Feature resolution in UM3?

[thalassa 9]

I'm in the midst of researching which 3D printer to buy for my library (located at a university marine biology research station). I haven't decided between an SLA or FDM printer. The Form 2 and the Ultimaker 3 are my leading contenders. They each have advantages and disadvantages.

Spec sheets aren't really comparable, as they use different jargon when referring to aspects of "accuracy" and "resolution." Some of my scientists need to print near the performance limits of desktop 3D printers with respect to feature resolution. Formlabs has done an excellent test on the Form 2 that shows exactly how well the printer performs at printing small features (they get it down to about 150 microns). See this blog post (here).

I'm wondering if Ultimaker has performed a similar analysis with the UM3? It would be very helpful to know how layer resolution and XYZ accuracy translate to feature resolution. Thanks!

[gr5 2,173]

Anything you can imagine measuring and finding an error you can improve.

So for example the printer tends to be pretty consistent if you use all the same settings in cura. If a key section of your part is 10 microns too big just reduce the part size by 10 microns in cad and it will probably come out perfect every time after that.

But change filament type or temperature or speed or infill speed or infill pattern and now you have to start over with the measurements.

fdm printers have a round nozzle hole - default for um2 and um3 is .4mm so you are going to get a radius of curvature on corners of 0.2mm. HOWEVER you can always put in a smaller nozzle such as a .25mm nozzle. But now your prints will be 4X slower.

The great advantage of fdm over sla printers is mostly materials. You have more materials to choose from with fdm and they tend to be less brittle. Also price. FDM materials are much cheaper than resin.

So this kind of measurement is difficult. It's best to just go to 3dhubs and find someone near you with a UM2 and ask them to print your part and just look at the results versus someone near you with a form2.

I'm not sure but my gut says the UM2 will probably print a little better than the UM3. Partly because people who have a UM3 couldn't have had it very long and probably aren't experts with it yet. Partly because the head is lighter.

[thalassa 9]

Thanks for the reply, @gr5. I appreciate your time!

 

Anything you can imagine measuring and finding an error you can improve...

I understand all of the caveats and the fiddling that can be done. This isn't much comfort to me, as I'll be the sole source of support for the station WRT 3D print jobs. I'm hoping to not have to tweak & reprint every job that comes through the door (and won't have time for that anyway).

My question was more practical. Has UM (or anyone) actually tested the limits of the performance of these printers in a robust way? Do all the tuning and tweaking you want (thought that's not quite fair to the consumer) - what can the printer really do? I think this should be done using the default nozzles and materials.

 

So this kind of measurement is difficult.

Formlabs did it and was completely transparent about the results. I really respect that transparency (especially since I wasn't impressed with their results).

 

It's best to just go to 3dhubs and find someone near you with a UM2 and ask them to print your part and just look at the results versus someone near you with a form2.

I mean, I don't mean to be such a Type-A scientist, but isn't this test something that Ultimaker should have done as part of R&D? IMHO, saying, "There are so many variables we can't really test for XY print feature resolution" is a cop out and doesn't lend much confidence in the machine.

Don't get me wrong - the UM3 is my leading contender right now. 8) It looks like an amazing FDM printer. I am just really surprised that this kind of characterization hasn't been done, and I think the results would be very helpful in assessing the true performance of the unit. ¯\_(ツ)_/¯

[meduza 191]

I can tell you without printing what would happen if you tried to print the test model that Formlabs used on a Ultimaker 3. The features would not show up at all.

Why? All the features they test are smaller than the 0.4mm nozzle, so the slicer (Cura) will just ignore them.

Edited December 21, 2016 by Guest

[neotko 1,417]

Thanks for the reply, @gr5. I appreciate your time!

 

Anything you can imagine measuring and finding an error you can improve...

I understand all of the caveats and the fiddling that can be done. This isn't much comfort to me, as I'll be the sole source of support for the station WRT 3D print jobs. I'm hoping to not have to tweak & reprint every job that comes through the door (and won't have time for that anyway).

My question was more practical. Has UM (or anyone) actually tested the limits of the performance of these printers in a robust way? Do all the tuning and tweaking you want (thought that's not quite fair to the consumer) - what can the printer really do? I think this should be done using the default nozzles and materials.

 

So this kind of measurement is difficult.

Formlabs did it and was completely transparent about the results. I really respect that transparency (especially since I wasn't impressed with their results).

 

It's best to just go to 3dhubs and find someone near you with a UM2 and ask them to print your part and just look at the results versus someone near you with a form2.

