US2364389A - Negative feedback amplifiers - Google Patents

Description

. 1944- A. H. ROCHE HAL 35 3 NEGATIVE FEEDBACK AMPLIFIERS Filed Nov. 29, 1945 2 Sheets-Sheet 1 s I By 7504: WI QM.

Dec. 5, 1944.

A. H. ROCHE EIAL NEGATIVE FEEDBACK AMPLIFIERS Filed NOV. 29, 1943 2 Sheets-Sheet 2 F/GS;

Logr

In venton Patented Dec. 5, 1944 hittttt NEGATIVE FEEDBACK"AMPLIFIERS Alleman Holly Roche and Thomas William Elliott, London, England, assignors to Standard Telephones and Cables Limited, London, England, a British company Application November 29. 1943, Serial No. 512,174

In Great Britain December 24, 1942 11 Claims.

The present invention relates to electric negative feedback amplifiers, and concerns particularly arrangements for preventing instability during switching operations.

In multi-channel carrier current signal transmission systems, it is necessary to provide convenient means for replacing amplifiers which may have become defective, or for other reasons, with the minimum of delay.

In co-axial cable systems carrying several hundred channels all of which pass through the same amplifier, it is essential that any switching operation involving the amplifiershall not produce any effective disconnection, otherwise all of the several hundred channels will be interruptedi and a break of one or two milliseconds may be sumcient to destroy the intelligibility of the signals. For this reason, when replacing any normal amplifier by a spare amplifier, it is the practice first to connect the two'arhplifiers in parallel to the cable conductors, and then to remove the normal amplifier. In this way interruption of the service is avoided.

Since, however, the amplifiers used are now almost exclusively of the negative feedback type, it is found that although any amplifier is stable when normally connected to the line, the connection of two amplifiers in parallel in the manner explained is liable to upset the transmission conditions of the feedback loops of both amplifiers so that singing often results, which will probably have a worse effect on intelligibility than a short disconnection. The principal object of the present invention is, therefore, to overcome this difficulty by providing a switching arrangement which does not cause the parallel connected amplifiers to be unstable, and which at the same time does not entail any appreciable sacrifice in the normal performance of individual amplifiers which would otherwise have to be made for the purpose of avoiding instability during parallel operation.

According to the invention, there is provided an arrangement for'connecting between two sections of an electric signal transmission line two similar negative feedback amplifiers in parallel without producing instability, each of the amplie fiers having input and output transformers, comprising means for causing the frequency of series resonance of the input and/or output transformer of one amplifier to differ by a predetermined amount from the frequency of parallel resonance of the input and/or output transformer of the sists in an arrangement for obtaining stable operation of two similar parallel connected neg ative feedback amplifiers inserted at a point in an electric signal transmission line comprising circuit means connected in tandem with one of the amplifiers, and adapted to prevent either amplifier from modifying the feedback loop of the other in such manner as to render it unstable.

According to a further aspect, the invention lies in an amplifying arrangement for an electric signal transmission line comprising a first negative feedback amplifier connected to the line and a second similar amplifier adapted to be connected in parallel with the first through circuit elements, the amplifiers and elements being designed so that the polar curve representing the transmission round the 11,6 path of either amplifier" does not enclose the point (1,0), whether the amplifiers are operated singly or in so parallel. L

According to still another aspect, the invention provides an amplifying arrangement for anelectrio signal transmission line comprising two similar negative feedback amplifiers connected in 25 parallel thereto, means being provided for rendering dissimilar the transmission characteristics of the s loops of the two amplifiers in such a manner that the parallel combination is stable.

