The purpose of this project is to deliver cost-effective pick and place machines to makers, businesses, laboratories, academic institutions, and others interested in electronics manufacturing. A pick and place machine automates the assembly of electronics by picking up tiny electronic components and placing them on circuit boards.
To get all the latest updates on this project please visit our Hackaday page https://hackaday.io/project/5200-pick-and-place-machine
Our automatic feeders are the biggest differentiator between us and other projects like ours. We laser cut the parts for our feeders locally. We have a demo video of our previous prototype of the feeders here.
Each feeder bus has a master feeder that talks back to the control computer with a USB cable. Each master feeder can be connected to multiple "daughter" feeders through a CAN (Controller Area Network) bus. The CAN bus is what is used in automobiles for reliable communication among all of the electronics in vehicles. The feeders are daisy chained together to establish communication from a feeder, to the master feeder, back to the control computer.
Various parameters stored on the feeder control the actions of the feeders, such as how many steps the stepper motor should take to advance the tape reel, how long the tape retraction motor should run, and so on.
We will release the details for all of this communication in case someone wants to use our feeders with a different machine.
The YouTube video https://youtu.be/Ny99KzsySHE
shows placements of 0603 and 0805 passives as well as TSSOP-28 (0.65mm pitch) and TQFP-44 (0.8m pitch) packages. These are not the limits of the machine! It's just what we use. We will upload videos to demonstrate the machine's performance with other packages. If you have a package you're curious about, let us know, or offer to send us a part to try!
In terms of components-per-hour: the average placement speed as shown in the video is 260 components-per-hour. Yes, we know this is on the slow side. This is with full vision processing on every part (worst case scenario). If you line up your parts square with your circuit board and don't need to do rotational vision correction on every part you can currently get closer to 850 components-per-hour.
We should be able to move our speed numbers up substantially before shipping. Our new head will enable this in several ways.
With our current pick up head our computer vision might tell us we need to rotate 3 degrees, then we tell the motor to rotate the part 3 degrees, but then we have to check to see if we actually rotated the part by 3 degrees. We have to check our work because of the backlash between our motor and our suction tube: with the backlash we might not actually rotate the part a full 3 degrees. The new head has zero backlash. So we'll be able to fly over the upward looking camera, grab a snapshot, immediately start on our way to the placement location, and apply the rotational correction as we're dropping the part after the computer vision has operated on the snapshot it took.
The new head will also have the cameras rigidly mounted to it, so we should be able to go faster without the cameras unintentionally moving from the acceleration.