Olympus-OM
[Top] [All Lists]

[OM] Re: Questions about Dynamic Range

Subject: [OM] Re: Questions about Dynamic Range
From: Tim Hughes <timhughes@xxxxxxxx>
Date: Thu, 28 Jun 2007 02:55:16 -0700 (PDT)
Sorry for the dense Enginering jargon.

Apart from the immensely important "marketing hype", there are some other more 
rational reasons to
have some more resolution than might be at first aparent, from considerations 
of just the sensor
dynamic range or say quantization noise issues. 

Having said that, it is no mean feat to make an A/D coverter with an even 
vaguely noise free
resolution of 22bits at data rates of many Mega Hertz and with low power 
consumption. I doubt very
much the converter mentioned, can actually do a low noise 22bit conversion ! 

If you don't seperate out the RGB channels where you can apply exposure (gain 
and offset
)adjustments for each color channel **seperately**, then having a high bit 
depth A/D can
substitute for not having seperated the color channels before a/d conversion.  
(How you do
offseting "zero" operations is rather important , to get unbiased estimates of 
black.) 

Say there is mainly green in the scene, then the camera sets the gain so the 
green signal will be
near maximum of the A/D, but the Red and blue signals if now passed through the 
common gain
channel to the A/D, will only use a very small part of the A/D converter range, 
so a noise source
like  quantization noise becomes more important. Having higer resolution A/D 
will then improve
these lower signal level channels.   Remember also, that the color matrix 
correction couples noise
from the now low A/D resolution on R & B channels to the cleaner Green channel 
because of sum and
difference operations, where noise is magnified because of lower A/D resolution 
for those
channels. Having "spare bits" can then help all channels in the color 
dematrixing/color correction
mathematical operations, even though output is decimated to a lower bit depth 
in final stored
image.

It may be the Foveon sensor is more noisey, partly because the color 
correction/dematrixing
operations add more noise than a similar Beyer sensor, because the color 
"filtering" of the Foveon
chip, causes a lot more overlap in the RGB channels.   


Tim Hughes

 
--- Moose <olymoose@xxxxxxxxx> wrote:

> Andrew Fildes wrote:
> > Mine too - I thought that the only languages used on this list were  
> > English with occasional forays into Spanish and French.
> > Andrew Fildes
> > afildes@xxxxxxxxxxxxx
> >
> >
> >
> > On 28/06/2007, at 2:01 AM, Winsor Crosby wrote:
> >   
> >> My brain hurts.  :-)
> >>     
> Makes me a little fuzzy too, but does make more sense out of Pentax' 22 
> bit A/D converter on the K10D when the final output will "only" be 16 bits.
> 
> Moose
> 
> ==============================================
> List usage info:     http://www.zuikoholic.com
> List nannies:        olympusadmin@xxxxxxxxxx
> ==============================================
> 


==============================================
List usage info:     http://www.zuikoholic.com
List nannies:        olympusadmin@xxxxxxxxxx
==============================================

<Prev in Thread] Current Thread [Next in Thread>
Sponsored by Tako
Impressum | Datenschutz