GB1595985A - Transversal filters - Google Patents

Description

(54) IMPROVEMENTS IN OR RELATING TO TRANSVERSAL FILTERS (71) We, SIEMENS AKTIENGESELLSCHAFT, a German Company of Berlin and Munich, German Federal Republic, 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 present invention relates to transversal filters of the type comprising a charge transfer device, wherein one surface of a substrate of doped semiconductor material has arranged thereon a series of capacitor storage elements, each element consisting of at least one insulating layer capacitor or blocking layer capacitor each having separate control electrodes for selective connection to a plurality of shift pulse trains which are mutually displaced in phase, said filter having a serial input terminal to which a signal to be filtered is applied, and each of said capacitor elements having a nonreactive amplifying output connected to a common filter output terminal from which a filtered version of the signal applied to said input terminal is obtained, the amplification factor of each respective element amplifying outputs being equal to a respective given value to determine the pulse response of the filter.

Transversal filters of the type described in the introduction are described, for example, in the IEEE Journal of Solid-State Circuits, Vol. SC-8, No. 2, April 1973, page 138 to 146; and in the Bell-Northern Research 11.4 page 240 to 243. Other details of an earlier proposed construction can be gathered from our co-pending United Kingdom Patent Application No. 40100/77 (Specification No. ). The charge transfer devices consist of CCDs or bucket-brigade circuits. The required non-reactive amplifying outputs are provided in the articles described in the cited literature references, either by means of so-called split-electrodes or by means of actual amplifier stages each having a high input impedance. In the case of the split-electrode method, each said capacitor storage element consists of two insulating layer subsidiary capacitor elements which are arranged next to one another and coupled by means of a doped zone in the substrate. The capacitance ratio of these two insulating layer subsidiary capacitors is selected in dependence upon the desired amplification factor. One nonreactive, amplifying output is provided for each of said specific capacitor storage elements, and in each case from that subsidiary capacitor storage element intended for one predetermined shift pulse train. All the non-reactive, amplifying outputs, one from each capacitor storage element, are connected to a common filter output terminal at which a filtered output signal is obtained.

The individual amplification factor of each of the non-reactive amplifying outputs has a respective value determined by the pulse response of the filter to provide the function which is to be realised by the filter.

The invention consists in a transversal filter in which a charge transfer device is formed on one surface of a substrate of doped semiconductor material by a series of main capacitor storage elements, each comprising two or more subsidiary capacitor elements of the insulating layer or blocking layer type provided with a plurality of connections by means of which each main capacitor storage element can be connected to the same set of shift pulse trains mutually displaced in phase; to effect a charge shift, when operating, the connections enabling the provision of the mutually different phase shift pulse trains to the respective subsidiary capacitor elements of each main capacitor storage element in the same spatial order within each main capacitor storage element, said filter having an input terminal to which a signal to be filtered is applied, when operating, and in each main capacitor storage element, one of the subsidiary capacitor storage elements having a non-reactive, amplifying output which is connected to a common filter output terminal at which a filtered signal is supplied, when operating, the amplification factor of each said amplifying output having an assigned value in the pulse response of the filter to provide the desired filter function, and wherein at least one of said main capacitor storage elements has at least one further nonreactive, amplifying output connected from another of its subsidiary capacitor elements to said common filter output terminal.

The fundamental advantages of an embodiment constructed in accordance with the invention are as follows: for a predetermined arbitrary filter function, in comparison to known or earlier proposed transversal filters it is possible to considerably reduce the surface space requirement, whilst if the same surface space can be utilised it is possible to considerably increase the reproduction accuracy of the filter function, by means of the one or more additional, non-reactive, amplifying outputs. Furthermore, different filter properties can be achieved simply by modifying the shift pulse train frequencies during operation.

