Dual drive geared extruder
The Flow Rate of an extruder is one of the main issues to reach high print speeds. Trying to push the limits of flow rate in direct-drive extruders leads to motor stall or filament smashing. A geared extruder, as Wade's, increases the torque available to the hobbed drive. The motor is less prone to stall, but smashing filament is still an issue. With two hobbed drivers pushing the filament, the force is distributed over larger area, so I started to try some projects that use dual drivers extrusion. Up to now I found two designs of double-drive extruders that work nicely, but I decided to develop another one with minor changes. This project describes a geared dual drive made with saved gears from an old laser printer.
 shows a large assortment of dual drive extruder designs. Those extruders have one hobbed drive connected to the motor shaft and other hobbed drive conducted by a spring loaded lever, as shown in the figure below. Two gears joined to the drivers are responsible for the dual drive characteristic: the filament is griped in two sides at same time.
But there are some objections to this design (and to the designs of direct drive extruders as well). The lever drive press the filament against the motor drive and this causes a radial force applied to the motor shaft. Motors in general are projected to exert only torque and do not cope well with radial forces. The company Applied Motion Products have an excellent documentation with lots of specs about the subject. In this document it is recommended that the maximum radial force applied to a Nema 17 motor shaft ranges from 20 to 30 N. A spring used in an extruder's lever can easily exert a force of 10 N. The distances between the spring, pivot and drive can result in a gain of momentum that generates forces well above 30 N over the motor shaft. Considering that the specifications given by Applied Motion Products are for good quality motors, cheap motors could be in real danger. Some issues about this problem have been reported in RepRap forum . A planetary gear box should be able to eliminate the radial force over motor shaft and increase the extrusion flow rate, but with that, things start to became expansive.
There is another objection to this design: The distance between gears depends on the filament hardness. For a soft ABS filament the driver's teeth could smash the filament so that the gears touch each other to the limit and avoid the driver's teeth to penetrate further. For a hard PLA filament the distance between the gears could be large enough to cause backlash. As I like gears touching in its primitive diameter or even a little bit tighter, I preferred to look for another solution.
Ryan Huang - yyh1002 - shows a Belted Dual Drive  that circumvents the gear distance problem described. The two drivers and the motor shaft are interconnected by a timing belt. The figure bellow shows the mechanical aspect of this design. The spring press the two bearings against the filament. The radial forces are applied to shafts suported by bearings and not to the motor shaft. But there is still a radial force over the motor shaft caused by the timing belt tension.
What annoys me in this design is that the filament is griped and smashed in just one side. After passing through the upper drive, the second drive will encounter an already ground filament.
Instead of using a belt, I wondered how I could user some gears savaged from old laser printers to make a geared dual drive and the main idea is shown is the figure bellow. In this design the filament is pressed in opposite sides by bearings supported by a lever. The drivers are coupled to gears that make them to turn in opposite directions. I used module 0.8 gears. A 19 teeth gear is attached to the motor shaft. Two 55 teeth gears are attached to 5 mm axis and coupled to MK8 hobbed drivers made from scratch as well .
The main advantages of this design are:
The gears distance does not vary. They are tightly coupled to have minimum backlash. The filament is griped in two opposite sides. Radial force on motor shaft is minimized.
The mounting sequence of the extruder is shown in sections ahead. All CAD files, images and STEP files can be downloaded from: .
The rear block of the extruder was printed in PLA. The first steps consist on inserting captive M3 nuts, attaching the Nema 17 motor and insert the bearings for the gears' axis.
The next three figures show the insertion of the 55 teeth gears and mounting of front pressing levers made in PLA.
The levers are attached to the front block (PLA as well). At the bottom there is an aluminium bracket to hold the hotend. Figure 9 shows the front and read block being joined. It may be easier to hold the lower hobbed driver in place before joining the front and rear blocks.
To finish, fix drivers' grub screws and mount the pressing spring. The complete set is shown in figure 12
Figures 13 to 16 show orthogonal views of the assembly.
The overall assembly is bulky of course, but no problem if its used as a Bowden extruder. I am working now in a more compact version. Anyway, I installed the extruder in my 3D printer made from scrap parts (I love to make things from scratch) just to make some tests. With ABS and hotend set to 220C and a 0.4 mm nozzle the new extruder yielded a flow rate of 25 mm3/s, a good result if you consider that I used a stepper motor model 42HS03 (holding torque = 0.45 Nm, current 0.7 A, inductance = 16 mH). When I tried to increase flow rate beyond 25 mm3/s the motor stalled, but there were no slipping or grinding of filament. I suppose that with a more powerful motor I could obtain higher flow rates.
With such a powerful extruder I could print 0.2 mm layers with feed rate greater than 200 mm/s ! Obviously my poor printer made from scrap parts could not resist. The X and Y motors were saved from old table scanners and have resistance and inductance (8 ohms and 7 mH) improper for high torque and high speed required. After checking the new extruder with the default calibrating parts (cubes etc) I made a print of a 3D model of Auguste Rodin - L'Éternelle Idole. The video below shows the result I got.