US2209955A - Wave translation system - Google Patents

Description

Aug. 1940- H. 5. BLACK 2.2o9,955

I WAVE TRANSLATIdN SYSTEM 7 Sheets-Sheet 1 Filed Dec. 5, 1936 VAYAVAF INVENTOR H. 5. BLACK BVJMW A 7' TORNEV 1940- H. 5. BLACK 2,209,955

WAVE 'mmswn'on SYSTEM Filed Dec. 5, 1936 7 Sheets-Sheet 2 FIG. 6

SULPHIDE\ A TTORNE V Aug. 6, 1940. s, BLAcK 2209,9555

WAVE TRANSLATION SYSTEM Filed Dec. 5. 1936 7 Sheets-SheetA FIG. [4 I44 0km orF/ksr runs a -//vs/0c AND OUTS/DE 2 FEEDBACK 00 E B fl 20 muvsulrre'o o n 1 I 200 1000 10,000 100,000 200,000 H g. g zz g FREQUENCY-IN crass PER SECOND 0 Wm sw m ATTORNEY Aug. 6, 1940. 5, BLACK 2.209,955

WAVE TRANSLATION SYSTEM Filed Dec. 5, 1936 7 Sheets-Sheet 5 \lllllllllllll IIHIIIIIIIII II 350 1 I lNVE/VTOR h. 5. BLACK AT TORNEV Aug. 6, 1940.

Filed Dec. 5, 1936 7 Sheets-Sheet 6 Q DE GRE E 5 PHASE SHIFT AROUND FEEDBACK PATH kuamwk E W8 bath .mwviu wukviu MNMQQMQ O m m w w m m w m m I00 l2u Q-DEGREES PHASE SHIFT AROUND FEEDBACK PATH lNl/ENTOR H. 5. BLACK FIG. I68

INNER PATH (us Y OUTER (3'PATH W W w OPERATING RANGE LOG OF FRE OUE NC Y F/G. I54

40/ INNER LOOP LOG OF FRE OUE NC Y ATTORNEY Aug. 6, 1940. I 5, BLACK 2,209,955

' WAVE TRANSLATION SYSTEM Filed Dec. 5, 1936 7 Sheets-Shget 7 GAIN FIG. /7A

lufl I/N db INVENTOR H. 5. BLACK BVJTMALQM ATTORNEY Patented Aug. 6, 1940 UNITED STATES WAVE TRANSLATION SYSTEM Harold S. Black, Elmhurst, N. Y., asslgnor to Bell Telephone Laboratories, Incorporated, New York, N. Y., a corporation of New York Application December 5, 1936, Serial No. 114,390- 17 Claims. (01.11844) This application is a continuation in part of my copending applications Serial No. 606,871 filed April 22, 1932, for Wave translation system, and Serial No. 663,317, filed March 29, 1933, for Wave translation systems which issued as U. S. Patent 2,102,671 December 21, 1937 and U. S. Patent 2,131,365 September 27, 1938 respectively.

This invention relates to wave translation and aims to control transmission of waves with regard to amplitude or phase relations or both.

The invention also aims to control transmission properties of wave translating systems, as for example, systems involving means for increasing the power level of waves.

1 Representative objects of the invention are .to

control modulation, stability, impedance relations, wave reflections, cross-talk, resistance noise, signal-to-noise ratio, heat transfer, temperature, transmission efficiency, gain-frequency 0 and phase relations in such systems.

A feature of the invention relates to effecting such control by-feedback.

Objects of the invention are also to control feedback and facilitate application of feedback.

25 The feedback may be, for example, feedback of a portion of the output wave of a system in gain-reducing phase and in amount sumcient to reduce distortion below the distortion level withcut feedback. Such feedback is disclosed, for

30 example, in the above-mentioned copending applications and in my article on Stabilized feedback amplifiers, published in Electrical Englneering.v January 1934, pages 114 to 120.

In one specific aspect the invention is an am- 35 pl fier having a path which produces such feedback, the input (or output) end of the amplifying path being joined to the feedback path and the incoming (or outgoing) circuit of the amplifier by an interconnecting network that has branches 40 w th. mutual impedance and causes the input (or cutout) impedance of the amplifier to stabilize around a fixed value which, with considerable amounts of feedback, is independent of variations within the amplifier, its gain, or the amount 45 of feedback. I

The interconnecting circuit at the amplifier inut, or the interconnecting circuit at the amplifier output, or each interconnecting circuit, may be for instance a hybrid coil or bridge transformer with four pairs of terminals-one pair connecting to the incoming (or outgoing) lines, a second pair to the feedback path, a. third pair to the input (or output) circuit of the amplifying element, and the remaining pair to a two- 55 terminal impedance, thehybrid coil net.

Considering, for example, the case in which the hybrid coil is .a three-winding transformer with a line winding connected between the line and the hybrid coil net, and a feedback winding connected between the feedback path and the hybrid coil net (the net thusbeing across the bridge points of the hybrid coil) it is shown hereinafter that, in the case of either an equality ratio or an inequality ratio between the turns of the line and feedback win as, the input or output 10 impedance of the am; according to whether the hybrid coil is the input hybrid coil or the output hybrid coil, can be made to equal the impedance of the hybrid coil net multiplied by one plus the turns ratio, regardless of whether the hybrid coil is balanced in the sense of giving passive conjugacy. Thus, for any chosen turns ratio between the line and feedback windings, the desired amplifier input or output impedance can be realized by choice of the impedance of 2 the hybrid coil net.

This same turns ratio controls the transmission loss from the incoming line to the amplifying path, in the case of the input hybrid coil, or I from the amplifying path to the outgoing line, in the case of the output hybrid coil; and if the turns ratio exceeds .unity say by several times,

varying the amplifier impedance by changing the net impedance will not materially change the transmission loss, which moreover can readily be kept below a few tenths of decibel, and, further, the amount of feedback will be influenced but little by the line impedance facing the amplifier input or output impedance and likewise the transmission between the line and the amplifier will be but slightly influenced by the magnitude of the impedance of the feedback path seen from the input hybrid coil or the output hybrid coil.

The input and output impedances of the amplifier are independent of one another and may have the same value or different values.

By making the input hybrid coil so that without feedbackthe amplifier input impedance is high compared to the impedance of the incoming circuit instead of being a match for the latter impedance, the ratio of signal to resistance-noise in the amplifier output energy can be made as much as three decibels greater than for the, matched impedance condition; and with proper choice of the value of impedance for the net of the input hybrid coil, feedback can make the amplifier input impedance match the impedance of the incoming circuit while the improvement in the ratio of signal to resistance-noise is retained. Indeed, the improvement that feedback can produce in the ratio of signal to resistancenoise is by no means limited to three decibels, and, as will now be explained, by the aid of feedback action an impedance from which to work the amplifier can be created which does not produce resistance-noise.

The relation of the amplifier input or output impedance to the impedance of the input or output hybrid coil net can be used for transforming impedances, i. e., for obtaining from the net impedance an impedance equal to the net impedance multiplied by a constant (unity plus the turns ratio), provided the amount of negative feedback is sufiicient. Thus, the impedance obtained is an enlarged copy of whatever impedance is 'used as the hybrid coil net (the impedance across the bridge points of the hybrid coil) and the latter impedance may be of any suitable type. For example, it may be an ordinary impedance such as a resistance, an inductance, a capacity or any complex form of (two-terminal) impedance constructed of resistances, inductances and capacities.

I have discovered that feedback action can produce resistances or generalized impedances free from resistance-noise-that feedback action can transform ordinary resistances (or generalized impedances) to resistances (or generalized impedances) that are free of all noise, including thermal agitation. For example, if the net across the bridge points of the input hybrid coil be a resistance, the feedback can transform this resistance by producing an enlarged copy of it as the amplifier input impedance, this copy having the remarkable property of freedom from resistance-noise. Thisproperty of freedom from resistance-noise or thermal agitation electrometive force obtains likewise for the case in which the input impedance is a generalized impedance (which may include inductance or capacity or both) produced from a corresponding generalized impedance across the hybrid coil bridge points by the feedback action. Thus, any desired simple or complex (two-terminal) impedances (resistances,

capacities, inductances or generalized impedances) with the extraordinary property that they are theoretically free from thermal agitation electromotive force and practically free at least to a considerable degree, can be obtained from similar smaller impedances which may if desired be of ordinary type, or which, on the other hand. may themselves in turn be impedances similarl obtained free of resistance-noise by feedback action.

