GB1596336A

GB1596336A - Charge transfer recursive filter

Info

Publication number: GB1596336A
Application number: GB1879478A
Authority: GB
Original assignee: Thomson CSF SA
Current assignee: Thales SA
Priority date: 1977-05-13
Filing date: 1978-05-10
Publication date: 1981-08-26
Also published as: FR2390857A1; DE2820836C2; DE2820836A1; FR2390857B1

Description

(54) CHARGE TRANSFER RECURSIVE FILTER (71) We, THOMSON--CSF, a French Body Corporate, of 173, Boulevard Haussmann, 75 Paris (8e) France, do hereby declare the invention, for which we pray that a patent may be granted to us, and the method by which it is to be performed, to be particularly described in and by the following statement: The invention relates to recursive filters using the transfer of electric charges in a semiconductor. More specifically, it relates to an input stage for a filter of the aforementioned kind, which realizes a new injection in the filter of the output signal of this last one.

As is known a charge transfer filter, often called a transversal filter, comprises a semiconductive substrate, covered with an insulant on which electrodes are disposed which, on periodic application of given potentials, transfer packets of electric charges representing the signal being processed. The electrodes are disposed paralled to one another and transversely with respect to the charge transfer direction. Some of the electrodes are transversely divided into two unequal parts, and the quantities of charges arriving under these electrodes are differentially read in order to weight the signal. The differential signal is the filter output signal which, in a recursive filter, is re-injected into the filter input and added to the input signal proper.

Separate elements can be used for performing the various afore mentioned operations, i.e. obtaining the difference between the signals -coming from the two parts of a divided electrode, re-injecting the differential signal and adding it to the input signal. The invention relates to a device having a simple structure, wherein the various components for performing the aforementioned operations are incorporated on a single substrate.

According to the present invention, there is provided a recursive charge-transfer filter comprising: -a semiconductive substrate; -an insulating layer deposited on the substrate; -electrodes deposited on the insulating layer and alternately undivided and divided into two parts and used, when given potentials are applied, for transferring charges in the substrate; means for reading the quantities of charges present under the parts of the divided electrodes: -an input stage, disposed in front of the electrodes, which comprises two electrically insulated, roughly parallel channels in which charges are transferred, each channel comprising, an injection diode for injecting charges into the channel, two roughly parallel electrodes or grids, and means for providing electric contact between the regions of the substrate under the grids, the first channel being used for sampling and injecting the input signal and the second channel receiving on its grids when operating the two reading signals corresponding respectively to the two parts of the divided electrodes, the injection diode of the second channel being controlled in phase with the first channel when operating and the charges injected into the two channels thus being summed by the first of the undivided electrodes.

For a better understanding of the present invention reference will now be made to the accompanying drawings in which: Fig. 1 is a block diagram of a recursive charge-transfer filter; Fig. 2 is a diagram of an embodiment of a filter according to the invention; Fig. 3 is a diagram of the signals applied to the device according to the invention, and Fig. 4 shows diagrams illustrating the operation of the filter according to the invention In the drawings, the same references relate to the same elements.

In the diagram in Fig. 1, a block 10 represents a conventional charge-transfer filter comprising alternately divided and undivided electrodes which, on application of suitable periodic potentials, transfer packets of charges representing the signal being processed.

In this kind of filter, as previously mentioned, it is necessary to read the quantities of charges passing under the divided electrodes in order to obtain the difference between the corresponding signals. This is represented by blocks 12 and 13, device 12 being used for reading the quantities of charges passing under the top parts, on the diagram (called negative electrodes for convenience) of the divided electrodes, whereas 13 is a device for reading the quantitites of charges passing under the bottom parts (called positive electrodes) of these electrodes. The reading can be realized by any known current or voltage method, advantageously as described in British Patent Application No.

17578/78 (Serial No. ) in the name of Thomson-CS F.

The signals S- and S+, produced by the reading devices 12 and 13 respectively, are sent to an element 14 which obtains the difference between them, which constitutes the filter output signal S.

In a recursive filter, the output signal S is sampled, so as to be re-injected into filter 10, at the same time as the input signal (E) to which it is added. In Fig. 1, a block 16 symbolises the operation of injecting the input signal E, i.e. the conversion of an analog electric signal into a sample represented by a quantity of charges. A block 17 represents the operation of reinjecting the output signal S, and a block 11 connected to filter 10 represents the addition of the two signals.

Fig. 2 is a top view of an embodiment of a recursive charge-transfer filter according to the invention, using charge-coupled devices or CCD.

