Tag Archives: IMEX

3D Printer: Multi Gantry Printer

by · published Technology , , ,

State: brain storming, early drafts

MG-MIEX (Multi Gantry & Multiple Independent Extruders) [early draft]

Updates:

  • 2024/12/20: early idea to add tool changer as well
  • 2024/12/16: adding more details with multiple extruders on same X gantry
  • 2024/12/06: early ideas about single layer segmenting
  • 2024/12/05: starting write-up

Introduction

While working on theoretical side of parallel 3D printing, I used the example of multi-gantry or Multi Gantry (MG) setup where a single bed is shared in Y – here a few rough sketches:

Multi Gantry with same order, Common Y Rails (CYR)
Multi Gantry with same order (closeup), Common Y Rails (CYR)
Multi Gantry with ability to change order, Dedicated Y Rails (DYR)
Multi Gantry with ability to change order (closeup), Dedicated Y Rails (DYR)

Gantry Details

There are several challenges, such as the Y- and the XZ motion system and making the thickness of each gantry as thin as possible to have as little “dead” or “unusable” space as it adds up and also limits the printable XY space.

Y motion system

We can share a single rail where each gantry rides hence Common Y Rails (CYR), then they can’t change the order, in that case the belt routing becomes the main issue of concern – whereas when each gantry has its own rail hence Dedicated Y Rails (DYR), and each gantry has a distinct height they can change the order (lift Z to the top and move over the lower one – hence each gantry has a different Z height, but with the filament and cabling the entanglement needs to be closely observed).

Single Rail, Multiple Gantries aka CYR
Dedicated Rails per Gantry aka DYR

XZ motion system

We require to have two positioning axes to integrate here, the X and Z:

  • two independent motion systems (X axis = 1 motor, Z axis = 1 motor)
  • Core XZ system also using two motors, but having XZ motion combined
Jon Schone’s (ProperPrinting) 2 Gantry System: X & Z axes with each their own motor, sharing a single Y rail

Slicing for Multi Gantry System

As I realized earlier, it makes no sense to statically segment Y-space for each gantry (e.g. Y-size/n) although it would make coding easier, but seamless printing would be impossible – in reality, we need slightly overlapping spaces to achieve zero space seams where the gantries print for the same piece, so we naturally choose flexible non-overlapping operations.

Layer Segmenting

Obviously one of the simple optimization is when we are looking at a single layer, and segment it area-wise to n-amount of gantries. For sake of consistency, a wall is assigned to a single gantry to have a seamless wall/perimeter.

Y Segmenting Single Layer for 4 gantries & nozzles (red, green, blue & yellow)
  • Infill areas are Y segmented
    • each area (e.g. A1-A4) is the same to distribute work equally
  • Walls/perimeters are assigned to a single gantry
    • if walls intersect each other in Y, those are printed sequentially not in parallel to avoid Y collision

The layer segmenting approach makes it rather easy so all nozzles have the same Z, thereby segmenting print jobs becomes easy as well.

Mixing Multi Gantry (MG) with IMEX (IDEX & ITEX etc)

We can also mix Multi Gantry (MG) with IDEX (Dual) or ITEX (Triple), or in general Independent Multi Extruders (IMEX), something like this:

Multi Gantry with IDEX: Independent Dual Extruders
per single XZ gantry
Layer Segmenting of Multi Gantry (4: red, green, blue, yellow) with IDEX (2) per Gantry

And like-wise segmenting layer areas in Y as before, and additionally segment in X for each printhead on Independent Multi Extruders (IMEX), in this example it’s IDEX (Dual).

As IMEX looks like scaling print parallelism further, one has to be aware of the spatial overhead for each printhead, e.g. a single printhead occupy ~40mm width in X, and similar in Y.

MG-MIEX 3×2

Early draft with more details based on Ashtar M, extended to MG, with Common Y Rails (CYR), with IDEX:

MG-MIEX 3×2
IDEX Detail

Challenges

  • 1) When doing a MG-MIEX 3×2 (3 gantry with 2 extruders each):
    • X: 3*2 motors = 6 motors
    • Y: 3 motors
    • Z: 3 motors
    • E: 3*2 motors = 6 motors
      we end up 6+3+3+6 = 18 motors to coordinate & control.
  • 2) Slim design of the gantry to reduce dead-space in Y.
  • 3) Slim printhead design to reduce dead-space in X.
  • 4) Slicer requires to coordinate 6 printheads (e.g. T0-T5) in non-colliding and efficient way.

Solutions

Slicer

A dedicated slicer is required which segments each layer into printable areas A1 to An whereas n is the amount of gantries, and those areas might be further segmented by m-amount of printheads on the same gantry, so they share the same Y position but can have distinct X position but not switch order or collide.

As proposed above, a single printhead of a gantry is then assigned to print within the same Y-segment the walls/perimeters to have a clean wall, hence, the infills are distributed among all the printheads.

Firmware & Controller

The controller with its firmware we have a few options:

  • Duet3D/RRF: Multiple motion systems
    • M400: wait for current moves to finish (when using T<n> with G1s together, applies to X, Y, Z, and also E)
    • M596 P<n> selects motion queue, prior each G0 or G1
    • M597: collision avoidance (at firmware level)
    • M598: sync up multiple motions (faster waits for the slower)
  • Klipper
  • Marlin
    • one controller per gantry, orchestrating between controller needed
    • simple setup, reliable

As the operation of the gantries are overlapping, they cannot have relative position, but rather have absolute Y position (relative position used to detect collision), the same for the X position.

Adding Tool Changer (TC)

In order to support a tool changer, the printheads need to reach a common position in order to deposit and pick up tools. In case of a single Y rail setup this seems at first sight challenging, only a dedicated tool change per gantry, at a particular a YZ position for example. The multi Y-rail setup allows any gantry to reach a common region in YZ, and so any printhead could deposit and pickup a tool from the common toolset.

