12-bit or 14-bit raw?

GhostInTheRuins

What do you use?

From what I've read online there's more detail information in the shadows and hilghlights in a 14bit raw file compared to a 12-bit, but is there any real world benefit to using 14bit?

pete4130

Better detail in shadow & highlight. If your scene has a wide dynamic range its one big plus for it. I always shoot 14bit,

Heebie

14-bit would have the potential for more detail in the entire range, not just in shadows or highlights. (but.. better across the range does mean better in shadows and highlights in general.)

In 12-bit/channel you have 36 bits representing the color (12 red, 12 green and 12 blue.) This makes for 2^36 possible colors (68,719,476,736 total possible shades of color, including 2^12 or 4,096 shades of grey) [grey being where R, G and B values are equal]

In 14-bit/channel you have 42 bits representing color 2^42 is 4,398,046,511,104 possible shades including 2^14 (16,384) shades of grey.

So.. for color.. there will be a LOT more detail to play with (68 million colors vs. 4 billion) Since the human eye can only distinguish about 10 million, there's already tons of shading that is imperceptible.. which is extremely valuable if you need to compensate for an imperfect expsosure.

The enormous difference though is in greyscale. 4 thousand vs 16 thousand is an enormous difference if you're doing black & white work. (and if you're shooting JPG's.. you're limited to just 256 shades of gray.)

The deeper bit depth you can have on your original source material, the better.. always.

Here's a real-world scenario as to why it's so valuable:

I had someone bring me a roll of film that he'd underexposed by several stops. I was able to get usable scans off of every frame.. including getting a usable image of some flowers that looked like a completely unexposed negative to the naked eye. (thanks 48-bit scanner!)

I've also given customers colour restorations of photos they brought in that they THOUGHT were black & white, but were just so faded that they looked like black & white photos with an artificial colouration to them. (usually looking like a cross between sepia.. and something red.)

I suppose that's WAYYYYY too much information just to say "Yes, 14-bit!" but it's one of those things that I always like to stress as important.

stcstc

the thing is though 12 or 14 bit is determined by the camera

8 or 16 is the choice when opening in raw converter

and where ever you can use 16, particularly for B&W. it means instead of 256 shades of grey you will have 65K

GhostInTheRuins

Great post Heebie!

Heebie wrote: »

So.. for color.. there will be a LOT more detail to play with (68 million colors vs. 4 billion) Since the human eye can only distinguish about 10 million, there's already tons of shading that is imperceptible.. which is extremely valuable if you need to compensate for an imperfect expsosure.

But surely if your eye can't even distinguish between these 4 billion shades, is it safe to say that you'll only ever notice the difference between 12 and 14bit if you're doing some extreme recovery of a really badly exposed image? (Like your underexposed film as an example, must be a savage scanner you have!)

Never the less, I've just switched my camera to 14bit, so thanks! I might do some of my own tests tomorrow and see if I can see any differences at all.

Heebie

If you're using 16-bit/channel.. you still only have 12 or 14 bits of source material, which is much more of a limiting factor... but yes, you should always edit in the highest bit-depth your source material realistically will support. (in the case of either 12 or 14 bit source.. 16-bit is a much better option than going down to 8!) Going to 32-bit/channel would just be silly from 14-bit source.

stcstc wrote: »

the thing is though 12 or 14 bit is determined by the camera

8 or 16 is the choice when opening in raw converter

and where ever you can use 16, particularly for B&W. it means instead of 256 shades of grey you will have 65K

stcstc

the thing is though 12 or 14 bit is determined by the camera

8 or 16 is the choice when opening in raw converter

and where ever you can use 16, particularly for B&W. it means instead of 256 shades of grey you will have 65K

Heebie

From the 2nd identical post now.

Your source doesn't have 64K (64K = 65536).. it has 16K shades.. but you're much better playing with those 16K shades within a 64K-capable space than having to downsample it to fit in the 256 shades.

stcstc wrote: »

the thing is though 12 or 14 bit is determined by the camera

8 or 16 is the choice when opening in raw converter

and where ever you can use 16, particularly for B&W. it means instead of 256 shades of grey you will have 65K

stcstc

ooops didnt mean to post twice


sorry why only 14k


16 bit is 65k ish per chanel

Hugh_C

stcstc wrote: »

16 bit is 65k ish per chanel

you'd make a great banker

stcstc

what ya mean hugh?

stcstc

16 bits per channel is 2 to the power 16 = 65536

so for B&W you have 65536 shades, ok it comes from a 12 or 14 bit raw, or 16bit from a hassie

