 |
12-bit vs 14-bit NEF |
Original +0EV
-8EV out of camera JPEG - non black set to white
recovered photo - 14-bit NEF (-8EV test)
recovered photo - 12-bit NEF (-8EV test)
|
Is 14-bit NEF worth it?: On the Nikon D300, 14-bit mode actually significantly
slows down how fast pictures can be taken (from 6/8 fps to 2.5 fps), and it
also introduces a subtle 20 to 40 ms shutter lag.
So, what is the quality difference between 12-bit and 14-bit mode? Is 14-bit worth it?
Well exposed photos: At first, I was testing 'well exposed photos'. I
could find NO visible difference between 12-bit and 14-bit mode, no matter how
much I poked and prodded, and pixel peeped. The 'problem' is that it is
near impossible to isolate the lower bits of a tonal range in a well exposed
pixel value in Capture NX -- because all you can do is push the pixel value to
255. It is very hard to see more tonal range on an 8-bit tonal range computer
display.
Underexposed photos: The solution to isolating lower tonal range bits?
Dramatically underexpose a photo on purpose so that the only thing remaining
is not the significant bits of a pixels value (which are then all zero), but
the lower bits, and then compare 12-bit vs 14-bit mode.
The test: Take a D300, on a tripod, and let it take
a NEF+JPG picture in 'P' mode. Take note of the exposure
settings and replicate them in 'M' mode. Then purposefully adjust
shutter speed minus 8EV. Then take a picture
in 12-bit mode, then again in 14-bit mode.
Results: The resulting OOC (out of camera) JPEG looked 100%
black, and essentially was. For full disclosure, to the right
is a map of every single color recorded in the OOC JPEG. A TIFF
produced within NX had far fewer non-black pixels, indicating that
most non-black pixels were just JPEG artifacts in the first place.
So job accomplished. All significant bits are effectively zero.
How to bring out the detail: Go back to the NEF's.
First, in RAW Adjustments, increase EV by +2. Then, bring
the pixel levels up the rest of the way with two back to
back gamma based Levels & Curves. The results you
can see to the right.
Clearly, there is a small and subtle difference between 12-bit
and 14-bit NEF.
Also, if you look carefully, you can see the 'columns' in 12-bit
mode, but not in 14-bit mode. These 'columns' can be seen even
in -4EV situations. Very strange. At least in the D300, the
implementation of 12-bit and 14-bit are very different.
High ISO 14-bit is actually better: I was first under the
impression that 14-bit in high ISO gained you nothing (only record
more noise). But I have changed my mind after pixel peeping at ISO
6400 in both 12 and 14 bit NEF on a D300. A difference (for the better) can be
seen in 14-bit modes. There is practically the same (subtle
difference) dynamic range between 12 and 14 bit NEF. The difference
is in the tonal values recorded, which in high ISO for some reason
made a difference? Technically, this should not be so (14-bit 'should'
only record more noise). But it may indicate larger implementation
differences between 12-bit and 14-bit modes. More research is needed
to find out why this is.
There does appear to be a major implementation difference in the D300
between 12-bit and 14-bit modes. Just look above at the columns that
can be seen in low light 12-bit mode, but not in 14-bit mode. I suspect
the only reason high ISO looks slightly better in 14-bit mode is because
of this unknown implementation difference between it and 12-bit mode.
Conclusions: NEF is far superior to JPG in recording detail in deep
shadows (black). 14-bit NEF is better than 12-bit NEF, but only if you
are going to heavily modify your photo and attempt to pull detail out of
very dark areas of your photo (or if you are going to be using high ISO).
Original -8EV files (NEF plus OOC JPEG's):
Why does 14-bit have (slightly) more dynamic range than 12-bit: When
you examine these two NEF, why does the 14-bit NEF capture more dynamic range
recorded than the 12-bit NEF? One possibility is a 'truncation' behavior in
the A/D (analog to digital) converter inside the camera. Namely, in 12-bit mode,
the 12-bit number of '4' is used to represent the 14-bit mode numbers of 16, 17,
18, and 19. Or looking at it another way (normalize the 14-bit numbers back to
12-bit numbers), 4.0, 4.25, 4.5, and 4.75. So for a particular number of photons
at a pixel location, the A/D converter might generate the number 4 in 12-bit mode,
but the number 19 (4.75) in 14-bit mode.
But based upon the 'column' behavior seen above, there is more going
on between 12-bit vs 14-bit than what we can deduce (only Nikon knows
the real reason).
A side note: From the test above, you can clearly see that even
though the JPEG was a lost cause (pure black pixels), we were
still able to go back to the original 12-bit NEF and recover a decent
recognizable image (recover black into something non-black). This
test directly contradicts crazy statements about RAW vs JPEG that
you can find on the web, like this one:
"Raw records usually with 12 bits, but a linear 12 bits.
JPG uses only 8 bits, but these are after the log and
gamma conversion, and thus preserves the 12 bit
precision at the shadow levels where it's important!" - Ken Rockwell
Use some common sense when browsing the web! Extraordinary claims require
extraordinary proof. If you blindly follow the claims of self-proclaimed
experts on the Internet -- without any supporting evidence -- you just might get
burned.
If NX101.com helps you, help this site by making a donation.
|
