3D LUT sizes and bit depths

[Tobias Schachinger]

Hello everyone,

I have some questions concerning LUT sizes and bit depths. For certain file formats, I can select an input and output bit depth. I want to use my LUT in DaVinci Resolve, so I need a cube file. For these, the bit depth selection is not available.

I’m a little confused as to what that means. What influence does the choice of bit depth in other file formats like 3dl have on the signal? If I understand this text from Light Illusion correctly, the only thing that matters to the accuracy of a 3D LUT is its size. Is the input and output bit depth just information that is necessary for the creation of a 3dl file, but not for the creation of a cube file? What does input and output even mean in this context, a LUT can’t convert signals between bit depths, can it?

It also confuses me what EIZO writes about LUTs on their website, because there, it sounds like bit depth does matter to the accuracy of a LUT. Should I be worried that generating a cube file in DisplayCAL to use as a display LUT in DaVinci Resolve might not be fully compatible with an accurate 10-bit workflow?

Unfortunately, DisplayCAL documentation does not seem to contain very detailed information about this. I think it all comes down to the fact that I don’t really know what the term “LUT bit depth” means as opposed to LUT size. But as I am currently writing my bachelor thesis on color management, I feel like I should understand this. Can anyone answer my questions and/or – even better – recommend some specific literature that I could use for my thesis? I would be really grateful!

Thanks in advance,
Tobias

[Florian Höch]

I want to use my LUT in DaVinci Resolve, so I need a cube file. For these, the bit depth selection is not available.

There is no bit depth with floating point numbers.

I’m a little confused as to what that means. What influence does the choice of bit depth in other file formats like 3dl have on the signal?

Floats are encoded as integers with n bits precision in some LUT formats.

the only thing that matters to the accuracy of a 3D LUT is its size

That’s a bit of an over-simplification. For formats that use integer encoding, you should use the highest supported bit depth to avoid quantization errors.

What does input and output even mean in this context, a LUT can’t convert signals between bit depths, can it?

It’s purely related to encoding. What the software using the 3D LUT does with the information, is up to the software – obviously output bit depth can’t be higher than what is encoded in the file (well, it can due to interpolation, but accuracy won’t be increased).

it sounds like bit depth does matter to the accuracy of a LUT

It does for LUT formats using integer encoding.

I don’t really know what the term “LUT bit depth” means as opposed to LUT size

Size is the number of grid points used in either direction (e.g. 65x65x65 = 274625 points). Resolution would be a better term.

• This reply was modified 4 years, 12 months ago by Florian Höch.

[Author]

Posts