Any physically realizable impedance can thus be reproduced free of resistance-noise; and of such impedances free from resistance-noice can be constructed all manner of impedance networks or active or passive transducers free from resistance-noise, for example, filters, equalizers, phase shifters, delay distortion correctors, impedance correctors, artificial cables or lines, amplifiers and systems in general.

Such impedances and networks that do not produce resistance-noise have a wide scope of practical application. For instance, in the case of the amplifier referred to above, a transmission equalizing network constructed of such impedances can be connected between the incoming line and the input hybrid coil; The incoming line may be, for example, a coaxial system or other cable delivering signals whose lower limit of transmission level is fixed by the cable resistance noise; the equalizer may be, for example, a constant resistance equalizer; the hybrid coil may be such that the amplifier input impedance without feedback would be high compared to the equalizer impedance from which the amplifier works; and the hybrid coil net may be a resistance or impedance of such value that the feedback matches the amplifier input impedance to the equalizer impedance from which the amplifier works. Notwithstanding the fact that the equalizer may include elements having resistance, the equalizer and amplifier need not materially increase the ratio of resistance-noise to signal; and thus the desired equalization and amplification of the signal can be accomplished without decreasing the ratio of signal to resistance-noise.

The resistances and impedances free from resistance-noise render possible large improvements in signal-to-noise ratio (by no means limited to improvements of three decibels) and the penetration of substantial amounts, theoretically as great as desired, below the noise level heretofore considered as an inevitable, natural, impenetrable limit established by thermal agitation.

Features of the invention include resistances, generalized impedances, networks of generalized impedances, and active and passive transducers, free from resistance-noise; and a further feature of the invention is production of such devices by feedback action.

I have discovered that feedback action can abstract heat from a body. When a resistance is connected to an amplifier, feedback action can be made to abstract heat from the resistance or cool it. For example, if an electric conductor or resistance be connected across resistance of the type described above as free from -resistancenoise, the effect of making the connection is to abstract heat from the ordinary resistance or cool it, the ordinary resistance receiving no energy from the other resistance but giving up energy of thermal agitation to the other resistance in the form of an electric current. To observe the cooling effect the resistance to be cooled can be heatinsulated. If it is not insulated, the small losses due to thermal agitation are readily replaced from the relatively vast reservoir of heat surrounding the unit.

The variation of the amplifier output impedance by changing the impedance of the net of the output hybrid coil can be accomplished without materiall changing the impedance into which the output tube (or the amplifying path) works; and thus the output tube, though its impedance may diifer from its optimum load impedance, can be worked into an impedance having substantially the optimum value and yet be made to appear from the outgoing line to be equal to the impedance of the outgoing line. In other words, the feedback cannot only transform the net impedance to the desired value of amplifier output impedance, but can do so without materially affecting the impedance into which the output tube works; and this ability of the feedback to change the difference between two impedances (such as the tube impedance and the impedance of the outgoing line) by a different amount for one direction of transmission between them than for the other direction, makes it possible to work the tube into its optimum load impedance and at the same time match the amplifier output impedance and the line impedance, without entailing the transmission loss entailed in doing this without feedback, (as for instance in shunting across the tube impedance aresistance which in combination with this parallel tube impedance is equal to twice the optimum load impedance of the tube and connecting the resistance and the tube to the line with a transformer that changes the difference between the impedances which it connects the same amount for both directions of transmission).

Since, as noted above, with considerable amounts of feedback the input and output hybrid coils can render the amplifier input and output impedances independent of the amount of feedback, the amplifier gain can readily be varied without varying the amplifier input and output impedances, byvarying the attenuation of the feedback path, as for example by an adjustable series or shunt resistance in the feedback path. The resistance may be, for instance, a thermoresponsive resistance, as for example a silver sulphide resistance adjustable by control of its tem-, perature. If desired, the attenuation change made in the feedback path for gain control can be made automatically, as for example in response to transmission level. For instance, a silver sulphide resistance unit may be connected in series in the feedback path, to render the amplifier a volume limiter by virtue of decrease in resistance of the silver sulphide due to heating of the silver sulphide resulting from increase of current through it in response to increase in output Voltage of the amplifying path. 4

Specific aspects of the invention also embrace feedback systems, including multiple feedback systems and systems involving repetition of the feedback process, with various forms of hybrid coil feedback connections.

Other objects and aspects of the invention will be apparent from the following description and claims.

In the drawings, Fig. 1 shows an amplifier of a specific form referred to above;

Fig. 2is a diagram for facilitating explanation of operation of such amplifiers;

Fig. 3 shows a specific form of network which may be used in such an amplifier to match the amplifier input or output impedance to the impedance of a specific type of attached cable circuit;

Fig. 4 shows a cable terminated in an equalizer working into such an amplifier, the equalizer and amplifier being constructed, in accordance with the invention, to avoid increasing the ratio of resistance-noise to signal;

Fig. 5 illustrates cooling a resistanceby feedback action produced for example by such an amplifier;

Fig. 6 shows a gain control for fier;

Fig. 7 shows a volume limiting circuit embodying a specific aspect of the invention;

Figs. 8 to 10 show feedback amplifier systems with a single hybrid coil or bridge transformer network interconnecting the input and the output such an ampliof an amplifying element with an incoming circuit and an outgoing circuit;

Fig. 11 shows a multiple feedback amplifier with, a hybrid coil feedback connection;

Fig. 11A shows an amplifier which is a modification of that of Fig. 11;

Fig. 12 shows a two-way transmitting, single loop feedback system with hybrid coil feedback connections through two feedback paths having a common portion;

Fig. 13 shows a triple loop feedback system with a hybrid coil feedback connection;

Figs. 14, 15 and 16 show feedback amplifiers with hybrid coil feedback connections and with repetition of the feedback process;

Fig. 15A shows gain-frequency characteristics of the amplifier of Fig. 15;

Fig. 153 shows a type of feedback connectionused in Fig. 15, for facilitating explanation of operation of such connections;

Figs. 16A and 163 show curves facilitating explanation of tie operation of the amplifier of Fig. 16;

Figs. 17 and 17A show curves for facilitating design of feedback systems;

Fig. 18 shows a specific form of amplifier of the general type of the amplifier of Fig. 1;

Fig. 19 shows an amplifier or system with feedback through hybrid coil connections that are modifications of those of Fig. 1.

The amplifier of Fig. 1 may be a stabilized feedback amplifier of the general type disclosed, for example, in the copending applications and published article mentioned above. It comprises an amplifying path or element shown as including tandem connected vacuum tubes I and 2, and comprises a feedback path I shown as including a transmissioncontrol network 3 of generalized impedances. The amplifying path or element may be referred to as the circuit, and the f'edback path may be referred to as the p-circuit, the significance of a and 5 being as indicated in the applications and article just mentioned. The network 3 may be referred to as the p-circuit network.

