Title: Mandelbox - some help needed... Post by: Kyzia on August 18, 2017, 03:15:53 PM I'm not sure whether this would be a bug, something I'm doing wrong, or something else, so I'd appreciate some help understanding what's going on...
I have a scene using a Mandelbox fractal type, and when I render it with default raymarching step size, I get a high percentage of wrong distance estimations (over 10%) If I try to run the optimization, even the 'low quality' optimization, I end up with a very small value for the raymarching step size (0.00128), and if I then try to render using that value I end up with an almost empty screen. Even if I set it higher (like 0.05), I'm getting large unexpected gaps. Here's the settings for anyone who wants to try it and see what I mean... Code: # Mandelbulber settings file Title: Re: Mandelbox - some help needed... Post by: mclarekin on August 18, 2017, 04:13:02 PM with this type of fractal it is best to keep the boxfolding value/limit ratio to 2: 1, so 1.4 : 0.7 will give less bad DE.
but even better, do not touch them at all, leave them at 2 & 1 and tweak the other parameters for Distance Estimation this formula should be set on preferred method and function , not deltaDE (too slow) BTW I never use those quality optimization button. For a mandelbox Just set the raymarching step to 0.7 and that should be good enough in most situations Code: # Mandelbulber settings file Title: Re: Mandelbox - some help needed... Post by: Kyzia on August 18, 2017, 04:31:39 PM Thanks for the reply - I'll try what you've outlined.
As for the DE method, I would typically leave it as Preferred, and had only set it to DeltaDE (linear) this once to see if it made a difference (apart from just making things slower), since adjusting the raymarching step size appeared to have no real effect on the wrong distance percentage - it stayed around 10% irrespective of whether the step size was 1, 0.1, or 0.05 - which is also why I tried the optimization button. Title: Re: Mandelbox - some help needed... Post by: taurus on August 21, 2017, 10:13:16 PM since adjusting the raymarching step size appeared to have no real effect on the wrong distance percentage - it stayed around 10% irrespective of whether the step size was 1, 0.1, or 0.05 - which is also why I tried the optimization button. I'm not quite sure, but when my observation is right, at very low step multiplyers the limitation of 1000 ray-steps prevent an improvement in error percentage, as the max ray steps are reached, before the algorithm is close to the surface. Use the statistics window to check. At scales close to 1/-1 maybe increasing the number of iterations might also help. The statistics window is useful there too. Title: Re: Mandelbox - some help needed... Post by: taurus on August 21, 2017, 10:21:19 PM Sorry, tried your parameter. None of my assumptions is true here. |