Markforged Onyx / NylonX applications

We received our Markforged Onyx One a week ago, and have it set up. From initial testing, it seems like it will live up to the promise of strong parts, reliability, and efficient workflow. We have had an issue with gaps between adjacent thin walls (see attached image) that are not there in the underlying model, so any suggestions (i.e. print/material settings) regarding how to avoid that are appreciated.

I looked around for information on inappropriate applications of the material, and I have not found too much (I’ve mainly found raves about applications that have worked), so I’m looking for advice about what NOT to use Onyx parts for.

One thing I want to ask about specifically is gears. There is an interest on the team in using Onyx for custom designs involving ring, bevel, herringbone, etc. gears. I think the material seems strong enough, but I’m not sure it is low-friction enough to not overheat. My intuition is Onyx gears may be OK at lower speeds but not higher. Pure Nylon seems slicker & probably more appropriate for gears spinning at higher speeds. Has anyone tested Onyx with this kind of application?

We also just got our Markforged recently so I cannot speak to experiences with it just yet.

I can say we used some printed SLS nylon gears last year on our swerve module. These were used for the steering of the modules, and for the encoder reading the steering position. The gears had no issues with this. That said I would not use them on a high speed application like the drive portion. The rpm rate is so much higher of those shafts and the gears would see a lot more wear. I suppose you never know u til you try it and I doing anyone has, but I don’t think it is worth the risk, and the small weight savings isn’t worth it, especially when there are so many aluminum gear cots options and printing gears would take much more time.

I would also love to hear from some other teams here about what is good and not good that they have found so far with the Markforged. Specifically I’d like to hear about if anyone has played around with the Turbo Build option so far, as well as when it is a good time to use the he or rectangle fill pattern.

Like most high performance things, there is a learning curve to the MarkForged printers and their proprietary slicing software. The double wall issue is one of those things. If you look closely in the Eiger slice files, you’ll see the separation --it shows up as a gap larger than the wall layer and is unfilled. There are various ways to mitigate it. The easiest is to reduce the number of wall layers to allow more room for fill. If you’re already down to a minimum for your application, you’ll likely need to increase the wall thickness in your CAD. Generally, you’ll need to change the wall thickness by an amount equal to about 1-2 wall layer thicknesses in order to alleviate the problem. In some cases, the opposite is true: a reduction in wall thickness will get the wall layers close enough that they bond with no fill. You’ll need to experiment to determine some design rules for wall thickness.

On the gears, we (2767) has only used printed gears for low speed application. They have worked very well. As far as friction goes, I can tell you that when tapping Onyx, the friction is so high that unless you’re very careful, the tap will get hot enough to melt the Onyx. For deeper tapped holes, we make sure we use sharp, high quality taps and dip the tap in water periodically to cool it down. If this is done, Onyx taps beautifully. I bring this up, because it does tend to validate your concern with heat build up due to friction. I’d approach high speed gears carefully. In addition to the heat issue you identified, I’d also point out that the double wall issue can rear its ugly head with gear teeth as well. Make sure you pay attention that you have good fill in the teeth, particularly down by the roots.

A few tips for the MarkForged in general: Make sure you periodically check the material and fiber routing paths for friction. Over time the tubing wears and things can bind up. This manifests itself as a clicking noise in the feed and leads to several print issues. Also, learn how to use the print bed level check test print utility and periodically check the bed level and spacing. When the gap gets off, bad things happen. Finally, recognize when to use brim and try to avoid tall support material that isn’t well contained by the part. Otherwise you’ll fight dislocations. Overall, we love the MarkForged printers and use the heck out of ours. You just need to get through the learning curve.

1 Like

Doug is pretty spot on with how to mitigate it. When using our X7, we find it much easier to decrease wall layers and help rigidity with continuous fiber. With the onyx only stuff, you can still decrease wall #'s because your use case in FRC probably doesn’t need to standup to what industry does.

Unless it is something like a 5/16 bolt, we are using thread inserts from Spirol. We use a therma-set machine but a soldering iron and carful hand will work for this.

You will love your onyx. Every FRC team should have one.


Whats been said above is the specific advice I’d give to solve the issue you’re seeing with unfilled walls. Basically mess with a combination of wall thicknesses + # of wall layers until you’re happy with the fill you’re seeing.

We’re constantly tinkering with Eiger and parts to develop some best practices. We’re often optimizing for print time, but structural integrity and strength is also something we play with a lot.

Our standard printing process for years has been to hog out parts to reduce wasted plastic and print time. See below for an example of a part affectionately labeled “Intake Module Pulley - 3mm - 68th - jenny craig”.

