[filmscanners] Re: CCD flexibility re: multiple concurrentscene exposures.

["Mats Petersson" <mats.petersson@btconnect.com>]

At 02:19 AM 1/29/2003, you wrote:
>Just wanted to figure out if my reading of the reasons why this is
>not currently possible at the same time.
>
>Specifically if we have either
>
>a) a filmstrip
>
>or
>
>b) a digital back
>
>is it currently impossible to have, in a single scan/scene capture,
>multiple sensitivities of the CCD/CMOS sensor arrays?
>
>Ideally this would offer the opportunity in high contrast situations
>(i.e. wedding photography, etc) where a digital sensor would be
>suitably (in)sensitive to the high white values of a scene's
>components (bridal dress) and sensitive enough to record shadow
>detail.
>
>I guess what I'm talking about is multiple exposure bracketing in the
>same picture. Given a 16 bit range for independent RGB channels, it
>should be possible to produce a picture of utmost range without
>clipping. As it is, given a negative/slide which has lovely detail in
>the bright areas (be it a bridal dress, or clouds etc) and detail in
>dark parts of the shadows, the only way to do it as far as I can see
>is either by re-scanning the area for both extremes (a la Vuescan's
>double-scan/long exposure technique) or to take two pictures.
>
>My guess (without any substantive knowledge of the mechanics of the
>sensors) is that the individual sensors in an array can't all be read
>individually and the exposure time has to be the same across the
>whole sensor array before the array as a whole can be 'closed' and
>made ready for data transmission. Is this correct?

I think, if you studied EXACTLY how a CCD works, that you'll find that
there is one (or significanly low number of) channels to "open" and "close"
the actual sensor and/or adjust it's sensitivity. This is simply a matter
of "how complex do you want to make it", and also "how expensive will it be".

As with any silicon-chip design, more control means more wires and
transistors, which means more cost in one way or another. Either you have
to reduce the feature-size and go a grade or two more expensive production,
or you make the thing bigger, which is going to cost more money. Add to
this a third dimension of increased risk of a transistor or wire not
working correctly, which is further increased if the feature-size is
reduced (it's more likely that ten atoms in the wrong place cuases a
problem if the item you're constructing is only a dozen or so atoms wide,
than if it's several dozen atoms wide).

All these are of course known factors in the chip-manufacturing industry,
but as with all parts of these systems, it's a question of cost vs.
performance vs. marketability. You have to be able to sell it, at the right
price, in the right volume. If you make a really wonderfull CCD, but it
costs a few thousand dollars just to buy the CDD in a volume of 10K units,
then it's not going to pay for itself, because it probbably cost a hundred
million dollars to produce, and you're not going to get 99% yields out of
the fab, no way. Too much cost (or pain?),  not enough little gain.

Of course, you'd still complain, because the CCD would then have square
blocks that are either over-exposed whites, because they have a majority of
background which is shaded or under-exposed backgrounds, when the majority
is the wedding dress. Unless you control each cell of the CCD, of course.
And it would be even harder to compensate on a white flower with a dark
shady background, like flowers in a forest at spring...

Further, the fuzzy logic, needed to figure out what you actually want, will
need to be pretty clever indeed... ;-)
And preferrably quick enough to take that shot while the bride has that
sparkly expression on her face, rather than half a second later when the
artificial intelligence has analyzed the picture and figured out what
you're doing, how to exposure it and digitally enhance the face of the
bride to look like Cindy Crawford ;-P

--
Mats



>Dieder
>
>----------------------------------------------------------------------------------------
>Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe
>filmscanners'
>or 'unsubscribe filmscanners_digest' (as appropriate) in the message title
>or body

----------------------------------------------------------------------------------------
Unsubscribe by mail to listserver@halftone.co.uk, with 'unsubscribe 
filmscanners'
or 'unsubscribe filmscanners_digest' (as appropriate) in the message title or 
body