The invention will be described with reference to the accompanying drawings, in which:

Fig. 1 shows a fragmentary schematic of a negative feedback amplifier to which the inven= tion is adaptable;

Figs. 2, 3, 4, 6 and 7 show schematic circuit diagrams used to explain the eflect of the transformers onthe feedback loop of the amplifier;

Figs. 5, 8, 9 and 10 show curves to illustrate the transmission conditions in the feedback loop;

40 Fig. 11 shows a block schematic diagram of an arrangement according to the invention. 7

In accordance with the practice which is no common in the art, a negative feedback amplifier will be regarded as comprising a forward amplitying path of which the. voltage transfer. ratio is denoted by 1., and a feedback path of which the voltage transfer ratio is denoted by ,8. Both a, and B are in general complex quantities and take account of the phase change as well as the corresponding galn or loss. The loop comprising the forward and feedback paths is known as the feedback loop, and s represents the voltage transfer ratio in magnitude and phase for one transit round the loop. For convenience, the feedback 455 loop is often called the as long and the gain the n? gain.

measured in decibels for a transit round the loop is proportional to log |;l,&|, and is often called In order to obtain the best advantage from feedback, is usually large compared with 1, so that the effective gain which the amplifier introduces into the circuit is substantially determined by 1/5.

A negative feedback amplifier of a type used on co-axial cables is shown diagrammatically in Fig. 1. This figure only shows the details necessary for an understanding .of the invention, and it will be understood that the amplifier may be otherwise designed in any appropriate way.

The amplifier comprises an input transformer IT which connects the line (represented by the impedance Z) to the control grid of the first thermionic amplifying valve V1, and an output transformer OT for coupling the anode of the last valve V2 to the line. The anode of V1 may be coupled in any suitable way, including further amplifying stages if necessary, to the control grid of V2; this coupling is indicated in Fig. 1 by the dotted line connecting this anode and control grid. If the amplifier should be a single stage amplifier, then the anode of V1 would be connected directly to the output transformer OT. It will be understood, of course, that V1 and V2 may all represent two or more valves connected in parallel, and may have any number of additional grids (not shown), appropriately polarised.

A feedback resistance R is connected, as shown.

in series with the anode circuit of the last valve V2, an appropriate blocking condenser K being included if necessary. The secondary wind ng of the input transformer IT is connected between the control grid of the first valve V1 and the high potential terminal of E. This is an ordinary current feedback arrangement, by which a potential proportional to the output current is fed back to the control grid of the first stage.

The phase shift for transmission once round the 5 loop should ideally be 180 or an odd multiple thereof at all frequencies. In practice, of course, this can never be achieved, and it is well known that the amplifier may become unstable if the phase shift reaches zero or a multiple of 360 at any frequency. The condition for stability in this case is that the 1 -8 gain must be zer or less at the frequency at which the phase shift is 360. Detailed information on the stabil ty conditions and design of these amplifiers may b found in the papers in the Bell S stem Techn cal Journal by H. Nyquist (Jan. 1932 pa es 126 to and H. W. Bode (July, 1940 pages 421 to In order to meet the stability reou rernents it is necessary to control the amplifier perf rmanc not only over the working frequency ra but also outside this range. For example. it is nieces sary to ensure that as the freouenc is raised above the working range, the :8 gain becomes zero or negative before the, phase shift has changed to 360 or a multiple thereof. The components which influence the up characteristics of the loop must therefore be proportioned accordingly. and the input and output transformers play an important part in this connection. The desired characteristics in the neighbourhood of the upper cut-off frequency may be substantially secured by suitably choosing the frequency at which the leakage reactance of each transformer resonates with its self capacity.

When a successful design has been achieved, by

aseasea which the best possible advantage has been secured from the negative feedback in the working range. consistent with a margin of stability which is sumcient to cover safely manufacturing variations, ageing, and the like, it will probably be found that when two similar amplifiers are connected in parallel to the line, both become unstable because the presence of each amplifier changes the transformer network eflective in the p loop of the other and the stability conditions are violated. To prevent this it has hitherto been necessary to sacrifice something in the design of the amplifiers, so obtaining a less eflicient arrangement. The reason for the instability caused by parallel operation will be presently explained, and it will be shown how the difiiculty may be avoided according to the invention without the necessity for sacrificing anything in the individual amplifier design.

