GENERAL INTEREST

digital cameras
how do they work?
Digital cameras may
not yet have captured
the imagination of the
consumer market
(although most pho-
tographic retail out-
lets stock them), but
with enhanced image
resolution made pos-
sible by recent devel-
opments, and price
falls they will take an
increasing slice of the
market over the next
few years. Market
research* indicates
that the digital cam-
era market in the USA
will grow from
US$240 million in
1997 to US$ 930 by
the end 2004.
photo: Philips
Europe, too, will see
a dramatic increase
in this market,
although spending All the well-known names are active in extended graphics adaptor type, fol-
on cameras has fallen the digital camera market: Agfa, lowed by vertical grid array cameras
Canon, Casio, Epson, Fuji, Kodak, and greater than megapixel types.
steadily over the past Mustek, Olympus, Ricoh, but few pro- Market researchers* expect that the
few years. This article duce models in each product segment.
There is currently no clear market
megapixel camera will retain its lead,
followed by greater than megapixel
takes a brief look at leader. Manufacturers are currently types and extended graphics adaptors
building up their product portfolios by 2004. The vertical grid array cam-
how a digital camera and concentrate on the increasing era’s market share will then have
works. sophistication of the image resolution dropped to below 1 per cent. This arti-
provided by their cameras. Prices in cle will concentrate on megapixel cam-
the consumer market at the time of eras.
writing (July 1998) vary from £200 to
By our Editorial Staff £700. DEVELOPMENTS
In terms of product types, the Digital cameras are about to become a
megapixel camera is leading the field. real force in the market. Owing to
*Source: Frost & Sullivan Ranking in second place is the recent developments, manufacturers

Elektor Electronics 10/98

22
 energy contained in the incident pho-
Photon Photon
1 tons by an amount equal to the energy
difference between the empty and
Gate valence bands. This means that for a
charge to be moved, the energy of the
SiO2
incident light must be greater than the
forbidden band. In CCDs, this nor-
n+ n+
mally means that photons can create
an electron/hole pair if their energy is
greater than 1 eV (electron-volt) and
their wavelength is <1 µm. After this
energy has been stored in the sub-
strate, the negative electrons and pos-
p-Si
itive holes must be separated. This is
effected by applying an electric field
a b 980081 - 11 across the substrate, whereupon the
electrons are freed and the holes dis-
Figure 1. Two photon-convert- appear into the substrate.
ing cells: (a) a metallurgical Unfortunately, it is virtually impos-
are finally able to np-junction, and (b) a voltage- lens into electrical sible to move a free electron since its
produce a digital induced np-junction. signals. In most energy is so tiny. Therefore, the
camera that is digital cameras energy, that is, the free electrons are
cost-efficient, has this task is per- gathered over a given period to make
a reasonably storage capacity, and has formed by a charge-coupled discrete packets of charge available.
a resolution that makes it suitable for device–CCD. These packets are then moved to the
a number of applications. A CCD is a semiconductor storage output, for which a small capacitor is
Although a breakthrough is in device in which an electrical charge is used.
sight, developments are nowhere near moved across the surface. Zeros and Figure 1 shows two variants of a
at an end. Two years ago, the few dig- ones are represented, respectively, by photo-sensitive cell: in (a) a metallur-
ital cameras on the retail market had a the absence or presence of a charge. In gical n-p junction and in (b) a voltage-
resolution of about 280×340 pixels; most digital cameras, the charge trans- induced n-p junction. Both use a
today, some of the better-quality digi- fer system, in which a charge is created p-type substrate. The separation of
tal cameras have a resolution by an impinging photon, is contained holes and electrons is effected by an
approaching one million pixels. Of in MIS (metal insulated semiconduc- electric field across the n-p junction.
course, compared with film, this reso- tor) or MOS (metal oxide semiconduc- When the fieldstrength across the
lution is low. A film has not less than tor) capacitors fabricated on a single junction is reduced, the capacitance of
100 lines per millimetre, which means crystal wafer. the capacitor diminishes and the den-
that a standard 35 mm film image has Photons, which may be considered sity of the charge carriers increases.
a resolution of some 2400×3600 pixels. as elementary particles of light, pass This means that the sensitivity of the
It is expected that within a few years through the lens of the camera on to sensor can be adjusted with a control
there will be digital cameras in the the CCD. The energy contained in a potential.
shops with a resolution of two million photon is converted by the CCD into
pixels, which is, of course, still signifi- an electron/hole pair. If the total CHARGE TRANSPORT
cantly less than the 8.5 million of con- energy is sufficient, electrons may pass The next link in the imaging chain is
ventional film. from the valence band to the empty the transport of the packets of charge
(conduction) band. This causes a hole from the integrating sites towards the
TECHNOLOGY in the valence band. This charge output of the device. There are two
The technology that has made digital movement ways of doing
cameras possible comes partly from reduces the this: either
the semiconductor industry and partly amount of Figure 2. Read-out structure to con- with a MOS
from the film industry. nect the photodiode to the outside
world: (a) a MOS switch, and (b) a
The semiconductor industry has
CCD shift register.
developed compact , high-resolution
image sensors, while the film industry
2
Photon
has developed advanced compression
algorithms, such as JPEG (Joint Photo-
graphic Experts Group), which are
able to convert the enormous amount n+ n+ n+
of data that form a picture into very
compact files.
Today, microprocessors used in dig- p-Si a
ital cameras are able without any dif-
ficulty to compress in a very short time Photon
a digital file of several Mbytes to some
hundreds of Kbytes. These processors
are complemented by memory cards
that can store these data in semi-per- n+ n+
manent form.

LIGHT CONVERTER p-Si b

A digital camera needs a converter that 980081 - 12
transforms the light incident on the

Elektor Electronics 10/98 23

switch containing a sense line or with
a CCD shift register (see Figure 2). In
both, the imaging cell or pixel consists
of a photodiode constructed on a
p-type substrate. The choice between
a MOS switch with a sense line and a
CCD shift register depends to some
degree on the application. Both have
their advantages and disadvantages.
For the MOS switch with sense line,
the fabrication technology is rather
simple, but the switch connects the
small capacitance of the pixel to the
rather large capacitance of the sense
line. This arrangement causes the sig-
nal-to-noise factor to be rather poor.
On the other hand, the technology
for the CCD shift register is more com-
plex in production and requires a clock
to shift the packets of charge to the

photo: Sony
output capacitor via various interme-
diate capacitors. However, the charge
packet taken from the small pixel
capacitance is transferred to the small
capacitance of the output diffusion
and this results in an excellent signal-
to-noise factor. camcorders. see Figure 3. The CCD shift register is
With reference to Figure 2, the con- shielded from any incident light,
version from a packet of charge to a FRAME TRANSFER which means that it can be used as a
voltage at the output pin of the imager Digital cameras use frame transfer (FT) buffer memory.
is done in a classical way: sensing of CCDs, which are different from other The cycle of operation is then as fol-
the voltage changes on a floating types of CCD in the way imaging data lows. In the mode in which light is
n+-region by means of a source-fol- are transported from the light-sensitive registered, all cells are set to the inte-
lower. cell or pixel to the output. In FT-CCDs, grating mode. One part of the CCD
MOS cells, that is, cells that use MOS cells is connected to a high direct volt-
IMAGER CONFIGURATIONS capacitors, are used. Since the imag- age, and another to a low direct volt-
So far, only the operation of a single ing elements, the light-sensitive sensor age. In this mode, photons create a
imaging cell or pixel is explained. In and the capacitors are fabricated in charge, which are gathered into pack-
practical applications, images can be MOS technology, they can be, and ets. At the termination of a defined
built up in a one-dimensional path, for often are, combined in a single design. integration period (in a camera, this is
instance, facsimile, or in a two-dimen- Each photo-sensitive CCD array is the shutter time), the CCD shift regis-
sional configuration, for example, extended by a CCD shift register of ters ensure that their charge is stored
home video or equal length— in the light-immune part of the array.
Figure 3. Device architecture of a The charge transport takes place as
frame transfer image sensor. quickly as possible to prevent mutila-
tion of the data.
3 When all packets of charge have
been transported, a start is made with
reading the CCD. During this phase of
Photo-sensitive CCD array

the process, the packets of charge on

one and the same horizontal line (but
on different vertical lines) are clocked
to a CCD output register, whereupon
the packets are shifted to the output
(parallel-to-serial conversion) and con-
verted into a direct voltage. Once a
line has been processed, the next one
is clocked to the output register. This
process continues until all lines have
been read. The video signal is then
available.
Memory array

In principle it is possible for the

photo-sensitive part of the CCD to be
active while data are read from its
light-immune part.

COLOUR
Since all CCD cells react to incident
light in a similar manner, the devices
are suitable for black-and-white imag-
Output Horizontal output CCD register ing only. For colour operation, the cells
980081 - 13
are combined with colour filters to

24 Elektor Electronics 10/98

 ccd chip formats
The roots for describing the size of imaging sensors in inches go back to the time when there were only vidicons. A
vidicon with a diameter of 1 inch (25.3 mm) had a rectangular, active window with a diameter of 0.6 in (16 mm). This
format has been retained until today.
CCDs are available in various sizes: 1 in, 2/3 in, 1/2 in, and 1/3 in. Nowadays, 1 in chips are used rarely, whereas
1/2 in and 1/3 in types have experienced a constant growth in applications, mainly in the field of surveillance, miniature
cameras, and home video cameras
Reducing the active sensor surface results in smaller pixels, eventually lowering the resolution. For most applica-
tions, a highly detailed picture is more important than the size of the CCD and thus more important than the size of the

1" 2/
3" 1/
2" 1/
3"
5
11 6.

4.8

3.3
16

6.6
8
9.6

4.4
6.4
8.8
980081 - 16
12.8
dimensions in mm

camera. For instance, the Olympus C-1400L uses a 2/3 in CCD containing 1.4 million pixels. The horizontal resolution is
1280 pixels, and the vertical, 1024 pixels. Note that this deviates somewhat from the usual 4:3 picture ratio.
roots for describing the size of imaging sensors in inches go back to the time when there were only vidicons. A vidi-
con with a diameter of 1 inch (25.3 mm) had a rectangular, active window with a diameter of 0.6 in (16 mm). This for-
mat has been retained until today.
CCDs are available in various sizes: 1 in, 2/3 in, 1/2 in, and 1/3 in. Nowadays, 1 in chips are used rarely, whereas
1/2 in and 1/3 in types have experienced a constant growth in applications, mainly in the field of surveillance, miniature
cameras, and home video cameras
Reducing the active sensor surface results in smaller pixels, eventually lowering the resolution. For most applica-
tions, a highly detailed picture is more important than the size of the CCD and thus more important than the size of the
camera. For instance, the Olympus C-1400L uses a 2/3 in CCD containing 1.4 million pixels. The horizontal resolution is
1280 pixels, and the vertical, 1024 pixels. Note that this deviates somewhat from the usual 4:3 picture ratio.

make them react to the green, blue or developed for them: the progressive with this arrangement, the camera is
red component of the incident light CCD. fitted with advanced software. The
only. Since the human eye is more sen- quality of this determines to a very sig-
sitive to green than to the other PROGRESSIVE CCD nificant degree the quality of the out-
colours, there are more green-sensitive Progressive CCDs use square pixels put image (picture). Finally, a progres-
cells than red and blue ones. which are filtered in the primary sive CCD captures the picture in one
The measured light intensity per colours: red, green and blue (RGB). operation, that is, it does not use two
cell is divided into 256 levels of bright- Moreover, each pixel is associated with halves (frames). Mutilation of fast
ness. In this way, each composite pixel only one primary colour. To ensure moving objects therefore does not
gives 2563 shades of colour, so that true that a perfect image is constructed occur.
colour operation is possible.
There are two types of CCD: one
for video cameras and the other for
film cameras. CCDs for video applica-
tions have rectangular cells and are fil-
tered with cyan, magenta and yellow
filters. Moreover, in these CCDs use is
made of the fact that television pic-
tures are built up from two halves
(frames).
It might appear as if this type of
CCD could also be used in film cam-
eras, but this is not so because in the
case of fast moving objects the differ-
ence between the two frames would
be so large that serious distortion
photo: Olympus

would ensue. Nevertheless, this type

of CCD is easy and inexpensive to pro-
duce and it is therefore used in inex-
pensive digital film cameras. Up-mar-
ket digital cameras use a CCD specially

