|
In film photography light focused by the lens strikes a thin chemical
emulsion containing a
silver halide compound. In proportion to the amount of light received at
each location on this film, metallic silver is broken away from the halide it was
previously bound to. Through the chemical processes of development and
fixing this effect is amplified and then stabilized against proceeding
further should additional light be received. In color photography various layers
in the film are dyed in response to their sensitivity to different wavelengths
of light. The image is both captured and stored on the film. After development,
fixing and dyeing the image is largely immune to further changes and can be
displayed, printed, etc. Part of the chemical process leading to
development of the image contributes to clumping of the silver particles. This
is a necessary part of making the image visible at all and the effect is called
grain which upon close inspection appears as an even distribution of fine
speckles or grit. There is a direct relationship between the speed or sensitivity
of a film and clumping or grain, with faster films showing more grain. If grain
is visible in the displayed image it is usually considered undesirable but if
you want speed in a film, grain is unavoidable. Film speeds (or
sensitivity to light) are given in numbers such as ISO 100, ISO 400 etc. with higher
numbers corresponding to greater sensitivity (and more grain).
As we know, there's no film in digital. Instead, there is a light sensitive
solid-state sensor in the form of a rectangular surface approximating
the proportions of a 35mm frame but generally smaller in actual size. This
sensor is usually a Charge Coupled Device (CCD) or it
may employ Complementary Metal Oxide Semiconductor (CMOS) technology. CCDs are
inherently better imagers but the CMOS fabrication process is much less expensive
and supports including additional electronics right on the sensor which can
compensate for shortcomings in the basic technology. Both types of sensor work
well if properly implemented but in higher-end cameras CMOS is more popular
because of the cost advantage (making large-size sensors affordable) and much lower power consumption.
The sensor
comprises millions of light responsive sites called pixels (picture
elements). Each pixel is further subdivided into units responding individually to each of the
three primary colors (in practice a "Bayer" arrangement of sub-pixels is usually
used which facilitates collecting separate luminosity information). Light is focused on the sensor by the lens and each
sub-pixel
then accumulates electrical charge in proportion to the amount of
exposure and strength of each of the primary colors. When the exposure is
complete the shutter closes and the charges in each pixel are converted into
numbers by an Analog to Digital (A/D) converter. These numbers are then written
to re-writable storage which is your SD card, Compact Flash (CF) card or
Memory Stick (Sony). The sensor is then drained of all charges in preparation for the
next shot. Immediately after the exposure a computer in the camera will almost
certainly apply various types of processing intended to remove imaging defects
and improve the appearance of the image (unless the chosen image format is RAW,
in which case it's up to you to apply appropriate processing at a later time with your computer).
All other things being equal, it is intuitively apparent the more
pixels the sensor is divided up into the more detail it should be able to
record. In practice there are limitations to how small a pixel can be made
before noise begins ruining the image. Numbers sell cameras and pixel counts are
easy to quantify, much like horsepower in the case of automobiles. Unfortunately, too many digital cameras are now sold on the
basis of pixel count. A 5 megapixel camera with a good lens may very well
produce images superior to those from a 10 megapixel unit having a sensor
of the same size.
ISO or sensitivity in a digital camera is made variable by controlling the
amplification or gain of pixel readout. This is a great convenience
compared to film where you literally have to change film to obtain a different
sensitivity. Beyond a certain ISO setting, however, sensor noise in digital
cameras intrudes and becomes visible. You either put up with it or try to remove
it with special software which may also destroy valuable detail. Noise in
digital imaging is a subject of much research, discussion and controversy. The plain fact is that digital is less noisy (grainy) than film at
similar ISOs but a certain fanaticism drives the pursuit of noise-free
images these days. As a consequence, we now sometimes see noise reduction being applied
to high ISO images at
the considerable expense of image detail - an unfortunate
engineering choice.
