Forgot your password?
typodupeerror
Build Hardware

A Look At the Firepick Delta Circuit Board Assembler (Video) 43

Posted by Roblimo
from the components-get-tinier-every-year dept.
From the Firepick website: 'We are developing a really cool robotic machine that is capable of assembling electronic circuit boards (it also 3D prints, and does some other stuff!). It uses a vacuum nozzle to pick really tiny resistors and computer chips up, and place them down very carefully on a printed circuit board.' There are lots of companies here and in China that will happily place and solder components on your printed circuit board, but hardly any that will do a one-off prototype or a small quantity. And the components have gotten small enough that this is really a job for a robot (or at least a Waldo), not human fingers. || There are obviously other devices on the market that do this, but Firepick Delta creator Neil Jansen says they are far too expensive for small companies, let alone individual makers.

The Firepick Delta Hackaday page talks about a $300 price for this machine. That may be too optimistic, but even if it ends up costing two or three times that amount, that's still a huge step forward for small-time inventors and custom manufacturers who need to populate just a few circuit boards, not thousands. They have a Haxlr8r pitch video, and have been noticed by TechCrunch, 3DPrintBoard.com, and Adafruit, just to name a few. Kickstarter? Not yet. Maybe next year. Open source? Totally, complete with GitHub repository. And they were at OSCON 2014, which is where Timothy found them. (Alternate Video Link)

Tim: All right. So, Neil, let’s talk about the FirePick Delta?

Neil: So this machine came out of my personal frustration of making small runs of circuit boards and having a very difficult time trying to get them manufactured. If you’re doing a couple of boards, normally you can place them by hand, the parts are getting smaller and smaller these days, so it’s generally you got to have good eyesight, you got to have steady hands, it’s very difficult. On the other side you could go to China and have boards manufactured over there, but you either got to do in high quantities or pay a lot of money and there’s really not been a decent attempt to trying to make very cheap sub-$1,000 pick and place machine that can manufacture things in your home without a lot of work. There’s been several attempts at doing DIY machines like this, but ours, we kind of started our philosophy on principles, so I mean we built this machine using the Delta configuration and a lot of the different other techniques that we used to kind of hit the different requirements that we have for the project. So we didn’t want to go with a Cartesian machine because we felt that with the linear bearings and linear rails and everything like that, that the price will be too high, so literally the entire Delta mechanism on this, 99% of it’s 3D printed. I mean there’s three stepper motors and some bearings, and that’s about it. So the frame is about $20 retail and most of the parts are 3D printed, so it’s a very affordable machine. I think it’s definitely going to be the most affordable pick and place machine out there, and we put a lot of thought into the vision system, the feeders and every other part of it to try to make it useful in a desktop environment.

Tim: Talk about the resolution or the accuracy of it?

Neil: So we did some simulations on the accuracy to try to make sure that it was going to be accurate enough to place the smallest parts that you can get nowadays. Because we went with the unique Delta configuration, we actually had to do a lot of simulations and trials to see that we could actually hit those accuracies. The machine itself as far as resolution is capable of placing [201]components which are basically like your – if you would open up like an iPhone or a android phone, those were going to be the super, super tiny parts that are found in there. So resolution-wise we’re able to do that, we’re still obviously going through systems level testing now and we’re kind of close to that accuracy, but we do have a little more work to do, we got calibration routines that we actually use the camera to do a spiral pattern on the bed and auto Z-probing to try to improve the accuracy for stuff like this. We use the camera for determining when we’re picking parts out of the feeders. We have a mirror that we’re using as a cheap up looking camera for looking at positions of the part on the nozzleand then we can also do fiducial recognition on the board to locate the parts properly, so the closed loop vision and software complexity kind of is a mitigation technique for not having so complicated of a mechanical system.

Tim: What kind of camera you’re using?

