Automatic Ad Lib has been upgraded to 5 flip card displays, controlled by an Arduino Yún. You can check it out at the ITP Fall 2014 Show!
Here's an outline of the process of making them.
I've always loved split flap displays. In fact, while I was looking for inspiration for thesis, I found notes about wanting to make one dating back to 2010 and earlier. In Nick Yulman's Automata class at ITP, I set out to explore the mechanism.
There is some awesome work out there from a BeachLab maker space in europe and the work of Tom Lynch. I initially worked off the Flip-Flap project which was is licensed as open source hardware and software under MIT license. Awesome!
I found that the open source CAD program they used to design some of the parts, Kokopelli, wasn't working well for my needs. Because of the setup at ITP, I needed to access and edit the models, not just get raw code for the laser cutter. Eventually I was able to build a convoluted workflow which included exporting the 3D model of the gears, flattening it back to 2D, editing the trace with illustrator and laser cutting.
It was an ordeal but I was able to make a working prototype.
Version 1 Questions
The version 1 left me with a lot of questions. It was pretty clear that the product created from files provided in the open source example were drastically different from the accompanying images on the blog post. A number of questions weren't answered. All considered, I greatly appreciated having a good place start.
- How do you mount the stepper motor?
- What's the best way to mount the wheel?
- Are the dimensions quite right?
- How to keep all the parts together?
Initial New Version for Automata Midterm
Timing belt + 3D printing > Gears
In class we looked at timing belts and timing pulleys. This seemed like it would work a lot better than the gears. I attempted 3D printing pulleys. In the final design, the larger pulley is 3D printed and the smaller one attached to the motor is purchased. Through trial and error, including ordering the wrong parts, I was able to get a pulley system working relatively well.
Eventually I switched out to large 3D printed pulleys printed by Shapeways which had higher resolution then the MakerBot version.
The motor mounts are 3D printed as well.
OpenBeam for structure
I built the structure out of OpenBeam. I believe the BeachLab version is using MakerBeam, essentially the same thing.
Some of the overly steppy stepping seems to be a result of the Adafruit Shield. When I saw the EasyDriver boards at Skycraft for $6.95 I knew that was the way to go. They retail from Sparkfun for almost $15!
Luckily the local plastic shop, T&T PlasticLand carries large Polypropylene sheets for less than $10. Unluckily, when I went to make 3x mope displays they were all out of black sheets. And they aren't getting any more for 4+ weeks. I tried to find the sheets everywhere but didn't have luck. A place in Spain offered sheets for just $400. $400! Eventually I ended up buying a white sheet, laser cutting it and spray painting it.
Polypropylene doesn't like spray paint.
Out came the sharpie and paint pen. I had to manually fix almost every card. All 135 or so of the new cards.
Choosing the right icons & making more displays
For the final, I wanted to make an array of 5 displays. This presented a big challenge. It didn't make sense for all the cards to be the same. I began thinking that the device had to have some other value as well. That's where the ability to tell time came in.
Perhaps the device would tell time 90% of the time and randomly go off into fits of icon based storytelling.
I also began to run into the horror of a billion wires. Each stepper motor uses 4 wires and the sensors require 3.
7 x 5 = 35 wires to make go.
This was easily resolved with ethernet cable. Ethernet cable is rated for a decent amount of power and having one cable out the back of each display is much cleaner.
The Easy Driver doesn't like it when a motor is disconnected while it's energized. I didn't realize it at first, but my connections with the ethernet cable were poor. I ended frying 3 boards.
I ended up with a mess of controller boards because of frying three of them. But using a protoshield, I was able to simplify connecting everything together. The ethernet cables from the displays connect to screw terminals on the boards.
I'm still having some instances where calibration isn't correct. The tight clearances within the displays sometimes lead to a step being skipped.