The invention will now be described with reference to the drawings, in which: Figure 1 (1) schematically illustrates one earlier proposed transversal filter for purposes of comparison; Figure 1 (II) schematically illustrates one exemplary embodiment of a filter constructed in accordance with the invention; Figure 1 (Ill) schematically illustrates one alternative exemplary embodiment; Figure 2 is a set of explanatory graphs showing shift pulse trains plotted against time t, and an exemplary input signal waveform which is to be filtered; Figures 3 to 7 illustrate explanatory matrix-like plan diagrams; and Figure 8 is a set of explanatory graphs showing the output signals of the earlier proposed transversal filter in comparison to the output signal of exemplary embodiment of the invention both plotted against time t.

In the construction in accordance with our earlier proposal, shown in Figure 1 (I) a transversal filter is formed by a charge transfer device 10, this being a two-phase operation CCD, which means that each capacitor storage element 1 to 4 of this CCD, consists of two adjacent ones of subsidiary capacitor elements 11, 12, 21, 22, 31, 32, 41 & 42. The first subsidiary capacitor element 11, 21, 31 and 41 in each element is intended for one shift pulse train, and the other subsidiary capacitor elements 12, 22, 32 and 42 are for the other of the two shift pulse trains, which are mutually displaced in phase. A serial input terminal E is provided for the CCD, and each subsidiary capacitor element which is provided for one shift pulse train has connected thereto a respective non-reactive amplifying output, K11, K21, K31 and K41. All these outputs are connected to a common filter output terminal A. In the known filter shown in Figure 1 (I) the outputs are connected to the subsidiary capacitor elements 11, 21, 31 and 41, but such outputs could equally well be connected instead to the subsidiary capacitor elements 12, 22, 32 and 42. The symbols K11, K21, K31 and K41 simultaneously represent the respective amplification factors of the relevant, non-reactive amplifying outputs. The CCD 10 can in general terms, be a CCD for phase operation, where n is any whole number greater than 1, and this can be expressed by stating that each subsidiary element of the CCD contains n adjacent subsidiary capacitor elements and n shift pulse trains all mutually displaced in phase are connected for operation of the filter.

The charge transfer device can also consist of a bucket brigade circuit. Each possibility is suitable with the use of respective nonreactive, amplifying outputs.

In the exemplary embodiment of the invention shown in Figure 1 (II), there are only two capacitor storage elements 1 and 2, in the charge transfer device, and a separate non-reactive, amplifying output K11, K2l, K31, K4" iS provided for each adjacent subsidiary capacitor element 11, 12, 21 and 22, all being connected to a common filter output terminal A, so that there are only four outputs employed, as in the filter shown in Figure 1 (I), but if the filter shown in Figure 1 (II) is compared it will be seen that only four subsidiary capacitor elements 11, 12, 21 and 22 are required, everything else remaining similar to the known filter.

Thus with the filter embodiment shown in Figure 1 (II) a surface space reduction of approximately 50% is achieved in comparison to that required for the filter shown in Figure 1 (I). When a CCD for n-phase operation, where n = 3 or more is used, then the surface space requirement for this embodiment of the invention can be reduced by the factor 1/n in comparison to that of the filter shown in Figure 1 (I).

The exemplary embodiment of a filter constructed in accordance with the invention and shown in Figure 1 (III) differs from the Figure 1 (1) filter in that four capacitor storage elements, 1 to 4 are used. as in the filter shown in Figure 1 (I), each having two of subsidiary capacitor elements 11. 12, 21, 22, 31, 32, 41 and 42, but the subsidiary capacitor elements 12, 22, 32 and 42. but the subsidiary capacitor elements 12, 22, 32 and 42 also have non-reactive, amplifying out puts K12, K22, K32 and K42, all connected to the common filter output A. Thus with the embodiment shown in Figure 1 (III) there is double the number of non-reactive, amplifying outputs, compared with the filter shown in Figure 1 (I). When a charge transfer device for n-phase operation is used, where n is 3 or more, then up to n times as many outputs can be provided, in comparison to the Figure 1 (I) filter.

