I routinely model objects for use in larger scenes based on real-world measurements. These measurements are given in millimetres. The easiest way to model therefore, is to set the scene scale to millimetres and input everything in millimetres. This avoids unnecessary button pushing (0's and .'s) and mental conversions.
The consequence of this is that the resulting scenes are very large in comparison to the default cube. What are the negative (and positive) implications of doing this?
In my experience, issues only tend to arise when you are performing any kind of simulation, such as:
- The gravity of the scene does not reflect any change in in the scene scale and therefore has to be manually adjusted to avoid objects falling very slowly.
- Similarly, the speed of sound does not reflect the change in the scene scale when calculating the Doppler effect for exported audio.
- Mass calculations may return bizarre results (such as zero mass and even negative mass)
On the plus side, Blender stores coordinates as floatingpoint and therefore no accuracy is lost when working with very large or very small objects. I have however read elsewhere, that there is a maximum scene size.
I am considering modelling individual objects in millimetres and then scaling them down at the end of the modelling process and then constructing scenes in a more reasonable scale. I think it would be beneficial if you could set the scene scale in Blender in one units and then set the input units in another scale. For example, you could keep the default scale of 1BU = 1m but change the input units to millimetres. In effect, typing 1234 when translating an object would be converted automatically into 1.234 m/BU.