US2412995A

US2412995A - Amplifier of electromagnetic energy

Info

Publication number: US2412995A
Application number: US458060A
Authority: US
Inventors: Levy Maurice Moise
Original assignee: Standard Telephone and Cables PLC
Current assignee: STC PLC
Priority date: 1941-06-06
Filing date: 1942-09-11
Publication date: 1946-12-24
Anticipated expiration: 1963-12-24
Also published as: CH272951A; FR939357A; BE476621A; GB581683A

Description

' M. M. LEVY 2,412,95

AMPLIFIER OF ELECTROMAGNETIC ENERGY Filed Sept. ll, 1942 3 Sheets-Sheet FIG.

| A Hill, m I W FIG. 2.

I I +nof F/GI 3.

. l l o f 2f sf 4] I FREQUENCX FIG. 4.

is @Ef I 1 4 lwjm A TTOPNEY ea, 24, 1945. MLEVY 2,432,9Q

AMPLIFIER ELECTROMAGNETIOENERGY Filed.Sept. 11, 1942 5 Sheets-Sheet 2 l' is l- 42 M/VENTOI? 40 M Dec, 24 mg. M, LEW 2,4123% AMPLIFIER OF ELECTROMAGNETIC ENERGY Filed Sept. 11, 1942 3 Sheets-Sheet 3 [ll-4f IIF Aha/Am) A 7' TOR/VF Y Patented Dec. 24, 1946 'UNlTED STATES PATENT OFFIE AMPLIFIER F ELECTROMAGNETIC ENERGY Maurice Moise Levy, London W. C. 2, England, assignor to Standard Telephones and Cables Limited, London, England, a British company Application September 11, 1942, Serial No. 458,060 In Great Britain June 6, 1941 11 Claims.

tage is gained by amplifying signals of very low power level on account of the noise produced in the amplifier itself, or in the preceding detecting equipment, or otherwise, when this noise is of the same order of level or higher than that of the signal. It is also well-known that the noise produced may be reduced if the receiver is provided with means to eliminate frequencies not required for reproducing the signal. A common arrangement is for the receiver to be made selective by tuning it more or less sharply to the carrier frequency of the trains of waves being received. The more sharply it is tuned the lower will be the noise, but the more will the detected impulses be distorted, until ultimately the distortion may become so great that the amplitude of the impulses will be reduced.

The object of the present invention, therefore, is to improve the signal-to-noise ratio of an amplifying system without at the same time causing appreciable distortion of the signals being transmitted.

Another object of the invention is to provide a selective amplifier in which currents of only those frequencies necessary for defining the outline of the signal (of of narrow bands of frequencies in the immediate neighbourhood of these frequencies), are amplified, other currents being substantially not amplified. In particular the currents to be amplified may be some fundamental frequency and any desired number of the harmonics thereof.

According to the principal feature of the invention, the desired type of selective amplification is produced by means of a delay network, which may be associated with a thermionic amplifier, and the variations of the change of phase with frequency in the network are employed to obtain the required properties. The network may be connected in tandem with the amplifier, or may form part of a feedback path therein. In most cases the delay network may have appreciable attenuation without destroying the selectivity, and accordingly can be constructed in a simple and convenient form; for example, it may consist of an artificial line.

According to a preferred embodiment of the in vention an amplifier is provided with two separate feedback paths, the first of which produces a fixed negative feedback independent of frequency, and the second produces a feedback the amplitude of which is preferably constant, but the phase of which varies with frequency, this phase being preferably opposite to that of the negative feedback for the fundamental frequency, and for all its harmonics, or alternatively for all the odd harmonics.

In another embodiment the amplifier is provided with a single feedback path including a Wheatstone bridge of impedances, one of which includes the input circuit of a delay network. In this case, a feedback, variable both in amplitude and phase, is produced and this, feedback should preferably be zero for the fundamental frequency, and for all its harmonics, or for the odd ones only.