I mean, I don't mean to be such a Type-A scientist, but isn't this test something that Ultimaker should have done as part of R&D? IMHO, saying, "There are so many variables we can't really test for XY print feature resolution" is a cop out and doesn't lend much confidence in the machine.

Don't get me wrong - the UM3 is my leading contender right now. 8)It looks like an amazing FDM printer. I am just really surprised that this kind of characterization hasn't been done, and I think the results would be very helpful in assessing the true performance of the unit.  ¯\_(ツ)_/¯

 

This is very interesting. The thing is that atm UM, afaik, focused on the important part to release a new machine, having it 'work' on a number of common and more complex scenarios (where some of us got in doing beta before the release) and the main goal was to help UM to help find possible issues so everything could be as smooth as possible. Said this, I really thing that once all the initial release is done, that kind of information like material expansion, per material / color / type, should be possible to compute and get on a nice chart so future releases of Cura can manage that data to try to adjust as much as possible to the goal of printing with just one button (I think this is the golden goose of 3d printing that will take time, even years, to be possible for FDM). R&D guys are amazing (saying this after having the chance to meet some of them) they do spend a lot of time doing really interesting tests, but mind this, atm the focus (or what I thing the goal was) is to be able to print as easy and as good as possible. After this initial part I think that, like you point, this kind of data is really important in order to get more accuracy on the prints. And since UM3 has the filament spool sensor (for UM brands atm) it should be possible to add the kind of information you want, but also mind this, on a SLA printer you can have much more control over a certain number of parameters, you don't have a moving head but only a Z motor and a light source to 'print'. So ofc, there's much less parameters to control, and also there's much less materials. Ofc, also SLA has some very interesting materials, but also it has, for what @IRobertI told on a chat, much more stuff after the print is done, for example (and about this I only remember parts from what @IRobertI told, you need to do Post Curing after printing, you need a cabinet to it faster (and I assume is a requirement to get more precision?) and most of the chemicals can have nasty odors and require gloves, mask, etc. So SLA could look cheaper, but also the materials are more expensive, and with a FDM you get more variables to control but cheaper and less (or close to zero) odors or particles in the air.

I think both technologies will evolve, and since SLC has less materials and a more controller environment the kind of data you want is easier to obtain, but with FDM you get more flexibility and cheaper materials, less ambient problems for most materials in exchange for more variables.

I must say that I have never used a SLC. And I also agree that we need more data to control more of the variables at hand and get more precision. But I bet that's somewhere on the UM to-do along the way.

[neotko 1,417]

Thanks for the reply, @gr5. I appreciate your time!

 

Anything you can imagine measuring and finding an error you can improve...

I understand all of the caveats and the fiddling that can be done. This isn't much comfort to me, as I'll be the sole source of support for the station WRT 3D print jobs. I'm hoping to not have to tweak & reprint every job that comes through the door (and won't have time for that anyway).

My question was more practical. Has UM (or anyone) actually tested the limits of the performance of these printers in a robust way? Do all the tuning and tweaking you want (thought that's not quite fair to the consumer) - what can the printer really do? I think this should be done using the default nozzles and materials.

 

So this kind of measurement is difficult.

Formlabs did it and was completely transparent about the results. I really respect that transparency (especially since I wasn't impressed with their results).

 

It's best to just go to 3dhubs and find someone near you with a UM2 and ask them to print your part and just look at the results versus someone near you with a form2.

I mean, I don't mean to be such a Type-A scientist, but isn't this test something that Ultimaker should have done as part of R&D? IMHO, saying, "There are so many variables we can't really test for XY print feature resolution" is a cop out and doesn't lend much confidence in the machine.

Don't get me wrong - the UM3 is my leading contender right now. 8)It looks like an amazing FDM printer. I am just really surprised that this kind of characterization hasn't been done, and I think the results would be very helpful in assessing the true performance of the unit.  ¯\_(ツ)_/¯

 

This is very interesting. The thing is that atm UM, afaik, focused on the important part to release a new machine, having it 'work' on a number of common and more complex scenarios (where some of us got in doing beta before the release) and the main goal was to help UM to help find possible issues so everything could be as smooth as possible. Said this, I really thing that once all the initial release is done, that kind of information like material expansion, per material / color / type, should be possible to compute and get on a nice chart so future releases of Cura can manage that data to try to adjust as much as possible to the goal of printing with just one button (I think this is the golden goose of 3d printing that will take time, even years, to be possible for FDM). R&D guys are amazing (saying this after having the chance to meet some of them) they do spend a lot of time doing really interesting tests, but mind this, atm the focus (or what I thing the goal was) is to be able to print as easy and as good as possible. After this initial part I think that, like you point, this kind of data is really important in order to get more accuracy on the prints. And since UM3 has the filament spool sensor (for UM brands atm) it should be possible to add the kind of information you want, but also mind this, on a SLA printer you can have much more control over a certain number of parameters, you don't have a moving head but only a Z motor and a light source to 'print'. So ofc, there's much less parameters to control, and also there's much less materials. Ofc, also SLC as some very interesting materials, but also it has, for what @IRobertI told on a chat, much more stuff after the print is done, for example (and about this I only remember parts from what @IRobertI told, you need to do Post Curing after printing, you need a cabinet to it faster (and I assume is a requirement to get more precision?) and most of the chemicals can have nasty odors and require gloves, mask, etc. So SLC could look equal, but also the materials are more expensive, and with a FDM you get more variables to control but cheaper and less (or close to zero) odors or particles in the air.

