Hi, Some notes regarding the scale of modelling:
- Rendering accuracy is best when the diameter of the scene is about 1 meter, but real world scale should work fine for most projects. Some analytic primitives precompute high powers of the dimensions of geometry to optimize and speed up rendering. This makes computations inaccurate. Even double precision floats are not enough when you examine a detail on a surface of of a real-world scaled globe sphere.
- The dynamic clipping of the perspective ground plane by drawing density works badly in large scale models. Increasing the default density of the ground grid from 0.5 meters to say 5 meters makes landscape modeling more comfortable. Currently the density is a defined in preferences (Asynchronous Drawing/Window Drawing/GroundDensity), and you have to restart the program after changing the setting, which is inconvenient. We will make it a regular view property in upcoming versions.
- Defining a proper native state for the view is recommended for unusual model scales: a proper clipping range say 1...10000 meters instead of 0.01...100 m, distance = 100 m etc. Also decimals are important because they eat up the available dynamical range. So, set clip start to one meter instead of 0.01 meters for large models. Drag-dropping a geometry item to view does not increase the lower limit of the clip range, only the upper one if necessary.
- There is one inconvenient issue when working with big models. Camera object's 'front frame' defines its field of view. The front frame is typically placed 1 meter in front of the camera position. This means that when you switch from perspective camera view to parallel camera projection and the distance to the scene is much higher than one meter, perpective camera shows much wider area of the scene. We will fix this problem in future releases.
Best regards, Vesa