Neil: We’re using a standard 5-megapixel Raspberry Pi camera. Normally when you get it, they give you a little ribbon cable, so you’ve got about a 6-inch reach from the Raspberry Pi camera, but we added – we made our own adapter boards and we’re actually using an HDMI cable to pipe the signal over. It’s not an HDMI camera obviously, but if you actually look at the Raspberry Pi camera, the electrical interface is close enough that we can actually use an HDMI cable and it just happens to work.

Tim: You just learned that from experimentation?

Neil: We just looked at it and just saw that they kind of looked identical and just decided to go for it, that might be something that we make and sell later on because there’s not really any other decent Raspberry Pi standards out there, and it does work pretty well and it’s EMI-safe, so.

Tim: Is this the one in the world or is this one of many that you built?

Neil: This is the first prototype. Basically I started working on it in January/February of this year, so I mean, with the 3D printing and all the other rapid prototyping techniques if we’ve been able to iterate a couple of times and make a pretty decent prototype so far, we have about 20 to 30 beta testers now. We’ve got maybe two or three that are starting to be built at this point. And we’re going to really try to, I guess, skip around a bit. We have some Kickstarter veterans, that have done unsuccessful Kickstarter projects that have actually decided to become a beta tester, so that they can use a machine like this to do their small production runs. They are a great resource because they’ve kind of seen the stuff first hand. They’re kind of our model customer and we’re going to take their advice and what they like and what they don’t like, refine it further and it’s going to be one of those gets-done-when-it-gets-done kind of projects, because we’re trying to refine it a lot before we release it to the public. But we’re hoping to maybe have something kind of towards the general public in six months to a year. But rather than your conventional pick and place machine where you’ve got to program the seeders and you got to do all this work before you start a job, this is going to be just a plug-and-play system and we’re using HTML5 Node.js Server with Express and Bootstrap and where you could pull it up with a tablet or a laptop and just go from there.

Tim: And as a prototype you’ve got a lot of things sort of stacked on top or you got the Raspberry Pi and then closer you’ve got some circuit boards up here, these are all going to be compressing to a more polished looking total?

Neil: Yeah, absolutely, so we based all the motion control on this machine off of a standard RepRap Marlin software and RAMPS 1.4 board. The cool part is that we can use that design which is stacked to bunch of circuit boards with a bunch of wires and we can use that as a prototype and then we’ve already started the process to kind of iron that down into a single board. It’s going to follow the same licenses as the RAMPS board which I think is GPL and we’re going to streamline that. All the wires will go away and we’ll have something that’s manufactureable and the cool part is we’re going to be able to manufacture the boards on this machine and 3D print the parts for the machine on itself, so we envision that as being one of the most replicatable RepRaps out there.

Tim: And keeping it Open Source is part of your philosophy?

Neil: Absolutely, we have about 90 to 100 dependencies in this project based on other Open Source projects. Some are kind of common and everybody has heard of, other ones are kind of the LGPL, like libfuse and zlib and just little things nobody ever thinks of, but they’re definitely a part of this. Our high-level software is BSD and MIT licensed. The machine itself is probably going to be GPL or Creative Commons ShareAlike license, but there’s a lot of dependencies that use all kinds of different licenses, so the takeaway there is that at the end of this project, it will be Open Source and stay Open Source because we couldn’t have done it without all of these different licenses and libraries and modules that we’re using.

Tim: Let me ask one more thing, what is it like developing a product like this in Florida versus some other part of the country? Is there a big maker ethos in your network?

Neil: It’s kind of funny, I think the internet... I mean I’ve always... I don’t know, I’m able to follow stuff for a little bit, I don’t know...

This discussion has been archived. No new comments can be posted.

A Look At the Firepick Delta Circuit Board Assembler (Video)

Comments Filter:
  • by maliqua (1316471) on Tuesday July 29, 2014 @06:00PM (#47561285)

    how can i make you take my money!

    • by gl4ss (559668)

      if they somehow manage to go 300 bucks, I'll buy it just for the steppers, alu extrusions and controllers.

      • by Skylinux (942824)

        this!
        A proper power supply and steppers are "expensive" but I would buy one if he can manage to make a unit for around 1500 EUR.

