E05 Handbook Image Sensors

Image sensors CHAPTER 05
1 CCD image sensors 6 Photodiode arrays with amplifiers

1-1 Structure and operating principle 6-1 Features
1-2 Characteristics 6-2 Structure
1-3 How to use 6-3 Operating principle
1-4 New approaches 6-4 Characteristics
6-5 How to use
2 NMOS linear image sensors

7 InGaAs linear image sensors
2-1 Features
2-2 Structure 7-1 Features
2-3 Operating principle 7-2 Structure
2-4 Characteristics 7-3 Operating principle
2-5 How to use 7-4 Characteristics
7-5 How to use
7-6 New approaches
3 CMOS linear image sensors

3-1 Features 8 InGaAs area image sensors
3-2 Operating principle and characteristics
3-3 New approaches 8-1 Features
8-2 Structure
8-3 Characteristics
8-4 New approaches
4 CMOS area image sensors
4-1 Features
4-2 Operating principle and characteristics 9 Applications
4-3 New approaches
9-1 DNA sequencers
9-2 ICP AES equipment
9-3 Optical emission spectrometers
5 Distance image sensors 9-4 Spectrophotometers
9-5 Grain sorters
5-1 Features 9-6 Optical channel monitors
5-2 Structure 9-7 Security and room access control,
5-3 Operating principle obstacle detection, and shape recognition
5-4 Characteristics 9-8 Detector for prime focus camera of Subaru
5-5 How to use telescope
9-9 Asteroid explorer Hayabusa
1
Image sensors
For many years Hamamatsu has developed image sensors for measurement in broad wavelength and energy regions from
infrared to visible light, to ultraviolet, vacuum ultraviolet, soft X-rays, and hard X-rays. We provide a wide range of image
sensors for diverse applications and meticulously respond to customer needs such as for different window materials,
filter combinations, and optical fiber couplings. We also supply driver circuits that are optimized for sensor evaluation or
installation in equipment, as well as easy-to-use multichannel detector heads.
Back-thinned CCD image sensors are suitable for low-level light detection because of their high UV sensitivity, high S/N, and wide
dynamic range. These sensors are extensively used in scientific and industrial fields such as DNA analysis, spectrophotometry,
and semiconductor inspection systems, as well as in the medical field.
Front-illuminated CCD image sensors are used for imaging and measurement in the visible and near infrared region. Their
applications have been recently expanded to include high-resolution X-ray imaging by coupling them to an FOP (fiber optic
plate) with scintillator for use in medical equipment such as for dental diagnosis and in industrial non-destructive inspection.
NMOS linear image sensors are suitable for precision spectrophotometry because of their high UV sensitivity and superb linearity.
CMOS image sensors are well suited for industrial applications that require small, low-cost, and low-power consumption
image sensors. Distance image sensors are CMOS image sensors that measure the distance to the target object using the
TOF (time-of-flight) method. We also provide photodiode arrays with amplifiers, which have a unique hybrid structure
comprised of a photodiode array with a freely changeable pitch and a CMOS amplifier array chip. These photodiode arrays
serve as sensors for identifying paper money. When combined with a scintillator, these photodiode arrays are also used for
non-destructive X-ray inspection of food and industrial materials.
InGaAs image sensors consisting of an InGaAs photodiode array and CMOS charge amplifier array are used for near infrared
spectrometry, DWDM monitoring, near infrared image detection, and the like.
Hamamatsu also provides flat panel sensors developed for X-ray detection, which combine a scintillator with a large-area
CMOS image sensor made from monocrystalline silicon. (See Chapter 9, X-ray detectors.)
Hamamatsu image sensors
Type Features Lineup
For spectrophotometry
Back-thinned type Image sensors with high quantum efficiency from the For scientific measurements
CCD linear/area image sensor visible region to the vacuum UV region TDI-CCD area image sensor
Fully-depleted area image sensor
Front-illuminated type
Image sensors with low dark current and low noise For spectrophotometry
CCD linear/area image sensor For scientific measurements
Image sensors with high UV sensitivity and excellent Current output type (standard type)
NMOS linear image sensor
output linearity suitable for precision photometry Current output type (infrared-enhanced type)
Voltage output type
Image sensors with internal signal processing circuits.
CMOS linear/area image sensor These are suitable for applications that require low power For spectrophotometry
consumption and device miniaturization. For industrial measurements
Sensors that measure the distance to the target object
Distance linear/area image sensor
using the TOF method. Used in combination with a pulse Distance linear image sensor
modulated light source, these image sensors output phase
difference information when light is emitted and received.
Distance area image sensor
Sensors combining a Si photodiode array and a signal
Photodiode array with amplifier processing IC. A long, narrow image sensor can also be Long and narrow type
configured by arranging multiple arrays in a row. For non-destructive inspection
Image sensors for near infrared region. Easy handling For near infrared spectrophotometry
InGaAs linear/area image sensor
due to built-in CMOS IC. For DWDM monitor
For near infrared image detection
Image sensor/photodiode arrays capable of acquiring CCD/CMOS area image sensor for X-ray radiography
X-ray image sensor high quality X-ray images when used in combination with TDI-CCD area image sensor
an FOS (FOP with scintillator) or phosphor screen non-destructive
Photodiode array with amplifier for
inspection
Flat panel sensor Sensors for capturing X-ray images in real time For X-ray non-destructive inspection
For radiography

2
Energy/spectral range detectable by image sensors (example)
InGaAs linear
image sensor
(long wavelength type)
InGaAs
linear/area
image sensor
Distance
image
1240 sensor
Wavelength [nm] =
Photon energy [eV]
CMOS area
image sensor
CMOS linear
image sensor
NMOS linear image sensor NMOS linear

(without window) image sensor
Back-thinned CCD
Back-thinned CCD (without window)
CCD for X-ray imaging Front-illuminated CCD (without window) Front-

il uminated CCD
Flat panel sensor
1 MeV 100 keV 10 keV 1 keV 100 eV 10 eV 1 eV 0.1 eV

Photon
energy
Wavelength
0.01 nm 0.1 nm 1 nm 10 nm 100 nm 1 m 10 m
KMPDC0105EH
Light level range detectable by image sensors (example)
High sensitivity film (ISO 1000)
InGaAs linear image sensor

(G9201/G9211 series)
NMOS/CMOS linear image sensor

(S3901/S8377 series)
Non-cooled CCD area image sensor

(large saturation charge type S7033 series)
Non-cooled type CCD area image sensor

Cooled CCD area image sensor

Irradiance (W/cm2)
Illuminance (lx)
KMPDC0106EC
3
(1) FT type
1. CCD image sensors The FT type CCD (FT-CCD) is comprised of two vertical
shift registers for the photosensitive area and storage
section, one horizontal shift register, and an output
1-1 Structure and operating principle section. Vertical shift registers are also referred to as
parallel registers, while the horizontal shift register is
CCD image sensors (referred to simply as CCD from now called the serial register or readout register. Transparent
on) are semiconductor devices invented by Willard Boyle electrodes such as made from poly-silicon are generally
and George Smith at the AT&T Bell Laboratories in 1970. employed as the electrodes for the photosensitive area.
CCDs are image sensors grouped within a family of charge When light comes through transparent electrodes into
transfer devices (CTD) that transfer charges through the the CCD semiconductor, photoelectric conversion occurs
semiconductor by using potential wells. Most current and a signal charge is generated. This signal charge is
CCDs have a buried channel CCD (BCCD) structure in collected into the potential well beneath the electrodes
which the charge transfer channels are embedded inside during a particular integration time. By utilizing the vertical
the substrate. blanking period, this signal charge is transferred at high
As shown in Figure 1-1, a CCD potential well is made by speed to the storage section for each frame. Therefore in
supplying one of multiple MOS (metal oxide semiconductor) the FT type, the vertical shift register in the photosensitive
structure electrodes with a voltage which is different from area acts as a photoelectric converter device during the
that supplied to the other electrodes. The signal charge integration time.
packed in this potential well is sequentially transferred The signal charge in the storage section is transferred to
through the semiconductor toward the output section. the output section through the horizontal shift register,
Because of this, the CCD is also called an analog shift register. while photoelectric conversion and signal accumulation
CCDs are essentially semiconductor devices through take place in the photosensitive area. The signal charge is
which a signal charge is transferred. Currently, however, transferred to the horizontal shift register for each line in
the term CCD has come to signify image sensors and the storage section during the horizontal blanking period.
video cameras since CCDs are widely used as image sensors. In the FT type, all areas other than the photosensitive areas
are covered with an opaque metal such as aluminum that
[Figure 1-1] CCD basic structure and potential well prevents light from entering.
P1 P2 P3
Metal
[Figure 1-2] Structure of FT type
Oxide film
Semiconductor
Photosensitive
area
Direction of transfer
Potential
Charge
Vertical
KMPDC0036EB shift
register
CCD types
Storage
section
Currently used CCDs are grouped by their transfer

method into the following five types.
FT (frame transfer) type (two dimensional)
FFT (full frame transfer) type (two dimensional)
IT (interline transfer) type (two dimensional) Horizontal shift register
FIT (frame interline transfer) type (two dimensional) KMPDC0037EA
One-dimensional type (linear image sensor)

(2) FFT type
Types except for the FFT and one-dimensional types
The FFT type CCD (FFT-CCD) basically has the same
are used in general-purpose video cameras. FFT and
structure as the FT type except that there is no storage
one-dimensional types are not suitable for use in video
section. Because there is no storage section, the FFT type
cameras because of their operating principle, and are
is usually used along with some type of external shutter
mainly used in measurement and analysis applications.
mechanism. This limitation makes it difficult to use the
FFT type in video cameras.
4
The operating principle of the FFT type is similar to that [Figure 1-4] Structure of IT type
of the FT type. The signal charge is collected in a potential Transfer gate
well in the photosensitive area during the integration Photodiode
time and then transferred to the output section via the
horizontal shift register during the external shutter closed
period and the like. Vertical
shift
Since there is no storage section, the FFT type can be register
fabricated with a larger number of pixels or with a larger

pixel size while using the same chip size, so the FFT type is
mainly used for measurement camera systems with a slow
Horizontal shift register
frame rate. Most Hamamatsu CCDs are the FFT type. KMPDC0039EA
[Figure 1-3] Structure of FFT type (4) FIT type

The FIT type CCD (FIT-CCD) was developed to solve the
problems of the IT type CCD. The FIT type is configured
basically by adding a storage section to the IT type. In the
Vertical FIT type, as soon as a signal charge is transferred from the
Photosensitive
shift
area
register photodiodes to the vertical shift registers, the charge is
transferred to the storage section at high speeds. The FIT
type therefore ensures reduced smear compared to the IT
type.
[Figure 1-5] Structure of FIT type

KMPDC0038EA Transfer gate
Photodiode
(3) IT type
The IT type CCD (IT-CCD) has an photosensitive area
Vertical
consisting of photodiodes or MOS structure diodes formed shift
register
separately from the transfer section. Recent IT types use
buried photodiodes with a low dark current. Vertical shift
registers are arranged along photodiodes, and horizontal
shift registers and output sections are also configured.
The signal charge produced by photoelectric conversion
Storage
in a photodiode is stored in the junction capacitance section
of the photodiode itself and others. This charge is then

transferred to the vertical shift register during the vertical
blanking period through the transfer gate which is provided
as a switch between the photodiode and the vertical shift
KMPDC0040EA
register. This operation differs from the FT type in that
the charge transfer from the photodiode to the vertical (5) One-dimensional type
shift register is performed for all pixels simultaneously.
In a one-dimensional type CCD, the signal charge generated
Subsequent operations are exactly the same as the FT type
by photoelectric conversion in a photodiode is collected
operation following signal transfer to the storage section,
in the adjacent storage gate. The signal charge is then
so the signal charge is transferred to the horizontal shift
transferred to the horizontal shift register through the
register for every line during the horizontal blanking period.
transfer gate provided as a switch between the storage gate
Figure 1-4 shows a simplified structure of the IT type.
and the horizontal shift register. Charge transfer from the
As with the FT type, areas other than the photodiodes
storage gate to the horizontal shift register is performed for
are light shielded with aluminum, etc. In the IT type,
all pixels simultaneously.
the signal charge is transferred from the storage section
Figure 1-6 shows the structure of a one-dimensional type
to the output section by using the period in which the
CCD. Signal charges from odd pixels in the photodiode array
charge is accumulated in the photodiode. This tends to
are transferred to the upper horizontal shift register, and
cause a phenomenon called smear due to the signal
signal charges from even pixels are transferred to the
charge leaking into the vertical shift register.
lower horizontal shift register. Those signal charges are
alternately detected by a single FDA (floating diffusion
amplifier, see FDA in section 1-1, Structure and operating
principle). Transferring the odd pixel signal charges
and even pixel signal charges to the separate horizontal
shift registers makes it possible to fabricate a photodiode
5
array with a small pitch and to form anti-blooming and (b) Potential
electronic shutter structures.
t1
[Figure 1-6] Structure of one-dimensional type Signal charge
t2
Photodiode array
Storage gate t3
Anti-blooming drain
Anti-blooming gate
Transfer gate
(c) Timing chart
KMPDC0280EA
P1
P2
Charge transfer operation
t1 t2 t3
KMPDC0042EB
CCDs using two electrodes (gates) for one pixel are called
2-phase (drive) CCDs or 2-gate CCDs. Figure 1-7 shows
the operating principle of a 2-phase CCD in which the FDA
signal charge is transferred by applying two clock pulses
with different voltage levels (high level and low level). FDA (floating diffusion amplifier) is the most popular
In a 2-phase CCD, the signal charge is transferred in method for detecting the signal charge of a CCD. As shown
the direction determined by the difference in potential in Figure 1-8, the FDA consists of a node for detecting
created in the semiconductor process. The signal charge charges and two MOSFETs (MOS1 for reset and MOS2 for
is stored beneath the storage electrode. In Figure 1-7 for charge-to-voltage conversion) connected to the node. The
example, the signal charge is stored beneath the storage charge transferred to the detection node is converted into
electrode for electrode P1 by setting electrode P1 to high a voltage by MOS2 via the relation Q = CV. The detection
level (setting electrode P2 to low level) at time t1. node is reset by MOS1 to the reference level (voltage on
It is important in 2-phase CCDs to optimize the overlapping RD) in order to read the next signal.
of clock pulses. As shown in the timing chart of Figure Noise accompanying the charge detection by FDA is
1-7 (c), clock pulses must cross each other (at time t2) determined by the capacitance of the node but can be
at a level higher than the midpoint of the high and low almost entirely eliminated by CDS (correlated double
levels for P1 and P2 (for example, if the high level is V and sampling) invented by White.
the low level is 0, the cross point should be higher than The signal charge output timing is synchronized with the
V/2). The signal charges can be transferred by setting timing at which the summing gate (SG) goes from high
the clock phase so that P1 alternately goes high and low, level to low level, which is the last clock gate for the shift
while P2 goes low and high. register.
[Figure 1-7] Operating principle of 2-phase CCD [Figure 1-8] CCD output section using FDA
RD
(a) Structure
P1 P2
RG MOS1 OD
N- N-
N
P OG
P1 SG Detection
node
MOS2
Direction of transfer
OS
Signal charge FD External load

resistance
Charge transfer
KMPDC0043EB
6
When used in combination with line binning, signals
Binning of signal charges
from 2 2 pixels for example can be summed as shown
in Figure 1-10. In this case, the signals of two vertical
During CCD operation, a signal charge accumulates in the
lines are first summed by line binning into the pixels of
potential well of every pixel during the integration time.
the horizontal shift register. Next, in the signal charge
In FFT-CCDs, this means that the charge information is
readout by the horizontal shift register, the signals of the
stored in two dimensions at the end of the integration
two horizontal pixels can be transferred and summed by
time as shown in Figure 1-9 (a).
applying just one clock pulse to the SG terminal during
Since clock pulses can be input separately to the vertical
each period of two P1H clock pulses.
shift register and horizontal shift register, an operation
This method is effective in detecting low level light. For
called binning can be performed. Binning is an operation
example, when the incident light level is too low to detect
unique to CCDs, and can be grouped into line (vertical)
with a CCD having 1024 1024 pixels, operating it as a
binning and pixel (horizontal) binning, depending on the
sensor having 512 512 pixels will acquire an image with
direction that the signal charge is added.
higher contrast, although the spatial resolution will be
(1) Line binning lower.
In line binning, the signal of each pixel is summed in the
[Figure 1-10] Pixel binning
vertical direction. As shown in Figure 1-9 (b), the signal
charge of each vertical pixel is sequentially transferred (a) Signal charge flow
and added to one corresponding pixel of the horizontal
shift register by applying a specified number of clock
pulses P1V and P2V to the vertical shift register while the
horizontal shift register clock pulse P1H is halted.
Line binning allows obtaining a signal which is equivalent
to that obtained from a one-dimensional sensor having
a very long photosensitive area in the vertical direction.
Noise intrusion resulting from signal readout can be
minimized since signal readout from the output section is
performed at one time.
(b) Timing chart
[Figure 1-9] Line binning
(a) Signal charge flow
At end of integration time
KMPDC0045EB
Signal charge injection
CCDs have input sources (ISV, ISH) and input gates (IGV,
IGH) each arranged at the respective heads of the vertical
shift register and horizontal shift register as the signal
input terminals for electrical tests. In normal operation,
(b) Timing chart
a specified bias (see datasheets) should be applied to
these test terminals. However, by applying a bias and
clock pulses other than the specified values to these
input sources and gates, a signal charge can be injected
into the shift registers. This will reduce radiation-induced
degradation of the CCD charge transfer efficiency. These
KMPDC0044EA
terminals can also make quantitative evaluations of the
saturation charge and the FDA linearity. A signal charge
(2) Pixel binning can also be injected into the shift registers by connecting
The last gate of the horizontal shift register in a CCD is an a current source to the input sources and shorting the
independent gate called the summing gate (SG). During input gates to P2 for clock pulse input [Figure 1-11]. The
operation not using pixel binning, the SG terminal is charge injected by this method equals the product of the
directly shorted to P2H (or the same clock pulses as P2H injection current value from the current source and the
may be input to SG without shorting SG to P2H). Pixel injection time (reciprocal of CCD drive frequency).
binning can be performed by supplying a different clock
pulse to SG.
7
Qinj = Iinj t (1) [Figure 1-12] Comparison of NMOS and CCD image sensors
Qinj: injection charge [C]

(a) NMOS image sensor
Iinj : injection current [A]
t : injection time [s] Digital shift register
[Figure 1-11] Structure of signal charge injection

section connected to current source Photodiode
(b) CCD image sensor (IT type and one-dimensional type)
Analog shift register

Transfer gate
Photodiode
KMPDC0136EB
KMPDC0137EA
Comparison with NMOS image sensor

Front-illuminated type and back-thinned type
The structure of CCD image sensors differs from NMOS
image sensors, so their specifications and performance In general, CCDs are designed to receive light from the
are also quite different from each other. front side where circuit patterns are formed. This type of
In NMOS image sensors, the signal charge accumulated CCD is called the front-illuminated CCD. The light input
in each photodiode is output to the signal line through surface of front-illuminated CCDs is formed on the front
a MOSFET switch by sequentially addressing the signal surface of the silicon substrate where a BPSG film, poly-
charge by a digital shift register. The operation at this silicon electrodes, and gate oxide film are deposited.
point is performed by supplying TTL level clock pulses Light entering the front surface is largely reflected away
at a constant timing to the digital shift register, so that and absorbed by those components [Figure 1-13 (a)].
NMOS image sensors operate with just a single 5 V The quantum efficiency is therefore limited to approx.
supply, except for the power to external signal processing 40% at the highest in the visible region, and there is no
sections. sensitivity in the ultraviolet region.
On the other hand, CCD image sensors (IT type and one- Back-thinned CCDs were developed to solve such problems.1)
dimensional type) transfer the signal charge accumulated Back-thinned CCDs also have a BPSG film, poly-silicon
in each photodiode to the analog shift register by electrodes, and gate oxide film on the surface of the silicon
turning on the MOSFET switch. The signal charge is substrate, but they receive light from the backside of the
then sequentially transferred by the analog shift register silicon substrate [Figure 1-13 (b)]. Because of this structure,
to the FDA in the final stage and is output. Operating a back-thinned CCDs deliver high quantum efficiency over a
CCD image sensor requires more than one power supply, wide spectral range. Besides having high sensitivity and low
and the clock pulse amplitude must match the specified noise which are the intrinsic features of CCDs, back-thinned
value. CCD image sensors exhibit a low readout noise level CCDs are also sensitive to electron beams, soft X-rays,
from a few to a dozen electrons (e- rms). They also allow ultraviolet, visible, and near infrared region.
a high-speed readout at a pixel rate of 10 MHz or more
[Figure 1-13] Schematic of CCDs
depending on the amplifier bandwidth for the FDA.
Although the noise level of NMOS image sensors runs as (a) Front-illuminated type
high as 3000 electrons (e- rms), they can handle a signal Incident light
charge over 100 times higher than CCD image sensors Poly-silicon electrode
that typically handle several hundred thousand electrons. BPSG film
In general, if the light level to be detected is sufficiently
Gate oxide film
high, using an NMOS image sensor is preferable because
it simplifies the measurement system. In contrast, CCD Potential well
Silicon
image sensors have low noise and ensure an adequate S/N
even at light levels that cannot be detected with NMOS
image sensors. CCD image sensors are therefore suitable KMPDC0179EB
for low-light-level detection.
8
(b) Back-thinned type [Figure 1-15] Device structure of back-thinned CCD
(schematic of CCD chip as viewed from top
Poly-silicon electrode
of dimensional outline)
BPSG film
Effective pixels
Gate oxide film Thinning

Effective pixels
Potential well
Silicon
22 23 21 20 15 14 13
Accumulation layer
2-bevel
Incident light
KMPDC0180EB
V
2 signal output
Thinning
In order for back-thinned CCDs to achieve high sensitivity, 5
4
4-bevel
it is essential to make the silicon substrate thin and to 3
2
12345 H
n
24
activate the photosensitive area. The photosensitive area is 1

12
activated by forming an internal potential (accumulation) 2 Horizontal 11

V=58, 122
shift register
H=512, 1024
slope so that signal charges generated near the backside
3 4 5 8 9 10
light input surface are smoothly carried to the CCD

n
potential wells without recombining. 2) 3) The internal 4 blank pixels 2 signal output 4 blank pixels
potential in an accumulation state is shown in Figure 1-14.
6-bevel 6-bevel
KMPDC0016EE
[Figure 1-14] Internal potential of back-thinned CCD
Front side
Back side
Near infrared-enhanced back-illuminated CCD
Accumulation state
Normal back-thinned CCDs have high quantum efficiency
Incident in the ultraviolet to visible region. However, since the
light
silicon substrate is about 15 to 30 m thick, the quantum
Backside well Poly-silicon efficiency in the near infrared (NIR) region is low. For
(when not in accumulation state) electrode example, the quantum efficiency at a wavelength of 1 m
Potential well
is approx. 20%. Thickening the silicon substrate improves
Gate oxide film sensitivity in the near infrared region, but the resolution
KMPDC0138EB decreases due to the charge diffusion in the neutral region
(undepleted region).
When the back-thinned CCD is viewed from the light input
Near infrared-enhanced back-illuminated CCDs were
side, the horizontal shift register is covered by the thick area
developed to solve such problems. To ensure high sensitivity
of the silicon (insensitive area), and short-wavelength light
in the near infrared region, the following two types are
hardly reaches the horizontal shift register. However, long-
available.
wavelength light may pass through the dead area of the
silicon and may be received by the horizontal shift register. (1) Fully-depleted type
If an external shutter is not used, light will enter the Fully-depleted types use an ultra-high resistance N-type
horizontal shift register during charge integration and wafer to form a thick depletion layer. When the silicon
transfer, and the false signal will be superimposed on top resistivity is the same, the dopant concentration in
of the actual signal. For example, if a time invariant signal N-type wafers can be reduced lower than that in P-type
enters the horizontal shift register, a constant offset will be wafers, so the depletion layer can be thickened even
added to the signal. This effect will be smaller with shorter when the same bias voltage is applied [Figure 1-16 (b)].
horizontal transfer time periods. On the other hand, MPP operation (see Dark current
As necessary, use an external shutter, adjust the light in section 1-2, Characteristics) is not possible because
input position, block the light, and take other effective the bias voltage is applied to the backside. Using a
measures. thick silicon substrate also causes the dark current to
increase, so the CCD must be cooled to -70 to -100 C.
Since the silicon substrate for standard back-thinned
CCDs is thin, it is difficult to fabricate large-area devices.
However, fully-depleted back-illuminated CCDs use a
silicon substrate whose thickness is about 100 to 300 m
over the entire area, so large-area devices can be easily
fabricated.
(2) Fine structure type

A special fine structure in the micron order is fabricated
on the light input surface of the silicon [Figure 1-16 (c)].
9
This structure traps in the silicon the near infrared light There is a trade-off between the readout frequency and
that would normally pass through the silicon to increase readout noise. Increasing the readout frequency shortens
the absorption, which in turn achieves high CCD the readout time but increases the readout noise [Figure
sensitivity. This also suppresses etaloning [Figure 1-51], 1-43].
which becomes a problem during the detection of near Using multiple CCD amplifiers (multi-port configuration)
infrared light. allows parallel readout of pixels and improves the frame
rate (number of frames acquired per second).
[Figure 1-16] Internal structure of back-illuminated CCD
(a) Standard type [Figure 1-18] Multi-port CCD structure
CCD on-chip amp
Incident light
Accumulation layer
10 to 30 m

P type silicon
Gate oxide film
BPSG film
KMPDC0281EA
Vertical
(b) Fully-depleted type shift
register
Incident light
Vbb
(bias voltage applied to substrate)
Accumulation layer
100 to 300 m
N-type silicon
Gate oxide film Horizontal shift register