The device, as is conventional in the case of charge-transfer devices, comprises a semiconductive substrate, e.g. of silicon, covered with a layer of insulant (e.g. silicon oxide) on which electrodes are disposed transversely to the direction of propagation OZ of charges in the semiconductor.

The device has two parts separated by an axis XX. The left part is called the input stage and the right part constitutes the filter proper, corresponding to block lü in Fig. 1.

The filter part is of the two-phase kind, i.e. its electrodes are connected to two periodic potentials in phase opposition. It thus comprises a set of undivided electrodes (3, 4, 5, 6 in the diagram) receiving a signal 0,, and a set of divided electrodes (31-32, 41v2, 51-52 and 6162) alternating with the previously-mentioned electrodes and receiving a signal Q, e.g. via reading devices 12 and 13.

By way of example, signals l and Q are illustrated in Fig. 3, diagrams a and b. The signals are roughly square-wave, having the same period T but in phase opposition. their amplitude varying between two values : i.e.

lb and 1H for 0,, and 2B and 2H for 8,, these values being expressed with respect to the substrate potential. Signals l and 2 are preferably identical, apart from the phase shift, and their lower value (1B or QB) is approximately zero.

The filter terminates in a diode Dc formed in the semi-conductive substrate and adapted to collect and evacuate the charges which have travelled through the device.

The input stage of the device comprises two parallel channels 1, 2 electrically insulated from one another, in which the charges can be transferred. The channels can be insulated by any known means, e.g.

by an excess thickness of insulant or by a local increase in the doping of the substrate.

Channel 1 comprises a charge-injecting diode D,l, two electrodes or grids Gl and G2, and a diode D" between grids Gl and G2. Diodes D,l and Dri are constituted by a PN junction in the semiconductive substrate. Similarly, channel 2 is made up of a diode D,2, two grids G3 and G4 and a diode D.2 between the two grids.

The two diodes Dil and D,2 are supplied by a signal V illustrated by way of example in diagram (C) in Fig. 3. The signal has the same period T as the preceding signals and its amplitude varies in squarewaves between VDB and V,,, its lower phase (VD=VDB) being in phase with the lower phase of l and its duration being less than T/2.

In channel 1, grid Gl is kept at a constant potential V1, which as shown in the drawing Is positive when the semiconductive substrate of the device is of type P. Grid G2 receives a constant potential V02 and the input signal E(t) to be filtered.

Channel 2 is used for re-injection. Its grid G3 receives the signal S-, supplied by the reading device 12, via a rheostat Rl, and its grid G4 receives the signal S+, provided by the reading device 13, via a rheostat R2.

The operation of the device will be described with reference to diagrams (a, b and c) in Fig. 4.

These diagrams are sectional views of the device along the axis OZ of charge propagation, level with channel 1. The diagrams show a semi-conductive substrate (20), a layer of insulant (21) covering it, the two grids G,, G2 of channel I and the first two electrodes (3 and 31) of the filter.

Diodes Dil and Dri comprise regions of the substrate having an opposite type of conductivity (e.g.N) from that of the substrate (e.g.P.).

Diagram (a) corresponds to an instant ta illustrated in Fig. 3, where 1=1B and VD=VDH. Opposite grids G1 and G2, broken lines 23 and 24 illustrate the levels of the potentials applied to these grids, i.e. V01 (line 23) nd V02+E (line 24). Line 25, opposite electrode 3, represents the value 1B of 81. Line 22 represents the level VDB of the potential VD applied to the injection diode D,l. During this phase, therefore, charge carriers are injected under grids G1 and G2 (the hachured region on the Figure) but cannot penetrate into the filter since they are stopped by electrode 3.

At the next instant tb, represented on diagram (b), l is still at its low value 0,,, but potential VD is equal to VDH, as illustrated by a line 26. The charge carriers thus remain trapped under grid G2, in a quantity which depends on the potential difference (VGZ+E), fixed by grid G2, and V01, fixed by diode Dri which provides an electric connection between the regions under grids G1 and G2. This quantity of charge (QA), denoted by A in the drawing, is equal to: QA=COX (V02+E-V01) where Cox is the oxide capacity. It represents the sampled input signal E(t). It can also be seen that, if the device is to operate efficiently, V02 must be greater than V01.

Diagram (c) in Fig. 4 illustrates the next instant (tc), when VD is still in the high state (Vn=VDH) and 01 is changing to the high state (line 27), which allows to only the quantity of charges QA (A) to be transferred under electrode 3, neglecting the extra charge under diode D,, since the next electrode (31) is controlled by the signal 2, which is then in the low state (#2=#2B, line 28).