I think both technologies will evolve, and since SLC has less materials and a more controller environment the kind of data you want is easier to obtain, but with FDM you get more flexibility and cheaper materials, less ambient problems for most materials in exchange for more variables.

I must say that I have never used a SLC. And I also agree that we need more data to control more of the variables at hand and get more precision. But I bet that's somewhere on the UM to-do along the way.

Also indeed the variables are quite a lot to control. And personally it took me a really good time to be able to create constant repetition on my prints. But ofc there are 'addons' from third parties that help achieve that goal (imo). Also that's a great advantage of the um/opensource, formlabs doesn't release their info about how is build and you depend only on them to get spare parts or upgrade (is the formlabs upgradable?).

[IRobertI 520]

what @IRobertI told on a chat, much more stuff after the print is done, for example (and about this I only remember parts from what @IRobertI told, you need to do Post Curing after printing, you need a cabinet to it faster (and I assume is a requirement to get more precision?) and most of the chemicals can have nasty odors and require gloves, mask, etc.

 

The only chemical you use, besides the actual resins, is IPA which can be quite smelly and you don't want to stand around breathing it for a long time (which I do, sometimes...). Some of the resin types have a bit of a smell as well. And you absolutely need to use gloves at all times when handling uncured resin. From the moment I open the lid, gloves go on.

And yes, all parts need to be post cured to reach their final and full strength/properties. For that you can use the sun (slow, unreliable, no control), build your own chamber, get a cheap cure chamber for nail polish, or go for a more professional/industrial unit for quicker curing time and added heat which helps as well.

Precision is a tricky subject. SLA prints can also be susceptible to warping and wherever you have support structures you'll get little blips that may need to be sanded away. For example, if you have a large flat surface it should be printed at an angle with support. Imagine picking up a square of cloth by pinching it with your fingers and you get an (very exaggerated) idea of what that flat side will look like.

There's a bunch of other considerations to take into accoutn but I'm way too tired right now to go through it

[gr5 2,173]

Although Ultimaker is trying to make printers plug and play and easy to use, if you want to hit the limits of it's capabilities you have to spend many hours. You won't have to print everything twice - mostly you will learn how much to adjust models and get it right usually the first time (with practice). However once you change color of filament you might have to start over. These machines are more like a milling machines in complexity and required knowledge than like a 2D printer. Did you get that? 3d printer complexity not like 2d printer even though both phrases are identical except for one digit.

[neotko 1,417]

Although Ultimaker is trying to make printers plug and play and easy to use, if you want to hit the limits of it's capabilities you have to spend many hours. You won't have to print everything twice - mostly you will learn how much to adjust models and get it right usually the first time (with practice). However once you change color of filament you might have to start over. These machines are more like a milling machines in complexity and required knowledge than like a 2D printer. Did you get that? 3d printer complexity not like 2d printer even though both phrases are identical except for one digit.

 

I think the main question is, why UM doesn't add material tests to each color and profile. Formlabs has the task easier since they have less materials to control on a machine that has less moving parts. But indeed the variables are many on a FDM since you are extruding plastic through a nozzle and there are many variables to would require so many sensors that the price would sky rocket. Also @gr5 some 2D printers are far more complex than an FDM (just try to make build a 2D printer). But the technology has evolved into mass market making all more accessible to users. FDM is becoming more accessible but to stay on the price range they are the amount of systems to keep all under control would make all many times more expensive.

So @thalassa I really think that if you want precision you might want a much more expensive machine. But also if you are patience, with little repetition and tests, you can get great results.

Edited December 21, 2016 by Guest

[thalassa 9]

I can tell you without printing what would happen if you tried to print the test model that Formlabs used on a Ultimaker 3. The features would not show up at all.

Why? All the features they test are smaller than the 0.4mm nozzle, so the slicer (Cura) will just ignore them.

