a significant cost (usually about $25).

Re:

[egcagrac0]

You're absolutely right on the layers & plated through holes. Anything you can run in-house is extremely likely to be single or double layer, with no plated through holes. That doesn't mean they're not useful, but it is darn hard to use a high pin count surface mount chip with them.

Etching your own board might be useful if none of your devices are more than about 28 pins. I guess I was thinking that if you're doing anything more complicated than that, there's enough board design to warrant a commerci

Re:

[Bassman59]

I haven't done a board which uses less than four layers in ages, so something that will etch or mill copper to make traces doesn't do much for me. I can wait a couple of days for a proto board house to send me a package.

Comment about the quality of the video:

[blind biker]

I can't hear shit! [ebaumsworld.com]

Re:

[ArcadeMan]

Look at my laser printer! It's like a 3D printer but it prints on paper using toner instead of depositing plastic in layers!
Look at my CNC mill! It's like a 3D printer but it cuts material instead of depositing plastic in layers!
Look at my pick-and-place machine! It's like a 3D printer but it places components on a circuit board instead of depositing plastic in layers!

Re:

[egcagrac0]

Look at my laser printer! It's like a 3D printer but it prints on paper using toner instead of depositing plastic in layers!

Just one layer on substrate is the limit with most of them; there are registration issues when you try to do multi-pass.

Toner IS plastic, isn't it?

Does anyone think about instruction?

[spopepro]

There's been an explosion of tools for creation coming out at low prices, and every time someone says "it's for schools!" like the only things that's keeping students from an engineering curriculum is the cost of the hardware.

The biggest obstacle is instructor support/training/professional devleopment/curriculum... basically everything except the hardware. So in the mean time you have university/foundation sponsored projects at indivudual schools that get everyone excited, all of which have absolutely no portability to any other context. So then we're back to individual people doing special things and you're lucky if your kid is at that school and screwed if they aren't.

But we get to feel good about "doing something for education", I guess...

Re:

[Grishnakh]

I don't see how that situation is going to change much. The people who are really into things like building their own 3D printers or PnP machines are not the type of people interested in being high school teachers and dealing with unruly teenagers. Just take a look at the video; does that guy look like someone who could (or would want to) handle a classroom full of disrespectful teens? Maybe if they stopped mainstreaming everyone and separated kids into different schools, the way they do in Germany, so t

Imagining, yay

[wonkey_monkey]

Okay, now imagine Lindsay Lohan, naked, doing a backwards crab walk.

Just do this for me!

hearing

[hackula]

guy needs to print himself a good hearing aid...

Imagine?

[ArcadeMan]

There's already other people working on a pick-and-place machine [buildyourcnc.com]. Granted the future goals of Board Forge are greater, but combining a multi-head CNC mill with a pick-and-place machine is not a new idea.

Re:

[Grishnakh]

He looked to me like he'd make an excellent Vulcan.

So much cheaper and simpler to use existing

[gnu-sucks]

There's a company called "Batch PCB" that will do small quantities of PCBs for reasonable costs if you don't mind waiting a bit. They just put several designs together onto one PCB, send it off where they are getting a bulk rate, and then cut the boards apart when they are done, and send them out.

The hassle of running a machine like that is really not worth it.

But yes, very interesting and impressive nonetheless.

Sucks

[goose-incarnated]

The video doesn't actually show the machine in action, only has a talking head explaining what it will do. Waste of time; you can't even tell if the thing works.