AoI Alternatives

Hi jonored,

... read here [brlcad.org] side 19, point 6 - difference between replicating or referencing objects, maybe it's helpfull?

Viktor

I used brl-cad once upon a time. I highly recommend it, especially now that I note it can output stl files. And I agree that reprap will need a CAD package that is friendly to engineers.

james Wrote:
-------------------------------------------------------
> And I agree that reprap will
> need a CAD package that is friendly to engineers.

LOL! This reminds me of the big turf war that was going on in the Pentagon over who ought to be allowed to operate unmanned aerial vehicles (UAV's). The air force wanted complete control over UAV's and wanted only qualified fighter pilots running 'em. The army was using just about anybody from the ranks who could run a games controller. The recent firings of the top two officers running the Air Force has, I hope, ended that little dust-up.

Anyhow, this "it's got to be something engineers like" debate on CAD choices strikes me as same same. Maybe you guys think that anybody using Reprap ought to become more like professional engineers in the way they think or approach problems. For my money, however, that's the surest way to drive Reprap into obscurity.

Edited 1 time(s). Last edit at 07/11/2008 12:12PM by Forrest Higgs.

Forrest Higgs Wrote:
-------------------------------------------------------

> Anyhow, this "it's got to be something engineers
> like" debate on CAD choices strikes me as same
> same. Maybe you guys think that anybody using
> Reprap ought to become more like professional
> engineers in the way they think or approach
> problems. For my money, however, that's the
> surest way to drive Reprap into obscurity.


You are vastly oversimplifying the objections to AoI if you've parsed it down to "engineers don't like it." I urge you to re-read some of my earlier comments in this thread if you don't know what I'm talking about.

That being said, how far do you think the open source software movement would have come if it wasn't supported by anyone who wrote software as a day job? Or if the major open-source projects (eg. Linux) forced their developers to use some arcane programming language because C just seems too commercial, and using it would give an unfair advantage to professional programmers?

As I've said before, I'm not an engineer, I'm just a hobbyist. But I think that a huge factor in the success of this project will be the number of engineers or hobby engineers it can recruit to collaborate on complex, useful, and (importantly) shareable open-source designs that haven't been developed yet. Maybe I'm committing heresy, but I would be willing to bet that your average RepRap user (or cheap 3D printer user in general - RepRap is just the tip of the iceberg) will have little interest in putting forth the effort to learn how to produce 3D models of their own, no matter how easy the tools are to use. They'll search the net for something that does what they want it to. Your average EeePC buyer isn't doing development on Firefox - they just want something that does the Right Thing.

Kyle Corbitt Wrote:
-------------------------------------------------------
> Forrest Higgs Wrote:
> --------------------------------------------------

>
> Maybe I'm committing heresy, but I
> would be willing to bet that your average RepRap
> user (or cheap 3D printer user in general - RepRap
> is just the tip of the iceberg) will have little
> interest in putting forth the effort to learn how
> to produce 3D models of their own, no matter how
> easy the tools are to use.
>

Well, you've certainly got a right to your opinion and indeed, many people may be content to passively print out things that other people have designed. I'm hoping for considerably more than that, though. After all, when word processors and cheap dot matrix printers emerged a few decades ago, the desktop publishing revolution, if you will, we didn't see people being content to download and print out professionally generated content. We saw an explosion of new authors and content generation. I'm hoping for the same effect with 3D printing.

The other thing to remember is that the PC-based word processing software that eventually emerged was nothing like the methods used by professionals before the revolution began, nor was it even like the early publishing software brought out by Wang and later Apple.

Edited 1 time(s). Last edit at 07/11/2008 02:18PM by Forrest Higgs.

I have to agree with Forrest, the early days of DTP vs command driven or embeded code stuff that profesional typesetters use, was is sorta where we are now.

Early DTP was realy bad, but look where the software is now.

Remember clip art libraries? I always avoided them, and I dont intrract enough with the corporate world to know, but I suspect it never took off in the way original developers thought it would. Content is generated now by digital camera and photoshop

Future Reprap users will use 3D scanners in the same way.

