Image Processing Reference
In-Depth Information
this structure, transfer direction is defined unambiguously. Charges are stored in the chan-
nel of the storage electrode and pass through the barrier electrode channel only in transfer
operations. The charge quantity that can be stored is that which fills the built-in potential
difference in the storage electrode channel. As shown in Figure 5.5, complementary pulses
are applied as driving clock pulses. At time
t
1−1
, the charge packet is stored in a storage elec-
trode channel. Since voltages of ϕ1 and ϕ2 switch places at
t
1−2
, charges in the storage gate
channel of ϕ1 are transferred to that of ϕ2 through the barrier gate channel of ϕ2. Thus, the
charge packet is transferred to the next stage by only one clock pulse time, and one period of
time is only two clock pulses. Therefore, the two-phase mode CCD is suitable for high-speed
transfer application, although it is not appropriate for use of large charge quantities.
The channel structure used in real CCD sensors is the buried MOS structure mentioned
in Section 2.1.4 to achieve high transfer efficiency by avoiding interference of interface
states. This is called a buried-channel CCD (BCCD),
3
while CCDs having normal MOS are
named surface-channel CCDs (SCCDs). Comparisons of structure and potential distribu-
tion between SCCD and BCCD are shown in Figure 5.6a and b.
Since the maximum channel potential (minimum potential energy for electron) of a
SCCD is at the surface of silicon, charges are transferred at the interface. Therefore, elec-
trons are trapped at the interface states and released behind them. This phenomenon
causes degradation of transfer efficiency. On the other hand, the maximum potential of a
BCCD does not reside at the surface but inside the substrate, as explained in Figure 2.14;
charges pass through inside the silicon. As charges are transferred without interference
by the interface state, a high transfer efficiency can be realized. Currently, CCDs have
BCCD channel types. In Figure 5.6c, gate voltage
V
g
dependence on channel potential ϕ
ch
for both BCCDs and SCCDs is shown. While the channel potential of SCCD is almost 0 V
at gate voltage of 0 V, that of BCCD is above zero because a positive voltage is applied to
the
n
-region to deplete the buried channel. The channel potential varies with gate voltage.
But when gate voltage is changed to be larger in the negative direction, holes begin to be
collected in the valence band to form a surface inversion layer at the surface at a specific
voltage. A further increase of negative voltage is used only to collect more holes, and chan-
nel potential is pinned at the voltage. This gate voltage is named pinning voltage. In almost
all real CCDs, applied voltage ϕ [H] is 0 V and ϕ [L] is approximately equal to pinning
voltage. Therefore, pinning voltage and channel potential
φ
c
0
at
V
g
of 0 V are important
design parameters.
Why have CCDs cornered the market for so long? The reason is the “complete transfer”
of the transfer operation, as transfer charges are highly efficient. Focusing on a specific
gate electrode, as shown in Figure 5.7, it receives 100 electrons from the left gate chan-
nel and passes on those 100 electrons to the right gate channel, that is, there is neither an
increase nor a decrease of electron number. This means no noise is generated in the trans-
fer operation. Therefore, a CCD can achieve a low-noise performance and consequently a
high signal-to-noise ratio (SNR), that is, high sensitivity. This is the most important feature
of CCDs and the origin of their advantage.
As a device, this means complete depletion and no charge remains after complete charge
transfer. This means that there is no channel capacitance after transfer completion, accord-
ing to the definition of capacitance
C
as the inverse ratio of potential voltage change Δ
V
caused by charge quantity change Δ
Q
. Therefore, kTC noise does not exist. In other words,
because no charge is left in the channel, the potential uncertainty formed by a stochastic
process does not exist.
A charge transfer device named a bucket-brigade device (BBD), proposed at a simi-
lar time, has a similar concept to CCDs, using an MOS-type gate electrode. Charges are
