I need help or at least ideas,i have decided to spend a little longer playing with this but i'm running into problems.
The original setup for these spheres was quite simple,first create a cube with sub-divided sides,place an object at the center position,point it at each vertex and move it the radius required plus a value for noise,then record the position of the object,this turns the cube data into sphere data,simple.
The hard part was taking all this data and making objects with it,and working out all the normals to keep the lighting smooth,but seeing as all the sides are using a shared array of data it was simply a matter of knowing if you go off the left side of the sphere(cube) which sides data to use to match up with it,corners were a bit trickier but not impossible,and as you can see from the previous video,i think it worked quite well,even my editing code was working well and fairly well optimized so as to only update the tiles that had changed,keeping the speed up,and because all the tiles where using a shared array all the objects lined up nicely with no seams or tares and could be rotated freely without any problems.
Now comes the problem,in order to make this usable the terrains needed to be bigger,so when you were closer to the surface you could barely see the curvature,at present each side was made up of 8 by 8 tiles each with a 17 by 17(0-16) vertex pattern and a step of 2 DBPro units between each vertex.But when you come to expand this you soon run into problems with culling,because each tiles position is still 0,0,0 making rotation easy,once 0,0,0 goes out of camera view tiles start to disapear.
So a new approach was needed,this turned out to be quite simple,create the data as before,but store the center position of each tile,when you create the tiles deduct this value and position the finished tile there,and once again everything is back to how it should be,and you can make truly massive spheres without the culling issues,with one problem, you cant simply rotate each tile any more as they all now have there own axis,they have to orbit the center point.This isn't a major problem either as using either EZRotate or offset limbs you can easily do this.
The problem is this,after spending all this time to make sure all the data is shared correctly to avoid any issues like this,after a few rotations the tiles start to come apart as can be seen in the image below,now i know its not the data,as when its first run the seams are not there,they only start to appear after its been rotated a few revolutions not imediatley,with either of the 2 methods,and the longer it rotates for the bigger the problem gets,i can only put this down to rounding errors,but cant see a way round this.
The only option i have come up with so far is to scale each tile slightly,but this has obvious consequences later and shows up the edges of each tile,and is not guaranteed to work,because the more the sphere is rotated the worse the problem gets.
I'm toying with idea of buying dark occlusion and going back to the original method of doing this,does anyone have any experience with it,thanks.