What we are realy looking for is a program that can edit chunks of 3D scanned meshes - lop bits off (~crop), bung holes in (~redeye removal) and bung it on the side of a scan grandma sent from Australia last week.

I was thinking - I like engineering CAD - I like dimensions. But only because the things I make need to fit to other things in the real world. If I could scan my wardrobe legs I wouldnt need to measure them, just define a new 'floor' and fill the gaps using 'WorldShop' program using the scan as a reference, and print.

Did you see what I did there? Do I claim copyright on the new software 'WorldShop' ? - Not just for modifying photos - for modifying the whole world.

Both

Both

Both

Both are needed. There needs to be a good entry level tool and their needs to be a tool or tools that experts can use. Ideally this would be the same tool but 3 pages of "discussion" does not seem to point in that direction.

Let's use the right tool for the right task. Already there have been a lot of good suggestions and named tools brought to the table. The reprap wiki would be better for having a rundown of the available tools and a blurb about each as well as one or two "officially" suggested tools for the person new to the 3d design world. IMHO it should be a front page link on reprap.org "3d design tools".

Viktor,

Imo, the problem with most of those programs is they're not solid modellers -- they're designed for working with meshes, which don't even have to define valid volumes, and tend to sacrifice a lot of quality, e.g. where a cylinder which was intended to be smooth would be manufactured with facets, simply because that's how mesh cylinders are.

Also, as-is, BRL-CAD's modelling interface, mged, requires a good bit of study to be effective with; it's no surprise that casual experimentation was not promising.

jonored, I imagine such a macro might exist (although I have yet to reach it in my expeditions through the documentation); I can't imagine anyone, say, modelling every segment of the treads on the Bradly model I linked by hand. If all else fails, mged's command line interface seems friendly to use by a script, and as it is CSG-based, meaning everything is stored as a set of boolean-combined primitives, making changes to a "macro" set of objects should cascade nicely without any extra effort, avoiding the repetition of operations you describe.
Edit: Viktor beat me to it, and did better research to boot. Looks like "macros" are supported after all.

Edited 1 time(s). Last edit at 07/11/2008 04:34PM by Ralith.

Just for fun, it's a good exercise to try to judge the size of the user communities for BRL, AoI and Blender using Alexa.

Blender has like a magnitude more people working with it.

Forest,

I'm actually a major Blender fan, but here I'm going to have to say that it's technically inappropriate. It's meant for CGI, not CAD, and its features reflect that. It is relatively polished, and very popular amongst opensource CGI people, but that's not really what we're looking for, and I think we'll only lose in the long term if we go for something like that. I'll steal your wonderful desktop publishing metaphor here: Would there have been such a revolution if, instead of word processors and desktop publishing tools such as MS Word or Indesign, people were instead encouraged to use things like the font tool of MS Paint or photoshop? Both types of software are useful, and you can even use one as the other in a pinch, but using the right tool for the right task almost always produces much better results. Mesh editing, as useful as it is, is not appropriate for CAD/CAM applications. Even storing data as a mesh, as STL does, leads to loss of precision.

Really, though, since when was a popularity contest an effective way of determining quality?

Ralith Wrote:
-------------------------------------------------------
> Forest,
>
> I'm actually a major Blender fan, but here I'm
> going to have to say that it's technically
> inappropriate. It's meant for CGI, not CAD, and
> its features reflect that. It is relatively
> polished, and very popular amongst opensource CGI
> people, but that's not really what we're looking
> for, and I think we'll only lose in the long term
> if we go for something like that.

Well, I must say that my initial assessment of Blender was that was indeed a CGI tool. The problem with that was that there are so damned many ways to use blender that what you had to do to get precision 3D Cad out of it wasn't obvious. Once Kaizer put me on to this set of tutorials, however,

[www.rab3d.com]

I found that I had to rethink my initial assessment.

> Really, though, since when was a popularity
> contest an effective way of determining quality?

There is a little thing called critical mass. If you don't have a sufficient number of people using and developing a technology it doesn't tend to get any better and instead sort of withers on the vine, so to speak.

Hi Ralith,

... yes, most programs i work with, are mesh-oriented.

They were designed with the main target of good looking screenwork and CG-animation.

