New Library Manager Approaching Trial Phase
The viewpoint renderer mixes pixels when the two colliding surfaces are
equal in position. Rotating the viewpoint of the object causes the
surface to flicker with both pixel colors (assuming different).
As little as 0.01 mm difference is enough for the viewpoint renderer.
The actual algorithm when rendering for export simply creates the
surface differently. The cut itself is likely clean in both cases.
Making sure that the viewpoint renderer shows the results you expect by
adjusting dimensions causes less confusion while designing.
Harvey
On 8/2/2024 3:13 AM, Peter Kriens via Discuss wrote:
Is there a reason you need to specify a slightly larger object for difference() than the parent object to make a clean cut?
If they are equal, the F5 renders it badly but the F6 manifold is more than fine. The code looks a lot cleaner and reduces potential collisions.
The question is probably already asked but I could not find it in searching the list. Pointers are thus also appreciated.
Kind regards,
Peter Kriens
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
Preview is black magic. It’s amazing that it works as well as it does.
Full render could in theory detect infinitesimal-thickness sections of the model and eliminate them, but I suspect that doing so efficiently would be a PhD thesis effort.
The thing is, the world is analogue, and openscad is digital. You can do
you wrt the ignoring of small values, but small is subjective. Then
there is the inaccuracies of floating point, representing 3d in a 2d
space, etc.. It all depends on your mindset. Openscad is more of a
construction tool, in my mind - 'difference' is drilling holes or
pocketing, 'union' is sticking stuff together, etc. I build objects from
parts, or machining from solid, in openscad, and reality. The main
difficulty (other than the syntax, and initially, terminology - cube
instead of cuboid, and cylinders that can be cones, etc.) is that it has
to be sort of written in the reverse order, to the way in which I think.
But, if I wanted, I could write functions/modules to overcome most of
that, or fork my own version.
I look at f5 rendering as an approximation of the object, and use it to
quickly identify small changes that I make. I do not believe i have
ever used f5 is as a final product - a coloured illustration.
If I was looking at something that already existed, and tried to model
it, then that is more difficult using openscad, in my way of thinking.
If you turn it inside out, instead of a shape with a hole in it, think
of it as as a space with stuff around it. Maybe linear_extrude is then a
better representation. e,g a cube with differenced cylinder, compared to
a linear extruded square with a differenced circle in it (but then 2D
does not exist, either).
On 02/08/2024 12:41, Peter Kriens via Discuss wrote:
Ok, thanks.
Just a bit surprised. I am a developer myself and it seems not too
hard to remove these types of faces that have a thickness less than
some very small value?
As a mechanic I know I have to handle real life and its messiness, but
as a developer it would be nice to take the more platonic world of a
digital model?
Am I still missing something?
Kind regards,
Peter Kriens
On 2 Aug 2024, at 12:36, Adrian Mariano via Discuss
discuss@lists.openscad.org wrote:
Actually if they are equal the F6 manifold is sometimes fine.
But not always. Sometimes you'll get a zero thickness layer over your
hole. I assume this depends on round off error. If the face is
orthogonal to the axes it will work OK. But if the face is at some
angle then you may end up getting the two "identical" faces in
slightly different places, such as by 10^-12 and if the hole cutting
ends up on the wrong side you then get the phantom extra face. This
doesn't have anything to do with using the z-buffer on your graphics
card---that's related to F5, which OP already notes will fail, but he
apparently doesn't care if F5 works.
On Fri, Aug 2, 2024 at 4:10 AM nop head via Discuss
discuss@lists.openscad.org wrote:
It's the same in real life. You have to start a drill above the
surface you are drilling and drill past the bottom of the object
to get a clean hole.
The reason you have to do it in OpenSCAD is it uses OpenCSG to
draw the preview and that uses the z-buffer of your graphics card
which has finite resolution.
