As we further explore Autodesk's Ember SLA DLP printer, we continue to learn more about the characteristics of the machine and the process. Consideration of these findings leads to higher quality and more successful prints. Over the course of much printing with Ember, our team realized that viscous forces in the resin can have a significant impact on the accuracy of the prints, particularly with respect to dimensions in the z (vertical) direction.
In a nutshell, as the build head approaches the resin tray and PDMS window for the first layer exposure, the resin tray tends to deflect downwards as the resin flows out from between the build surface and the PDMS. By default, Ember starts exposing the first layer while the system is still equilibrating. This causes a thicker than intended first layer, which leads to the errors in the dimensions of features in the z direction.
We rectify this issue by adding a delay after the build head has approached the PDMS but before exposure begins. Currently, opening the door while the build head is traveling to the start position enables this delay with the current firmware. However, we plan to release new firmware that respects the "delay before approach" setting any time the build head approaches the PDMS window.
With this in mind, I revisited my previous attempts (see https://www.instructables.com/id/3D-Printing-Threaded-Features-with-Autodesk-Ember-/) to print threaded fasteners with Ember. Additionally I explored the impact of printing the features at various orientations.
I used the same process as my previous attempts to prepare the models for printing. I downloaded models of hex nuts ranging in size from 1/4-20 to 4-40 from the McMaster-Carr website. These were imported into Inventor and arranged into Ember's build volume.
To evaluate variations in feature resolution due to orientation, I printed four instances of each part. One with the thread axis perpendicular to the build surface (horizontal) and three with the thread axis parallel to the build surface but at 0 (vertical II), 45 (vertical III), and 90 (vertical I) degrees relative to the x-axis. I've annotated a picture of the 3D model to label the orientations.
I added a 3 second delay between approach of the build head and start of exposure via the "First Layer Wait (After Approach)", "Burn In Layer Wait (After Approach)", and "Model Layer Wait (After Approach)" advanced settings. Due to limitations in the current firmware, there is no delay before the first exposure, even with these settings set to non-zero values. To add the crucial delay before the first exposure, I opened the door as the build head was moving to the start position, waited approximately 10 seconds after the build head came to rest, and then closed the door to allow the first exposure to begin. As I mentioned before, we are currently working on an revision to the firmware that will cause the desired delay to be present any time the build head approaches the PDMS window. Keep an eye out for available firmware upgrades at emberprinter.com.
In comparison to my previous attempts to print threaded fasteners, I got favorable results using the modified printing techniques. I extensively evaluated the 1/4-20, 10-32, and 4-40 nuts with bolts printer on the Ember and off the shelf, steel versions. Although the fit was initially somewhat snug, repeated threading resulted in the ability to successfully mate the 4-40 and 10-32 printed nuts and bolts and the 1/4-20 printed nut with an off the shelf metal bolt.
I did notice variations in fit between the orientations. In general, the nuts printed in vertical orientation III seemed to have the best fit, followed by vertical orientation II, horizontal orientation, and vertical orientation I. At this point I do not have enough data to confirm that the orientation does indeed have a signification effect or if the observed differences are due to random variation. Suggested future work includes determining if there is an optimal orientation of the thread grooves with respect to the projector's vertical and horizontal directions.
I would also like to try modifying the 3D models in an attempt to achieve a better fit with the parts as printed. It is often necessary to make adjustments in tolerances to accommodate the manufacturing process and it is impressive that the unmodified models from McMaster printed as well as they did.
In conclusion, I have shown that it is possible to leverage the dimensional accuracy of Ember to print mechanically functional parts with surprisingly little effort. Additionally, the discovery of the necessity of a delay before exposure suggests that there is still much to be learned about Ember and the DLP SLA process. Happy exploring!