Sophisticated
technology goes into the design and manufacture of your digital camera.
Understanding some of that technology can help you to predict how the
camera will fare when you’re shooting in a variety of situations. For
our annual How-To issue, we’re including this article on the anatomy of
an image sensor. The overall image quality in your photographs is
dependent upon a number of factors in addition to sensor resolution.
Lens quality, in-camera processing, compression algorithms and just
plain old shooting technique all work together. Still, the image sensor
is the core around which the rest of the camera is built.
CCD
Let’s examine how CCDs operate. The CCD is comprised of a series of
photosites in a grid pattern. Each photosite contains a light-sensing
device (a photodiode) and a storage area to hold the charge created by
the photodiode. When light hits the photodiode, it converts the light
into electrons, or a charge. The more light that hits the photodiode,
the greater the charge.
The next step in the process is to move the charge out of the
photosites and into another area of the chip called a transfer
register. Row by row, each row of charges is moved vertically and
passed into the transfer register. When a row reaches the transfer
register, the photosite charges are read out (horizontally) one by one,
converted into a voltage and amplified. Once one row has been read out,
its charges are deleted and the next row drops down into its place to
be read out. This coupling of one row to the next is where the device
gets its name. Further processing of the signal is required to convert
the analog voltage into a digital signal. This A-to-D processing is
done on a separate chip.
In order to sample the color information of the light hitting the
photodiodes, a series of filters are embedded into the chip so that
some photosites are reading red light, some blue and others green.
CMOS
Now let’s look at a CMOS. Like a CCD, a CMOS image sensor is also
comprised of a series of photosites arranged in a grid pattern;
however, the makeup of each photosite differs from the CCD. A CMOS
photosite contains a photodiode for converting photons or light to
electrons, but rather than storing the charge and processing the data
in another part of the chip (or even outside of the chip), some of the
processing is done at the photosite itself. Each photosite contains a
converter for changing the charge to a voltage, an amplifier to
increase the very low signal coming from the photodiode, and circuitry
to reduce noise in the signal. There’s a series of grid connections
among the photosites in order to read out the data, and while they
might be arranged in rows and columns, the design allows for accessing
each photosite directly as opposed to reading out row by row.
As far as sampling the color information, with most CMOS image sensors,
a similar technique to the CCD is used to filter the photosites so that
they receive the appropriate colors of light. One manufacturer, Foveon,
uses the properties of silicon itself to filter out the various colors
of light. This process allows each photosite to capture red, green and
blue rather than just one color. (The Foveon image sensor is currently
only available on Sigma digital SLRs, but we expect to see other camera
manufacturers employing the sensor in the future.)
Comparing Design Differences
So how do these design differences compare? Let’s break it down into a few specific topics.
Power Consumption. It turns out that in order to read out all of
that data from the photosites in a row-by-row and then pixel-by-pixel
fashion in a CCD, there are all sorts of sophisticated timing signals
that need to be generated. And while it’s technically possible to put
all of those timing generators and other necessary processing functions
on the CCD, it’s not very economical to do so.
With a CMOS image sensor, many of the processing functions can be built
into the chip during manufacture much like Pentium CPUs have onboard
memory built right into the chip. So a CCD requires additional
integrated circuit chips compared to a CMOS image sensor in order to
process the data. Anytime you add circuitry, you add power consumption.
Therefore, a CCD will consume more power than a CMOS image sensor.
Speed. By having to shift charges around the CCD and parse out
the data one row at a time, a CCD can take a long time to create an
entire image. With the CMOS sensor having individual charge converters
and amplifiers in each photosite, the whole image can be read out much
faster.
Yet another speed increase is found with the CMOS sensor if you use
special capture functions that utilize only a part of the sensor. Let’s
say that you’re capturing an image using only one half of the sensor.
Since the CMOS chip can access only the pixels it needs, the frame rate
of the image capture is increased. With a CCD, even if you’re only
using the middle part of the sensor, the chip still has to shift the
data from all the photosite rows, one by one, until it gets the data it
needs.
Sensitivity. If you examine each photosite of a CMOS, you’ll see
that the power-saving and speed-enhancing circuitry of converters and
amplifiers takes up space. That extra space can’t be used for capturing
light. On the other hand, a CCD sensor might capture 100 percent of the
light hitting the photosite.
Noise. CCDs essentially have a single amplifier through which
all of the data travels. That and the fact that extra processing is
done off the image sensor means the noise on a CCD is largely kept at
bay.
A CMOS chip may contain millions of tiny converters and amplifiers in
each photosite. In a perfect world, all of those converters and
amplifiers would work exactly alike, no matter the operating
environment. But this isn’t a perfect world. As all of the amplifiers
begin working at different efficiencies, they induce fixed-pattern
noise into the data that they’re amplifying. (What’s fixed-pattern
noise? Imagine taking a picture with the lens cap on and having some of
the pixels register black and some register a little lighter than
black. That’s an approximation of fixed-pattern noise.) Also, having
that additional circuitry all on the same chip can add noise.
Blooming. Blooming occurs when the electron charge leaks from
one photosite to the other. It can be seen in the image as streaks
anchored at small high-brightness areas. CCDs are prone to blooming
when individual photosites are oversaturated.
Since CMOS sensors convert the charge to a voltage right at the
photosite, it’s more difficult for the charge to leak to adjacent
photosites. CMOS chip builders also can add pixel-reset circuitry right
at the photosite to minimize overloading (remember all that extra
circuitry they can add directly to the photosite?). So by its inherent
design, CMOS sensors generally are immune to blooming.
Of course, the question is, Which sensor is better? It’s an impossible
question to answer definitively. Obviously, there are pluses and
minuses with each technology, but the manufacturers aren’t ignoring
these differences. For example, CCD designers have gone to different
read-out methods such as reading out the entire sensor at once in order
to increase speed, while CMOS designers may incorporate special
noise-reduction circuitry to reduce photosite noise, or they may add
lenses to the individual photosites to increase sensitivity. CMOS chips
also are being produced with more on-chip functions that take advantage
of the flexibility of CMOS technology.
Although the image sensor is the heart of the digital camera, there’s
more to a digital camera than just the sensor. The data coming out of
the chip needs extensive processing to achieve an image to store on the
media card. Image-processing software algorithms are sophisticated
custom software embedded into the camera to minimize the minuses and
take advantage of the pluses of the image sensors.
Much like you wouldn’t buy a car based only on the engine under the
hood, you wouldn’t want to decide on a camera without considering the
whole picture.