BPSG film
KMPDC0282EA
KMPDC0140EA
(c) Fine structure type

Incident light Thermoelectrically cooled CCD
P+ Accumulation layer
CCD dark current varies with temperature; namely dark
10 to 30 m
P- P type silicon
Gate oxide film current is reduced by approx. one-half for every 5 to 7 C
BPSG film
decrease in temperature. As with MPP operation (see
N Poly-silicon electrode
KMPDC0456EA
Dark current in section 1-2, Characteristics), cooling
a CCD is an effective way to reduce the dark current and
[Figure 1-17] Spectral response of back-illuminated CCDs enhance the detection limit.
(without window, typical example) Thermoelectrically cooled CCDs contain a thermoelectric
100
(Ta=25 C) cooler (Peltier element) in the package, which efficiently
Fully-depleted type
90 (Td=-100 C) cools the CCD. The cooling temperature is determined by
maximum heat absorption and heat dissipation capacities
80
of the thermoelectric cooler. The following parameters
Quantum efficiency (%)
70
differ depending on the thermoelectric cooler.
60
50 Standard type Maximum current (Imax)

Maximum voltage (Vmax)
40
Maximum heat absorption (Qmax)
30 Fine structure type
20 To avoid damaging a thermoelectric cooler and CCD,

10 always operate them within the ratings specified in the
0 datasheets. Heat dissipation methods are important
200 400 600 800 1000 1200
when using a thermoelectric cooler. The cooling might
Wavelength (nm) be inadequate unless the heat is sufficiently dissipated.
KMPDB0295EC
This is because the temperature on the hot side of the
thermoelectric cooler becomes high. In those cases, an
Multi-port CCD optimal heatsink, forced air/water cooling, or the like is
required. Thermoelectric coolers are designed to cool a
The readout time of a CCD is determined by the number CCD efficiently when used at less than or equal to 60% of
of pixels and readout frequency, and readout usually the maximum current.
takes a long time. For example, when reading out signals As a rough guide, CCDs are cooled to the following
from 1024 1024 pixels at a readout frequency of 100 temperatures when the ambient temperature is 25 C.
kHz, the readout time will be 10 seconds or longer.
10
One-stage TE-cooled type: about 0 to -10 C to the number of vertical stages of the CCD pixels.
Two-stage TE-cooled type: about -20 to -30 C In TDI operation, the signal charges must be transferred
Four-stage TE-cooled type: about -50 to -70 C in the same direction at the same speed as those of the
object to be imaged. These speeds are expressed by
To ensure stable and reliable operation, the thermoelectric
equation (2).
cooler current and heat dissipation condition should be
determined according to the surrounding environment. v = f d (2)
[Figure 1-19] Cooling characteristics of one-stage v: object speed, charge transfer speed
thermoelectrically cooled type (S7171-0909-01) f : vertical CCD transfer frequency
d: pixel size (transfer direction)
(Typ. Ta=25 C)
6 30
Voltage vs. current
CCD temperature vs. current In Figure 1-20, when the charge accumulated in the first
5 20 stage is transferred to the second stage, another charge
produced by photoelectric conversion is simultaneously
CCD temperature (C)

4 10
accumulated in the second stage. Repeating this operation
Voltage (V)
continuously until reaching the last stage M (number of

3 0
vertical stages) results in a charge accumulation M times
2 -10
greater than the initial charge. This shows that the TDI
operation enhances sensitivity up to M times higher than
1 -20 ordinary linear image sensors. (If the number of vertical
stages is 128, the sensitivity will be 128 times higher than
0
0 0.5 1.0 1.5 2.0
-30 ordinary linear image sensors.) Since the accumulated
signal charges are output for each column from the CCD
Current (A) horizontal shift register, a two-dimensional continuous
KMPDB0180EA
image can be obtained. TDI operation also improves
sensitivity variations compared to two-dimensional
TDI-CCD operation mode.
Back-thinned TDI (time delay integration)-CCDs [Figure 1-20] Schematic of integrated exposure in TDI operation
allow acquiring high S/N images even under low-light
Object movement
Charge transfer
First stage
conditions during high-speed imaging and the like. TDI
operation yields dramatically enhanced sensitivity by
integrating the exposure of a moving object. The back- Last stage M
thinned structure ensures high quantum efficiency over
Charge
a wide spectral range from the ultraviolet to the near

infrared region (200 to 1100 nm).
KMPDC0139EA
TDI operation
In CCD operation, a signal charge is transferred to the [Figure 1-21] Imaging examples in TDI operation
output section while being held in potential wells so as (a) Imaging of fast moving object
not to mix with other individual charges. TDI operation TDI-CCD
Direction of
makes good use of this CCD charge transfer principle, (2048 128 pixels)
charge transfer
and it is an effective technique for detecting weak light
and for imaging a moving object or a still object while
scanning it with a CCD sensor that is itself being moved.
Normally, an image focused on the CCD sensor is output
as a signal corresponding to the focused position. This
means that the image focused within the integration
time must stay in the same position on the CCD sensor. Direction of
If, for some reason, the focused position is shifted, object movement
then the image S/N will deteriorate. When an object is

moving, the focused position will shift, causing the image
to blur or, in some cases, no image to appear.
The TDI operation, in contrast, is a unique operation that
captures images of a moving object. In FFT-CCD, signal KMPDC0266EB
charges in each column are vertically transferred during

charge readout. TDI operation synchronizes this vertical
transfer timing with the movement of the object, so
signal charges are integrated by a number of times equal
11
(b) Imaging of fast rotating object
Resistive gate structure
TDI-CCD
(2048 128 pixels)
Direction of charge transfer
In ordinary CCDs, one pixel contains multiple electrodes
and a signal charge is transferred by applying different
clock pulses to those electrodes [Figure 1-24]. In resistive
gate structures, a single high-resistance electrode is
formed in the photosensitive area, and a signal charge is
Direction of transferred by means of a potential slope that is created
object rotation
by applying different voltages across the electrode
[Figure 1-25]. Compared to a CCD area image sensor
which is used as a linear sensor by line binning, a CCD
linear image sensor having a resistive gate structure in the
photosensitive area offers higher speed transfer, allowing
Drum
readout that leaves behind fewer unread charges even if
KMPDC0267EA
the pixel height is large.
In Figure 1-21 (b), when the CCD is put in two-dimensional
[Figure 1-24] Schematic and potential of ordinary
operation and the drum is imaged while in idle, a clear 2-phase CCD
image with no blurring is obtained as shown in Figure P1V P2V P1V P2V
1-22 (a). However, when the drum is rotating, the image is
blurred as shown in Figure 1-22 (b). Shortening the shutter
time captures an unblurred image, but the image becomes
dark as shown in Figure 1-22 (c). Using a TDI-CCD acquires N- N N- N N- N N- N
clear, continuous images with no blurring as shown in P

Figure 1-23 since charge transfer is performed in the same
direction at the same speed as those of the rotating drum.
[Figure 1-22] Imaging in two-dimensional operation

(a) When drum is in idle
KMPDC0320EA
(b) When drum is rotating [Figure 1-25] Schematic and potential of resistive gate
structure
REGL REGH STG TG
(c) When drum is rotating (with shutter time shortened)

Resistive gate
P+ N N- N
P
[Figure 1-23] Imaging in TDI operation
(continuous image during drum rotation)
Potential slope
KMPDC0321EB
Buried photodiode
In the case of a CCD linear image sensor that uses a

photodiode structure in the photosensitive area, low
dark current can be achieved by buried photodiodes.
The buried photodiode has a P + N + P structure with a
thin P + diffusion layer formed on the surface of the
photosensitive area [Figure 1-26]. Since the depletion
layer is distant from the Si-SiO 2 interface, the dark
current can be reduced to a level equivalent to that of a
CCD in MPP operation.
12
[Figure 1-26] Cross section (buried photodiode) that poly-silicon gate electrodes be formed on the
Photosensitive area effective photosensitive area because of their structure,
Si-SiO2 interface which makes the CCD almost insensitive to ultraviolet
P+
light at wavelengths shorter than 400 nm. To make it
N+
sensitive to ultraviolet light, a scintillator material called
P
Lumogen is coated on the CCD surface of some front-
Depletion layer illuminated types. Back-illuminated CCDs, on the other
KMPDC0461EA hand, deliver a high quantum efficiency from ultraviolet
region to near infrared region and also exhibit excellent
stability even during exposure to ultraviolet light.
1-2 Characteristics In the near infrared region at wavelengths longer than 700
nm, the quantum efficiency of standard front-illuminated
CCDs is not so high (this depends on the depletion layer
Conversion factor thickness), but the near infrared enhanced front-illuminated
CCD delivers high quantum efficiency even in the near
The conversion factor is the charge-to-voltage conversion infrared region [Figure 1-28].
ratio of an FDA. When a CCD is cooled during use, it should be noted that
The FDA converts a signal charge into a voltage, which is the sensitivity drops at wavelengths longer than approx.
output from the output end OS as the voltage Vout. 800 nm [Figure 1-29].
Vout = Av Q / Cfd (3) [Figure 1-27] Spectral response (without window)

Av : voltage gain of charge-to-voltage conversion MOSFET (Typ. Ta=25 C)
100
Q : signal charge [C] Fully-depleted
Cfd: node capacitance [F] 90 back-illuminated CCD
80
The conversion factor (Sv) is expressed by equation (4).
70 Standard
back-thinned CCD
60
Sv = q Vout / Q [V/e-] (4)
50
q: electron charge Front-illuminated CCD
40
When the S7030/S7031 series is used: Sv=2.2 V/e-
When the S11071 series is used: Sv=8.0 V/e- 30
20
The node capacitance (Cfd) is expressed by equation (5).
10
0
Cfd = q Av / Sv [F] (5) 200 400 600 800 1000 1200
When the S7030/S7031 series is used: Cfd=48 fF Wavelength (nm)

When the S11071 series is used: Cfd=12 fF KMPDB0280EB
[Figure 1-28] Spectral response of front-illuminated CCDs

Spectral response (without window)
(Typ. Ta=25 C)
50
Figure 1-27 shows the spectral response of front-
NIR-enhanced
Standard type
illuminated CCDs and back-illuminated CCDs. Front- type
illuminated CCDs have no sensitivity in the ultraviolet 40

region, and the maximum quantum efficiency in the

visible region is approx. 40%. In contrast, standard back- 30
thinned CCDs deliver very high quantum efficiency,

which is 40% or higher in the ultraviolet region and 20
approx. 90% at a peak wavelength in the visible region. Lumogen-coated
type
The fully-depleted back-illuminated CCDs use a thick 10
silicon substrate which allows higher sensitivity in the
wavelength range from 800 to 1100 nm than standard
0
back-thinned CCDs. The fully-depleted back-illuminated 200 400 600 800 1000 1200
CCDs also have high sensitivity in the visible region from Wavelength (nm)
400 to 700 nm due to use of a special AR (anti-reflection) KMPDB0205EC
coating process; however the ultraviolet sensitivity is low.

The spectral response range is determined on long
wavelengths by the silicon substrate thickness and on
short wavelengths by the sensor structure on the light
input surface side. Front-illuminated FFT-CCDs require
13
[Figure 1-29] Temperature characteristics of sensitivity (S7031 series) on the CCD effective photosensitive area by vacuum
(Typ.)
sublimation.
5
Lumogen absorbs light at wavelengths shorter than 480
nm and emits light whose center wavelength is around
4
Temperature coefficient (%/C)
530 nm. In other words, ultraviolet light striking the

3 CCD surface is converted to visible light by the Lumogen
scintillator, and this light is then detected by the CCD.
2 Care should be taken when using a Lumogen-coated
front-illuminated CCD because its life under ultraviolet
1 light is extremely short and its sensitivity is highly
dependent on temperature when compared to back-
0
thinned CCDs.
-1
200 400 600 800 1000 1200
Selecting window materials
Wavelength (nm)
KMPDB0383EA
Back-thinned CCDs provide high quantum efficiency
over 90% at the peak sensitivity wavelength around 700
(1) Back-thinned CCD with optimized spectral response
nm, but this is a value measured without a window.
With the back-thinned CCD, various spectral responses The quantum efficiency of CCDs is affected by the window
can be achieved by optimizing the anti-reflection film on material. Hamamatsu mainly uses three types of windows
the photosensitive surface [Figure 1-30]. for CCDs: AR (anti-reflection) coated sapphire (S type),
quartz (Q type), and windowless (N type).
[Figure 1-30] Spectral response (back-thinned CCDs,
without window, typical example) Sapphire is mechanically strong and scratch-resistant
compared to quartz and is stable in high humidity
(Ta=25 C)
100 environments. In addition, since the thermal conductivity
For detecting
90 320 to 560 nm of sapphire is close to that of metal, it is less likely to cause
Standard type
80 moisture condensation and can be hermetically sealed
70 within a metal package, making it an excellent choice as

60 window material. The transmittance of sapphire having
50 For detecting
no AR coating is not so high, but AR coated sapphire has
320 nm and shorter good transmittance which is higher than quartz in the
40
visible region. Hamamatsu thermoelectrically cooled
30
CCDs such as the S9971/S7031 series use sapphire as a
20
standard window material.
10
Quartz is available in two types: synthetic quartz and
0
200 400 600 800 1000 1200 fused quartz. Synthetic quartz is more frequently used
as the window for CCDs since it contains fewer metallic
Wavelength (nm)
KMPDB0385EA
impurities. Even if there is no AR coating, quartz has high
transmittance which is approx. 94% in the visible region.
(2) Front-illuminated CCD linear image sensor with stable Quartz transmits light down to a wavelength of 200
spectral response in the ultraviolet region nm or shorter and so is suitable as a window material,
especially when detecting ultraviolet light. However, in
The spectral response in the ultraviolet region of front-
the past when quartz was used as a window of a cooled
illuminated CCDs that are designed to be sensitive in the
CCD, consideration had to be given to condensation
ultraviolet region had been inconsistent depending on
that may form within the package due to the moisture
the element. We have developed the S11151-2048 front-
permeability of the adhesive resin. Using new technology,
illuminated CCD linear image sensor that suppresses
we have made it possible to hermetically seal the package
inconsistencies between different elements in the
without using adhesive resin for the quartz window.
spectral response in the ultraviolet region by fabricating
Naturally, CCDs not having a window exhibit the highest
a special structure in the photosensitive area.
quantum efficiency. Windowless CCDs are sometimes used
(3) Lumogen-coated front-illuminated CCD especially in the vacuum ultraviolet region at wavelengths
The photosensitive area of front-illuminated CCDs is shorter than 160 nm because appropriate window materials
covered with poly-silicon electrodes. Ultraviolet light are not available in that region.
is almost totally absorbed by the poly-silicon, so the Other window materials include borosilicate glass
quantum efficiency in the ultraviolet region is nearly which is less expensive than quartz. Borosilicate glass is
zero. To make the front-illuminated CCDs sensitive to mainly used to detect visible and longer wavelength light
ultraviolet light, a scintillator material Lumogen is since its transmittance sharply drops from wavelengths
sometimes used for coating. Lumogen is directly coated around 300 nm. To detect X-rays, aluminum or beryllium
14
is used as the window material which allows X-rays to [Figure 1-32] Photon transfer curve
transmit through but blocks out light (use caution since (S9974-1007, typical example)
beryllium is toxic). (Td=-45 C)
103
Fixed pattern noise
[Figure 1-31] Spectral transmittance of window materials Shot noise
Noise (digital number)

102
Quartz
Transmittance (%)
AR coated sapphire
101
100
100 101 102 103 104 105
Signal (digital number)

KMPDB0210EA
Wavelength (nm)
Saturation charge
KMPDB0110EA
The saturation charge for a CCD indicates the number

Photoresponse nonuniformity of signal electrons that can be transferred by a potential
well. This saturation charge is expressed in units of e-.
Photoresponse nonuniformity specifies the variations The saturation charge for CCDs is determined by the
in sensitivity between pixels of a CCD, and is caused by following four factors:
variations in the light input window and the wafer process.
Vertical shift register saturation charge (vertical full well
Noise accompanying the photoresponse nonuniformity is
capacity)
proportional to the signal level.
Horizontal shift register saturation charge (horizontal
A photon transfer curve [Figure 1-32], which plots the
full well capacity)
relationship between the noise and the input signal level
Summing saturation charge
varying with light exposure, is acquired by illuminating
Output section saturation charge
the effective photosensitive area with uniform light
and setting a measurement area of about 50 50 pixels. In two-dimensional operation mode, the signal charge of
Photoresponse nonuniformity (PRNU) is then defined by each pixel is output individually, so the saturation charge is
equation (6). determined by the vertical shift register. On the other hand,
the horizontal saturation charge is designed to saturate
Noise
PRNU = 100 [%] (6) at a higher level than the vertical saturation charge so as
Signal
to enable line binning. The summing saturation charge
Here, noise is a statistical value indicating the standard formed by the summing gate, which is the last clock gate,
deviation of pixel signals. The signal is the average signal is designed to be greater than the horizontal saturation
of each pixel in the effective photosensitive area. When charge in order to add signals from the horizontal shift
the signal level is low, the PRNU is affected by shot noise register (pixel binning).
(see Noise in section 1-2, Characteristics). However, The saturation voltage (Vsat) of an output signal is given
when the signal level is sufficiently high, the PRNU by equation (7).
becomes a constant value. The PRNU specified in our
Vsat = FW Sv (7)
datasheets was measured at a signal level that is 50% of
the saturation charge. A PRNU of typical FFT-CCDs is FW: full well capacity (saturation charge)
Sv : conversion coefficient
approx. 1% rms or 3% typ. (peak to peak).
Linearity
Linearity of CCD output characteristics deviates slightly

from the ideal line =1. The cause of this deviation is related
to the output stage and results from capacitive variations
in the reverse-biased PN junction constituting the FDA and
from fluctuations of the MOSFET transconductance.
The extent of linearity deviation is expressed in terms of
linearity residual (LR) as defined by equation (8).
15
LR = ( 1 - Sm/Tm ) 100 [%] (8) [Figure 1-34] CTE evaluation method by Fe-55 stacking
S/T
Ideal CCD single event line
Number of pixels (n=1024)
Sm : signal level at one-half the saturation charge
Tm : exposure time at one-half the saturation charge
S : signal Charge transfer loss
T : exposure time
CTE=0.99995
1620 e-, Fe-55

Signal charge
CTE=0.9995
[Figure 1-33] Linearity (S9971-1007, two-dimensional
operation mode, typical example)
(Td=-10 C)
5
4
Leading Trailing
3 KMPDC0046EA
2
Linearity error (%)
1
Interline CCDs experience image lag in the order of several
percent due to the incomplete transferring of signals from
0
the photodiodes to the shift register. On the other hand,
-1
in the case of an FFT CCD in which the shift register itself
-2
receives light, image lag is generated due to the trapping
-3
and discharging of signal charges by the traps (see Damage
-4 by radiation in section 1-3, How to use). As a result, this
-5
101 102 103 104 105 106
image lag is observed as a CTE deterioration. Here CTE
image lag is described briefly using line binning as an
Signal (e-)
example.
KMPDB0211EA
In line binning, signals are acquired for each line
corresponding to the number of horizontal shift registers.
Charge transfer efficiency If the CTE is 1 (the ideal case), the signal charge in the
readout after a line signal is equal to the dark level. But,
Ideally, there is no loss in the charge transfer process if the CTE is less than 1, unread signal charges will be left
of CCDs. In actual operation, however, 100% charge behind as shown in Table 1-1 depending on the number of
transfer is not attained due to traps resulting from the transfers.
semiconductor materials and wafer process. A very small
amount of charge is not transferred and is left behind. [Figure 1-35] CCD image lag in line binning
Charge transfer efficiency (CTE) is defined as the ratio of Signal level
charge that is transferred from one pixel to the adjacent Signal of

one line 1st 2nd
pixel. (In a 2-phase CCD, 2 charge transfers are required
Dark level
in gate units in order to transfer the signal charge Blanking Ideal lag: 0
per pixel, but those two transfers are specified as one KMPDC0047EB
transfer.)
[Table 1-1] Charge transfer efficiency and ratio of image
An X-ray stimulation method is effective in measuring
lag in line binning
the transfer efficiency of a small charge because X-ray
incident on the CCD causes an ideal spot charge to be CTE S9971-0906 S9971-0907
input in a pixel without using electrical means. 0.99995 0.0032 0.0064
In this measurement, the signal of each line in the 0.99999 0.00064 0.00128
horizontal direction is stacked (horizontal stacking). 0.999995 0.00032 0.00064
By means of this horizontal stacking, the CCD output
depicts a single event line according to the X-ray energy
as shown in Figure 1-34. In an ideal CCD with a CTE Dark current
equal to 1, the signal height of the leading and trailing
edges would be the same. In actual use, however, the Dark current is an output current that flows when no
CTE is less than 1, so a loss of the signal charge transfer light is input. This is generally expressed in units of A
occurs at the trailing edge. If we let the signal charge at (ampere), A/cm2, and V (volt). In CCDs for measurement
the leading edge be 1, then the charge transfer loss is applications, e-/pixel/s or e-/pixel/h units are generally
expressed by equation (9). used, which indicate the number of electrons generated
in one pixel per unit time. Dark current nearly doubles
Charge transfer loss = n CTI (9)
for every 5 to 7 C increase in temperature.
n: number of pixels Three major causes that generate CCD dark current are
CTI ( charge transfer inefficiency ) = 1 - CTE
as follows:
The CTE of standard Hamamatsu CCDs is 0.99999.

16
Thermal excitation and diffusion in undepleted region same potential as the substrate, even if a further negative
Thermal excitation in depletion layer voltage is applied.
Thermal excitation by surface level In the state where the oxide film interface is inverted by
holes, the generation of thermally excited electrons is
Among these three causes, item is most dominant.
drastically suppressed. This therefore allows attaining a
MPP (multi-pinned phase) operation reduces the dark
state with low dark current.
current and is also referred to as inverted operation.
MPP operation is performed by setting the portions [Figure 1-38] Dark current vs. gate voltage
under all MOS structure gates, which constitute the CCD (S9974-1007, typical example)
electrodes, to the inverted state. (Td=-10 C)
10000
The dark current can be significantly lowered by MPP
operation since it suppresses the effect of .
1000
Dark current (e-/pixel/s)

[Figure 1-36] Dark current vs. temperature (S9970/S9971 series)
100
10
1
-10 -9 -8 -7 -6 -5 -4 -3 -2 -1 0
Gate voltage (V)

KMPDB0212EA
In MPP operation, the dark current can be greatly reduced

by applying the optimum pinning voltage. However, if
Temperature (C) the voltage does not reach the optimum pinning voltage,
KMPDB0152EB the inverted layer is not fully formed by holes, so the dark
current cannot be minimized. In contrast, if the voltage
In a 2-phase CCD, a potential difference is applied between
is increased as a negative value in excess of the optimum
the barrier phase and the signal charge collecting phase
pinning voltage, not only is extra clock amplitude required,
by way of ion implantation and the like. The 2-phase CCD
but the dark current may also increase due to excessive
therefore provides the potential wells for accumulating
charges called spurious charges (see Spurious charges
charges even when all gates are set to the same voltage.
in section 1-2, Characteristics). Dark current can be
MPP operation can be performed by applying a bias to
minimized by adjusting the voltage to an optimum value
invert all phases of the CCD.
near the gate voltage listed in the datasheet.
When the dark current must be reduced, using the MPP
operation and cooling the CCD are very effective.
Noise
[Figure 1-37] Potential distribution in MPP operation
Oxide film
CCD noise is classified into the following four factors:
(1) Fixed pattern noise (Nf)

Collecting phase
This noise is caused by variations in sensitivity between
Barrier phase
Potential
CCD pixels (variations in sensitivity between pixels is

caused by nonuniformities in the aperture area and film
thickness). When the signal is large, the fixed pattern
noise is proportional to the amount of exposure (number
Distance from surface in depth direction of signal electrons). The fixed pattern noise Nf can be
regarded as zero based on the noise from one pixel.
Pinning voltage
(example: -8 V)
(2) Shot noise (Ns)
KMPDB0055EB
Shot noise is the noise generated by statistical changes
in the number of photons incident on a CCD. Shot noise
As shown in Figure 1-37, during MPP operation, both is expressed by equation (10) according to the Poisson
the collecting phase and barrier phase are pinned in the distribution.
inverted state. In the pinned state, the CCD surface is
inverted by holes supplied from the channel stop region. Ns = S (10)
The potential at the oxide film interface is fixed at the S: number of signal electrons [e - ]
17
For example, if a CCD receives photons that generate a The CCD dynamic range is a value obtained by dividing
signal electron quantity of 10000 e- inside the CCD, then the saturation charge by the readout noise.
the shot noise will be 100 e- rms.
Saturation charge
Dynamic range = (12)
(3) Dark shot noise (Nd) Readout noise
Dark shot noise is caused by dark current and is proportional

Dynamic range is also given by equation (13).
to the square root of the number of electrons generated
in a dark state. To reduce the dark shot noise, the dark
current itself must be reduced. The variation in dark
Dynamic range = 20 log ( Saturation charge
Readout noise )
[dB] (13)
current between each pixel is larger than variations in the

sensitivity. The dynamic range varies with operating conditions
such as operating temperature and integration time.
(4) Readout noise (Nr)
At around room temperature, the dark shot noise
This is electrical noise from thermal noise caused by determines the lower detection limit. Under operating
the MOSFET used as the amplifier in the CCD output conditions where the dark shot noise can be ignored
section. It also comes from the readout circuit and (by cooling the CCD sufficiently), the readout noise
eventually determines the lower detection limit of the determines the dynamic range.
CCD. Readout noise is determined by the CCD output In two-dimensional operation, the saturation charge will
method and is not affected by the amount of exposure. equal the charge that can be transferred by the vertical
Readout noise is also frequency dependent [Figure 1-43]. shift register. In line binning, the saturation charge
will equal the charge that can be transferred by the
Total noise (Nt) is expressed by equation (11). horizontal shift register.
Nt = Nf 2 + Ns 2 + Nd 2 + Nr 2 (11) [Table 1-2] CCD specification examples
Parameter S9736 series S7170-0909