These objects and features will be more clearly understood by a reference to the following detailed description and the accompanying drawings, in which:

Fig. 1 shows some periodically repeated trains of waves;

Fig. 2 shows the frequency spectrum for such periodically repeated trains of waves;

Fig. 3 shows part of the gain frequency characteristic of an amplifier in accordance with the invention;

Figs. 4 and 5 show block schematics of amplifiers with feedback;

Fig. 5A shows a modification of a detail of F Figs. 6, 7 and 8 are vector diagrams of the feedback voltages;

fi Fig. 9 shows the schematic circuit of an ampli- Fig. 10 shows the schematic circuit of an artificial line;

Fig. 11 shows another vector diagram of feedback voltages; and

Figs. 12, 13 and 14 show circuits for compensating the effect of the attenuation of a delay network.

In the valve circuits of Figs. 9, 13 and 14 certain well understood coupling and operating arrangements' are indicated for completeness, but these arrangements are unessential as regards the invention, and may be modified to suit particular cases. Thus, grid and plate batteries are indicated by the usual symbols without implying that these supplies must necessarily be provided in this manner. Rp represents a plate circuit resistance of suitable value and g a grid resistance valves with any number of electrodes could be used, with appropriatearrangements; f

Fig. 1 shows the form of a number of repeated trains of high frequency waves of the type frequently used in obstacle detection. Such trains of waves are reflected from the obstacle and have to be received and detected with a suitable receiver. It will be assumed that the frequency of the waves is F and the frequency of repetition of the trains of waves is f. Y

In Fig. 2 is shown the frequency spectrum of such a system of repeated trains of waves and it consists of a central frequency F accompanied on either side by a large number of components whose frequencies may be designated by the formula Fi-nf. where n can be any integer. As n increases so the amplitude of the corresponding component tends on the whole to decrease and components may be eliminated without appreciably affecting the form of the received signal provided that n is not too small.

The usual method of receiving such a train of impulses is to amplify them with a selective high frequency receiver tuned to the frequency F, then to detect them by means of a low frequency detector. A certain amount of noise will be produced in the receiver and the detector, and also from atmospherics which will be picked up together with the impulses. If the impulses are at a very low power level the noise may be of the same or even higher level, and in order to minimize the effect of this noise it has been the practice to tune the receiver very sharply to the frequency F. This has the effect of cutting off the components shown in Fig. 2 on either side of the frequency F. a This may be done so long as the band of frequencies passed by the receiver is not made too narrow; for example, there will be some value no of it below which any further elimination of components will produce an undesirable distortion of the detected impulses. This is indicated in Fig. 2 by the dotted lines at F+nuj and F-nof. Accordingly, if the noise is still excessive no further advantage can be gained by reducing the band width because then the impulses received will become so distorted-that their amplitude is reduced.

The method characteristic of the present invention is to include in the receiver a selective amplifier so designed that it substantially only amplifies appreciably currents of frequencies actually contained in the impulses. This amplifier may be located in the high-frequency part of thereceiver before detection, in which case it will be designed to amplify all the frequencies Fin where n is zero or any integer up to and including no. Alternatively, the amplifier may be connected in a position in the circuit subsequent to the detector, when it will be designed to amplify all the frequencies of where n has all the same values except'zero. Thus the frequencies not concerned with defining the impulses will be largely eliminated and with them all the corresponding noise; so that the only noise which remains is that associated with frequencies immediately adjacent to those which are selectivelyamplified. A large improvement in signal-to-noise ratio will thus be obtained. 1 InFig. 3 is shown thegain frequency characteristic of such an amplifier, which amplifies the r up to a frequency of, say, 1 megacycle.

fundamental and all the harmonics. It will be seen to consist of anumber of very sharp peaks occurring at frequencies f, 2f, 3 etc. These peaks should be as narrow as possible in order that the maximum amount of noise may be eliminated.

