Sandau R. Digital Airborne Camera: Introduction and Technology

Подождите немного. Документ загружается.

4.4 Opto-Electronic C onverters 183

depending on the glass ﬁlter, lies in the range>λ/T

∼

0.02 − 0.25 nm/Kin the

temperature range 10–90

◦

4.3.2 Interference Filters

These ﬁlters are made using vapour-deposition by combining multiple thin layers

of dielectric material with different indices of refraction. This is done in such a

way that speciﬁc interferences are generated in transmitted light to obtain speciﬁc

band-pass ﬁlters. It is also possible to deﬁne the transmission curve: steep rising

and falling edges can be placed on the selected wavelengths. These are the ﬁlters of

choice to meet tight tolerance parameters.

The central wavelength shifts linearly towards longer wavelengths with increas-

ing temperature. The shift factor depends on the original central wavelength and

varies between 0.016 and 0.03 nm/K in a temperature range of −50 to +70

◦

(Oriel, 2004).

Owing to the structure of the interference ﬁlters, there is a shift of the wave-

lengths to shorter wavelengths with increasing angles of incidence. In the case of

small angles, this effect can be used to adjust a ﬁlter to a selected central wave-

length, as shown by Equation (4.3-2). In the case of small angles of incidence, there

is no distortion of the pass band or reduction of maximum transmission. Using

= λ



1 − (

∗

)

sin

α, (4.3-2)

the shift of the central wavelength can be determined for α <10

◦

.In this

equation,λ

is the central wavelength at angle of incidence α;λ

, the central wave-

length, normal incidence; n

, the index of refraction of the surrounding medium

= 1.0 for air); n

∗

, the effective index of refraction of the ﬁlter; and α, the angle

of incidence of light.

Since interference ﬁlters are hygroscopic, they must be protected from moisture.

Interference ﬁlters should be chosen if narrow-band ﬁlters with deﬁned steep edges

are required for remote sensing applications.

4.4 Opto-Electronic Converters

Electronic images are formed through illumination and by converting radiation

intensities into electrical signals. Semiconductor image sensors, i.e., CCD and

CMOS components made of crystalline silicon in matrix or line arrays, are normally

used as opto-electronic converters. In general, the advantages of crystalline silicon

image sensors include the excellent electrical properties of this material and good

control of the technologies involved. Both technologies are based on the (inner)

photoelectric effect in which light impinging on a material excites electrons.

184 4 Structure of a Digital Airborne Camera

The CCD is photosensitive, fast and has a large dynamic range. It is relatively

expensive, however, and its manufacture is complicated. Additional components are

required for its drive and for eliminating undesirable effects such as smearing and

blooming. Since CCD sensors require more space and consume more power, they

are better suited to environments in which the emphasis is on high image quality

rather than compactness and portability. They are particularly well suited for use in

astrophotography, research and industry. CCD-based line sensors are used mainly

in scanners.

CMOS sensors are economical with energy and can be made using efﬁcient

standard processes. Since image-processing electronic elements can be integrated

on chips, they are compact in size. Blooming and smearing effects are prevented

through individual pixel drive. This direct pixel drive through pixel match enables

individual image sections to be read out at a high image repeat rate (windowing).

Moreover, the CMOS sensor is designed for a greater temperature range than the

CCD. Recent technological advances have made it possible to mass-produce CMOS

chips at low prices. In view of ongoing development, CMOS technology can be

expected sooner or later to catch up with CCD in terms of the image quality and,

in view of its other advantages, crowd it out. But currently, the CCD sensor is the

proven technology in terms of image sensing and it provides better image quality at

high resolutions.

A distinction is made between digital cameras with area array CCD sensors

(matrices) and those with linear CCD sensors. Depending on the application

(moving or non-moving images, required resolution, etc.), lines or matrices are

preferred.