Elektor Electronics 10/98 25

4 5 source file
white-white-bright red-pink-red-dark red-red-pink etc.

source file light compression

white-white-white-white-white-red-red-red-yellow-yellow etc. 2x white-2x dark red-3x red-1x bright red etc.

heavy compression
lossless compression
2x white-6x red etc.
5x 3x 2x

980081 - 14 980081 - 15

Figure 4. Picture data may be compressed loss- Figure 5. Much more severe compression is pos-
less, for instance, by LZW compression as sible if certain data can be made redundant.
sketched. Illustrated is how the degree of compression
determines the loss of detail.

To ensure optimum results from the compressed in the camera with the aid made redundant. In such compression
CCD, taking account of various prop- of a microprocessor. Of a picture con- algorithms, for instance, JPEG, use is
erties, the device has twice as many sisting of 1028 × 768 pixels, each pixel made of the fact that the human eye
green-sensitive cells as red- or blue- must be stored with a resolution of 24 can perceive only about 2000 shades of
sensitive ones. bits. This is equivalent to 2.25 Mbyte of colour, which is appreciably fewer
It should be noted that in (good) digital data. In other words, the inter- than the 16.7 million that are regis-
digital cameras a separate sensor is nal memory of a camera, usually tered.
used for each primary colour. The 2–4 Mbyte, would be able to contain How such an algorithm analyses a
specified resolution does therefore only a few pictures. series of colour shades and replaces it
conform to the actual number of sen- The solution to this problem is an with a much more compact series is
sors. It might be thought that each integral compression algorithm. A dis- shown in Figure 5. The higher the
pixel is built up by three sensors, each tinction must be made between loss- compression, the more detail is lost
reacting to a different primary colour. less compression and redundant-bit and the poorer the quality of the
However, manufacturers have taken a compression. With loss-less compres- reproduced image. Nevertheless, with
different route by computing the sion, for instance, the TIFF (Tagged the use of this kind of algorithm and
desired colour data with the aid of Image File Format) with LZW (Lempel without much discernible loss of qual-
refined algorithms. The colour infor- Ziv Welch – a Unisys patent) com- ity, a 2 Mbyte file can be reduced to
mation for each pixel is therefore the pression, use is made of the data struc- 100 Kbyte or less. The data so obtained
result of an arithmetic analysis in ture. Sequential series of identical may be stored in the internal memory
which the data of adjacent pixels are information are clustered as shown in or on the added memory card.
also taken into consideration. Figure 4, which results in a significant [980081]
compression of data.
COMPRESSION Much better efficiency is provided
When the digital data representing a by redundant-bit compression, in
picture have been gathered, they are which, as the name implies, data are

MOS or CMOS?
Currently, CCDs are produced in MOS technology, which has several disadvantages. For example, it is not suitable for
energy-saving circuits, and it is a deviant production pro-cess.
The semiconductor industry pre-fers energy-saving CMOS technology and researchers are therefore working
on the development of CCDs in this technology. Recently, it was announced that the first CMOS CCDs had been produced.
This will, in time, bring down the price of CCDs and, perhaps more importantly, it will become possible to add intelli-
gence to the device. For example, it will then be possible for the data of each and every pixel or cluster of pixels to be
processed on the CCD.
Moreover, a single pixel may be accessed so that new functions, such as picture analysis, can be provided by
the CCD.
The Fraunhofer Institute for Microelectronic Circuits recently announced the design of a CMOS CCD with
120,000 imaging cells and associated logic circuits.
Another research institute has succeeded in developing a single-row photosensor 2048 pixels wide and an
exposure-time range from 100 ns to 4 s. Each imaging cell in this design has its own read-out amplifier, a buffer and
dark compensation. This enables the CCD to be used even at very low light intensities.

26 Elektor Electronics 10/98