Inaccurate Results with UM Nylon with Cura on S5; General dissatisfaction with default profiles

[P3D 46]

I am printing a fairly basic part in Nylon, and the final dimensions are very inaccurate in X and Y directions. For outer dimensions, the part is 0.3mm too big in X Y, for inner dimensions (hole in the XY plane) it is 0.7mm(!). When compensating with negative "horizontal expansion", the inner hole cuts into the wall in layer view at -0.35, so I have to compensate this in the CAD file... this is not a viable long-term solution! Does anyone know if there is a way to fix this in Cura? 

 

Also, the default UM material profiles in Cura generally seem to be rather unrefined and crude, judging from my experience - see the recent thread on CPE issues. When paying this amount of money for a machine, I expect much more, honestly - especially when seeing what quality for example a Prusa i3 for nearly 1/10 the price is capable of. If Ultimaker is serious about gaining industrial customers in the long term (and who else, apart from wealthy hobbyists, will pay ~6500€ for a FDM printer?), they really have to offer a solution that works without endless tinkering.

 

On the positive side, at least in Z direction the accuracy is good so far.

Edited September 7, 2018 by P3D

[Smithy 1,144]

Inner holes are always a problem and always too small. The only solution is to make a correction in the CAD file, but bad if you want to use this CAD file after the prototype for real production on CNC machines. The reason is melting/cooling/expansion of the material and at the hole there is nothing on one side, so the material can easily expand "in the hole". And different materials have different expansions. With PLA there is the same effect, but not so much.

 

So you can adjust the outer dimensions with a negative horizontal expansion and for the hole you can only adjust in the source file and make the hole bigger.

 

Regarding the material profiles, I fully agree. I do not expect, that a 3rd party filament is perfectly tuned for the UM machines, but UM material should. And you know what we tested in the other thread with the UM CPE material. 

[P3D 46]

Thanks for your feedback! Correction in CAD is far from ideal in our companys workflow unfortunately.

 

For now, I made the adjustments in CAD, and while there is still some error (seems to be non-linear?), the accuracy is OK for the intended use of the part.

 

In my opinion, this is a prime example of what *should* be corrected in a material/printer profile. This should not be too difficult for Ultimaker, as they have in-depth knowledge of their materials, hardware and software. In other words, they are in full control of all the parameters.

 

A turn-key "print the STL file exactly as it is" solution would be a huge differentiator, especially for industrial customers

 

On the other hand, they don't even seem to bother about simply ticking a box that is already there, like disabling the combing for CPE....

 

Sorry if I seem to be quite negative, I really like the S5 for its flexibility and general reliability, as well as the very useful material range. However, I expected much more from the combination of HW/SW/Materials coming from a single manufacturer.

 

 

 

 

[Smithy 1,144]

The corrections are depending on different parameters like layer height, speed, temperature, and a lot of other parameters, so when you make try and error rounds to find your setting, do it with the settings you want to have for the final print.

 

Of course UM could make this "automatic" in Cura, but due the fact, thats depending on so much parameters, it will not be so easy, to fit all circumstances. But yes, they have everything under control and it would be more less possible.

 

On the other hand, I have a colleague here with an Anet A8 and watching him, modifying, tuning and operate his printer, then you actually see how small the UM problems are. Working with such a China thing is really pain in the ass, but ok, they cost a fraction of an Ultimaker.

[gr5 2,167]

Nylon is difficult to print and you shouldn't compare the S5 with nylon versus the prusa i3 with PLA.

 

But overall I agree with most of your comments.  I think UM should come out with an "engineering" profile that prints with higher dimensional accuracy.  Here are the steps I do to improve dimensional accuracy over default profiles:

1) Set all printing speeds to be the same - anytime you speed up or slow down you over/under extrude briefly.  Slower speeds will give better results.  I usually print at 35mm/sec (ALL 6 or so printing speeds) and 0.2mm layers and the parts are both accurate and gorgeous and just as fast as 100mm/sec (primary speed) at 0.1mm layers.