Heebie

Not 14K.. 14-bit/channel. It's 16K not 64K because the source is only 16K greys (or some tens of millions of colours.) Bits are 0's or 1's, and depending on where they are they correspond to different decimal values.
The first bit has a value of 1, second a value of 2, 3rd a value of 4, 4th a value of 16 (exponential. 1,2,4,8,12,32,64,128,256,512,1024,2048,4096,8192,16384,32768,65536,131072 up to 16-bits)
The 1 is the "least significant bit" as it being set changes the value the least, the highest bit, in 16-bit 131072 is the "most significant bit"
If that bit is set to 1, then you add it's value to the total value of the number.. if it's set to 0, you don't. The least significant bit (LSB) is the furthest to the RIGHT as written the MSB in the left-most position
So..
00 = 0
01 = 1
10 = 2
11 = 3
0000 = 0
0010 = 2
0100 = 4
1000 = 8
1010 = 10
1011 = 11 and so on..

The source material is 14-bit.. you're just mapping it to a 16-bit color space. Generally it would get mapped to the centre of the 16-bit space, so the least significant and most significant (lowest & highest) bits would end up not being used. Having the extra 2-bits/channel as "headroom" for editing purposes is a very very good thing.

with 14-bit you have a range of:

00000000000000 << 14 possible bits, all 0's (absolutely black)
11111111111111 << all 1's (absolutely white)

This gets mapped into 16-bit space like this:

-00000000000000- << the bits above mapped to the centre of 16-bits
-11111111111111- << same with all the bits set to 1 instead of 0.

It's still the exact same data when it gets into 16-bits., there's just extra space. (those extra bits would generally actually bet set to 0)

Some converters might map it all the way to the bottom of the space, and some to the top like below.

--00000000000000
00000000000000--

If mapped to the "bottom", white becomes a bright grey and black gets REALLY black.
If mapped to the other exreme..white becomes REALLY white and black becomes a dark grey.
Mapped to the middle, white becomes a brighter gray, but not true white, and black becomes a dark grey but not a true black, so you can adjust in both directions.. so it tends to be preferable to putting it at either end.
Some algorithms might do fancy things to try & maximise highlights and shadows.. but they still don't really make those 14 bits into 16, they generally do something similar to pixel doubling. (CD players have been doing this type of trickery on digital audio since the late 80's.. it is still trickery though.)

stcstc wrote: »

ooops didnt mean to post twice


sorry why only 14k


16 bit is 65k ish per chanel

stcstc

sorry i understand what ya mean now

i was actually talking about real 16 bit files, where ya have 65k shades, like what comes out of a hassie

but the point is, its a huge diff compared with 256 from a 8 bit jpg for example

Hugh_C

stcstc wrote: »

what ya mean hugh?

'2^16 is 56k-ish', it's all those ish-es that have the economy in sh1te

Heebie

Well, not all of us have Hasselblad kit!

and yes.. even 4K (12-bit) is a tremendous improvement over 256 (8-bit).. by orders of magnitude. 16K the same again, and 64K the same again.. again.

It is interesting that switching to a sepia tone, or some other type of shading to it, can really open up the shading possibilities a great deal. More shades of grey = YES!! =D

stcstc wrote: »

sorry i understand what ya mean now

i was actually talking about real 16 bit files, where ya have 65k shades, like what comes out of a hassie

but the point is, its a huge diff compared with 256 from a 8 bit jpg for example

stcstc

it is 65k is, or actually 65536

Heebie

No.. that's 64K (1K = 1024.. 1024 * 64 = 65536)

stcstc wrote: »

it is 65k is, or actually 65536

stcstc

sorry what i mean is 65k as in 65 thousand

Hugh_C

Heebie wrote: »

The source material is 14-bit.. you're just mapping it to a 16-bit color space. Generally it would get mapped to the centre of the 16-bit space, so the least significant and most significant (lowest & highest) bits would end up not being used.

I can't understand why it would be mapped to the middle of a 16-bit space, why would you sacrifice half your greyscale?

For the sake of argument, let's say you've captured something that's full-white on your 14bit system (i.e. 11111111111111 for all channels), bringing it into the centre of a 16-bit system isn't going to ket you "find" any more detail in something that's already at clipping point. Adding another empty bit at the MSB is not going to allow more white information, it just hasn't been captured.

I could more easily understand adding 2 LSBs to allow you stretch the blacks. Or add dithering (another audio trick) to give you smoother gradients.

Hugh_C

melekalikimaka

big is ALWAYS better