Figure 1-39 shows the interrelation between these four
factors and the amount of exposure. The CCD detection Type Front-illuminated type Back-thinned type
limit is determined by the dark shot noise and readout Number of pixels 512 512
noise. This means that the CCD detection limit can be Pixel size [m] 24
lowered to the readout noise level by reducing the dark Saturation charge (vertical) [ke-] 300 320
current and lowering the dark shot noise below the Conversion factor [V/e-] 3.5 2.2
readout noise. Readout noise [e- rms] 4 8
The S/N is mainly determined by the fixed pattern Dynamic range 75000 40000
noise when the amount of exposure is high, and it is
Dark current (0 C) [e-/pixel/s] 10
determined by the shot noise when the amount of
exposure is low.
[Figure 1-39] Noise vs. exposure

Resolution
Signal (slope 1)
The ability of an image sensor to reproduce the contrast
at a spatial frequency in an image is called the spatial
resolution and is quantified by the MTF (modulation
Fixed pattern noise
(slope 1) transfer function) for sine waves. Since the pixels of a CCD
Noise, signal
are individually separated, there is a limiting resolution

Shot noise determined by the Nyquist limit due to the discrete
(slope 1/2)
sampling theorem. For example, when a black-and-
white stripe pattern is viewed with a CCD, the difference
Dark shot noise between the black and white signal levels decreases as the
(slope 0)
Readout noise
stripe pattern becomes finer, and finally reaches a point
(slope 0) at which the stripe pattern cannot be resolved. The ideal
MTF of CCDs is expressed by equation (14).
Exposure
KMPDB0057EB
MTF = sinc {( f)/(2 fn)} (14)
f : spatial frequency of image

fn: spatial Nyquist frequency
Dynamic range
Dynamic range generally specifies the measurable range Because optical sine waves are difficult to generate, a test
of a detector and is defined as the ratio of the maximum chart having square wave patterns is commonly used.
level to the minimum level (detection limits). The spatial frequency response measured using this test
chart is called the contrast transfer function (CTF) which
18
is different from the MTF. (The CTF can be converted [Figure 1-42] Standard deviation of charge diffusion
into the MTF by Fourier transform.) vs. bias voltage
The actual CCD resolution is determined by the extent of (Silicon thickness 200 m, =450 nm)
16
diffusion occurring when the signal charge is collected
Standard deviation of charge diffusion (m)

inside the silicon. Since the incident photons are absorbed
14
within the depletion layer, the generated electrons do not
diffuse and is collected by the corresponding pixels, so the
12
resolution does not deteriorate. The resolution also varies
depending on the depth in the silicon where the incident
photons are absorbed. The longer the incident photon 10
wavelength, the deeper the position where the photons

are absorbed, causing the resolution to deteriorate. 8
[Figure 1-40] MTF vs. spatial frequency at different

6
wavelengths of input photons 0 10 20 30 40 50 60
(S9970/S9971 series, calculated values)
Bias voltage (V)
KMPDB0281EB
Frequency characteristics of noise
When dark current and spurious charges are sufficiently

MTF
small, the readout noise determines the eventual number

of noise electrons generated in a CCD. The readout noise is
determined by the thermal noise of the MOSFET comprising
the FDA in the readout section. MOSFET thermal noise
includes white noise and 1/f noise, both of which should be
reduced to achieve low noise. White noise can be reduced
by increasing the MOSFET mutual conductance (gm). In
Spatial frequency (f/fn) MOSFET built into CCDs for measurement applications, the
KMPDB0206EA
corner frequency of 1/f noise is reduced to as low as a few
[Figure 1-41] CTF calculation method kilohertz.
MOSFET thermal noise depends greatly on the bias
CTF = (Max. - Min.)/(Max. + Min.)
White level conditions. To achieve the readout noise specified in our
Max.
CCD image sensor datasheet, the bias must be applied
according to the recommended operating conditions. Even
Signal when the recommended bias conditions have been set,
the signal processing circuit still has a great effect on CCD
Min. readout noise. Since a CDS circuit is commonly used for
Black level
CCD signal processing, optimizing the transfer functions
KMPDC0048EA for the CDS circuit and the LPF (low-pass filter) installed in
the preceding stage of the CDS circuit will result in reduced
CCD readout noise. If the effect of the 1/f noise corner
Point spread function
frequency can be reduced versus CCD readout frequency,
then the output noise of the CCD system including the
The standard deviation ( D) of charge diffusion in the
signal processing circuit will be determined by the white
depletion layer is defined as shown in equation (15). Here,
noise and noise bandwidth.
D is proportional to the square root of the depletion layer
To summarize the above, CCD readout noise depends
thickness and also of the absolute temperature in the
on the readout frequency. The readout frequency must
silicon, and is inversely proportional to the square root of
be low (less than 100 kHz) to achieve a readout noise of
the bias voltage applied to the backside.
a few electrons (e- rms) which is a noise level required in
2 Xdep 2 k T measurement applications. If the signal readout frequency
D = (15)
Vbb q becomes higher, the readout noise increases sharply.
Xdep: thickness of depletion layer

k : Boltzmanns constant
T : absolute temperature
Vbb : bias voltage applied to backside
q : electron charge
19
[Figure 1-43] Readout noise vs. readout frequency [Figure 1-44] Spurious charge vs. temperature
(S9737-01, typical example)
Readout noise [e- rms] (typical example)
Spurious charge (e-)

Clock rising edge
Readout frequency (kHz) Temperature (C)

KMPDB0207EA KMPDB0208EA
Spurious charge Anti-blooming
Spurious charges are generated by clock pulses during Blooming (overflow) is a phenomenon that occurs when
operation such as in MPP mode and do not result from high-intensity light enters the photosensitive area and
signals produced by the incident light. In MPP operation, the resulting signal charge exceeds a specific level. This
the vertical clock pulse is set to low, and during this excess charge then overflows into adjacent pixels and
low period, the region under the gate of each pixel is transfer region. A technique to prevent this is called
in an inverted state. In this state, holes move from the anti-blooming which provides a drain to carry away the
channel stop region to a point under the gate, and the excess charge.
surface potential in that region is pinned at the substrate Anti-blooming structures for CCDs are roughly divided
potential. At this point, some holes are trapped along into a lateral type and a vertical type, and our CCDs
the oxide film interface, and the gate phase of each pixel use the lateral type. The lateral type structure has an
becomes a non-inverted state when the clock pulse goes overflow drain formed along the pixels or charge transfer
to high level. The trapped holes have high energy after channels. This structure has the drawback that the
being released and generate a spurious charge which is fill factor is reduced when used for front-illuminated
then collected in a potential well. The CCD output is the CCDs. However, this problem can be avoided when used
sum of the signal, dark current, and this spurious charge. for back-thinned CCDs. The vertical type structure is
Spurious charges can be reduced by delaying the rising designed so as to carry away the excess charge into the
edge of clock pulses or decreasing the voltage difference inside of the substrate. The fill factor is not reduced, but
between high and low clock levels. When a CCD is cooled there is a problem in that the sensitivity drops at longer
to a sufficiently low temperature where the signal level wavelengths.
approaches readout noise level, it is important to set the When controlling the anti-blooming function by means
clocking conditions by taking the spurious charge into of the overflow drain voltage (VOFD) and overflow gate
account. voltage (VOFG), these applied voltages may decrease the
saturation charge.
[Figure 1-45] Schematic of anti-blooming
Blooming
occurs.
Vertical
low level
Current flows.
Vertical
Potential high level
KMPDC0285EA
20
[Figure 1-46] Anti-blooming structure and potential [Figure 1-48] Saturation output voltage vs. anti-blooming
(lateral type) gate voltage (typical example)
1 pixel
Poly-Si
SiO2
Saturation output voltage (V)

N- channel
Charge drift
Barrier
(clock: low)
Storage
(clock: high)
Anti-blooming gate voltage (V)

Potential KMPDB0282EA
KMPDC0286EA
[Figure 1-47] Imaging examples Electronic shutter

(a) Without anti-blooming (b) With anti-blooming
In general, the integration time for CCD linear image
sensors is equivalent to the interval between two clock
pulses to the transfer gate. Using an electronic shutter
function allows setting an effective integration time that
is shorter than the transfer gate clock pulse interval.
When the anti-blooming gate voltage is set high, all signal
charges generated in the photodiodes are carried away
into the anti-blooming drain. It is therefore possible to
set an effective integration time shorter than the normal
integration time by providing a period during which the
Anti-blooming function of CCD linear image sensors anti-blooming gate voltage is high and a period during
which the anti-blooming gate voltage is low. In addition,
In CCD linear image sensors, an anti-blooming drain
the start timing of the integration time can be synchronized
and anti-blooming gate are formed in the vicinity of
to an external trigger pulse.
the storage gate. The anti-blooming function works by
applying appropriate voltage to the anti-blooming drain [Figure 1-49] Timing chart for the CCD linear image
and anti-blooming gate. The anti-blooming gate voltage sensor (electronic shutter function)
controls the saturation output voltage. Moreover, when Effective
integration time
the anti-blooming gate voltage is set high, all signal
Readout cycle (integration time)
charges generated in the photodiodes can be extracted
into the anti-blooming drain to make the output zero. Anti-blooming gate
This function is used to activate the electronic shutter Transfer gate
described next. Horizontal shift register*
* Indicates readout period

KMPDC0287EA
21
[Figure 1-50] Output voltage vs. effective integration time
(CCD linear image sensor, typical example)
Cosmetics
The term cosmetics refers to the extent of CCD defects

(blemishes and scratches). Blemishes and scratches are
divided into two categories: white spots that appear
bright in a dark state and black spots that appear dark
Output voltage (mV)
when light is incident on a CCD.

White spots are usually caused by lattice defects or
metal impurities in the substrate material or by pattern
failures from mechanical damage or dust during the
wafer process. Black spots are mainly caused by irregular
reflections due to dust on the CCD surface during the
wafer process or partial defects of the surface insulation
film, or by contamination such as dust on the device
surface or window. It is difficult to completely eliminate
Effective integration time (s)
these white and black spots. The larger the photosensitive
KMPDB0283EA
area and the smaller the pixel, the more obvious the
effects from these white and black spots become.
Etaloning Hamamatsu defines white and black spot specifications
and inspects every CCD to check the number of these
Etaloning is an interference phenomenon that occurs spots.
while the light incident on a CCD repeatedly undergoes Cosmetic specifications are defined as described below.
reflection and attenuation between the front and back These definitions may vary depending on the manufacturer,
surfaces of the CCD, and causes alternately high and so use caution when comparing specifications.
low sensitivity. An etalon commonly refers to an optical
(1) Point defect
element consisting of two parallel planes with highly
reflective films facing each other. The light entering a CCD White spot
repeatedly undergoes imperfect reflection, transmission,
White spots are pixels that generate dark current in excess
and absorption within the CCD as if the light has entered
of 3% of the saturation charge after charge integration for
an etalon. This phenomenon is therefore called etaloning.
1 second at a cooling temperature of 0 C.
When long-wavelength light enters a back-thinned CCD,
etaloning occurs due to the relationship between the silicon Black spot
substrate thickness and the absorption length [Figure Black spots are pixels that provide an output of less than
1-51]. Using our advanced device technology, we have 50% of the average output value calculated when the
succeeded in producing back-thinned CCDs that offer CCD is illuminated with uniform light so as to generate
reduced etaloning. Note that etaloning is a phenomenon a charge equal to 50 to 90% of the saturation charge.
unique to back-thinned CCDs and does not occur in We usually perform this test using uniform light that
front-illuminated CCDs. generates a charge equal to 50% of the saturation charge.
[Figure 1-51] Etaloning characteristics (typical example) (2) Cluster defect

(Ta=25 C) A cluster consisting of two to nine continuous pixel defects
110
100 is called a cluster defect and is distinguished from point

90 Etaloning-improved type defects. Cluster defects appear vertically in most cases, but
appear as a two-dimensional cluster if originating from
Relative sensitivity (%)
80
70 black spots of back-thinned CCDs or front-illuminated

Previous product
60 CCDs coupled to an FOS (fiber optic plate with scintillator).
50
(3) Column defect
40
30
A cluster consisting of ten or more continuous pixel
20
defects (larger than a cluster defect) is called a column
10
defect and is viewed as different from cluster defects. As
0 with cluster defects, column defects also appear vertically
900 920 940 960 980 1000
in most cases, but may appear as a two-dimensional
Wavelength (nm) cluster if originating from black spots of back-thinned
KMPDB0284EB
CCDs or front-illuminated CCDs coupled to an FOS.
22
Front-illuminated CCDs with a small photosensitive [Figure 1-53] Cross section of CCD (back-illuminated)
area, such as Hamamatsu S9970/S9971 series, have no that has received cosmic rays
point defects, cluster defects, or column defects. When (a) Standard type
CCDs are coupled to an FOP (fiber optic plate) or FOS,
1 pixel
CCD side
defects might occur due to other factors not originating
in the CCDs, so the shape and number of defects will
- - - -
differ from those occurring only in CCDs. - - - -
Photosensitive
side
Effects of cosmic rays on CCDs
Cosmic rays Cosmic rays
If cosmic rays enter the CCD, these rays may be detected
KMPDC0412EA
as shown in Figure 1-52. A certain ratio of cosmic rays
reaches the Earth's surface. The main portion of such (b) Fully-depleted type
rays consists of particles (up to several GeVs), and they
generate signal charges inside the CCD silicon along CCD side 1 pixel
their tracks. With the standard type back-thinned CCD,

the silicon is thin, and therefore false signals due to - -
cosmic rays will appear in only several pixels at most, -
-
but with the fully-depleted back-illuminated CCD, false - -
-
-
signals may appear in many pixels.
-
-
The frequency at which cosmic rays are detected
- -
depends on the sensor structure or environment, but a -
- -
rough estimate is 150 counts/(cm2h). It is known that
-
the amount of cosmic rays increases as you move further -
-
away from the Earth's surface. - -
-
If false signals occur temporarily due to cosmic rays,
their effects can be reduced by acquiring multiple images Cosmic rays Photosensitive side
KMPDC0413EA
and averaging them or by reducing the integration time.
By interacting with silicon atoms, cosmic rays may
cause lattice defects, and this phenomenon leads to
white spots and charge traps. It is recommended that
1-3 How to use
a correction function be provided in the device at the
outset to deal with white spots.
Timing
[Figure 1-52] Image example of false signals caused by cosmic
rays on fully-depleted back-illuminated CCD Operating a CCD requires seven types of signals: 2-phase
512 pixels clock pulses (P1V, P2V ) for the vertical shift register, a
(Td=-100 C, integration time 900 s) transfer gate pulse (TG), 2-phase clock pulses (P1H, P2H)
for the horizontal shift register, a summing gate pulse
(SG), and a reset pulse (RG). The TG electrode utilizes a
part of the last P2V electrode but is recommended to be
used as a separate terminal where the clock pulse should
be input at the same timing as P2V. However, operation
is also possible by shorting the TG and P2V terminals. To
512 pixels
find timing charts of pulses needed to operate a CCD,

refer to our datasheet.
FFT-CCDs can be operated in any of four modes: line
binning, two-dimensional operation, pixel binning, or TDI
operation. The desired operation mode can be chosen by
simply adjusting the timing of each clock pulse.
(1) Line binning

The number of bits that should be binned is first
transferred in the vertical direction. This permits signal
charges to be added to the corresponding horizontal shift
register. Then all horizontal signal charges are transferred.
The summing gate pulse should be exactly the same as
the clock pulse (P2H) for the horizontal shift register.
23
[Figure 1-54] Timing chart of line binning
Integration period (shutter open) Vertical binning period (shutter closed) Readout period (shutter closed)
Tpwv
P1V
Tovr
P2V, TG
Tpwh,Tpws
P1H
P2H
SG
Tpwr
RG
OS
KMPDC0050EB
Parameter Symbol Min. Typ. Max. Unit

S703*-0906 1.5 2 -
Pulse
S703*-0907/-1006 Tpwv 3 4 - s
P1V, P2V, TG*1 width
S703*-1007 6 8 -
Rise and fall times Tprv, Tpfv 10 - - ns
Pulse width Tpwh 500 2000 - ns
P1H, P2H*1 Rise and fall times Tprh, Tpfh 10 - - ns
Duty ratio - - 50 - %
Pulse width Tpws 500 2000 - ns
SG Rise and fall times Tprs, Tpfs 10 - - ns
Pulse width Tpwr 100 - - ns
RG
Rise and fall times Tprr, Tpfr 5 - - ns
TGP1H Overlap time Tovr 3 - - s
*1: Symmetrical clock pulses should be overlapped at 50% of maximum pulse amplitude.
[Figure 1-55] Timing chart of two-dimensional operation

(a) Low dark current mode
Integration period (shutter open) Readout period (shutter closed)
Tpwv
P1V
P2V, TG
P1H
P2H
SG
RG
OS
Detail
Tovr
P2V, TG Tpwh, Tpws
P1H
P2H
SG
Tpwr
RG
OS
KMPDC0049EB
24
(b) Large saturation charge mode
Tpwv
P1V
P2V, TG
P1H
P2H
SG
RG
OS
Detail
Tovr
P2V, TG Tpwh, Tpws
P1H
P2H
SG
Tpwr
RG
OS
KMPDC00145EA
Parameter Sensor Symbol Min. Typ. Max. Unit

Pulse width *2 S7171-0909-01 Tpwv 6 8 - s
P1V, P2V, TG
Pulse width*2 Tpwh 500 2000 - ns
P1H, P2H Rise and fall times Tprh, Tpfh 10 - - ns
Duty ratio - 40 50 60 %
All series
Duty ratio - 40 50 60 %
RG
TG-P1H Overlap time Tovr 3 - - s
[Figure 1-56] Timing chart of pixel binning (2 2)

(a) Low dark current mode
Tpwv
P1V
P2V, TG
P1H
P2H
SG
RG
OS
Detail
Tovr
P2V, TG Tpwh, Tpws
P1H
P2H
SG
Tpwr
RG
OS
KMPDC0051EB
25
(b) Large saturation charge mode
Tpwv
P1V
P2V, TG
P1H
P2H
SG
RG
OS
Detail
Tovr
P2V, TG Tpwh, Tpws
P1H
P2H
SG
Tpwr
RG
OS
KMPDC0146EA

S703*-0906 1.5 2 -
Pulse
S703*-0907/-1006 Tpwv 3 4 - s
P1V, P2V, TG*1 width
S703*-1007 6 8 -
Pulse width Tpwh 500 2000 - ns
P1H, P2H*1 Rise and fall times Tprh, Tpfh 10 - - ns
RG
TGP1H Overlap time Tovr 3 - - s
26
[Figure 1-57] Timing chart of TDI operation
(a) 1 1
Tpwv
P1V
P2V, TG
P1H
P2H
SG
RG
OS
Tovr Detail
P2V, TG Tpwh, Tpws
P1H
P2H
SG
Tpwr
RG
OS
KMPDC0147EA
Note: For the timing chart in low dark current mode, see P1V, P2V, and TG in Figure 1-55 (a).
(b) 2 2, pixel binning

Tpwv
P1V
P2V, TG
P1H
P2H
SG
RG
OS
Detail
Tovr
P2V, TG Tpwh, Tpws
P1H
P2H
SG
Tpwr
RG
OS
KMPDC0148EA
Note: For the timing chart in low dark current mode, see P1V, P2V, and TG in Figure 1-56 (a).
Parameter Sensor Symbol Min. Typ. Max. Unit

P1AV, P1BV Pulse width tpwv 30 - - s
P2AV, P2BV, TG*2 *3 Rise and fall times tprv, tpfv 200 - - ns
Pulse width tpwh 125 - - ns
P1AH, P1BH
Rise and fall times tprh, tpfh 10 - - ns
P2AH, P2BH*3
Pulse width S7199-01 tpws 125 - - ns
SG Rise and fall times tprs, tpfs 10 - - ns
Pulse width tpwr 10 - - ns
RG
Rise and fall times tprr, tpfr 5 - - ns
TG-P1AH, P1BH Overlap time tovr 10 - - s
*2: Apply the same pulse as P2AV to TG.
27
(2) Two-dimensional operation (area scan) CCD saturation charge. If the low level voltage of the
reset clock pulses becomes high, the charge that can be
This operation transfers all horizontal signal charges each
stored in the FD decreases because the potential has
time one bit is transferred in the vertical direction. When
not lowered sufficiently in a state where the reset switch
the transfer in the vertical direction is fully complete, a
is off. This may cause an overflow before all transfer
frame transfer is complete. At this point, the summing
charges are converted into voltage. For this reason, the
gate pulse should be set exactly as the clock pulse (P2H)
low level of the reset clock pulses must be set to a voltage
for the horizontal shift register.
low enough not to affect the saturation charge level of
(3) Pixel binning the output section.
One bit is first transferred in the vertical direction. Then Set the pulse width of the reset clock pulses to about 10
all the horizontal signal charges are transferred. At ns to 100 ns (there will be no problem if longer than 100
this point, by halting the summing gate pulses for the ns).
number of bits required for summing, the signal charges (3) Transfer clock pulse generator
are added to the summing well.
Figure 1-58 shows an example of a transfer clock pulse
Note: Line binning and pixel binning can be performed at the same time.
generator. As stated above, clock pulses with high and
low levels of voltage amplitude are required to operate
(4) TDI operation a CCD. These clock pulses must drive the vertical shift
As explained in TDI-CCD in section 1-1, Structure and register and horizontal shift register at high speeds,
operating principle, TDI operation allows imaging of a which have an input capacitance of several hundred
moving object. To do this, the CCD vertical transfer clock picofarads to several nanofarads. For this purpose, MOS
pulse must be synchronized with the speed at which the driver IC is commonly used to drive a CCD since it is
object moves along the photosensitive area surface of the capable of driving a capacitive load at high speeds.
CCD. Normally, the timing signal generator circuit uses a TTL
or CMOS logic level IC. The operating voltage for these
Clock pulse and DC bias adjustment ICs is +3.3 V or +5.0 V, so a level converter circuit must be
connected to the MOS driver IC.
The clock pulse and DC bias must be adjusted properly In 2-phase CCD operation, the clock pulses for driving
to make fullest use of CCD performance. the vertical and horizontal shift registers must overlap
with each other (see Charge transfer operation in
(1) Transfer clock pulse section 1-1, Structure and operating principle). For this
The low level of the clock voltage for the vertical shift reason, a resistor Rd with an appropriate value (damping
register affects the CCD dark current. If it is set to a resistor: a few to several dozen ohms) should be placed
voltage higher than the pinning voltage that initiates between the MOS driver IC and the CCD in order to
MPP operation, the dark current will not lower as adjust the rise time and fall time of the clock pulses.
expected. The pinning voltage differs according to the To minimize noise intrusion to the CCD from digital circuits,
individual CCD due to variations in device production. it is recommended that the analog ground and digital
Ideally, it should be adjusted for each product. ground be set to the same potential by the transfer clock
After determining the low level of the vertical clock pulse generator.
voltage, adjust the high level. The clock pulse amplitude
[Figure 1-58] Example of transfer clock pulse generator
should be large enough to maintain the desired is set too
VDD
large, the spurious charges become large, causing the +Vcc
dark current (Nb) during the readout time to increase 10 F

100 pF
6 20 0.1 F
and resulting in an offset that appears in the entire 5 4 18 2
P1V 17
Y1 A1
3
P1V_in
Rd 2.2 k 16
Y2
Y3
A2
A3
4 P2V_in
output signal. Normally, the spurious charges cannot EL7212 15

Y4 A4
5
7 2 2.2 k 14
13
Y5 A5
6
7
P2V 12
Y6
Y7
A6
A7
8
be distinguished from the dark current at around room Rd 11

Y8 A8
9
3 G1
1
19
temperature, but they may cause problems when the CCD 100 pF
74HCT540
G2
10
is cooled. Therefore, the vertical clock pulse amplitude 2.2 k

2.2 k
should be adjusted to a minimum as long as other
characteristics are not impaired. 10 F
VEE
(2) Reset clock pulse

VDD: high level voltage of clock pulse
VEE: low level voltage of clock pulse
The reset clock pulse is applied to the reset gate (RG) Rd: damping resistor (a few to several tens of ohms)
to periodically reset the signal charge flowing into the KMPDC0052EC
FD (floating diffusion) to the reference voltage (V RD).