  • by zAPPzAPP (1207370) on Tuesday July 29, 2014 @06:36PM (#47561475)

    How do I feed it parts?

    I can see two rolls kind tucked to the side there, with tape just loosely hanging out. Would it somehow take parts out of those?

    What about designs with lots of different parts?
    Parts that don't come on rolls (this is about small numbers and prototypes after all)?
    How much time goes into preparing all the parts for pick up?

  • by Charliemopps (1157495) on Tuesday July 29, 2014 @06:42PM (#47561505)

    These are around $6000 but it's the real deal: https://www.youtube.com/watch?... [youtube.com]
    My fathers got a few where he works. They're so fast a crucial part of setup is laying out where the parts are in the feed tray. You put the most used parts in the center so the head has less travel, and the least used parts near the ends. It's so important that the machine calculates this for you and tells you where to place what.

    It's amazing to watch, that video doesn't even do it justice.

    • These are around $6000 but it's the real deal: https://www.youtube.com/watch?... [youtube.com]

      You may have lost a zero there--I wasn't able to find pricing on that machine specifically, but most pick and place machines of that caliber cost upwards of $50,000. There are some chinese ~$3,000 desktop machines like this http://dangerousprototypes.com... [dangerousprototypes.com], but they have poor accuracy and no optical feedback.

      A $300 pick and place machine would be awesome, but honestly anything around $2000 that can accurately place 0402 components with optical feedback would be amazing.

    • by labnet (457441)

      Not sure where you got $6k from.
      We have a Yamaha & Juki machines and they cost between $60k and $120k new + feeders.

      For $6k you get somthing more like this.
      http://www.alibaba.com/product... [alibaba.com]

  • Not so sure... (Score:4, Insightful)

    by Anonymous Coward on Tuesday July 29, 2014 @06:52PM (#47561567)

    I'm right in the target market for this device and I don't know if I'm really interested. We always do a couple small batches of prototyping on a new product, several times a year. However, I find that our circuit assemblers are all very helpful and more than willing to do this for us for a reasonable fee. If your business partners are not interested in helping develop your product (there's FAR more to it than just quick prototype assembly!) then I'd look for somebody else who isn't in it just for the repeat high volume stuff.

    Sure, this $300 machine *seems* cheaper than paying your assembly house for the service, but here it's not free either. You pay your employees and you have to train them (not just "here, this is how you load the machine", but ideally also training like IPC certifications for operators), and you'll have to assume the cost of all potential errors, from having placed the wrong part or having it placed reversed (loaded the machine incorrectly), dealing with all the placement/soldering defects that might arise, managing the inventory levels (keeping enough parts on hand of everything necessary on small reels, often at higher cost), etc. Nobody knows yet how good the feeders are, what parts it can or can't handle, board size limits, etc. There's WAY too many unknowns still...

    It's probably gonna take a good while to fine tune your process to know its limits and to get reliable results, and that time is money. Setting up a pick and place machines (and stencils and what not) is typically too labor intensive and too expensive for small runs, and I don't see how a cheaper machine will change that. It could end up being very nice but we're more than happy to outsource all that trouble for now :)

  • by Change (101897) on Tuesday July 29, 2014 @07:33PM (#47561827)

    This project appears to have computer vision for parts alignment, which is a HUGE deal for a pick-and-place. You need your machine to know if a component is oriented improperly in the reel, and to provide positive feedback on board position by referencing fiducials for accurate placement of fine-pitch components. Other pick-and-place projects I've seen in the past have been just standard 3-axis CNC gantries with a vacuum pickup, the addition of CV means it potentially can truly compete with the high-cost units.

    • by plover (150551)

      Watch the vid. Not only does it use a RasPi 5.0MP camera, but he was able to cheap out and use a mirror instead of a second upward looking camera. And yes, it aligns to fiducials. I was just disappointed the interview didn't show it in operation.

  • This means pretty soon we'll have more "robotic machines that are capable of assembling electronic circuit boards"
    than we have electronic circuit boards.

  • Eventually the machines will start designing and building better revisions of themselves autonomously, at which point it will really become their world, not ours.