Linear CCD sensors have proved themselves in scanner technology and, as a

rule, they allow a markedly higher resolution at lower cost. CCD lines are arranged

in the image plane of an object especially for CCD image acquisition from aircraft

or satellites, making it possible to acquire all pixels of an image line oriented at right

angles to the direction of ﬂight. Through the sensor carrier’s own motion, a ground

strip is imaged line by line at an appropriate imaging frequency, hence the term

pushbroom scanning. 24,000-pixel line CCDs consisting of staggered 2 × 12,000

pixel CCD lines (offset by

pixel) have been developed speciﬁcally for the ADS40

airborne camera.

4.4.1 Operating Principle

Since detailed descriptions of operating principles and physical aspects of semicon-

ductor CCDs have already been published in the literature, only brief summaries are

given here.

An image is a two-dimensional sample of light intensity of different colours.

Light impinging on a photocell or a pixel of a (CCD) sensor produces an electri-

cal charge in the form of free electrons (internal photoeffect) and, owing to the

temporary absence of electrons in the crystal, positively charged holes. The number

of electrons increases in proportion to light intensity, i.e., to the number of photons

4.4 Opto-Electronic C onverters 185

(light quanta). The free electrons are then collected in the silicon substrate potential

sink while the holes move out of the silicon substrate. The potential s inks can trap

only a certain maximum number of charges, which in the end determines the dynam-

ics of the CCD sensor. The colour of the light – the energy of the photon plays

only an indirect role. The optical properties (spectral transmission) of the electrode

material determine the quantum yield (photo responsivity) at different wavelengths.

In its simplest form, the CCD is a linear array of tightly packed MOS diodes,

the bias voltage of which is set such that there is a marked depletion of majority

charge carriers on the surface. Basically the same process takes place in each MOS

diode – the principle is shown in Fig. 4.4-1. Incident ultraviolet, visible or infrared

light penetrates a very thin transparent electrode (1) and the translucent oxide layer

(2) and then impinges on the semiconductor material (3). When a voltage is applied

(2)(1)

(3)

Control

voltage

MOS series

Time

P silicon (3)

potential sink (4)

translucent silicon oxide (2)

translucent electrode (1)

P silicon (3)

potential sink (4)

translucent silicon oxide (2)

P silicon (3)

potential sink (4)

translucent silicon oxide (2)

translucent electrode (1)

Fig. 4.4-1 Generation of charge carriers in silicon by incident light

4.4 Opto-Electronic C onverters 187

source follower circuits are used as charge-voltage transformers to ensure this lin-

earity. Longer readout registers (>500 pixels) are frequently split for reading out in

opposite directions from the centre and arranging output ampliﬁers at either end of

such a split readout register.

Figure 4.4-3 shows a typical output stage. It is part of a typical Kodac CCD line

array as shown in Fig. 4.4-7. The principle of the output stage is brieﬂy explained

as follows. The reset clock φ

is “ON” to reset the N-doped side of the ﬂoating

diffusion diode to a positive potential by means of transistor Q1. As soon as reset

is switched to “OFF”, this side of the diode ﬂoats and the output signal charge

can become effective by pushing the charge into the ﬂoating diffusion point using

push clocks φ

and φ

. This generates a voltage that can be evaluated and thus the

output signal has reached its new value. In the next phase, the ﬂoating diffusion

point is reset again and the process starts anew. The level that is produced at reset

is termed reset level (typically 6–9 V; see also the clock diagram in Fig. 4.4-4) and

is superimposed by a characteristic reset noise, part of the CCD noise, which is

explained in Section 4.4.3.

Fig. 4.4-3 Typical CCD output stage (Kodak, 1994)

Fig. 4.4-4 Clock diagram of a CCD output stage (Kodak, 1994)

188 4 Structure of a Digital Airborne Camera

4.4.2 CCD Architectures

For a better general understanding, let us ﬁrst review the various types of CCD

sensors and their practical signiﬁcance in imaging.

Line sensors consist of a single row of adjoining CCD cells (pixels). They sample

the image line by line while either the object, such as a conveyor belt or CCD scan-

ner, is in motion. For colour images, the red, green and blue colour portions must

be acquired in three different exposure phases. Very high resolution is obtained with

line sensors.