Adjusting the low and high levels of this clock pulse
changes the saturation charge level of the output section.
When these are properly adjusted, the saturation charge
level of the output section is sufficiently larger than the
28
[Figure 1-59] Potentials in CCD output section level, and the difference between the OG potential and
Charge amplifier the potential at reset becomes the factor that determines
SG OG RG RD
the amount of signal charge that can be handled. As seen
from Figure 1-59, the amount of signal charge is limited by
VOG VRD
either the potential under OG or the potential of the reset
gate at low level. As VOG decreases, the saturation charge
SG Low
FD RD level of the output section increases. If VOG is too low,
RG Low however, the signal charge is unable to flow into the FD
SG Low
during the low level of SG, so VOG must be adjusted to an
OG
SG High appropriate value.
SG High RG High VRD Bias voltage generator circuit

The bias voltage is mainly applied to the peripheral section
KMPDC0183EA of the CCD output amplifier, so use a stable power supply
with relatively low noise. It is also important to note the
(4) DC bias
voltage accuracy, voltage fluctuation, ripple, and output
Bias (VOD) applied to OD current.
Figure 1-60 shows an example of a bias voltage generator
VOD is the bias voltage applied to the output transistor.
circuit for the OD terminal. The reference voltage is
When using a one-stage source follower output amplifier,
generated from the power supply IC and is set to a specified
apply approx. 20 V as VOD. In a source follower circuit
voltage value by the amplifier making up the low-pass filter.
using a load resistance of about 20 k connected to
This allows obtaining a highly stable and accurate voltage
the MOSFET source, the source DC level is about 15 V.
with low noise.
Therefore, although several volts are applied across the
source and drain of the MOSFET via the voltage applied [Figure 1-60] Example of bias voltage generator circuit
to the OD, the following phenomena will occur unless
this voltage is sufficiently high. 51 k 51 k
0.1 F
Voltage gain (Av) in the source follower circuit is lowered.
+24 V +24 V
MOSFET does not operate in the saturation region.
1 F 10 F
VOD=+20 V
These phenomena adversely affect CCD performance. -
VOD Vref=+10 V +VI
For example, they may cause a decrease in the conversion 100 + VO
10 F TRIM N.R.
factor (unit: V/e-), an increase in the readout noise, or 1 F
GND 1 F
deterioration in the linearity.
When using a multi-stage amplifier such as a two-stage
source follower output type, set VOD to approx. +15 V,
KMPDC0303EC
which is lower than VOD for a one-stage type. As in the
case of one-stage type, phenomena and occur in the
two-stage source follower output type. Signal processing circuit
Bias (VRD) applied to RD Major sources of noise from a CCD are the well-known
VRD is the bias voltage applied to the reset drain. It determines kT/C noise and 1/f noise. The kT/C noise is generated
the reset level of the output section and also serves as the by a discharge (reset operation) in the FDA (see FDA
gate voltage of the output transistor. The VRD determines in section 1-1, Structure and operating principle). This
the voltage gain and MOSFET operating region the same as noise is inversely proportional to the square root of the
with VOD and also affects the saturation charge level of the node capacitance (Cfd) of the FDA and makes up a large
output section. Increasing the VRD also raises the potential percentage of the total noise of a CCD. The 1/f noise is
in the FD and the amount of signal increases. However, generated by the MOSFET constituting the FDA and is
it must be set to an optimum value in consideration of inversely proportional to the frequency.
phenomena and which may occur in the output These noises degrade the S/N in the CCD system and
transistor. therefore should be reduced as much as possible in the
signal processing circuit. A typical circuit for this purpose
Bias (VOG) applied to OG
is a CDS circuit.
VOG is the bias voltage applied to OG that separates the FD The operating principle of the CDS circuit is described
arranged at the last stage of the horizontal shift register below. Figure 1-61 shows an output waveform from a CCD.
from the last clocking gate (summing gate). The signal As stated above, kT/C noise occurs during a reset period
charge is output to the FD in synchronization with the in the FDA. At the point where the reset period has ended,
falling edge to the low level of the summing gate pulse the voltage level varies due to kT/C noise. Therefore, if
(SG) which is the last clocking gate. The OG potential data is acquired at time T2, the S/N deteriorates by an
therefore becomes smaller than the SG potential at low amount equal to the kT/C noise variation. In contrast,
29
acquiring data at times T1 and T2 on the output waveform A circuit example of Type 1 is shown in Figure 1-64.
and then obtaining the difference between them will The preamp gain should be set high in order to sufficiently
extract only a signal component V with the kT/C noise amplify the CCD output signal. Since the CCD output
removed. DC components such as the offset voltage signal contains DC voltage components, a capacitor is
component and reset feed-through are removed at the used for AC coupling. Note that this capacitor can cause
same time. a DC voltage error if the preamp bias current is large.
Therefore, a preamp with a small bias current must be
[Figure 1-61] CCD output waveform
selected. A JFET or CMOS input amplifier is generally
Signal output
Reset period period used. It is also necessary to select a low-noise amplifier
with a bandwidth wide enough to amplify the CCD output
kT/C noise Reset level
waveform.
V
The clamp circuit is made up of capacitors and an analog
Signal output level
switch. For the analog switch, we recommend using a
high-speed type having low ON resistance and small
T1 T2
KMPDC0304EA charge injection amount.
For the preamp, the last-stage amplifier is AC-coupled via
There are two types of CDS circuits: Type 1 that uses a capacitor, so a JFET or CMOS input amplifier should be
a clamp circuit in combination with a sample-and- selected. In addition, a non-inverted amplifier must be
hold amplifier (SHA), and Type 2 that uses a SHA in configured to allow high input impedance.
combination with a differential amplifier. Type 1 has a Incidentally the CCD provides a negative-going output
very simple circuit configuration [Figure 1-62]. But if the while the last-stage amplifier gives a positive-going
ON resistance of the switch used in the clamp circuit is output to facilitate analog-to-digital conversion. For
large, the amount of noise that can be removed will be this reason, an inverted amplifier is connected after the
small or a DC voltage error will occur. Ideally, the ON preamp.
resistance should be 0 .
[Figure 1-62] CDS circuit block diagram High-speed signal processing circuit
(using clamp circuit and SHA)
Sample signal (T2) For a CCD signal processing circuit that requires high-
Preamp (LPF) Buffer amp
CCD Video speed readout at several megahertz or faster, it is difficult
output output
for a circuit constructed only of discrete components to
Clamp signal (T1) achieve high-speed clamp operation and fast capacitor
charging/discharging response.
Clamp circuit Sample-and-hold amp (SHA) A high-speed signal processing circuit can be constructed
by using an analog front-end IC (a single IC chip consisting
of CDS, gain, and offset circuits, A/D converter, etc.)
CCD output
optimized for CCD signal processing.
Clamp signal (T1)
Sample signal (T2)

Measures against light emission on the output circuit
Video output
If the operating conditions are not suitable for the CCD
KMPDC0305EB
output circuit employing a two-stage MOSFET source
follower, the amplifier may emit light. If this light is
Type 2 [Figure 1-63] uses a larger number of components
received by the resistive gate, storage gate, or horizontal
but removes noise more effectively than Type 1. However,
shift register, the output for the first pixel that is read out
since Type 2 makes an analog calculation of the SHA
will be large even in a dark state [Figure 1-67].
output, the noise of the SHA itself may be added, resulting
To reduce this effect, the following measure is effective.
in increased noise in some cases. The SHA noise should be
small enough so that the kT/C noise can be ignored. Apply +1 V typ. to the Vret terminal (if a Vret terminal is
available).
[Figure 1-63] CDS circuit block diagram
(using SHA and differential amplifier) Intersect the horizontal shift register clock pulse (P1H,
Sample-and- P2H) at the amplitude of 50% 10% [Figure 1-68].
Preamp (LPF) hold amp (SHA)
Horizontal 2-phase drive: P1H, P2H
Differential amp
CCD output
Horizontal 4-phase drive: P1H, P3H and P2H, P4H
Video output
After reading out all pixels, perform horizontal dummy

readout up to immediately before TG is set to high level.
Sample signal 1 (T1) Sample signal 2 (T2)
KMPDC0306EB
30
[Figure 1-64] CDS circuit example (using clamp circuit and SHA)
33 pF 47 pF 47 pF
1k 1.5 k 1k 1k
-15 V 1k -15 V
-15 V
1 F 1 F
+20 V 1 F UI
24 IG1V RG
1
2 - - -
23
22
IG2V
ISV
RD
OS
3
+ + Video Out
21 NC OD
4
100 1k + 100 1 nF
20
19
SS
NC
OG
SG
5
6 0.1 F
18
17
NC
NC
Th1
Th2
7
8 22 k 100 k
16 9
15 TG P2H 10 1 F 1 F 1 F
14 P1V
P2V
P1H
IG2H
11 -15 V
13
ISH IG1H
12 +15 V +15 V +15 V
S9970-1007 1 F
V+ IN
COM NO
Clamp
V- GND
1 F
+15 V
KMPDC0053ED
[Figure 1-65] High-speed signal processing circuit example (using S11155/S11156-2048-01 and analog front-end IC)
To FPGA +5 V +3.3 V
10
10 0.1 F
10
15 SDATA D0 (LSB) 14 10
+VOD 16 13 10
SCLK D1
10 F 17 SLOAD D2 12 10
0.1 F 18 AVDD D3 11 10
0.1 F 19 10 10
0.1 F AVSS D4
20 CAPB D5 9 10
10 F 0.1 F 21 8 10
CAPT D6
22 VINB D7 (MSB) 7 10
0.1 F 0.1 F 23 6
CML DR VSS
0.1 F 24 5
0.1 F VING DR VDD
0.1 F 25 4 330
1 OFFSET OEB
26 3 330
VINR ADCCLK
24 RG OS 1 27 2 330
AVSS CDSCLK2
23 TG OD 2 28 1 330
AVDD CDSCLK1
22 NC OG 3
21 NC SG 4 0.1 F AD9826KRS To FPGA
20 STG Vret 5 2.2 k
19 NC RD 6 0.1
18 RD REGL 7
17 SS REGH 8
16 NC P2H 9
15 ISH P1H 10 330
14 ARD IG2H 11
13 ARG IG1H 12
S11155/S11156-2048-01
KMPDC0457EB
[Figure 1-66] Timing chart (S11155-2048-01)

Output period of one line
Tpwar Tinteg
(Electronic shutter: closed) (Electronic shutter: open)
ARG
REGH, REGL (REGH=-4 V, REGL=-6.5 V)
Tpwv Tovr
TG
Tpwh, Tpws
P1H 1 2 3..2067 2068 2069 2070... N*
P2H
SG
Tpwr
RG
OS
D1 D2 D19 D20
D3..D10, S1...S2048, D11..D18
Normal readout period Dummy readout period
* Apply clock pulses to appropriate terminals during dummy readout period.

Set the total number N of clock pulses according to the integration time.
KMPDC0382EC
31
[Figure 1-67] Effect of light emission on the output circuit [Figure 1-69] Element temperature vs. operation time
(horizontal profile in a dark state, typical example) (S11155-2048-01, using our evaluation circuit,
typical example)
3000
50
2500
45 Signal output frequency
Output (digital number)
10 MHz
Element temperature (C)

2000
40 Signal output frequency
5 MHz
1500
35
1000
30
500
25
0
0 50 100 150 200 250 300 20
0 20 40 60 80 100 120
Number of pixels in horizontal direction
KMPDB0328EA Operation time (min)
KMPDC0382EA
[Figure 1-68] Waveform of horizontal shift register clock pulses

P1H
Correction
50%
Image sensors generally have two nonuniformities: 1)
P2H
photoresponse nonuniformity (PRNU) that is variations
KMPDC0381EA in sensitivity to photons between pixels, and 2) dark
current nonuniformity (DCNU) that occurs under the set
When integrating over a relatively long period, to discard
operating conditions. At least these two nonuniformities
the charge generated by the horizontal shift register, after
must be corrected to collect highly accurate data. Since
reading out all pixels, a dummy readout is performed
these nonuniformities vary with temperature, the
up to immediately before beginning the transfer to
correction must take the temperature into account.
the transfer gate. This method is effective also when
discarding the charge generated by the horizontal shift (1) Dark current correction
register during the integration time. Dark current differs from pixel to pixel and must therefore
be handled at the pixel level to make accurate corrections.
Element temperature When no light is incident on the CCD, the dark current
(Nt) is expressed by equation (16).
Figure 1-69 is a measurement example showing the
relationship between the element temperature and Nt(x, y, t, T) = Nd(x, y, T) t + Nb(x, y, T) (16)
operation time when the S11155-2048-01 is operated x : horizontal direction address
using our evaluation circuit (the circuit system is sealed y : vertical direction address
t : integration time
and without any heat dissipation measures). The element T : CCD temperature
temperature increases significantly when operated at high Nd(x, y, T): dark current of each pixel [e -/pixel/s]
Nb(x, y, T): dark current when integration time is zero
speeds. Since an increase in element temperature causes
an increase in dark current, it is recommended that heat
When the integration time is zero, the dark current Nb(x,
dissipation measures be taken such as by providing a
y, T) is also called the offset or bias. This value varies with
heatsink or a fan.
the operating conditions. The dark current values listed
in our datasheets are the dark currents of Nd(x, y, T)
averaged over a certain region and are different from the
dark current actually output from the CCD. To correct the
dark current, both Nd and Nb must be acquired. Nd and
Nb can be acquired from a single data readout, but more
accurate correction image data that excludes the effect
of disturbing noise can be obtained by acquiring a few or
up to a dozen images and taking their average.
(2) Flat field correction

As described in Photoresponse nonuniformity in section
1-2 Characteristics, the sensitivity of each pixel in a CCD
is not uniform, so it must be corrected at the pixel level
just as with the dark current. An uncorrected output I(x, y)
32
measured under certain exposure conditions is given by and disappear. However, some of the holes with less
equation (17). mobility than electrons are trapped in the oxide film
and produce a voltage that can shift the CCD operating
I(x, y) = Nt(x, y, t, T) + i(x, y) r(x, y) (17) gate voltage. This causes the CCD pinning voltage to
shift toward the negative side (amount of shift may be
i (x, y): original image output
r(x, y): sensitivity of each pixel up to several volts in some cases). When high energy
electrons or photons enter, both ion damage and bulk
To acquire the original image output i(x, y), not only the damage occur. On the other hand, heavy particles such
dark current (Nt) but also r(x, y) must be known. Normally, as protons and neutrons also generate charges in the
the sensitivity r(x, y) can be acquired by illuminating a CCD gate insulation film. Electrons and photons generate a
with extremely uniform light and measuring the output. new interface state at the oxide film interface. Since the
However, uniformly illuminating the entire surface of energy level of that new interface state is within the band
a CCD is difficult. Moreover, the sensitivity may vary gap, the dark current will increase.
with wavelength depending on the position of the CCD Bulk damage occurs when energized charged particles
photosensitive area. To accurately correct this within 1% or like protons interact with the silicon atoms. If protons
less, it is necessary to acquire correction data while paying have enough energy, they displace the silicon atoms from
attention to the optical systems and temperature, etc. The their lattice positions to interstitial locations (energy of
sensitivity r(x, y) can be acquired from a single data readout, approx. 100 eV is required to cause this). The displaced
but more accurate calibration data can be obtained by silicon atoms then collide with other silicon atoms to
acquiring a few or up to a dozen images and taking their further displace more atoms. The resulting defects serve
average to eliminate the effect of external noise. as electron traps. If many electron traps are created
inside the charge transfer channels of the CCD, the
Coupling to FOS charge transfer efficiency (CTE) will deteriorate. Those
defects will become pixels with a large dark current.
Coupling an FOS (fiber optic plate with scintillator) to
[Figure 1-71] Damage by X-rays (S9970-0907, typical example)
a front-illuminated CCD allows detecting X-rays up to
(Td=-20 C, X-ray source 60 kVp)
several dozen keV or higher. See section 3 CCD area 105
image sensors in chapter 9 X-ray detectors for more

information. 104 6220 roentgens
622 roentgens
Damage by radiation 103
Ion damage and bulk damage can occur in a CCD due to 102
Before
radiation, the same as with other devices made of silicon. irradiation
This must be considered before attempting to use a CCD

101
for X-ray detection or in space environments.
100
[Figure 1-70] Effects of radiation on CCDs -11 -10 -9 -8 -7 -6 -5
Low-energy High-energy High energy Gate voltage (V)

photons photons particles
KMPDB0243EA
Gamma-rays
Electrons
UV light X-rays
Protons
Neutrons
Heat dissipation
Ion damage Bulk damage (1) Heatsink selection

When using a thermoelectrically cooled CCD, you must
Charge trap Phosphorus vacancy
Interface state Atom between vacancy and lattice select a heatsink with sufficient heat dissipation capacity.
Increase in dark current Deterioration in charge transfer

Change in operating conditions efficiency
Increase in dark current
Generation of white spots
RTS (random telegraph signal)
Increase in readout noise
KMPDC0206EA
Ion damage occurs when photons with energy higher

than a certain level (energy levels roughly higher than
ultraviolet light) enter a CCD and the resulting electron-
hole pairs are generated in the gate oxide film. Most
electron-hole pairs generated by photons will recombine
33
[Figure 1-72] Temperature characteristics of one-stage Dow Corning Corporation, thermally conductive sheet
thermoelectrically cooler 5580H by 3M) between the product and heatsink. Such
(S7031-1006S/-1007S, CCD undriven)
material should be applied over the entire junction area
(Typ. Ta=25 C) of the product and heatsink with uniform thickness.
7 30
Voltage vs. current
CCD element temperature vs. current
When a mica sheet is used, it must also make contact
6 20
with the entire heat-dissipating surface of the package.
CCD element temperature (C)

5 10
The cooling effect will degrade if the sensor package
is fastened to the heatsink with screws while the mica
Voltage (V)
4 0 sheet is still too small to cover the screw positions.

3
This may also warp the package base, causing cracks
-10
between the sensor and the package base [Figure 1-74].
2 -20 Never press the window when inserting the product
into the circuit board. Pressing the window may crack
1 -30
or break the window or cause the window to fall out,
0 -40 resulting in a malfunction [Figure 1-75].
0 1 2 3 4
When fastening a heatsink to the product with screws,
Current (A) set the tightening torque to 0.3 Nm or less, and tighten
KMPDB0179EA
the two screws alternately to apply uniform stress on
[Figure 1-73] Temperature characteristics of two-stage the product.
thermoelectrically cooler
(S7032-1007, CCD undriven) [Figure 1-74] Method of mounting to heatsink
(Typ. Ta=25 C)
7 30
Voltage vs. current
CCD element temperature vs. current Screws
6 20 Base
CCD element temperature (C)
(package heat-dissipating surface)

Sensor Sensor
5 10 Mica sheets
Heatsinks
Voltage (V)
4 0
KMIRC0023EA
3 -10
2
[Figure 1-75] Method of mounting to circuit board
-20
1 -30
0 -40
0 1 2 3 4 5 6
Current (A)
KMPDB0384EA
(2) Device design

The device must be designed so that the heat generated
by the heatsink is adequately dissipated. Provide good air
ventilation by installing air fans and ventilation ducts. (4) Current supplied to thermoelectric cooler
(3) Heatsink mounting method To protect the thermoelectric cooler and to maintain stable
To allow the thermoelectric cooler to exhibit fullest operation, set the current supplied to the thermoelectric
cooling capacity, the heatsink must be mounted correctly cooler to no greater than 60% of the maximum current
onto the product. Mount the heatsink while taking the specified in the datasheet.
following precautions.
Check that the heat-dissipating surface of the product
Electrostatic and surge measures
and the attachment surface of the heatsink are clean
CCDs may become damaged or deteriorate if subjected
and flat.
to static electrical charges and voltage surges. Take the
If the heat dissipation by the thermoelectric coolers
following precautions to avoid trouble.
during cooling is insufficient, the element temperature
will increase and may cause physical damage to the (1) Handling precautions
product. Provide sufficient heat dissipation during
When taking a CCD out of the packing box, always do
cooling. As a heat dissipation method, we recommend
so in locations where anti-electrostatic measures are
inserting a highly thermal conductive material (e.g.,
provided. For example, lay a grounded conductive sheet
thermally conductive silicone gel GR-d by Fuji Polymer
(1 M to 100 M) on the work bench or work floor.
Industries, thermally conductive silicone SE 4490 CV by
When handling CCD sensors, always wear a wrist strap
34
and also anti-static clothing, gloves, and shoes, etc. The
wrist strap should have a protective resistor (about 1 M) 1-4 New approaches
on the side closer to the body and be grounded properly.
Using a wrist strap having no protective resistor is Hamamatsu will continue to develop new CCDs by
hazardous because you may receive an electrical shock if applying new technologies in addition to the technologies
electric leakage occurs. accumulated in the past.
Always ground the soldering iron so no leakage voltage is
(1) Fully thinned CCD
applied to the device.
Do not bring charged objects (VDT for a PC or insulator Except for a portion of the near infrared-enhanced types,
materials such as plastics and vinyl) close to the CCD. our back-thinned CCDs are partially thinned types in
The CCD may become electrically charged just by which a portion of the silicon is made thin. Since the
being brought close to a charged object, causing ESD silicon thickness of the photosensitive area of partially
(electrostatic discharge) damage. thinned CCDs is approximately 10 to 30 m, it is difficult
to fabricate large-area devices. In addition, because
(2) Usage precautions the thin silicon must be enclosed with thick silicon to
Measurement devices and jig tools must be properly provide support, it is difficult to fabricate light-shielding
grounded so no surges are applied to the CCD by a leakage parts or mount FOP on the photosensitive surface.
voltage. Do not allow a voltage exceeding the absolute To solve these problems, Hamamatsu is developing
maximum rating to be applied to the CCD from the fully thinned CCDs. To create a fully thinned structure,
measurement device or tester, etc. (This tends to occur a supporting substrate is pasted to the wafer and then
during ON/OFF switching of power sources, so use caution.) the entire wafer is made thin. Through-hole electrodes
If there is the possibility of a surge voltage, insert a filter are fabricated on the supporting substrate side. The
(made up of a resistor or capacitor) to protect the CCD. adoption of this fully thinned structure strengthens
When installing the CCD, be extremely careful to avoid the chip and makes fabrication of large-area devices
reverse insertions, wrong insertions, or shorts between possible. This also makes it easy to form light-shielding
terminals. parts and mount FOP on the photosensitive surface.
Do not attach or detach any connector that is connected
[Figure 1-76] Structure of partially thinned CCD
to the power supply line or output line during operation.
Incident light
(3) Carrying precautions
P+ Accumulation layer
When carrying a CCD, place it on a conductive mat by
10 to 30 m
P- P type silicon
inserting the lead pins into the mat (for shorting leads) Gate oxide film
BPSG film
and then put it in a conductive case. The PC board for Electrode
N Poly-silicon electrode
mounting the CCD should also be put in a conductive KMPDC0458EA
case for carrying. Avoid using plastic or styrofoam boxes

since they may generate static electricity due to vibration [Figure 1-77] Structure of fully thinned CCD
during shipping and cause device deterioration or Incident light
breakdowns.
Accumulation layer
P+
(4) Storage precautions P type silicon
30 m
P-
10 to
Gate oxide film

BPSG film
When storing a CCD, place it on a conductive mat by
300 m
N Poly-silicon
100 to
Support substrate
inserting the lead pins into the mat (for shorting the electrode
Through-hole electrode
leads) and then put it into a conductive case. The PC
KMPDC0459EA
board for mounting the CCD should also be put in a
conductive case. (2) TDI-CMOS CCD
Avoid placing CCDs near equipment that may generate
high voltage or high magnetic fields. The TDI-CMOS CCD is an image sensor that combines
the feature of the TDI-CCD that allows acquiring of high
S/N images even under low-light conditions during high-
It is not always necessary to provide all the electrostatic
speed imaging and the features of the CMOS sensor that
and surge measures stated above. Implement these
provides digital output (an A/D converter is built into
measures according to the extent of deterioration or
the chip) and low voltage operation. Signal charges are
damage that may occur.
integrated and transferred in the charge state through
TDI operation, converted into voltage by a readout
amplifier in each column, converted into a digital
signal by the on-chip signal processing circuit and A/D
converter, and then output.
35
(3) Back-thinned CCD with suppressed UV-light-
irradiation-caused sensitivity deterioration 2. NMOS
We are developing back-thinned CCDs with high
sensitivity in the ultraviolet region and low sensitivity
linear image sensors
deterioration (deterioration becomes a problem during
UV light irradiation) by applying ingenuity to the NMOS linear image sensors are self-scanning photodiode
production method and sensor structure. arrays designed as detectors for multichannel
spectrophotometers. The scanning circuit of these image
sensors is made up of N-channel MOS transistors and
operates at low power consumption, making them easy
to use. Each photodiode has a large photosensitive area
and high UV sensitivity, yet the noise is extremely low so
high S/N signals can be obtained even at low light levels.
Current output type NMOS linear image sensors also
deliver excellent output linearity and wide dynamic range.
2-1 Features
Wide photosensitive area
High UV sensitivity and stable characteristics under

UV light irradiation
Low dark current and large saturation charge
Allows a long integration time and wide dynamic range

at room temperature
Excellent output linearity and sensitivity uniformity

Low power consumption
CMOS logic compatible start pulse and clock pulse
Infrared-enhanced type: high sensitivity in infrared

and soft X-ray regions
Thermoelectrically cooled type: highly reliable

package sealed with sapphire window
2-2 Structure
NMOS linear image sensors consist of a photodiode array
that performs photoelectric conversion and stores the
resulting charges, address switches connected to each
photodiode, and a digital shift register. Each address switch
[Table 2-1] Hamamatsu NMOS linear image sensors (current output type)
Pixel height Pixel pitch

Type Type no. Number of pixels
(mm) (m)
S3901 series 50 128, 256, 512, 1024
2.5
S3904 series 25 256, 512, 1024, 2048
Standard type
S3902 series 50 128, 256, 512
0.5
S3903 series 25 256, 512, 1024
Infrared-enhanced S8380 series 50 128, 256, 512
2.5
type S8381 series 25 256, 512, 1024
Note: We also offer voltage output types with built-in current integrating circuit and thermoelectrically cooled types for low-light-level detection.
36
is a MOS switch with the source connecting to a photodiode, start time and end time for each pixel will differ. The next
the gate handling the address pulses from the shift register, start pulse cannot be input until the last pixel is scanned.
and the drain connecting to the video line (common output
line). Figure 2-1 shows an equivalent circuit of a current
output type NMOS linear image sensor (S3901 to S3904 2-4 Characteristics
series).
[Figure 2-1] Equivalent circuit (S3901 to S3904 series) Input/output characteristics

Start st
Digital shift register
Clock 1
(MOS shift register)
End of scan Figure 2-2 shows input/output characteristics of current
Clock 2
output type NMOS linear image sensors (S3901/S3904
Active Active video
photodiode series). The horizontal axis indicates the exposure
amount which is the product of the incident light level
Vss and the integration time. Since no photodiode can store
Saturation
control gate a charge larger than its junction capacitance bias
Saturation
control drain Dummy video voltage, there is an upper limit on the output charge.
Dummy diode
This upper limit on the output charge is referred to as
the saturation charge, and the exposure at this point is
KMPDC0020EA referred to as the saturation exposure.
In an over-saturated state, an excess charge flows into
the overflow drain connected to the photodiode and will
2-3 Operating principle not affect other pixels. When a charge amplifier is used to
read out the signal, the actual output deviation from the
Photodiodes are initialized to a certain potential from an ideal line =1 is within 1% which indicates satisfactory
external circuit when their address switches turn on. When input/output characteristics.
light enters a photodiode, a photocurrent proportional
[Figure 2-2] Output charge vs. exposure (S3901/S3904 series)
to the incident light level flows in the photodiode, and
a charge (Qout = I L Ts) equal to the product of the 102
(Typ. Ta=25 C, Vb=2 V, V=5 V, light source 2856 K)
photocurrent (IL) and the integration time (Ts) is stored

in the junction capacitance (Cj) of the photodiode. The
101 Saturation charge
integration time is the period between the times that the
Output charge (pC)
photodiode address switch turns on. It corresponds to the S3901 series

100
time interval between the start pulses applied to the shift
S3904 series
register. This photodiode operation is called the charge
integration mode and allows acquiring a large signal even 10-1
from image sensors with small pixels by integrating the Saturation exposure
charge. In low-light-level detection, making the integration 10-2
time longer yields a large output.

