When do you scan at 48-bit?

[napper]

My Epson 3170 can scan at 24 bit or 48 bit color. I tried both out with
some photos. Don't see much difference other than 48 bit creates a
bigger filesize. So when would we scan something at 48 bit in
practical?

 

[Wayne Fulton]

My Epson 3170 can scan at 24 bit or 48 bit color. I tried both out with
some photos. Don't see much difference other than 48 bit creates a
bigger filesize. So when would we scan something at 48 bit in
practical?

48 bit images are 16 bit mode (three 16 bit RGB channels is 48 bits).

It's a moot question unless you have a photo editor that supports 16 bit
mode, you would have zero use for it otherwise. Such a program could be
identified by having a menu to change images between 8 bit and 16 bit
mode, and its tools would work on 16 bit images. Most programs dont
support 16 bits, and just convert to 8 bits to continue.

Moderate to extreme tonal adjustments might be better done in 16 bit
mode, but it is not a large effect, seems more realistic to say not even
visible, but some do quibble about this. This is perhaps of more
interest for film, which sometimes needs larger tonal adjustment, and
some people do like to use 16 bits. When editing is complete, then you
would change the image to 8 bit mode (24 bit image) to use it, because
our printers and video system are 24 bits.

 

[Robert Feinman]

It's a moot question unless you have a photo editor that supports 16 bit
mode, you would have zero use for it otherwise. Such a program could be
identified by having a menu to change images between 8 bit and 16 bit
mode, and its tools would work on 16 bit images. Most programs dont
support 16 bits, and just convert to 8 bits to continue.

I have a discussion of this, with examples, on my web site. You can judge
for yourself if the benefits from using 16bits are worth it.
Follow the tips link on the home page.

 

[Roger Riordan]

......
Moderate to extreme tonal adjustments might be better done in 16 bit
mode, but it is not a large effect, seems more realistic to say not even
visible, but some do quibble about this. This is perhaps of more
interest for film, which sometimes needs larger tonal adjustment, and
some people do like to use 16 bits. When editing is complete, then you
would change the image to 8 bit mode (24 bit image) to use it, because
our printers and video system are 24 bits.

http://www.scantips.com "A few scanning tips"

I was very interested in your demonstration of the differences between scans
taken with 8 bit resolution, and with 16 bit resolution, and particularly in the
two photos of the fellows face. This was seriously underexposed in the
original, and this is where I would expect that any difference would be most
obvious. According to all the textbooks, the eye has a logarithmic response,
whereas (I think) all photoelectric detectors have a linear response. So, with
eight bit resolution, we can only get 256 linearly spaced levels of brightness
for any given colour, and theoretically the differences in intensity between the
first few steps up from black should be readily visible

When, as in this photo, we increase the gain for lower intensities, to
compensate for under exposure, these differences in intensity should become even
more obvious. If the sensor was really capable of 16 bit resolution, we would
be able to get 256 steps between each of the steps given by the eight bit scan,
and this would almost certainly be enough to give a smooth curve, even in
seriously underexposed regions.

I have not been able to find any datasheets for CCDs which give any meaningful
data on resolution, but I would be very surprised if the A-D onverters used in
them were capable of giving more than 10, or at the outside 12, meaningful bits,
corresponding to either 4 or 16 steps for each step with eight bit resolution.
But even with this value I would be very surprised if it were not possible to
detect a significant difference in the almost blacks. However, when we examine
these two photos of yours, there is no evidence of banding in the almost black
region, and no significant difference between the two scans. There is, however,
a great deal of noise, which could be due either to grain in the film, or to
noise in the CCD. The most likely explanation for the lack of banding in the
eight bit scan is that it has been masked by the noise. If the noise was in the
film, it implies that there is not the slightest virtue in working at higher
resolution with this film, processed in the way it was. On the other hand, if
the noise was in the CCD, the higher resolution mode becomes completely useless.

I was interested in this demonstration for its possible application to digital
photography. The situation MAY be different there, but I am not feeling at all
hopeful.


Roger Riordan AM