22.3 Photodiodes
617
Fig. 22.38 Surface
potential as a function of
the signal charge Q sig for
various values of the bias
V G − V FB as labeled for a
SiO 2 /p-Si diode with
N A = 10 15 cm −3 and an
oxide thickness of 100 nm.
The dashed line represents
the limit for inversion given
by s = 2 B ≈0.6 V.
Adapted from [1852]
10
5
0
0
1
2
3
Q (10 e/cm )
sig
12
2
s (V)
Si
15V
10V
5V
τ th =
4 L
2
π 2 D n
,
(22.58)
where L is the length of the electrode. For a sufficiently large charge packet the self-induced drift due
to Coulomb repulsion is important. The decay of charge is then given by
Q(t)
Q(0)
=
t 0
t + t 0
,
(22.59)
with t 0 = π L
3 W e C i /(2μ n Q(0)). W e is the width of the electrode. This dependence is shown as a dashed
line in Fig. 22.40a. The last electrons are efficiently transferred by the drift induced by the fringing
field of the electrodes (solid line in Fig. 22.40a). The origin of the fringing field is schematically shown
in Fig. 22.40b; the minimum fringing field shown is 2 × 10
3 V/cm. In about 1–2 ns practically all
(1 − 10
−5 ) charges are transferred. This enables clock rates of several 10 MHz.
For the clocking of the CCD the lateral variation of potential depth with the applied gate voltage
is used. In Fig. 22.41 it is shown how a lateral variation of doping or oxide thickness creates a lateral
potential well. Such structures are used to confine the row of pixels against the neighboring rows
(channel stops, Fig. 22.42). In order to avoid carrier loss at the interface between the oxide and the
semiconductor a buried-channel structure is used (Fig. 22.43).
For front illumination parts of the contact electrodes shield the active area of the device. Higher
sensitivity (in particular in the UV) is achieved for back illumination. For this purpose the chip is thinned
(polished). This process is expensive and makes the chip mechanically less stable. For red/infrared
wavelengths typically interference fringes occur for such thinned chips due to the small thickness. An
increase in efficiency for front illumination can be achieved with an on-chip microlens (Fig. 22.44).
For color imaging the CCD is covered with a three-color Bayer mask [1858] (Fig. 22.45a). On
average there are one blue and one red pixel and two green pixels since green is the most prominent color
in typical lighting situations. Thus each pixel delivers monochromatic information; RGB images are
generated using suitable image software. Alternatives in high-end products are the use of a beam splitter,
static color filters and three CCD chips, one for each color (Fig. 22.45b), or the time-sequential recording
of three monochromatic images using one CCD chip and a rotating color-filter wheel (Fig. 22.45c).
Another method is to shift (by piezoelectric actors) the CCD by half or full pixels and take sequential
exposures, increasing spatial and color resolution for still objects.
617
Fig. 22.38 Surface
potential as a function of
the signal charge Q sig for
various values of the bias
V G − V FB as labeled for a
SiO 2 /p-Si diode with
N A = 10 15 cm −3 and an
oxide thickness of 100 nm.
The dashed line represents
the limit for inversion given
by s = 2 B ≈0.6 V.
Adapted from [1852]
10
5
0
0
1
2
3
Q (10 e/cm )
sig
12
2
s (V)
Si
15V
10V
5V
τ th =
4 L
2
π 2 D n
,
(22.58)
where L is the length of the electrode. For a sufficiently large charge packet the self-induced drift due
to Coulomb repulsion is important. The decay of charge is then given by
Q(t)
Q(0)
=
t 0
t + t 0
,
(22.59)
with t 0 = π L
3 W e C i /(2μ n Q(0)). W e is the width of the electrode. This dependence is shown as a dashed
line in Fig. 22.40a. The last electrons are efficiently transferred by the drift induced by the fringing
field of the electrodes (solid line in Fig. 22.40a). The origin of the fringing field is schematically shown
in Fig. 22.40b; the minimum fringing field shown is 2 × 10
3 V/cm. In about 1–2 ns practically all
(1 − 10
−5 ) charges are transferred. This enables clock rates of several 10 MHz.
For the clocking of the CCD the lateral variation of potential depth with the applied gate voltage
is used. In Fig. 22.41 it is shown how a lateral variation of doping or oxide thickness creates a lateral
potential well. Such structures are used to confine the row of pixels against the neighboring rows
(channel stops, Fig. 22.42). In order to avoid carrier loss at the interface between the oxide and the
semiconductor a buried-channel structure is used (Fig. 22.43).
For front illumination parts of the contact electrodes shield the active area of the device. Higher
sensitivity (in particular in the UV) is achieved for back illumination. For this purpose the chip is thinned
(polished). This process is expensive and makes the chip mechanically less stable. For red/infrared
wavelengths typically interference fringes occur for such thinned chips due to the small thickness. An
increase in efficiency for front illumination can be achieved with an on-chip microlens (Fig. 22.44).
For color imaging the CCD is covered with a three-color Bayer mask [1858] (Fig. 22.45a). On
average there are one blue and one red pixel and two green pixels since green is the most prominent color
in typical lighting situations. Thus each pixel delivers monochromatic information; RGB images are
generated using suitable image software. Alternatives in high-end products are the use of a beam splitter,
static color filters and three CCD chips, one for each color (Fig. 22.45b), or the time-sequential recording
of three monochromatic images using one CCD chip and a rotating color-filter wheel (Fig. 22.45c).
Another method is to shift (by piezoelectric actors) the CCD by half or full pixels and take sequential
exposures, increasing spatial and color resolution for still objects.