Digital imaging sidesteps the grain issue plaguing film but introduces a
problem of its own. This is the matter of electronic noise and if it is
visible in the displayed image takes on the appearance of film grain or - even
worse - a colored mottling and blotching of darker areas. Digital noise, when it
does appear, is usually more unpleasant than film noise because of this color
mottling. Nikon recognizes this fact and their cameras apply noise reduction
mostly to the color channels and less to the luminance. For this reason, "Nikon
noise", when visible, is more film-like and less objectionable than might
otherwise be the case (or so we Nikonians like to think).
Sensor noise arises
for many reasons and the following list is only to suggest the range of causes:
- It is impossible to fabricate millions of pixels all having identical
response under identical circumstances. There is always going to be some
random irregularity of sensitivity and this will contribute a background
pattern of variability to the image.
- At any temperature above absolute zero random motion of electrons will
disturb the imaging charges accumulated by pixels and produce
changes unrelated to the image itself.
- Environmental electrical noise (generated by the camera itself) may
intrude on the accuracy of charge accumulation. If you visit dpreview you
will find reports concerning cameras that exhibit "banding" and other
artifacts of sloppy electrical design.
- At short exposures or in dark areas of the image there is liable to be some increased effect from
random variation in the arrival of photons at the pixel sites.
- The small sensors used by snapshooters and superzooms require
very tiny pixels in order to preserve resolution. The smaller a pixel the noisier it
will be. This problem is made much worse by "the pixel game" wherein
manufacturers crowd more and more (smaller and smaller) pixels onto
under-size chips, producing more noise but no improvement in image detail.
The good news is that electronic sensors are liable to prove
responsive to research efforts on behalf of improving their performance. Digital snapshooters and superzooms use the smallest sensors (and the smallest pixels)
so have the most noise. This usually becomes visible in prints at around
ISO 200 or even lower. DSLRs use large sensors and generally deliver clean images up to
ISO 800
or even higher, which is actually a big improvement over film.
So why doesn't everyone just use large sensors?
- Cost rises exponentially with size.
CMOS technology partially addresses this issue which is why you see it in
higher-end cameras, being used in the large sensors these units are equipped
with. CCD sensors are much more costly per unit area but deliver better
native
performance with fewer electronics. They make economic sense in smaller
sizes where their "per pixel" high power consumption is also less likely to prove a
liability.
- Lens design
tradeoffs. A sensor the size of a 35mm film frame
("FF" or "FX" size)
would require a physically enormous and heavy lens if the design objective
were to provide high speed (i.e., high light gathering ability) with a 12X zoom range
and acceptable freedom from imaging defects. You would need a wheel-barrow
or a thick-witted assistant to lug it around.
Some high-end DSLRs do offer 35mm frame-size sensors but are very expensive.
These Full Frame (FF/FX) sensors are especially attractive to photographers with
an inventory of legacy film lenses and wanting to utilize the full covering
power these offer. Noise levels are exceptionally low. Apart from that, there is nothing inherently desirable about the FF format.
After all, 35mm was never intended to be a high quality photographic medium.
Many DSLRs and one or two superzooms now feature mid-size "APS" sensors
("DX" format in Nikon terminology) which,
although a bit smaller than 35mm, avoid most of the noise problems associated with
smallest devices while supporting the design of reasonably powerful zoom lenses
weighing less than a ton.
Nikon equips most of its DSLR models with "DX" format sensors in both CCD
and CMOS technologies. These are around 2/3
35mm frame size. This compromise also facilitates the design
of reasonably compact lenses (and reasonably practical zooms to 11X or greater) while ensuring low
noise at high ISO settings. Nikon offers a series of lenses
designed specifically for DX sensors. Because these do not have to cover so
large an imaging area they are smaller, lighter and (usually) less expensive
than full-format lenses.
