Note: Descriptions are shown in the official language in which they were submitted.
108~)355
This invention relates to a charge transfer
device for use in transversal filters and the like, and more
particularly to one which utilizes a double split in the
sensing electrodes of the device to reduce the total electrode
capacitance and hence the common mode signal component of the
sensed voltages.
Background of the Invention
Charge transfer devices (CTDs) such as CCDs
(charge-coupled devices) and BBDs (bucket brigade devices) can
be readily used in signal processing applications such as for
split-electrode transversal filters. A discussion of the
design considerations for such a filter can be found in a
paper by Richard D. Baertsch et al entitled "The Design and
Operation of Practical Charge-Transfer Transversal Filters"
IEEE Transactions on Electron Devices, Vol. ED-23, No. 2,
February 1976, pp 133-142. This paper outlines a number of
difficulties encountered in constructing such a filter using
split-electrode charge transfer techniques. In such a
transversal filter, the output signal is derived by
repeatedly differentially summing the total of the sensed
signals from one segment of each split-electrode from that
of the others as mobile charges are transferred along the
device. From Figure 15 of this paper it can be seen that in
a typical filter, for a large majority of the split electrodes,
only a small differential signal is developed between the two
segments. As the paper points out, this results in the
.. ;. :
' differential signal getting negligibly larger while the total
electrode capacity continues to increase thereby resulting in
a net decrease of signal voltage and a loss in signal-to-noise
ratio.
~803SS
Statement of the Invention
The present invention substantially decreases
this effect by utilizing a double split-electrode structure
in which the sensing electrodes are divided into three
segments of which only two are differentially sensed. The
third non-sensing segment is clocked separately to insure
that all mobile charge in each stage is transferred along
the device. With this structure, the total electrode area
connected to each sensing line is minimized thus reducing
clock transient pickup, capacitive loading and the common
mode signal which is applied to the differential amplifier.
In addition to providing improved performance, this puts
less constraints on the peripheral circuits for detection
i of the output signal.
Thus, the present invention relates to an
improvement in a charge transfer device which is particularly
adapted for use with transversal filters and the like and
comprises: a charge storage body; a dielectric layer disposed
over the body; and a plurality of storage electrodes disposed
' 20 in a contiguous relationship over the dielectric layer for
controlling the sequential transfer of mobile charges along
a channel in the body in response to clock voltage signals
applied thereto. The improvement in such a device comprises
at least some of the storage electrodes being split into three
segments along the direction of charge transfer whereby only
two of every three segments are differentially summed.
. I .
One of the inherent problems in any
-1 split-electrode device is the ambiguity with which the charge
divides beneath the sensing electrodes. In one method of
forming such a device, a localized field oxide region or
island is formed beneath the split in each electrode. An
',~
. , .
~0803S5
improvement in such a structure can be made by utilizing a
narrow tongue on the leading edge of each localized region
which extends partially under the previous storage electrode ,
so that charge splitting commences earlier during the transfer
operation. This narrow tongue also reduces the width of the
ambiguous area and enhances the field in the splitting region.
The addition of this localized field oxide
region reduces the active portions of the sensing electrodes.
To compensate for this, one embodiment of the invention
incorporates a notch at the side of the channel. The area
of this notch is equal to one-half of the area under the
localized oxide region, so that when the charge splits it
occupies the same surface area beneath the split electrode '
that it occupied beneath a non-split electrode. This
insures that the charge density and hence the surface '
potential of the split electrode remains the same under each
sensing segment of the electrodes, thus minimizing the output
distortion.
Brief Description of the Drawings
An example embodiment of the invention will
now be described with reference to the accompanying drawings ~' '
in which:
Figure 1 illustrates a top plan view of a
portion of a charge coupled device; and
Figure 2 illustrates a cross-sectional view
of the portion of the device taken along the line II-II of
Figure 1.
Description of'the'Preferred Embodiment
The fabrication of the charge coupled device
described herein utilizes technologies well established and
known in the semiconductor fi;ld. It is therefore -
.. . . .
' ~ .
. ~
1~80355
considered unnecessary to describe in detail the individual
steps for forming the device. However, Canadian Patent
No. 941,072 issued January 29, 1974 to James J. White,
describes one method of constructing a two-level poly-silicon
charge coupled device which is the basic structure of the
device disclosed herein. Also, it is evident that the figures
shown in the drawings are exemplary of the construction of
the invention and are not necessarily drawn to scale.
In the following detailed description and
accompanying drawings, basic reference numbers will be
assigned to individual elements of the device. Where it is
necessary to distinguish between such elements, additional
reference characters will be added to the base number. In
general, reference will be made only to the base number.
Referring to Figures 1 and 2, the illustrated ~-~
portion of the charge coupled device comprises a p-type
silicon substrate 10 having a variable thickness silicon
dioxide insulating layer 11 deposited thereon. A row of
interleaved polysilicon storage electrodes 12 and transfer
gates 13 laterally disposed so as to overlap adjacent ones,
are deposited on the insulating layer 11.