This-kinder selective or comb amplification maybe produced by making use of the property of suitably designed delay networks by which the phase change obtained by propagation therethrough may be made to increase progressively with frequency, and so to pass through values which are multiples of 1r. This property may be employed either by direct transmission through the network, or by making use of reflections at the distant end, the impedance of which is mismatched, usually by open or short circuit. Without special means, however, good selectivity cannot be obtained ecause of the attenuation which accompanies the phase change in any practical form of delay network due principally to the resistance of the inductance coils, which cannot. generally be reduced sufi'iciently even by the use of extremely bulky coils. In accordance with certain features of the invention, however, it becomes possible to use delay networks with high attenuation without any loss of selectivity, and they may therefore be given convenient forms; for example artificial lines may be used.

In order to explain the reason for the low selectivity ordinarily produced by attenuation in the delay network, an example will be given.

One way in which the desired properties may theoretically be produced in any amplifier is to connect in the plate circuit of one of the valves, in parallel with the load, the input circuit of a delay network havin its output terminals short circuited, or left unconnected, whereby reflections will be obtained at the output terminals- As is well known, the input impedance of the delay network can then be designed to pass through an infinite value for some frequency f and for all its harmonics, and through a zero value for certain intermediate frequencies, provided the network is made up of pure reactances, or in other words, has no attenuation. As already explained this can never be achieved in practice and accordingly the impedance never becomes even approximately zero or infinite, and the selectivity is accordingly bad. This will be better appreciated from the following numerical example.

Suppose that a delay network is required to produce a delay of 200 microseconds, for use It could, for example, be made up of about 1000 sections of a simple low pass filter. If this be constructed of elements of reasonable dimensions, its attenuation could easily be 30 to 50 decibels. The selectivity would in this case be almost inappreciable; the diiference between the maximum and minimum impedance would in the worst case be only 0.3 and in the best case only 3%. According to one feature of the invention, the effect of the attenuation ofthe netw'orkin reducing the selectivity may be eliminated in the manner indicated in Fig. 12. f

A 'delay network D (which may, for be in the form of an artificial line) is provided 'with an input transformer I! havin a tapping point t on the secondary winding. The output of the network D is terminated by an impedance Z equal to the image impedance thereat, and an output transformer T2, the primary winding of which is connected in parallel with Z. The

example secondary winding of T2 has the lower end connected to the tapping point t, the upper end being connected to one output terminal 3, the other 4, being connected to ground. T! and T2 are supposed for simplicity to be ideal unity ratio transformers, though this limitation is not essential. Let V1 be the potential applied to the input terminals l and 2 at some given frequency, and let V2 be the corresponding potential developed across the impedance Z. Then where and a are respectively the attenuation and the phase change for transmission through the network, assuming, for simplicity, that it is symmetrical. Since the secondary winding of the transformer T2 is connected back to the tapping t on transformer Ti, the difference of potential V between the

terminals

3 and 4 will be A1.V1- -V2, where A1 is the fraction of the applied potential V1 tapped 01f at t. The choice of sign will depend upon the poling of the connections of the secondary winding of T2. Thus For the best results ,8 should be independent of frequency, and on should be proportional to frequency. If A1 be made equal to e-", then, taking the positive sign, V will assume a zero value whenever a is an odd multiple of 1r and a maximum value 2e-".V1 whenever a is an even multiple of 1r; Or taking the negative sign the zero and maximum values will be interchanged.

If therefore the arrangement of Fig. 12 be incorporated in an amplifier so that, for example, the

terminals

3 and 4 are connected in the grid circuit of a valve, currents applied to the input terminals l and 2 will be selectively amplified and it can be arranged so that, with the positive sign, for instance, the values of on which are even multiples of 11' occur for the fundamental frequency and all its harmonics.

If ,8 should not be independent of frequency, a suitable correcting network may be connected in series with D provided that it is designed so that the phase change is substantially independent of frequency, or an amplifier with appropriate gain characteristic could also be used.