Here is a sample where i designed a short clip with an ant toying with assemblies - constructing parts and building the scene:


Rendered scene:



I made this in maybe half a day with a predefined 3DS-object of an ant i reassembled for converting in an animated 'rag-doll' and the other parts from scratch.

Simpler objects made from some primitives and boolean combinations (see the atached images "Lego-l.jpg" and "3D-Monitor-l.jpg") needs some minutes or maybe an hour to construct in every accuracy i want (and my PC can handle).

So for 'normal' use i didn't see the need for 'recyling' geometries with parametric solids and CSG-programs - and here the meshed (and mostly cheaper) programs are more common then CSG- and engineering CAD-programs ...

Viktor

Forest,

That would make sense if BRL-CAD were some run of the mill garage project. It's not. I suspect that the U.S. Army alone provides more than a critical mass, and the website suggests that it's got multiple users in the industrial sector, too. There's more to this project than website visitors; until recently, it didn't even have a respectable internet presence; for most of its history it's been an entirely real-world based piece of software.

All that aside, if this thing's made it for 20 years *without* an internet following, now that it's the first true open source CAD suite, chances are it's only going to get better.

Now, I'm well aware that blender can do relatively precision modeling when correctly used; so can almost any modeling tool, no matter its design purpose. However, the question isn't whether it can model something of specific desired dimensions. Blender, as a mesh modeler, is fundamentally inappropriate. In almost every case, a CSG model will be infinitely more accurate than a mesh version, no matter how high-resolution it is; it's perfectly analogous to vector vs. bitmap. Bitmaps are great for things like graphics and logos and photos and everything we see every day on the web, but there's a reason that the lasercutting diagrams are done in SVG, a vector format. No matter how high resolution a bitmap or mesh may be, a vector or CSG model will by definition provide *infinite* resolution. The only case where this is irrelevant is objects which are entirely composed of infinitely sharp edges, and may thus be accurately represented by a set of planes, and thus may be represented by a mesh or CSG with equal accuracy. Meshes are thus universally of equal or lesser accuracy than CSG equivalents.

CSG has other nice benefits, too; if you want to make a shape by, say, subtracting a circle from a sphere, you can do that (approximately) in a mesh, or (accurately) in CSG. However, if you want to then lower the radius of the cut-out after the fact, in a mesh you'll have to start from scratch again; make a new box and a new sphere of the desired radius. In CSG this is much simpler; all you have to do is modify the radius of the original sphere, meaning that instead of rebuilding your entire object you simply change a the radius value of one existing portion. This might not seem like a huge big deal with such a simple example, but what if your object is far more complex? Starting from scratch could mean hours of work you'll have to repeat.

To clarify, I'm not arguing Blender vs. BRL-CAD so much as meshes vs. CSG. I stand behind BRL-CAD in particular simply because it is the only complete open source CSG CAD system I'm aware of, and even were there others, BRL-CAD stands a good chance of remaining superior, thanks to the proven nature implied by 20 years of use by at least one highly demanding user, with active development continuing throughout. Were we debating mesh modelers, instead of CAD systems, I'd be behind you on Blender straight away; it's a great tool for what it does, but in the end is prevented from becoming a truly powerful CAD system because of its fundamental nature as a mesh modeler.

Viktor,

As I state above, we do have a free CSG editor; BRL-CAD. Yeah, it's UI is unfriendly, but that's being actively remedied; there's quite a bit of interest in making it more people friendly. I realize that CSG is not absolutely necessary for many simple and presumably common tasks, but it's what we should standardize on, lest we be beset by issues whenever someone tries to do something remotely complex.

Edited 1 time(s). Last edit at 07/11/2008 05:10PM by Ralith.

CSG has the potential advantage that traversing curved surfaces could be described algorithmically, ie using gcode arc drawing codes, rather than expressed as a series of short straight lines, so you'd get a nice smooth finish without faceting. Possibly.

I had thought that it might reduce the final print job size too, but of course the bulk of any layer is probably going to be infill instructions.

Ru, actually, this is an interesting point. Have you seen wildcat ?
[wildcat-cad.blogspot.com]
the author proposed generating slices from his software. I assume the slices could end up with nice b-spline edges. Are there G codes for those? hmm. Would be nice though.