 

2) Disable jerk control and acceleration control.  When enabled it prints slower and prints have more globby corners because it takes longer to slow down and overextrusion can be quite prominent on the corners.  jerk control and acceleration control remove "ringing" which can be ugly but is too small to measure with a micrometer so in "engineering mode" or "accuracy mode" we should embrace the ringing in exchange for higher accuracy of parts.

 

Back to nylon.  Make sure you keep the nylon very dry.  Even just printing it for a single 12 hour print in a humid (70% humidity or worse) environment is enough to degrade it and it needs drying again.  I recommend 2 25gram dessicant packs stored with the filament that color-change when wet and recharge the packs in a microwave (very easy to do with some experience) every time they turn to the "wet color".  I leave the spool in a 2 gallon zip lock with the dessicant and open the zip lock only enough to let the filament out and print it like that - spool in zip lock with filament leaving the small hole to the feeder.  Nylon can be reset/dried by putting the whole spool on the heated bed of your printer at 95C overnight with a towel thrown over it.  Or in a cardboard box.  100C can warp some spools a bit but the nylon will be fine.  95C is safer.  Very wet nylon makes a hissing/popping sound and comes out snowy/spongy/foamy.  Dry nylon comes out glossy and natural color nylon comes out clear when dry - bottom layer should be transparent.

 

Also cover the front and top of the printer and use a heated bed temp of 100C to get the air temp up to 35C inside.  Set the fan low to get good layer adhesion (default nylon profile fan speed is probably okay for S5 but much too fast on my UM3).  

 

[P3D 46]

Thank you all for the feedback and the suggestions! 

 

As for the "print dry" suggestion, I am well aware of just how moistures sensitive many Nylons and other filaments are, so I'm in the process of building a drybox with a mount for the original NFC Reader. 

Interestingly, UM Nylon seems to be less moisture sensitive than UM CPE, not to speak of TPU - seems that the claims about reduced moisture uptake are more than marketing.

 

Thank you @gr5 for your profile insights! I tried them out, and parts come out much more accurate than with the default profile. After tweaking the horizontal expansion setting a bit, too, I am quite happy with the resulting dimensions, inner holes being near-perfect, and outer dimensional innaccuracy of max. 0.1mm.

 

Ultimaker, if you're listening - dimensionally accurate "engineering profiles" would be most certainly appreciated by your customers, and would be one more reason to stay with UM materials...

[mgibeault 0]

There's definitively a need for profiles that produces accurate parts without fiddling into settings. I'm quite disappointed in our S5 to get such inaccurate parts. Holes are too small and posts are too big, by a lot and on all the materials we used up to now.

We're not supposed to go through a trial-and-error process with a new machine, using Ultimaker's materials, and simple parts that needs to be assembled together.

[PaulK 14]

I think it is more then profile needed. IMO, slicers need to be come aware of two things (well, ok at least two ) 1) Thermal mass and 2) Expansion/contraction. (1) can solve some problems by doing dynamic cooling instead of fixed for everything. (2) is needed to fix the whole hole thing, and general dimension issues. It would need to do geometry analysis and "correct" the geometry based on estimated contraction/expansion. I expect this will be very difficult. Part of the problem is, I suspect, that STL is a really really bad idea, because you want to know about solids but are given information about surfaces, so you basically have to reverse-engineer the solid.

 

But, in my mind, this problem is solved by improving the slicer, not by some fixed expansion value in the profile. On another printer I played with those parameters and it was impossible to get holes and posts and perimeters all right at the same time; it needs to be smarter. Maybe the next generation of slicers? I keep hoping some super smart Ph.D project will take it on and add it to Cura 🙂

[mgibeault 0]

I used a lot a small Zortrax printer and, while it had other problems, the parts it produced were of excellent dimensional accuracy. Parts fitted well together without any post-processing.

I don't think it made fancy thermal control, they just got very good default settings for each material they sell, probably by doing a bunch of tests...