15 miles on the Erie Canal

Started by sboerner, January 25, 2018, 11:05:04 AM

Previous topic - Next topic

sboerner

QuoteOnly RT reflection under water will increase rendertime, so I always add another layer for the actual reflection with a minimum altitude (water), as child of the max altitude surface/distri layer with the darkening stuff.

Yes, I can see how that would help. Thanks much for the tip – let me give it a try.

QuoteI agree with Dune about the water line. That is a great idea. Otherwise if you need help with cpu power for rendering I might able to help. Got 2 ice-cold hexacore Xeon here doing nothing.

Thanks! I appreciate the offer. I think I'm OK. I'm at 3-4 hours for a 1920x1080 rendering, which seems reasonable given the scene. MPD=7.5 and AA=12.0. Most everything else left at default. Render times increase geometrically with larger sizes, though I think I can reduce MPD and maybe AA to compensate.

For larger renders of this type of scene, what kind of range for these values should I be looking at?

Past time here to add a second machine for rendering.


Oshyan

Given how much of the ground is covered by vegetation, which are ray-traced anyway, I'd reduce MPD. 0.6 is probably fine.

- Oshyan

Dune

Yes, for my big renders I standardly use 0.6 and AA6. Usually they end up on walls or big panels, so either needn't be much bigger, as resolution isn't higher then 100dpi, and for the biggest walls even 50dpi. I don't think AA of 12 would make much difference on that low resolution, and if people stand away a bit anyway.

sboerner

Thanks Oshyan, Dune. I appreciate your guidance here. I had to increase the AA on the small renderings (1920 wide) to avoid noticeable aliasing on the gently curved utility pole wires. I suspected that I would be able to reduce that as well as the MPD as the size of the renderings increase. I'll run some test strips with the lower values.

Oshyan

Yeah, that's a bit of a unique case, very fine lines are hard to anti-alias. I wonder if the new adaptive anti-aliasing options (New Curve and Robust Sampling) would help there. You can test lower AA with Max Samples in a crop of the utility pole wires to see what the *minimum* AA level is for smooth lines there. Then at higher levels if you use adaptivity (e.g. 1/64 first samples), it's just up to the adaptive sampling system to make correct choices for how to assign samples there.

- Oshyan

sboerner

Quote
Quote
Only RT reflection under water will increase rendertime, so I always add another layer for the actual reflection with a minimum altitude (water), as child of the max altitude surface/distri layer with the darkening stuff.

Yes, I can see how that would help. Thanks much for the tip – let me give it a try.

After I replied to this it occurred to me that the canal surface does not use a water shader. It's just a surface layer with a reflective shader in the node tree. There is no transparency. The minimal amount of transparency that I was getting with a water shader didn't seem worth the additional (and substantial) render time. So this suggestion is moot here? If so I'll file it away and will definitely use it on future projects.

sboerner

#217
Here are two high-resolution test strips, each 3840x432 pixels. I can see a few details now that need to be fixed. MPD=0.6 and AA=8. The wires look perfectly fine to me. But there's a noticeable moiré in the foreshortened vertical slats on the interurban car, so I assume AA will have to be increased to fix this. I'll run some more tests on this area.

Running a much higher resolution test strip at full poster-size resolution, 9000x1000 pixels. That should be ready by this evening.


sboerner

QuoteYeah, that's a bit of a unique case, very fine lines are hard to anti-alias. I wonder if the new adaptive anti-aliasing options (New Curve and Robust Sampling) would help there. You can test lower AA with Max Samples in a crop of the utility pole wires to see what the *minimum* AA level is for smooth lines there. Then at higher levels if you use adaptivity (e.g. 1/64 first samples), it's just up to the adaptive sampling system to make correct choices for how to assign samples there.

Upgrading my maintenance plan soon so I can run TG 4.3. Looking forward to trying this.

bobbystahr

That should make a difference with the(I've yet to experience)Path Tracing....
something borrowed,
something Blue.
Ring out the Old.
Bring in the New
Bobby Stahr, Paracosmologist

archonforest

i love the high res render. the amount of details are just awesome!
Dell T5500 with Dual Hexa Xeon CPU 3Ghz, 32Gb ram, GTX 1080
Amiga 1200 8Mb ram, 8Gb ssd

Hannes

Truly amazing!! The lighting looks so beautiful. It's quite difficult when the sun is more or less behind the camera, but you did a great job here. And the details are fantastic!

Dune

Awesome detail! Everything comes out very nicely. I don't think you need to bother about the RT reflection 'under' water indeed. But I'd check anyway. Most important issue is the highlights that may occur under the waterlevel, but likely there are not parts that would make them in this setup.
May I make a few more suggestions; dirty up the poles, especially the lower end, where they sink into ground. Rain splattered dirt/mud againts them and such. Same goes for the buildings at front. If the poles are a pop, it's easy, if it's all individual it's more work, if it's one object, well easy too.
And I would give the plant objects alongside the road just a tiny tilt of 3º or so. They have some kind of evenness to them (but this is really nitpicking).
And perhaps roughen up the stretched lines on the dirtroad a bit more, some bigger, some smaller?
And a last thing (sorry); give the waves a bit more variation, like wind patches. They seem very even.

sboerner

QuoteTruly amazing!! The lighting looks so beautiful. It's quite difficult when the sun is more or less behind the camera, but you did a great job here. And the details are fantastic!

Thank you, Hannes. The light has been a challenge from the start, because I want to use a legitimate angle. (It would be much easier if I could just put the sun wherever I liked.) The time frame is very narrow, early fall between 1914 and 1917. At the moment the angle is set for 6:45 a.m., September 24, 1915. The light seems a bit harsh to me now, but there will be scattered cloud cover that should soften it.

sboerner

QuoteMay I make a few more suggestions; dirty up the poles, especially the lower end, where they sink into ground. Rain splattered dirt/mud againts them and such. Same goes for the buildings at front. If the poles are a pop, it's easy, if it's all individual it's more work, if it's one object, well easy too.
And I would give the plant objects alongside the road just a tiny tilt of 3º or so. They have some kind of evenness to them (but this is really nitpicking).
And perhaps roughen up the stretched lines on the dirtroad a bit more, some bigger, some smaller?
And a last thing (sorry); give the waves a bit more variation, like wind patches. They seem very even.

These are all great suggestions, thank you. I've added them to my running list of things to fix. I also want to fix the diffuse shader on the railroad ties – it repeats and creates a pattern when the ties are foreshortened. Add flashing and soot stains to the post office chimney. And dirty up the interurban car's windows – they are far too clean. Detail work like this is easy and kind of fun, and the cumulative effect makes a huge difference.

I've run into a snag with the print resolution test strip. The renderer was chugging right along but hit a small area about 75 percent of the way through that stopped it dead in its tracks. I need to investigate the surface shaders of that area to see what's holding things up.