CGS versus Brep
Reinoud Zandijk
out of curiosity as a (systems) programmer, i'm currently evaluating a
number of (solid) modellers and their implementation techniques. What i am
seeing is the almost absolute dominance of boundary representation
schemes. I would expect to see more CSG based modellers since they are far
better suited for CAD/CAM IMHO. Shelling, filling and rounding etc. can
also be expressed in CSG so why the dominance of B-rep? And are there real
CSG modellers around still even? BRL-cad comes to mind but it comes over
to me as ancient and with a horrible user interface.
SolidWorks, CATIA, IronWorks, Unigraphics NX to name a few all seem to
work on the ParaSolid kernel with different user interfaces and tools and
thus are basicly boundary based though annotated on how the solid was
created.
Well, since you folks are working with CAD on a daily basis, do you think
that a non B-rep CSG based system could be feasable or would even be
desirable if confined to industrial design and manufacturing i.e. not for
pretty pictures but for machinable parts? Would NURBS free form
primitive support be crucial? or is it overrated in this field of work?
With regards,
Reinoud
ber...@swiftdsl.com.au
CATIA is not Parasolid based, they've got their own, but anyway...
CSG has serious limitations. Even without going too far, a simple
fillet is a big task when you want to construct it from primitives. B-
rep handles complexity well. Also, there are tasks when any system can
only approximate, like intersection of free form surfaces, etc.
Today's practical CAD applications with robust solid modeling are
mostly feature based, because the natural thing is to build things
from / with features, not from primitives. If you want free form
stuff, then your primitives are not so primitive anymore.
For industrial strength applications one needs something robust, not
something what looks good in theory.
Parasolid is well developed and it does more than just a kernel, such
as preparing geometry for graphics, etc. You can sign up to get a
developer's kit, develop your app around it and pay a modest license
fee on it (Same with ACIS). Before I forget, it is also well proven.
But if you want something different look at OpenCascade.
Regards
Attila
Edward T Eaton
different user interfaces. Excepting interface changes and certain
modeling algorithms (like SolidWorks Boundary and Surface Fill), they
are mostly interchangeable. And even these goodies are modifications
beholden to the BREP. And when it comes to NURBS, there are still
unrealized applications.
The only thing I could find on a quick search about CSG based
modellers is that ' constructive solid geometry is performed on
polygonal meshes'.
Please let me know if I missed something, but as described meshes are
crap dead-ends for manufacturing. Give that up - they are old, dead
technologies, way behind the curve.
The thing that bugs me as a professional product developer is that the
BREP, as a product, feels like a dead end.
It's good - we get our jobs done, don't get me wrong about that. There
is a lot of history behind that statement that would take a long time
to get into.
But after being at SolidWorks World 2008 (and prior history) for a
wild night where I was meeting with a number of folks from a number of
established and startup companies it was clear that most CAD we see
today is being driven by folks pursuing the still unrealized aspects
of the Computervision dream.
At the core of everyone's development is the BREP, and they are all
chasing the same dog.
As someone who is very conscious of the BREP, I know that most of our
modeling pain come from the limitations of the BREP.
I can say this... 20 years from now we won't be talking BREP anymore
(excepting legacy applications). There are new techs I've been hinted
to, and there better be lots of new techs that "I am too small a
player" to be hinted to. The BREP is near its end. My gut says that
in five years we will see the first real commercialization of the new
paradigm, and in 20 years we will be saying BREP... what's that???
Reinoud Zandijk
> The only thing I could find on a quick search about CSG based modellers is
> that ' constructive solid geometry is performed on polygonal meshes'.
CSG modelling is not a new paradigm, it was implemented in PADL-2 before
CATIAv1 was even out but its not used that much since rendering CSG
models is inherently slow; its more like raytracing. A CSG description is
purely mathematical; every surface patch is mathematically described and
when combined its mathematically soundly connected. Some applications like
SolidWorks (but also Blender) seem to do boolean operations on polygon
meshes whereas IMHO BRep/polygon meshes should only be used for displaying.
> Please let me know if I missed something, but as described meshes are crap
> dead-ends for manufacturing. Give that up - they are old, dead
> technologies, way behind the curve.
I think you might be right :-S
> At the core of everyone's development is the BREP, and they are all
> chasing the same dog.
> As someone who is very conscious of the BREP, I know that most of our
> modeling pain come from the limitations of the BREP.
Also in SolidWorks? What problems do you run into?
> I can say this... 20 years from now we won't be talking BREP anymore
> (excepting legacy applications). There are new techs I've been hinted to,
> and there better be lots of new techs that "I am too small a player" to be
> hinted to. The BREP is near its end. My gut says that in five years we
> will see the first real commercialization of the new paradigm, and in 20
> years we will be saying BREP... what's that???
Would you care to elaborate a bit on that new paradigm? if by email if you
prefer?
With regards and thanks in advance,
Reinoud
Reinoud Zandijk
On Wed, 20 Feb 2008 15:46:43 -0800, bertok wrote:
> CATIA is not Parasolid based, they've got their own, but anyway... CSG has
> serious limitations. Even without going too far, a simple fillet is a big
> task when you want to construct it from primitives. B- rep handles
Well, without a form of macro's it sure would be a big task, but i was
assuming that a fillet on a CSG based system would be implemented using
CSG modelling but with the UI of today's fillet operations i.e. automated
generated operations that together perform the fillet.
> Parasolid is well developed and it does more than just a kernel, such as
> preparing geometry for graphics, etc. You can sign up to get a
> developer's kit, develop your app around it and pay a modest license fee
> on it (Same with ACIS). Before I forget, it is also well proven. But if
> you want something different look at OpenCascade.
Thanks for the link to OpenCascade! I'll try it out too though i think its
BRep too. True, Parasolid looks like well developed and it does do a good
job for todays systems.
With regards,
Reinoud
