View Soft shadows
Soft-shadows, Quality vs Rendertime.
CV 2006-2011: http://www.cineversity.com/forum/forum_posts.asp?TID=3547&PN=11
Soft shadow is a technique to create shadows, it is an indirect way to produce shadow information based on a data-base.
A common misconception by using the Soft-shadows is given in the fact that it would have any kind of pixel-based texture or is at all perfect. It is more or less a relict from a time when computer had no real power. Precise set up saves render time even with a good result for stills.
The soft-shadow is a simulation, which takes “rays” from the light source out into the scene space.
If you use a spot light, then this shows the limitation in which direction these “rays” send out.
On the way, they check from time to time if there was an object, if not then they continue to select information (Bias).
In the case of the 250x250 you get this amount of rays. The points where an object occurs will be noted. Which is not precise as it is not done continuously, it is done based on a distance given in your settings.
The points will be then interpolated among all others in a certain distance (Sample Radius) to the place where a polygon is used to render a pixel. In that way you get an average value among a group of “shadow points”. A reason why it is soft. It is interpolated and that in a spatial way.
If the light source doesn’t move, the points are stable in “space”, but if the objects move, there is a chance that “you” run into problems.
This technique was used in a time when computer were slow and before ray-tracing was available to applications on the market. It is faster and that for a reason.
These parameters need to be understood in detail to get good results. It is suggested to limited the area where the shadow is needed, to get the most out of the “map” which is not a texture or image map BTW., as some people like to introduce it. It is pure raster data. More importantly, the way these points are taken and calculated gives you a good or bad result, in terms of animation.
Worse case, you like to use an omni light and have the map set to 250x250. To my knowledge this results into 6 maps. (In all three axis + and -) All very small in resolution is the result.
In the case of something like a cloth object, which might moves fast, it travels among all the shadow points. Flickering is more often than not the case, especially with low values.
If time allows, avoid this kind of shadow, especially with the new Physical render, to me knowledge, it can even increase the render time, as it needs to be converted. But I have to test more before I have something that fits in a nutshell.
Raytrace is razor sharp and not really everyones taste, also here, the project makes it good or bad in use.
Yes, Area shadow is the way to go, but it takes longer.
However, light in 3D/4D is different than in reality. Here we can select an object and have a light source working on it in- or exclusively. (Scene Tab).
In that way you can apply the most expensive (render-time) light source only where you need it and the less expensive light source to anything else. For backgrounds where you have no change (animation) that could alter the shadow, the Soft-shadow option could be an idea. Especially for these areas if used no other soft shadow lights in the scene, the Render option has a calculate once check box (Cache Shadow Maps).
Sometimes a scene works best with no shadow from any illuminating light source, but with some few Shadow Caster “light” source to the places where you need those. Well, each scene is different and with that, each option to optimize the scene as well. (Baking, camera-mapping, background replacements etc.)