Found some time at last to get back to this model again.

So far, I've been able to stick to solid CSG modelling.

The boiler fairing was quite straightforward as expected, being 
basically just a spline lathe. The firebox fairing is a spline prism 
(actually two copies tilted a bit along the boiler axis) and to my 
surprise fits snugly with the lathe.

The smokestack is a blob construct; the upper rim turned out differently 
than planned, but I think it's perfect as it is. I'm a bit worried 
though whether I can use the same approach for the steam dome and sand boxes

Boilerplate "decoration" is just a bunch of rounded boxes and -cylinders.

Well, still a long way to Tipperary...

Post a reply to this message

Attachments:
Download 'steam.png' (709 KB)

Preview of image 'steam.png'

clipka wrote:

> Well, still a long way to Tipperary...

...crossin' some plains?

-Ive

Post a reply to this message

clipka <ano### [at] anonymousorg> wrote:
> Found some time at last to get back to this model again.
>
> So far, I've been able to stick to solid CSG modelling.
>
> The boiler fairing was quite straightforward as expected, being
> basically just a spline lathe. The firebox fairing is a spline prism
> (actually two copies tilted a bit along the boiler axis) and to my
> surprise fits snugly with the lathe.
>
> The smokestack is a blob construct; the upper rim turned out differently
> than planned, but I think it's perfect as it is. I'm a bit worried
> though whether I can use the same approach for the steam dome and sand boxes
>
> Boilerplate "decoration" is just a bunch of rounded boxes and -cylinders.
>
> Well, still a long way to Tipperary...

Cool work from what I can see.
Perhaps test render in a flat white texture so it's easier to see details.

-tgq

Post a reply to this message

Ive schrieb:

>> Well, still a long way to Tipperary...
> 
> ...crossin' some plains?

Absolutely :-)

Though they won't be this green...

I even got a perfect-fit quote from the book.


BTW, the steam dome and sand box "humps" were actually easier than I 
would have thought. Not exactly to the blueprints, but could you tell?

(For some reason though they remind me of German WWII helmets :-P)

I was really worried about those; now that I got them, I'm confident 
that I'll indeed be able to make it an all-CSG object.

Post a reply to this message

Attachments:
Download 'steam.png' (455 KB)

Preview of image 'steam.png'

Trevor G Quayle schrieb:
> Cool work from what I can see.

Thanks - though such things are usually supposed to be steaming hot :-)

> Perhaps test render in a flat white texture so it's easier to see details.

Well, I didn't want to bore you, but since you ask for it... I think the 
black version is more impressive though ;-)


(BTW, someone knows how to get rid of the artifacts at the smokestack?)

Post a reply to this message

Attachments:
Download 'steam.png' (492 KB)

Preview of image 'steam.png'

clipka schrieb:
> (BTW, someone knows how to get rid of the artifacts at the smokestack?)

Ah well, never mind...

(Lesson learned: When punching a hole into a blob with a smaller but 
stronger blob, make sure that it leaves a substantial wall thickness... >_<)

Post a reply to this message

Looks great so far.

Question for you, how do you control blobs into such a defined shape? =) I
always considered them limited to "organic" shapes and myself have virtually
never used them. But with what you are doing, they look quite powerful.

I applaud your detail and devotion to an sdl-only model. I'm trying to keep my
in process IRTC entry sdl and I sometimes think that whipping out the 3D modeler
would make some things a lot easier. There's something magical about a detailed
complex model that is built using pure sdl.





clipka <ano### [at] anonymousorg> wrote:
> clipka schrieb:
> > (BTW, someone knows how to get rid of the artifacts at the smokestack?)
>
> Ah well, never mind...
>
> (Lesson learned: When punching a hole into a blob with a smaller but
> stronger blob, make sure that it leaves a substantial wall thickness... >_<)

Post a reply to this message

P Brewer schrieb:
> 
> Question for you, how do you control blobs into such a defined shape? =) I
> always considered them limited to "organic" shapes and myself have virtually
> never used them. But with what you are doing, they look quite powerful.