Hi Ralith,

... i want to make a second try with BRL-Cad and Blender when i'll find some time ...

In the meantime i use what's there, as most people around.

I received most of my programms for free as outweared old software or i've found them for 4 to 7 Euros as atachment to PC-magacines and such ...

Some of my friends work with professional CAD-programs, others didn't use any or didn't even know how to make a 3D-object

In my actual surrounding maybe 2 percent of the people i know are capable of some computer-artwork, but none of them is willing to learn a completelly different piece of software, when they used to work with Acad or PoEngineer.

So they're more focussed on finding some free Acad-clones with (hopefully) the same GUI/command-list and didn't want to play around with Blender, BRL-Cad or such.

Only one friend was experienced with CGI -- and he seems to be an unchangeable fan of Lightwave

Viktor

Ru,

That's part of my point. Data which simply cannot be retained by a mesh is trivial to keep in CSGs. I believe slicing might also be simplified as there's no worry about 'holes' as you can have in a mesh; any volume is guaranteed to actually be a volume.

Richard,

I'm pretty sure Gcode can only represent portions of circular arcs. A somewhat intelligent pathing algo could use sets of these to approximate almost any curve to a pretty reasonable degree, though, and there's always the possibility of implementing custom M-codes with more precise support for that sort of thing.

Ralith :
A bit of googling found
[www.makino.com]
But I dont have time to read it ATM.

BRL-CAD's original focus was on the the "will this artillery barrel explode or not?" side of things, and not so much on the drafting and design side, as per this
usage diagram:
[brlcad.org]

From the website, it seems as though they are working on the drafting and design side a bit, but I don't think it is mature and user-friendly yet.
[brlcad.org]

Sebastien,

I'm not sure where you got that impression; it's about as mature as CAD suites get. Admittedly, there's no question about its lack of user-friendliness, but that's being actively remedied by a GSoC project.

BRL-CAD's original focus included things like radar profiling and ballistic vulnerability assessment; things like "will this artillery barrel explode or not?" are much more complicated questions which have never been covered. Said original focus is conducive to the creation of a good CAD suite, and BRL-CAD is developed with flexibility in mind (thus the relatively trivial nature of creating a new GUI). Finally, given the already-discussed importance of CSG for our uses, and as it's the only mature, let alone industry-proven open-source CSG modeling suite available, I think the fact that the current interface has a sharp learning curve is a relatively minor issue, and a soon-to-be-remedied one at that. Given the choice of short-term usability issues (which might well be resolved by the time any decisions made here come into effect) associated with BRL-CAD versus the long-term functionality issues associated with an immature system or a mesh modeler, I don't believe there's any question as to which is preferable.

Edit: Had a thought: Most users are going to be coming in with no 3D modeling experience whatsoever, and I'm not aware of any powerful modeler, CSG or otherwise, open and free or otherwise, which is really userfriendly and lacks a steep learning curve; perhaps this shouldn't be such a major concern.

Edited 1 time(s). Last edit at 07/11/2008 08:25PM by Ralith.

I meant that the interface for designing things wasn't mature yet. (Although I should download it and give it another try before saying that.)

Frankly, I should buy Rhino and run it under WINE and get used to a professional tool while waiting for things to improve.

In other CAD news, CollabCAD is promising.
[www.collabcad.com]
The Indian government is helping fund a project to create a closed source, inexpensive CAD program that will run on Windows and *nix. Might work out. Wish it was going to be GPL.

I'm not sure how many of you saw this link, but I think that this is interesting [brlcad.org]

As a CAD user, those improvements practically have me jumping out of my seat. If they're able to integrate any single one of those with their project it will be huge, but if all four get into the trunk it will absolutely revolutionize the usability of this program.

Sebastien,

Ah, that makes more sense. It's a very powerful interface, but, having made my way through the first few tutorials, I agree that as-is it's probably the most awkward out there for a new user who's not willing to spend hours familiarizing himself with the documentation. However, based on that same experience, I'll say that it's nevertheless a very powerful interface, and worthy of its long history. That said, it's somewhat inappropriate for a project like reprap where we want to ensure a low barrier for entry to encourage rapid adoption and extensive modification of designs, which is why the GSoC project for a better GUI is so exciting. If that achieves the success that seems likely, we may well be left with the perfect toolset: a lossless and compact CSG file format accessible through a powerful backend which has both a straightforward and friendly and a powerful but arcane frontend, enabling users to work with whatever fits their needs and time the best.

