As Jeremy wrote about in our last Behind-the-Scenes post, testing is an area we are a bit obsessed with at Shaper. This time, I’d like to share a story about some of our most recent nitty-gritty product testing, and how we’re using Origin to quickly build automated machines to make the process faster and more repeatable.
One of the most easily overlooked — but in my opinion one of the coolest — aspects of Origin is the domino tape used by the vision system to locate the tool. (For those of you who haven’t watched our videos, this is the tape you put down before cutting that creates a field of black-and-white dominos on the workpiece and allows the tool to very precisely know its position). The accuracy of Origin depends on the accuracy of the tape. When we first started on the academic project that grew into Shaper, accuracy was the only property of the tape we really cared about. We even went so far as to build our own reel-to-reel tape printer so that we could carefully control the printing process (this is a story for a future post — one that involves more homemade machines and the aroma of solvent ink).
Now that we are in the midst of commercializing Origin, we’re putting a lot of effort into learning how to mass-produce our domino tape. We have found a great vendor to work with, and using an industrial printing press we are now producing tape about 400 times faster and with much higher quality. Naturally, we’re also sweating many more details of the tape than ever before. Accuracy is still essential, but we also care about adhesion to the work surface, how easily the tape can be removed without leaving residue, and making sure that the tape is robust enough to resist being scratched during use. It’s this last consideration of abrasion resistance that has been the focus of our most recent round of tape prototypes.
A couple months ago we ordered 64 rolls of tape with a variety of abrasion-resistant coatings. We had been hoping that an obvious winner would emerge over the many hours of tool use needed to run through this shipment of tape. We did in-house testing at Shaper HQ and then much more by deploying the tape to our beta users. But at the end of the day, when all but a few rolls of the tape had been used up, we still couldn’t draw a definitive conclusion. The problem was that our in-vivo experiment just wasn’t controlled enough. And a bigger problem was that we were out of tape and urgently about to place a significantly larger order. But which coating to go with? In order to answer this question, we quickly built an automated abrasion-testing machine to very repeatably scratch away at each type of tape and tell us which version was the most robust. Like most of the things we craft around the office nowadays, we naturally picked up Origin to help us shape all of the pieces. There’s also something poetic about Origin being used to improve itself :-)
In order to test the abrasion resistance of the tape, we made what is essentially a single-axis CNC machine with an equally singular purpose in life: to rub sandpaper over the surface of the tape, over and over and over. We run each type of tape in the tester for a certain number of cycles, and then compare the samples to determine which tape coating held up better. We evaluate the abrasion resistance by looking at the severity of the white smudges where the ink has been rubbed away. It’s a pretty rough metric but sufficient for now because it turned out that one coating was the clear winner.
Setup starts by placing some tape on the table of the machine. Sandpaper is double-sticky-taped to the bottom of the "abrasion head" — which is just a shaft collar at the end of a precision steel shaft. This shaft is free to move up and down inside recirculating ball bushings (overkill, but it’s what we had around), but is otherwise constrained to move with the carriage of the tool. The reason for floating the abrasion head in this way is to ensure that a constant force (the weight of the head) is always applied to the sandpaper regardless of any imperfections in the surface of the machine’s table. To start the test you simply press the start button to the right of the display. A stepper motor drives the motion carriage back-and-forth by way of a timing belt. Each time the direction changes, the cycle counter increments by one. The test can be paused at any time by pressing the start button again, or the counter reset by pressing the left reset button.
The control system is super simple... just an Arduino and a stepper driver. It was tempting to use the control framework pyGestalt that I wrote for my master’s thesis (which we used for our home-built tape printer years ago), but sometimes KISS is the way to go. One upgrade we’d like to make eventually is to connect a webcam to automatically make videos of the tape degrading, and to determine exactly when the dominos become unreadable by our CV algorithm. This would benefit from a computer hook-up.
I started by designing in Solidworks. For something simple like this (and sometimes for complex things too) I use a multi-body modeling approach where each component is a distinct body within a single "master" model. This makes it really easy to relate geometry and ensure that everything will fit properly. (The task of design was made easier by the fact that my friend Nadya Peek and I had already given some thought as to how Origin might be used to build CNC machines. But more on this some other time!)
I then put down a piece of 1/2" plywood onto my workbench and started cutting out all the pieces. For features like the display pocket, it’s really nice to be able to test-fit things and then quickly make minor adjustments on-tool if necessary. This is in stark contrast to the standard CNC workflow where to offset the tool path you’d need to go back into your CAM software.
For the display I bought a nice four-digit I2C module from Adafruit. Unfortunately the PCB wouldn’t fit into the narrow thickness of the plywood sandwich that comprises the machine’s base, so I had to manually wire the 7-segment display to the controller PCB. In order to create the holes for the buttons I clamped the sandwich together and drilled the holes with a hand drill.
My favorite part of the design is the belt clamp. For some strange reason I have a soft spot for the minutiae of machine design, and these annoying but essential elements certainly fit in that category. I used Origin with a 3/64" bit to create an MXL belt tooth pattern in a small piece of wood. I then glued this piece into a slot in the carriage. The belt gets clamped by screwing one wedge into another, which causes the first wedge to press on the belt and forces the teeth of the belt to mesh with the toothed wood piece.
A coping saw was used to make the wedges.
The window with the Shaper logo isn’t just decorative — with the back removed it provides access to the belt clamp for assembly.
One of the tricky bits of many linear motion systems is ensuring that the guide shafts are parallel. This is essential to ensure smooth motion (unless extreme measures are taken to float one set of bushings). In this design parallelism is pre-set by cutting small grooves into the carriage for the bushings to sit in. Origin’s precision makes it perfect for accurately cutting these grooves. But even though the shafts are now parallel, the exact spacing between the shafts is very difficult to accurately replicate in the end-support blocks. This issue is addressed by “floating” one of the shafts, meaning that its distance from the other shaft isn’t set in stone. The floated shaft sits in pockets in the end blocks that are a bit wider than they need to be to accommodate the shaft, giving a bit of wiggle room for things to align themselves. If you look at the right-hand picture above, you’ll see that the upper guide shaft (on the right) has some extra room in the end block, while the lower guide shaft (on the left) is tightly constrained.
The process of building the abrasion tester was surprisingly quick. From identifying a pressing need to pick the best tape coating, to having data, was just a couple days of effort over the course of a week. It helped that before joining Shaper full-time I had spent the prior six years working on tools to enable the rapid prototyping of automated tools, and also that my attic is filled with stepper motors, linear guides, etc. Without sounding too academic, I do believe that there is a lot of exciting work to be done to explore how the emerging field of “handheld robotics” will enable people to quickly build their own automated machines. As an aside, one of the conundrums of machine building is how to accurately create tools bigger than the tools used to in their creation. Origin, with it’s nearly unlimited work envelope, is one solution to that problem. Imagine showing up in an empty warehouse with a stack of materials and an Origin, and building all of the CNC tools for a makerspace or Fablab. (I can dream!)
I am looking forward to all that is in store as we continue developing and producing Origin, and the many upcoming opportunities to build more testing machines!
~Ilan, Shaper co-founder and mechanical engineering lead