Maxwell Render | Official public forum

I am revisiting an old issue that I haven't tried for a long time.
It seems to still be there despite several new versions...

I'm trying to make a good sea surface.

The geometry in this image is extremely simple: Two planar rectangles. One is the water surface, which has a standard water material plus MPDM, the other one is the bottom which sits a few meter below and has a simple lambertian material with a colormap.

There is however some kind of artifact going on with the micropolygons. It looks like they sometimes kind of oscillate their angle for every two polygons.
This kind of pattern is absolutly not there in the map image, which is very high resolution and very smooth (3000x3000x24bit PNG).

Top is lowres version of final image, bottom is magnification of one of the problematic areas. (It hasn't rendered very long, but this is not any artifact that disappears with longer cooking)



Of course the problem is less visible if I increase the precision of the MPDM, but this image is already at my limit in terms of the voxelization time needed (about 2-3 minutes on 8 core 3.2GHz machine).
This precision would be fine if it wasn't for these strange patterns.

This is with 1.7.1 64bit on WinXP Pro/64.

Any ideas on how to get rid of them?

/j

Good looking sea surface but how many subdivs has your plane got ? you need more than 2 polys on such a plane to get good enough results. Have u tried adaptive displacement ? that might also help..

/ Max

I would try:

-increasing mesh density (as Max said); it not only may correct the problem but also increase rendering speed;
-if speed gets better maybe you can also increase precision;
-also turning texture filtering on in texture picker may help if it has something to do with pixelation of the texture.

Maximus3D wrote:but how many subdivs has your plane got ?

500x500 subdivs in this image.

Actually the number of subdivs does not change anything.

More subdivs -> fewer texture tiles AND lower subdiv precision -> same result and same voxelization time.

Fernando Tella wrote:turning texture filtering on in texture picker may help if it has something to do with pixelation of the texture.

It was off, but turning it on changes nothing. There are many pixels of texture on each of those jagged polygons. Maybe 10-20 texture pixels per polygon.

Here is roughly the same area of the texture as the cropped part above:

The plane is not an instance, right?

Is smoothing on on displacement parameters?

I think I said something wrong before: if you increase mesh density you can lower precision and thus the speed increment (better benchmark) Also the starting of the render after the voxelization should be faster, but voxelization time is geometry dependant and should increase with polycount.

Back to topic...

Changing precision changes nothing? That's odd.

Maybe your surface is just too big or your texture resolution is to low for such a big surface and those artifacts are unavoidable no matter what you try.

/ Max

Hi JesperW

Apart from the jaggies that is a fine water texture. How did you make it if I may ask?

I've been trying to do good sea surface for a while now, with varying success. I've also noticed jaggies on my seawater in some images, as long as I dont get too close to it, they're not noticeable. Not a solution, I know, but choosing a view to avoid jaggies is sometimes possible.

This image shows some jaggies:




However same texture from different angles you cant see them:





My waves are too rounded, I like yours with their sharper peaks, much more realistic of wind blown ripples.

I've tried out your texture that you posted, see images below. Works very well, original render was done at 6000 pixels wide, only got to SL 8. The water object is a 10 m cube, the top surface meshed to 500mm and triangulated. Texture applied as planar, with displacement, precision raised to 32. Smoothng on.



detail of above image


I'll do some more tests...

I'll be watching with interest to see if anyone comes up with some solutions to the jaggies.


Regards,
Bim Daser
www.fluid-future.com

flower wrote:Apart from the jaggies that is a fine water texture. How did you make it if I may ask?

It is made completely in Genetica. Maybe a couple of weeks full time work in it

Your closeup has some traces of jaggies in it as well, as you notice. With smoothing on, the jaggies become less visible, but they are still there, which might point to there being a problem with the actual MPDM polygons.

Smooting makes the surface look much more like someone poured oil on it. The waves look much more natural without smoothing.

Fernando Tella wrote: Is smoothing on on displacement parameters?
The plane is not an instance, right?
Changing precision changes nothing? That's odd.

No. See above post.

No, it's not an instance.

I didn't say that I said the sum of changes to precision, tiling and subdivision needed to achieve the same render looks ends up with roughly the same result in terms of jaggies and render time.

I guess that's because to get the same looks you need about the same size and number of micropolys...

Is the physical sun enabled in the scene? If not, what happens if you disable it and use only physical sky?

Frances wrote:Is the physical sun enabled in the scene? If not, what happens if you disable it and use only physical sky?

Yes, but the results are identical in terms of jaggedness with only sky.

OK, here is a futher test:

The texture here is just a very simple pyramid, plus it is smoothed with a gaussian blur to make shure it has no sharp edges at all anywhere.

It looks like the normals of the polygons are pointing the wrong way or something.

Could it be a problem caused by the combination of MPDM and a refractive material?

If this was a normal surface with just very rough and unsmoothed tesselation you would see the polygons, but not with the jagged edges