The selection method in which the pulse train frequency of the shift pulse train for the charge transfer device in the exemplary filters shownxin Figure 1 (II) and Figure 1 (III) is of fundamental significance. Two situations are of particular significance: the pulse train frequency may be selected to equal the pulse train frequency f0 of the earlier filter, resulting in a doubling of the output frequency at which the filtered signal samples are fed to the filter output terminal A in the embodiment shown, and if we say a charge transfer device for n-phase operation is used, we can say the increase will be n-times; the second possible situation being that the pulse train frequency of the shift pulse train is selected to be equal to fro/2 (or in general terms f"/n), whereby the output frequency at which filtered samples of the input signal are fed to the output terminal A is equal to fO. The output frequency of samples to the output terminal A is in any case double (generally n times) the sampling frequency with which the signal to be filtered is sampled at the input terminal E.

Thus it is possible to differentiate between a total of four alternative cases for the exemplary filters shown in Figures 1 (II) and 1 (III). If the pulse train frequency of the shift pulse train for the filter shown in Figure 1 (I) is referenced f(, whilst for the filter embodiment shown in Figure 1 (II), a pulse train frequency of f1i is used and a frequency fill is used for the embodiment of the invention shown in Figure 1 (III); the four possible combinations fil = fo, fill = fil = f)n and f111 = fwn exist for the illustrated exemplary embodiments where n = 2. The shift pulse train is to be understood as the pulse train with which an information charge is forwarded from element to element of the charge transfer device.

The set of graphs shown in Figure 2 consist of a shift pulse train plotted against time tin row (IV), for comparison with a typicalinput signal to be filtered, shown in row (V), with the shift pulse train for the charge shift device 10 schematically superimposed by way of explanation, these pulses have a repetition period To. The pulse train frequency of the shift pulse train is governed by f0 = lIT11. The signal to be filtered is sampled in the filter I at the pulse train repetition frequency f(). The sampled signal values are refenced S11, S21, S31 and S41. If the filter shown in Figure 1 (I) were operated with a pulse train frequency 2fo, additional signal values would be sampled, and these additional signal values are indicated in Figure 2 row (V) by broken lines S12, S22, S32, and S42. However, these additional signal values are of no special significance in the following discussion of general principles.

In the filter shown in Figure 1 (I), the capacitor storage elements 1 to 4 may be more generally identified for larger scale embodiments by the use of two-digit references xy, where x indicates the serial number of the storage element of the charge transfer device, counting towards the input E, and y indicates the serial number of the subsidiary capacitor element in that storage element, again counting towards the input E, so that in the charge transfer device 10, y only has the values 1 and 2 although in general it can be considered to pass through the numbers 1 to n to cover the possibility of more elaborate embodiments. The nonreactive, amplifying outputs are referenced Kxy2 , to identify their respective output positions, for example the output K21 being the output provided by the subsidiary capacitor element 21.

The instantaneous signal values which have been evaluated by the respective amplification factors can be generally represented as, kxy Suv identifies successive sampling times, u signifying the serial index of the sampling times tl, t2 etc., (see Figure 2 row (IV); and v being a number from 1 to n, and additionally indicating sampling times between tu and tU+l when the filter is operated with a pulse train frequency of n f1). Only the situation n = 1 is of significance in the following discussion.

Figures 3 to 7 illustrate respective matrixlike plans of the weighted signals Kxy Suv In each diagram only the columns for v = 1 are of significance, in this discussion, and the columns for v = 2 could be omitted. In each plan diagram the related weighted signal values , Kxy Suv which together form any output signal are surrounded by a respective circle and linked by a diagonal line, indicating that the related values kxy suV in those circles connected by a line are added to form a particular output signal.

The plan diagram in Figure 3 relates to the filter shown in Figure 1 (I), assuming that the signal value S11 fed in at a time t, has reached the subsidiary capacitor element 11 in the charge transfer device. This subsidiary capacitor element 11 then contains signal value SI1 the element 21 contains the signal value S21, the subsidiary capacitor element 31 contains a value S31 and the subsidiary capacitor element 41 contains a signal value S41. These signal values are read out from the charge transfer device in parallel and the sum of the values Kll sll, K21's21, K31 s3l and K41 s4l is fed to the common filter output terminal A. The next pulse train causes all these signal values to be shifted towards the left by one subsidiary capacitor element, whereby the subsidiary capacitor element 41 contains a newly added signal vaue S5. Naturally the charge transfer device must be provided with an output stage following the last subsidiary capacitor element, in order that each signal value contained therein can be read out on the arrival of the next pulse train. This also applies to the exemplary filter embodiments shown in Figures 1 (II) and I(III). A suitable output stage is an electrode having an implanted barrier, for example. The signal values S21, S31, S41 and S51 are again read out in parallel, and the sum of K11s21, K2l-s3l-K3l s4l and K41-ssl is fed to the common filter output terminal A.