'Another arrangement is shown in Fig. 13 in which the input transformer is removed and replaced by a tapped potentiometer, and the output transformer 'is replaced by the grid-cathode circuit of a thermionic valve. The operation will be practically as described with reference to Fig. 12, but taking the negative sign in the eX- pression for V.

Fig. 14 shows a modification of Fig. 13 employing a pair of similar valves. The plates are connected together and are fed through a common resistance Rp from the plate battery, and the control grid of one valve is connected to the tap t on the potentiometer, the control grid of the other being connected to one of the output terminals of the delay network D. In this case, the positive sign will be taken in the expression for V and the addition of the two terms will occur in the common resistance Rp.

In Figs. 13 and 14, the valves may be considered as part of the selective amplifier itself and may be followed by such other amplifying stages as may be found convenient.

It will be understood that the arrangements shown in Figs. 12, 13 and 14 are only three examples of the application of this particular meth- Ill od of compensating the attenuation of the network. Many other arrangements conforming to the principles explained in connection with Fig. 12 are obviously possible.

According to another feature of the invention still better selectivity may be obtained with a delay network having attenuation by associating it with a feedback path in the amplifier.

An embodiment employing this arrangement is shown schematically in Fig. 4. An amplifier A has a pair of input terminals I and'a pair of output terminals 2. It is also provided with two

separate paths

3 and 4, the first of which produces negative feedback N which is constant as the frequency varies and the second produces a feedback which is constant in amplitude but varies continuously in phase with frequency, and may contain a delay network D.

Suppose that Eis the input voltage applied to the grid by the signal. Let G be the ratio which defines the gain of the amplifier in the absence of feedback, then the output voltage will be GE in the absence of feedback. Let the constant negative feedback be GEFM and let the variable feedback be It can be shown that the eifective gain of the amplifier with both the feedbacks operating will be defined by the ratio If H is chosen equal to F, then the denominator of this expression will be a minimum equal to 1 whenever the angle 5 is a multiple of 271', and accordingly the gain of the amplifier will then be a maximum defined by the ratio G. If GF and GH are chosen to be large compared with 1, then a very small change in the angle due to a small frequency change will make a large increase in the denominator of the above expression; in other words, the gain of the amplifier will be reduced by a large amount, and the selectivity will be good. This can be seen more clearly with reference to Figs. 6, '7 and 8.

Considering first Fig. 6, 0A is a vector representing th feedback GF and this vector is drawn in the negative direction since GF is negative with respect to the input voltage E. The vector AB represents the variable feedback GH and makes an angle with the positive direction A0. The resultant of these two vectors is the vector OB which is equal to J and at an angle 0 with A0. As has already been mentioned, the vector GH has been assumed to be constant, for simplicity, and accordingly as the frequency varies the point B will describe a circle of which A is the centre. The maximum and minimum values of the vector J are given by OX and CY. The maximum gain of the amplifier will occur at the point Y and the minimum gain at the point X, but unless OY is very small compared with OX the variation in the gain of the amplifier will be small. Accordingly, the arrangement of Fig. 7 is preferable in which the vectors OA and AB have been made equal, in other words, making F equal to H. In this case the vector OY becomes zero and the selectivity is accordingly high if OX is large compared with 1. As the frequency is varied continuously in the same direction the point B rotates continuously around the circle and the corresponding gain produced will have the form of Fig. 3.

Fig. 8 shows another arrangement in which the vector GF is less than the vector GI-I. In this case thevector OB is never zero and can have a positive component at certain frequencies. Thus it will be seen that in Fig. 6 the feedback is never zero, but always. has a negative component; in Fig. 7 it is zero at each of the harmonic frequencies and has a negative component at all other frequencies; and in Fig. 8 the feedback sometimes has a positive and sometimes a negative component.

