Thursday, January 17, 2013

The Vexing Issue of FDM Edge Accuracy


I opened my email this morning to find this plea from a Cube owner.

" I was wondering if you could help me with a problem – every time I print something with my Cube it turns out a little bit bloated. That is 10mm holes turn out to be 9.5mm, and 10mm shafts turn out to be 10.5mm. This means that printing things like gears requires a lot of sandpaper work – I printed a small planetary gear system from Thingverse and it took me 5 hours to make it even fit. Do you have the same problem? Maybe you have some tips?

I would like to print something like a wankel engine model - http://www.thingiverse.com/thing:42579 - but I'm afraid that I will spend hours with sandpaper...

Best Regards

Andrzej T....."

Andrzej, you are not alone. 

And, in fact, Cube owners are not alone. This is a problem that affects most, if not all FDM (Fused Deposition Model i.e. Extrusion) printers. Here is a video that explores this issue on a huge printer costing more than $20,000! Here is an image of the printer...


Dimension BST 768

And, here is a video documenting the BST 768's issue common on smaller printers...




Now, I have to note that this was the experience of one user FIVE years ago. But, comments on the video seem to indicate that this behavior is shared by other users as well. It is possible that accuracy has increased in these high end models over the last 5 years. Even so, it is instructive to those of us with less expensive platforms.

We've written a little about this in the past and shown how we can alter our designs to allow for the fact that holes end up being slightly smaller than designed and outside edges (posts) end up being larger than designed.

The problem is quite acute when we download an STL file. If we try to reduce the size of the object in the Cubify client software then holes are reduced even more dramatically. If we try to expand the model to open the holes, then the outside edges are expanded more dramatically.

Some time ago, I suggested that the STL file, while convenient, was probably NOT the best sharing format among the 3D printing community because there are very few options for conversion to a format that can be modified with any finesse. Among Cube owners, the native Cubify Invent format would be a good alternative.

At some point, I will be visiting 3D Systems again to get a hands on look at the CubeX printer. When I actually have the opportunity to talk to the designers face to face, I want to try to gain some more insight on this issue to see if there are any realistic expectations of being able to deliver the kind of accuracy we expect from milling machines.

Remember, the material is being melted. So, that, in and of itself, might be part of the problem that 3D printer designer's face. But, the fact that we are bringing this up at all is a testament to how far we have come in just a few years. Back when I started blogging about the POTENTIAL for low cost desktop 3D printers, we felt lucky to be able to print a recognizable object!! That problem as long been solved and now we are talking about tenths of an inch. :)

This discussion is good because it sets higher goals for all of the 3D printer manufactures, The fact that 3D Systems is touting the increased accuracy of CubeX means they ARE listening and are having some success at pushing the accuracy envelope further.

Thank you Andrzej.

I wish I had a solution for STL files. But I don't. Let's hope that somebody out there might have one and share that with us. But, for now, "vexing" is the best way to describe it.

An Observation...

Orientation seems to have a bearing on accuracy. I think that I see that Y Axis, or VERTICAL accuracy, which depends on layer accuracy, is much closer than X and Y Axis accuracy. That may be because the physical tip of the print jet acts as a mini-trowel controlling the form of the molten plastic whereas there is no restriction on HORIZONTAL edge flow after the plastic leaves the print jet orifice. But, that is all conjecture on my part. LOL!

UPDATE TO THIS ARTICLE...

This is welcome news from 3D Systems in response to Andrzej's question...
"The 2nd gen Cube is 2X more accurate than the 1st gen
0.5mm dimensional tolerance over the length of PrintPad
And you are right...it is a key focus."
I will be with some members of the Cubify team this weekend for an event in Wilmington, DE at the Hagley Museum. I will try to run my Hole vs. Post test on one of the 2D Generation machines. This is truly great news!
For those that are 1st Generation Cube owners, I can only tell you what my attitude is toward the fact that those buying the 2nd Generation Cube now are getting a more advanced machine. Having been an "early adopter" of most of the major consumer technologies from the late 1960's to now, I have learned that the secret to minimizing angst at rapid improvements is to use the life out of anything I purchase. In the time between I purchased my 5D MK II Canon and the announcement of the 5D MK III, with greatly improved video and low-light performance, I took more than 10,000 photos. Likewise, since I first opened my Cube, I believe that I have printed at least 100 items and probably way more. Even if I had to replace it today, it would have been a great investment. I'm just happy to see things progressing. :)

8 comments:

  1. From my point of view the 2nd gen Cube has not been advanced mechanically if you look at the posted photos.
    I suspect that 3D Systems has used the same molds to produce the outer shape and used the same GT2 belts and pulleys to drive all axis.

