Introduction: How to Configure Ember for High Speed 3D Printing

One of the principal advantages of 3D printing is that it can be used to manufacture parts that cannot be made using any other technique giving the designer great freedom and permitting them to produce highly optimized parts. A typical example would be a strut optimized for minimum weight while maintaining adequate strength for the application.

Despite this advantage, one of the factors holding back the adoption of 3D printing in manufacturing, is speed. The output of today's 3D printers (across all technologies) is much slower than that of other manufacturing processes such as CNC milling, injection molding or forging. As a result, the cost to manufacture 3D printed parts is prohibitive and often outweighs any benefit from the optimized part (there are some exceptions, such as dental and hearing aid industries, where 3D printers have replaced manual labor and thus led to significant cost savings). If the speed of 3D printing increases, then it can be transformed into a viable manufacturing technique and open up a host of opportunities.

In this Instructable, we're going to look at how to increase the speed of a Digital Light Processing Stereolithography (DLP SLA) 3D printer, specifically the Autodesk Ember 3D Printer. The techniques that we describe here apply to the whole class of DLP SLA printers and can be replicated on many different systems.

The Ember printer is open which means it can be easily used to explore the limits of DLP SLA 3D printing. Through optimization of the printer settings, software, and material (without hardware modifications) it is possible to increase the standard print speed of Ember from 18mm/hour to 440mm/hour an increase by a factor of 24 for a particular class of geometries.

So why is this important and why as a software company is Autodesk conducting this research?

  1. By only tweaking a few things in the machine settings, software and material you can print high speed on a system that's readily available.
  2. We want to continue advancing the state of additive manufacturing and we expect the best advances in manufacturing processes to come from approaches that combine hardware, materials, and software.

This research is the first step towards realising high speed 3D printing in a production environment.There are unique design rules that apply for high-speed DLP SLA that are beyond the capabilities of the current generation of design software. By researching in this field, our goal is to drive the additive manufacturing industry forward by developing a connected ecosystem that can provide designers and manufacturers the software they need unlock this class of technology.

We also want to demonstrate the power of an open approach to technology. If Ember were a closed system, then researchers would be unable to explore the limits of additive manufacturing. With Ember, we have created a powerful research platform that gives scientists, engineers and designers the opportunity to explore the future of additive manufacturing.

If this sounds interesting, read step 1 to learn about the science behind Ember. If you're already familiar with how DLP SLA works skip ahead to step 2 to learn how to configure Ember for high speed.

Step 1: Science of Ember

Ember is a DLPStereolithography 3D printer. It uses a digital projector to expose, cure and harden photosensitive liquid plastic (we call it resin) into a solid part.

This is how the process works:

  1. The 3D model is sliced into cross-sectional layers. Each layer is saved as an image and transferred to the printer
  2. The projector exposes the resin and it solidifies in the shape of the image. The first layer sticks onto the build head and then subsequent layers stick onto the layer above.
  3. The build head lifts up and then the next layer is printed, this process repeats until the part is finished

You may have noticed in the GIF above, that after each exposure the resin tray rotates back and forth 60 degrees, lets look at this in more detail.

As you expose and create each layer the hardened resin acts as glue, binding the build head to the optical window in the resin tray. The resins that are used in Ember are acrylates and methacrylates photopolymers that cure through a free radical photopolymerization process. To prevent the printed layer binding to the optical window we coat the window with a thin layer of Polydimethylsiloxane (PDMS), which is an oxygen rich silicon rubber. Free radical polymerization is inhibited by the presence of oxygen thus the oxygen in the PDMS prevents a very thin layer of resin, around 5 microns thick, from curing at the surface of the PDMS. This means that the printed layer is not adhered to optical window.

With thin, uncured layers of resin, there would be enormous suction forces exerted on the printed layer if you were to lift up the build head directly. These suction forces are inversely proportional to the thickness of uncured resin, in other words, the thicker the uncured layer of resin the lower the separation force. The suction forces are also proportional to the surface area of the part, the larger the part, the greater the forces.