In the next stage, 0,=0,,, VD=VDB and 2=2Hs i.e. we return to diagram (a) in Fig.

4, a new sample representing a new value of the input signal E(t) being prepared at the same time as sample A is transferred from electrode 3 to electrode 31.

It can be seen from Fig. 2 that, by a conventional mechanism in charge transfer filter, a packet of charges representing a sample of the input signal E(t) (e.g.QA) is transferred from one electrode to the other at each half-period (T/2) of signals 8, and .

As is known, to ensure that the charges are always transferred in the same direction (+OX) it is necessary to provide some asymmetry at each electrode.

As previously stated, one of each pair of electrodes is divided into two parts, the ratio between the surfaces of which expresses a weighting coefficient applied to the signal in order to obtain the desired filtering. The quantity of charge present under the divided electrodes must therefore be read and, to this end, the electrodes are combined with the reading devices 12 and 13 which each provide a reading signal Sand S+ respectively. In a conventional filter, signals S and S+ supply a differential amplifier 15 which prepares the filter output signal S.

In the recursive filter according to the invention, signals S- and S+ are sampled at the outlet of devices 12 and 13 and sent to grids G3 and G4 of channel 2 of the input stage.

Channel 2 operates in similar manner to channel 1, signal S replacing the potential V01 and signal S+ replacing the potential (V 2+E). Channel 2 therefore is a difgerential stage providing a quantity of charges representing the difference (S±S-).

Since the samples are prepared in both channels 1 and 2 under the control of a single signal V,, they arrive in phase under the first electrode (3), where they are added.

The advantage of this structure is its symmetry, i.e. the same structure is used for injecting the input signal and re-injecting the output signals, thus inter alia reducing interfering signals to a minimum, since the output signal is obtained by difference.

However, channel 2 can be connected to signals S+ and S via rheostats (R2 and R1), if weighting is necessary in practice.

The preceding description relates to a method of differential injection of quantities of charges using a diode (Dri and Dr2) in each channel to maintain electric contact between the two grids of the channel. This is only an exemplary embodiment, and contact can be provided by any known means, inter alia by extending the second grid (G2 and G4 above the first (Gl and G3) after covering it with an insulating layer.

The device, therefore, has a simple structure, in which the filter input stage contains the various components for obtaining the difference between signals S+ and S- with a view to re-injecting them, and for re-injecting and adding the re-injected signal to'the input signal in a recursive filter.

This compact structure can be used in conventional manner to reduce the consumption and bulk of the device.

The filter according to the invention is particularly suitable for obtaining frequency responses having steep fronts which, of course, require a large number of weighting stages, i.e. a large number of electrodes in a conventional filter. The invention therefore, is applicable inter alia to telephone transmission.

WHAT WE CLAIM IS:- 1. A recursive charge-transfer filter comprising: -a semiconductive substrate; -an insulating layer deposited on the substrate;

**WARNING** end of DESC field may overlap start of CLMS **.

Claims

**WARNING** start of CLMS field may overlap end of DESC **.

Diagram (a) corresponds to an instant ta illustrated in Fig. 3, where 1=1B and VD=VDH. Opposite grids G1 and G2, broken lines 23 and 24 illustrate the levels of the potentials applied to these grids, i.e. V01 (line 23) nd V02+E (line 24). Line 25, opposite electrode 3, represents the value 1B of 81. Line 22 represents the level VDB of the potential VD applied to the injection diode D,l. During this phase, therefore, charge carriers are injected under grids G1 and G2 (the hachured region on the Figure) but cannot penetrate into the filter since they are stopped by electrode 3.

At the next instant tb, represented on diagram (b), l is still at its low value 0,,, but potential VD is equal to VDH, as illustrated by a line 26. The charge carriers thus remain trapped under grid G2, in a quantity which depends on the potential difference (VGZ+E), fixed by grid G2, and V01, fixed by diode Dri which provides an electric connection between the regions under grids G1 and G2. This quantity of charge (QA), denoted by A in the drawing, is equal to: QA=COX (V02+E-V01) where Cox is the oxide capacity. It represents the sampled input signal E(t). It can also be seen that, if the device is to operate efficiently, V02 must be greater than V01.

Diagram (c) in Fig. 4 illustrates the next instant (tc), when VD is still in the high state (Vn=VDH) and 01 is changing to the high state (line 27), which allows to only the quantity of charges QA (A) to be transferred under electrode 3, neglecting the extra charge under diode D,, since the next electrode (31) is controlled by the signal 2, which is then in the low state (#2=#2B, line 28).

