Z-channel, theoretical accuracy

[tomtnt]

Hi Dears
Just a little question.
What is a theoretical depth of Z-channel? For example. I have got a scene in mm units and I would like have 1mm precision in Zdepth channel. So, if I am using EXR render output, which is 16bit float, than in one channel is 65535 colors, so theoretically can a z-channel store 65.535 mm depth in channel? Is that right?
Because sometimes looks for me, if I have a big scene...round 1 km square size, the Z channel with 1000m distance is so rough in details in front of camera.
Also, what does it mean Clamp Zdept? Does it respect Zdepth MAX value or just respect the size of BOX SCENE?

Best regards
Tomas

[instinct]

K. Lemme try and sum up real quick:

- by default VRay for Max stores images in 16bit halffloat which is way sufficient for image layers. If you need accuracy in aux channels you might wanna consider going 32bit full float
- clamping is clamping to 1.0. Depth is remapped to have min/max be 0.0-1.0. So if clamping is on, everything beyond max will be clamped. Using 0/1, invert and clamp off you get an "as is" distance to the camera in scene units
- 16bit float does not have a maximum value, that's the whole point of floating point vs. integer. With growing numbers your accuracy behind the comma gets worse tho. (Imagine it as loosing a digit behind the comma vor every digit you need in front of the comma. e.g. 1.000 -> 10.00 -> 100.0. That's simplifying quite a bit tho heh) 16bit half float (as used by default is a special floating point representation. For accuracy you might consider going 32bit full float tho.

Hope that helps,
Thorsten