The shift register starts scanning when a start pulse is 10-3
10-5 10-4 10-3 10-2 10-1 100
input with a clock pulse train applied to the shift register.
The address switch connected to the first pixel photodiode Exposure (lxs)
KMPDB0042EB
first turns on in synchronization with the clock pulse and
the charge stored in the first pixel photodiode is output
to the external circuit via the video line. The first pixel Spectral response
address pulse then turns off, and the second pixel address
pulse turns on, allowing the charge stored in the second Figure 2-3 shows spectral response characteristics of
pixel photodiode to be read out to the external circuit in standard type and infrared-enhanced type NMOS linear
the same way. By sequentially repeating this operation image sensors. Both types have sensitivity ranging from
until the last pixel is scanned, the position information 200 nm to 1000 nm.
of light irradiated onto the linearly arranged photodiode Sensitivity varies linearly with temperature. At wavelengths
array is read out to the external circuit as time-series shorter than the peak sensitivity wavelength, however,
signals including a time differential. sensitivity is stable and is not significantly dependent on
The minimum integration time is the product of the temperature. The longer the wavelength region, the larger
readout time per pixel and the number of pixels. In NMOS the temperature dependence, and these types exhibit a
linear image sensors, the address switch on/off operation temperature coefficient of approx. 0.7%/C at 1000 nm.
determines the integration time of the photodiode, and NMOS linear image sensors provide stable operation
its timing shifts for each photodiode. So even though the even during ultraviolet light measurement because
integration time is the same for all pixels, the integration their structure is designed to prevent sensitivity from
deteriorating due to ultraviolet light.
37
[Figure 2-3] Spectral response (typical example) noise. The charge amplifier reset noise can be reduced
by using a clamp circuit. In NMOS linear image sensors,
amplifier noise is predominant and increases as the video
Infrared enhanced type line capacitance becomes larger. Noise is defined by the
standard deviation of the output. In the case of the S3901-
Photosensitivity (A/W)
128Q NMOS linear image sensor, noise is 3000 e- rms in

terms of the number of electrons and 0.5 fC rms in terms
of charge. When defining the dynamic range as the ratio
of the saturation charge (upper limit) to the noise (lower
limit), the S3901-128Q dynamic range is 10 5 since the
saturation charge is 50 pC and the noise is 0.5 fC. Effective
Standard type methods for reducing the noise are to insert a low-pass
filter in the signal processing circuit, perform low-speed
readout, and average the data after acquiring ten or more
Wavelength (nm) pieces of data.
KMPDB0161EA
Photoresponse nonuniformity
2-5 How to use
Image sensors contain a large number of photodiodes in An external driver circuit for NMOS linear image sensors
arrays, and each photodiode is different from the others in includes a digital circuit for generating input clock pulses
terms of sensitivity. This may be due to crystalline defects and an analog circuit for converting output charges into
in the silicon substrate or variations in processing and voltage signals. The digital circuit consists of a clock
diffusion during the manufacturing process. In NMOS oscillator circuit and a timing control circuit. Clock pulse
linear image sensors, these variations are evaluated signals should be input at CMOS logic levels. The analog
in terms of photoresponse nonuniformity (PRNU) by circuit consists of an output processing circuit and an
illuminating the image sensor with uniform light emitted amplifier circuit. The output processing circuit normally
from a tungsten lamp and measuring variations in the uses a charge integration circuit including a charge
output from all pixels. PRNU is given by equation (1). amplifier. This method has the advantages that signal
detection accuracy is high and it produces easy-to-process
PRNU = (X/Xave) 100 [%] (1) boxcar waveforms. Figure 2-4 shows a recommended
block diagram of an external current integration method,
Xave: average output of all pixels
X : difference between Xave and maximum or minimum pixel output and Figure 2-5 shows the timing chart.
The PRNU of our NMOS linear image sensors is specified [Figure 2-4] Recommended block diagram
(external current integration method)
as 3% maximum.
Control signal generator D.GND
st
1, 2 Start, CLK
Dark output Buffer PLD Buffer
EOS, Trigger
EOS Video signal processor VCC
Dark output is an output that is generated even when Reset

Video
no light strikes the image sensor. This is caused by
Amp Buf Data video
recombination current on the photodiode surface and Sensor
C-V Clamp
within the photodiode depletion layer. Although the
magnitude of dark output differs from pixel to pixel, it is
A.GND
a fixed pattern so that the dark output component can be +2 V
+15 V
removed by measuring dark state and light state data
-15 V
and obtaining the difference between them by means Voltage regulator
of software. Dark output is temperature-dependent, so KACCC0149EA
it nearly doubles for every 5 C increase in temperature.

The dark output will increase during a long exposure but [Figure 2-5] Timing chart
can be reduced by cooling the image sensor. 50 ns min.
st
Noise and dynamic range

1, Reset
Noise is random output fluctuations over time and 2
determines the detection limit in the low-light-level

KMPDC0024EA
region. Noise includes dark current shot noise, incident
light shot noise, charge amplifier reset noise, and amplifier
38
Immediately before the address switch connected to a
photodiode turns on, a reset pulse discharges the feedback 3. CMOS linear image sensors
capacitance in the charge amplifier. When the address
switch turns on, the charge stored in the photodiode is
In the CMOS process, unlike NMOS process technology,
stored in the feedback capacitor. The relationship between
digital and analog circuits can be fabricated onto the chip.
the output voltage ( Vout) of the charge amplifier and
CMOS linear image sensors have signal processing and
the integrated charge (Qout) of the photodiode and the
timing control circuits fabricated on the image sensor
feedback capacitance (Cf ) is expressed by equation (2).
chip and so need only a simple external driver circuit.
Vout = Qout/Cf (2) Functions difficult to implement in external circuits can be
built into the image sensor to make it more sophisticated.
A capacitor of about 10 pF is used as the feedback A/D converters, for example, can be fabricated on the chip
capacitance. A clamp circuit is connected to the latter to output video data as digital signals. We also welcome
stage of the charge amplifier. The output of the clamp requests for custom devices, so feel free to consult us for
circuit should be connected to ground in the period special orders.
immediately after the feedback capacitance is reset. This
drastically reduces noise components generated when the
feedback capacitance is reset. 3-1 Features
Precautions when building driver circuits Incorporating a signal processing circuit into the sensor
chip to match the required specifications integrates the
Separate the analog circuit ground and the digital following features into the sensor. This allows downsizing
circuit ground. the photo-sensing systems and upgrading their functions.
Connect the video output terminal to the amplifier High-speed response
input terminal in the shortest possible distance.
Avoid crossing of analog and digital signals as much as
High gain
possible. Low noise
Use a series power supply having only small voltage
fluctuations. Digital output mode (with built-in A/D converter)
Low voltage drive (3.3 V drive)
[Table 3-1] Hamamatsu CMOS linear image sensors
Pixel height Pixel pitch Video data rate

Type Type no. Number of pixels
(m) (m) max. (MHz)
S9226 series* 1024 125 7.8 0.2
PPS (Passive Pixel Sensor) S8377 series 128, 256, 512 50
500 0.5
S8378 series 256, 512, 1024 25
S9227 series* 512 250 12.5 5
S10453 series 512, 1024 500 25
S11106-10 64 127 127 10
APS Compact
S11107-10 128 63.5 63.5
(Active Pixel Sensor)
High S11108 14
2048 14 10
sensitivity S11639 200
High speed S11105 512 250 12.5 50
Digital output S10077 1024 50 14 1
S10121 series 128, 256, 512 50
2500 0.25
S10124 series 256, 512, 1024 25
Current output
S10122 series 128, 256, 512 50
500 0.5
S10123 series 256, 512, 1024 25
* Surface mount type compact plastic packages are also available.
39
In the S8377/S8378 series, since the signals are read out
3-2 Operating principle and characteristics by a single charge amplifier in the last stage, the charge
amplifier must be reset each time one pixel is read out.
Here we introduce the following five types among our The video data rate of the S8377/S8378 series is 500 kHz
CMOS linear image sensors. maximum.
Standard type S8377/S8378 series High-speed type S11105 series
The S8377/S8378 series CMOS linear image sensors The S11105 series CMOS linear image sensors have
have on-chip circuits that are built in an external circuit a simultaneous charge integration function for high-
section for NMOS linear image sensors. A block diagram speed readout. Compared to the video data rate of 10
is shown in Figure 3-1. Like NMOS linear image sensors, MHz maximum on the previous high-speed type (S10453
these CMOS linear image sensors consist of photodiodes, series), the S11105 series has achieved a video data rate
address switches, and shift registers. A timing generator of 50 MHz maximum. The photosensitive area consists
is formed on the input side, and a signal processing of 512 pixels at a height of 0.25 mm, arrayed at a 12.5 m
circuit made up of a charge amplifier and clamp circuit pitch. In NMOS linear image sensors and S8377/S8378
forms the readout circuit on the output side. series CMOS linear image sensors, a lag occurs in the
pixel charge integration start/end times. However, the
[Figure 3-1] Block diagram (S8377/S8378 series)
S11105 series has simultaneous integration and variable
integration time functions (shutter function) controlled
7 EOS by an internal CMOS signal processing circuit, so charge
CMOS digital shift register 3 Vg integration in all pixels can start and end simultaneously.
Address switch Charge Clamp
amplifier circuit 6 Video The S11105 series has a CMOS amplifier array to convert
Photodiode array charges to voltages. The conversion gain is determined
1 2 3 4 5 N-1 N by the charge amplifiers feedback capacitance. A small
feedback capacitance of 0.1 pF allows a high output
voltage.
Timing generator
Each photodiode pixel is connected to a charge amplifier.
4 8 1 2 There is no switch between the photodiode and a charge
Vdd Vss CLK ST amplifier. Since the photodiodes act as a current source, the
KMPDC0150EC
generated signal charge is not stored in the photodiodes
The S8377/S8378 series operate only on a single 5 V but is stored in the charge amplifiers feedback capacitance.
power supply, ground, a clock pulse, and a start pulse. The output voltage from the charge amplifier changes in
All pulses necessary to operate the shift register, charge proportion to the incident light level during the integration
amplifier, and clamp circuit are generated by the timing time. A hold circuit is connected following the charge
generator. An analog video output with boxcar waveform amplifier of each pixel. The charge amplifiers of all pixels
and an end-of-scan pulse are the output signals. The are simultaneously reset. By inputting a hold pulse to each
charge-to-voltage conversion gain can be adjusted in hold circuit immediately before the charge amplifiers
two steps by switching the charge amplifiers feedback are reset, the charge amplifier outputs from all pixels
capacitance via the input voltage to the gain selection are simultaneously held in their respective hold circuits.
terminal. The time from when the reset switch for each charge
The peak sensitivity wavelength is 500 nm, which is amplifier is turned off to when the hold pulse is input is
shorter than NMOS linear image sensors. Since the the integration time. Charge integration therefore starts
CMOS linear image sensors operate on a single 5 V and ends simultaneously for all pixels. An address pulse
power supply, their dynamic range is narrow compared from the shift register is next input to the switch in the
to NMOS linear image sensors that operate on 15 V stage after the hold circuit to allow the output signals being
supply. However, there is almost no difference in basic held to be sequentially output as a time-series signal from
characteristics such as linearity accuracy and dark the video output terminal. Since this signal readout from
output compared to NMOS linear image sensors. CMOS the hold circuits is performed in a circuit separate from
linear image sensors are suitable for use in compact the operation for integrating the photodiode charges,
measurement systems since they need only a simple the photodiodes and charge amplifiers can start the
external driver circuit. next charge integration while video output readout is in
The S8377/S8378 series are available in six types with progress.
different pixel pitches and number of pixels. A variant
type, S9226-03 series, is also available having the same
block configuration but a different pixel format with 7.8
m pitch, 0.125 mm photosensitive area height, and 1024
pixels.
40
[Figure 3-2] Block diagram (S11105 series) series. The reset pulses for the charge amplifiers, hold
pulses for the hold circuits, and a start pulse for the
shift register are all generated by the internal timing
Trig Shift register EOS
generator. Switching the start pulse from high to low
Hold circuit Video
initializes the timing generator which then sequentially
CLK
Timing
generator
generates the various control pulses. First, hold pulses
ST Charge amplifier array
are generated to hold the charge amplifier outputs in
Bias
Photodiode array generator the hold circuits. Next, the reset pulses for the charge
amplifiers are switched on to reset the charge amplifiers.
KMPDC0312EA No signal charges are integrated while the charge
amplifiers are in a reset state. A start pulse is then input
[Figure 3-3] Equivalent circuit (high-speed readout
circuit of S11105 series) to the shift register to sequentially read out the video
output as a time-series signal from the first pixel. When
Signal component
the start pulse changes from low to high, the reset pulses
Buffer circuit for the charge amplifiers are switched off, or in other
Differential circuit
words, charge integration starts. When the start pulse
Peak hold
circuit
again changes from high to low, the timing generator is
initialized as described above, and one cycle of operation
is complete. Strictly speaking, charge integration starts
0.5 clocks after the start pulse has changed from low
Reset component
KMPDC0462EA
to high, and ends 0.5 clocks after the start pulse has
changed from high to low. Therefore, the integration time
The S11105 series employs a high-speed readout circuit is equal to the high period of the start pulse. If the length
that uses a peak-hold circuit to increase the video data of one cycle is fixed, then the integration time can be
rate [Figure 3-3]. The signal and reset components enter adjusted by changing the ratio of high to low periods.
the differential circuit. Only the signal component is With the S11105 series, in addition to reading out all 512
output and enters the peak-hold circuit, and there the pixels, it is possible to read out a portion of the pixels
output waveform is held at the peak value of the signal (e.g., 32 pixels from the first to the 32nd pixel) [Figure
component. Since there is no need to reset after reading 3-6]. The line rate when reading out 512 pixels is 88.65
each pixel signal as in a normal circuit, signal fluctuation kHz and 595 kHz when reading out 32 pixels.
is reduced, making high-speed readout possible. Figure
3-4 shows the video output waveform displayed on an [Figure 3-6] Operation example
oscilloscope. (a) When reading out all 512 pixels
tlp(ST)=1.04 s thp(ST)=10.24 s
[Figure 3-4] Video output waveform (S11105 series)
ST
Voltage (V)
CLK
50 MHz tpi(ST)=11.28 s (line rate: 88.65 kHz)

(1 pixel)
Video
When the clock pulse frequency is maximized (video data rate is also maximized), the time of
one scan is minimized, and the integration time is maximized (when reading out all 512 pixels).
20 40 60 80 Clock pulse frequency=Video data rate=50 MHz
Start pulse period=564/f(CLK)=564/50 MHz=11.28 s
High start pulse period=Start pulse period - Minimum low start pulse period
Time (ns) =564/f(CLK) - 52/f(CLK)=564/50 MHz - 52/50 MHz=10.24 s
The integration time corresponds to high start pulse period, which is 10.24 s.
KMPDC0407EA
Two types of input pulses consisting of a clock pulse
and a start pulse are required to operate the S11105
[Figure 3-5] Timing chart (S11105 series)

1 2 3 17 18 19 46 47 48 49 50 51
CLK
Integration time
tlp(ST)
ST
thp(ST)
tpi(ST)
512 1 512
Video
Trig
EOS
KMPDC0463EA
41
(b) When reading out all 1 to 32 pixels the analog signal into a digital signal which is serially output
tlp(ST)=1.04 s thp(ST)=0.64 s from the MSB (most significant bit). Even though it has
a small photosensitive area size, it delivers a high output
voltage since the charge amplifiers feedback capacitance
ST
of the readout circuit is set to a small value of 0.05 pF. A
tpi(ST)=1.68 s (line rate: 595 kHz)
switch and a hold circuit are connected to each photodiode
pixel. During the integration time, the signal charge of each
When the clock pulse frequency is maximized (video data rate is also maximized) and photodiode, which is proportional to the incident light level,
the integration time is maximized (when stopping the output at channel 32).
Clock pulse frequency=Video data rate=50 MHz is transferred to the hold circuit and held there. All pixels are
Start pulse period=84/f(CLK)=84/50 MHz=1.68 s
High start pulse period=Start pulse period - Minimum low start pulse period simultaneously reset. The integration time is from when the
=84/f(CLK) - 52/f(CLK)=84/50 MHz - 52/50 MHz=0.64 s
The integration time corresponds to high start pulse period, which is 0.64 s. reset switch for each photodiode is turned off to when the
KMPDC0408EA hold pulse is turned on and then off. An address pulse from
the shift register is next input to each hold circuit to allow
The previous S10453 series was available only in DIP
the output signals being held to be sequentially output as
packages, but the S11105 series is available in two package
a time-series signal from the video output terminal. Since
types: DIP (S11105) and surface mount (S11105-01).
this signal readout from the hold circuits is performed in
a circuit separate from the operation for integrating the
Digital output type S10077 photodiode charges, the photodiodes can start the next
charge integration while video output readout is in progress.
The S10077 is a low power consumption CMOS linear Two types of input pulses consisting of a clock pulse and
image sensor incorporating a simultaneous integration start pulse are required to operate the S10077 series.
function and internal A/D converter. It provides an 8-bit The reset pulses for the photodiodes, hold pulses for the
or 10-bit digital output which is switchable. The video data hold circuits, and a start pulse for the shift register are
rate is 1 MHz maximum, and the S10077 can operate from all generated by the internal timing generator. Switching
a single supply voltage of 3.3 V at a power consumption the start pulse from high to low initializes the timing
of 30 mW. The photosensitive area consists of 1024 pixels generator which then sequentially generates the various
at a height of 0.05 mm, arrayed at a 14 m pitch. The control pulses. First, hold pulses are generated to hold
simultaneous integration and variable integration time the photodiode charges in the hold circuits. Next, the
functions (shutter function) are controlled by an internal reset pulses for the photodiodes are switched on to reset
CMOS signal processing circuit. the photodiodes. No signal charges are integrated while
the photodiodes are in a reset state. A start pulse is then
[Figure 3-7] Block diagram (S10077)
input to the shift register to sequentially read out the
CLK ST D.Trig A.Trig Vdd Vss
video output as a time-series signal from the first pixel.
22 23 2 4 8 21 9 20
At the time when the start pulse changes from low to
Timing generator Bias generator high, the reset pulses for the photodiodes are switched
off, or in other words, charge integration starts. When
Shift register 16 EOS the start pulse again changes from high to low, the
timing generator is initialized as described above, and
Address switch Readout circuit 5 AO
one cycle of operation is complete. Strictly speaking,
charge integration starts 0.5 clocks after the start pulse
Hold circuit A/D converter 3 DO
has changed from low to high, and ends 7.5 clocks after
Photodiode array the start pulse has changed from high to low. Therefore,
19 17 the integration time is equal to the sum of the high
Vsel EOC period of the start pulse and the period of 7 clock pulses.
KMPDC0293EA
Within one cycle, the integration time can be adjusted
[Figure 3-8] Equivalent circuit (S10077) by changing the ratio of high to low periods of the start
pulse.
Start
Reset Shift register

Hold
Photodiode
Hold circuit
KMPDC0292EA
In the S10077, the signal charge stored in each photodiode

is transferred to the hold circuit, and the resulting analog
voltage is then sent from the readout circuit to the A/D
converter via the address switch. The A/D converter converts
42
[Figure 3-9] Timing chart (S10077, 10-bit mode) [Figure 3-10] Contrast transfer function characteristics
(a) Vicinity of start pixel 1 line pair
T1(CLK)
Incident
012345 10 15 20 25 30 35 40 45 50 55 light
CLK
T3(ST)
ST
Output
T4(ST)
T2(ST)
Pixel pitch
7
EOS KMPDC0070EA
22 34 46
AO AO1 AO2 AO3 AO4
21 33 45 57 The fineness of the black and white lines for input
Trig(A)
pattern is given by the spatial frequency of the input
D9
34
D0 D9
44 46
D0
56
image. The spatial frequency is the number of black and
DO DO1 DO2 white line pairs per unit length. The higher this spatial
34.5 43.5 46.5 55.5
Trig(D) frequency, the finer the input pattern, causing a drop
33 45 57 in the contrast transfer function. Figure 3-11 shows
EOC
an example of the S10077 contrast transfer function
(b) Vicinity of last pixel measurement.
CLK
[Figure 3-11] Contrast transfer function vs. spatial
ST
frequency (S10077, typical example)
T4(ST)
(White light)
T2(ST)
12322
EOS
12286 12298
AO
Contrast transfer function
AO1023 AO1024
12285 12297
Trig(A)
D0 D9 D0 D9 D0 D9 D0
12284 12286 12296 12298 12308 12310 12320
DO DO1022 DO1023 DO1024
12283.5 12286.5 12295.5 12298.5 12307.5 12310.5 12319.5
Trig(D)
12285 12297 12309 12320
EOC
KMPDC0226EA
The resolution of the S10077 is indicated as a contrast

transfer function. Resolution is a measure of the
degree of detail to which image sensors can reproduce Spatial frequency (line pairs/mm)
KMPDB0291EA
an input pattern. Since the photosensitive area of an
image sensor consists of a number of regularly arrayed
photodiodes, when an image pattern of alternating High sensitivity type S11108
black and white lines as shown in Figure 3-10 is input, an
output corresponding to the input pattern appears. As the The S11108 is a 2048-pixel high sensitivity type CMOS
pulse width of the input pattern becomes narrower, the linear image sensor with 14 14 m pixel size. Its
difference (contrast) between the black and white level operation is easy since it only uses a 5 V supply voltage
outputs becomes smaller. In such a case, the contrast and two types of drive pulse signals.
transfer function (CTF) is given by equation (1). The S11108 is designed to be used in commercial and
industrial applications such as barcode readers and
VWO - VBO
CTF = (1) encoders. Compared with the previous product, it offers
VW - VB
smaller pixel size, more pixels, and high sensitivity
VWO: output white level
as well as high-speed response. In comparison with
VBO : output black level
VW : output white level (when input pattern pulse width is wide) the previous PPS (passive pixel sensor) type, which
VB : output black level (when input pattern pulse width is wide)
performs charge-to-voltage conversion in the last-stage
amplifier, the S11108 is an APS (active pixel sensor)
type, which performs charge-to-voltage conversion in
the amplifier available in each pixel. This achieves high
sensitivity due to high efficiency. Moreover, to increase
the conversion efficiency of charges generated in the
photosensitive area, the charge integration capacity in
the charge-to-voltage converter has been reduced. For
the photosensitive area, unlike the previous surface type
43
photodiode structure, a buried photodiode structure was (b) Buried photodiode
employed to reduce the dark current and shot noise in
the dark state. 1 pixel
[Figure 3-12] Schematic (CMOS linear image sensors)

Gate
(a) PPS type
LOCOS P+
N-
Charge storage section N+
Shift register Shift register Shift register
P-
Video
KMPDC0467EA
Amp
[Figure 3-14] Spectral response (S11108, typical example)

PD PD PD
(Ta=25 C)
100
1 pixel
KMPDC0464EA
80
Relative sensitivity (%)

(b) APS type
60
Shift register Shift register Shift register 40
Video 20
0
400 600 800 1000 1200
Hold Hold Hold Wavelength (nm)

circuit circuit circuit KMPDB0308EB
set set set

Current output type S10121 to S10124 series
Amp Amp Amp The S10121 to S10124 series are current output type
CMOS linear image sensors consisting of a photodiode
PD PD PD
array that performs photoelectric conversion and stores
1 pixel the resulting charges, address switch array connected
KMPDC0465EA
to each photodiode, and a shift register with a readout
control function. Each address switch is a MOS switch
[Figure 3-13] Structure of photosensitive area
with the source connecting to a photodiode, the gate
(a) Surface type photodiode handling the address pulses from the shift register, and
the drain connecting to the video line (common output
1 pixel
line). The shift register comprises D (delayed) type flip-
flops and NOR gates.
LOCOS
N+ [Figure 3-15] Block diagram (S10121 to S10124 series)
Charge storage section
CLK
P-
ST Shift register EOS
INT
Active Video
Address switch array
Dummy Video
KMPDC0466EA
Vdd
Photodiode array
GND
Vofd
Overflow drain
Vofg
KMPDC0232EC
44
[Figure 3-16] Equivalent circuit (S10121 to S10124 series) [Figure 3-17] Timing chart (variable integration time function)
ST D Q D Q D Q D Q CLK
ST
C Q C Q C Q C Q EOS
CLK Readout 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4 1 2 3 4
timing
INT
INT
Active Video 1 ch integration time