As shown in Figure 1, the silicon dioxide
insulating layer 11 includes a gate oxide region 15 which
define~ the channel along which mobile charges are transferred
under control of clock voltages applied to the electrodes 12
and 13. These gate oxide regions are delineated by thicker
field oxide regions 16 which define the boundaries of the
channel. The field oxide regions 16 are sufficiently thick
that the portions of the semiconductor substrate 10
immediately beneath them do not invert in response to the
application of clock vo1tages to the electrode~ 12 and 13.
108C~355
Consequently, minority carrier charges are only carried along
the substrate 10 in the channel defined by the gate oxide
regions 15. In addition, small islands or regions 17 of
field oxide are formed beneath each of the splits in the
storage electrodes 12. In addition, for ease of manufacture
field oxide regions 18 are also used where centre segments
12F" and 12H't of the split storage electrodes 12 are
interconnected by a conductor 19. Each of the islands 17 -
and 18 has a narrow tongue 20 of field oxide which extends
forward beneath a portion of the preceding electrode 12.
The purpose of this tongue 20 is to initiate the division
of charge beneath the split electrodes as it is being
transferred from beneath the previous electrode so as to
minimize any ambiguity and minimize the width of the
splitting area.
In a typical split-electrode charge coupled
device, charge sensing occurs beneath every second storage
electrode. Consequently, storage electrodes 12A, 12C, 12E
and 12G are non-split. Electrodes 12B and 12D are typical
of the known single split electrodes which can be utilized
in conjunction with the double split electrodes 12F and 12H
of the present invention. Thus, where a large differential
is required, the single split electrode (e.g. 12B) may be
used whereas when a small differential exists the double
split electrode (e.g. 12F) performs the same function.
Referring more specifically to Figure 2,
a typical CCD of the present invention has electrodes 12
and 13 which measure 8~m in the direction of charge flow
that are separated from each other by a gap of 4~m, thus
providing a 2~m overlap between adjacent storage electrodes 12
and transfer gates 13.
- 5 -
: . . . ~ . . -: ~ :
10803SS
As can be seen from Figure 1, the
delineation 15' between the gate oxide region 15 and the
field oxide region 16 is stepped slightly wider beneath each
of the split electrodes than beneath the non-split electrodes.
The purpose of this is to allow for the reduced areas of the
gate regions resulting from the field oxide islands 17 in
the split electrode. The area of this step or notch is equal
to one-half of the area under the preceding storage electrode
adjacent to the splitting islands 17 so that when the charge
splits it occupies the same surface area under the sensing
electrodes 12B, 12D, 12F and 12H as it did under the previous
non-split storage electrodes 12A, 12C, 12E and 12G. This
insures that the charge density and surface potential remains
; the same under each sensing segment of the electrodes thus
minimizing output distortion.
In operation, packets of charge are transferred
down the channel 15 from left to right under control of clock
ges 01~ 02' 02 ' 02 ~ 03 and 04 by appropriately shifting
the locations of the potential minima in a well known manner.
Differential sensing of the change in voltage on lines 02 and
02' ?esulting from the transfer charge beneath the segments of
the electrodes connected thereto may be achieved using
conventional differential amplifier techniques such as
described in Section IV entitled "Clocking and Signal Recovery"
of the article by Baertsch et al.
Clock 02" functions in unison with
clocks 02 and 02' so that the mobile charge is transferred
simultaneously beneath the three segments of the double split
electrodes. However only the two end segments 12F, 12H and
12F', 12H' are differentially sensed while the centre
segments 12F" and 12H" functlon solely to transfer the balance
-- 6 --
.,
1(~8~)355
of the mobile charge along the device so as to maintain a
uniform charge density in the channel 15 beneath the double
split electrodes. However this differential is achieved with
a marked decrease in total capacitance of the sensing
electrodes 12F, 12F', 12H and 12H'.
This is particularly evident in the double
split electrode 12H where the differential between the two
end segments 12H and 12H' is comparatively small, and the
non-sensed centre segment 12H" covers over 80% of the width
of the channel 15.
As pointed out previously, where a very large
differential is required such as in electrode 12B, the
conventional single split electrode is used instead of one
that is double split. Thus a single device may utilize both
types advantageously. ~ -
The Baertsch et al article also discloses the
use of a parallel channel on one side (Fig. 15) to balance
the total capacity between the two sets of sensing electrodes.
Since etching techniques are not perfect, an imp~oved balance
can be achieved by including a narrow channel on the other
side. Again the areas of these channels are such as to
balance the total capacity of the two sides. However, any
over or under etching of the device during the formation of
one parallel channel is compensated for by an equal over
or under etching in the other.
-- 7 --