I purchased the CAD (SolidWorks) files from Chris’s website so I could make some design modifications and then 3D print. I wanted to be able to export high-res STLs, so curved surfaces were curved rather than triangulated. I printed everything on my Prusa i3 MK3 (8” square build surface) with IC3D Red PET-G. All in, not considering reprinting any parts, it was just north of 70 hours of print time and under 2kg of filament.
Maybe the most visible design modification is at the base. I integrated the base and the J1 Turret, as well as the support ribs. This took the longest to print, at around 15 hours.
Due to the build surface constraint, I had to cut the J2 arm in half. For increased rigidity, I made the walls taller, similar to the supports Chris recommends when printing. I made some overlaps and printed a separate, bolt-on stringer to tie them together. It’s still a little wobbly, so I may revisit the design.
The modification with maybe the most design work is with the J5 actuation. I changed from the lead-screw-driven system to a bevel-gear-driven system. Initial (no load) testing shows pretty good prospect. By tensioning the belt, the J5 extended arm is slightly bent (it’s plastic...), so I may revisit this and go to a more direct-drive system. But for now I’m happy with how it’s working.
By changing to the bevel gear drive, I used a geared stepper, which sticks out the back of the J4 arm further. To utilize this space, I incorporated the J4 hub into the J5 motor mount.
For the J5 drive pivot on the J6 assembly, I had a hard time getting enough torque on the J5 drive pulley set screw without cracking the plastic. I modified the whole J6 design for ease of printing (modular instead of all-in-one) and used the leftover 8mm keyed shaft from J3 to fix the pulley. So far so good!
As for the electronics...that’s the next part of the fun. I plan to make a custom PCB that acts like a shield for the Arduino Mega. It will have screw terminals for the motor driver control wires, as well as for external IO. I’ll also include 2 DC-DC step-down converters to take the 24V from the power supply and output 12V for the Arduino and 5V for the RPi (see next paragraph). That way I can eliminate the 5V power supply. This concept needs validation, so I’ll update this post once I get it sorted out.
For actual control, I’ve modified the AR2 software to work on the RaspberryPi. I’m planning to use an RPi Zero W. I have it set up as a wireless access point, so I can access the GUI with a tablet via the free VNC Viewer app (android and iOS). Initial testing shows good prospect, but I need to validate that the RPi can handle the processing. If it works, I’ve got a headless unit - no Windows computer required!
That’s all for now, I’ll continue to update this post as I make more progress. Let me know if you have any questions!
-Zach
Nice work. Its awesome to see what you have done. Thanks for posting this. Please keep us updated on your progress. I’ve had a number of people asking about adapting the software for raspberry pi - if you could post a how to, a lot of people would be very interested. Thanks again.
Absolutely, I’ll get that together. It’ll take a few days, but I’d be happy to share. I’ll email you the source files so you can post on github.
Nice work!
Update for today:
I finally got everything wired up temporarily to test the functionality before continuing. One important note is that I didn’t wire the limit switches per Chris’s instructions. I used only 2 wires, one connected to COM and the other to NO. For the limit switch to work properly with the code, it needs to be pulled to ground (LOW) when open, and at 5V (HIGH) when closed. The 3rd wire (connected to NC) to ground accomplishes that. I did it slightly differently. Since I didn’t have a ground connection, I changed the Arduino code such that the switch closed would be to ground (LOW) and open is at 5V (HIGH) using the Arduino’s internal pull-up resistor. The code was pretty easy to modify - all of the calPin digitalRead commands had to have HIGH/LOW flipped. This reduced the physical wiring. I’m not sure if I’ll keep it this way, or if I’ll incorporate pull-down resistors on the shield I’m going to start working on shortly. Probably the latter...but this got me going along in the meantime.
I’ve been testing everything on Windows, but once I get a little further along I’ll test the full functionality on the RaspberryPi. And then I’ll provide a tutorial on how to get the RPi going.
I saw that Chris posted the 2.0 software, so I may go through that to make it RPi compatible, but for now the focus will be on 1.5. I’m pretty sure I won’t include the Xbox controller functionality on the Pi, but it’ll be helpful to have the trajectory mapping that’s he’s included.
Awesome. Great feedback, thank you
I made some progress with the electronics. I designed a custom shield for the Arduino to incorporate some different components. I have the 24V coming to the board at the bottom, then I have 2 buck converters to get 12V to the Arduino and 5V to the Pi. I ran out of headers so I still need to solder the Pi to the board. I know I had some I/O I wanted to add (buttons, switches, sensors, etc) so I implemented 4 molex connectors also. And then I used screw terminals for the calibration, Pul/Dir wires, and additional I/O. I’m waiting for the resistors to arrive for the limit switches, and then I’ll be able to put it through the paces.
On another note, I’m thinking about reworking the whole J1 assembly. It will be introducing more cost (I ordered a $140 bearing), but it may give me better stiffness with the 3D printed assembly at J1. More to come on that later...
looks good. The shield looks really good. I was thinking about making a shield to step all logic to 3.3v so I could use a Due down the road but haven't had time to mess with it. Ive also wondered in the past if I could run solely on a Pi and us the GPIO instead of the arduino but im not sure if step timing would be reliable. Let me know if you can provide instructions for getting the software to run on the Pi -had a few people asking. Thanks for your help and for posting your progress.
Sorry for the delay in getting out instructions for the Pi. I was waiting for this board to arrive to make sure the pi worked before sharing the source file. Work and the holidays reduced my evening hours haha.
I‘m not familiar with the Pi at all, so I’m not sure if it can do all the step commands. Since my shield uses adjustable buck converters, it can do whatever voltage is needed. I can pull out the headers for my personal customization, update the layout (I forgot to label all of the I/O on the right side...) and share the eagle files, if anyone is interested.