To take advantage of this in Ember we use a shear separation mechanism. The resin tray rotates 60 degrees until the build head is no longer above the optical window with the uncured resin layer acting as lubrication and minimizing the shear force. After the rotation, the build head is directly above a channel that is deeper than the optical window. At this point, there are over 1000 microns of resin between the printed layer and the bottom of the resin tray, this means the suction force is reduced by a factor of 200 and thus becomes negligible, and you can lift up the build head with a minimal suction force exerted on the printed part. The tray rotates back 60 degrees and then next layer is printed.

We call this process Minimal Force Mechanics, and it allows Ember to reliably, produce parts with incredible detail, like the peacock feather above. BUT it takes around 2-3s per layer and thus represents about 50% of the print time and limits the print speed at 25-micron layers to 18 mm/hour.

If you're interested in learning more about the Ember mechanics, you can download the mechanical CAD and it explore it, the Ember CAD is shared under a Creative Commons Attribution-ShareAlike license.

I'm now going to show you that by optimizing the software and materials you can eliminate this separation step and print at 440mm/hour.

Step 2: Printing at 440mm/hour

440mm/hour is 24 times greater than Ember's typically printer speed and we achieve this through optimization of three things:

  1. Material - we've designed a resin that cures quicker and at thicker layers
  2. Process - we've changed the print process by eliminating the separation step and printing at 250micron layers
  3. Geometry - we've chosen a lattice structure that reduces the surface area per layer


First, we need to prepare a variation on our PR48 resin that will cure quicker and to a deeper depth. We call this resin PR48-High-Speed and the formulation is listed below.

  • Oligomer: Allnex Ebecryl 8210 39.8238%, Sartomer SR 494 39.8238%
  • Photoinitiator: Esstech TPO+ (2,4,6-Trimethylbenzoyl-diphenylphosphineoxide) 0.4005%
  • Reactive diluent: Rahn Genomer 1122 19.9119%
  • UV blocker: Mayzo OB+ (2,2’-(2,5-thiophenediyl)bis(5-tertbutylbenzoxazole)) 0.0400%

The UV blocker concentration in PR48-High-Speed has been reduced by a factor of 4 compared to PR48 to allow it to cure quicker and to a deeper depth.

If you want to learn more about how to tune your own resins from Ember, check out this Instructable.

Printer Configuration:

Next we need to configure the printer settings on Ember, you can do this either through or by SSH into the printer and editing the file /var/smith/config/settings.

On Mac using terminal you can SSH into the printer with follow commands (remember to change the IP address if not connecting over USB)

ssh -l root

Navigate to settings file and edit it

cd /var/smith/config
nano settings

Edit the following settings:

  • "ImageScaleFactor" : 1.0,
  • "DetectJams" : 0,

Next measure the irradiance output of Ember with a fresh clean resin tray (I recommend using a G&R UV Light Meter Model 220 with a 420nm probe or an ILT 1400 with SLE005/U detector) and configure the "ProjectorLEDCurrent" so that the output is 20 mW/cm^2

If you have edited the print settings file over SSH then remember to enter the following command to make the changes take effect

echo refresh > /tmp/CommandPipe

Print Job Setup:

Now that the material and printer are setup its time to prepare the print job. Open Print Studio and import the model 12-15-14_full_rigid_lattice.stl that is attached. For help on how to use Print Studio refer to this user guide.

Now create a new custom material, you can start this by duplicating the Autodesk CYMK 25 micron profile. Configure the profile as per the screenshots above. The main settings changes are

Burn in Layers

  • Number of layers: 10
  • Wait (before exposure): 0.5 s
  • Exposure time: 3 s
  • Separation Slide Velocity: 20 RPM
  • Approach Slide Velocity: 20 RPM

Model layer:

  • Wait (before exposure): 0 s
  • Exposure time: 1.2 s
  • Z-axis overlift: 0.25 mm
  • Separation Z-axis Velocity: 5 mm/s
  • Angle of rotation: 0 degrees

In the object browser, turn off the automatic support generation then slice and send the job to your printer.

I've also attached the job file to the instructable in case you can't be bothered with the above!


Follow the Pre-Print Checklist then sit back and watch your Ember print at 440mm/hour.

Step 3: Explanations, Limitations and Future Work

So that was pretty cool! Lets look at why the optimizations worked, the limitations of the system, what that means in practice and how it could be improved on in the future.


Direct pull (printing without separation) worked in this instance principally because we used software to optimize the geometry and material.

