[HUG] Re: Digi-Blads Beware!
- From: Sauerwald Mark <mark_sauerwald@xxxxxxxxx>
- To: hasselblad@xxxxxxxxxxxxx
- Date: Wed, 22 Aug 2007 09:01:26 -0700 (PDT)
As an engineer who makes his living in the
semiconductor industry, I would agree with some of
your statements, but disagree with others. We in the
semiconductor industry have been continually wondering
if Moores Law would hold for another generation -
since 1964 it has (Moores Law is an empirical
observation that the number of transistors that we can
put on one chip doubles every 18-24 months).
There are limits - as pixels get smaller, noise
increases. As pixels get smaller they have less light
gathering capability. I also have confidence that we
are far from the physical limits to improving on the
size/noise/sensitivity of the sensor elements
themselves.
What I suspect will happen is that they will continue
to build larger resolution sensors, even if it means
less efficient pixels, and will compensate with image
processing that will generate the image pixels by
processing a number of physical pixels. This will
allow the marketing guys to go crazy, advertizing 50
gazillion pixels, and will shift much of the burden of
the image quality from the more analog sensitive
sensor elements to the more cost effective digital
processing. Already we are in a realm where very few
people actually store their images in RAW, there is
almost always some sort of processing that takes place
between sensor and storage. If nothing else, the most
common technology uses different color sensors in an
array and then uses digital processing to generate a
final RGB image with equal numbers of R, G and B
pixels (most sensors do not have equal numbers of
sensors for each of the primary colours).
One thing that does baffle me is the drive to make
sensors which are the same size and aspect ratio as
35mm film - does anybody really like that 3:2 aspect
ratio? It makes lens design less efficient than a
square or near square sensor, and does not generally
result in a pleasing image. If I were to design a
digital camera from scratch, I would use a square
sensor, and have the ability to do in-camera cropping
to generate 4x5 aspect ratio vertical or horizontal
shots, as well as square images for those who like to
keep things more balanced.
--- Jim Brick <jim@xxxxxxxxx> wrote:
> Pixels cannot be made any smaller than they
> currently are being made.
> The smaller the pixel, the more noise. five square
> microns is about
> the limit for a quality photograph. P&S cameras have
> smaller pixels
> in order to pack in the consumer hyped 'magapixels'.
> But as you know,
> P&S cameras are noisy at higher ISOs. The best
> sensors have from
> seven to nine square microns per pixel. Each pixel
> requires a
> capacitance (electron bucket) in order to capture
> the electrons that
> the light intensity has allowed through the photo
> diode junction. The
> smaller the pixel, the smaller the electron bucket.
> The smaller the
> bucket, the less dynamic range the pixel has. This
> is why ALL
> professional level digital cameras and backs
> typically use pixels
> around nine microns square. This gives them large
> enough electron
> buckets to capture an extended dynamic range which
> has the side
> effect of greatly reducing noise, which has the side
> effect of
> allowing superb photographs to be taken at ISO 1600
> and higher.
>
> ALL of the visualized increase in digital camera
> image quality has
> been provided by software/firmware engineers. Just
> look at what
> Genuine Fractals can do. You don't need a 39mp
> digital back to
> produce 30x40 prints. A good DSLR, properly used,
> and the resulting
> file pumped up via CS3 or Genuine Fractals, will
> give you a result
> that looks as good as the o/p of a H3D. Of course
> H3D O/P could be
> pumped up to mural size. But it's all 'software',
> NOT hardware. To
> increase the megapixel output, think interpolation.
>
> The five to nine micron pixel size is a wall that
> probably will not
> be torn down with current technology. Just as
> optical microscopes hit
> the wall decades ago, a new technology had to be
> invented, the SEM,
> in order to be able to look at smaller stuff. The
> same with digital
> sensors. A new technology will have to be invented
> in order for an
> image sensor to ever approach the physical data
> capabilities of film.
> With millions of software engineers working on image
> data, the need
> may never come. Just look at what software produces
> with MRI data.
> Creating an image out of random oscillating atoms.
>
> :-)
>
> Jim
>
> PS... they make sensors with one square micron
> pixels. They basically
> are graphic devices. Black or white. Nothing in
> between.
>
>
> At 11:49 PM 8/21/2007 -0700, Richard Schiff wrote:
>
> >In theory, as technology gets better and better it
> should be
> >possible to get the resolving power of an 8X10 into
> a 24X36 inch
> >detector... the problem is that optical science has
> a long way to go
> >to get optics to match the detectors....
>
>
>
=============================================================================================================
> To unsubscribe from this list, go to
> www.freelists.org and logon to your account (the
> same e-mail address and password you set-up when you
> subscribed,) and unsubscribe from there.
>
>
____________________________________________________________________________________Ready
for the edge of your seat?
Check out tonight's top picks on Yahoo! TV.
http://tv.yahoo.com/
=============================================================================================================
To unsubscribe from this list, go to www.freelists.org and logon to your
account (the same e-mail address and password you set-up when you subscribed,)
and unsubscribe from there.
Other related posts:
