This diagram illustrates the general layout of the most common type of CCD array, the Interline Transfer CCD. The CCD is composed of precisely positioned light sensitive semiconductor elements arranged as rows and columns. Each row in the array represents a single line in the resulting image. When light falls onto the sensor elements, photons are converted to electrons, the charge accumulated by each element being proportional to the light intensity and exposure time. This is known as the integration phase. After a pre determined period of time the accumulated charge is transferred to the vertical shift registers.
In cameras conforming to the video standards mentioned above the charge transfer to the vertical shift registers is accomplished in two stages. Initially the charge in the odd numbered rows is transferred, followed by the even rows. Next the charges in the vertical registers are shifted into the horizontal shift register and clocked to the CCD output. Consequently all the odd rows are clocked out first (odd field) followed by all the even rows (even field). The rate at which the charge from the horizontal shift registers is clocked out is governed by the number of elements (pixels) per row and the video standard the camera complies with.
An
inherent problem associated with the interline transfer CCD lies in the
fact that the vertical shift registers running across the array are
insensitive areas and as such act as blind spots. One way of overcoming
this is to fabricate micro lenses over each element thereby increasing
the effective area of the cell. The lenses also help with the smaller
format CCD. Because of the electrical characteristics of the
semiconductor substrate on which the CCD is formed each cell has an
absolute minimum separation from adjacent cells. Therefore smaller CCDs
require smaller cells. Reducing cell size reduces the amount of
accumulated charge, using lenses increases the incident light.
Another way of overcoming the problem caused by the vertical shift registers is to do away with them and utilize a different charge transfer mechanism. Frame Transfer CCDs do exactly that. This type of CCD has a separate storage area into which the charge is directly transferred from each cell. This process has to be performed rapidly in order prevent blurring as transfer occurs during the exposure time. Once in the storage area the charge can be clocked out in a similar manner to the interline transfer device.
CCD Characteristics
There are a number of key CCD variables that
characterise a particular CCD. These are the size classification, pixel
size, spectral sensitivity and signal to noise ratio.
CCD arrays come in a number of size formats as shown in the table below:
Size
|
Width (mm)
|
Height (mm)
|
Diagonal (mm)
|
1"
|
12.7
|
9.525
|
15.875
|
2/3"
|
8.8
|
6.6
|
10.991
|
1/2"
|
6.4
|
4.8
|
8.0
|
1/3"
|
4.8
|
3.6
|
6.0
|
As you can see from the table the CCD size is
generally specified in inches. It is a measure of the length of the
diagonal. However just because you have say a 1" CCD it does not mean
you will have a CCD with a 1" diagonal. The size classification is
another throwback to the Vidicon tubes used in sensors of old. Instead a
CCD is assigned to the class whose diagonal is at least the length of
the sensor. For example the Dalsa 2M30 mega pixel camera has a CCD with a
14.8mm diagonal, this will therefore be classified as a 1" CCD. The CCD
size classification is important when selecting an appropriate lens and
when considering the required field of view.
The pixel size is important on two fronts. Firstly
the smaller the pixel the less light will fall upon it. This has the
effect of reducing the sensitivity and may have implications on the type
of lighting system used or the integration time. Secondly the aspect
ratio needs consideration. For example, in a gauging application where
the objective is to determine the distance between two features in the
image a square pixel could easily render different results to a
rectangular pixel, the rectangular pixel effectively stretches the image
in the x direction. The image with rectangular pixels can obviously be
compensated for but it increases the complexity of the gauging algorithm
and consequently the inspection time increases too. In these
applications it is best to use cameras whose CCDs have square pixels.
The spectral sensitivity is a measure of the response
of the CCD to differing wave lengths. The response of a CCD is
different to that of a human eye. Human eyes respond to wavelengths in
the range 400 to 700nm whereas a CCD typically responds from 200 to
1100nm. This obviously includes infra red light. In applications where
you are interested in the visible spectrum care needs to be taken to
shield the work piece from excessive infra red otherwise the image could
become over brightened or the sharpness could be compromised due to
chromatic aberrations in the lens. On the other hand your application
may be best suited to inspection in the infra red region. By using an
infra red light source and a suitable band pass filter on the camera
lens the system would become very tolerant to changes in ambient light
levels.
The signal to noise ratio is a measure of the dynamic
range of the CCD. Most cameras operating in normal conditions have a
signal to noise ratio of just under 50dB. This is generally represented
by an 8 bit brightness level. (8 bits gives 256 levels, or roughly 48dB)
The signal to noise ratio is often limited by the Dark current produced
by the CCD. The dark current is caused by thermal excitation and
typically doubles for every 7°C rise in temperature at the sensor and
increases linearly with integration time. When using cameras with high
signal to noise ratios it is important to carefully consider thermal
management if the full range is to be reliably utilized.