After DMC’s second Fed Ex Day, Eric highlighted our 3D Printer. In May, we had an opportunity to show it off at the Tres de Mayo party and I wanted to go into a little more detail about what we’ve learned since the last blog update.
As many tinkerers who have played around with 3D printing will tell you, configuring a machine to produce a high-quality print is part art and part science. Like the original DIY personal computers of the 1970s (we’re looking at you, Altair 8800), every unit in the field is unique in some way. As a result, there isn’t a one-size-fits-all configuration that can be distributed. On the upside, this presents a wonderful opportunity to learn how each “knob” affects the machine and its product. On the downside, there are about 200 knobs to tweak.
If you think back to Eric’s blog entry, one of our first successful builds was a hammer-and-sickle shaped cookie cutter (I bet you can’t guess which DMC employee inspired that). One of the reasons why this build was successful was that we printed a “raft” underneath the actual 3D object. When the plastic material is extruded through the nozzle, it’s heated to 225 °C in order to melt and flow properly. However, there is a risk of the plastic deforming after extrusion as it cools down. This is mitigated somewhat by heating the actual build platform (to about 120 °C), but the first few layers tend to deform anyway. On top of that, the first layer of plastic often times won’t “stick” to the platform, which results in a botched build.
One of the most common methods of dealing with this build irregularity is to very slowly extrude a thicker layer of plastic onto the build platform first and then build the model on top of this “raft” layer. This raft provides a much better surface for the subsequent layers to deposit and extruding a thick layer slowly will significantly reduce the risk of deformation.
This approach worked very well initially to mitigate the aforementioned issues, but one thing became readily apparent: Detaching the model from the raft was very difficult. Since the layers of the model melt into the raft, the raft often has to be cut away from the model, rather than snapped off. Boris, however, developed a very nifty way to improve the raft within the constraints of the build software.
Instead of printing a single raft layer, a second layer was added. Instead of extruding a solid, thick layer, this next deposit was thinner and was printed with an infill of about 20% (which means about 80% of the layer volume was air). In addition, each “rib” of the second raft layer was extruded orthogonally to the first layer. Looking down from above, a crosshatched pattern can be seen as shown below:
This new, improved raft is much easier to detach from the rest of the model, owing to the significant reduction in contact surface area (driven by the 20% infill crosshatch). At the same time, it still offers protection against deformation and allows the plastic to stick to the build surface.
Many 3D Printers and CNC machines accept build instructions in a format known as “G Code,” and the MakerBot Thing-O-Matic is no exception. A G code file, which is generated for each 3D print, will tell the build platform how to move its motors and control the onboard tools (extruders, heaters, etc.) in order to create the 3D model of interest.
One of the most popular G code generators is the Python-based Skeinforge. Like many generators, it is intended to work with a wide spectrum of devices and can thus support anything from CNC machines to 3D printers with multiple extruders. As one might expect from such a generic program, there are a lot of parameters that must be configured, and many that are not used on a specific type of machine. Configuring this manually is, needless to say, difficult.
Luckily, there is a program called ReplicatorG, endorsed by MakerBot Industries, that serves as a bridge between Skeinforge and a large number of Arduino-based printing platforms. ReplicatorG also has a handy 3D model viewer that can perform basic manipulations, such as scaling and rotation. Skeinforge presets are included for a few of the popular 3D printing models, including our Stepstruder Mk. 7. As I touched on earlier, these presets are simply estimates and aren’t always the best settings for each device. At DMC, we’ve had an opportunity to refine our settings to create better and better models.
One of the most critical parameters for any build is the feed rate of the extruder. As the plastic is heated beyond its glass transition temperature, it begins to flow. If the extruder feeds the plastic through the extruding tip too fast (at a given tip temperature), the plastic may not reach the appropriate transition temperature and will not bond with the rest of the 3D model. On the other hand, if the feed rate is too slow, the plastic will collect on the nozzle and degrade the resolution of the printed model. In both cases, the entire build may be compromised.
To verify the feed rate defined by Skeinforge, there’s a simple calibration routine that can be performed as described by this MakerBot blog. Another, easier calibration is to simply extrude plastic for a known period of time (using a stopwatch or smartphone), and then dividing the length of the extruded material by the extrusion time. If this number doesn’t match the parameter defined in Skeinforge (which is certainly possible, given the generic nature of Skeinforge), you can make a correction that should improve the quality of the final build.
Finally, there’s another tradeoff in the 3D printing process that we’ve become familiar with as we continue to build models. When Skeinforge / ReplicatorG generate GCode, they’re actually “slicing” the software-defined 3D model into layers. Roughly, the tool path of the extruder is generated for each layer, and the printer knows to raise the extruder tip by a certain amount before starting the next layer.
Naturally, the thickness of the layers during slicing affects the “resolution” of the print. Finely sliced models will better capture subtle changes in the surface of the model compared to a more coarsely sliced model. The tradeoff, however, is that the more finely a model is sliced, the longer it takes to generate the G Code and print. Since even small builds can take 20-30 minutes, this is a tradeoff that should be weighed on a case-by-case basis. Building a simple box, for example, won’t require thinner slices, as the faces don’t change significantly with height. Our model of Chichen Itza, on the other hand, turned out nicely with a fine layer slicing, despite its hour-long build time.
Check out some of our builds here below. If you have some experience with 3D printing, would like to share some tips of your own, or just want to share your comments about the post, please let us know in the comments!
Learn more about DMC's company culture.