2 ch integration time
Pixel 1 Pixel 2 Last pixel
Output
Dummy Video Effective
Invalid data data
KMPDC0233EC
KMPDC0279EB
The saturation charge of the CMOS linear image sensors
The structure of the S10121 to S10124 series is similar is more than twice that of NMOS linear image sensors
to the S3901 to S3904 series NMOS linear image sensors of the same pixel size [Table 3-2]. Thus, the upper limit
but different in that a shift register with readout control of light level that the CMOS linear image sensors can
function is used. On the S10121 to S10124 series, the detect is higher. Figure 3-18 shows the spectral response
shift register output can be controlled using INT pulses. of the S10121 to S10124 series. The spectral response
The output from the D type flip-flops and INT pulses range is 200 to 1000 nm, featuring high UV sensitivity.
enter the NOR gates. The output from the NOR gates They provide stable operation even during ultraviolet
is connected to the address switch gate. To perform a light measurement because their structure is designed to
readout, both the output signal from the D type flip-flops suppress sensitivity from deteriorating due to ultraviolet
and the external INT pulse must be set to low level. light.
The D type flip-flops start operating when a start pulse
[Figure 3-18] Spectral response
is input while a clock pulse is being applied to the D (S10121 to S10124 series, typical example)
type flip-flops. The D type flip-flops synchronize to the
(Ta=25 C)
falling edges of the clock pulses and output low level 0.4
logic signals in order starting from the first pixel. Set the
INT pulses for pixels that you want to read out also to low
0.3
level. For pixels that you don't want to read out, set the
INT pulses to high level. By controlling the INT pulses in

this manner, you can selectively read out the pixels that
0.2
you want. This allows the integration time to be varied
for each pixel.
The timing chart of the variable integration time function 0.1
is shown in Figure 3-17. Here, an example is provided for a
case where the integration times of the second, third, and
0
fourth pixels are set to twice, three times, and four times 200 400 600 800 1000 1200
the integration time of the first pixel. The integration time
Wavelength (nm)
can be varied for each pixel by applying INT pulses so that
KMPDB0399EA
their durations are one cycle of the start pulse for the first
pixel, two cycles for the second pixel, three cycles for the
third pixel, and four cycles for the fourth pixel.
[Table 3-2] Comparison table of saturation charges (NMOS linear image sensors, CMOS linear image sensors)
Product name NMOS linear image sensor CMOS linear image sensor
S3901 S3902 S3903 S3904 S10121 S10122 S10123 S10124
Type no.
series series series series series series series series
Pixel pitch (m) 50 25 50 25

Pixel height (mm) 2.5 0.5 0.5 2.5 2.5 0.5 0.5 2.5
Saturation charge (pC) 50 10 5 25 140 28 14 70
45
3-3 New approaches 4. CMOS
area image sensors
Increasing the pixel size of buried photodiode structures
The production of Hamamatsu CMOS area image sensors
CMOS linear image sensors employing the buried photodiode
started off with sensors used in X-ray imaging such as
structure have relatively small pixels with a size of 14
intraoral imaging for dental diagnosis. They featured
14 m or 14 42 m. Currently, we are developing larger
relatively large pixel sizes. In recent years, thanks to
pixel sizes such as 7 125 m, 14 200 m, and 127 127
the development of buried photodiode structures,
m. We are also in the process of incorporating additional
miniaturization of pixels, and improvements to readout
features to the photosensitive area such as high sensitivity
circuits, we have been providing CMOS area image
and high durability in the ultraviolet region and high
sensors that support special functions such as position
sensitivity in the near infrared region.
and pattern detection for industrial and measurement
applications, in addition to the normal imaging function.
Digital output
Most of the current Hamamatsu CMOS linear image sensors 4-1 Features
have analog output, but we are planning products with a
12-bit pipeline type A/D converter. This will result in an High near infrared sensitivity (S11661, S11662, S12524)
all-digital I/O interface making it easy to use and high
speed. Global shutter readout (S11661, S11662, S12524)
High-speed partial readout (S11661, S11662, S12524)

Furthermore, we are also developing CMOS linear image
sensors that integrate these endeavors. Position detection (S9132)
Single power supply operation
Custom devices available (various additional functions)
4-2 Operating principle and characteristics

This section introduces Hamamatsu's representative CMOS
area image sensors for industrial and measurement
applications.
Near infrared-enhanced APS (active pixel sensor) type

S11661, S11662
These CMOS area image sensors have achieved high

sensitivity in the near infrared region. We offer two
types: S11661, an SXGA type (1280 1024 pixels) and
S11662, a VGA type (640 480 pixels). Their pixel size is
7.4 7.4 m. These sensors employ a buried photodiode
structure.
In a typical CMOS image sensor, four transistors are
used to form a pixel, but the S11661 and S11662 use
five transistors to form a pixel in order to allow global
shutter readout (simultaneous integration of all pixels).
To achieve wide dynamic range, voltage that exceeds 5 V
is used to drive the transistors in the pixels, and the built-
in charge-pump type booster circuit makes it possible to
drive the image sensor using a single 3.3 V power supply.
Moreover, the built-in timing circuit allows the sensor
to be driven continuously with only a clock signal. The
sensor is equipped with a partial readout function for
reading a certain region of pixels. The reduction in the
46
number of pixels to be read out enables the frame rate to [Figure 4-3] Block diagram (S12524)
be increased. The partial readout region can be changed

for each frame.
The S11661 and S11662 contain a pipeline type A/D
converter that converts analog signals into 12-bit digital Port 1

signals for output.

Port 2

[Figure 4-1] Near infrared-enhanced APS types S11661 and S11662
512 512 pixels
Port 3
Port 4
Port 5

Column hold circuit Port 6

Port 7
Port 8
KMPDC0468EA
[Figure 4-4] High-speed sensor head of intelligent vision sensor
Multi-port type S125244)
The S12524 is a multi-port (eight-port) CMOS area image

sensor that outputs 512 512 pixel analog video signals
in parallel. Like the S11661 and S11662, this sensor offers
high sensitivity in the near infrared region. Its pixel size
is 20 20 m. It employs a buried photodiode structure.
To make global shutter readout possible, each pixel is
made up of five transistors. The built-in timing circuit
consists of the minimal necessary circuitry, and both the Profile sensor S91325)
vertical scanning circuit and horizontal scanning circuit
can be directly controlled externally. External control
The profile sensor is a CMOS area image sensor designed
makes it possible to specify where and how many regions
specifically for acquiring projection profile data. Projection
to be read out, and a moving object can be tracked by
profile data contains a full set of data of a line of output;
changing the partial readout region for each frame.
position information and luminance information of
A block diagram is shown in Figure 4-3. The sensor is
the input light spot can be acquired from this data. The
contrived to keep the readout load of the eight ports
photosensitive area of the profile sensor is divided into
uniform even during partial readout. During a full-
a photosensitive area for the X direction and that for the
screen readout of 512 512 pixels, imaging is possible
Y direction [Figure 4-6]. For the photosensitive area for
at 250 frames/s maximum. During a partial readout of
the X direction, the photosensitive area of one column is
four regions of 64 64 pixels, imaging is possible at 880
connected with metal wiring. And, for the photosensitive
frames/s maximum. Note that the S12524 is used in the
area for the Y direction, the photosensitive area of one row
high-speed sensor head of Hamamatsu intelligent vision
is connected with metal wiring. This allows the projection
sensor [Figure 4-4].
profile data of the X-axis and Y-axis to be read out. A
[Figure 4-2] Multi-port type S12524 readout amplifier and A/D converter are provided for each
direction, X and Y, allowing each direction to be controlled
separately.
Since the data size of the projection profile data for
the X and Y directions of the profile sensor is small,
position detection and moving object detection of light
spots can be performed faster than normal area image
sensors. The pixel rate is 833 kHz. When 256 pixels are
read out, the frame rate is 3200 frames/s maximum.
Typically, a two-dimensional PSD is used for light spot
detection. However, the profile sensor is advantageous
in that it improves output linearity, it can detect multiple
light spots, and it simplifies the external driver circuit.
47
Since the sensor chip contains a timing generator, bias Hamamatsu has mainly been producing CMOS area
voltage generator circuit, and successive approximation image sensors with a single A/D converter on a chip,
type 10-bit A/D converter, it can be operated with an but in recent years, we have been developing various
extremely simple external driver circuit and external types of column parallel type ADCs. We are also
signal processing circuit. developing readout circuits for monolithic, high-speed,
low-noise CMOS area image sensors as well as hybrid
[Figure 4-5] Profile sensor S9132 image sensors that combines a CCD and compound
semiconductor (e.g., InGaAs).
2-port parallel output type
Hamamatsu is developing CMOS area image sensors

with a variety of built-in functions. As an example, a
2-port parallel output type is described below. The 2-port
parallel output type has a readout circuit on the top and
bottom of the photosensitive area as shown in Figure
4-7, and each circuit can be operated independently. A
video line for each circuit is connected to each pixel, and
the data of any pixel can be read out from either circuit.
[Figure 4-6] Operating principle (profile sensor) This circuit structure allows, for example, the top half of
the photosensitive area to be read using the top circuit
and the bottom half of the photosensitive area to be read
using the bottom circuit. As another example, one circuit
could read out a specific small region at a high frame
Y-direction shift register
rate, and the other circuit could read out the remaining
regions at a normal frame rate. When this sensor is used
as an optical communication image sensor that performs
optical communication and imaging in parallel,
numerous readout modes shown below can be used.
Normal imaging mode (640 480 pixels, 40 frames/s,

the remaining port is paused)
Y-direction projection data
Double-speed imaging mode (top half of 640 240
pixels: 80 frames/s, bottom half: 80 frames/s)
Light spot Blinking light source detection mode (odd lines of 640
X-direction shift register
240 pixels: 80 frames/s, even lines: 80 frames/s)
Imaging and communication (1 location) parallel mode
(640 479 pixels: 40 frames/s, 32 1 pixels: 100 kHz)
X-direction projection data Imaging and communication (3 location) parallel mode (640
KMPDC0197EA
477 pixels: 40 frames/s, 136 1 pixels 3 regions: 20 kHz)
Communication (2 location) parallel mode (32 1
pixels: 100 kHz, 32 1 pixels: 100 kHz)
4-3 New approaches [Figure 4-7] Block diagram (2-port parallel output type)
Digital input signal
Column parallel ADC type Readout circuit (top)

Timing dout2
Horizontal control 2
generator 2 12-bit Vsync2
If we categorize image sensors by the number of A/D ADC Hsync2
Hold 2 Amp Pclk2
converters, we have a type that has a single high-speed
A/D converter on a chip, and another type that has a low-
Vertical control 2
Vertical control 1
speed A/D converter for each column (column parallel

640 480 pixels
type ADC). With advances in CMOS micromachining Bias circuit
technology, there is a growing number of cases where

image sensors with column parallel type ADC are used Hold 1 Amp dout1
Vsync1
in high-speed low-noise applications. The performance 12-bit Hsync1
Timing ADC Pclk1
advantages of column parallel type ADCs become generator 1
Horizontal control 1
apparent especially when the image sensor has large Readout circuit (bottom)
number of pixels. Digital input signal

KMPDC0469EA
48
[Figure 5-1] Block diagram
5. Distance image sensors (a) S11961-01CR
Vdd(A) GND Vr Vsf Vpg
15 14 1 17 16
Distance image sensors are image sensors that measure the
distance to the target object using the TOF (time-of-flight) VTX3 2
VTX2 3 272 pixels Bias
method. Used in combination with a pulse modulated light VTX1 4 (number of effective pixels: 256) generator
source, these image sensors output phase (time) difference p_res 5
information of emitted and received light. Distance data Vout1

13
phis 6 Sample-and-hold circuit
can be obtained by performing calculation on the output Vout2 CLA
12
signal with an external signal processing circuit or on a PC. Buffer amp
dclk CLA
mclk 7 9
trig 8
10 11 CLD
5-1 Features Vdd(D) GND
KMPDC0430EA
High-speed charge transfer (b) S11963-01CR
Wide dynamic range and low noise by non-destructive CLTX CLTX CLTX
readout (S11961/S11963-01CR) GND Vdd(A) GND Vdd_tx VTX1 VTX2 VTX3 GND Vdd_tx
3 2 1 44 43 42 41 37 36
35 sel2
Built-in column gain amplifier (S11963-01CR) 34 sel1
Vertical shift register

33 sel0
Gain: 1, 2, 4 32 Vdd(A)
168 128 pixels 31 GND
(number of effective pixels: 160 120) 30 Vpg
Fewer detection errors even under fluctuating
29 Vref2
dis_read 4 28 Vr
background light (charge drain function) ext_res 5 27 Vref
reset 6 Vout1
Column gain amplifier circuit 26
Real-time distance measurement vst 7 Timing
25 Vout2
generator CLA
hst 9 Buffer 24 Vdd(A) CLA
mclk 10 amp 23 GND
oe 8 11 dclk 12 13 19 20 21 22
GND Vdd(D) GND Vdd(D) nbias pbias
5-2 Structure CLD CLD CLb CLb
GND Vdd(A)
KMPDC0443EA
Distance image sensors consist of a photosensitive area,
shift register, output buffer amplifier, bias generator,
timing generator, and so on. The block diagram is shown
in Figure 5-1. Distance image sensors are different from
5-3 Operating principle
typical CMOS image sensors in the following manner.
The timing chart of the photosensitive area of the distance
Pixel structure that allows high-speed charge transfer image sensor is shown in Figure 5-2. Output voltages V1
Outputs two phase signals representing distance information and V2 obtained by applying charge-to-voltage conversion
from two output terminals on accumulated charges Q 1 and Q 2 based on their
Like a typical CMOS image sensor, the output signal from integration capacitances Cfd1 and Cfd2 are expressed by
the photosensitive area is processed by the sample-and- equations (1) and (2).
hold circuit or column gain amplifier circuit, scanned V1 = Q1/Cfd1 = N Iph {(T0 - Td)/Cfd1} (1)
sequentially by the shift register, and read out as voltage
output. V2 = Q2/Cfd2 = N Iph (Td/Cfd2) (2)
Cfd1, Cfd2: integration capacitance of each output

N : charge transfer clock count
Iph : photocurrent
T0 : pulse width
Td : delay time
[Table 5-1] Hamamatsu distance image sensors
Pixel size Dimensions

Type no. Type Number of pixels Package
[m (H) m (V)] (mm)
S11961-01CR Linear image sensor 256 20 50 10.6 5.8 2.0

Surface
S11962-01CR 64 64 40 40 mount type 8.18 9.26 2.0
Area image sensor PWB
S11963-01CR 160 120 30 30 11.85 9.5 2.0
49
Delay time Td when Cfd1=Cfd2 in equations (1) and (2) is [Figure 5-3] Structure and surface potential of
expressed by equation (3). photosensitive area
(a) VTX1: on, VTX2: off (in the case of Figure 5-2 )
Td = {V2/(V1 + V2)} T0 (3)
VTX1 VTX2
Using the values (V1, V2) output according to the distance,

distance (L) is expressed by equation (4). Vpg
L = 1/2 c Td = 1/2 c {V2/(V1 + V2)} T0 (4)
Cfd1 Cfd2
c: speed of light (3 108 m/s)
PG
[Figure 5-2] Timing chart of photosensitive area
Pulsed light T0
-
Reflected light Td
VTX1 Q1

Q1 Q2
- - - - - -
VTX2 Q2 KMPDC0471EA

(b) VTX1: off, VTX2: on (in the case of Figure 5-2 )
KMPDC0470EA
VTX1 VTX2
The structure and surface potential of the photosensitive
area of the distance image sensor are shown in Figure
Vpg
5-3. Typical CMOS image sensors can be driven with a
single power supply, but the transfer time needed for
the charge to move from the photosensitive area to the Cfd1 Cfd2
integration area is in the microsecond order. On the PG
other hand, high-speed charge transfer (nanosecond
order) is possible on CCD image sensors, but they require
multiple voltage inputs including high voltage.
To achieve the high-speed charge transfer (several tens of
nanoseconds) needed to acquire distance information,
we have developed a pixel structure that enables high-speed -
charge transfer like the CCDs in the CMOS process. This has
allowed distance image sensors to achieve the high-speed
charge transfer needed for distance measurement.
The number of electrons generated in each pulse emission
is several e-. Therefore, the operation shown in Figure 5-3 Q1 Q2
- - - - - -
is repeated several thousand to several tens of thousands KMPDC0472EA
of times, and then the accumulated charge is read out.

The number of repetitions varies depending on the [Table 5-2] Distance measurement range and VTX1,
VTX2, and light-emission pulse widths
incident light level and the required accuracy of distance
measurement. Distance measurement range VTX1, VTX2,
max. light-emission pulse widths
(m) (ns)
4.5 30
6 40
9 60
Note: Light travels approximately 30 cm in 1 ns.
50
(ext_res, reset, vst, hst, mclk, VTX1, VTX2, and VTX3)
5-4 Characteristics needs to be generated with an external circuit and
applied (see the datasheet). Note that because VTX1,
Figures 5-4 and 5-5 show distance measurement examples VTX2, and VTX3 are high-speed clock pulse signals, high-
under the following conditions using the S11961-01CR speed buffer ICs must be provided near the distance
and a light source. image sensor. Bias voltages (Vr, Vpg, Vref, and Vref2) must
Integration time=30 ms (effective integration time=180 s) be applied within the ranges specified in the datasheet.
Charge transfer clock width (VTX1, VTX2)=30 ns
Light receiving lens F=1.2, angle of view=37.5 27.7 Configuration example
Light source output=10 W, duty ratio=0.3%,
light emission pulse width=30 ns, =870 nm A configuration example of a distance measurement
Light projection angle=10 10 system using the distance image sensor is shown in
Background light: room light level Figure 5-6. The system consists of the distance image
sensor, light source and its driver circuit, light emitting/
[Figure 5-4] Measurement distance vs. actual distance receiving optical system, timing generator, and arithmetic
(S11961-01CR, center pixel of photosensitive
circuit for calculating distance. The distance accuracy
area, typical example)
depends greatly on the light source emission level and
(Ta=25 C)
5 the light emitting/receiving optical system.
Gray object (reflectance 18%)
White object (reflectance 90%)
[Figure 5-6] Configuration example of distance
4
measurement system
Measurement distance (m)
Distance image sensor Object

3
Optical system
Timing generator
Arithmetic circuit for Light source, driver
1 calculating distance circuit for light source
Measurement distance
0
0 1 2 3 4 5 KMPDC0473EA
Actual distance (m)

KMPDB0389EA Light source selection
[Figure 5-5] Distance accuracy vs. actual distance
When the distance image sensor is used to measure distance,
(S11961-01CR, the center pixel of the
photosensitive area, typical example) a light source (LED or pulse laser diode) suitable for the
pulse width of the distance image sensors charge transfer
(Ta=25 C)
30 clock must be selected. For example, to measure up to
White object (reflectance 90%) 4.5 m, the pulse width of the charge transfer clock and
the light emission pulse width must be set to 30 ns. Thus,
the response speed of the light source needs to be around
Distance accuracy (cm)
20
10 ns or less for rise and fall times. Since the light source
must be irradiated in a line in the case of the S11961-01CR
distance linear image sensor and over an area in the case
10 of the S11962-01CR and S11963-01CR distance area image
sensors, large output power is required. For this, multiple
light sources are sometimes used. When multiple light
sources are used, a driver circuit for driving the multiple
0
0 1 2 3 4 5 light sources at high speeds and high output is also required.
An estimation example expressing the relationship between
Actual distance (m)
KMPDB0390EA
the light source pulse peak output and the distance accuracy
is shown in Figure 5-7.
Conditions
5-5 How to use Integration time=16 ms (effective integration time=32 s)
Charge transfer clock width (VTX1, VTX2)=30 ns
To operate the distance image sensor, supply voltages, Light receiving lens F=1.2, focal distance f=2.8 mm
timing signals, and bias voltages must be applied. For the Light emission pulse width=30 ns, =870 nm
S11963-01CR, two types of supply voltages are required: Light projection angle=20 20
Vdd at 5 V and Vdd_tx at 3 V. In addition, timing signals Background light: room light level
51
[Figure 5-7] Distance accuracy vs. pulse peak output
(estimation example) 6. Photodiode arrays with
(a) Measurement distance: 1 m
amplifiers
(Ta=25 C, duty ratio 0.1%)
50
White object (reflectance 90%) Photodiodes with amplifiers are a type of CMOS linear
40
image sensor designed for long and narrow area detection
systems using an equal-magnification optical system.

30 These devices combine a Si photodiode array with a CMOS
signal processing IC. The CMOS signal processing IC contains
20 a timing generator, shift register, hold circuit, and charge
amplifier array, so only a simple external circuit is needed.
10
A long and narrow image sensor can be configured by
arranging multiple devices in a row. For X-ray detection
0
applications, we also offer photodiode array/amplifier
10 100 devices with a phosphor screen attached directly to their
Pulse peak output (W)
photosensitive area.
KMPDB0391EA
(b) Measurement distance: 2 m 6-1 Features

50
Gray object (reflectance 18%) Long and narrow sensor can be configured by
White object (reflectance 90%)
40
placement of multiple devices.
5 V operation
30
Simultaneous integration by charge amplifier
20 Time-series signal readout by shift register
(data rate: 1 MHz max.)
10
Low dark current due to zero-bias photodiode operation
0
10 100 Internal clamp circuit achieves low noise and wide
Pulse peak output (W) dynamic range.
KMPDB0392EA
Internal timing generator allows operation with two
(c) Measurement distance: 4 m types of input pulses (reset and clock).
50 X-ray detection type available with a phosphor screen
White object (reflectance 90%) attached to the photosensitive area
40
Usable with wide variety of photodiode specifications
30
(custom order product)
20
6-2 Structure
10
As shown in Figure 6-1, a photodiode array with amplifier
consists of two chips: a Si photodiode array chip for light
0
10 100 detection and a CMOS signal processing IC chip.
Pulse peak output (W)
KMPDB0393EA
52
[Figure 6-1] Structure diagram (S11865 series)
CMOS signal processing IC chip
6-3 Operating principle
The CMOS signal processing IC chip consists of a timing
generator, shift register, hold circuit, and charge amplifier
array [Figure 6-2]. Each pixel of the photodiode array is
connected by wire bonding to the charge amplifier in
the CMOS signal processing IC chip. The light-generated
Si photodiode array chip Board charge (Qout) in a photodiode, which is expressed by the
KMPDC0186EA product of the photocurrent (IL) and the integration time
(Ts), is converted into an output voltage (Vout = Qout/Cf )
Signals from 64 pixels or 128 pixels at a time are handled
by the charge amplifier feedback capacitance (Cf ). The
by the CMOS signal processing IC. This makes the entire
output signal, which is sent to the hold circuit before the
system configuration very simple compared to methods
charge amplifier is reset, is read out by the shift register as
that connect each pixel on the photodiode array to an
time-series voltage signals. In the S11865/S11866 series,
external signal processing circuit.
signals from all pixels are read out by the simultaneous
Sensor devices with a phosphor screen attached to the
integration method. The S11865/S11866 series also have
photosensitive area are intended for X-ray detection.
a shutter function capable of adjusting the integration
When X-rays irradiate a sensor device, the phosphor
time. The video data rate is 1 MHz maximum.
screen converts the X-rays into visible light which is then
detected by the photodiode array. These devices are used [Figure 6-2] Block diagram (S11865-64/-128, S11866-64-02/-128-02)
as line sensors for X-ray non-destructive inspection tasks
requiring long X-ray detectors.
Timing generator
Shift register
Hold circuit
Charge amplifier array
Photodiode array
KMPDC0153EA
[Table 6-1] Hamamatsu photodiode arrays with amplifiers
Pixel height Pixel pitch Line rate

Type no. Number of pixels
(mm) (mm) (lines/s)
S11865-64 0.8 0.8 64 14678

S11865-128 0.6 0.4 128 7568
S11865-256 0.3 0.2 256 3844
S11866-64-02 1.6 1.6 64 14678
S11866-128-02 0.8 0.8 128 7568

Note: We also offer X-ray detection types with a phosphor screen attached directly to their photosensitive area.
53
Besides fabricating image sensors with a long detection
6-4 Characteristics length, photodiode arrays with amplifiers also allow
downsizing of detection systems. Using our standard
Figure 6-3 shows the spectral response of the S11865/ CMOS signal processing IC chips makes it easy to create
S11866 series. custom image sensors by just changing the Si photodiode
array chips or circuit boards to a desired shape.
[Figure 6-3] Spectral response
(S11865/S11866 series, typical example) [Figure 6-5] Long and narrow image sensor consisting of
multiple photodiode arrays with amplifiers
Wavelength (nm)
KMPDB0220EA
6-5 How to use

The S11865/S11866 series are designed so that the effective
area length equals the board length. This allows easily
configuring a long and narrow, one-dimensional detector
system by arranging multiple sensors in a row. This type
of system is difficult to create with normal image sensors.
When multiple sensors are arrayed, the output signal from
each sensor can be read out in parallel using multiple
external circuits or read out serially using a single circuit.
To read out signals serially, set the external voltage so that
the preceding-stage end-of-scan signal is used as the next-
stage start signal.
[Figure 6-4] Timing chart (S11865/S11866 series)

1 2 3 4 5 14 15 16 17 18 19 20 1 2 3
CLK
tpw(RESET1)
RESET
tpw(RESET2)
8 clocks 8 clocks
Integration time
Video output period
Video 1 2 n-1 n
Trig
EOS
KMPDC0294EB
54
7. InGaAs 7-2 Structure
linear image sensors InGaAs linear image sensors consist of an InGaAs photodiode
array and a CMOS IC (ROIC) including a charge amplifier
InGaAs linear image sensors are designed specifically for array, sample-and-hold circuit, shift register, readout
near infrared detection. These image sensors successfully circuit, and timing generator. The InGaAs photodiode array
minimize adverse effects from dark current by driving is connected to the CMOS IC by wire bonding or via bumps.
the InGaAs photodiode array at zero bias, and they Available packages include a ceramic package for room
deliver a wide dynamic range in the near infrared region. temperature operation and a metal package with a built-
in thermoelectric cooler, which are selectable according to
application. A typical block diagram for TE-cooled InGaAs
7-1 Features linear image sensors is shown in Figure 7-1. An analog
video output (Video) and digital outputs (AD_trig, AD_sp)
Wide dynamic range for sample-and-hold can be obtained by supplying analog
inputs of +5 V (Vdd), GND (Vss), a charge amplifier reset
Low dark current due to zero bias operation voltage (INP), pixel voltage (PDN), and readout circuit
reset voltages (Vinp, Vref, Fvref ), as well as digital inputs of
Wide spectral response range
master clock pulse (CLK) and integration time control pulse
High gain due to charge amplifier (Reset).
Low noise due to CDS circuit
Internal saturation control circuit
Internal timing generator allows simple operation.
Low crosstalk
Selectable gain
Hybrid type: includes a back-illuminated chip for

different spectral response ranges
[Table 7-1] Hamamatsu InGaAs linear image sensors
Spectral response range Pixel size