An input hybrid coil 5 couples the incoming circuit 6 and the feedback path I to the input end of the amplifying path; and an output hybrid coil 1 couples the output end of the amplifying path' to the outgoing circuit 8 and the feedback path. One of the important advantages of this type of feedback circuit, with considerable amounts of feedback, is that the input and output imped'ances of the amplifier stabilize around fixed values. that are independent of variations within the amplifier, its gain, or the amount of feedback, regardless of whether, in the passive condition of the amplifier, the hybrid coils are balanced, or in other words, regardless of whether the impedance Zn of the hybrid coil net (i. e., the impedance 9 or 60 across the bridge points of the hybrid coil) .is such as to give passive conjugacy (i. e., conjugacy in the absence of feedback) between the line and the feedback path. vInasmuch as the case of relatively large amounts of feedback (relatively large values of p) in a case of great practical importance, the derivation of the amplifier input impedance for that case ,is

given by reference to Fig. 2. The equation for the output impedance is similar.-

In Fig. 2, for simplicity, the hybrid coil is shown in the unsymmetrical form (i. e.,as unbalanced to ground). The source of electromotive force E and impedance C represent or replace line 6 of Fig. 1. The capacity Go represents the effective grid-cathode capacity of tube l. The voltage across the input terminals of the amplifieris designated e. The input impedanceof the amplifier is designated Zn. The impedance of the hybrid coil net is designated ZN. The numbers of turns in the line and feedback. windings of the hybrid coil are designated m and m, respectively. The turns If [15 is very large compared with 1, there will be current i2 return to the feedback side of the coil suflicient to reduce the flux in the coil 5 by a factor equal to A '-#B r This means that the voltage across terminals I, 2 will approach zero as compared to the value of e. The fed back current, 2'2, to accomplish this flux cancellation will be equal to If the voltage across I, 2' is negligible, e approaches (i1i2). ZN=drop across the network.

.'.e=i,Z i,(l 5% Z 2 and, for large amounts of feedback,

E Z p (I 2)Z 1v It is thus seen that, for large values of 49 zF=(m)zN:kzN,

is representing the constant This gives the very valuable result that the impedance of the amplifier is equal to the impedance of the net connected across the hybrid coil bridge-points multiplied by one plus the turns ratio of the' equality ratio or usually inequality ratio hybrid coil. Thus, assuming #5 is large, the impedance of the amplifier can be made to approach what is wanted as closely as can the net. Using this procedure, amplifiers have been built and used having remarkably-good impedances. It has been observed that the input or output impedance of the amplifier can easily be varied, for example, in the ratio of 100:1 by merely changing the impedance of the net. Moreover, if the hybrid coil is an inequality ratio hybrid coil with the turns ratio t sufficiently large, so that n1 n2 say by several times, then varying the impedance in this manner will hardly vary the input or output loss at all, and further, each of these losses can readily be kept below a few tenths of a decibel (instead of the usual three decibels for an equality ratio hybrid coil). The ability to vary the amplifier input or output impedance (or both) in such an easy manner without much affecting the transmission is a highly desirable feature. Zn is a surprisingly accurate copy of ZN, either enlarged or attenuated according as m is greater than or less than 112. Thus, if ZN is for example a capacity, the (input or output) impedance of the amplifier is a capacity, etc.

Regarding conjugacy relations with the feedback through hybrid coils, the transmission between the (incoming or outgoing) line and the and the feedback path is affected by the impedance of the (incoming or outgoing) connecting line; and as a corollary the amount of feedback obtained for any setting of the [El-circuit network such as network 3 is somewhat dependent upon the impedance of the connecting line. However, by making the hybrid coils inequality ratio hybrid coils with one of the quantities m or m sufliciently exceeding the other, as for example, with one of these quantities several times as large as the other, the effect of the line impedances upon the amount of feedback can be made negligibly small and likewise the effect that the value of the impedances of the feedback path as seen from the hybrid coils has upon the transmission between the incoming or outgoing line and the amplifying path can be made negligibly small.

Thus, regarding conjugacy relations with feedback through hybrid coils, an indication of the degree of conjugacy may be obtained by noting the change in if as the p-circuit impedance is changed. Changing the B-circuit impedance (generator impedance producing in) will change the value of p. But it can be seen from equation (2) above, that Zr is dependent on due to changes in all are nil, or Zr is independent of /.I.,B. Therefore, regardless of the turns ratio,

the input impedance presented to the connecting circuit is rendered independent of the feedback path. But the converse is not true, i. e., the value of ,ufi is not independent of the connectin circuit, unless the hybrid coil possess a passive impedance balance.

The incoming and outgoing lines are in conjugacy with the feedback path when the proper hybrid coil balancing nets 9 and ID are used. For example, the amplifier output impedance is not a function of the impedance of the feedback path by direct transmission, but is controlled by the apparent plate impedance resulting from the feedback that the impedance of the feedback path provides. That is, feedback causes the apparent plate impedance to approach such a value that the output hybrid coil will be in dynamic balance. It is this property of hybrid coils, in conjunction .with sufficient negative feedback, that causes the nets 9 and Ill of the input and output hybrid coils to determine the amplifier input and output impedances.

Since either an input hybrid coil or an output hybrid coil can render the amplifier input and output impedances independent of one another (regardless of whether the hybrid coil is in passive balance), it results that if an input hybrid coil is used, or an output hybrid coil is used, or both are used, the amplifier input and'foutput impedances can be given equal values or altogether different values, at will.

Particularly in the communication field, there are many applications of amplifiers requiring the input (and also the output) impedance of the amplifier to match the impedance of the circuit it joins. If the input impedance of the amplifier is required to match, instead of being permitted to greatly exceed the impedance of the cable or circuit to which it connects, then for the same insertion gain the amplified noise that is due to thermal agitation and appears in the output of the amplifier will be about three decibels more than for the case in which the amplifier input impedance is relatively high, assuming a proper and well designed input circuit. However, with an input hybrid coil for example as shown in Fig. 1 or Fig. 2, the coil (and its terminating impedance,

if any, across its winding attached to the grid and cathode of the first tube) can be, chosen so that the amplifier input impedance without feedback is high, and at the same time the value ZN of the impedance of the net of the input hybrid coil can be chosen so that the negative feedback will cause the amplifier input impedance to be improved and stabilized around a proper value that will match the input connecting circuit. As a result, when the insertion gain is the same as that of an amplifier without feedback but whose input impedance is high, the amplified resistance-noise at the output of the two amplifiers will be the same (because in either amplifier the noise in question depends upon the resistance component of the passive impedance without feedback between the grid and cathode of the first tube), and yet the feedback amplifier will have a matched impedance instead of a very high impedance. For systems of this character whose general noise level is of the order of magnitude of resistancenoise, it can be shown, other things being equal, that under certain circumstances this may amount to a 2:1 saving in the amount of output power on the basis of comparable signal-to-noise ratios. Moreover, as will now be explained, the improvement that feedback can produce in the ratio of signal to resistance-noise is by no means limited to three decibels, and feedback action in circuits such for example as those of Figs. 1 and 2 can produce stable impedances which do not create resistance-noise and from which the amplifier can be worked.

In circuits such for example as that of Fig. 1 the relation of the amplifier input or output impedance to the impedance Zn of the input or outputhybrid coil net 9 or it! can be used for transforming impedances, or in other words for producing from the net impedance an impedance equal to the net impedance multiplied by the con stant provided the amount of negative feedback is sufficient. Thus, the impedance produced as the amplifier input or output impedance is an enlarged copy of-whatever impedance ZN is used as the net 9 or ill of the input or output hybrid coil, and the impedances 9 and I may be of any suitable types. For example, either may be an ordinary impedance such as a resistance, an inductance, a capacity or any complex form of (twoterminal) impedance constructed of resistances, inductances and capacities.

Further, in circuits such as those of Figs. 1 and 2, the feedback action can transform ordinary resistances or generalized impedances that produce resistance-noise to corresponding resistances or generalized impedances that are free of all noise, including thermal agitation. For example, if the net 9 across the bridge-points of the input hybrid coil be a resistance, the feedback can, as just noted, transform this resistance by producing an enlarged copy of it as dance is a generalized impedance produced from a corresponding generalized impedance across the hybrid coil bridge-points by the feedback action.

For instance, the impedance Zn of the net 9 of hybrid coil 5 in Figs. 1 and 2 may be the impedance Zn shown in Fig. 3, which, with and s 1 causes the stabilized input impedance of the amplifier to be an extremely close match, over the 12 kilocycle to 60 kilocycle frequency range, for the impedance of a 19 gauge, non-loaded, .062 capacity standard toll cable; and then this input impedance (the input to the amplifier) 'is free from noise due to thermal agitation.

In general, in thus building an impedance equal to (t+1)ZN, free from thermal agitation, there are no restrictions on ZN, which may be any combination of coils, resistances or condensers, or a generalized impedance. Moreover, of such impedances free from resistance-poise canbe constructed all manner of impedance networks or active or passive transducers free from resistance-noise.