  • Their presentation for investors quotes a sale price of $1000, not $300. At that price they might be able to do it. How well they'll do it remains to be seen.

    Their presentation is all about their XY positioning mechanism. But that's not the problem. The hard problem is dispensing solder paste reliably and precisely, sticking the component down, and using hot air to solder it into place. As with low-end 3D printers, most of the problems are where the weld/soldering action takes place. They don't say much

    • by linearZ (710002)

      Their presentation for investors quotes a sale price of $1000, not $300.

      The numbers seems a bit off. He said it cost $20 for the frame, but that looks closer to $40-$60 worth of 80/20. Not clear they really know how much this thing will cost yet.

      But this machine is open source, so who is going to invest?

    • by gl4ss (559668)

      I guess you're supposed to stencil the paste in first and the put it in a heat oven as if you had done the pick and placing by hand.

      it's a pick and placing machine after all so I don't know why you would expect it to do the things the machines left and right to it on a proper "factory" setup do... 1000 sounds more reasonable for the cost though. it's just a xy gantry and vacuum hose anyways, but even that 300 is not enough..

      • by Animats (122034)

        I guess you're supposed to stencil the paste in first and the put it in a heat oven as if you had done the pick and placing by hand.

        Their FAQ [hackaday.io] contains:

        • TBD - Solder paste dispensing
        • TBD - Selective Reflow via custom ATC head

        That's what would make the machine useful for prototyping. Printing a solder paste stencil can be done on a laser cutter, but you need access to one, or you must send the job out. Laying down solder paste by hand with a little syringe on each pad (probably under a microscope) takes longer than manually placing parts and is Not Fun.

        Printing solder paste with an ink-jet printer type head has been done. [mydata.com] If they c

  • by Mr.CRC (2330444) on Wednesday July 30, 2014 @02:23AM (#47563795)

    For runs of 1 to a few dozen pieces of a board which would take me more than an hour to hand assemble, I just send my stuff to Advanced Assembly: http://www.aa-pcbassembly.com/ [aa-pcbassembly.com]

    The hidden costs involved in assembling boards by hand are staggering, mostly in time. I've built an entire electronics lab, which is 3/4 storage. The buying, organizing, and storing of parts takes a big chunk of my time. If I were to set up a reflow oven, stock solder paste, etc., that would eat up more money and time. Consequently, I've become very skilled at building any type of SMD (except BGA) with just wire solder and an ordinary soldering iron. Then there is the hazardous waste management and chemical inventory overhead, and the entire day down the drain ordeals several times a year when I use the corporate application to do the waste tickets, which tells me I have to install a new Java version (different from the one in the corp. standard desktop--WTF?!?), which works after 4 hours of installing uninstalling and reinstalling, but then breaks all my other corp. apps (accounting) so I have to do the ordeal in reverse to reconcile my CC later.

    Thus, even for smaller batches of fairly simple boards, I am going to be sending all my boards to places like Advanced Circuits. Even if it costs 50%-100% more than my time is worth, it's still worth it, because my co. will let me do it, and then I can use my time for stuff that actually matters for performance. I'm sick of building more than 1 of anything.

  • I don't have my 'phones handy, so had to watch with the sound off. The video doesn't show the machine actually working, which makes it a pretty boring watch. Hell, even the guy in the checked shirt looks bored talking about it.

    I know us geeks aren't great at PR, but if you've got a machine that does something and you're at a trade show, then make sure your machine runs 24x7 - even if it's not your machine and it's just an incidental part of what you do, it's still better than leaving it idle.

  • While this looks like an interesting and cheap device to populate empty PCBs, it is only a small part of the total sulution. The PCB has to be made, solder paste added (maybe this device could be extended to do so), and most importantly: heated. Of all these steps, the pick-and-place may be the least enjoyable, but also the one that _could_ be done manualy, if needed. Still, if this device saves two days of manual labor, it already pays itself.

Do not simplify the design of a program if a way can be found to make it complex and wonderful.

Working...