*ahem*

Kyle,

I'm excited to hear that someone with real CAD experience is interested too. I may think it's the coolest thing since sliced bread, but I'm just another geek with a lot of spare time; my conjecture could be a lot less solidly grounded than I think it is.

Having talked with the folks on #brlcad, the sort of thing I'm talking about is somewhere between partially implemented and on the list of things to do sometime, and closely related to stuff that they're doing for the SoC. I suspect that we may end up with BRL-CAD just because of it becoming the best tool for the job in a relatively short timeframe.

Of course, it won't exactly be hard to deal with, because it's got an explicit design goal of easy import from and export to almost any format out there. Lots of converters, templates to write more if we need them.

The only possible problem with this is that the latest version is 7.12.4 but the lastest Linux download is 7.8.4. If we are going to use new features, we need to be sure that they are going to be available to Linux users as well.

davidgoodenough Wrote:
-------------------------------------------------------
> The only possible problem with this is that the
> latest version is 7.12.4 but the lastest Linux
> download is 7.8.4. If we are going to use new
> features, we need to be sure that they are going
> to be available to Linux users as well.


Considering all development for this package took place exclusively on Linux or other Unix-based operating systems for the vast majority of its existence (it was ported to Windows relatively recently), I assume that building from source, at least, should still work to get the latest version on Linux. But, that's just a guess, I haven't tried.

Lord! Linux and "building from source"! I'd rather get tangled in poison oak.

Forrest Higgs Wrote:
-------------------------------------------------------
> Lord! Linux and "building from source"! I'd
> rather get tangled in poison oak.

To build 90% of programs in Linux, you type 3 commands:

./configure
build
build install

and that's taking the slow and steady courteous approach.

Try that in windows

Yeah, it's that they haven't been keeping their binary package for debian up to date. That doesn't matter much to the devs, or to the powerusers/testers, or the gentoo users, etc... It seems worth pointing out that the "linux download" you are referring to is a debian-derivative-specific binary package that will work on only x86 machines. To include binary packages for all configurations in use would be impractical. Partly for that reason, the standard linux installer /is/ the source tarball, and normal practice is for /distributions/ to take the tarballs and compile them for their customers. It's not unusual for a developer to not keep the convenience packages for distributions that don't include their project in their repositories up to date, and there's no real reason for them to bother - that's /not their job/. That said, if anyone wants an ebuild for BRL-CAD, I did one once and can do one again if it's not already done.

greenarrow Wrote:
-------------------------------------------------------
> Forrest Higgs Wrote:
> --------------------------------------------------
> -----
> > Lord! Linux and "building from source"! I'd
> > rather get tangled in poison oak.
>
> To build 90% of programs in Linux, you type 3
> commands:
>
> ./configure
> build
> build install
>
> and that's taking the slow and steady courteous
> approach.
>
> Try that in windows


LOL! I'm reminded of the cartoon years ago where Bill Gates and the Mac freak are together and the Mac freak says, "name something that I can't do on my Mac that you can do with Windows". Bill replied, "right click".

First off, once you do a build as a developer and let your development environment do a release (I'm talking about Visual Studio here), that's it. End users never do builds. Indeed, when you try to talk about a build to most windows users (not developers) they don't have a clue what you're talking about. Problem I've seen over and over in these forums and before these forums existed in Reprap is that it seems like EVERYBODY including end users of Linux lives and breathes builds. It seems weird, creepy, geeky and worst of all silly.

As to your "try that" taunt. I've been using Visual Studio to bang out a quick app this morning. Ordinarily, I run in my IDE, but when I want to run fully compiled code I simply go to the Build menu and select "build", "rebuild" or "publish" as indicated.

I'm trying to remember the last time I got an error code on a build. Mind, you don't try to build when you have outstanding errors in debug mode, so that's probably a bit unfair I suppose.