In broad band high frequency amplifiers it is practically necessary to employ pentode valves. For this reason the output impedance of the last valve is very high and the output current is practically independent of the load, assuming that the feedback resistance R is small compared with the impedance of the whole anode circuit loop.

As a result the input and output transformers of such amplifiers affect the #3 loop in substantially the same way, and it is therefore necessary only to consider one of them and the conclusions will then equally apply to the other. Only the output transformer will therefore be considered in detail, with reference to Figs. 2, 3 and 4.

In Fig. 2 is shown the equivalent circuit of the output transformer from the terminals I, 2 as seen from the anode circuit of the last valve V2.

In this figure the symbols have the followin meanings:

Fig. 2 is similar to part of Fig. 101!) shown on page 189 of Radio Engineering by 1". E. Terman. second edition, 1937.

In the neighbourhood of the cut-off frequency, the quantities Rh, Re, L-p and Cm can all be neglected in comparison with the other quantities, and the circuit then becomes simplified a shown in Fig. 3 in which all the quantities have been referred to the primary side of the transformer and in which:

in which n is the impedance transformat on ratio.

Fig. 4 shows the network presented to the output current i1 (Fig. 1). In this figure Co is the output capacity of the last valve stage, and R is the feedback resistance, whichappears in series with the circuit of Fig. 3. If i: is the current through R, and since R is negligible compared with the impedances in either of the other branches of the network, it can be shown that, approximately (i=w'C'i'Rs (where w is 2w times the frequency) m=La/'1'R:i

Fig. shows the curves relating log 1' (curve A) and e (curve B) to log a, for a typical amplifier. It will be seen that r is practically equal to l at low frequencies, rises slightly and then falls rather steeply at mid-frequencies and then rises again to a practically constant value less than 1 at high frequencies. The final constant value is due to the fact that the network reduces at high frequencies to the two condensers C1 and C2, the impedance of L: having risen to such an extent that its effect in shunting C1 is negligible.

When another amplifier is connected to the line in parallel with that shown in Fig. 1, the transformer OT will have another similar transformer bridged across its secondary winding at the terminals 3 and t. Since the second amplifier-is the same as the first, the bridging impedance can be determined from Fig. 2. Fig. 6

shows a simplified circuit of the second transformer, in which C: has been omitted because its impedance is efi'ectively reduced to a negligible value by the feedback. R has accordingly plied by n is thus shunted across the nZ portion of Rs Fig. 4, but since nZ is large compared with R1+n Rz, the error in considering the whole of Ra so shunted will be small. The arrangement is then substantially as shown in Fig. 7. Rather complicated expressions for r and tan similar to (1) and (2) above may be derived from simple network theory, but it is simpler to exhibit the effect in a diagram.

It will be seen that the seriesresonance of the path shunting Ra occurs at substantially the same frequency as the parallel resonance of the network as a whole, and since 19B; is small,

the resistance R3 is practically short-circuited. 60

The curves of Fig. 5 now appear as shown in Fig. 8. The phase change c is now 1 This greatly sharpens the resonance of the first transformer.

greater than 90 over a range of mid-frequencies, and the ratio r is greater than 1 over part of this range, so that the conditions result in serious instability.

Fig. 9 shows the 14,8 gain expressed on a decibel scale (curve A) and phase shift (curve B) related to 10g f, where j is the. frequency, for the conditions of Fig. 7. It will be seen that the phase shift has become 360 or more over a short mid-frequency range, and there is a net gain round the loop over part of this range. The amplifier will therefore be expected to sing.

, Fig. 10 is similar to Fig. 9 and shows the curves which would be obtained for either amplifier alone, that is, when not shunted by another amplifier. It will be seen that the phase shift does not reach 360 until well after the gain has become negative so that there is a very good margin of stability.

It may be pointed out in connection with Figs.

-9 and 10 that if they be drawn in the form of Nyquists polar diagrams, the point (1,0) is enclosed in the case of Fig. 9 and not in the case of Fig. 10, so that his condition for stability is satisfied in the second case and not in the first.