For example, Fig. 4 shows an equalizer M and an amplifier 52 which are substantially free from resistance-noise notwithstanding the fact that the equalizer may, if desired, include resistance. The equalizer may be, for instance, a constantresistance equalizer for equalizing the cable attenuation. The amplifier'may be of the type shown in Fig. 1, and may have any desired number of stages, G and P designating the grid of the first tube and the plate of the last tube (in this figure and also in other figures of the drawings). The incoming line 6 may be, for example, a coaxial system or other cable delivering to the equalizer signals .whose lower limit of transmission level is fixed by the cable resistance-noise. The input hybrid coil 5 may be such that the amplifier input impedance without feedback would be high compared to the equalizer impedance from which the amplifier works; and the net 9 may be a resistance or impedance of such value that the feedback matches the amplifier input impedance to the equalizer impedance from which of the hybrid coil 51s large, giving low loss for transmission from the equalizer to the grid G, (and correspondingly a high loss results for transmission from the feedback path f to the grid G). Then if the impedance elements of which the equalizer is constructed, shown for example as a series arm of impedance Zn and a shunt arm of impedance Zrz, are quiet impedances of the type described above (impedances free from resistance noise, which may be obtained as inputimpedances of negative feedback amplifiers), the equalizer can produce the desired equalization without adding resistance-noise. The amplifier then restores the signal level, amplifying the signal without introducing resistance-noise. Thus the desired equalization and amplification of the signal can be accomplished without decreasing the ratio of signal to resistance-noise.

It will be understood that the impedances Zn and ZF2 are input impedances of suitable feedback amplifiers. For instance, they may be input impedances of amplifiers such as the amplifier of Fig. 1 or Fig. 2, and may be obtained by giving the nets of the input hybrid coils of the amplifiers the impedance values l+t respectively,

Fig. shows how a suitable feedback amplifier such, for example, as amplifier l2 cancool a resistance l5 by the feedback action. The resistance is of the ordinary type producing thermal agitation electromotive force. It is shown in a heat-insulated chamber l6, and a switch [1 is shown by which the resistance can be connected to the input impedance of the amplifier. Since, as indicated above, the input impedance of the amplifier can be made a resistance free from thermal agitation by having the net 9 of the input hybrid coil 5 a resistance, it results that with switch I'I closed power is abstracted from resistance 15 without power being returned, and hence resistance l5 loses heat.

Since, with considerable amounts of negative feedback in the amplifier of Fig. 1, the impedance of net ID will not materially affect theimpedance into which tube 2 works, the tube, though its n t and "impedance may differ from its optimum load tration of the advantageous nature of such operation is when the output tube 2 is a pentode. Suppose the output impedance R0 of the pentode is 1,000,000 ohms, and it delivers maximum power when the output impedance it works into is 25,000 ohms. Then if it were to connect through a transformer, say a two-winding transformer, to an output cable or line having a resistance of 100 ohms, a 25,000z100 impedance ratio output transformer would be required. Usually, in such cases, it is also a requirement that the output impedance of the transformer on its low side match the impedance. of the connected line or amplifier load. In the example given, without feedback this requirement could not be met because the output impedance of the amplifier would be the low side impedance of the coil when its high side was terminated by 1,000,000 ohms or practically open, whereas the high side winding of the coil would have to be terminated by 25,000 ohms for the low side impedance to be 100 ohms. Of course, by using a 50,000z100 impedance ratio output transformer and by adding a 50,000 ohm resistance across the high side winding in parallel with the tube, both these requirements (a 100 ohm output impedance for the amplifier and a 25,000 ohm load for the tube) could be satisfied simultaneously, but in this case, one-half of the output power delivered by the tube would be wasted.

However, by using at the output of the amplifier an inequalityratio hybrid coil connection which in itself introduces only a very slight insertion loss, both these requirements can be met simultaneously and practically all of theavailable tube power is useful. For example, in the amplifier circuit of Fig. 1, if the impedance of the net I 0 were such as to cause the amplifier output impedance to equal 3,500 ohms, and the tube 2 were, say a coplanar grid tube having its impedance R0 equal to 3,500 ohms and working into 3,500 ohms, and if then the coplanar grid tube were replaced by a power pentode whose impedance R0 was 75,000 ohms and whose optimum load impedance was 3,500 ohms, this pentode would work into 3,500 ohms, thus satisfying the first requirement. It would also be found that substituting the pentode produced practically no change in the amplifier gain, because the amount of negative feedback was already great and substituting a still higher gain tube merely further increased the feedback. The amplifier output impedance if measured would likewise be found not to have changed, thus satisfying the second requirement, because as explained above, the value of the amplifier output impedance is set by the impedance of the net H], which was not changed.

Thus, it can be seen that by using the hybrid coil connection, the pentode can be worked into its optimum load impedance and at the same time the output impedance as presented by the amplifier can be kept on a matched impedance basis and hence, since practically no power is wasted, the available output is doubled as compared to the case without feedback.

Especially in amplifiers without feedback, if the gain-load curve of the amplifier is scrutinized very closely, it will usually be discovered even i at light loads, that the gain of the amplifier varies slightly with changing load. When a great number of such amplifiers are in tandem as, for example, on a telephone circuit having many repeaters, this efiect, if systematic, tends to be proportional to the number of amplifiers. It leads to a degradation in the quality of the speech transmitted and this particular kind of deterioration has been termed pep effect. Even when the number of amplifiers in tandem is very large and the characteristics of the amplifiers without feedback are unsuitable from this standpoint, by using considerable amounts of negative feedback the practical difiiculties of pep effect are readily avoided. When the number of amplifiers involved is very large, the transmission characteristics of the input and output transformers may show this effect and in such instances the design of these coils is simplified by feeding back around the transformers as, for example, in the circuit of Fig.1 or the circuit of Fig. 14 described'hereinafter.

Since, with considerable amounts of negative feedback in the amplifier of Fig. l, the impedamplifier gain changes independent of frequency,

or a network adjustable for giving variable equalization or changes of amplifier gain dependent on frequency. When it is desired that gain characteristics be parallel and fiat, they are preferably made as parallel as required or possible, and, in addition, as flat as possible. This distinction is made because for applications requiring great refinement, if the curves are parallel and almost fiat, then the slight lack of flatness can be corrected by the addition of a fixed equalizer.

Fig. 6 shows an amplifier similar to that of Fig. 1,' but shows the B-circuit network by way of example as a thermo-sensitive shunt element, preferably a resistance -23 of silver sulphide for controlling the gain of'the amplifier. G and P in this figure, and wherever appearing in other figures of the drawings, designate the first grid of the first stage and the plate of the last stage of the amplifier, and indicate that the amplifier may have any suitable number of stages. To

vary the resistance, for changing the amplifier 1934; E. I. Green Patent 1,918,390, July 18 1933;

gain, the temperature of the resistance is varied.

This temperature variation is accomplished by adjusting a resistance 24 which controls heating current supplied from an alternating currentor direct current power source 25 to a heating element 26 for the silver sulphide resistance 23.

Ordinarily the amount of power required to heat the silver sulphide resistance so as to produce a gain change of a few to as much as to 100 decibels would not exceed a small fraction of a watt and in many instances could be as little as one milliwatt or even a fraction 'of a milliwatt. Used in this manner it would often be required that the silver sulphide be stable with time and humidity. Where efiects of room temperature variations tend to be objectionable, the silver sulphide resistance may be enclosed in a heat insulated chamber 21. The heat insulated container, possessing sufficient heat capacity with respect to the size and heat capacity of the silver sulphide resistance, is very helpful in reducing efiects of variations in ambient or room temperatures upon the silver sulphide resistance. If desired, to compensate for effect of room temperature on operation of the silver sulphide unit, the heat insulated container can be maintained at a constant temperature, above the highest room temperature.

