Print 3D

Edit:10 nov. 2019, Cre:04 oct. 2015

'D-Box' 3D printer

Enclosed delta 3D printer ‘D-Box’

Enclosed with air recycling through activated carbon filter
The first version of this printer had a particularity: tower angles were not equal (120°) as on usual deltas.
It started with angles 135°, 90°, 135° but this have been modified to equilateral (120°), see below.
It started from the idea of a ‘square’ delta by Ryan Carlyle (90°,90°,180°), but following his comments of lack of stability, I choose more regular angle, then go back to the tradition.

It was an entirely new design done in 2015 and no parts from another printer were reused.
After some use and development, I reworked entirely the box around nearly identical mechanics and electronics, which became the Lily “big” (published on Github in 2016).

Really usable area diam 286mm x height 400mm
Overall size : Width 570 x Depth 420 x height 840 mm.
Box is made of 10mm thickness plywood but should have been done in melaminated chipboard 18mm, as this is safer in case of fire, lower cost, and easier to build with standardized width. Supplied plans are for melaminated chipboard.

Control board

Control board is the 32bit arm-processor new Duet (V0.8.5). There is one stepper driver more for a second extruder on this Duet version.
Printing is done from an SD card. Files are transfered to the SD-card via an ethernet link at an amazing speed (~around 1 MB/sec with present firmware versions). There is an incorporated web server in the board, which is used to control the printer through any browser. If there is Wifi on your network, you can control or check the printer through a smartphone or tablet.
I started it without difficulties as this is nearly the same board as used on the Fisher.

Carriages and columns

The carriages are rolling on ball bearings on galvanised steel angle. This is a bit more noisy than other system, but the rolling noise is not the dominating noise of this printer, which is more the belt/motor humming.
There is no endstop switches, the carriages are stalled on the roof at low current. This is noisy and not precise, but calibration will further correct the position. The angle used are 35mm x thickness 1.5mm and lack a bit torsional stifness.


The arms in aluminium are ended with 8mm balls, which articulate in printed cups on carriage and effector. They are maintained in contact with high tensile polyethylene wire (Dyneema/Spectra) and a spring. The aluminium does not give the lightest arms, but this is easy to supply in any DIY store.


The hotend is a Prometheus by Distech, which have the particularity to be fully adjustable. Its stainless steel nozzle is integral and does not need any re-torquing when hot. Fan is a bit noisy, which is limited by the enclosure. The hotend is delivered with 1m cables, which fits perfectly this setup. So, there is no local plugs, but this is not a problem with the magnetic effector mount.

Hotend positioning

The hotend is positioned on the effector with a kinematic coupling maintained in place with magnets. This system is also used as a calibration sensor.
This is very practical for hotend maintenance or exchange.
The idea is taken from the Fisher, but set on the effector instead of the bed, which makes the calibration much easier to adjust (there is only one offset) and is more efficient for mechanic protection as the inertia is lower.

A video of the hotend ‘jumping over’ a plastic strip of 0.85 mm thickness. You see on the right of the part that the realignment is perfect.


Belts are twisted to have their back rolling on the idlers (flanged bearings F623).


Motors are 0.9° Wantai motors 42BYGHM809 torque (also for extruder motor, for supply simplification)
They are installed with cork dampeners

Power supply

Power supply voltage is 24V, which allow higher speed and is required notably due to the use of 0.9° motors.
The heated bed will be installed on the mains, so the power supply will not be used for the heatbed. Its rating is 350W, which offer a huge margin, but the extra-cost over a 150W power supply was low and this kind of power supply is not of good quality so a good margin may help to improve reliability.
To allow the use of old 12V computer fans (recycling and board cooling fan), I installed an adjustable buck converter, set at 10V to reduce fan noise.

Direct drive extruder

Direct drive extruder is generally sufficient for 1.75mm filament and skip steps while having extrusion problem, which is better than grinding filament. The skipping noise warn you of problems.
Its simple design, with articulation being a plastic cylinder set in a printed socket is designed for dual extrusion.
The extruder is positioned to have the minimum ‘Bowden’ tube length, which is 640mm, short for this printer size.
Filament is maintained very close to the hobbed insert.
Pressure is adjustable and cleaning is easy.
Bowden is screwed in an ordinary M4 nut, which was chosen after tests of different pushfits.
As there is a pushfit on the hotend, you don’t need to disassemble the bowden tube from extruder frequently.
I am using a stepper (same as movement steppers), but it shall be preferable to use a 6.5 stepper.

Due to vibrations, extruder is installed on an antivibration support.

Wire guide with cleaning/oiling sponge and room for a filament detection switch.

Air recycling

An air recycling system flow through activated carbon for removal of pollutants and odors.

Activated carbon filters are easy to supply as they are used in kitchen hood fans.

Delta kinematics

The unequal angles is not optimal. The best stability of an effector is obtained with the largest arm space with the smallest offset. Having irregular angles limit the possible arm space and oblige to increase offset. The actual arm spacing value (60mm) is correct in comparison with similar deltas, because of the compacity of the articulation system, but could be improved.
Manual manipulation shows that the stability of the effector is irregular. The flexibility around y axis is more than twice that around X axis. While near the right column, the flexibility is much higher than near other columns. And that show in prints, where it is near perfect on left side and with irregularities on right side.
So, unequal angles have been abandoned and I modified the printer back to equilateral angles (120°), with a small usable diameter loss (280mm instead of 286mm), which allow an arm spacing of 84mm for an offset of 30mm and a much better stability.


directly on the bottom of the box !

Other prints


Before starting the machine, I was quite afraid to have resonances on the panel, as others experienced. Hopefully, this was not the case except for some retract speed on the extruder, which was easily solved.
The box significantly reduce the fan noise, but the overall noise is not reduced a lot and this machine is not very silent. The noise is relatively deep, so this is less aggressive than other printers.


The columns are prepared and drilled for assembly on bottom and top supports.
The supports are positioned on the bottom and top plate with templates.


After some use and the design of the first Lily printer, I decided to not publish the D-Box design “as is” and entirely reworked the box around the same mechanics. This became the Lily “big”
Simulation is made on OpenScad.
Stronger magnets (diam 6×6) have replaced initial diameter 5×5, as the hotend was too sensitive to cable and bowden tube tension which was troubling the calibration.
There was many attempts for the belt attach system and first design inspired from the Fisher have been since modified for a system a bit more complex, but reliable.
I was intending to have a dual extruder with a tool changer, possible easily with the magnetic system but the crossing of the bowden tube may not be easy to manage. Such system may need to have an extruder on each side of the printer, which will take a lot of space on a printer already quite large.

Credits :

- Ryan Carlyle, for the idea that a delta does not need to be equilateral, even if I abandoned later…
- Tim Jacobsen, for the rod system with wire tensioning system.
- RepRapPro company, for the idea of kinematic positioning system used as a sensor, but here I set it on the effector instead of the bed.

(c) Pierre ROUZEAU
Privacy - Vie privée - Imprimable - Rechercher
Page mise à jour le 10/11/2019 20:39