It may further be mentioned that the parallel connection of the amplifiers may not always result in actual singing, but evenso, the state of the circuit is one of conditional stability in which although the point (1,0) is not enclosed, the configuration of the polar diagram is such that some small change in s such as might be due to ageing, replacement of valves and the like would cause the point to be enclosed, and thelamplifier's would sing. Such a condition cannot be regarded as satisfactory.

According to the invention, the instability caused by connecting the second amplifier in parallel with the first is prevented by separating the resonance frequencies, whose coincidence is the cause of the trouble, by an amount calcu-' lated to produce a sumcient margin of stability. One way in which this may be done is to connect the second amplifier at each end through an impedance network which slightly modifies the impedance presented to the transformer of the first amplifier so that it resonates at a different frequency. This is shown in block schematic form in Fig. 11. In this figure, A1 is the ampli her which is normally connected in the circuit and A2 is a spare amplifier which is to replace it for a short period. Two networks N1 and N: are connected in tandem with the amplifier A2 at the input and output, respectively. S1 and S: are intended to represent any suitable switching or jacking system for connecting the amplifiers in parallel. After the switches have been closed, the amplifier A1 may be removed by other jacks or switches not shown, so that the amplifier A: carries the tramc in the meantime. N1 and N2 should therefore be designed to introduce a very small loss. If the amplifier A1 is defective or othewise expected to be switched out for a long period, another normal amplifier may be switched in parallel with A2, which may afterwards be removed by opening the switches S1 and S2, and it will then be available for use again. v

Since it is necessary to change the resonance frequency of the impedance bridging the amplifier A1, the networks N1 and Ni will generally be composed of reactive elements, and in one preferred arrangement each consists of a single inductance connected in series with the conductor corresponding to the central conductor of the co-axial cable; or ina balanced system, two equal inductances would be used, one connected in series with each wire. The effect is to increase one of the inductances L3 in Fig. '1, (that is. effectively to increase the leakage inductance of the corresponding transformer) so that the two resonance frequencies are separated. It will be seen that this will increase Lo in shunt with R: for A1 and L3 in series with R3 for A1.

It has been found that by choosing a'suitable inductance for connecting in series with the input and output terminals of the amplifier A:

arrangement is thus obtained. In a certain.

practical case of amplifiers used in a co-axial cable system, the inductance connected in series with the central conductor was about 4% microhenrys.

It should be noted that according to the invention, the means for preventing instability are connected externally to the spare amplifier and accordingly it does not require the provision of any special amplifier for switching purposes. This is an important consideration in the interests of standardisation.

What is claimed is: v p

1. An arrangement i'or connecting between two sections of an electric signal transmission line two similar negative feedback amplifiers in parallel without producing instability, each of the amplifiers having input and output transformers, comprising means for causing the frequency ofseries resonance of one of said transformers of one amplifier to differ by a predeter-' mined amount from the frequency of parallel resonance of the corresponding transformer of the other amplifier.

2. An arrangement for obtaining stable op eration of two similar parallel connected negative feedback amplifiers inserted at-a point in an electric signal transmission line comprising circuit means connected in tandem with. one of ass-sass signed so that the polar curve representing the transmission ,round the up path of either amplifier does not enclose the point (1.0) whether the amplifiers are operated singly or in parallel.

4. An amplifying arrangement for an electric signal transmission line comprising two similar negative feedback amplifiers connected in parallel thereto, means being provided for rendering dissimilar the transmission characteristics of the p13 loops of the two amplifiers in such manner that the parallel combination is stable. I

5. An arrangement according to claim 4 in which the said means is adapted effectively to operate at frequencies outside the signal frequency 18.1185.

6. An arrangement according to claim 4 in which the said means comprises an impedance network connected externally to one of the amplifiers. I

7 An arrangement according to claim 1 in which the said means comprises means for eflectively increasing the leakage inductance of one of the transformers.

8. An arrangement according to claim 4 in which the said means comprises two impedance conductors leading to and from the amplifier.

ALLEMAN HOLLY ROCHE. 4 THOMASWIIHAM ELLIOTT.

Images