Why is AMF not yet used for additive manufacturing?

[Abdul Hadi Azman]

Hi,

I've been doing extensive research recently about AMF and the needs for a better format for data representation dan exchange data for CAD/CAM in additive manufacturing.

I see that AMF fits nicely for the needs of additive manufacturing and is better than STL.

But, I can't really understand why it is not yet widely used for additive manufacturing. What are its disadvantages?

Can anyone help or give any references the problems facing AMF?

Thanks in advance

Abdul Hadi AZMAN

GSCOP, 

University of Grenoble,

France

+33649898307

[Hod Lipson]

Hi Abdul,

 

I am wondering about the same question. I think there are three reasons:

 

1.      STL files are not bad enough, and given the amount of investment in legacy code and files, they will stay for as long as possible

2.      AM Equipment manufacturers are dragging their feet on implementing AMF import (why? It’s not a priority, for reason #1)

3.      The standard is not freely available (although it is available to all equipment manufacturers)

 

--hod

 

Hod Lipson

Associate Prof. of Mechanical & Aerospace Engineering and Computing & Information Science

Cornell University, 242 Upson Hall, Ithaca NY 14853, USA

Office: (607) 255 1686 Lab: (607) 254 8940 Fax: (607) 255 1222

Email: Hod.L...@cornell.edu

Web: http://www.mae.cornell.edu/lipson

Administrative Assistant:  Craig Ryan  cd...@cornell.edu

Calendar: http://www.mae.cornell.edu/lipson/calendar.htm

--
You received this message because you are subscribed to the Google Groups "STL 2.0" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stl2+uns...@googlegroups.com.
To post to this group, send email to st...@googlegroups.com.
Visit this group at http://groups.google.com/group/stl2.
For more options, visit https://groups.google.com/groups/opt_out.

[Charles Overy]

One of the reasons that stl is pervasive is that it is dead simple to code. I can almost do it.

So the installed base of STL is huge.  

Also, stl is very predictable (predicably bad in many ways)

Many if not most of the population of people now using AF have arrived in the last 24 months.  They are still realizing that stl is part of the problem. 

I think AMF adoption will come and when it does it will come quickly.  

Charles 

--

You received this message because you are subscribed to the Google Groups "STL 2.0" group.
To unsubscribe from this group and stop receiving emails from it, send an email to stl2+uns...@googlegroups.com.
To post to this group, send email to st...@googlegroups.com.
Visit this group at http://groups.google.com/group/stl2.
For more options, visit https://groups.google.com/groups/opt_out.

--

---

Charles Overy Director - CEO Voice: 970 827 5274 ext 201 SMS: 970 827 5760 | Free: 800 448 8808 cw...@lgm3d.com | www.lgm3d.com to see our latest work

[Abdul Hadi Azman]

Hi,

I see.. Thanks very much Charles and Hod for your answers and opinions.

Does anyone else see other problems why AMF has not been able to replace STL yet as the de facto format for Additive manufacturing.

As for STEP, it seems, it is better and more equiped than STL for additive manufacturing, but why does STEP too has not been able to replace STL? What are the disadvantages of STEP?

Thanks,

Hadi,

France

Abdul Hadi AZMAN

GSCOP

0649898307

[Jeremy Pullin]

One of the things that is currently holding back a wider scale adoption of AMF is because it carries lots of features which are not applicable to many of the system manufacturers.

 

Consider the STL file. It carries the same value to all AM systems because it describes the form of the part. There are many different technologies which build parts in lots of different ways but needing the size and physical form described in the input file is something that they all have in common.

 

Now consider the AMF file. It carries information such as colour and material properties which is not applicable to all systems. Indeed it is only applicable in a minority of cases.

 

The net result of these differences is that the majority of system manufacturers will look at the AMF file and say “the extra functionality offered here is of no interest to us because it doesn’t give us anything we need 0that we haven’t already got with STL. If you look at extrusion based processes for example they do not produce parts with different material sections so do not need this information. They also do not produce full colour parts so neither do they need that information. That is just one example of a process that doesn’t need AMF I could also have chosen SLS, EBM, DLP, SLM in fact the vast majority of processes.

 

Don’t get me wrong I am not arguing the wide scale adoption of AMF for a single second. I’m simply trying to explain why so many system manufacturers are not exactly rushing to adopt it. None of the 3D printing manufacturers have the programming resources of companies such as Microsoft or Apple so software changes whilst by no means impossible to make do however mean that the resource required to make changes (as a fraction of their software budgets) are large and I’m very much afraid to say that if there is nothing in it for them and it offers nothing in terms of competitive advantage or unique selling points then you can perhaps see why the change to AMF sits so low on their already loaded priority lists.

 

Jez Pullin.

--------------------------------------------------------------------------------------------------
This email and any attachments are confidential and are for the use of the
addressee only. If you are not the addressee, you must not use or disclose the
contents to any other person. Please immediately notify the sender and
delete the email. Statements and opinions expressed here may not
represent those of the company. Email correspondence is monitored by
the company. This information may be subject to export control
regulation. You are obliged to comply with such regulations.

