Winsor,
There is what is called quantization noise which adds to the sensor
noise and is A/D bit
depth dependent. In a good design the sensor noise is a lot bigger than
quantization noise, so A/D
resolution then adds very little quantization noise , exactly as you suggest.
Generally random noise adds as RMS quantities, so quantization noise is a small
effect,provided
sensor noise is greater than the A/D LSB by about a factor of 3x (at >3x the
quantization adds <
~10% to sensor noise) . So ideally you want the A/D to have 2-3 bits more
resolution than the
sensor dynamic range,and having a convertor noise level itself of less than
0.5LSB, so that it
then does not contribute significantly to the sensor's inherent noise limit.
Take the hypothetical extreme case of the sensor having a much wider dynamic
range than the A/D
bit depth (!) and the A/d having converter noise less than 0.5LSB, then the
noise would be solely
limited by quantization noise of A/D. Say we then had a sensor voltage that is
sitting at the edge
of any A/D bit transition: a 12 bit A/D quantization noise floor would then be
worse than the
16bit A/D noise floor by about 16x (4bits). In reality the higher resolution
A/D likely has more
noise above it's quantization limit, so the difference would be smaller.
Tim Hughes
--- Winsor Crosby <wincros@xxxxxxxxxxx> wrote:
>
> Dynamic range is unrelated to bit depth. Roughly stated, dynamic
> range on a sensor is the range between unacceptable shadow noise and
> blown out highlights. It remains the same whether that range is
> mapped to 12 bits or 16 bits. The bits in 16 bit would just divide
> the dynamic range into smaller ranges giving you smoother transitions
> in areas of relative uniformity, like the sky, or more differentiated
> detail in shadows where the fewest bits are available.
>
> Hope that makes sense, Garth.
>
>
>
> Winsor
> Long Beach, California, USA
>
>
>
>
> On / June 25, 2007 CE, at 8:33 AM, Garth Wood wrote:
>
> > What is the basis for this claim? From what I can see from the
> > published specs, all three cameras produce 12-bit RAW files, and I was
> > under the (possibly erroneous) impression that the bit-depth of the
> > RAW
> > file was a more-or-less reliable indicator of the dynamic range of the
> > captured image.
> >
> > Is this correct? What am I missing?
>
>
>
> ==============================================
> List usage info: http://www.zuikoholic.com
> List nannies: olympusadmin@xxxxxxxxxx
> ==============================================
>
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|