Douglas Burns wrote:

> On 11th March 2004 David Kay wrote:
> 
>> Electrons are generated proportionate to the intensity of light.
>> So when twice as many electrons are generated, up to the saturation point of
>> the CCD, we have one f-stop more exposure. And when half as many are
>> generated, we have one f-stop less exposure. Agreed?

> I don't see how increasing the sampling rate from 8-bit
> to 16-bit makes the sensor any more sensitive.

It doesn't in terms of end points, they get mapped to black and white in
both devices. However in a 16-bit device, the higher bit depth allows the
device to describe progressively lower and lower subject values with
improved tonal differentiation because of the CCD's proportional response to
light. In the 8-bit device, once we hit the noise floor of the CCD, all
lower subject tones are mapped to black or 0. The overall contrast is the
same in both devices, however, there are more f-stop equivalents resolved in
the higher bit depth device.


> When we're using terms such
> as "f-stop" we're just describing a mathematical relationship. Surely just
> because you can make the device more accurate in its description of what
> it's recording doesn't make it able to actually record more. I think someone
> else posted the example of a monitor displaying an image: whether it's
> displaying 16 colours, 256 colours, or millions of colours (the bit-depth)
> it's dynamic range (the difference between the blackest black and the
> whitest white it can display) doesn't change.
> 

>From the example above, the 8-bit device does a damn fine job and we see a
smooth progression of tones, however, if we were to apply any kind of
destructive curve to our pristine  "input = output" data, then the chances
of getting tonal degrades caused by rounding errors is likely.
In virtually all high-end camera or scanner workflows, you work in high bit
depths, do you "toning" there and export a finished 8-bit file mapped from
the 16-bit or whatever data. My high-end scanner operates like. It's only
when I change my mind after exporting the 8-bit file, apply a curve
adjustment layer or two, hoping to improve the image, that I sometimes see
evidence of banding. Naturally, with PS CS I now keep the data in 16-bit for
any adjustment such as this.

David Kay 




===============================================================
GO TO http://www.prodig.org for ~ GUIDELINES ~ un/SUBSCRIBING ~ ITEMS for SALE

Reply via email to