On Fri, 2 Aug 2024 at 08:13, Peter Kriens via Discuss
<discuss@lists.openscad.org> wrote:
Is there a reason you need to specify a slightly larger
object for difference() than the parent object to make a
clean cut?
If they are equal, the F5 renders it badly but the F6
manifold is more than fine. The code looks a lot cleaner and
reduces potential collisions.
The question is probably already asked but I could not find
it in searching the list. Pointers are thus also appreciated.
Kind regards,
Peter Kriens
_______________________________________________
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
_______________________________________________
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email todiscuss-leave@lists.openscad.org
I think you will find the world is digital on the smallest scales because
everything is quantized. That, together with the fact that quantum stuff is
only probabilities until you measure it smacks of lazy evaluation, makes me
think we live in a computer simulation.
On Sat, 3 Aug 2024 at 11:56, Raymond West via Discuss <
discuss@lists.openscad.org> wrote:
The thing is, the world is analogue, and openscad is digital. You can do
you wrt the ignoring of small values, but small is subjective. Then there
is the inaccuracies of floating point, representing 3d in a 2d space, etc..
It all depends on your mindset. Openscad is more of a construction tool, in
my mind - 'difference' is drilling holes or pocketing, 'union' is sticking
stuff together, etc. I build objects from parts, or machining from solid,
in openscad, and reality. The main difficulty (other than the syntax, and
initially, terminology - cube instead of cuboid, and cylinders that can be
cones, etc.) is that it has to be sort of written in the reverse order, to
the way in which I think. But, if I wanted, I could write
functions/modules to overcome most of that, or fork my own version.
I look at f5 rendering as an approximation of the object, and use it to
quickly identify small changes that I make. I do not believe i have ever
used f5 is as a final product - a coloured illustration.
If I was looking at something that already existed, and tried to model it,
then that is more difficult using openscad, in my way of thinking.
If you turn it inside out, instead of a shape with a hole in it, think of
it as as a space with stuff around it. Maybe linear_extrude is then a
better representation. e,g a cube with differenced cylinder, compared to a
linear extruded square with a differenced circle in it (but then 2D does
not exist, either).
On 02/08/2024 12:41, Peter Kriens via Discuss wrote:
Ok, thanks.
Just a bit surprised. I am a developer myself and it seems not too hard to
remove these types of faces that have a thickness less than some very
small value?
As a mechanic I know I have to handle real life and its messiness, but as
a developer it would be nice to take the more platonic world of a digital
model?
Am I still missing something?
Kind regards,
Peter Kriens
On 2 Aug 2024, at 12:36, Adrian Mariano via Discuss
discuss@lists.openscad.org discuss@lists.openscad.org wrote:
Actually if they are equal the F6 manifold is sometimes fine. But not
always. Sometimes you'll get a zero thickness layer over your hole. I
assume this depends on round off error. If the face is orthogonal to the
axes it will work OK. But if the face is at some angle then you may end up
getting the two "identical" faces in slightly different places, such as by
10^-12 and if the hole cutting ends up on the wrong side you then get the
phantom extra face. This doesn't have anything to do with using the
z-buffer on your graphics card---that's related to F5, which OP already
notes will fail, but he apparently doesn't care if F5 works.
On Fri, Aug 2, 2024 at 4:10 AM nop head via Discuss <
discuss@lists.openscad.org> wrote:
It's the same in real life. You have to start a drill above the surface
you are drilling and drill past the bottom of the object to get a clean
hole.
The reason you have to do it in OpenSCAD is it uses OpenCSG to draw the
preview and that uses the z-buffer of your graphics card which has finite
resolution.
On Fri, 2 Aug 2024 at 08:13, Peter Kriens via Discuss <
discuss@lists.openscad.org> wrote:
Is there a reason you need to specify a slightly larger object for
difference() than the parent object to make a clean cut?
If they are equal, the F5 renders it badly but the F6 manifold is more
than fine. The code looks a lot cleaner and reduces potential collisions.