I'm not sure yet. I'm still learning (you might also say "toying around").

One thing I did was write macros to get a different parameterization for 
the blob components. As my initial goal was no more than to get a 
rounded transition between cylinders, I wrote blob macros that allowed 
me to specify the "idle" radius of the blob component (i.e. the radius 
it will have when not anywhere near another component) and the distance 
from its "idle" surface to its limit of influence (to directly control 
the size of the transition region).

Apparently this must be a quite intuitive blob parameterization scheme, 
as I'm surprised myself how quickly I got good results not only for the 
smokestack, but also for the sandboxes, by basically just experimenting 
with the parameters.


> I applaud your detail and devotion to an sdl-only model. I'm trying to keep my
> in process IRTC entry sdl and I sometimes think that whipping out the 3D modeler
> would make some things a lot easier. There's something magical about a detailed
> complex model that is built using pure sdl.

Well, actually I'm more of a mesh-model person normally, using SDL only 
for quite simple models. And usually I don't even model the mesh objects 
from scratch, but just customize 3rd party models to fit my needs.

In this case, however, I feel that I'd be lost with this approach, 
because I expect to need a close-up of the engine, at a detail level 
that a pre-fabricated mesh-based model could probably provide only at a 
humongous polygon count.

I might have attempted to try selectively adding detail to an existing 
mesh model in the areas I intend to show close-up - but here's another 
problem: I'm not really sure yet what perspective to settle on. And 
given the highly technical, non-organic structures of a steam engine, 
being basically an assortment of cylinders, pipes, bars and bolts, I 
figured it would actually be easier to pull this off using CSG.

Post a reply to this message

Growing a chassis...

Problem #1 that had to be solved: Develop a framework to generate nicely 
beveled prism-like objects.

I wasn't sure whether I could pull this off by combining sphere_sweep 
and prism objects in some mathematically smart manner, so I ultimately 
settled for a nice framework to specify polygons with basically linear 
outline, but rounded off at the vertices with circular arcs of arbitrary 
radius. This also makes the whole thing easier to control.

The downside is that the heavy use of CSG already starts taking its toll 
on rendering times, so I needed a quality parameter to switch beveling 
off for quick test renders.

Problem #2 to overcome: Decipher the blueprints. (fortunately I do have 
some at all :-))

Post a reply to this message

Attachments:
Download 'steam.jpg' (80 KB) Download 'steam_1.jpg' (55 KB) Download 'steam_2.jpg' (51 KB)

Preview of image 'steam.jpg'

Preview of image 'steam_1.jpg'

Preview of image 'steam_2.jpg'

Impressive and interesting. Nice texture too. I am curious to see how long 
you will be able to sustain the CSG construction before render times become 
unmanageable...

Thomas

Post a reply to this message

Thomas de Groot schrieb:

> Impressive and interesting. Nice texture too. I am curious to see how long 
> you will be able to sustain the CSG construction before render times become 
> unmanageable...

I guess that will not happen at all.

I actually found that the detail stuff isn't so much of a problem - it 
was the big holes punched into the main struts that bogged down render 
times for some reason I didn't fully grasp. I'm using a different 
approach for them now, which is much more render-time-friendly: 
Previously, I had defined those holes separately from the struts' basic 
outline, each of these being a complex compound based on a prism shape. 
Changing this to use just one multi-loop prism made quite some difference.

Post a reply to this message

clipka a écrit :
> Thomas de Groot schrieb:
> 
>> Impressive and interesting. Nice texture too. I am curious to see how 
>> long you will be able to sustain the CSG construction before render 
>> times become unmanageable...
> 
> I guess that will not happen at all.
> 
> I actually found that the detail stuff isn't so much of a problem - it 
> was the big holes punched into the main struts that bogged down render 
> times for some reason I didn't fully grasp. I'm using a different 
> approach for them now, which is much more render-time-friendly: 
> Previously, I had defined those holes separately from the struts' basic 
> outline, each of these being a complex compound based on a prism shape. 
> Changing this to use just one multi-loop prism made quite some difference.

