What Ultrafractal formula to use for rendering mandelbulbs ?!??!

[fractalrebel]

Hi everyone,

I have added a tutorial for 3D Fractal Raytrace (UF5) to my tutorials pages on my website. There are snapshots of the user interface, three Mandelbulb examples and several other 3D/4D fractal examples, complete with parameter files that the user can copy and paste. The general link to the tutorials is:

http://www.hiddendimension.com/Tutorials/Tutorials_Main.html

While the specific link to the 3D Fractal Raytrace (UF5) tutorial is:

http://www.hiddendimension.com/Tutorials/3DFractals/3DFractalRaytraceTutorial.html

 

[cKleinhuis]

i have corrected the tutorials link to your site, i would like to put the formula .upr files on to the wiki with images, are you ok with that ?!
as it shall also serve as parameter collection for ultrafractal
 

[fractalrebel]

i have corrected the tutorials link to your site, i would like to put the formula .upr files on to the wiki with images, are you ok with that ?!
as it shall also serve as parameter collection for ultrafractal
 

Please feel free to use any any of the contents of the tutorial for posting to the wiki.

[cKleinhuis]

thank you, but i have to insist: setting of iteration depth is not working, the mandelbulb 8th order gets low resolution inner bulbs, wenn zoomin in, i have entere an iteration depth of 250, and nothing changes, do i have to increase the additional accuracy ?! 

[David Makin]

thank you, but i have to insist: setting of iteration depth is not working, the mandelbulb 8th order gets low resolution inner bulbs, wenn zoomin in, i have entere an iteration depth of 250, and nothing changes, do i have to increase the additional accuracy ?! 

My formula or Ron's ?
In mine:
No, not the accuracy, you have to reduce the distance estimate threshold distance - when using distance estimator solid is not based on iteration but based on the distance to inside and is solid when the distance to inside is less than the threshold you specify.
You only need to *increase* the accuracy if there are errors in the render such as missing bits - often an accuracy of 2 is sufficient but sometimes up to 4 is required (or higher on some Julia Sets).

In Ron's again for more detail you'll need to reduce the distance estimate threshold or at least modify the parameter that will do so.

[fractalrebel]

thank you, but i have to insist: setting of iteration depth is not working, the mandelbulb 8th order gets low resolution inner bulbs, wenn zoomin in, i have entere an iteration depth of 250, and nothing changes, do i have to increase the additional accuracy ?! 

Trifox,

Try setting iteration number to 3. You will see it is working. As Dave has pointed out, greater detail in my formula is obtained by decreasing the closeness parameter. Iteration number does not affect how close the surface is approached. Approach to the surface stops when the estimated distance to the surface is less than closeness. In the code, closeness is called epsilon. This follows the Hart, Sandin and Kauffman algorithm.

When you zoom in you will probably have to decrease the closeness to maintain a good resolution level.

[fractalrebel]

The first image shows a zoom into an 8th power mandelbulb with closeness = 1 (the default value). Pretty nasty looking. The second image has closeness = 0.045. This is more what we might expect. Trifox, I think you just did a bit of beta testing. I will modify the formula so that closeness depends upon the zoom. This formula has built-in version history so old uprs will still render correctly, while new one will use the modified closeness. I will put a message on this thread when the modification has been tested and uploaded to the UF database.

[cKleinhuis]

shame on me !

i just played around with the excellent gpu version made by subblue, and i understand, your iterations work just fine, i wasnt used to the behaviour of the iteration increase, was
hoping the would get more detail, but you have to zoom in ! i think it is another side effect of the whippey cream behaviour

so, i will now subclass your 3d renderer, and quat classes, to provide them with an "iterationStart" event, i try to hold it as compatible as possible, to use
all your existing formulas in an formulat alternation process... i have played with it within the realtime gpu version, had was much fun and needs to be implemented
in ultrafractal to make mystic 3d objects movies

very cool! thx for patince! 

[fractalrebel]

shame on me !

i just played around with the excellent gpu version made by subblue, and i understand, your iterations work just fine, i wasnt used to the behaviour of the iteration increase, was
hoping the would get more detail, but you have to zoom in ! i think it is another side effect of the whippey cream behaviour

so, i will now subclass your 3d renderer, and quat classes, to provide them with an "iterationStart" event, i try to hold it as compatible as possible, to use
all your existing formulas in an formulat alternation process... i have played with it within the realtime gpu version, had was much fun and needs to be implemented
in ultrafractal to make mystic 3d objects movies

very cool! thx for patince! 

I have updated 3D Fractal Raytrace to autocorrect the closeness on zooming. I also added a bit of explanation to the tutorial.

[fractalrebel]

Here is a 400x zoom into an 8th power Mandelbulb. Its comprised of two layers. A distance layer provides the color and the raytrace (Phong) layer provides the highlights.

[fractalrebel]

Here is a little demo that shows the effects of maximum iterations of an 8th power mandelbulb. The top two panels show the mandelbulb with no zoom. The upper left panel is with maximum iterations = 6 and the upper right panel is with maximum iterations = 1000. Not a lot of difference, although the higher iteration panel shows more detail. With a 9x zoom substantial differences are seen. The lower left panel has maximum iterations = 6. Stair steps can be seen from the low iteration value. The lower right panel has maximum iterations = 1000. There is considerable more detail, and regions which in the left panel had little detail other than the stair steps, now have fractal detail.

[cKleinhuis]

yeah, thank you very much for explaining, i think this is one of the essentials we have to learn about the behaviour ( when browsing ) by now i had some time to experiment with a gpu version of those things, and
it is enjoyable, but sometime dissappointing that iteration increase does only affect (non-whipped-cream ?) areas ...