References

Parallel 3D Printing / Additive Manufacturing – Part 1

by · published Technology , , , , , , , , ,

Updates:

  • 2024/12/09: ready for publication
  • 2024/12/04: completing first table on printheads & nozzles
  • 2024/07/30: starting write up

Introduction

When depositing material through a nozzle, the variables to compose a workpiece depends on the amount of nozzles and their operational spaces – let’s lay out the different methods which gives us the foundation to tackle then parallel procedures in the next part in the series.

Printheads, Nozzles & Operational Space

PrintheadsNozzles per PrintheadNozzle Size [mm]Layer Height [mm]MaterialOperational Space
Single nozzle FDM/FFF110.1-1.00.1-0.6Polymer (PLA, PETG, ABS)100%
Dual nozzle FDM/FFF aka IDEX210.1-1.00.1-0.6Polymer (PLA, PETG, ABS)2x 50%; horizontally separated
Duplex F2210.1-1.00.1-0.6Polymer (PLA, PETG, ABS)2x 50%; vertical separated
CM3P Dual Conical210.1-1.00.1-0.6Polymer (PLA, PETG, ABS)2x 50% of negative cone
Resin SLA (UV Laser)110.1500.050-0.150Resins100%
Resin MSLA (UV & LCD)150M-100M0.020-0.0400.050-0.150Resins100%
Quantica NovoJet1960.050Resins100%
Stratasys J55 PolyJet1192*)0.2*)0.18Resins100%
Selective Laser Sintering/Melting1+10.10.05-0.10Polymer or Metal Powder100%
  • Stratasys J55: nozzles & nozzle size based on J850 specs, J55 details specs seem not publicized (2024/12)

Print Base

A print base is where the nozzle can extrude on. For the first layer, there is the print bed, after the first layer the workpiece or support structure can be build upon. One can alternatively use a stabilizing medium like silicon and extrude in such liquid medium which operates as bed or foundation like Rapid Liquid Printing (RLP) does:

Extruding into supporting material
Extruding into supporting material

The extruded material just has to stay where it was put, either a solid bed or a medium which prevents it to float out of position, or as traditionally printed on a print bed or base, very similar does Xolography where the solidified resin stays put as well.

Massive Parallel Nozzles Printhead

Resin printing with a printhead may have hundreds or even thousands of nozzles, yet, they share the same operational space, but due the parallel setup the print speed multiplies direct with the amount of parallel nozzles on the same printhead.

Xaar 128 printhead printing high viscous material

Massive Parallel Nozzles: MSLA

As mentioned above, we can also view Masked Stereolithography (MSLA) resin printing as a massive parallel nozzle setup, where each pixel is either an active or inactive nozzle depositing a voxel.

Anycubic Mono M7 Max MSLA Setup

Separated vs Shared Operational Space

Disclosure: I have been contracted to work on the Duplex F2 software stack (2022-2024).

Let’s take a look at the Duplex F2 printer where space is separated vertically, or the CM3P dual conical printer where the cone space is separated horizontally, or the Multi Gantry 3D printer by Proper Printing (Jon Schone). We have two printheads which never collide due their separated operational space, the firmware is simple and path planning is simple, both heads pretty much can operate independently.

Duplex F2: Dual Delta
CM3P Dual Conical Printer
ProperPrinting (Jon Schone): Multi Gantry

When using more than two printheads it is beneficial to share the operational space, yet assume 6 or 8 printheads, each printhead needs rods to keep the printhead and position and orient the nozzle(s), so overlapping operational space requires extreme well planned tool paths avoiding any collision of the printheads.

Regular Operation Space Separation

We can segment or separate the space evenly or according the reach of the printheads, and each separated space can be printed without colliding. Yet, in reality the printheads mounts limit that operational space into slightly smaller spaces, but ideally:

nvolumes = volumetotal / volumeprinthead

If the individual printhead volumes aren’t regular, then we end up with arbitrary amount of printheads to cover a given print volume:

volumetotal = sum( volume1..n )

In reality, we require (slightly) overlapping operation to get seamless operation, so the “regular operation space separation” is only theoretically, but not practically.

Overlapping Operational Spaces

When the printheads can reach each other operational space, they become overlapping and controlling tool path generation needs to take care no collision is occurring (same place at the same time).

The Static Non-Overlapping Operation has static defined operational spaces where the operators can function – it’s quite obvious such solution is impractical, as in real life there would be space which cannot be reached, a kind of blind seam not reachable by either operator.

The Flexible Non-Overlapping Operation is flexible defined operational spaces, in the illustration above those spaces are co-dependent.

The Static Overlapping Operation is when those operational spaces are overlapping, yet, prefixed or static operational spaces.

The Flexible Overlapping Operation is flexible operational spaces, yet due the nature of the setup these operators cannot occupy the same space at the same time which would result in physical collisions.

Now, the last part of the last sentence may sound obvious to even mention, but bear in mind you can have two projectors shining light into a resin bath, and expose and solidify a 3D model, then these two lights acting as operators indeed occupy the same space at the same time as part of their function. So, the operators functioning with light can occupy the same space at the same time, whereas solid operators, such as robotic arms, cannot.

Print Speed in Parallel Setups

total print speed [mm3/s] = nprintheads * nnozzles * vextrusion [mm3/s] * parallelfactor

whereas parallelfactor is 1.0 if printheads can print parallel, or is less if the operational space is overlapping and preventing printheads to operate parallel thereby.

~ * ~

“Parallel 3D Printing / Additive Manufacturing” Part 2 follows later (will be linked when published)

References