48-bit, 32-bit, 24-bit, 16-bit, or what?

[Lucile Davis]

When one scans anything (photo, film, transparency), intending to work on it in PhotoShop, what bit to choose?

[Doug Nelson]

It's a bit of a mishmash, since different sources are referring to different things when they say "X-bit".

But for our purposes, we talk in terms of "X-bits per channel", with RGB having 3 channels. So if we're working in 8-bits per channel, some references (driven by marketing, I suspect) will call that 24-bit (8bit/channel X 3 channels).

As to what it all means, the number of bits is the exponent to the number 2. So an 8bit channel can store 2^8 tones (256 tones, referred to oddly as tones 0-255) (tones is all there is when discussing channels, no colors. Colors come from combining channels).

Generally, scan at 8 or 16 bit (per channel). 16 is "better", but can be a waste of resources for many purposes (esp. if you have a slow machine or small HD).

[byRo]

I personally work this way:

1) Scan in 16-bit (which for your scanner may appear as 48-bit [i.e. 3 x 16]).

2) Do the inicial levels and curves processing in 16-bit.

3) Switch to 8-bit from there on...

Rô