This sum is shown in the plan diagram of Figure 3. By continuing to develop the diagram in this manner, all the successive signal values of the filtered signal can be determined simply. In the plan diagram shown in Figure 3, the first three signal values of the filtered signal are referenced A1, A2 and A3. The output frequency fA at which the filtered signal samples are withdrawn is governed by fA = fl The plan diagram shown in Figure 4 relates to the filter shown in Figure 1 (II), and applies where fit = f0. If the signal value S11 fed to the input terminal E at a time t1 has been forwarded to reach the subsidiary capacitor element 11, the subsidiary capacitor element 31 then contains a signal value S21. The two signal values are read out in parallel and the resultant instantaneous value Al = Kll-s + K31. s2l of the filtered signal is fed to the common filter output terminal A. These signal values are then shifted towards the left by one subsidiary capacitor element after half the pulse train repetition period, so that when the subsidiary capacitor element 12 contains signal value S2, and the subsidiary capacitor element 22 contains a newly input signal value S31. Both signal values are read out at this time, and the resultant value A, = K21#s21+K41#s31 of the filtered signal is fed to the common filter output terminal A.

these two signal values are then again displaced towards the left by one subsidiary capacitor element in the second half of the full pulse train period, so that at this time the resultant instantaneous value A3 = Kll-s2g+K3l-s31 of the filtered signal is fed to the common output terminal A. Here the output frequency fA = 2f(,, whereas the sampling frequency fE = The plan diagram shown in Figure 5 also applies to the filter shown in Figure 1 (II), for a case where fit = f0/2. If the signal value S11 fed to the input terminal E at time tl has reached the subsidiary capacitor element 11, the subsidiary capacitor element 31 contains value S31, as in this case sampling is carried out only following every second timing instant. At this time the resultant value Al = K1l-sll+K3l-s3l of the filtered signal is fed to the common filter output terminal A.

Following the time To = 1/fo a shift by one subsidiary capacitor element towards the left has occurred, as a result of which the resultant value A2 = K21-s3l+K4l-ssl is fed to the common output terminal A at this time. The output frequency fA = fo, whereas the sampling frequency fE = fo/2.

The plan diagram in Figure 6 relates to the filter shown in Figure 1 (III) for the case where fiii = f0. The diagram illustrates the formation of the signal values of the output signal for the first three signal values A1, A2 and A3. The following signal values can easily be determined by continuing the diagram accordingly. Here fA = 2fo and fE = fo.

The plan diagram in Figure 7 also applies to the filter shown in Figure 1 (III), for the case where f111 = fo/2. Again the formation of the first three signal values Al, A2 and A3 can be gathered by study of the diagram, which is formed in the manner described.

By continuing the diagram accordingly, it is easily possible to determine all the following resultant signal values of the filtered signal.

In this case fA = fg and fE = f()I2.

In the set of explanatory graphs shown in Figure 8, diagram (VI) illustrates an example of a filtered signal formed by samples fed from the common filter output terminal A of the filer shown in Figure 1 (I) during operation with a pulse train frequency f(). By way of comparison, the diagram VII illustrates the output signal at the common filter output terminal A of the filter shown in Figure 1 (III) for the same applied input signal, if this filter is operated with a pulse train frequency fill = f(). It can be seen that a significant smoothing of the filtered signal will be achieved with the filter shown in Figure 1 (III), in comparison to the filter shown in Figure 1 (I). The information appears with double the frequency in the embodiment shown in Figure 1 (III), but would be n-times the frequency in an embodiment with n subsidiary capacitors per storage capacitor element. The filter contains no more information, but is presented in a more favourable form. thus simplifying further processing. Thus smoothing with a RC-element can be sufficient to enable the signal to be satisfactorily reformed at its basic frequency i.e. to eliminate the higher frequency components produced in the sampling of the signal.