You'll notice from the graph above that the lattice structure that the global surface area (the sum of all the white pixels in a given slice) never exceeds 15% of the slice. The global surface area must remain below 15% so that the suction forces, which remember are proportional to surface area, do not become greater than the strength of the cured resin, the tear strength of the PDMS window and the normal force that the linear drive and motor can deliver. If the suction forces exceed any of these then the failure modes are as follow:

  • Suction force > strength of cured resin: the printed object is pulled apart
  • Suction force > tear strength of the PDMS: the PDMS so torn apart
  • Suction force > normal force delivered by the linear drive and motor: the z-axis jams

You can see from the graph and the video at the top of this step that the geometry changes rapidly from layer to layer showing that fluid can easily flow into the areas that need to cure. If we were to print a vertical column, then after a few layers all the fluid between the part and the PDMS would be used up and it would be difficult to get more fluid into the curing area.

We also optimized the material to make it cure quicker and to a deeper depth by reducing the about of photo-inhibitor, this allowed us to print deeper layers. Technically, you could call this out, because printing at 250 micron layers is 10 times faster than 25 micron layers. But with the optimization of the geometry and process, we were able to make Ember print 24 times faster.


There are four principal limitations to the geometry that you can print

  1. Global surface area
  2. Local surface area: The surface area of individual parts of the slice. For example, a strut in the lattice.
  3. Rate of change of position of local surface area: How the position of local surface area changes from layer to layer
  4. Strength of the cured material

Global Surface Area:

The suction forces generated by the global surface area of the part must not exceed the normal separation force of the system.

Local Surface Area:

The maximum length of the center of each local surface area to the boundary should be less than maximum distance that a fluid particle could move from the boundary to the center at a given print speed and resin viscosity. Essentially, if the local surface area of a strut is too big, then resin will not be able to reach the center.

Rate of Change of local surface area:

The rate of change of position local surface area should be such that no pixels are exposed in X consecutive layers.

Strength of the cured material:

At a certain speed, the normal forces will become greater than the strength of the cured material causing the printed part to pull itself apart.

Future Work

So how could you make a faster system?

  1. Make it stiffer: the z-axis, the resin tray, the optical window and the resin
  2. Make the inhibition layer thicker
  3. Make the resin cure quicker and lower viscosity

Make it stiffer:

The stiffer the system, the quicker you can pull and the faster you will print. Every component of the system will need to be stiff enough to withstand the suction forces; this includes the cured resin, the optical window, and the Z-axis. But be careful, if you make the resin too stiff and strong, then it will become difficult to remove from the build head and remove any supports.

Make the inhibition layer thicker:

At 5 microns the inhibition layer just isn't that thick. If you could get the inhibition layer up to 500-1000microns thick, then the suction forces would be negligible, the holy grail, but more challenging than it seems.

Make the resin cure quicker and lower viscosity:

A lower viscosity resin that cures in milliseconds would increase print speed but would not overcome the limitations outlined above.

What do these limitations mean in practice?

For a start, you can't print standard DLP SLA parts like dental restorations, hearing aids or rings. Even thin walled parts like ear shells and dental copings have too much surface area per layer to work (at least on Ember). We have found that all the parts printed using this technique need to be thin strutted lattices.

The Spark team have developed a tool to allow you to create lattice structures from solid models. For example, if we take the ubiquitous Stanford Bunny we can create a lattice representation and then use Print Studio to slice it for Ember, but it 's hard to control the end product using this technique. For example, if you download the bunny models you'll see that some parts of the lattice in the ears are not connected to the main body. To successfully design for high-speed DLP, you need design software that understands the process, the hardware and materials.

At Autodesk, we're

researching, building and testing solutions that will change the future of making. In the future, you may not sit down at a workstation and sketch, extrude and form a part. You could be using a generative design tool like Dreamcatcher, where you input a set of high-level goals including how you want to manufacture the product and the computer iterates through thousands of designs options until it finds one that meets all your goals. The output would be a functional part that is optimized for high-speed DLP.

The key to unlocking high-speed DLP as a manufacturing process isn't just new hardware or materials but, in fact, rests on developing new design software that can take full advantage of the capabilities on offer. That’s why we're building a connected ecosystem of hardware, software and materials so we can deliver production ready additive manufacturing workflows.