I hadn't thought about this until now but I've noticed for a long time
that sensor specs (such as those shown on dpreview for the E-1) always
show more pixels available than actually used. I used to think that
perhaps there is an area on the outer perimeter where circuits might not
be completed or bayer patterns might not be complete.
However, looking at the E-1 specs on dpreview I see: 5.6 million total
pixels (2738x2044), 5.1 million active pixels (2614x1966) and 4.9
million output pixels (2560x1920). Looking at the numbers I think there
are far too many pixels that aren't used to support my simple
hypothesis. Something else is afoot and I know not what.
But I have recently learned of one area where different software can
produce quite different results from raw files. For example, it might
be that the red channel on a particular chip tends to blow out before
the green and blue channels. The manufacturers software may consider
this to be an unrealiable pixel and call the entire pixel white by
boosting the green and blue channels to max as well. They'd rather
report the pixel as blown than report an inaccurate color. Other, more
daring software may show the pixel as recorded or may attempt to be even
smarter and interpolate a different value for the "blown" red channel
and fill it in. I think this may be the reason that some raw conversion
software is reported to be able to pull a bit more shadow or highlight
detail than some other converter.
Chuck Norcutt
Richard Lovison wrote:
> On 2/11/06, C.H.Ling <chling@xxxxxxxxxxxxx> wrote:
>
>>
>> Just done know why they don't use the maximum size, it
>>gives more FOV.
>
>
>
> That's what I feel as well but then maybe the "cites" in the outer rows and
> columns vary in quality from chip to chip and the safe way for Oly would be
> to ignore them in the conversion.
==============================================
List usage info: http://www.zuikoholic.com
List nannies: olympusadmin@xxxxxxxxxx
==============================================
|