This could be done by a thermostatic control, but in Fig. 6 is accomplished by a chamber-heating element 28 supplied with heating current from alternating current or direct current power .source 29 through a regulating network such for example as network 30. The network 30 is shown as having a series resistance arm 3! and a shunt arm comprising a resistance 32 in series with two parallel resistances 33 and 3d, resistance 33 being a silver sulphide resistance. The constants of the network depend upon the thermal and other properties of the particular silver sulphide'unit 33. With the temperature of the chamber 2'8 elevated above the highest room temperature by the chamber-heating unit 28, if the room temperature rises the resistance of the silver sulphide element 33 falls sufiiciently to reduce the heating current in the element 28 so that the temperature of the chamber 21 is held constant.

The voltages or currents from the power supply sources 25 and 29 should be relatively stable.

With the gain control in the feedback path of a stabilized feedback amplifier, as indicated for example at 3 in Fig. 1 and at 23 in Fig. 6, reducing the gain is accomplished by correspondingly increasing the amount of negative'feedback; and this improves the amplifier performance accordingly, for example reducing modulation, increasing gain stability, and, in the case of hybrid coils that are in passive unbalance, increasing the independence of the amplifier input and output impedances with respect to the impedances of the' network such as 3. Since the amplifier gain practically equals the loss in the feedback circuit and the working gain is usually appreciable, a loss is usually required in the feedback path; and consequently, considerable loss may occur in the gain control device in the feedback path without necessarily being a disadvantage.

The gain control 3 in Fig. 1, or the adjustable gain-control contact of resistance 24 in Fig. 6, may be operated manually; or if desired it may be operated automatically, for example asthe gain-varying element is operated by the pilot apparatus in the transmission control system disclosed in H. S. Black Patent 1,956,547,,May 1,

or J. R. Fisher-C, O. Mallinckrodt Patent 2,116,- 600, May 10, 1938.

Silver sulphideis preferred as the temperature responsive transmission control element because of its large (negative) temperature coefiicient of resistance, constancy and uniformity of performance as disclosed more fully in the Fisher and Mallinckrodt application just mentioned. The specific form of the silver sulphide element may be, for instance, the form disclosed therein; and the preparation of the element may be, for example, as disclosed in J. R. Fisher Patent 2,082,102, June 1, 1937.

Instead of having the silver sulphide element shunted across the feedback path, it can be placed in series'in the feedback path in the manner indicated in Fig. '7, about to be described; and if it is. then desired that the circuit be symmetrical (balanced to ground), two such ele ments can be used, as indicated in Fig. 7. Both can be in the same heat chamber, both heated by the same heater 26 oreach heated by an individual heater such as 26.

Fig. '7 shows an amplifier which may be a stabilized amplifier, having considerable negative feedback, similar to the amplifiers of Fig. 1 and Fig. 6. However, the c-circuit network is shown as two silver sulphide resistances 4d and 45 and two ordinary resistances 46' and 41. This network automatically controls. the amplifier gain to maintain constant output. It causes part of the output wave to vary the amplifier gain so that the amplifier output level is maintained constant although the signal input level may vary over a considerable range of values. The resistances M and 46 are in series in one side of the feedback path ,f and the resistances 45 and 41' are in the other side. Where symmetry (balance to ground) is not required, the resistances in either side may be omitted.

The resistance of silver sulphide has such a large negative temperature coefiicient that passage of even a small current will heat it up sufficiently to reduce its resistance as compared to the value of its resistance at ambient temperatures. In the operation of the system, at sufficiently light loads the feedback currents do not heat the silver sulphide appreciably; and; as a result, the silver sulphide resistance introduces nearly a constant insertion loss and, therefore, will not appreciably vary the amplifier gain.

As the applied input is further increased, due to self-heating, the resistance of the silver sulphide decreases; and this reduces its'insertion loss and therefore the amplifier gain. Thus, there is a tendency for the gain to decrease as the input increases, which efiect in the proper proportion can hold the output constant. By adding small resistances 46 and 41 in series with the silver sulphide resistances, this proper proportion is maintained, so that there is a considerable range of values of input for which the output is independent of the input, and in these instances the gain of the feedback amplifier goes down decibel for decibel for each decibel in-' independent of the input level to a precision of 0.2 decibel over a range of 1 to 15 decibels, and to a precision within 1.05 decibel over a range from +1 to +14 decibels, and to a precision within 1.01 decibel from +2 to +9.5 decibels. This was for a particular frequency, 5000 cycles per second. However, inasmuch assilver sulphide resistances are independent of frequency from very low frequencies to very high radio frequencies, the characteristic effect described as obtaining for one frequency is the same for other frequencies in the useful range of the feedback amplifier. Therefore, in a well-designed amplifier, the gain-frequency curves for various specified values can be made parallel and flat over the useful range.

To obtain the precisions mentioned above, the

constant resistance in series with the silver sul-, phide resistance was given an optimum value,

which in this particular instance was 210 ohms and amounted to about 3 per centof the resistance of the silver sulphide unit at room temperature.

While silver sulphide resistance and a constant resistance in series, bridged across a circuit, can hold the circuit voltage constant over a considerable range of variation of the internal voltage of the generator feeding the circuit, the range over which the volume can be limited is greater in the case of a circuit including an amplifier in the general manner indicated in Fig. 7.

. That is, the amplifier increases the useful range of the volume limiting circuit. Further, the useful range of operation, and also the precision of the control, can be considerably extended by refinements in the design of the silver sulphide resistance and in the design of the feedback circuit associated with it. 1

The output voltage at which regulation takes place in Fig. '7 can be reduced by decreasing the volume of the silver sulphide unit. Another way to increase the sensitivity is to elevate the temperature of the unit, for example to a temperature just below that at which it begins to regulate, by an auxiliary or indirect heater such for instance as the heater 26 shown in Fig. 6.

The useful range of the volume limiting system of Fig, '7 can be increased by connecting volume limiting devices, such as the device shown, in tandem with each other in the feedback path I, and adjusting one to operate when the limit of the useful range of the preceding device has been reached; or the resistances 46 and 41 may each be made a silver sulphide resistance and a constant resistance in series, these additional silver sulphide resistances then being adjusted to begin their regulating or limiting action when the volume limiting device shown has reached the limit of its useful range.

Aside from its simplicity the circuit has a desirable property not possessed by the ordinary current or volume limiter type of circuit. In the circuit of Fig. 7, as the limiting increases, i. e., gain decreases, the harmonic level also decreases. In other words, the greater the limiting, the better the transmission performance generally. This is a very desirable action and is opposite to the action of the ordinary type of circuits of this class.

Since modulation originating in the feedback path or B-circuit is not improved by feedback and appears in the amplifier output, it might be questioned whether, as the alternating current flowing through the silver sulphide in Fig. 7 varies from instant to instant, the resistance of Fig. '7

of the silver sulphide might not vary correspondingly and thereby produce serious modulation. Fundamentally, the answer depends upon the speed with which the silver sulphide heats up and which includes a frequency as low as 1000 cycles per second.

The speed with which a circuit of this character may be made to operate automatically to control the gain of the feedback system may be very fast or very slow, for example, apparently from as fast as of the order of 1/10,000 second to as slow as more than ten minutes. The speed of operation is controlled chiefly by changing the physical size, and mechanical and thermal design of the silver sulphide resistance unit itself.

Silver sulphide changes its resistance with temperature apparently faster than any other known material which simultaneously would satisfy the additional practical requirements that such a resistance be stable, that its temperature effects and behavior be accurately reproducible and that it be capable of being manufactured cheaply tofall within closely specified limits. Silver sulphide meets these latter specifications and in addition, for all temperatures below 179 0., its resistance. is halved every time its temperature is increased approximately 13.9 C. At 179 C. its resistance is abruptly divided by more than 40, and for still higher temperatures its resistance increases.

Due, for temperatures below 179 C., to the large and negative temperature coefficient of the resistance of silver sulphide, after a voltage applied to such a resistance exceeds a certain critical value depending upon the temperature and material surrounding the silver sulphide and upon the thermal properties of the silver sulphide unit itself, the current flowing will commence to increase abruptly, due to excessive self-heating of the silver sulphide, and this increase will continue until finally limited by any resistance or impedance in series with the applied voltage or until the temperature of the unit has reached 179 C.