When you advance to the next frame in a film camera you get a new, pristine
imaging surface to work with. A digital camera re-uses its imager (the
sensor). This puts digital at a significant disadvantage regarding the
accumulation of dirt that might get in the way of light reaching the sensor. DSLRs, having
removable lenses, are particularly vulnerable to dust penetrating to the camera
interior where some of it will inevitably find its way to the sensor and stick there. If the
particles are big enough they may be visible in the final print, monitor display
etc. It is unrealistic to expect there will never be a dust problem with
your DSLR. Furthermore, there may be semi-opaque, rounded blemishes
arising from condensation. Many new cameras are delivered with dust and blobs
already in-place. If these particles are small enough and seldom visible it is
best simply to do nothing or occasionally remove them from critical images
during processing with the clone brush. If they appear regularly (most noticeably
in even-toned areas such as the sky) then something has to be done.
Some DSLRs now have internal cleaning mechanisms which act by vibrating the
sensor either on command or automatically. The consensus in photo forums is that
this remedy is generally ineffective. This has led to an abundance of
over-priced sensor cleaning systems on the market. Using these is fraught with
risk because they involve placing a bit of pressure on the most delicate part of
the camera - the sensor itself (or, more properly, the anti-aliasing filter
covering the sensor). If you are a bumbler with manual skills inferior
to those of a neurosurgeon you might want to entrust dust removal to an expert.
There is little danger, however, in exposing the sensor for cleaning and gently
applying some compressed air or better, perhaps, using a mini-vacuum (at a
distance) from someone like "Green Clean". If you decide to make physical
contact with swabs and chemicals just be sure to follow the instructions
precisely. You are on your own here. I have had very good results using the
Green Clean product but there are many others. This is a real problem for DSLR
owners the manufacturers need to come to grips with.
Even if you change lenses frequently, you can avoid most dust problems by keeping the outside of the camera clean and by taking care to
switch lenses in as uncontaminated an environment as you can find. In very risky
conditions, plan ahead by fitting a zoom lens which can handle all your project requirements without having to be removed.
There is no such thing as a perfect imaging sensor. Out of the millions of
microscopic pixels comprising the sensing surface a few will always perform
outside of their design specification. The worst ones may be visible as tiny
white or colored specks if you zoom in to 100% or so.
In the worst instances
these defects may even be visible in larger prints.
Unless you are very
lucky with a particular unit, all digital cameras suffer from this defect to
some degree. Fortunately, the consequences are usually invisible in
the final
presentation of an image.
The number of defective pixels may increase over time. They are
most likely to be problematic in shots with long shutter speeds in poor light.
Your options are as follows:
- If the problem is really obvious the camera manufacturer can replace the
sensor under warranty. Manufacturers will expect you to accept a certain
number of these defects. Do not expect perfection because there is none.
It is
not unheard of for a camera serviced for this defect to be returned
with even more stuck pixels.
- If there are only a small number of these and you are seldom, if ever, aware of
them in prints, ignore them or remove the occasional one with the clone brush in
processing. 99% of the time these tiny specks simply disappear in image detail
or are removed by noise reduction.
- Acquire software designed to identify stuck pixels and replace their
values in the image with a blend of surrounding correct pixel information. I
don't have a visible stuck pixel problem with my Nikons but have experimented with "PixelFixer",
a free program, and it does a great job on test images. At the types of
exposures likely to invoke visible pixel defects some Nikon dSLRs automatically
deal with the problem. Silkypix, Nikon "Capture NX" and some other
processing software automatically or, as an option, attempt to "cloak" stuck
pixels under all circumstances.
This may all sound bad but is largely a non-issue. Defective pixels are most
often reported by "pixel-peepers" forensically examining their images at 200% or
looking at frames taken with the lens cap on (the best way to spot pixel
defects). In the real world this is seldom a problem. Don't obsess over
it. If a couple of dots are visible in large prints or on monitor display,
just learn how to manage the situation. For a small number of annoying
incidents, use the clone brush. If you really need to get rid of stick pixels,
use one of the automated software aids such as
"Pixelfixer". This works
extremely well and can be configured to support more than one camera body
through use of stored pixel defect maps.
Back to top
of page
|