Renishaw plc (company number 1106260), Wotton Travel Limited (company
number 01973158) and Renishaw Advanced Materials Limited (company number 04632041),
are companies registered in England and Wales with a registered office
at New Mills, Wotton-under-Edge, Gloucestershire, GL12 8JR,
United Kingdom, Telephone +44 1453 524524.
--------------------------------------------------------------------------------------------------

[Markus Hitter]

Am 06.11.2013 13:26, schrieb Abdul Hadi Azman:
> why does STEP too has not been able to replace STL?

Complexity.

Most CAD applications have surface triangulation already, because that's
needed for 3D screen rendering anyways. Based on this I managed once to
write an STL exporter for Catia V4 within a single day.

Compare this to STEP, where one would expect something like a man-year
of coding to cover most features. Also, the STEP standard is closed/to
be paid, while STL is simple enough to be re-engineered within minutes.


Markus

--
- - - - - - - - - - - - - - - - - - -
Dipl. Ing. (FH) Markus Hitter
http://www.reprap-diy.com/
http://www.jump-ing.de/

[Hod Lipson]

I agree. AMF is also a triangulations (unless you add color and material). I bet you can write a basic AMF export in an afternoon too.
See open source code at STL2.org

--hod

-----Original Message-----
From: st...@googlegroups.com [mailto:st...@googlegroups.com] On Behalf Of Markus Hitter
Sent: Wednesday, November 06, 2013 9:44 AM
To: st...@googlegroups.com
Subject: Re: Why is AMF not yet used for additive manufacturing?

[Neil Sewell]

Hi all,

I'd like to add my 10p worth to this at the moment - I've tried to stay out of the conversation(s) a little bit as I'm unsure as to where I'm officially supposed to stand with respect to AMF.  I'm the conveynor for ISO TC261 WG4 which is all to do with data exchange for AM.  I knew of AMF from a long while ago and have been pushing to get it adopted as a joint ISO/ASTM standard - mostly because I think it just makes sense to have one worldwide standard!

However, there's another reason which actually I think addresses Jeremy's point rather well.

I can appreciate that machine vendors don't necessarily have resource to commit to creating new routines which read in different file formats - it can be an expensive, time consuming job and actually the fact that every CAD system in the world (probably!) can output STLs means that is a good standard for them to adhere to.  However, the other reason I believed that AMF was a good target for a new standard going forward is that it left backwards compatability open whilst extending functionallity for vendors who wanted or needed to make use of an ever growing feature set.

Putting it frankly, it's a very easy job to use an AMF file as a "bucket" for an STL file.  A couple of dozen lines of code will import an AMF, pull out the STL and present it to the rest of the system for parsing just like normal.  So vendors who didn't want or couldn't afford to do complex work on inputting new file formats were, I felt, at liberty to simply strip out the STL file held inside.  But, if they wanted to know that the part should be red in colour, or whatever else was appropriate for their specific technology, they could get at it.

My hope is that if ISO and ASTM can ratifiy AMF as a dual standard then there is a good reason that machine vendors and CAD vendors alike should start to consider adopting it.  The work needed to do the basic adoption of AMF is very light-weight - those who want to get more data can, and those who are savvy will probably just use freeware librarys and snippets of code from the internet to do some of the work.

Having said all this Jeremy's right - for lots of machine vendors, AMF gives no competititve advantage at the moment.  For those who can make use of the colour functionallity, or the different types of material, there is an advantage and so it is probably those companies who will first find most use for AMF.

Just thoughts!

Neil Sewell.

Conveynor for ISO TC261/WG4

[Leonid Raiz]

In reality things are not that simple. 

It is indeed easy for CAD vendors to output stl since they already do triangulation for display purposes. However display triangulation does not necessarily make airtight meshes and both OpenGL and Direct3D are capable of dealing with it.  Triangulation coming from CAD (even from solid modelers)  is typically airtight in the middle of individual surfaces but often has gaps and overlaps at surface edges. Fixing those imperfections is not easy, current generation of solid modelers is not really good at it. Current 3D printing software on the other hand can't deal with imperfect meshes. and printer manufactures spend considerable time developing software to fix stl files. AMF files would not have this problem because by design vertices are shared between triangles. Acceptance of AMF would make life much easier to printer vendors but would present big headaches for content generators. 

Does it explain why AMF implementation  is stalling?

[Markus Hitter]

Am 06.11.2013 14:58, schrieb Neil Sewell:
> Putting it frankly, it's a very easy job to use an AMF file as a "bucket"
> for an STL file. A couple of dozen lines of code will import an AMF, pull
> out the STL and present it to the rest of the system for parsing just like
> normal.

Sure. But this only works if the AMF is actually sort of an STL
container and contains flat triangular surfaces only. If a machine
vendor claims to support AMF by implementing it this way and a CAD
vendor uses these new curved triangles to reach sufficient accuracy, you
have unhappy users.

[Leonid Raiz]

​BTW - It is unrealistic to expect CAD vendors to output curved ​rectangles. They would much rather increase the number of flat rectangles. Neither it is realistic to expect printer vendors to be able to slice curved rectangles reliably. 

I think inclusion of curve rectangles into spec was a mistake and contributes to industry reluctance to accept AMF.  