For example, for a fixed ambient temperature, if a very small voltage be applied in series with asilver sulphide resistance, a very minute current will flow in response to the applied voltage. This flow of current will cause power i R, to be dissipated as heat which will raise the temperature of the silver sulphide and thus reduce its resistance, the offset of this can be shown to increase the current and, therefore, further raise the temperature of the silver sulphide. The total energy this represents in any arbitrary interval of time will be in the form of heat energy, and, if by conduction, convection and radiation this heat is transferred to the surroundings at a faster rate than it is supplied, the dynamic system will be stable and the silver sulphide will assume an equilibrium temperature slightly in excess of its ambient temperature. For further increase in voltage, the equilibrium temperature is further increased and the rate of increase of the latter can be shown to exceed the former. The net result of allthis is that if the current is plotted against voltage applied, the resulting curve is concaveupward and lies above the straight line For many applications this effect is not enough representing the relationship between current and voltage for a fixed resistance whose value is equal to that of the resistance of the silver Sill phide at its ambient temperature.

However, as the applied voltage is further inerates to prevent further increase in tempera. ture, reduction of resistance, and increase of current.

Appropriate analysis disclosed that the trigger action above described is duplicated in a general sort of way if impedance is inserted in series with the applied voltage source provided the added impedance is not obviously too great. Theoretical analysis also showed that a resistance, such as 45 or .47, of proper value and in series with the silver sulphide greatly improves the operation of the circuit of Fig. '7, and experi: ment demonstrated the improvement.

The input or trigger voltage at which the automatic regulating system of Fig. 7 operates and, therefore, also the value of fixed output, is somewhat reduced as the room temperature increases.

to be troublesome. Where the operation needs to be precise, the silver sulphide resistance can be housed in a container maintained at constant onds and release in more than 0.8 second. In 1 a circuit such as shown in Fig. '7 this fast-operate time would be obtained by making the volume of the silver sulphide unit sufficiently. small and the release time adjusted by the amount of heat insulation immediately surrounding the imit. When the surrounding heat insulation is sufflciently far from the silver sulphide it. merely has the eifect of putting the silver sulphide in a heatinsulated room and the times of. heating and cooling are not affected. However, placing the silver sulphide in an insulated chamber sufliciently large (relative to 'the silver sulphide unit itself and the power it dissipates) greatly reduces the efiect upon it of fluctuations in room temperature, and materially reducesits slow temperature changes, also.

The volume limiting circuit of Fig. '7 is of general application. For example, it is suitable for use as a receiving amplifier in a carrier transmission system, automatically keeping the voice frequency equivalent constant independent of high frequency variations; as a substitute for the volume control circuits now used in radio receiving circuits; as a volume limiter in front of a loud speaker, to prevent overloading; as a volume limiter to prevent overloading a detector, a group modulator or a broad band amplifier carrying a number of channels (the effect in this case resulting in an important increase in theuseful operating level of the amplifier or repeater); as a control to keep the pilot current fixed at the sending end of a pilot channel system; as a control to maintain theroutput of multi-frequency carrier supply systems at a constant voltage; as acontrol to limit the volume of voice frequency telegraph signals superimposed on carrier telephone systems; or as acontrol for an alternating current supply voltage to render the voltage constant for operation of alternating current apparatus or for testing or measuring purposes.

- The hybrid coils in Fig. '7 not only render the amplifier input and output impedances independent of the connecting circuits, but, in this volume'limiter, the hybrid coils prevent the amplifier impedances from being changed by changes in the amount of" feedback that result from changes of input level.

Itis noted as to the operation of the hybrid coils in a circuit such as Fig. 1, that when'the e-circuit network 3 is a network for amplitude or phase equalization. or correction of distortion (in the general manner disclosed for example in my above-mentioned article, Patent 1,956,547, or copending application 606,871. or in British Patent 371,887) the hybrid coils render the equalization independent of the flexibility of the amplifier input and output impedances, the impedances of the amplifier being adjustable through very wide ranges by means of the hybrid coil nets 9 and it] without affecting the equalization or distortion correction. This is especially noteworthy since locating the equalizing or transmission controlling network in the feedback path or c-circuit has important advantages. ,For example, whereas in certain instances it is practical to design a corrective network having the same transmission characteristic as the apparatus or system to be equalized, the design of a network havingthe inverse characteristic may involve negative elements; and therefore by equalizing in the feedback path compensation can be effected for transmission distortion which, without resort to feedback action, could not be corrected in any known manner using physically realizable elements.

Another advantage concerns effect upon signal-to-noise -ratio.' If the received signal has been propagated over a transmitting medium I having more loss at some frequencies than others,

and if the general noise level is of the order of the noise due to thermal agitation, a substantial improvement results from performing the equalback path. This can be appreciated by observing that the introduction of a loss of :c decibel in I front of the input to the amplifier degrades the signal-to-noise ratio :r decibel, assuming the introduction of such loss leaves theresistance component of the equivalent passive series impedance from grid to filament of the first tube unchanged.

be required to have 30 decibels more loss at the,

lowest than at the highest frequency. In a constant-resistance type of equalizer, this would degrade the signal-to-noiseratio of the lower frequencies something in excess of 30 decibels. These relations only hold for resistance noise which is unaffected by insertion of a constantresistance network in front of the amplifier.

A further advantage is increase of allowable ization or other transmission control in the feedhead-end loss of the equalizer or corrective network. In practical designs, at the frequency at which the insertion loss is required to be least, actually there is a finite decibel loss. Usually, when the equalizer is located in the line, this loss is required to be a minimum. However, if

the equalizer is located in the B-circuit, in most instances this loss can be considerable without affecting the overall transmission performance. Generally speaking, increasing head-end loss will permit appreciable economies in size and cost of the parts.

In the case of pre-equalization there is an advantage with respect to load rating. If the equalizer is located in the c-circuit and, in addition, the cable or transmitting medium is free from practically all noise but resistance noise, pre-equalization will result in a substantial number of decibels of improvement in the level rating of the amplifier depending upon the band width, number of channels, and attenuation-frequency characteristic of the medium. As compared to locating the equalizer in front of the amplifier, the input levels can be reduced by the total amount of loss of the equalizer for either high or low frequencies. For the case of a cable this would mean reducing the levels at the lower frequencies by amounts corresponding to the difference between the loss at the highest frequency and the lesser loss at the lower frequencies plus an additional reduction at all frequencies equal to the head-end loss. As a result of reducing the level of some frequencies more than others, the

load rating of the amplifier is improved.

Another advantage is reduced modulation requirements. If the pre-equalization is one-half of the cable slope, attenuation vs. frequency, in stead of the full amount as just discussed, a worthwhile portion of the improvement in level rating of the amplifier is still retained. However, with re-equalization, if and only if the equalizer is in the fi-circuit, the modulation requirements are uniformly reduced at all frequencies in the transmitted band by an amount approximately equal to one-half the difference in the cable attenuation at the highest and lowest used frequencies. For the same performance as before, this leads to a worthwhile reduction in the gain without feedback which is more important the wider the frequency band transmitted and the higher the top frequency.

Even without pre-equalization, the perform ance can be improved by the p-circuit location of the equalizer. If the cable noise varies with frequency and increases for each frequency the same number of decibels the cable attenuation decreases, pre-equalization is not possible. In this case the improvement in level rating and reduction in modulation requirements as just mentioned cannot be obtained. However, the performance can be improved, or for equal performance less feedback will be required. For example, if the attenuation is 30 decibels less at the lowest than at the highest frequency, the negative feedback is 30 decibels greater at this frequency due to the presence of an equalizer in the ii-circuit. The result, as compared to not locating the equalizer in the fl-circuit, is the gain Without feedback for equal performance does not have to be held at so high a value over so broad a frequency band. This leads to a higher gain per stage for the top frequency and this gain increase is comparable to that of the preceding paragraph.