Trilinear line sensors consist of three parallel sensor lines, each of which carries a

red, a green and a blue colour ﬁlter on its top side. This enables a colour image to be

captured in a single scanning operation. The ﬁlters are mounted on the chip by the

CCD manufacturer or subsequently on the cover glass according to the customer’s

speciﬁc requirements. Trilinear line sensors provide the maximum resolution and

colour reproduction quality.

TDI line sensors can be regarded as a cross between line sensors and area array

sensors. Unlike linear sensors, they have not one but several photosensitive lines

arranged side by side. The shifting of the image data from one line to the next

is synchronised with the motion of the object being scanned (or of the scanner,

for example, a satellite) and the data are analogously cumulated. This gives TDI

sensors advantages over conventional line sensors, which include reduced noise and,

above all, higher responsivity. This is important in applications with fast-moving

objects requiring exposure times that are too short for normal line sensors, since,

owing to the fact that the lines are superimposed, exposure times of TDI sensors are

effectively longer than those of linear sensors.

Area array sensors (matrices) capture all image pixels simultaneously, enabling

moving objects to be photographed at (virtually) any shutter s peed. They are

optionally ﬁtted with a colour ﬁlter matrix (Bayer mosaic) enabling a colour

image to be captured in one step. The disadvantage here is that the resolution is

diminished.

Interline and frame transfer CCDs use two separate surfaces for recording images

and for transferring the charge. In this way, the readout operation and exposure of

the next image can take place simultaneously. They are used primarily in digital and

video cameras for recording moving objects and for full videos.

Full-frame transfer sensors use almost their entire surface for light conversion,

thereby providing a better optical resolution than interline or frame transfer CCDs.

X3 image sensors (presented for the ﬁrst time in 2002) represent an entirely new

CCD technology, which makes it possible to make full use of the entire complement

of pixels for colour images. X3 is based on the fact that that the depth to which light

penetrates silicon depends on its wavelength. X3 has three superimposed layers with

colour receptors embedded in silicon, so that the entire set of colour data can be

recorded on each pixel. In this way, the colour resolution is virtually three times that

of a conventional area sensor.

Driven by numerous special applications, a wide variety of line and matrix sensor

architectures has come into being. The maximum number of pixels in the case of

4.4 Opto-Electronic C onverters 189

line sensors lies between 10 k and 20 k (unstaggered) and, in the case of matrix

sensors, 10 k × 10 k. Since the technological complexity increases at an enormous

rate, it is extremely difﬁcult to manufacture larger sensors as monolithic units.

4.4.2.1 Lines

As already mentioned, line sensors usually consist of a single photosensitive line

(except in the case of TDI). The horizontal shift register is situated beneath the

line. A charge barrier prevents charges from ﬂowing out into the shift register pre-

maturely during the integration process. The principle is illustrated in Fig. 4.4-5.

light-sensitive pixels

charge barrier

horizontal

shift register

Fig. 4.4-5 Principle of the

line sensor (Engelmann and

Ahner, 2004)

The enormously high data rate is a key problem with larger CCD sensors. For

instance, a line sensor 12,000 pixels long is to be read out at 2,000 lines per second.

This corresponds to a frequency of 24 MHz if only one output is used (Fig. 4.4-7),

which is close to the physical limits. Various methods are used to solve this problem:

in the simplest case, one horizontal readout register can be placed above and one

below the line (Fig. 4.4-6). If the pixels are numbered consecutively from 1 to x,

theoddpixels(1,3,5,7...) are read out downwards, the even ones (2, 4, 6, 8...)

upwards. In this way the readout frequency is halved. This is referred to as the

interleaved mode. The two data streams are processed separately and sorted only

after undergoing analogue-digital conversion.

horizontal

shift register

charge barrier

horizontal

shift register

Fig. 4.4-6 Line sensor with

interleaved pixels

(Engelmann and Ahner,

2004)

190 4 Structure of a Digital Airborne Camera

Fig. 4.4-7 Block diagram of a line sensor made by Kodak (IAI FZK, 2004)

The rate can be increased further by dividing the two readout registers into two

halves that can be read out in opposite directions (in this way the frequency is halved

again) or by subdividing the line into several smaller segments, which are then read

out simultaneously and separately (this method is termed multi-tapping).