Fig. 9 shows an example of a three stage ampliher in which the principles just described are applied. The signal is applied to the grid through a transformer Ti and the constant negative feedback is obtained by connecting the cathode of the first Valve to a resistance BC in the cathode circuit of thethird valve. The variable feedback is produced by connecting a delay network or artificial line L to the plate circuit of the third valve, through a transformer T2, for example. The output of the artificial line is terminated by a potentiometer P, the moving contact of which is connected to the grid of the first valve through the secondary winding of the input transformer Ti. The artificial line L should preferably have an attenuation substantially constant over the frequency range concerned and the phase change should also preferably be substantially proportional to the frequency. However, if the artificial line L has a variable attenuation, it would be possible to compensate this by means of a correcting network or with an amplifier having a suitable gain characteristic. The electrical length of the artificial line L should'also preferably be chosen so that the phase change for the fundamental frequency which has to be amplified is a multiple of 211-. Thus for every harmonic of the fundamental frequency, it will also have a phase u change which is a multiple of 211'. Fig. 7, the feedback GF produced by the resistance RC in Fig. 9 will be represented by the vector A, and the feedback GI-I produced by the ar- Referring to tificial line L will be represented by the vector AB. I

H can be adjusted to be equal to F by adjusting the potentiometer P, for example, or by other means. The sum of the feedbacks produced by resistance RC and by the artificial line L is represented by the vector OB in Fig. 7. The output may be taken from a third winding of the transformer T2 as indicated, or in several other ways which may be more convenient in certain cases.

It will be seen by this arrangement the artificial line can have relatively large attenuation and accordingly it could be constructed with elements of reasonable dimensions. For example, suppose the amplifier has a gain of 60 decibels in the absence of feedback, and suppose that the attenuation of the artificial line is decibels, the value of the feedback GH can be such as to correspond to 40 decibels. The difference between the maximum and the minimum amplification will be about 46 decibels, because the maximum negative reaction OX in Fig. 7 is equal to G(F+H) and since these two vectors have been chosen to be equal, the vector OX will be 6 decibels greater than GH.

It will be clear from the explanation just given that the attenuation of the artificial line L is compensated by part of the gain of the amplifier.

This is accordingly another means whereby an artificial line with high attenuation may be used as a delay network without any reduction in the selectivity of the amplifier.

In Fig. 5 is shown a different arrangement for producing a variable feedback in the amplifier. In this case the amplifier A has a pair of input terminals I to which a signal voltage E is applied and a pair of output terminals 2 where an amplified signal voltage EG appears in'the absence of feedback. In this case there is only one feedback path which involves a, Wheatstone bridge KLMN; the diagonal points KM are connected to the output terminals 2 and the diagonal points LN provide the desired feedback. The arms KL, LM and MN are composed of impedances Z3, Z4 and Z2, respectively and the arm KN is composed of an impedance Z! in series with the input terminals of the delay network D, the output terminals of which are left unconnected for example. This delay network should preferably have substantially no attenuation. The impedances Zl to Z4 will preferably be pure resistances and in the following explanation this will be assumed.

Referring to Fig. 11, the vector KM represents the voltage applied to the diagonal points K, M of the Wheatstone bridge. It is composed of two vectors KL and LM in the same line which rep-- resent respectively, the voltage drops across Z3 and Z6. The vectors KQ and QP represent, respectively, the voltage drops across Zl and Z2, which since ZI and Z2 are preferably bothpure' resistances, will be in the same direction. The vector PM represents the voltage drop across-the delay network D which will be at right angles to the vecto KQP because the network will have an impedance which is substantially a pure react'ance at all frequencies, the output being opencircuited or short-circuited. The point P will thus describe a circle with KM as the diameter, as the frequency changes. From the point Q is drawn a vector QN parallel and equal to PM so that NM will be equal to the drop across Z2. LM being equal to the drop across Z 'l, it follows that the resultant of LM and MN, namely, LN, must be the difference of potential between L and N, and accordingly must be the feedback voltage. It can be seen that as P moves round the circle, QN will cut the vector LM in a fixed point 0 and the point N will describe another circle whose diameter is OM. Thus the vector LN will have the same properties as the vector J in Fig. 6. By adjusting the relative values of Z! and Z2 or Z3 and Z4 the point 0 may be moved along the vector LM, and, for example, can be made to coincide with L. In this case a result equivalent to Fig. '7 is produced and this will be the preferred arrangement.

