We have found the most common issues of hot-end performance are related to jamming and extreme force to extrude (back pressure). As hot-end developers, we wanted to gain a better understanding of what is actually happening inside hot-ends and the extrusion process. So much of what you see or read online may sound logical, but can actually be misleading or even false. We needed more in our repertoire of design tools and REAL, quantitative metrics to take us to the next level of hot-end design, and this is where we saw the need to develop an extrusion force test stand. A simple device that will extrude filament into a hot-end supported by a load cell and give real-time feedback for analysis.
In 2013, we used thermal simulations early on in the design process for the original Pico (Gen1) hot-end, which propelled us miles ahead of the competition. With the quantitative data from the Force Test Stand, we have brought ourselves galaxies ahead of the competition. The following is an explanation of how we did it and what we found—Even we were shocked by the results!.
After an initial test lineup of a select few hot-ends, the results show Pico Hybrid (PTFE & All-Metal) requires the least amount of force to extrude.
So, what does this mean?
Since the beginning, size and weight have always been important design goals for us (that's why we called it Pico!) And the small footprint of Pico (Gen1) has always been a point of excitement for consumers and printer manufacturers from a design standpoint. Smaller, lighter gantries can move faster and resonate less. Now, with Pico Hybrid's even smaller footprint, and reduced force to extrude, printers won't need large, torquey stepper motors or gear reducers. So, printer performance will increase again. We can't wait to see what you'll print.
The basic blueprint of the force test stand, built and coded by Branden Coates, is a standalone device that monitors the amount of force, in grams, that an extruder is exerting onto the filament while flowing through a hot-end at temperature. A live feedback load cell is placed in between the extruder and hot-end to accurately measure the force of extrusion; excluding any other variables such as slip from the drive mechanism, the tension of the drive mechanism and hysteresis from Bowden tubes.
The extruder motors and hot-ends are controlled through normal host software via a SmoothieBoard. A button on an Arduino Uno has been assigned to a macro that activates the pre-programmed test run. The Arduino also monitors readings from the load cells and displays them live through the serial monitor of a connected computer.
A series of three extruders exist on one stand to simultaneously test three hot-ends for side-by-side comparison. Preceding each test session, all load cells are calibrated with a precision calibration weight and a tare macro is executed between each session to account for the differences in weight for each hot-end.
The line up of hot-ends included Pico Hybrid and Pico Hybrid Pro, E3D V6, V6 Lite and Volcano, Hexagon AO hot-end and the J-head each in 1.75mm and 3.00mm setups (excluding 1.75mm Hexagon AO).
The filaments tested were provided by IC3D in Columbus, Ohio. Manufacturers of the First-Ever Open Source 3D printer filament.
Natural 1.75mm and 3.00mm PLA
Natural 1.75mm and 3.00mm ABS
Blue 1.75mm ABS
The initial tests, conducted by Coates, consisted of setting each hot-end at a fixed temperature, 200°C, for IC3D PLA and 230°C for IC3D ABS. Then extruding a set volume rate of 10mm^3/sec of filament through each hot-end once the temperature is reached. See graph. The approach of using a volumetric extrusion rate, instead of a set length and speed, is attributed to the ability to equally compare results across all hot-ends regardless of filament sizes of 1.75mm and 3.00mm. For comprehension purposes, 10mm^3/sec using 1.75mm filament at 249.5mm/min(at entry) is equivalent to using 3.00mm filament at 84.88mm/min(at entry). Each test was run 10 times. The graphed results show the average force across 10 runs for each hot-end.
The effects of varying printing temperatures vs. force to extrude:
Having experience in hot-end manufacturing, we understand the placement and construction of the temperature sensor affect true static and dynamic internal molten plastic temperatures. Other effects include melt chamber and heater block construction material. Creating a control for all types of scenarios between hot-end construction materials, infinite sensor positions, control board/temperature table variations and static or dynamic conditions would have required a next to impossible experiment. This is why we chose to abide by most practiced printing temperatures for each material tested: 200°C for PLA and 230°C for ABS.
These suggested printing temperatures are very popular, which is why we conducted the test based on the common practice. However, we did not want to ignore the effects of true internal molten plastic temperature and its relation to the force to extrude. To empirically understand how temperature affects force to extrude across select hot-ends, we conducted the same test but varied the printing temperatures of IC3D’s blue 1.75mm ABS in our Pico Hybrid, E3D’s V6 and the J-head (Set A). We began the test at 220°C, stepping up in increments of 10° until reaching 260°C. The test is identical to the ones above: 100mm of filament at 249.45 mm/min. See graph. You will notice the drop in force to extrude across all hot-ends, Pico Hybrid being the least amount of force required in all temperature tests. Each test was conducted twice. The graphed results show the average force across 2 runs for each hot-end.
Future Direction:The excitement level at our design center has spiked with the new data we have been pulling from the stand. We can use this data in more ways than we originally imagined. We hope to gain insight into dynamic characteristics of how hot-ends are reacting such as retraction, residual back pressure, laminar and inconsistent flow, static friction of extrusion, surface finishes of polished vs unpolished, nozzle geometry, cold-zone design and maximum flow capabilities due to heat flux rates attributed from different nozzle materials.
Beyond hot-ends, we believe this stand will be very useful for filament manufacturers to refine filament formulas for 3D printing and recommend more accurate printing temperatures based on filament color and specific hot-ends. We hope to continue expanding the research even further to slicer software enhancements, optimized extruder design, and more. Please leave a comment to let us know what you are more interested in!
In the near future, we have plans to expand the testing parameters to include variables such as the effects of the nozzle distance to the print surface and object, the effects of slip at the extruder drive mechanism, Bowden tube hysteresis effects and more. To accomplish this, we plan to outfit a normally operating printer with the live load cell setup to actively monitor force feedback during true printing.
Our promise: Staying True to You:In no way has this data been altered to favor one hot-end over the other. All tests have been done methodically and equally. It would seem obvious for us to skew the results in our favor, but we promise you, in the sake of open source and the betterment of RepRap 3D printing, this data is original. The only exception has been noted below for graphing purposes. If you do not believe us, we have made all the plans and code for the test stand open source. We welcome you to replicate it and observe for yourself. Thank you, from William M. Moujaes and Michael A. Koury.
Note: Single spikes due to noise in the load cell or wiring have been noticed during the logging of data across all hot-ends. We went through and removed these spikes from the data for graphing purposes. For example: for a given set of consecutive data logs, one instance of an unusually high spike, between 10 to 25 times greater than its surrounding values, would be recorded. Such spikes are physically impossible as no ramp-up or ramp-downs were evident in the preceding or following values. We treated these as anomalies and have removed them from the graphed data (we replaced these anomalies with an average of the two values before and after).
Note: All the original data with the spikes can be found here on Google Documents.