    I believe that the improvements will come from the firmware of the 2nd Cube in conjuction with a soon to be released Cubify desktop software that controlls the slicing part (like all advanced settings in Slicr).

    Hope that all existing user will benefits from this as well. If not is there somebody out there that is able to take a look at the current 1st gen software and compare with the 2nd gen software once it is available (or the first update of the 2nd gen software as it has to be loaded via computer). Maybe with some coding this would enable more user to benefit from the advanced accuracy

    Let me know what you think.

    ReplyDelete
    Replies
    1. I think you are absolutely right. I posted in a different article that really the only differences are a different print pad, and different sized extruder. The main difference between the two cubes are the firmware and the software.

      While Cubify may be fast about support and whatnot. They will not be updating the 1st Cube, and they will only be focusing on the latest technology they can develop. Basically if any problems with the Cube 2 come up... they will be fixed on the Cube 3. Even if those problems are mainly just software and firmware (which they are).

      Delete
  2. While, like you, I suspect that improved accuracy is a function of software rather than hardware, we can't be sure of that until we get a chance to make a comparison. I have found that the Cubify team truly does (1) care about Cube owners and (2) want the best possible result from their products.

    So, if all it takes is a firmware upgrade then we'll probably see it. If it would take something more, then I would rather they spend resources moving us forward as rapidly as possible rather than be held back dealing with legacy issues.

    I know I have a lot of friends who would DEFINITELY be Cube owners if they could print in Chocolate! LOL!

    Nice contribution!

    ReplyDelete
  3. Hi Tom,

    I was about to email you or send a message on facebook about the accuracy when I read this blog entry. Thanks for the write up, at least it answers a few questions I had about my cube. I originally wondered if it was merely that it prints "on the lines" as opposed to "inside the lines" of an STL file. Combining that with the 0.5 resolution, you'd get the anomaly. A bit of a pain.

    That being said, wanted to print a thingyverse object yesterday and got the dreaded Error 14 code. So far Cubify support has been very prompt and will be sending me a new jet to replace the broken part.

    Again, thanks for the blog entries, I'm getting more help here than anywhere else.

    TvZ

    ReplyDelete
  4. Great to hear from you, TvZ,

    While all things mechanical (or electrical) can break, the thing that is really important is how quickly the manufacturer responds. And, from the emails that I get from people, the Cubify support team has really got it together. I'm glad to hear that.

    Thanks for writing. I'm glad this blog is helpful. And, it's people, like yourself, that participate, that make it so. :)

    ReplyDelete
  5. Tom: I have some much newer FDM equipment in my lab, and in conjunction with my spring rapid prototyping class, we are comparing the results from a test block on the Cube and a very recent Dimension 1200es and uPrint machine. I also think the real problem is in software, not hardware, so updates to both the client (slicing adjustments, etc.) and firmware will be useful. I'll let you know what we discover. - Eric

    ReplyDelete
  6. That would be great Eric! Please keep us updated! Part of the joy of this journey is the discovery process. :)

    ReplyDelete
  7. Another vote for the software impact on accuracy. I ran your "HolePostMatchChampfer.stl" on both my Cube and Afinia with some interesting results. Holes ranged from -1.5% to -8.2% for the Afinia vs. -8.5% to a whopping -26% on the Cube. Posts were -1% to +2% Afinia vs. -3.2% to +14% an the Cube. But wait a minute. The numbers on the sample were fragmented on the Afinia but sharp and well formed on the Cube sample. Similar observation when printing stage two of the Eiffel Tower (from Thingiverse). Cube prints the thin railings interior and exterior. Afinia does not!
    Suggests to me there may be a tradeoff in software between accuracy and resolution?
    My statistics are suspect since they are from only one printing of the samples and only one set of measures with digital calipers. By the way the 5.5 inch sample was deemed too wide for the Cube using current Cube client. Had to trim a bit off the right end to print. Accepted by Afinia software and measured 5.529".

    ReplyDelete

countercounter