Delta mechanism design and Frame construction

A project log for FirePick Delta, the Open Source MicroFactory

An affordable electronics manufacturing system for hobbyists, students, & small businesses. Inspired by RepRap. Powered by OpenPnP/FirePick.

Neil JansenNeil Jansen 06/07/2014 at 16:472 Comments

This post will give the reasonings behind the frame and the delta mechanism design.  We're using a pretty unique design, in fact I'm not aware of any successful pick and place machines that use a delta mechanism.  We'll explain why we went against conventional wisdom and chose the delta mechanism.  We'll also explain what the heck a delta mechanism is for the more casual readers.

Here's what your 'status-quo' SMT pick and place looks like, with  Hackaday judge LadyAda for size comparison:

Ian from DangerousPrototypes is no stranger to a Pick and Place machine either.  Pictured below, he has a NeoDen TM-220A entry-level pick and place machine sitting on his desk.

TM220A table top pick and place overview

Here's another Hackaday judge, Dave Jones from EEVBlog, giving a great walkthrough of electronics manufacturing that features his uCurrent widget getting assembled on a professional pick and place machine.  The video is highly recommended for those that are not electronics gurus.

I know that there's also a picture of Bunnie Huang standing in front of a pick and place machine somewhere, but I haven't found it yet :-/  Here's a great video with Bunnie and Ian talking about some of the headaches of manufacturing, like the hot dog bun deficit problem, and how it relates to component reels.

You get the idea.  The average pick and place machine is quite huge, and hard (if not impossible) to fit in a house or apartment.  The notable exception being the NeoDen TM-220A that Ian demonstrates above.  But it is a really a crappy machine to put it bluntly. 

Cartesian machines like these are really simple to understand.  You have three axes in which the end effector can move: X, Y, and Z.  You use linear rails or slides for the effector to move smoothly, and use some method of belts, ballscrews, or something else to link the motors to the other pieces.  In practice, these rails and all of the metal or other structure to support them gets heavy and expensive.  

Furthermore, most RepRaps that tried to save cost by moving the bed on one axis and then moving the end effector in two other axes.  This works great for 3D printing.  However it does not work for a pick and place machine.  Why? Because we want to keep the bed stationary and move the end effector in all three axes. 

Some recent 3D printers (Rostock and Kossel derivatives) use a variant of the delta robot geometry.

Here's the original drawing from patent US 4976582.  

It's kind of a bad picture, so let's watch some videos instead!

Holy crap.  Did you just see what that thing did?  

Delta robots are the fastest robots in the world.  They are fast because the heavy motors stay stationary, and a super-light end-effector is attached via six (or more) rods.  There's less moving mass and therefore the rest of the machine doesn't have to be as heavy to keep it from bouncing around.  Seems like a perfect fit for a desktop SMT pick and place machine. 

The astute reader will notice that our design uses the conventional (rotational) delta design, rather than the Kossel and Rostock "linear tripod" delta design.  This allows us to use a rectangular frame and place component feeders on all four sides.  It also allows us to chain several machines together in an assembly line fashion to make a distributed networked manufacturing system.  It gives us less Z height than the linear delta, but that's an acceptable tradeoff for our machine.  We feel that it also makes for a cheaper design.  We don't require Maker Rails, 8mm smooth rods, or linear bearings of any sort.  Pretty much everything can be 3D printed except six 608 bearings, a handful of cheap hardware fasteners, and some hobby RC rod ends.  We're trying to make it even cheaper by going with ball bearing and magnet ball joints.  We'll obviously post updates as to how that works.

Here's our top-level proposed specs (the target cost and price was more in this slide, because it also included other accessories that the Hackaday version of the machine didn't initially come with.  You should still be able to make your own for about 300 bucks if you're good at sourcing parts).

We feel that the size of the frame is a perfect size for desktop use.  You get the same bed size as a standard prusa i2 or i3, possibly even more because some of these machines can't print to the boundary of the bed platform.

The frame construction is extremely simple.  It's made from 12 pieces of Misumi HFS5-2020 aluminum extrusion.  Here's a great article on using aluminum extrusion in your own project:

Extrusion, frame, top/bottom, 260mm
Extrusion, frame, sides, 500mm
Corner braces, cast metal
Extrusion, nuts100MisumiHNTT5-5

The extrusion materials for the frame work out to about $38 USD if bought directly from Misumi.  They will cut the pieces for you to the exact sizes, which is really nice of them. The corner brace pieces can be 3D printed from plastic, or you can buy the metal version for about $0.75 each, or $12 total.  It's up to you if you're going the cheap ($300 total cost) route, or if you're making a fancy one.  There's also a few pieces of acrylic.  The size of the machine was actually made to fit on a Full spectrum Engineering Hobby Laser cutter, which has a bed size of 12" x 20".  The first prototype that I made was bigger than that, and I couldn't make my laser cut pieces :(  Anyway, the cost of acrylic is pretty cheap, and you can safely use wood or other materials that you have laying around.  You can even hand-drill and cut the material if you don't have a Makerspace or Hackerspace close by.

Here is the delta motor assembly.  Three of these assemblies get mounted to the frame.

Here's what it looks like when three of them are mounted together, with the linkages and and effector:

Which looks pretty crazy, but here's what it looks like when it all comes together:

Thanks, and stay tuned for more build log entries.  And let me know in the comments if there's anything else that I missed.


barawn wrote 06/12/2014 at 17:59 point
I really would recommend considering adding a second up-looking camera to the design - once the part is picked up, you'd really like to make sure that its orientation and location is identical to what you thought it was when you picked it up. For small parts this isn't that big a deal if you're accurate enough on the pickup, but for larger, fine pitch parts (specifically BGAs) a small angle error can correspond to balls not being placed right.

The pick and place I have access to (a Madell DP2006-2, which is a relatively low-cost full-vision setup) actually has 2 uplooking cameras, one narrow-field and one wide-field for larger parts.

I personally feel that the most important use for a pick and place is for BGAs, because you only really have 1 shot at getting those right.

Anyway, just a recommendation.

  Are you sure? yes | no

Neil Jansen wrote 06/13/2014 at 21:33 point
I agree that up-looking vision is extremely important and can't be overlooked. It is our intention to support it fully.

For our cheap $300 version, we are using a mirror mounted to the table in order to use the down-looking camera to see the bottom of the part. This is sort of a cheap compromise but is quite elegant for most situations. We also intend to offer a second up-looking camera for those that want a higher performance option, and have a lot of big high-pin-count devices to place.

We're working to make our system extremely modular so that any number of cameras can be added, for any purpose. Our FireREST implementation makes this trivial. We are certainly not boxing ourselves into any corners here :)

  Are you sure? yes | no