I am working on a project to make a sorting and retrieval machine. Our current idea is to make a frame with T-slot framing and use 2 NEMA23 stepper motors to move a nest vertically and a NEMA17 to move the nest horizontally to be able to match position with pretty much a vertical grid. Any suggestions on how to wire the motors? (motor drivers, power source, etc.) I have very little experience with Arduino and our engineering mentor has none. So we are going in blind. Any advice would be very helpful.
I think you should use a 3d printer with an arduino-compatible board running the Marlin firmware (like Creality CR10 and many others) and modify it for your needs
I recommend that you start with analysing your requirements (which you may have done, but did not mention here):
Dimensions and mass of objects (incl. nest) to sort -> Will allow to define the size and stability of the grid and how powerful the motors have to be
Any time restrictions? -> Will give you a hint how quick and powerful your motors must be
Any need for a transmission for the steppermotors? -> May also influence power consumption and timing
Usually off-the-shelf driver boards are available that fit to the motors and are controlled from the microprocessor and there are usually also libraries available that fit to those boards
Just google and you will find lots of sources, like these
I use these motor controllers a lot because they are cheap, easy to wire and work pretty well. The screw terminals are surprisingly good quality for the low cost.
It takes a step pulse and direction signal from the arduino and they are actually pretty easy to use.
Your design has two challenges.
The first mechanical, providing a two dimensional positioning device that is able to position the ‘chute’ in front of the many target bins.
The second (easier) project is the controller.
You really need to start with a design and prototype of the mechanism (optimised for x-y motorised control) and get that moving smoothly.
Then your controller software - which I’m sure we can all help with.
With the latter, you need to specify the inputs that will determine the required output actions. (colour, size, weight, user inout or whatever)
That may introduce another sub project- identifying the colour, size, weight etc before the positioner can make any ‘decisions’.
Fully agree: The prototyping will also show which tolerances have to be considered.
3D printers as mentioned by @guix seem to be a good reference, so that zeroing of the nest position in x,y,z with just some switches may be enough, but requires to zero at least after every restart. Depends on dimensions and tolerances ...
Not necessarily.
If you were to sprinkle a few hall or opto sensors around the mechanics apparatus, the controller could figure out and recalibrate on the fly before the first programmed operation, and keep recalibrating itself repeatedly during normal operation - simply by passing those known sensor locations.
Agree, that is a valid possibility to accelerate the position detection! It goes with the cost of additional hardware and therefore will depend on requirements, buget and available hardware ...
Depends on the precision needed. I worked on a similar machine (actually a storage subsystem inside a larger machine). We had about 50 X,Y positions inside a structure that was about 24"x36" and we needed to position to within about 0.005" in order to pick up the items properly or they could fall out of the gripper. There was no Z axis and this complicated the motions needed.
We did it by putting encoders on the axes and having a table of each position. Due to manufacturing tolerances (we were building thousands of these), sometimes that precalculated table wasn't enough and some positions needed to be adjusted manually, so we had to make provision for an additional offset.
One of the interesting problems we ran into is that because of the distance we needed to travel and the short time available to pickup & deliver at each location, the system was driven by timing belts instead of leadscrews. Over that distance, belt stretch became an issue and on every power-up, they controller would have to measure the belt stretch and add that compensation to the X/Y lookup table.
Actually, if I had to design something today within the same constraints, I'd seriously consider using machine vision. Use the encoders to get close and then a camera for the final alignment before picking up the container. They were all identical, so that should make it easier.
Definitely use a 3D printer like @guix said. There's plenty of cheap CNC boards out there for control. If you want to split hairs you could design a PCB.
I second that ... Especially, as you may have to consider "wear over time", tolerances in production and environmental impacts like temperature, humidity, ... . On the other hand, machine vision will most probably require reproducable illumination and unambiguous marks ... So as everything -- it depends on the circumstances of the project.
Anyway: 0.005" accuracy within 36" sounds easier than it is
Come to think of it, we may have considered a camera and dismissed it due to cost (the initial design was quite some time ago). Since it's a closed structure that we had complete control over, marks or illumination wouldn't have been a problem.
However, the average temp in the box was about 5 degrees C and it had a door that opened from time to time, so condensation on the lens might have been an issue.
Every comments like ours do not directly support the TO but may give a hint to make up his/her mind what the special requirements of the specific project may be. Finally everybody learns mostly from the failures done ...