The 5V buck converter is exclusively for powering the Pi. The 5V for the Pul/Dir and the calibration are all from the 5V pin on the Arduino. I didn’t want to disrupt all of that logic, I just didn’t want another cable plugging into the already cramped Pi Zero W.
I’m not familiar with the Due; for this application, what are the benefits over the mega?
its nearly the same board with much faster processor - it uses an ARM processor instead of AVR so it runs at 3.3v which is a pain since most stepper drivers are 5v.
Gotcha. Yeah, 12 step-ups for the pul/dir doesn’t sound like a great thing.
Could you load the arduino mega skecth directly into the arduino Due or do you have to make a new sketch? To be able to excite the drivers, since the drivers of the AR2 are excited with 4v minimum, could be used for example DRV8825 drivers, which reach up to 2.5A and use 3.3V as well, or could optocouplers such as 6N136 be used to make Relay between the Due to 3.3 and the drivers to 5v? The truth is that using a Due, the movements would be more fluid, as happens with the Delta type printers, the 8-bit arduino are a bit short in calculation speed to interpolate the axes.
@Víctor Manuel Castañares Martín yes you can load the sketch to the due but as you point out you would have to use 8825 drivers or Gecko drives.
I've never used GitHub before, so I'm still trying to learn how it all works. But I made a fork of the AR2 project and uploaded the board Gerber files as well as images of the board layout:
https://github.com/AllenPrototype/AR2
I ordered my boards through JLCPCB.com, and got 10 boards for $20 shipped to Ohio, USA.
If anyone has any questions, let me know!
Update on the RPi AR2 controller... I’m planning to test with the robot today, and if all works I’ll put together a tutorial.
I got it wired up and it works! I’m ecstatic. I tried to upload a video to YouTube, but there was an error. I’ll try again later. @Chris Annin, I’ll get to work on the how-to
that sounds great. thank you.
Hi, Everyone!
I haven't posted in a while, but if anyone is following this and is wondering about how to get the software to work on their Raspberry Pi, @Chris Annin pulled my code/documentation into his repository: https://github.com/Chris-Annin/AR2/tree/master/RaspberryPi
Updates:
I'm still working through my somewhat-modified 3D printed AR2 robot, but have recently shifted focus to the stepper drivers. In @Max Favre's build log, we discussed using Trinamic's motion control drivers for the stepper motors. There's a product on the market called SlushEngine, which is an open-source project by Roboteurs - you can check out their website and GitHub repo here: https://roboteurs.com/, https://github.com/Roboteurs/slushengine. This product seems like a great option, but at a pretty steep price, especially compared to the drivers @Chris Annin picked from StepperOnline.
I'm working on making a driver system similar to the SlushEngine, but with TMC5160 drivers by Trinamic. They offer a lot of impressive functionality, but I won't go into all of that in this post. They offer a small form-factor, as well as high amp-per phase current. Additionally, they offer 256 microsteps per step, which gives them very smooth and precise operation. One of the best selling points for this application, which the SlushEngine also offers, is direct positioning. As in, the motion controller is built into the chip, and you only have to give it a position command and it does all the hard work internally. That means no step-direction, no Arduino required, and no real-time OS (e.g. RPi Raspbian OS).
I've modified the SlushEngine project for the TMC5160s, but it's very much a work in progress. It works entirely (and exclusively) on the Raspberry Pi, which aligns with my goal for the project - drive the AR2 with only a Raspberry Pi, preferably the $5 RPi Zero W. My milestone tonight was getting the library at a point where a very short Python script would make the motor run 5 rotations (arbitrarily picked) back and forth.
While I want to tap into the functionality of the drivers more, I'm going to start a parallel path of adapting my already published RPi code with this new driver. Overall I think the AR2 software, written in Python, will work great with these drivers. The drivers have inputs for limit switches, so the home position can be recorded by the driver. This should greatly simplify the homing scheme.
One of the trickiest parts I believe I will deal with is syncing all the motors. Meaning getting all the motors to run the same amount of time, regardless of how many steps they have to take. There's no problem getting them all to start at the same time but for precise and complex motions, having some sort of CNC-esque logic will make this a lot better.
Overall, I'm quite impressed by these drivers, and hopefully I can get this library matured and integrate it into the AR2 software quickly.
Another important piece of the pie is obviously the driver circuit board. Now that I have a working prototype with Trinamic's pre-built breakout board, I will begin to tackle designing the PCB, sourcing the parts, and assembling it all. I'm not sure the price-point will get below the StepperOnline drivers, but it will hopefully be cheaper than the SlushEngine.
Since I'm building off the SlushEngine project, instead with Trinamic's famed silent stepper drivers, I'm calling the project "ShushEngine". I'm bad at coming up with names, so let me know if anyone has any better ideas. You can find my GitHub repo where I'll be constantly updating the code at: https://github.com/ZJAllen/ShushEngine
Here's a quick video showing the motor spinning with the code written behind it.
ShushEngine is super cool. I remember that you mentioned somewhere in the forum that you dev in eagle. Do I have that Correct? Would love to help work on this with you If you need / want a contributor. Please DM me at some point. Regardless I’d love to do a hangouts session with @Julian Holcroftyou and @Max Favre. Also, so glad your tackling this project; excellent contribution and very excited for this development.
@matthew flego I made the Arduino shield in Eagle, but will probably move all future work to Kicad. I’ve never used Kicad, but hopefully I can pick it up quickly. The reason for the switch is the free version of Eagle limits to 80x100 mm (hopefully I remembered that that correctly) and the ShushEngine board will need to be larger than that.
@Zach Allen makes sense to me. I have the oportunity to learn KI CAD shortly through a makerspace in town. we’ll see what happens. Are you farmilliar with CAElinux?