 

So, is there a different kind of test print that could be made on the UM3 to test feature resolution, or does the software need to be tweaked to enable these kinds of tests? If the 0.4 mm nozzle can lay a 20-micron Z-layer, seems like it should be able to do better than 400 microns in the X/Y.

But, this gets back to my original question: why hasn't anyone actually characterized the performance of this amazing machine? I don't get it.

[thalassa 9]

These machines are more like a milling machines in complexity and required knowledge than like a 2D printer.  Did you get that?  3d printer complexity not like 2d printer even though both phrases are identical except for one digit.

 

Yes, I understand that 3D printers are more complex than 2D printers, and I don't expect perfect prints every time or to be able to print without tweaking here and there. All I really wanted, before spending $3500, was a better understanding of the performance of the machine. Like I said before, I think this is an amazing printer, and I'm probably going to buy one. I'm not here to criticize the performance of the UM3, but to understand it.

[thalassa 9]

Precision is a tricky subject.

 

People keep saying that, with lots of hemming and hawing about materials and settings. Leaving SLA out of the equation, has anyone ever seen a feature resolution test performed on ANY FDM printer? If not, why do you think no one has done this? Is anyone interested in trying?

Now I just have a bee in my bonnet about it!

[gr5 2,173]

Well it's just so hard to say what's the best one can do. Even if you restrict to unmodified UM3 and only the 0.4mm nozzle (there will be 0.25mm nozzles in the future) then there is still just a huge amount of variability. With prints that look like crap and prints that look great. If anyone even tries to do this I will not believe they came even close to the quality that can be achieved. It's just too complicated for anyone to do where I would trust the results of the test.

So I guess my point is even if someone did this test it would be worthless.

SLA printers have their own quirks. You can't print something with a flat overhang like a "chair" print standing on it's 4 legs. It doesn't "bridge" well. So they always tilt things like that. FDM printers bridge quite well but don't tilt well (severe overhangs are tougher I think on FDM). SLA printers use the same material for support and have lots of support attachment points which are "ugly". UM3 has dissolvable support.

In general SLA is going to give you better resolution and parts will look better. But I can make a print better and with equal resolution than an SLA print if I use a .1mm nozzle (I've done it).

Then again there are SLA printers that go below .1mm resolution at least in X,Y.

People on this forum will say things like "ultimaker gray filament sucks - I get much better quality with color XXXXX" and they post lots of pictures and analysis and I'm almost convinced until someone else does the same analysis and says "gray is the best - it's white that sucks" and then I find that the guy who like white so much was printing 20C cooler and so on. It is just too damn complicated to do a good scientific comparison. But you can do a less scientific comparison by walking around a show and taking free samples and comparing the printers. I think if you do that you will agree that SLA has higher quality looking prints. But there are many downsides to SLA (price of resin, lack of colors, lack of materials, dangerous chemicals, brittle parts and more).

[gr5 2,173]

So maybe it is a cop out but I really don't know how to do this test in a way that I would trust the results. And I wouldn't trust anyone else's test very much.

[thalassa 9]

So I guess my point is even if someone did this test it would be worthless.

 

I guess we'll have to agree to disagree. You wouldn't even trust a test conducted by the manufacturer?!

If FDM printers were so inconsistent that they can't even be reliably tested for performance, then they would be useless for professional use. But, people are using these professionally - Ultimaker themselves used them to make parts! - so we know they can perform reasonably well.

This failure to characterize them in a robust way is only because we haven't been clever enough to design a good test, or no one has made it a priority.

[tomnagel 126]

XY nozzle movement has a resolution of 12.5microns. Z movement has a resolution of 2.5 microns. But you probably figured that out already.

The UM3 currently only features 0.4mm nozzles, other sizes will follow. The size of your nozzle party determines feature size, very much depending on what you want to print.

The slicer will play a big role too. If you need a wallthickness of 0.6 and you use the 0.4mm nozzle, I think the slicer will give you 1 wall of 0.4mm. You may not like that.

There are many error sources. In mechanics there is Eccentric pulleys, not-straight shafts, steppermotor non-lineairity, in electronics we have stepperdriver non-linearity. And then there is influence of materials and printing process: warpage and/or material shrinkage. And there is influence of the software: e.g. the slicer will make a circle from many straight lines, which makes holes a little bit smaller. Etc. Etc.

Some errors may hurt you, some will not. I can design a test like FormLabs for the Ultimaker in which some error sources will not show up. That makes the UM look good. Or should I design a test in which all errors do show up? That makes UM look bad, and maybe worse than you will see in practice.

If you could share some typical designs you want to print, together with info on what features are important to you, we may be able to tell you whether or not the UM3 is able to perform according to your wishes.

