Coming up with a 3D printing performance metric

[Reece Arnott]

Carrying on from my thoughts a couple of weeks ago about my guess of the
exponential improvement of 3D printing cost/performance, I want to come
up with a performance metric so that you can actually compare different
3D printers and see which is 'better' for a certain price.

I have come up with one which I call "voxels per second" (vps) but I'm
unsure of whether it encompasses everything relevant and there are
already a few issues with measuring it on my Makerbot.

I am defining a voxel to be the smallest volume that can be printed by
the printer so once you have that volume, you can calculate how much
volume can be printed out in a certain time period (I'm using seconds)
and divide it by the voxel volume. I wanted to be able to compare big
concrete printers with Repraps and commercial high precision printers so
this concept of counting voxels allows this although I'm not sure we
don't also want to mention voxel volume explicitly as well.

With Digital Light Processing (DLP) projector based photo-initiated
polymer resin printers this will be simple: the resolution of projector
gives an area of the smallest pixel multiplied by a layer height giving
a voxel volume. As the entire layer can be cured at once you don't
actually need this layer height figure, you just calculate how long a
single layer takes and divide it by the resolution of the projector.

For example the resin printer described in
http://www.kickstarter.com/projects/b9creations/b9creator-a-high-resolution-3d-printer
says it maxes out at 20mm height per hour with layer height of
100microns (i.e. 0.05555 layers per second) and the resolution of the
DLP projector is 1024x768 so it can do a maximum of 43690.666vps (the
other figure of 12mm/hr gives 26214.4vps).

With powder based printers the voxel is the average volume of the
powder. If the powder is glued together then you probably want to add
in the volume of glue to the voxel volume. In the case of metal powder
that is melted by focused lasers then the voxel volume is the sphere
around the focal point that is above the melting point.

With the Reprap/Makerbot extrusion techniques I've got myself a bit more
confused, probably as I'm overthinking it.

If you assume the extrusion hole size doesn't change and the filament
feed rate is also constant then the voxels per second is simply based on
the speed and accuracy of the X/Y/Z motors:

For example, from the Makerbot wiki:
X/Y Positioning Resolution: 0.085mm
Z Positioning Resolution: 3.125 microns
Maximum X/Y Feedrate: 5000mm/minute
Maximum Z Feedrate: 200mm/minute

which means it takes approximately 1/1000 of a second to move the
resolution for each of the axes to give approximately 1000 vps.

However, it is normally suggested you don't run the Makerbot this fast
and I am running mine at 29.7-30mm/s to give approximately 350 vps.

If we want to take into account the accuracy of start/stop for the
extruder motor this goes down even further. From wikipedia:
"Early extruders for the RepRap used a geared DC motor driving a screw
pressed tightly against plastic filament feedstock, forcing it past a
heated melting chamber and through a narrow extrusion nozzle. However,
due to their large inertia, DC motors cannot quickly start or stop, and
are therefore difficult to control with precision. Therefore, more
recent extruders use stepper motors (sometimes geared) to drive the
filament, pinching the filament between a splined or knurled shaft and a
ball bearing"

My Makerbot uses one of these geared DC motors and I have done some
tests and I don't think it can really do a start/stop cycle of less than
0.1 seconds and if we take that as limiting factor means a maximum of 10vps.

Finally to compare apples with apples you have to multiply or divide
this voxels per second performance to give a theoretical performance per
$US1000 (or any amount really - my Makerbot was approximately $US1000 so
I thought it was a good round number to use).

This means that the DLP resin printer I linked to is a much better deal
than my Makerbot, for the Kickstarter backers at least, as you could get
it for a pledge of $US3775 which gives it a range of 6944.2-11573.68vps
per $US1000 compared to my Makerbot which is something less than 1000vps
(I'm leaning towards 350 as being most representative of what I want to
measure).


Does this make sense? Have I missed anything important? Which measure
seems a better performance metric for my Makerbot: <10vps ,350vps, or
1000vps?


--
-------------------------
Reece Arnott
Dunedin
New Zealand
-------------------------