As already explained-however, the desired result can only approximately be attained in practice because of the attenuation of the delay network D which cannot be reduced to zero. Accordingly, a variation of the method may be adopted which is shown in Fig. 5A, In this case the diagonal of the Wheatstone bridge KN is composed of an impedance Zi, connected in parallel with which are the input terminals of the delay network D which in this case is supposed to have attenuation, and may thus be in the form of an artificial line. It is desirable that the impedances Zl, Z2, Z3 and Z4 should be chosen so that the delay network may bev terminated in its characteristic impedance in order to avoid undesirable reflections at the input terminals. The ex, planation of the operation of the circuit in terms 9 of a vector diagram is'in'this case rather complicated and can be more easily understood in the following way. Assume first of all that the delay network is infinite in length. The impedances in the bridge can be adjusted so that a fraction of the output voltage is transmitted as a feedback to the input of the amplifier and so that the feedback is negative. This feedback then is equivalent to the feedback GF in Fig. '7. Now assume that the network has a finite length, reflection will take place at the output terminals and the reflected current will be transmitted back to the input and a part will be transmitted to the input of the amplifier. This reflected current then plays the part of the vector GH in Fig. '7.

By adjusting the impedances in the bridge the effects corresponding to Figs. 6 and 3 can also be-produced.

Experience shows that a circuit combining the principles of Figs. 5 and 5A gives good results in practice if the values of the impedances of the bridge are correctly chosen.

The delay network 01' artificial line used for these circuits should preferably have a constant attenuation and a phase change which varies in proportion to the frequency, as already stated above. One form of the delay network D may be similar to that of an artificial line such as is shown in Fig. consisting of a number of sections having series elements of inductance and resistance and shunt elements of capacity and resistance, and having mutual inductance between adjacent sections, as indicated. Various other forms are, however, also possible and Fig. 10 has been given just as an example.

The artificial line is only one example of a delay network that can be used in these circuits. Other types of network will occur to those skilled in the art, as also will other arrangements in accordance with the principles of the invention. The arrangements shown in Figs. 4, 5, 9, 12, 13 and 14 have been given by way of illustration, and the invention is not intended to be limited thereto.

What is claimed is:

1. A selective thermionic amplifier circuit comprising a thermionic valve amplifier, a delay network having attenuation and terminated at its output terminals with an impedance substantially equal to its image impedance thereat, means for Coupling said output terminals to an input circuit of said amplifier, means for applying said periodically repeated impulses to input terminals of said delay network, and circuit means for algebraically adding the output voltage of said delay network to a proportion of the input voltage applied to its input terminals, in which the aid circuit means comprises afirst transformer having a secondary winding connected to the input terminals of the said delay network, said secondary winding being provided with an intermediate tapping point, and a second transformer, the primary winding of which is connected to the output terminals of the said delay network, one terminal of the secondary winding of the said second transformer being connected to the said intermediate tapping point.

2. A selective thermionic amplifier circuit comprising a thermionic valve amplifier, a delay network having attenuation and terminated at its output terminals with. an impedance substantially equal to its image impedance thereat, means for coupling said output terminals to an input circuit of said amplifier, means for applying said periodically repeated impulses to input terminals of said delay network, and circuit means for algebraically-adding the output voltage of said delay network to a proportion of the input voltage applied to its input terminals, in which the said circuit means comprises apotentiometer, having an intermediate tapping point, the said potentiometer being connected across the input terminals of the said delay network, and said thermionic valve amplifier comprises a thermionic valve, the control grid of which is connected to an output terminal of the said delay network, and the-cathode of which is connected to the said intermediate tapping point.

