Image Processing Reference
In-Depth Information
φSH
S
+
C
S
Canceled
output
Differential
circuit
φSH
0
-
C
0
(a)
φ
SH
C
C
Canceled
output
φ
CL
C
SH
V
clamp
(b)
FIGURE 5.42
Examples of an offset variation canceling circuit: (a) differential circuit; (b) CDS circuit.
full-fledged digital still cameras (DSCs) for the first time by Toshiba, is offset cancelation
in the charge domain by a cancelation circuit formed at each column,
33
which is similar to
the CDS circuit method.
Variations caused by the arrangement of an amplifier at each pixel are not only thresh-
old voltage, as shown in Equation 5.10, but gradient, which voltage sensitivity also varies.
Contrary to expectation, it is very rare for variation of voltage sensitivity to be a problem.
Because variation of the threshold voltage is randomly determined by microscopic condi-
tions at each transistor channel, correlation between adjacent pixels is low and differences
are large, so image quality is degraded drastically. On the other hand, it is thought that
mutual conductance
g
m
, which is seen in the equation of
G
V
, and the summed capacitances
of
pn
-junction, gate electrode and stray capacitance
C
in
, which mainly rules the voltage gain
of the SFA, are macroscopic parameters determined as averaged factors. Since they have a
certain level of correlation between adjacent pixels, they are not very different from each
other. Moreover, the voltage gain of the SFA also has a character that tends to compensate
for the influence of
g
m
variation.
A schematic diagram of a 3-Tr pixel CMOS sensor is shown in Figure 5.43. In operation,
the row selection pulse generated by the vertical access circuit is applied to row select
transistor RS as well as the MOS sensor. This makes the SFA active, and the vertical signal
line is set to the voltage of the SFA output. Multiplied charge quantity corresponding to
the voltage is generated on it. Thus, amplified signal charges in the vertical signal line are
supplied by the SFA. This is quite different from the MOS sensor, in which signal charges
on the vertical signal line are optically generated signal charges themselves in the PD.
Since signal charges in the PD do not move but remain at readout operation, this type of
readout is called nondestructive readout. To start the next exposure, the reset pulse gen-
erated by the vertical access circuit is applied to reset transistor RS to reset the PD. Then
the column selection pulse generated by the horizontal access circuit is applied to column
select transistor CS. Potential of the vertical signal line is output by the output part, after
the offset variation is canceled by the FPN cancelation circuit. Columns are accessed to be
output in series. After all columns in one row are output, the next row is accessed as well
as the MOS sensor.
Figure 5.44 is a schematic diagram of a readout operation of one pixel in a 3-Tr pixel
CMOS sensor. Figure 5.44a and b show cross-sectional view and potential distribution in
