C.H. ,
unforunately most of the user testing you see on noise, is overall system
noise of which the
sensor pixels may be dominant or may not, and signal processing tricks are used
to mask noise
effects. If the amplifier and electronics following the sensor is better than
the sensor (if well
designed) then you may see the real differences between the sensors. Even the
way the sensor clock
is implemented (sloped edges etc) can affect noise in CCD's. It is
interesting for example that
niche companies like Nucore make modified signal processing chains that
significantly reduce noise
implying many normal post sensor implementations add significant noise over and
above sensor
noise. (Nucore also greatly improves cross color coupling associated with
dynamic range effects by
seperating the color channels)
The sensor area, noise-tradeoff issue is, in theory a law of diminishing
returns: For equal pixel
fill factor and same technology the noise improves only as sqrt(area), in other
words linearly
with chip linear dimension. So a sensor has to be 4 times larger in area (2X
linear) to have half
the noise. A number of different types of noise are lumped together in user
comparisons and
vendors use software to first order compensate for issues like dark current
noise (pixel
leakage)and differences in pixel to pixel sensitivity (pattern noise), which
are all "noise" to
the user, so the inherent differences in sensing technology may not be so
obvious depending on how
well these additional signal processing steps are done or controlled. Dark
current issues are
important at low light levels while pattern noise is important at any light
level. In any of
these sensors the noise performance can vary depending on who actually
manufactures the chip, as
manufacturing cleanliness and contamination trap noise generating charge near
chip surfaces. The
scaling of this "excess noise" with area is unkown, so may improve at a
different rate from normal
sqrt law with area. In any case, this is a "manufacturing problem" rather than
inherent to the
technology, but remains just as negative an effect to users.
The assumption people tend to make, is the implentation of all these additional
factors is equal
for all vendors, so only the inherent chip sensor technology is a variable.
This is likely a bad
assumption! Nucore's website has some comparisons, showing how their post
process chips improve
the same sensors performance.
Regards,
Tim Hughes
The olympus mailinglist olympus@xxxxxxxxxx
To unsubscribe: mailto:olympus-request@xxxxxxxxxx?subject=unsubscribe
To contact the list admins: mailto:olympusadmins@xxxxxxxxxx?subject="Olympus
List Problem"
|