3. A selective thermionic amplifier circuit comprising a'thermionic Valve amplifier, a delay network having attenuation and terminated at its output terminals with an impedance substantially equal to its image impedance thereat, means for coupling said output terminals to an input circuit of said amplifier, means for applying said periodically repeated impulses to input terminals of said delay network, and circuit means'for algebraically adding the output voltage of said delay network to a proportion of the input voltage applied to its input terminals, in which the said circuit means comprises a potentiometer having an intermediate tapping point, the said potentiometer being connected across the input terminals of the said delay network, and said thermionic valve amplifier comprises two thermionic valves, the plates of which are fed from the plate supply through a common resistance, the control grid of one of the said valves being connected to the said tapping point and the control grid of the other valve being connected to an output terminal of the said delay network.

4. A selective thermionic amplifier circuit for amplifying periodically repeated electrical impulses comprising a thermionic amplifier having input and output circuits, a first feed-back circuit extending between said output and input circuits including a delay network arranged to vary the phase of the feedback voltage dependent upon frequency and a second feedback circuit extending between said output and input circuits and arranged to produce a constant negative feedback voltage independent of frequency.

5. A selective thermionic amplifier circuit according to claim 4 in which the said delay network has an attenuation substantially independent of frequency and in which the amplitud of the feedback voltage produced thereby is equal to the said constant negative feedback voltage produced in the said second feedback path, whereby the effect of the said attenuation tending to reduce the selectivity of the said amplifier may be Substantially eliminated.

6. A selective thermionic amplifier circuit comprising a thermionic amplifier having input and output circuits, a feedback circuit extending between said output and input circuits and phase changing means in said feedback circuit arranged to produce an overall amplifier gain which is a maximum for the fundamental frequency of said impulses and for a plurality of harmonics of said fundamental frequency, in which said feedback circuit includes a Wheatstone bridge, three arms of which consist of impedances and the fourth arm consists of a fourth impedance and the input circuit of a delay network, opposite diagonals of said bridge being connected respectively to said input and output circuits.

7. A selective thermionic amplifier circuit according to claim 4 in which the said delay network comprises an artificial line.

8. A selective thermionic amplifier circuit for amplifying periodically repeated electrical 'impulses comprising a thermionic valve amplifier, a delay network having attenuation and terminated at its output terminals with an impedance substantially equal to its image impedancethereat, means for coupling said output terminals to an input circuit of said amplifier, means for applying said periodically repeated impulses to input terminals of said delay network, and circuit means for algebraically adding the output voltage of said delay network to a proportion of the input voltage applied to its input terminals.

9. A selective thermionic amplifier circuit com-. prising a thermionic amplifier having input and output circuits, a feedback circuitextending between said output and input circuits and phase changing means in said feedback circuit arranged to produce an overall amplifier gain which is a maximum for the fundamental frequency of said impulses and for a plurality of harmonics of said fundamental frequency, wherein said feedback circuit includes a Wheatstone brid e, three arms of which consist of impedances, and the fourth arm consists of a fourth impedance in series with the input of a delay network.

10. A selective thermionic amplifier circuit comprising a thermionic amplifier having input and output circuits, a feedback circuit extending between said output and input circuits and phase changing means in said feedback circuit arranged to produce an overall amplifier gain which is a maximum for the fundamental frequency of said impulses and for a plurality of harmonics of said fundamental frequency, wherein said feedback circuit includes a Wheatstone brid e, three arms of which consist of impedances and the fourth arm consists of a fourth impedance in parallel with the input of 'a delay network.

11. A selective thermionic amplifier circuit according to claim 4 wherein said delay network comprises an artificial line, means for coupling the input terminals of said line to the output circuit of said amplifier and means for coupling the output terminals of said line to the input circuit of said amplifier.

- MAURICE MOISE LEVY.