In the two exemplary embodiments illustrated, all the subsidiary capacitor elements are provided with non-reactive, parallel outputs, but it should be noted that this is not essential. In order to modify the filter properties, it is only necessary to provide at least one of the other subsidiary capacitor elements of a storage capacitor element to possess a non-reactive, amplifying output.

In many embodiments, including the described embodiments, the amplification factor of this further output corresponds to an assigned value of the pulse response of the filter.

WHAT WE CLAIM IS: 1. A transversal filter in which a charge transfer device is formed on one surface of a substrate of doped semiconductor material by a series of main capacitor storage elements, each comprising two or more subsidiary capacitor elements of the insulating layer or blocking layer type provided with a plurality of connections by means of which each main capacitor storage element can be connected to the same set of shift pulse trains mutually displaced in phase, to effect a charge shift, when operating, the connections enabling the provision of the mutually different phase shift pulse trains to the respective subsidiary capacitor elements of each main capacitor storage element in the same spatial order within each main capacitor storage element, said filter having an input terminal to which a signal to be filtered is applied, when operating, and in each main capacitor storage element, one of the subsidiary capacitor storage elements having a non-reactive, amplifying output which is connected to a common filter output terminal at which a filtered signal is supplied, when operating, the amplification factor of each said amplifying output having an assigned value in the pulse response of the filter to provide the desired filter function, and wherein at least one of said'main capacitor storage elements has at least one further non-reactive, amplifying output connected from another of its subsidiary capacitor elements to said common filter output terminal.

2. A transversal filter as claimed in claim 1, in which the amplification factor of said at least one further output has a respective assigned value in the pulse response of the filter to provide said desired, filter function.

3. A transversal filter as claimed in Claim 1 or Claim 2, in which there is a non-reactive, amplifying output connected to said common output from each one of said subsidiary capacitor elements.

4. A transversal filter substantially as described with reference to Figure 1 (II) or Figure 1(111).

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

Claims (4)

**WARNING** start of CLMS field may overlap end of DESC **. are provided with non-reactive, parallel outputs, but it should be noted that this is not essential. In order to modify the filter properties, it is only necessary to provide at least one of the other subsidiary capacitor elements of a storage capacitor element to possess a non-reactive, amplifying output. In many embodiments, including the described embodiments, the amplification factor of this further output corresponds to an assigned value of the pulse response of the filter. WHAT WE CLAIM IS:

1. A transversal filter in which a charge transfer device is formed on one surface of a substrate of doped semiconductor material by a series of main capacitor storage elements, each comprising two or more subsidiary capacitor elements of the insulating layer or blocking layer type provided with a plurality of connections by means of which each main capacitor storage element can be connected to the same set of shift pulse trains mutually displaced in phase, to effect a charge shift, when operating, the connections enabling the provision of the mutually different phase shift pulse trains to the respective subsidiary capacitor elements of each main capacitor storage element in the same spatial order within each main capacitor storage element, said filter having an input terminal to which a signal to be filtered is applied, when operating, and in each main capacitor storage element, one of the subsidiary capacitor storage elements having a non-reactive, amplifying output which is connected to a common filter output terminal at which a filtered signal is supplied, when operating, the amplification factor of each said amplifying output having an assigned value in the pulse response of the filter to provide the desired filter function, and wherein at least one of said'main capacitor storage elements has at least one further non-reactive, amplifying output connected from another of its subsidiary capacitor elements to said common filter output terminal.

2. A transversal filter as claimed in claim 1, in which the amplification factor of said at least one further output has a respective assigned value in the pulse response of the filter to provide said desired, filter function.

3. A transversal filter as claimed in Claim 1 or Claim 2, in which there is a non-reactive, amplifying output connected to said common output from each one of said subsidiary capacitor elements.

4. A transversal filter substantially as described with reference to Figure 1 (II) or Figure 1(111).