4.4.2.2 Matrices

The essential operating principles of the various matrix architectures mentioned in

the foregoing are brieﬂy described below (Engelmann and Ahner, 2004).

Interline Transfer Matrices

This sensor type is the prototype of the CCD matrix and is no doubt the most

common sensor type worldwide. It is used in almost all commercial cameras and

provides very good results even in professional image processing. In the interline

transfer sensor, there are columns with vertical shift registers between the photo-

sensitive pixels, which virtually constitute parallel readout memories for each pixel

(Fig. 4.4-8). These registers are protected from incoming light rays by metal masks.

When the exposure time expires, the image is written at high speed in vertical

registers which are provided as intermediate memories. Then, while these vertical

registers are being read out, the pixels are already collecting photons of the next

image, so neither a shutter nor a synchroniser is required.

The entire sensor is now read out line by line by transferring the charges of all

vertical shift registers step by step, according to the bucket chain principle, down-

wards to the horizontal readout register. The charges are now read out according to

the line architecture described in Section 4.4.2.1 and, depending on the application,

processed into an analogue or digital signal. In this way, the entire sensor is read out

line by line.

4.4 Opto-Electronic C onverters 191

Fig. 4.4-8 Interline transfer

CCD principle according (IAI

FZK, 2004)

Frame-Transfer Matrices

The architecture of frame-transfer CCDs is similar to that of full frame CCDs.

Frame-transfer matrices have a huge parallel shift register, which is divided into

two areas of equal size. These areas are referred to as image array and storage array.

The image array consists of a photosensitive photo diode register that serves as

an image plane and collects the incoming photons on the CCD surface. After the

image data have been collected and converted into electrical charges, the charges

are immediately shifted to the storage array, which is normally non-photosensitive,

to be read out by a serial shift register there (line architecture). The transfer time

from the image array to the storage array depends on the size of the storage array.

Matrices of this type operate in either full-frame or frame-transfer mode. By using

mechanical shutters, a frame transfer CCD can be used to record two images in very

quick succession, a feature that is occasionally used in ﬂuorescence microscopy.

As can be seen in Fig. 4.4-9, the storage array must be covered to prevent interac-

tion with the incoming photons. The image array collects new photons for the next

image while the storage array is being read out. The ability also to do without a shut-

ter or synchronisation is an advantage of this architecture, which makes it possible

to increase the readout speed and increase the image rate.

In some cases, frame-transfer CCDs have problems with image smear effects,

which are caused by simultaneous imaging and storage. Smear effects are limited

to the time the matrix requires to shift the recorded image to the storage array.

Generally, matrices of this type are more expensive than interline matrices, because

it takes twice as much silicon to make them. The essential difference between frame

transfer and interline transfer sensors is that the latter have no separate vertical

192 4 Structure of a Digital Airborne Camera

Fig. 4.4-9 Frame-transfer

CCD principle (IAI FZK,

2004)

readout registers. Instead, the photosensitive pixels are themselves used as vertical

readout registers, so there are no “blind spots” in the pixel ﬁeld.

Full-Frame-Transfer CCD Matrices

Full-frame CCDs have very large pixel ﬁelds, which can provide images with the

highest resolution currently possible. Owing to its simple structure, reliability and

simple production technique, this CCD architecture is in widespread use. The fact

that there are no blind spots in the pixel ﬁeld is important (Fig. 4.4-11). The pix-

els cover the entire area on to which light impinges during the exposure phase

(Fig. 4.4-10). The data are read out ﬁrst line by line in parallel and then in serial

manner.

The principal advantage of this matrix is its 100% light efﬁciency relative to

the surface. To simplify storage and image processing operations, the resolution of

many full-frame CCDs is to the power of two (512 × 512 or 1024 × 1024). To

Fig. 4.4-10 Frame-transfer

CCD architecture

(Engelmann and Ahner,

2004)