[Markus Hitter]

Am 06.11.2013 22:07, schrieb Leonid Raiz:
> I think inclusion of curve rectangles into spec was a mistake and
> contributes to industry reluctance to accept AMF.

Fair opinion. Perhaps an open source all-triangle-types to
flat-triangles-only converter would help. A tool not simply stripping
all curvatures, but approximating curvatures with additional triangles.

Even if it rarely does something, printer manufacturers can move to the
safe side by simply running every file through such a tool.

[Martin Wicke]

I believe that the open-source AMF interpreter that Jon Hiller wrote
provides that functionality. It has its shortcomings (it's Windows only,
and it's a program, not really a library), but all the functionality is
there.

Probably the biggest contribution to adoption of this standard would be
an open-source, cross-platform library implementing (all of) AMF's
features with a sensible API.

As a thought experiment, how many people have written their own PNG file
reader/writer? Close to zero. That's the reason PNG is successful: you
don't have to write a whole bunch of code to use it.

Sadly, I am not at liberty to share my AMF implementation. But starting
from the code already there (in form of the editor), the main challenge
is to design a proper API that exposes AMF to programmers in a less
painful manner, extracting it, and making it into a cross-platform library.

Martin

[Jonathan Hiller]

A couple updates on the AMF tool Martin is referring to:

1) It has the framework for an API in place. All core amf functionality including import/export of stl and curved triangle division are accessible from a single header file include.
2) It is cross-platform. The main amf library is all plain c++. The gui is a thin qt layer on top of the main library. Windows and Linux are a go. Mac shouldn't be much different than linux, but I don't have one to try on.

If anyone who has more experience than I at how to package up an API wants to jump in, that's why it's open source. Otherwise I'll just be taking educated guesses at how and where people will want to use it as I have time.

  ~Jon

http://sourceforge.net/projects/amftools/

--

[Leonid Raiz]

One can either use open source or implement the AMF prescribed subdivision algorithm for curved rectangles. Not a big deal. But how much to subdivide? What criteria to use if originating surface is not available and there is no way to estimate deviation. Just pick an arbitrary number? 

Controlling accuracy of of manufactured model is a big concern is . From content producer point of view they want to know how accurate the printed model is going to be. CAD software is able to evaluate deviation of flat triangles from true CAD surfaces and if necessary adjust number of triangles. On the other hand AMF curved triangle is a mathematical abstraction. CAD vendor can't control or even evaluate how close they will be to ideal surfaces. 

If I am to print my model and take responsibility for it I would rather make more triangles and know exactly how closely they describe my model rather then reduce the number of triangles by making them curved and just hope that subsequent subdivision will make a nice surface that will be close enough (without a numeric measurement) to CAD surface. 

  

[Mark Burhop]

"BTW - It is unrealistic to expect CAD vendors to output curved rectangles. They would much rather increase the number of flat rectangles."

I'm just curious if this is speculation or you have talked to a CAD  vendor that said this. My experience has been that mechanical CAD vendors support the formats that customers ask for, And if they don't, someone writes a 3rd party translator. 

On the accuracy point, I would think a curved AMF triangle would be more accurate than an STL one and that you could calculate the tolerance as well. If that is not true, could someone explain it? I mean, getting the tolerance distance between a surface and a facet plane is probably easier than the distance form a surface and a curved facet but it hardly seems impossible. Maybe I'm not understanding.. 

If there is a problem with AMF acceptance, I don't think it is is the CAD world.

Mark

[Hod Lipson]

All CAD companies need to do in order to output curved triangles is to output the normals at the vertices. That's not very difficult... especially that these normals are already calculated for graphics rendering.

 

As is shown in the AMF specification, for a sphere, curved triangles are 1000 times more accurate than flat triangles. That means that you need 1000x times fewer triangles for the same accuracy, or get 1000x smaller error with the same number of triangles. For other curved surfaces the factor might be different, but I still expect a 10x-100x order improvement.

 

But as long as equipment manufacturers don't import AMF, there is little incentive for anyone else to support it.

 

--hod

 

 

 

From: st...@googlegroups.com [mailto:st...@googlegroups.com] On Behalf Of Mark Burhop
Sent: Wednesday, November 06, 2013 6:30 PM
To: st...@googlegroups.com
Subject: Re: Why is AMF not yet used for additive manufacturing?

 

"BTW - It is unrealistic to expect CAD vendors to output curved rectangles. They would much rather increase the number of flat rectangles."

[Hod Lipson]

The specification says each triangle needs to be subdivided exactly five times. That means that a single curved triangle is temporarily converted into 1024 flat triangles for slicing.

 

--hod

 

From: st...@googlegroups.com [mailto:st...@googlegroups.com] On Behalf Of Leonid Raiz
Sent: Wednesday, November 06, 2013 5:48 PM
To: st...@googlegroups.com
Subject: Re: Why is AMF not yet used for additive manufacturing?

 

One can either use open source or implement the AMF prescribed subdivision algorithm for curved rectangles. Not a big deal. But how much to subdivide? What criteria to use if originating surface is not available and there is no way to estimate deviation. Just pick an arbitrary number?