Asof Aug 2020 ::: This was originally going to be a Kickstarter project but it was too much work, too much support, and not enough interest… so instead I’m releasing this as an open-source design on Github.
I’ve posted a version onto the autodesk online viewer here: https://a360.co/2D8yZbf that lets you zoom in and download the entire model.
I’ve been working on this for a few years and finally the software (OpenPnp) got good enough and the hardware (many 3d models later) seems reliable enough.
Here’s the entire machine (minus controller and power). Click here for a high-res photo.
Here’s the controller.
I’ve been working on a small project (see the PC board in the machine) and just couldn’t deal with the surface mount chips. The design doesn’t go smaller than 0805 but between the bypass caps and other chips and then a module and switches and … all it took was a slight mistake to cause an hour of tweezer fiddling and possible tombstoning and shorts and … so I finished my pick and place machine.
The machine shown here places 0805s and the modules virtually perfectly. It should easily do 0603 and maybe 0402 (I can barely see them to test with). It also has a great semiautomatic mode that supports a game controller (or keyboard) to move/rotate/place.
Here are some specs:
- 10"x14"x1.5" bed (usable 8x14 for camera coverage)
- dual Y motors
- interchangeable Juki nozzles (manual currently although automatic is possible)
- Two 5MP cameras (resolution=0.0002")
- 32 bit controller
- 15A 24V supply
It’s slow — maybe 5 secs per part with optical recognition. The video below shows the head running at 5,000 mm/m (mm per minute). It runs ok at 15,000 mm/m but I’m not convinced parts would stay on the nozzle.
Here’s a later video placing some large components. Starts placing at 1:33.
Here’s a simple video from the top (the bottom recognition is code I’m working on and really slow at the moment so it looks at the camera for a while). Volume of about 20% sounds near the real thing (you can barely hear the background hissing of the vacuum generator) on my system.
Here’s a video showing the bottom recognition in OpenPnp (Ubuntu). The head was a slight bit misplaced so normal placement is actually better than this — and this is very adequate.
Both videos are real-time so no speedup/cropping.
I’m looking for feedback / constructive comments? Do folks think there’s an interest in something like this? I’m using mine and love it :)
I’ll show comments.