The question is probably already asked but I could not find it in
searching the list. Pointers are thus also appreciated.
Kind regards,
Peter Kriens
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
The entire issue can be avoided in many cases by linear_extrude-ing a 2D
object with a hole in it, e.g. replace
difference() {
cube(10,true);
cylinder(h=10,r=3,center=true);
}
with
linear_extrude(height=10,center=true)
difference() {
square(10,true);
circle(r=3);
}
which will be a perfect manifold in all cases irrespective of any
rotations or other transformations.
Bonus: it also renders /much/ faster.
On 8/2/2024 6:36 AM, Adrian Mariano wrote:
Actually if they are equal the F6 manifold is sometimes fine. But
not always. Sometimes you'll get a zero thickness layer over your
hole. I assume this depends on round off error. If the face is
orthogonal to the axes it will work OK. But if the face is at some
angle then you may end up getting the two "identical" faces in
slightly different places, such as by 10^-12 and if the hole cutting
ends up on the wrong side you then get the phantom extra face. This
doesn't have anything to do with using the z-buffer on your graphics
card---that's related to F5, which OP already notes will fail, but he
apparently doesn't care if F5 works.
not so easy if you'remaking a hole though an existing object, eg an stl
generated by someone else, you have to drill it right through, then.
On 03/08/2024 14:37, Douglas Miller via Discuss wrote:
The entire issue can be avoided in many cases by linear_extrude-ing a
2D object with a hole in it, e.g. replace
difference() {
cube(10,true);
cylinder(h=10,r=3,center=true);
}
with
linear_extrude(height=10,center=true)
difference() {
square(10,true);
circle(r=3);
}
which will be a perfect manifold in all cases irrespective of any
rotations or other transformations.
Bonus: it also renders /much/ faster.
On 8/2/2024 6:36 AM, Adrian Mariano wrote:
Actually if they are equal the F6 manifold is sometimes fine.
But not always. Sometimes you'll get a zero thickness layer over
your hole. I assume this depends on round off error. If the face is
orthogonal to the axes it will work OK. But if the face is at some
angle then you may end up getting the two "identical" faces in
slightly different places, such as by 10^-12 and if the hole cutting
ends up on the wrong side you then get the phantom extra face. This
doesn't have anything to do with using the z-buffer on your graphics
card---that's related to F5, which OP already notes will fail, but he
apparently doesn't care if F5 works.
OpenSCAD mailing list
To unsubscribe send an email todiscuss-leave@lists.openscad.org
On 03/08/2024 12:58, nop head via Discuss wrote:
I think you will find the world is digital on the smallest scales
because everything is quantized. That, together with the fact that
quantum stuff is only probabilities until you measure it smacks of
lazy evaluation, makes me think we live in a computer simulation.
In that case, AI is the real world. Anyway , at the smallest scale it is
all binary - it either exists or not, or it's alive or dead. Any ideas
more than that is over-complication, that is the problem these days.
Now, how many angels can dance on the head of a pin?
It depends on whether they are spiritual or have a material existence.
The answers are remarkably different.
A rather old question.....
Harvey
On 8/3/2024 7:20 PM, Raymond West via Discuss wrote:
On 03/08/2024 12:58, nop head via Discuss wrote:
I think you will find the world is digital on the smallest scales
because everything is quantized. That, together with the fact that
quantum stuff is only probabilities until you measure it smacks of
lazy evaluation, makes me think we live in a computer simulation.
In that case, AI is the real world. Anyway , at the smallest scale it
is all binary - it either exists or not, or it's alive or dead. Any
ideas more than that is over-complication, that is the problem these
days. Now, how many angels can dance on the head of a pin?
OpenSCAD mailing list
To unsubscribe send an email to discuss-leave@lists.openscad.org
Can we stay on topic please. Getting metaphysical can lead down paths
which are not good.
Michael
OpenSCAD Admin