An outstanding advantage of the fl-circuit phase distortion.

location for the corrective network relates to speeding up transmission. As shown in my above-mentioned article or copending application 606,871, or in British Patent 371,887, the amplification of a feedback amplifier with pfl 1 is approximately B so when the fi-circuit is made such that its propagation is the'same as that of the line or cable between a source of voltage 6 and the input to the -circuit, the transmission from the source of voltage e to the output of the -circuit is (as brought out for example in the disclosures just mentioned or in my copending Patent 2,002,499, May 28, 1935, or F. A. Cowan Patent 2,017,180, October. 15, 1935), and hence there is no delay nor distortion and the amplified signal appears instantly at the output of the it-circuit as a replica of the sign-a1 e applied to the line or cable by the source, except reversed in sign. (There is no restriction on other than that the amplifier comply with Nyquists rule.) This result is obtained from theoretical considerations. The equations which areavailable for arriving at this conclusion are rigorous only for systems containing lumped constants. The transmission through systems made up of continuous elements is only approximated from the lumped constant equations. Hence to this extent the above procedure will correct for amplitude distortion and (Applications are encountered where there are likewise important advantages in having the B-circuit correct solely for phase distortion.)

Compared to the customary way of improving cable distortion this procedure speeds up transmission. The customary way, wherethe network for correcting attenuation and phase over the frequency band of interest is not in a feedback path, delays the time of transmission by a period that exceeds the time required for a particular frequency to travel down the cable, this particular frequency generally being that one in the band of interest which is most slowly transmitted. In contrast to, this, for the described procedure of correcting-in the fi-circuit, the time is the time for a particular frequency, which is the one most rapidly transmitted, or in other words, it is the time required for current no matter how trivial to make its appearance at the receiving end. All other velocities for all other frequencies for which the circuit is properly operative are made equal to this most rapid speed. In the case of a cable, this speed apparently corresponds to a speed less than the velocity of light in a vacuum, depending upon the dielectric constant and permeability of the cable. It should be noted that this time is independent of the wave form impressed at the sending end. It is also independent of lumped series inductance, either positive or negative, and likewise is independent of lumped shunt resistance, either positive or negative. Thus, by adding apparatus (the feedback amplifier) to a heavily loaded voice frequency cable, a wave can be propagated ,over the cable at the same speed as over the same cable non-loaded.

Assuming the cable is many wave-lengths long,

' preferably the fl-circuit instead of being a like section of cable may have a difierent indicial admittance, as now to be explained.

Considering indicial admittance of a transmission system as a transfer admittance, it is the current as a function of time at the far end in response to a unit voltage suddenly applied at time t=0 at the sending end, and two systems having like indicial admittances will behave identically at the receiving end for like excitations at the sending end. If two systems have like indicial admittances' their steady state amplitude and phase characteristics are identical. If two systems have like amplitude and phase characteristics (that is, steady state attenuation.

and phase vs. frequency characteristics) over a specified band of frequencies, for example ii to is, then the received currents within this band probably will be the same for each system for like excitations at the sending end.

Nevertheless two systems can have very nearly identical steady state attenuation and phase characteristics from i=0 to f= and yet the time taken for some current, no matter how little, to arrive at the receiving end can be entirely different. n

For example, two such systems are on the one hand a cable of length l-where Z is many wavelengths and on the other hand a lattice network of one section whose four elements are two impedances each equal to the cable impedance of length with far end shortcircuited and two other impedances each equal to the cable impedance of length where v is the velocity of propagation referred to above, 1. e., 1- is the time for the first current to appear at. the distant end.

When a feedback amplifier is to correct for a section of cable in front of the amplifier in the general manner referred to above, preferably the p-circuit should have an indicial admittance equal to the imiicial admittance of the cable when, referring to the admittance characteristic of the latter,,t is replaced by (i''r). If this is done and pfl 1, the output of the amplifier will be a replica of the input at the sending end except it will appear 7' seconds later.

If the p-circuit be a like section of cable (assuming Nyquists rule is satisfied by appropriby making additions) the time of transmission is still 1' but for an interval 1' at the beginning and a corresponding interval at the end, the wave form apparently will be incorrect. This effect would be something more than multiplying the time of transmission by two as compared to the method of the preceding paragraph. It would make the received wave distorted in a manner analogous to that of a loaded cable.

To give a physicalconcept of the method described in the second preceding paragraph for correcting phase distortion the response without feedback to the excitation canbe viewed as the forced or steady state response plus the free flow or transient response. As a result .of feedback action the amplifier does not amplify the transient response for the conditions above described. Thus, the first currents to arrive of necessity carry information as to the signal impressed so that the amplifier sends out a copy of the signal wave form applied at the sending end, and, when the main body of the transmitted signal arrives later and appears at the input to the-amplifier, it just is not amplified.

When an attenuation equalizer or other corrective network -is located in the feedback path between the hybrid cells as in the case of the ,B- circuit network 3 of Fig. 1, since the corrective network does not afiect the amplifier input or output impedance the equalization or correction can be accomplished by merely inserting a simple series or shunt arm between two arbitrary fixed impedances. Usually this requires less than half as many elements as the corrective network requires when it is so located that it affects the amplifier impedance seen from theline.

Moreover, when an attenuation equalizer or phase corrector is to be located between the hybrid coils as in Fig. l, the impedances between which it is to insert a specified loss or phase shift can be conveniently caused to assume a very wide range of values, as muchas lflllii'zi. This element of flexibility is an aid improducing a low I cost design.

In Fig. 2,'if p 1, assuming C is made equal to Zr there is a decibel gain from C to Zn equal to C 10 lOgmTN Therefore, a system, such for example, as that shown in Fig. -8 (which is dominated by R. S. Caruthers application Serial-No. 114,409, Patent 2 .166,929,-issued July 25, 1939, entitled Electric wave amplifying systems, filed of even datehere- -with) can be built, comprising an amplifier (with pfl 1) that works on a matched impedance basis from a -resistance R1 into a greater, resistance C2 and has its gain In Fig. 8 the amplifying element is designated 50. It may be of any suitable type, as for example, a

vacuum tube device such as the amplifying deand ately modifying a, or the cable itself is modified In transmitting in the opposite direction through the system, that it, from R2 to R1, the loss is precisely the same as the previous gain (i. e., from R1 to R2). In either direction, the gain or loss, as the case may be, is independent of frequency by virtue of the method of operation of the circuit.

The circuit of Fig. 8 can be rearranged as in Fig. 9, wherein to work from a resistance R1 into a resistance R2 that is greater than R1, giving a gain and giving a loss from R2 to R1 equal to this gain Gp. In Fig. 9 the amplifier 59' and the hybrid coil 5! correspond to the amplifier 50 and hybrid coil 5| of Fig. 8.

Figs. 8 and 9 can transform impedances, in the manner indicated in the discussion above of Figs. 1 and 2, the circuit of Fig. 9 dividing an impedance by a number greater than unity.

Fig. 10 shows an amplifier circuit obtained by combining two circuits such as those of Figs. 8 and 9. The input impedance Z02 and output impedance Z01 of the amplifying system equal respectively the impedance Zc2 of the output connecting circuit and the impedance Z01 of the input connecting circuit. Without the pad 52 shown in the path joining the two amplifying elements, the amplifying system hasa gain (from west toeast) equal to and has a loss (from east to west) equal to If desired, to reduce the gain without afiecting the impedance relationships, the pad 52 of impedance level Zn can be inserted as shown.

The circuits of Figs. 8 to 10 show that power is dissipated in stabilizing an impedance by feedback as in Fig. 2. In the case of Figs. 8 to 10 (with network 52 omitted) all such power is dissipated in the output load.

Certain of the circuits described hereinafter employing hybrid coil feedback connections, such for example, as those shown in Fig. 1, employ types of feedback that will be called multiple feedback and repetition of the feedback process. The significance of these terms will now be indicated.