[thalassa 9]

... I can design a test like FormLabs for the Ultimaker in which some error sources will not show up. That makes the UM look good. Or should I design a test in which all errors do show up? That makes UM look bad, and maybe worse than you will see in practice.

If you could share some typical designs you want to print, together with info on what features are important to you, we may be able to tell you whether or not the UM3 is able to perform according to your wishes.

 

Well, it would be nice to see tests that show performance across strengths and weaknesses. TBH, I thought the results from the Form2 were suprisingly bad, and I respect their transparency in sharing the test and the results.

No such thing as a "typical design" around here. Our reseachers will be printing everything from scaled-down whale baleen plates to custom laboratory and intertidal research equipment. Prints will run the spectrum on shape, size, and function. This is the main reason why I've had a hard time comparing 3D printers - I have to balance a wide range of needs and use cases. Ultimately, I think the UM3 is maybe my best choice for meeting these needs. I appreciate your input!

[meduza 191]

Well, i'll think the Formlabs test had a pretty good looking result.

I cannot see how the Form2 could have gotten a better result, it's laser dot does have a half-width of 140 microns, so it is not surprising at all that you cannot really make a positive feature smaller than about 150 microns.

If you did the same test in the X-Y plane for any FDM printer with a 0.4mm nozzle, it will even out at somewhere around that if the software even would try to print it, since you cannot put out lines much finer than the hole in the nozzle, but as i said earlier, most FDM slicers will simply ignore features smaller than the line width instead. But as @tomnagel said, you can of course position that 0.4mm line in much finer steps.

On the other hand, in another plane, it will give a different result.

Edited January 6, 2017 by Guest

[tomnagel 126]

Well, it would be nice to see tests that show performance across strengths and weaknesses. TBH, I thought the results from the Form2 were suprisingly bad, and I respect their transparency in sharing the test and the results.

Out of curiosity: what exactly do you find bad on their results?

[peggyb 170]

that must be a really hard task to find a printer that does everything...

I think you will have to accept that each printer has its limits and that they each have their own specialities and live perfectly next to each other. Depending on your needs of size, material and detail you choose your machine. The Ultimaker sits in the middle of the range, it can do a lot but also has its limits.

Form2: max. size - 145 × 145 × 175 mm (your model is positioned diagonally most of the time, support takes a lot of space), laserpoint 140 micron, layer height 25 micron, limited and very expensive materials, not very good at flat surfaces, support scars.

UM3: max. size 215 x 215 x300 mm (extended), nozzle 400 micron (for now), layer height 20 micron, duo-materials, open to a lot of materials-different prices, capable of complex shapes with PVA support, top/bottom layer lines more visible than wall lines

In the photo:

left: UM2Go, MoldLay (wax like) 100 micron layers, 1 hour

right: Form2, Black resin, 25 micron, 2,5 hours

[tomnagel 126]

This is what Stratasys publishes for one of their professional systems. It gives a lot of info, but still I have many questions. Why not publish the drawing of the object that they have actually measured? What dimensions were measured?

http://www.stratasys.com/~/media/Main/Secure/White-Papers/WP_FDM_Fortus360mc400mcAccuracyStudy.pdf?la=fr

Edited January 6, 2017 by Guest

[thalassa 9]

 

Well, it would be nice to see tests that show performance across strengths and weaknesses. TBH, I thought the results from the Form2 were suprisingly bad, and I respect their transparency in sharing the test and the results.

 

Out of curiosity: what exactly do you find bad on their results?

 

I think I had unreasonable expectations about how well it would perform. With all of the hype about SLA printers, I thought that 150-microns wasn't very good feature resolution. Now that I understand a bit more about how they work, WRT the laser dot size, I shouldn't have been so surprised.

It was honestly a nice discovery that SLA printers don't have much of an edge over FDM printers (depending on what you want to print). I can't have big buckets of isopropyl alcohol sitting around (I work in a library), so SLA printers are OUT. UM3 it is!

[thalassa 9]

This is what Stratasys publishes for one of their professional systems. It gives a lot of info, but still I have many questions. Why not publish the drawing of the object that they have actually measured? What dimensions were measured?

http://www.stratasys.com/~/media/Main/Secure/White-Papers/WP_FDM_Fortus360mc400mcAccuracyStudy.pdf?la=fr

 

Thanks for sharing this! It is interesting, but I agree - still many questions. Also, I find it maddening when they test things in inches. mm or microns would be so much more useful.

I've found that most companies share information regarding XYZ accuracy, which is important, but I've only seen Formlabs address the issue of feature resolution. That, in itself, is interesting!