Illumination type Type Number of pixels Package
(m) [m (H) m (V)]
128 50 250 Ceramic
Standard type 256 50 250, 50 500 Ceramic,
512 25 250, 25 500 Metal
0.9 to 1.7
256 50 50
High-speed type 512 25 25 Ceramic
1024 25 25, 25 100
Front-illuminated 256 50 250
type 0.9 to 1.85
512 25 250
0.9 to 2.05 256 50 250
Long wavelength 256 50 250
0.9 to 2.15 Metal
type 512 25 250
0.9 to 2.25 256 50 250
256 50 250
0.9 to 2.55
512 25 250
256 50 50, 50 500 Ceramic,
Back-illuminated Standard type 0.95 to 1.7
512 25 25, 25 500 Metal
type
Hybrid type 0.95 to 2.15 512 (256 + 256) 25 250 Metal
55
[Figure 7-1] Block diagram (TE-cooled type) becomes saturated yet the charge integration still continues,
CLK Reset Vdd Vss INP PDN Vinp Fvref Cf_select
then the charges that are no longer stored in the charge
amplifiers feedback capacitance will flow out to the adjacent
pixels, degrading the purity of the signal output (this is
Timing generator Bias generator
known as blooming). To avoid this blooming, each pixel
AD_sp
Shift register has a circuit for stopping the charge integration by sensing
AD_trig
Address switch Readout circuit Video whether the charge amplifiers feedback capacitance is
saturated.
Charge amp + sample-and-hold circuit CMOS IC
To extract continuous signals, the integration capacitance
of the charge amplifier must be reset. A drawback of this,
however, is that a large reset noise occurs. This reset noise
InGaAs photodiode array
must be removed to make measurements with high accuracy.
In the CDS circuit for InGaAs linear image sensors, the
TE+ TE- Temperature monitor integration start output is held in the signal processing
KMIRC0033ED
circuit immediately after reset and the integration end
output is then held to obtain the difference between the
7-3 Operating principle two outputs to eliminate the reset switching noise.
Incidentally, high-speed type image sensor circuitry gives
priority to high-speed readout while standard type image
In the CMOS IC for InGaAs linear image sensors, a charge
sensor circuitry gives priority to a wide dynamic range.
amplifier and sample-and-hold circuit array is formed
To achieve even higher speeds, the multi-port types employ
and connected one-to-one to each pixel on the InGaAs
a multiport readout format that reads out the data in parallel
photodiode array. Figure 7-2 shows an equivalent circuit
by dividing the pixels into multiple ports.
for one pixel.
[Figure 7-3] Multi-port example (8-port)
[Figure 7-2] Equivalent circuit (for one pixel)
AD_st_odd
Reset
Vout+1 Vout-1 Vout+7 Vout-7
CLK
Timing
generator
Sample-and- Reset
Readout circuit
hold circuit
Vref Vdd
Vss
INP INP
KMIRC0034EA
Vref1 to 3
When light enters the photodiodes of an InGaAs linear Cfa to c

image sensor, electric charges are generated and flow
into the feedback capacitance of the charge amplifier.
Timing
This differential-input charge amplifier can operate generator
photodiodes at nearly zero bias, which suppresses the dark

current. In actual operation, however, the amplifier has an Vout+2 Vout-2 Vout+8 Vout-8
Port AD_st_even
offset voltage and the supply voltage to each pixel varies by KMIRC0039EA
several millivolts, so the video output of some pixels goes

positive versus the reset level, while that of other pixels
goes negative. Therefore, when no light is incident on the 7-4 Characteristics
image sensor, extending the integration time increases the
output on both the positive and negative sides. However,
this output caused by the dark current is a fixed pattern,
Input/output characteristics
so only the output signal resulting from light input can be
extracted by subtracting the output due to the dark current The relation between the light level incident to the image
from the output signal obtained with light incident on the sensor and the signal output is referred to as the input/
image sensor. Since InGaAs photodiodes are made from output characteristics. Since InGaAs linear image sensors
a compound semiconductor, there are lattice defects and operate in charge amplifier mode, the incident light
the dark current has relatively large absolute values and exposure (unit: J) is expressed by the product of light
variations compared to Si photodiodes. The maximum level (unit: W) and integration time (unit: s).
integration time (integration time needed to reach saturation The output from an InGaAs linear image sensor is
due to dark current) of InGaAs photodiodes therefore represented in voltage. Figure 7-4 shows a schematic
varies among the pixels. If a pixel with high dark current graph of input/output characteristics. The slope in the
56 figure can be expressed by equation (1).
[Figure 7-4] Schematic graph of input/output characteristics [Figure 7-5] Linearity error
(log graph)
(a) G9204-512S
(Typ. Ta=25 C, Vdd=5 V, INP=4.5 V, Vref=1.26 V, CE=16 nV/e, f=250 kHz)
Saturation voltage 30
20
Output voltage (V)
y = ax + b
Saturation output voltage

10
Linearity error (%)

0
-10
-20
Dark output
-30
Saturation exposure
-40
Exposure (J) 1 10 100 1000 10000
KMIRC0019EB
Output voltage (mV)
y = ax + b ............ (1)
KMIRB0081EA
y: output voltage (b) G12230-512WB (c=1.7 m chip)

a: sensitivity (ratio of output with respect to exposure)
x: exposure (Typ. Ta=25 C, Vdd=5 V, INP=PDN=Vinp=4 V, Fvref=1.2 V, CE=160 nV/e, f=5 MHz)
30
: slope coefficient
b: dark output (output when exposure=0)
20
Since the upper limit of the output voltage is determined 10

Linearity error (%)
by the output voltage range of the charge amplifier, the

0
input/output characteristics will have an inflection point
even if the incident light exposure is increased linearly. The -10
incident light exposure at this inflection point is referred
to as the saturation exposure, the output voltage as the -20
saturation output voltage, and the amount of charge stored

-30
in the charge amplifier as the saturation charge.
-40
In our InGaAs linear image sensor datasheets, the saturation 1 10 100 1000 10000
output voltage (Vsat) is defined as the saturated output
Output voltage (mV)
voltage from light input minus the dark output. Under the
KMIRB0082EA
condition that no light is incident on the sensor (dark state),
extending the integration time causes the output to increase
on the positive and negative sides. The voltage going negative Spectral response
becomes saturated at approx. 0.5 V.
The saturation charge is calculated from the equation Q = C V When light energy incident on the photosensitive area
based on the saturation output voltage. If the integration formed with a PN junction is greater than the InGaAs band
capacitance (Cf ) is 10 pF and the saturation output voltage gap energy, the electrons in the valence band are excited
is 3.2 V, then the saturation charge will be 32 pC. into the conduction band, generating electron/hole pairs.
This generated charge diffuses toward the photodiode
depletion layer where the electric field accelerates the
Linearity error
charge to pass through the PN junction, resulting in a signal
for readout. If the light energy is smaller than the band gap
The slope coefficient () of input/output characteristics
energy, it cannot be detected. The cutoff wavelength (c) is
shown in the preceding section corresponds to the slope
given by equation (2).
plotted on the logarithmic graph. This value is 1, but during
actual measurement, the input/output characteristics will 1.24
c = [ m] ............ (2)
slightly deviate from this. This deviation is known as the Eg
linearity error and is expressed in percentage. Eg: band gap energy [eV]
Figure 7-5 shows the linearity error obtained by random
sampling. The linearity error of the G9204-512S at 95% or The band gap energy for InxGa1-xAs (x=0.53) is 0.73 eV at
below the saturation exposure is within 3%, which is quite room temperature, so the cutoff wavelength will be 1.7
small. The G12230-512WB is an improved type designed to m. On the long wavelength type InxGa1-xAs (x=0.82),
reduce the linearity error at high gain (CE=160 nV/e-). the band gap energy is 0.48 eV at room temperatures, so
the cutoff wavelength will be 2.6 m.
57
The light absorption coefficient for InGaAs differs
depending on the light wavelength. The longer the light
wavelength, the smaller the light absorption coefficient,
InGaAs linear image sensors contain a large number of
and near the cutoff wavelength it decreases abruptly.
InGaAs photodiodes arranged in an array, yet sensitivity of
The incident light at longer wavelengths penetrates
each photodiode (pixel) is not uniform. This may result from
deeper into the InGaAs substrate, generating carriers
crystal defects in the InGaAs substrate and/or variations in
in deep positions within it. Since these carriers have
the processing and diffusion in the manufacturing process
a limited life, they can only diffuse a certain distance
as well as inconsistencies in the CMOS charge amplifier
(diffusion length) after being generated. This means that,
arrays. For our InGaAs linear image sensors, variations in
even when the same amount of light enters the InGaAs
the outputs from all pixels measured when the effective
linear image sensor, the probability that the generated
photosensitive area of each photodiode is uniformly
carriers can reach the depletion layer and eventually be
illuminated are referred to as photoresponse nonuniformity
detected as an output signal depends on the wavelength.
(PRNU) and defined as shown in equation (3).
Moreover, how the incident light undergoes interference,
reflection, and absorption on the surface passivation film PRNU = (X/X) 100 [%] ............ (3)
of the photodiode (such as the insulation film) depends
X : average output of all pixels
on the wavelength and affects the sensitivity.
X: absolute value of the difference between the average output
Figures 7-6 and 7-7 show examples of spectral response. X and the output of the maximum (or minimum) output pixel
The spectral response varies with the temperature. This is

because the band gap energy is temperature-dependent. In our outgoing product inspection for photoresponse
The InGaAs band gap energy increases as the temperature nonuniformity, the output is adjusted to approx. 50% of
drops, causing the peak sensitivity wavelength and cutoff the saturation output voltage and a halogen lamp is used
wavelength to shift to the short wavelength side. as the light source. Since InGaAs linear image sensors use
a compound semiconductor crystal for light detection, the
[Figure 7-6] Spectral response photodiode array may contain crystal defects, resulting in
1.5
(Typ.) abnormal output signals from some of the pixels (defect
Td=25 C
G9206-256W G9207-256W pixels). The photoresponse nonuniformity specification is
Td=-10 C
Td=-20 C G9208-256W 5% to 20%.
G9205-256W
G9201 to G9204/ Scratches and dust on the light input window may also
G9211 to G9214/
1.0 G9494 series cause the sensitivity uniformity to deteriorate. So caution
should be exercised on this point when handling image
G11135/G11620
series sensors. Figure 7-8 shows typical examples of photoresponse
nonuniformity. These data were obtained by random
0.5
sampling.
[Figure 7-8] Photoresponse nonuniformity

(typical example)
0
0.5 1.0 1.5 2.0 2.5 3.0 (a) G9204-512S
Wavelength (m) (Td=25 C, INP=4.5 V, CE=16 nV/e, integration time 13.544 ms, f=250 kHz)
3500
KMIRB0068EB
Output value after subtracting dark output
3000
[Figure 7-7] Spectral response (G12230-512WB)
2500
Output voltage (mV)
(Typ.)
1.5
Td=25 C
Td=-20 C 2000
1500
1.0
1000
500
0.5 0
0 100 200 300 400 500 600
Pixel
KMIRB0026EC
0
0.8 1.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
Wavelength (m)
KMIRB0065EB
58
(b) G9207-256W Figure 7-9 shows the temperature characteristics of the
(Td=-20 C, INP=4.5 V, CE=16 nV/e, integration time 3.736 ms, f=250 kHz) G9204- 512S dark current (random sampling).
3500
Output value after subtracting dark output
[Figure 7-9] Dark current vs. element temperature
3000
(G9204-512S, typical example)
2500
Output voltage (mV)
2000
1500
Dark current (A)

1000
500
0
0 50 100 150 200 250 300
Pixel
KMIRB0027EC
Dark output Element temperature (C)

KMIRB0028EA
The dark output is the output generated even when no

incident light is present. This output is caused by the dark Noise
current (sum of diffusion current, recombination current,
and surface leakage current) of the photodiode, which InGaAs linear image sensor noise can be largely divided
flows to charge the charge amplifier and is converted into into fixed pattern noise and random noise.
a voltage output. Since the upper limit of the output is Fixed pattern noise includes photodiode dark current which
determined by the saturation output voltage, a large dark is current noise from the DC component. The magnitude of
output narrows the dynamic range of the output signal. The the fixed pattern noise is constant even if readout conditions
output signal is the sum of the output generated by light are changed, so it can be canceled by using an external
and the dark output, so the purity of the output signal can signal processing circuit.
be improved by using signal processing to subtract the dark Random noise, on the other hand, results from fluctuations
output from each pixel. in voltage, current, or charge that are caused in the signal
The dark output is given by equation (4). The integration output process in the sensor. When the fixed pattern noise
time must be determined by taking the magnitude of the has been canceled by external signal processing, the random
dark output into account. noise will then determine the InGaAs linear image sensors
lower detection limit or lower limit of dynamic range.
Vd = I D (Ts/Cf) + Voff ............ (4)
Random noise includes the following four components:
Vd : dark output [V]
ID : dark current [pA]
Dark current shot noise (Nd)
Ts : integration time [s] Signal current shot noise at light input (Ns)
Cf : integration capacitance [pF]
Voff: ROIC offset voltage [V] Charge amplifier reset noise (Nr)
CMOS charge amplifier readout noise (NR)
When rewriting the above equation in terms of integration In general, the reset noise that occurs when the charge
time (Ts) by substituting the saturation output voltage amplifier is reset is predominant. This reset noise can be
(Vsat) for the dark output (Vd), the maximum integration greatly reduced with a CDS circuit, so the dark current
time (Tsmax) is expressed by equation (5). shot noise and CMOS charge amplifier readout noise
become significant sources of noise. Dark current
Tsmax = Cf Vsat/ID ............ (5)
shot noise results from erratic generation of the output
charge due to dark current. This noise becomes larger
The band gap widens as the temperature decreases, so the
as the output charge due to dark current increases, and
number of carriers thermally excited into the valence band
therefore varies depending on operating conditions such
from the conduction band decreases, causing the dark
as integration time and temperature.
current to reduce exponentially with the temperature. In our
Signal current shot noise is caused by fluctuations due
InGaAs linear image sensors, the temperature coefficient
to incident photons arriving randomly at the sensor.
of the dark current is 1.06 to 1.1. If the dark current at
The total noise (N) is expressed by equation (7).
temperature T1 (unit: C) is IDT1 (unit: A), then the dark
current IDT at temperature T is given by equation (6). N= Nd2 + Ns2 + NR2 ...................... (7)
IDT = IDT1 (T - T1) [A] ............ (6)
59
The dark current shot noise (Nd) and signal current shot (b) G9207-256W
noise at light input (Ns) can be expressed as the root (Td=-20 C, INP=4.5 V, CE=16 nV/e, integration time 0.736 ms, f=250 kHz, integration count 50)
0.50
square of the generated charge by representing them as an
0.45
equivalent input noise charge which is a value converted
0.40
to a charge quantity for input to the image sensor.
Noise output (mV rms)

0.35
2ID
Nd = TS [e- rms] ............ (8) 0.30
q
0.25
2IS 0.20
Ns = TS [e- rms] ............ (9)
q 0.15
0.10
Is: signal current by light input
q : electron charge 0.05
0
0 50 100 150 200 250 300
We specify the noise level in InGaAs linear image sensors
as fluctuations in the output voltage of each pixel by Pixel
using root-mean-square noise voltage ( V rms) units. KMIRB0030EB
Converting equations (8) and (9) into voltage therefore

gives equations (10) and (11), respectively.
7-5 How to use
Ts
Nd = 2q ID [V rms] ............ (10)
Cf
This section explains how to use and operate InGaAs linear
Ts image sensors including handling precautions and setting
Ns = 2q IS [V rms] ............ (11)
Cf operating conditions.
When the CMOS charge amplifier readout noise (NR) is

Setups
measured at 25 C using an integration time of 1 ms, a
data rate of 250 kHz, and an integration count of 50, the
There are two types of InGaAs linear image sensors: a
standard deviation in the G9201 series is calculated to
thermoelectrically cooled type that contains a thermoelectric
be 180 V rms at Cf=10 pF and 400 V rms at Cf=0.5 pF.
cooler and thermistor, and a non-cooled type. Both types
The noise levels listed for the G9201/G9211 series in our
basically operate with the same drive method except for
datasheet are measured at 25 C using an integration
cooling operation.
time of 10 to 20 ms, data rate of 250 kHz, and integration
count of 50 in order to calculate the standard deviation. (1) Terminal description
In the G9205 to G9208 series, this measurement is made Make connections by referring to Figures 7-11 and 7-12
at -20 C using an integration time of 1 ms or less, data and to Table 7-2.
rate of 250 kHz, and integration count of 50. Figure 7-10
shows output noise fluctuations measured with the [Figure 7-11] Pin connections (G9201 to G9204 series, top view)
G9204-512S and G9207-256W with no incident light. 256 pixels 512 pixels
These data were obtained by random sampling.
[Figure 7-10] Noise output fluctuations (typical example) Cf select Cf select

Reset Reset_even Reset_odd
TE+ TE- TE+ TE-
(a) G9204-512S AD_trig AD_trig_even AD_trig_odd
Therm Vdd Therm Vdd
Therm Vss Therm Vss
(Td=25 C, INP=4.5 V, CE=16 nV/e, integration time 13.6 ms, f=250 kHz, integration count 50) Case INP Case INP
0.25 CLK CLK_even CLK_odd
Vref Vref
0.20 Video Video_even Video_odd

Noise output (mV rms)
0.15
KMIRC0013EA
0.10
0.05
0
0 100 200 300 400 500 600
Pixel
KMIRB0029EB
60
[Figure 7-12] Setup and wiring (G9201 to G9204 series) [Figure 7-13] Temperature characteristics of one-stage
thermoelectrically cooled device
TE+ (Typ. heatsink 0.5 C/W)
4 70
TE- Temperature
Voltage
Thermoelectric cooler Therm controller circuit
Temperature difference 60
Therm
Temperature difference (C)

3 50
Vdd 40
Voltage (V)
INP
2 30
Vref Power supply
Vss 20
Cf select
1 10
CLK
Timing signal generator 0
Reset
0 -10
0 1 2
AD_trig
Signal processing circuit
Video Current (A)
KMIRB0031EC
InGaAs linear image sensor
KMIRC0035EA
[Figure 7-14] Temperature characteristics of two-stage
(2) Heatsink thermoelectrically cooled device
(Typ. heatsink 0.4 C/W)
Selecting a heatsink 7 60
Voltage
Temperature difference
When cooling a one-stage thermoelectrically cooled device 6 50
Temperature difference (C)

to -10 C, select a heatsink of 0.5 C/W or less including a
5 40
safety margin. When cooling a two-stage thermoelectrically
Voltage (V)
cooled device to -20 C, select a heatsink of 0.4 C/W or less. 4 30

Equipment should be carefully designed so that the
3 20
heatsink is not placed where heat builds up. Provide good
air ventilation to allow heat emitted from the heatsink to 2 10
sufficiently dissipate by installing air fans and ventilation
ducts. Note that the heatsink thermal resistance varies 1 0
according to forced air cooling. 0 -10

0 1 2 3
Current (A)
KMIRB0032EC
[Table 7-2] Terminal function and recommended connection (G9201 to G9204 series)
Terminal name Input/output Function and recommended connection

CLK Input (CMOS logic) Clock pulse for operating the CMOS shift register
Reset pulse for initializing the feedback capacitance in the charge amplifier formed on
Reset Input (CMOS logic)
the CMOS chip. The width of the reset pulse is the integration time.
Vdd Input Supply voltage for operating the signal processing circuit on the CMOS chip
Vss Ground for the signal processing circuit on the CMOS chip
INP Input Reset voltage for the charge amplifier array on the CMOS chip
Cf select Input Voltage that determines the feedback capacitance (Cf) on the CMOS chip
Case This terminal is connected to the package.
Therm Thermistor terminal for monitoring temperature inside the package
TE+, TE- Power supply terminal for the thermoelectric cooler for cooling the photodiode array
AD_trig Output Digital signal for A/D conversion; positive polarity
Video Output Analog video signal; positive polarity
Vref Input Reset voltage for the offset compensation circuit on the CMOS chip
61
Heatsink mounting method
Drive method
To allow the thermoelectric cooler to exhibit fullest
cooling capacity, the heatsink must be mounted correctly Sensor operation should be checked in a dark state.
onto the sensor package. Mount the heatsink while Block the light falling on the photosensitive area before
taking the following precautions. checking operation.
Check that the heatsink attachment surface and the (1) Turning on power to the driver circuit
heat-dissipating surface of the InGaAs image sensor
First check the voltages (Vdd, INP, Vref, etc.) supplied
package are clean and flat.
to the sensor, and then turn the power on. At this point,
Mount the heatsink so that it makes tight contact with
also check that the current values are correct. If excessive
the entire heat-dissipating surface of the package.
current is flowing, the power supply line might be
The heat-dissipating surface area should be large to
shorted so immediately turn off the power and check the
improve the cooling efficiency and prevent possible
power supply line.
damage.
Apply a thin coat of heat-conductive grease uniformly (2) Inputting control signals from the pulse generator
over the attachment surface in order to lower thermal While referring to the timing chart shown in Figure 7-16,
resistance between the package heat-dissipating input the control signals from the pulse generator to the
surface and the heatsink. Fasten the sensor package InGaAs linear image sensor (G9201/G11135 series). Two
to the heatsink with screws using equal force so that control signals (CLK and Reset) are input to the image
the grease spreads more uniformly. When a mica sheet sensor and must be H-CMOS level inputs. The image
is used, it must also make contact with the entire sensor may malfunction if other control signal levels
heat-dissipating surface of the package. The cooling are used. In the G9201/G9494 series, set the Reset signal
efficiency will degrade if the sensor package is fastened pulse width to at least 6 s. The CLK signal frequency
to the heatsink with screws while the mica sheet is still determines the video signal readout frequency, and the
too small to cover the screw positions. This may also Reset pulse interval determines the integration time.
warp the package base, causing cracks between the Normal operation is performed whether the CLK
sensor and the package base [Figure 7-15 (a)]. and Reset signals are synchronized or not. When the
Do not press on the upper side of the package when Reset pulse rising edge is synchronized with the CLK
fastening the sensor package to the heatsink or printed pulse falling edge, the integration starts at the falling
circuit board. If stress is applied to the glass faceplate, edge of the CLK pulse following the Reset pulse rising
this may cause the faceplate to come off or may impair edge. When not synchronized, the integration starts
airtightness of the package [Figure 7-15 (b)]. at the falling edge of the second CLK pulse from the
Reset pulse rising edge. When the Reset pulse falling
[Figure 7-15] Sensor mounting method
edge is synchronized with the CLK pulse falling edge,
(a) Example 1 the integration ends with the falling edge of the CLK
pulse following the Reset pulse falling edge. If not
synchronized, the integration ends with the falling edge
Screws
Base of the second CLK pulse from the Reset pulse falling
(package heat-dissipating surface)
Sensor Sensor edge.
Mica
(3) Setting the drive timing
Heatsinks
Example 1: When operating an InGaAs linear image sensor

KMIRC0023EA
G9201-256S at CLK frequency of 1 MHz

(b) Example 2 Since the video signal readout frequency is 1/8 of the
CLK signal frequency, the readout time (tr) per pixel is
Insertion 8 s. The time required for one scan (tscan) is therefore
given by equation (12).
Sensor Sensor tscan = (tc 14) + (tr N) ....................... (12)

Printed = 1 [s] 14 + 8 [s] 256
circuit board
= 2062 [s]
KMIRC0024EA
tc: CLK period

(3) Video signal monitoring N : number of pixels
The image sensor output end does not have drive capability,
In this case, the reset time [low period of Reset in
so in order to monitor the video signal, the sensor output
Figure 7-16 (a)] must be longer than the time required
should be amplified by a buffer amplifier and then fed to an
for one scan, so set the scan time (tscan) longer than
oscilloscope.
2062 s. Note that the scan time becomes slightly
62
[Figure 7-16] Timing chart
(a) G9201 series
tr(clk) tf(clk)
CLK
t1 t2
Blank period
tr(RES) tf(RES) (10 clocks or more)
Reset Integration time
Actual integration time
AD_trig
Reset Output
t3
Video
Readout time Readout time

Readout time for channel 1 for channel 2 for last channel
KACCC0224EB
Clock pulse frequency f 0.1 - 4 MHz

Clock pulse width tpw(clk) 100 - - ns
Clock pulse rise/fall times tr(clk), tf(clk) 0 20 100 ns
Reset pulse width tpw(RES) 6000 - - ns
Reset pulse rise/fall times tr(RES), tf(RES) 0 20 100 ns
Reset (rise) timing t1 50 - - ns
Reset (fall) timing t2 50 - - ns
Output settling time t3 - - 600 ns
(b) G11135 series

g ( )
6 CLK
4 CLK 8 s
CLK
Blank period
5 CLK (3 clocks or more)
Reset Integration time
5 CLK Actual integration time

AD_sp
AD_trig
n CLK
Video
1 2 n-1 n
Actual integration time = Integration time - 8 s + 1 CLK

tf(clk) tr(clk)
tpw(clk)
CLK
tr(RES) tf(RES)
Reset
tpw(RES)
KMIRC0050EC
63
Clock pulse frequency f 0.1 - 5 MHz