In the next stage, 0,=0,,, VD=VDB and 2=2Hs i.e. we return to diagram (a) in Fig.

4, a new sample representing a new value of the input signal E(t) being prepared at the same time as sample A is transferred from electrode 3 to electrode 31.

It can be seen from Fig. 2 that, by a conventional mechanism in charge transfer filter, a packet of charges representing a sample of the input signal E(t) (e.g.QA) is transferred from one electrode to the other at each half-period (T/2) of signals 8, and .

As is known, to ensure that the charges are always transferred in the same direction (+OX) it is necessary to provide some asymmetry at each electrode.

As previously stated, one of each pair of electrodes is divided into two parts, the ratio between the surfaces of which expresses a weighting coefficient applied to the signal in order to obtain the desired filtering. The quantity of charge present under the divided electrodes must therefore be read and, to this end, the electrodes are combined with the reading devices 12 and

13 which each provide a reading signal Sand S+ respectively. In a conventional filter, signals S and S+ supply a differential amplifier 15 which prepares the filter output signal S.

In the recursive filter according to the invention, signals S- and S+ are sampled at the outlet of devices 12 and 13 and sent to grids G3 and G4 of channel 2 of the input stage.

Channel 2 operates in similar manner to channel 1, signal S replacing the potential V01 and signal S+ replacing the potential (V 2+E). Channel 2 therefore is a difgerential stage providing a quantity of charges representing the difference (S±S-).

Since the samples are prepared in both channels 1 and 2 under the control of a single signal V,, they arrive in phase under the first electrode (3), where they are added.

The advantage of this structure is its symmetry, i.e. the same structure is used for injecting the input signal and re-injecting the output signals, thus inter alia reducing interfering signals to a minimum, since the output signal is obtained by difference.

However, channel 2 can be connected to signals S+ and S via rheostats (R2 and R1), if weighting is necessary in practice.

The preceding description relates to a method of differential injection of quantities of charges using a diode (Dri and Dr2) in each channel to maintain electric contact between the two grids of the channel. This is only an exemplary embodiment, and contact can be provided by any known means, inter alia by extending the second grid (G2 and G4 above the first (Gl and G3) after covering it with an insulating layer.

The device, therefore, has a simple structure, in which the filter input stage contains the various components for obtaining the difference between signals S+ and S- with a view to re-injecting them, and for re-injecting and adding the re-injected signal to'the input signal in a recursive filter.

This compact structure can be used in conventional manner to reduce the consumption and bulk of the device.

The filter according to the invention is particularly suitable for obtaining frequency responses having steep fronts which, of course, require a large number of weighting stages, i.e. a large number of electrodes in a conventional filter. The invention therefore, is applicable inter alia to telephone transmission.

WHAT WE CLAIM IS:- 1. A recursive charge-transfer filter comprising: -a semiconductive substrate; -an insulating layer deposited on the substrate;

-electrodes deposited on the insulating layer and alternately undivided and divided into two parts and used, when given potentials are applied, for transferring charges in the substrate; means for reading the quantities of charges present under the parts of the divided electrodes; -an input stage, disposed in front of the electrodes, which comprises two electrically insulated, roughly parallel channels in which charges are transferred, each channel comprising, an injection diode for injecting charges into the channel, two roughly parallel electrodes or grids, and means for providing electric contact between the regions of the substrate under the grids, the first channel being used for sampling and injecting the input signal and the second channel receiving on its grids when operating the two reading signals corresponding respectively to the two parts of the divided electrodes, the injection diode of the second channel being controlled in phase with the first channel when operating and the charges injected into the two channels thus being summed by the first of the undivided electrodes.
2. A filter according to claim 1, wherein the injection diodes are controlled by the same periodic signal.
3. A filter according to claim 2, wherein the undivided electrodes are connected to a signal l having a period T, wherein the two parts of the divided electrodes are connected to a signal 2 having the same period T, but phase-shifted by T/2 with respect to l and wherein the signal controlling the injection diodes has its lower phase in phase with the lower phase of 8,.
4. A filter according to claim 1, wherein the two grids of the first channel are respectively connected to two constant potentials, the second grid also receiving the filter input signal
5. A filter according to claim 1, wherein, in each channel, the means for providing electric contact between the regions of the substrate on the grids comprise a second diode formed from a region of the substrate having the opposite type of conductivity from the rest of the substrate.
6. A filter substantially as described with reference to Figure 1.
7. A filter substantially as described with reference to figures 2, 3, 4.