The Isis One Linear System Design Story Sept 10 2013
Today's post will share the story of our linear systems: what we tested, what we discovered, and why our machine works so well.
Our experience began with the first scratch-built MM1.5 we made last summer. It used the old 8 mm rod/SDP bushing system off the Prusa. The linear motion didn't work well, and the bed was nearly impossible to level. We tried replacing this with a dual makerslide system involving laser cut wood, but this somehow managed to be worse rather than better. In January, we tried MTW's Igus rail upgrade kit. It was a huge improvement at first, although it still wasn't perfect. One central carriage in the center seemed a little insufficient. It seemed slightly less than rock solid, and the layer stacking was good but not perfect. We ran the machine for hundreds of hours in this configuration, and began to notice powder on the rail. The bearings were wearing down, and the slop was increasing. The bed began to wobble noticeably side to side. It developed this weird vibration we dubbed the "Igus shimmy," which manifested itself in horrible looking wobbly perimeters.
We figured that maybe the one central carriage was insufficient, and that perhaps the torsional loads on it when the bed reversed direction were too much. We switched to a dual rail system--I believe we were using the Igus WSQ-10 rail at first--so as to get better support closer to the corners of the bed. This helped the wear situation, but the layer stacking just wasn't great. There was too much slop. We ended up going through several other dual rail configurations on our Y axis--WS-10 and N27, with different combinations of standard, preloaded, and floated carriages. At this point we also tested a number of rail options on the X: single N27 (we were trying a Bowden extruder at the time), double N27, double N17, single N80, and the minirail from PBC. Nothing worked well, and everything was expensive.
At this point, we took a step back and looked at a number of commercially available products that involved linear systems. We looked at inkjet printers, scanners, and optical drives. We noticed that all of them--even DVD drives that had to read tracks spaced several hundred nanometers apart--used rod based systems. We figured that everybody couldn't be wrong, and that perhaps we, and the Reprap community in general, had been doing something wrong with rods.
The breakthrough for us came when we were repairing our old MM1.5's x carriage. It was binding terribly, worse than it had been before, and we were skipping steps way too often. We decided to replace the SDP bushings with LM8UUs (foolish in retrospect, but we didn't know) but since our only working printer was down and we couldn't print a new X carriage, we decided to make the smaller new bearings fit in the old holes by filling the gaps with hot glue. Hacky, I know, but desperate times called for desperate measures. Amazingly, though, it worked. Really well. The hot glue served as a float system, eating up the hundred or so microns of inevitable misalignment, and thus allowed the carriage to move freely and accurately.
The community moved away from rods in the first place because (I'm assuming) they were too flexible and they were prone to binding. Both these problems are fixable, however. If you use 12 mm rod, flexibility is not an issue on a machine the size of ours. If you use precision ground rod (available from McMaster) and SDP teflon filled acetal (quieter than sintered bronze, and going strong at 700+ hours) bearings, you have very smooth movement and no detectable slop. And if you use a float system whereby the bearings on one rod are fixed and the ones on the other are free to move a few hundred microns, the binding issue can be totally solved. The float system we use on our Y system uses a printed floated bearing holder (see the STL) that allows flexing in the X direction, but not the Z. On the X, we use the simpler expedient of putting the top bearings in short pieces of rubber tube. This does provide some Y flexibility, but gravity provides an effective preload that prevents this from being a problem, and this allows the printer to compensate for too-thin first layers.
Expect an update on the progress of dual extruder/soluble support project in the next few days. Our experiments have led us in a fairly different direction than the one we originally had in mind, but I think we've almost got it cracked.
Stay tuned! Team Isis3D