Clock pulse width tpw(clk) 60 100 - ns
Clock pulse rise/fall times tr(clk), tf(clk) 0 20 30 ns
High *1 - *2
Reset pulse
tpw(RES) clock
width Low Number of channels + 12 - -
Reset pulse rise/fall times tr(RES), tf(RES) 0 20 30 ns
*1: (8 s + 5 CLK) or (18 s - 1 CLK), the longer of the two
*2: 1.008 ms - 1 CLK
longer depending on the Reset pulse width and the (4) Turning on the power supply for the thermoelectric cooler
synchronization with the CLK signal.
Use extra caution to avoid damaging the image sensor
The G9201 series performs simultaneous integration
when turning on the power to the thermoelectric cooler.
of all pixels and sequential readout, so the line rate
Take the following precautions when designing a power
is calculated from the integration time (high period
supply circuit for the thermoelectric cooler.
of Reset) and the time required for one scan (tscan).
Since the maximum CLK frequency is 4 MHz and the Never exceed the absolute maximum ratings for the
minimum integration time is 6 s for the G9201 series, thermoelectric cooler.
the maximum line rate of the G9201-256S is expressed by Make sure that the power supply voltage and connection
equation (13). polarity are correct. Turning on the power supply with
the wrong voltage or polarity will damage the image
Maximum line rate = 1/(Integration time + tscan) .......... (13) sensor.
= 1/(6 [s] + 515.5 [s]) A power supply with the lowest possible noise and
= 1917 [lines/s] ripple should be used. Also, use power supply wires
thick enough to keep impedance as low as possible.
The longer the integration time, the lower the line rate. The TE+ and TE- wires in particular must be sufficiently
thick.
Example 2: When operating an InGaAs linear image sensor
Be sure to provide an over-current safeguard circuit to
G11135-512DE at a CLK frequency of 5 MHz protect the thermoelectric cooler from being damaged.
Since the video signal readout frequency equals the CLK Provide a protection circuit that monitors the temperature
signal frequency, the readout time (tr) per pixel is 0.2 s. on the heat-emitting side of the heatsink to prevent the
The time (tscan) required for one scan is therefore given heatsink temperature from exceeding the specified level
by equation (14). due to excessive cooling.
While referring to Figures 7-13 and 7-14, set the optimum
tscan = (tc 12) + (tr N) ....................... (14)
voltage and current values that maintain the target
= 0.2 [s] 12 + 0.2 [s] 512
temperature.
= 104.8 [s]
tc: CLK period

N : number of pixels
7-6 New approaches
The reset time must be longer than the scan time, so set
the reset time longer than 104.8 s. Note that the scan InGaAs image sensors are widely used for near infrared
time becomes slightly longer depending on the Reset spectrophotometry. When a sample is irradiated with
signals pulse width and the synchronization with the near infrared light, it reflects or transmits the light. A
CLK signal. characteristic spectrum is obtained by separating the
The G11135 series performs simultaneous integration reflecting or transmitting diffused light into individual
of all pixels and sequential readout, so the line rate is wavelengths. This spectrum contains multiple pieces of
calculated from the integration time (high period of information regarding the components of the sample,
Reset) and the time required for one scan (tscan). Since so measuring this spectrum allows for rapid, non-
the maximum CLK frequency is 5 MHz and the minimum destructive quantitative/qualitative analysis. Near infrared
integration time is 17.8 s for the G11135 series, the spectrometers have been downsized and are widely used
maximum line rate of the G11135-512DE is expressed by not only in laboratories but also on production lines or in
equation (15). outdoor locations.
Maximum line rate = 1/(integration time + tscan) .......... (15) This expanded field of applications means that infrared
= 1/(17.8 [s] + 104.8 [s]) image sensors will need higher cost performance. Meeting
= 8156 [lines/s] this requirement will require fabricating larger wafers and
smaller chips in order to yield a greater number of chips
The longer the integration time, the lower the line rate.
from one wafer.
64
To solve this challenging problem, back-illuminated
structures will help reduce the chip size as well as decrease 8. InGaAs
the wiring capacitance to achieve higher speeds. In
the future, the development of a narrow-pitch charge
area image sensors
amplifier array will help suppress output variations caused
by differences in the video line, which have long been a InGaAs area image sensors detect near infrared rays
problem with conventional devices. In addition, we will invisible to the human eye and convert the weak light
make InGaAs image sensors which improved the sensitivity into images. They have a hybrid structure consisting of a
in the short wavelength region, enabling light detection two-dimensional back-illuminated InGaAs photodiode
over a wide range from the visible region to the near array and high-gain low-noise CMOS readout circuit
infrared region. (ROIC: readout integrated circuit) that are connected by
In bumps. A pixel is made up of one InGaAs photodiode
element and one ROIC. The ROIC has a built-in timing
generator that makes it possible to produce analog
video outputs and AD_trig digital outputs with a simple
application of an external master clock (MCLK) and
master start pulse (MSP).
8-1 Features
Cutoff wavelength: 1.7 m or 1.9 m
High quantum efficiency
High sensitivity: 1 V/e- min.
Readout mode: Global shutter mode, rolling shutter

mode
Simple operation: built-in timing generator
Hermetic seal package: compact, high reliability
8-2 Structure
InGaAs photodiode
The two-dimensional back-illuminated InGaAs photodiode

array built into the InGaAs area image sensor provides high
quantum efficiency in the near infrared region [Figure 8-1].
We also offer a type with built-in thermoelectric cooler for
controlling the photodiode temperature.
[Table 8-1] Hamamatsu InGaAs area image sensors
Cutoff wavelength Number of Pixel pitch

Type ROIC Cooling Package
(m) pixels (m)
Wide dynamic range, compact 64 64 One-stage TO-8
1.7
50 CTIA TE-cooled
Wide dynamic range 128 128 28L metal
Long wavelength, compact 1.9 64 64 Two-stage TE-cooled TO-8
High resolution, compact 128 128 TO-8

1.7 20 SF Two-stage TE-cooled
High resolution (VGA) 640 512 28L metal
65
[Figure 8-1] Spectral response [Figure 8-2] Block diagram
1.2
(Typ. Td=25 C) (a) CTIA type
Reset switch
Cutoff wavelength
1.0 1.9 m
0.8 Cf
0.6
Sample switch
Source follower
0.4
Cutoff wavelength
1.7 m InGaAs photodiode
0.2 Hold capacitance
0
0.8 1.0 1.2 1.4 1.6 1.8 2.0
KMIRC0078EA
Wavelength (m)
KMIRB0084EA
(b) SF type
ROIC
The ROIC in the InGaAs area image sensor is manufactured InGaAs photodiode
to suit the characteristics of the InGaAs photodiode

using CMOS technology. Multi-functionality and high Sample switch
performance as well as cost reduction in constructing Source follower
systems are accomplished by mounting the analog circuit
for signal processing and the digital circuit for generating Reset switch
Parasitic capacitance
timing signals on a single chip.
The ROIC comes in two types: CTIA (capacitive trans-
impedance amplifier) and SF (source follower). The
proper ROIC type must be selected according to the KMIRC0079EA
application. Figure 8-2 shows block diagrams of each

There are two signal readout modes: global shutter mode
type.
and rolling shutter mode. In global shutter mode, all
The advantages of the CTIA type are that (1) the charge-
pixels are reset simultaneously, and integration of all
to-voltage converter takes on an amplifier structure and
pixels begin at the same time. Therefore, data integrated
(2) it has superior linearity since the voltage applied
over the same time duration is output from all pixels. In
to the InGaAs photodiodes can be kept constant. The
rolling shutter mode, a reset occurs every line, the data
disadvantages are that (1) the power consumption by
is output, and then integration starts immediately. If you
the amplifier is large and (2) temperature control using
want to prioritize the frame rate, select the rolling shutter
the TE-cooler is necessary because of the temperature
mode.
increase in the sensor caused by the large power
consumption. In addition, with the CTIA type, the pitch
is larger because the size of the amplifier is larger than In (indium) bumps
that of the SF type.
The SF type provides high gain and high resolution. Since In bumps electrically connect the InGaAs photodiode
the charge-to-voltage converter has a simple structure and ROIC. Since the Youngs modulus of In is lower than
and is small, the pitch can be made smaller and higher that of Au, Cu, and Al and the melting point is 157 C,
mounting density is possible. In addition, it does not distortion caused by heat can be suppressed. Thus, In is
require cooling due to its low power consumption. suitable for connecting metals and semiconductors with
Moreover, high sensitivity can be attained because the different thermal expansion coefficients.
parasitic capacitance can be reduced. The disadvantage
is its narrow linearity range.
66
[Figure 8-3] Schematic of InGaAs area image sensor In our InGaAs linear image sensor datasheets, the saturation
output voltage (Vsat) is defined as the saturated output
voltage from light input minus the dark output. The
saturation charge is calculated from the equation Q = C V
based on the saturation output voltage. If the integration
capacitance (Cf ) is 0.1 pF and the saturation output voltage
is 2.0 V, then the saturation charge will be 0.2 pC.
Back-illuminated
ROIC (Si) In bump InGaAs photodiode array
InGaAs area image sensors contain a large number of

Front end board
InGaAs photodiodes arranged in an array, yet sensitivity of
KMIRC0036EB each photodiode (pixel) is not uniform. This may result from
crystal defects in the InGaAs substrate and/or variations in
the processing and diffusion in the manufacturing process
8-3 Characteristics as well as inconsistencies in the ROIC gain. For our InGaAs
area image sensors, variations in the outputs from all pixels
Input/output characteristics measured when the effective photosensitive area of each
photodiode is uniformly illuminated are referred to as
The input/output characteristics express the relation photoresponse nonuniformity (PRNU) and defined as
between the light level incident to the image sensor shown in equation (2).
and the signal output. Since InGaAs area image sensors PRNU = (X/X) 100 [%] (2)
operate in charge amplifier mode, the incident light X : average output of all pixels
exposure (unit: J) is expressed by the product of light level X: absolute value of the difference between the average output X and the
output of the maximum (or minimum) output pixel
(unit: W) and integration time (unit: s). Figure 8-4 shows
a schematic diagram of the input/output characteristics. In our outgoing product inspection for photoresponse
The slope in the figure can be expressed by equation (1). nonuniformity, the output is adjusted to approx. 50% of
the saturation output voltage and a halogen lamp is used
[Figure 8-4] Schematic graph of input/output characteristics
(log graph) as the light source. Since InGaAs area image sensors use
a compound semiconductor crystal for photoelectric
conversion, the photodiode array may contain crystal defects,
Saturation voltage
resulting in abnormal output signals from some of the pixels
(defective pixels). Moreover, scratches and stain on the light
Output voltage (V)
y = ax + b
input window may also cause the sensitivity uniformity to
Saturation output voltage
deteriorate. So caution should be exercised on this point when

handling image sensors. Figure 8-5 shows typical example of
photoresponse nonuniformity (random sampling).
[Figure 8-5] Photoresponse nonuniformity

(G12242-0707W, typical example)
Dark output (Ta=25 C)
Saturation exposure 1.6
Exposure (J) 1.4

KMIRC0019EB 1.2
1.0
Voltage (V)
y = ax + b (1)

0.8
y: output voltage 0.6
a: sensitivity (ratio of output with respect to the exposure) 0.4
x: exposure 1
: slope coefficient 0.2
32
b: dark output (output when exposure=0) 0 64
1 Vertical pixels
32 96
64
Since the upper limit of the output voltage is determined Horizontal pixels 128
KMIRB0086EA
by the output voltage range of the ROIC, the input/output
characteristics will have an inflection point. The incident
light exposure at this inflection point is referred to as the
saturation exposure, the output voltage as the saturation
output voltage, and the amount of charge stored in the
charge amplifier as the saturation charge.
67
Dark output Noise
The dark output is the output generated even when no InGaAs area image sensor noise can be largely divided
incident light is present. This output is the sum of the dark into fixed pattern noise and random noise. Fixed pattern
current (sum of diffusion current, recombination current, noise includes ROIC DC offset voltage and photodiode dark
and surface leakage current) of the photodiode and the current which is current noise from the DC component.
ROIC offset voltage. Since the upper limit of the video The magnitude of the fixed pattern noise is constant even
output is determined by the saturation output voltage, if readout conditions are changed, so it can be canceled by
a large dark output narrows the dynamic range of the using an external signal processing circuit.
output signal. The output signal is the sum of the output Random noise, on the other hand, results from fluctuations
generated by light and the dark output, so the purity of the in voltage, current, or charge that are caused in the signal
output signal can be improved by using signal processing output process in the sensor. When the fixed pattern noise
to subtract the dark output from each pixel. The dark has been canceled by external signal processing, the random
output is given by equation (3). The integration time must noise will then determine the InGaAs area image sensors
be determined by taking the magnitude of the dark output detection limit for low-level light or lower limit of dynamic
into account. range. Random noise includes the following four components:
Vd = ID (Ts/Cf) + Voff (3) Dark current shot noise (Nd)

Vd : dark output [V] Signal current shot noise at light input (Ns)
ID : dark current [pA]
Ts : integration time [s]
Charge amplifier reset noise (Nr)
Cf : integration capacitance [pF] CMOS charge amplifier readout noise (NR)
Voff : ROIC offset voltage [V]
Dark current shot noise results from erratic fluctuation in
The band gap widens as the temperature decreases, so charge. This noise becomes larger as the output charge due
the number of carriers thermally excited into the valence to dark current increases, and therefore varies depending
band from the conduction band decreases, causing the on operating conditions such as integration time and
dark current to reduce exponentially. In our InGaAs area temperature. Signal current shot noise is caused by
image sensors, the temperature coefficient of the dark fluctuations due to incident photons arriving randomly at the
current is 1.06 to 1.1. If the dark current at temperature sensor. The total noise (N) is expressed by equation (5).
T1 (unit: C) is IDT1 (unit: A), then the dark current IDT at
temperature T is expressed by equation (4). N= Nd2 + Ns2 + Nr2 + NR2 ......... (5)
IDT = IDT1 (T - T1) [A] (4)

We specify the noise level (unit: V rms) in InGaAs area
image sensors as fluctuations in the output voltage of
Figure 8-6 shows the temperature characteristics of the
each pixel.
G12242-0707W dark current (random sampling).
[Figure 8-6] Temperature characteristics of dark current

(G12242-0707W) 8-4 New approaches
(Typ.)
10 pA
InGaAs area image sensors used for infrared detection
significantly increase the information obtained through
1 pA detection as compared to single-element InGaAs PIN
photodiodes or InGaAs linear image sensors. Its field of
Dark current
application is expanding including night vision for security

100 fA purposes, plastic sorting (pet bottles and other disposables),
farm produce sorting (e.g., grains), semiconductor analyzers,
and academic research (astronomy and satellite). Night
10 fA
vision and semiconductor analyzers require detection of low
level emissions from the object with high reproducibility.
1 fA
This requires high sensitivity and high resolution, which
-20 -10 0 10 20 30 40 50 60 70 must be accomplished by increasing the conversion gain and
Element temperature (C) decreasing the pitch at the same time. Hamamatsu is working
KMIRB0085EA to optimize the ROIC and developing fine pitch bump
bonding. Sorters use different wavelengths depending on
the object. Long wavelength is used in farm produce sorting,
and high-speed processing is also required. Hamamatsu will
increase the area of long wavelength type InGaAs area image
sensors and achieve high-speed processing by increasing the
speed of buffer amplifiers and using multiple ports.
68
[Figure 9-2] Sensitivity comparison between various
9. Applications types of analytical technique
100% 1000 ppm 1 ppm 1 ppb
9-1 DNA sequencers Gravimetric method,

volumetric method
Here we introduce the DNA sequencer, which is one typical

Atomic absorption method
application for back-thinned CCDs. The DNA sequencer
uses laser light to excite DNA fragments labeled with
ICP AES equipment
fluorescent dye and separated according to molecular
KMPDC0205EA
weight, and then sequentially reads out their base sequence
by detecting the resulting fluorescent light.
The DNA sequencer must be able to detect the faint
fluorescent light representing the four DNA bases (adenine,
9-3 Optical emission spectrometers
thymine, cytosine, and guanine) with good accuracy, and it
utilizes back-thinned CCDs having high quantum efficiency The optical emission spectrometer (OES) applies energy
close to the detection limit in the visible light range. to samples in various ways and performs spectroscopic
analysis on the emitted light to qualitatively and
[Figure 9-1] Configuration of DNA sequencer quantitatively analyze the sample components. The
emitted light spectrum varies depending on the elements
Mirror
Laser contained in the sample and spans from the ultraviolet
region to the near infrared region. Hamamatsu back-
Lens
thinned CCD linear image sensors feature high UV
sensitivity suitable for OES.
9-4 Spectrophotometers
The S9971 series of front-illuminated CCD with an internal
Back-thinned
CCD area Sample thermoelectric cooler is used in spectrophotometry over a
image sensor
broad spectral range (400 to 1200 nm). The S9971 series is
Autosampler
low cost compared to back-thinned CCDs with the same
KMPDC0204EA
photosensitive area size, making it suitable as a sensor in

compact spectrophotometers.
9-2 ICP AES equipment Back-thinned CCD area image sensors with high sensitivity
from the ultraviolet region to near infrared region can
be used as linear image sensors of vertically long pixels
The ICP AES (inductively coupled plasma atomic emission
during binning. Since these sensors are highly sensitive
spectroscopy) equipment performs qualitative and
over a wide spectral range and have a sensor structure
quantitative analysis of trace metals within a liquid. The
suitable for spectrophotometers, they are used in a variety
demand for ICP AES equipment is increasing rapidly as an
of spectrophotometers that require highly accurate
effective means of making environmental measurements.
measurements such as LED testers and light interference
ICP AES equipment requires high detection sensitivity and
type thickness meters for semiconductor process
accuracy, and therefore uses back-thinned CCDs with low
monitoring.
noise and high sensitivity. The spectral range to be detected
We also provide PMA (photonic multichannel analyzer)
partly includes ultraviolet light, so our back-thinned CCDs
and mini-spectrometers that incorporate our CCD image
with high ultraviolet sensitivity are found beneficial in
sensors or InGaAs linear image sensors.
enhancing the sensitivity of ICP AES equipment.
9-5 Grain sorters

Grain sorters irradiate light onto the falling grain, identify
unwanted items from the reflected or transmitted light,
and then remove those from the grain by high-pressure
air spray. Using InGaAs linear image sensors in the grain
sorter allows simultaneously identifying multiple grain
types while analyzing grain components.
69
[Figure 9-3] Schematic of grain sorter [Figure 9-5] Security (tailgate detection)
Distance area image sensor
Top view
Approx.
2 to 3 m
InGaAs linear
image sensor
Approx. 2 m
Approx. 2 m
KMPDC0474EA
Spray nozzle
[Figure 9-6] Obstacle detection by automobiles

Unwanted
grains Good grains
KMIRC0037EA
9-6 Optical channel monitors Approx. 5 to 10 m

Approx. 2 m
Devices called optical channel monitors (OCM) fulfill

an important role in monitoring the signal wavelength
and power on wavelength division multiplexing (WDM)
networks that carry huge quantities of information.
InGaAs linear image sensors used in the OCM detect
Distance linear image sensor
light spatially separated by spectral-dispersion elements. KMPDC0475EA
[Figure 9-4] Schematic of optical channel monitor

Output
Input
[Figure 9-7] Shape recognitions by robot
Relay station
Add/drop device
Distance area
image sensor
Optical tap
Feedback
Optical signal
diagnosis
Approx. 2 m
KMIRC0038EA
9-7 Security and room access control,

obstacle detection, and shape recognition
Approx. 50 cm
Approx. 50 cm
The light from a light source is incident on the subject, KMPDC0478EA
and a distance image sensor detects the reflected light.

The distance to the subject is then calculated using the
TOF (time-of-flight) method. The distance measurement
system combined with image processing is a promising
device for use in security and room access control (tailgate
detection), obstacle detection (unmanned vehicles, robots,
etc.), shape recognition (logistics, robots, etc.), and motion
capture.
70
[Figure 9-9] Andromeda galaxy captured by the Subaru
9-8 Detector for prime focus Telescope
camera of Subaru Telescope

The near infrared-enhanced back-illuminated CCD
(fully-depleted CCD with thick silicon) is being used
as a detector for the prime focus camera of the Subaru
Telescope (National Astronomical Observatory of Japan)
that is installed on the top of Mauna Kea on the island of
Hawaii. This CCD featuring high quantum efficiency at
1000 nm is expected to contribute greatly in the leading
field of observational astronomy, such as in the research
of dark energy and distant space (e.g., discovery of the
first heavenly body created in the universe).
In 2008, Suprime-Cam containing ten of our CCDs
(S10892-01) began its operation. In 2013, observation
using Hyper Suprime-Cam began. It uses 116 CCDs
(S10892-02) with improved sensitivity in the blue region.
This CCD not only has high sensitivity from the visible
region to the 1000 nm near infrared region but also in
the soft X-ray region (up to 20 keV). Due to its superior
performance, this CCD has already be chosen to be used 9-9 Asteroid explorer Hayabusa
in Japans X-ray astronomical satellite ASTRO-H.
Hamamatsu CCD area image sensor was chosen as the
[Figure 9-8] Subaru Telescope
detector of the Hayabusas fluorescent X-ray spectrometer
Prime focus
(a device for investigating materials on the surface of the
asteroid from above). Elements in matter on the ground
emit fluorescent X-rays of a certain wavelength in response
to energy from X-rays from the sun. The wavelengths of
the fluorescent X-rays are defined for each element, so, by
measuring the fluorescent X-rays emitted from the surface
of an asteroid, it is possible to determine about how much of
what elements are there. The fluorescent X-ray spectrometer
succeeded in measuring the material composition of the
asteroid Itokawas surface, which included magnesium,
aluminum, silicon, etc.
In addition, our InGaAs linear image sensor was chosen
for the Hayabusas near-infrared spectrometer owing to
the sensors high sensitivity in the near-infrared region as
(Courtesy of National Astronomical Observatory of Japan)
KMPDC0476EA well as its high reliability and durability. The near-infrared
spectrometer (NIRS) on Hayabusa was a device to disperse
and detect infrared rays from the sunlight reflected off the
surface of the asteroid, in order to analyze the minerals
on the ground and the form of the terrain. When 0.8 to
2.1 m light reflected from Itokawa was spectroscopically
measured, it was found that reflectance dropped at the
regions of 1 m and 2 m. Thus, it was concluded that the
minerals on the surface include olivine and pyroxene.
71
Reference
1) Masaharu Muramatsu, Hiroshi Akahori, Katsumi Shibayama, Syunsuke Naka-
mura and Koei Yamamoto, Hamamatsu Photonics K. K., Solid State Division:
"Greater than 90% QE in Visible Spectrum Perceptible from UV to near IR
Hamamatsu Thinned Back Illuminated CCDs", SPIE, Solid State Sensor Arrays:
Developments and Applications,3019 (1997), P2
2) M. P. Lesser, Steward Observatory, University of Arizona: "Chemical/Mechanical

Thinning Results", SPIE, New Methods in Microscopy and Low Light Imaging,
1161 (1989),P98
3) James Janesic, Tom Elliott, Taher Daud, Jim McCarthy, Jet Propulsion Labora-
tory California Institute of Technology, Morley Blouke, Tektronix. Inc.,: "Back-
side charging of the CCD", SPIE, Solid State Imaging Arrays, 570 (1985), P46
4) Y. Sugiyama, et. al., "A High-Speed CMOS Image Sensor With Profile Data Ac-
quiring Function", IEEE Journal of Solid-State Circuits, Vol.40, No.12, pp.2816-
2823, (2005)
5) Y. Sugiyama, et. al., A 3.2kHz, 14-Bit Optical Absolute Rotary Encoder with a
CMOS Profile Sensor, IEEE Sensors Journal, Vol.8, No.8, pp.1430-1436, (2008)
72

E05 Handbook Image Sensors

Uploaded by

Copyright:

Available Formats

E05 Handbook Image Sensors

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

E05 Handbook Image Sensors

Uploaded by

Copyright:

Available Formats

What are some of the different types of image sensors discussed?

What are some of the different types of image sensors discussed?

What are some applications of CCD and CMOS image sensors mentioned?

What are some applications of CCD and CMOS image sensors mentioned?

Image sensors CHAPTER 05

1 CCD image sensors 6 Photodiode arrays with amplifiers

2 NMOS linear image sensors

3 CMOS linear image sensors

Hamamatsu image sensors

Type Features Lineup

NMOS linear image sensor NMOS linear

Back-thinned CCD (without window)

CCD for X-ray imaging Front-illuminated CCD (without window) Front-

1 MeV 100 keV 10 keV 1 keV 100 eV 10 eV 1 eV 0.1 eV

Light level range detectable by image sensors (example)

High sensitivity film (ISO 1000)

InGaAs linear image sensor

NMOS/CMOS linear image sensor

Non-cooled CCD area image sensor

Non-cooled type CCD area image sensor

Cooled CCD area image sensor

Currently used CCDs are grouped by their transfer

One-dimensional type (linear image sensor)

fabricated with a larger number of pixels or with a larger

[Figure 1-3] Structure of FFT type (4) FIT type

[Figure 1-5] Structure of FIT type

of the photodiode itself and others. This charge is then

Horizontal shift register

Signal charge FD External load

Signal charge injection

Qinj: injection charge [C]

[Figure 1-11] Structure of signal charge injection

(b) CCD image sensor (IT type and one-dimensional type)

Analog shift register

Comparison with NMOS image sensor

for low-light-level detection.

Gate oxide film Thinning

activate the photosensitive area. The photosensitive area is 1

activated by forming an internal potential (accumulation) 2 Horizontal 11

light input surface are smoothly carried to the CCD

(2) Fine structure type

Horizontal shift register

Gate oxide film Horizontal shift register

(c) Fine structure type

50 Standard type Maximum current (Imax)

20 To avoid damaging a thermoelectric cooler and CCD,

CCD temperature (C)

continuously until reaching the last stage M (number of

a wide spectral range from the ultraviolet to the near

then the image S/N will deteriorate. When an object is

charges in each column are vertically transferred during

clear, continuous images with no blurring as shown in P

[Figure 1-22] Imaging in two-dimensional operation

(c) When drum is rotating (with shutter time shortened)

In the case of a CCD linear image sensor that uses a

Vout = Av Q / Cfd (3) [Figure 1-27] Spectral response (without window)

When the S7030/S7031 series is used: Cfd=48 fF Wavelength (nm)

[Figure 1-28] Spectral response of front-illuminated CCDs