As discussed in my above-mentioned copending application 606,871, it is often desirable, especially in analysis and design of feedback systems, to distinguish between systems restricted to a single feedback loop and systems having a plurality of loops or in other words multiple loop feedback systems. Afeedback loop with more than one path from the same output to the same input is'a multiple loop, but as noted in that copending application, if the multiple loop can be theoretically replaced by a single loop, the feedback is regarded as parallel feedback, except that a special case of parallel feedback is termed repetition of the feedback process (referred to below). As also noted in that copending appliincluding those cases of feedback in which a single ip-loop includes additional local feedback loops and thus comprises a collection of feedback systems having elements in common and satis fying the condition that when the over-all system is viewed analytically in the form of an equivalent circuit, it is not possible to obtain further reduction in the number of local feedback loops surrounded by a single B-loop. As further noted in that copending application, a repetition-of the feedback process is considered to occur whenever a complete feedback system (single loop or multiple feedback) may be viewed and treated as a unit and, in addition, is used to form a new ,ufl-path which path in every way is independent of the first feedback system except in so far as it utilizes the original over-all properties of the first feedback system. This requires conjugacies or their equivalent and the use of transformers, assuming no new unilateral devices or their equivalent be added. As shown hereinafter, repetitions of the feedback process are but a special case of parallel feedback, namely, with an added restriction regarding conjugacy. Accordingly, unless active elements are used to separate the feedback paths the various paths or loops theoretically can be replaced by a single path or loop. However, repetition of the feedback process can have practical advantages, for example, as pointed out hereinafter.

Suppose, for example, it is desired to feed back through the input and output transformers and suppose further that decibels of feedback is required but that with the particular transformers being used only 25 decibels of feedback is'possible through the coils without getting into singing troubles. Now by putting 35 decibels of feedback inside the coils and 25 decibels outside the amplifier can be designed so as not to sing. Suppose further it is desired that the input and output impedances be exceptionally good. By using repetitions of the feedback process instead of some other type of multiple feedback, the deviations of these impedances from the ideal sought will be improved roughly fifty-fold.

Fig. 11 is a simple example of a'multiple feedback amplifier employing a hybrid coil feedback connection, the feedback path I connecting to the output of the amplifier through hybrid coil 1 and being in series with the secondary winding of a two-winding input transformer 50 that connects the incoming line 6 to the amplifier. Thus this feedback through path I is a series or current feedback at the input of the amplifier and a hybrid coil feedback at the output of the amplifier and may be referred to as a series-hybrid feedback. The amplifier has tubes 55 and 56 I coupled in tandem through an interstage coupling circuit comprising tuned circuit 6|, stopping condenser 62 and grid leak 63. A resistance or impedance 64 common to the input and output circuits of tube 55 provides a local or internal series-series feedback around tube 55. Thus this local series-series feedback loop or system is surrounded by the main and single ,ufi-loop. The series-series internal loop around tube 55 is paralleled by a second loop comprising tube 56 and its output hybrid coil, etc. Therefore, with respect to the input and output of tube 55, the circuit of Fig. 11 is an example of a double loop multiple feedback system. Although double loop systems are often multiple feedback systems, it is evident that from some aspects this particular double loop system around tube 55' could .be-replaced by an equivalent single 1 loop system. However, in the case of an equivalent single loop system, the modulation atthe output of tube 55 would be different and, of course, with respect feedback with m3 l for this main loop; or, if

desired, it may be made positive feedback by the poling of the hybrid coil. The positive feedback may be desired, for example, to produce over an appreciable frequency interval an increase in .gain or to cause the change in phase due to feedback to increase at low frequencies in order to reduce delay distortion. The internal negative feedback, (and also the external feedback,

-if negative) may be desired, for example, to

flatten the gain-frequency characteristic, as for example, in the case of a voice frequency amplifier.

The procedure for derivation of equations for the transmission performance of multiple feedback systems can be along exactly the same lines as followed, for example, in my above-mentioned article and copending application 606,871.

tems, to avoid singing of the over-all system, or in other words to meet the requirement that the free response of the system as a whole will not be a wave expanding with time, it is necessary that the 3 characteristic of an individual single loop do not inclose ,a unit circle about the point l, 0 as a center instead of the point l, 0 itself as given by Nyquists rule for a single loop. In other examples, certain loops will be unstable if other loops are opened, etc.

Fig. 11A shows an amplifier which is a modification of the circuit of Fig. 11, the potential, with respect to ground, that is fed back from the hybrid coil. to the grid of the first tube being applied across resistance 64. The voltage across resistance 65 is transmitted also to the grid of the second tube through the plate-cathode path in the first tube and the interstage coupling circuit, before phase reversalin the first tube, but the voltage thus applied to the grid of the second tube is small in view of the high impedance of that plate-cathode path. The amplifier is suitable, for example, as a high quality voice frequency amplifier for program transmission circuits. The tubes are shown as a screen grid tube 55 and a coplanar grid tube 56' which may be Western Electric Company 259A and 281A tubes, for example. The space current for tube 55' is supplied through a choke coil 6| and the negative bias for the grid of tube is supplied partly by battery C1 and partly by the resistor 64 which is unby-passed and, therefore, produces local negative feedback around tube 55'. Battery 02 supplies the screen biasing voltage for this tube, the control grid and coplanar grid biasing voltages for tube 56' being supplied by batteries C3 and C4. As in the case of Fig. 11, the outer feedback may be made either negative or positive by poling of the hybrid coil.

However, the criterion for singing may be different. For example, in certain multiple feedback systween network I8 and hybrid coil 61.

13 common to two feedback paths. Viewing the system as a single loop feedback system, the value of e is other than zero for each direction of transmission, clockwise and counterclockwise. REis an amplifier for transmission from line LW to line LE; and RW is an amplifier for transmission from line LE to line LW. Transformers 65 and 66 are connected to operate as a hybrid coil 6?; and, similarly, transformers 68 and 69 form a hybrid coil 10, transformers H and 12 form a hybrid coil I3; and transformers l4 and 15 form a hybrid coil 16. A fi-circuit network 11 may provide any suitable control for transmission through the repeater. Networks 18' and 18 are balancing networks of hybrid coils l3 and 16, respectively; and networks 19 and are balancing networks for hybrid coils 61 and 10, respectively.

' Transmission from line LW arriving at hybrid coil 13 is partly dissipated in network 18" and partly transmitted through-hybrid coil 73 to hybrid coil 67. The transmission arriving at hybrid coil 6! is partly dissipated in the output of amplifier RW and partly transmitted through the hybrid coil 55 to amplifier RE and thence to hybrid coil it. Part of the transmission thus reaching hybrid coil '19 is dissipated in network 80 and part is transmitted to hybrid coil 16. Of this latter part, one portion goes through the hybrid coil to line LE, and another portion goes through the hybrid coil and the p-circuit-network TI to the hybrid coil 13, there dividing be- Thus, waves from the output of amplifier RE are fed back through hybrid coil 16, fi-circuit network 11 and hybrid coil 13 to the input of amplifier Similarly, transmission from line LE is amplified by amplifier RW and transmitted to line LW, and the amplifier RW feeds back through hybrid coil 13, fi-circuit network Ti and hybrid coil 16 to amplifier RW.

Fig. 13 shows a second example of a multiple feed back amplifier with a feedback connection from an output hybrid coil i. There is a triple loop path (and, as in the case of the circuit of Fig. 11, it is not apparent how the multiple loop could be replaced by a single loop connection).

The amplifier has a two-winding input transformer 8! with primary and secondary windings 82 and 83. Winding 82 forms one arm of an input bridge 8%, the other arms being resistances or impedances 85, 85 and 81. The incoming line 6 is in one diagonal of this bridge, and the other diagonal includes a feedback path, which will be called the pa path.

The grid-cathode impedance of the first tube of the amplifier forms one arm of a second input bridge 88, the other arms being resistances 89, 9d and SI. The winding 83 is in one diagonal of this bridge, and the other diagonal includes a feedback path, which will be called the ,81 path, this feedback path comprising a two-winding transformer 92 and also comprising a p-circuit network 93 for controlling transmission through this feedback path in the same general manner as explained above for the case of network 3 in the feedback path in Fig. l.

Bridges 88 and 94 need not be balanced. .As

pointed vout jn my above-mentioned copending application Serial No. 663,317, there is a G-decibel advantage (3 decibels at the input'and 3 decibels at the output) in unbalancing the bridges. Considerable amounts of negative feedback around the er path and th up: path can give practical

Images