We’ve found over time that Markforged is trying to help us and the algorithm it uses actually is faster sometimes if we don’t mass reduce the part. Theres several factors to that: # of wall layers, wall thicknesses, infill %, # of layers, etc. You can see example below. The part on the right actually has a decent amount more material (volume) but gets finished 5 minutes faster. We’ve found other ones like this with large discrepancies that significantly reduce time.

And the parts themselves assembled, just because its an interesting part:

We have a Mark 2 and lately have been playing with the continuous fiber settings as well. We’re particularly curious about fibers in things like gear, sprocket and belt teeth. We’ve been looking at what minimum size pitch is required to get fibers into the teeth.

Overall its an awesome machine. As with all high end machines, you need to work it and become one with it.


1 Like

Can you give us more information on the thread inserts? Which ones do you use? How do you design parts for them?



Check this out for threaded inserts

The biggest issue to deal with is always the quality of layer adhesion achieved during a print. Most parts printed on a MarkForged (not unlike any 3D printer) that I have seen fail did so along a layer plane. We have decided not to print gears because all those we needed are standard COTS parts and, as mentionned the weight savings is only minimal for those gears.

We considered printing some parts of a Swerve Drive we are developing on a MarkForged printer using Kevlar reinforced nylon fiber. Our parts geometry, however, required that we use the larger X7 printer, and none were available in our area to perform a prototype print on.

Considering the robot will not be subjected to a very long life or tremendous shocks for many of the parts, we are investigating using a Prusa i3 Mk3 with a Fiber Force Nylforce carbon-fiber reinforced filament. That set-up is used to print 5 of the larger parts of the caster portion of the Swerve Drive. We certainly liked the fact that using 3D printing allowed us to come up with a design for the caster that is much more “organic” than what would have been possible with standard fabrication techniques.

Another downside of the MarkForged, I found, is the considerable amount of time taken to print a part. The largest and most complex part of our Swerve Drive caster assembly is a 17-hour print on the Prusa i3 Mk3. On the MarkForged, using essentially the same slicing parameters, we are told the print would take no less than 40 hours. Right now, the team can produce 4 Serve Drive caster 5 parts sets in about 6 days. On the MarkForged, we are talking about a 12-day process… Our team has access to 3 Prusa i3 Mk3 printers to accelerate the fabrication process, if need be. I don’t know that we could afford more than 1 MarkForged printer. Those factors need to be taken into consideration given the tight Build schedule that is the bane of FRC.

As the 2019 Build season is coming upon us, our Swerve Drive development will wind down until another off-season R&D push.

Thanks for the responses. Regarding gears and friction, I was just reading this article on lubricants (, and I’m curious: has anyone tried graphite powder or any of the other dry lubricants with gears made of Onyx (or other Nylon+CF) to reduce friction & noise? Or would there be risks of contamination damage to other parts of the robot?

Given the number of posts on CNC and 3D printing, manufacturing seems conspicuously missing from CD’s forum structure (could even support a few sub-forums). Also, I noticed there are several dozen FRC-related designs on Thingiverse. Maybe it’d be good to have a Slack or something on FRC CNC (both additive and subtractive) as a place to share FRC-specific designs, general applications, practices, problems/solutions. People always seem interested in talking about it on here.

Thanks again! Hope all are enjoying holiday time… the calm before the storm rolls into Deep Space. :ahh:


I’ve had success making voids in parts to accept mcmcaster " Low-Profile Narrow-Base Weld Nuts with Projections"; but this requires a pause for a fully blind insert, or a hole carried through the part bottom. One time I took too long to insert a nut, and the layer adhesion where it was paused was greatly reduced.

We’re printing some hex hubs ( directly from the provided models to see how they fare. I’d like to see more finite control of infill and layer height changes across a single part in eiger. It seems to be quicker to get parts to print than other slicing programs (Cura, Slic3r), but with the tradeoff of less control.

Using a Mark Two at work, the turbo build option is a huge time saver, and I have not seen it result in any lower quality parts.

1 Like

We only have a Pulse (prusa clone) FDM printer. You will get those gaps if you don’t design the parts right. Each wall thickness is usually specified by the nozzle width there is a little play with over and under extrusion but if you lets say have a .8mm nozzle and the design specifies a 3mm wall then you will have a .6mm gap as the slicer will lay down the lines at the 3mm border and 0mm border and then work in (even if it might print inside out. 3mm is not an integer multiple of .8. Same comes down with layer height. just imagine if you are building something out of 2x4 and can’t cut them in the 2x4 direction only length wise then your width has to be a multiple of 4 and your height a multiple of 2 or there are going to be gaps or something will be too tight.