A prism is not the more effecient primitive: It's relatively hard to 
optimize a "free form" object.
Punching holes in any object result in bounding inefeciency, that 
increases the render time.
Each and every ray that encounter the bounding box of one of the struts 
needs to be tested against the strut AND every object used to make the 
holes.

With the complexe prism, you now only need to test that object.


Alain

Post a reply to this message

Just to show off the hardest single part so far.

The locomotive has 6 of these - plus 2 more that are similar, but a 
little less complex.

Still all pure CSG.

Post a reply to this message

Attachments:
Download 'steam.jpg' (63 KB)

Preview of image 'steam.jpg'

Alain schrieb:

> Punching holes in any object result in bounding inefeciency, that 
> increases the render time.

I know that, but...

> Each and every ray that encounter the bounding box of one of the struts 
> needs to be tested against the strut AND every object used to make the 
> holes.

... for some reason the many, many small cylinders I'm cutting away to 
round off inward edges don't seem to matter that much.

Post a reply to this message

clipka a écrit :
> Alain schrieb:
> 
>> Punching holes in any object result in bounding inefeciency, that 
>> increases the render time.
> 
> I know that, but...
> 
>> Each and every ray that encounter the bounding box of one of the 
>> struts needs to be tested against the strut AND every object used to 
>> make the holes.
> 
> ... for some reason the many, many small cylinders I'm cutting away to 
> round off inward edges don't seem to matter that much.

A cylinder is much more effecient than a prism to cut holes.
If you cut ceveral holes using prisms, you need to evaluate each prism 
for every rays.

Use two prisms, one for the faces and another for the edges, and add 
small cylinders to do the beveling between the faces and the edges.
You'll get an union with no or very few differences or intersections.

Do you use a prism and 3 cylinders to punch those rounded triangular 
holes. If you use a besier spline, you can have straight and round 
parts, and do away with a good part, even all, of the cylinders. Even if 
the prism takes slightly longer to render after that, you'll gain more 
by no longer having that difference.


Alain

Post a reply to this message

Alain schrieb:

> A cylinder is much more effecient than a prism to cut holes.

Um... yes sir, but how would you cut /straight/ lines with a cylinder?

> If you cut ceveral holes using prisms, you need to evaluate each prism 
> for every rays.
> 
> Use two prisms, one for the faces and another for the edges, and add 
> small cylinders to do the beveling between the faces and the edges.
> You'll get an union with no or very few differences or intersections.

'Scuse me, but you're not telling me anything new. I didn't want to go 
into that much detail, but that's exactly how I've been doing it right 
from the start, except that I didn't make use of multi-loop polygons.

(You're forgetting the spheres and tori, btw... you can't properly bevel 
in 3 dimensions without those.)

 > If you use a besier spline, you can have straight and round
> parts, and do away with a good part, even all, of the cylinders. Even if 
> the prism takes slightly longer to render after that, you'll gain more 
> by no longer having that difference.

While this may be so, splines are (a) more cumbersome to work with than 
the approach I'm using, as they are not so straightforward, to control; 
(b) very difficult to smoothly interface to standard box-&-cylinder 
geometry, and (c) I'd be lost with the math required to displace the 
spline for the beveling job on the outline (if you know what I mean).

Post a reply to this message

"clipka" <ano### [at] anonymousorg> schreef in bericht 
news:4ab90267@news.povray.org...
> Just to show off the hardest single part so far.
>
> The locomotive has 6 of these - plus 2 more that are similar, but a
> little less complex.
>
> Still all pure CSG.
>
>

wow

Thomas

Post a reply to this message

Thomas de Groot schrieb:

>> Still all pure CSG.
> 
> wow

Absolutely positively possible (at least for me) only due to a 
"rounded-off corners thingumajig" framework.

There's absolutely /no way/ I would go through all the pain of computing 
and hacking in the parameters for every individual torus, cylinder and 
prism in there :-P

Post a reply to this message