CHAPTER 7 FEEDBACK Feedback can be either negative (degenerative) or positive (regenera, tive, design, negative feedback is applied to effect one or more of the followi, "8 Prop 11. Desensitize the gain: that is, make the value of the gain less sensit the value of circuit components, such as might be caused by chan ‘ive toh BES in f . Reduce nonlinear distortion: that is, make the output Proportional to thet other words, make the gain constant, independent of signal level), 4 Reduce the effect of noise: that is, minimize the contribution to the Output electric signals generated, either by the circuit components themselves, orby interference. 4. Control the input and output impedances; that is, raise or lower the ‘input ang impedances by the selection of an appropriate feedback topology, Pl 5. Extend the bandwidth of the amplifier. All of the desirable properties above are obtained at the expense of a eduction in » be shown that the gain-reduction factor, called the amount of. feedback, is the j which the circuit is desensitized, by which the input impedance of a Volta, increased, by which the bandwidth is extended, and so on. In short, the basie tive feedback is to trade off gain for other desirable properties. This chapter is q the study of negative-feedback amplifiers; their analysis, design, and char (Under certain conditions, the negative feedback in an amplifier can become such’a magnitude as to cause oscillation. In fact, in Chapter 11 we will study the use gt tive feedback in the design of oscillators and bistable circuits Here, in this chapter hy, we are interested in the design of stable amplifiers) We shall therefore study the sunt” problem of negative-feedback amplifiers and theif potential for oscillation, re It should not be implied, however, that positive feedback always leads to insta fact, positive feedback is quite useful in a number of nonregenerative applications, Such a the design of active filters, which are studied in Chapter 10. Before we begin our study of negative feedback, we wish to remind the reader thar have already encountered negative feedback in a number of applications, Almost all opus circuits employ negative feedback. Another popular application of negative feedba the use of the emitter resistance Rg to stabilize the bias point of bipolar transistors and ‘increase the input resistance, bandwidth, and linearity of a BJT amplifi n addition, the eee nitter follower both employ a large amount of negative feedback. The question then arises about the need for a formal study of negative feedback. As will be ciated by the end of this chapter, the formal study of feedback provides an invaluable tool for the analysis and design of electronic circuits. Also, the insight gained by th terms of feedback can be extremely profitable. F Ces with GENERAL FEEDBACK STRUCTURE Figure 7.1 shows the basic structure of a feedback amplifier, Rather than sho} and currents, Fig. 7.1 is a signal-flow diagram, where each of the quantities x ca either a voltage or a current signal. The open-loop amplifier has a gain A: thus related to the input x; by X= Ax,7.1 THE GENERAL FEEDBACK STRUCTURE & 713 RE 7.1, General structure of the feedback amplifier. This i signal, “ag Seat either voltage or current signals, Saat flow diagram, and the quantities The output x, is fed to the load as well as to a feedback i network, which produces a sample of the ouput. This sample xis related to x, by the feedback taeny B OB. a2) The feedback signal xy is subtracted from the source signal x . '» Which is the input to the com- plete feedback amplifier,’ to produce the signal x,, which is the input to the basic amplifier, Xi = X,-2, Here we note that it is this subtraction that m: tive feedback reduces the signal that appears Implicit in the description above is that do not load the basic amplifier. That works. In practice this will not be the case, and we shall have to find a metho for casting a real circuit into the ideal structure depicted in Fig. 7.1. Figure 7.1 also implies that the for- ward transmission occurs entirely through the basi ic amplifier and the reverse transmission occurs entirely through the feedback network. The gain ofthe feedback amplifier can be obtained by combining Eqs. (7.1) through (7.3): (7.3) akes the feedback negativé. In essence, at the input of the basic amplifier] the source, the load, and the feedback network is, the gain A does not depend on am nega- 4) The quantity AB is called the loop gain, a name that follows from Fig. 7.1{For the feedback to be negative, the loop gain AB should be positive; that is, the feedback Signal x, should have the same sign as x,, thus resulting in a smaller difference signal x, Equation (7.4) indi- Lit that for positive AB the gain-with-feedback A, will be smaller that! the open-loop gain A by the quantity 1 + AB, which is called the amountof feedhack _ «a il:asis the case in many circuits, the loop gain ABis large, AB > 1, then from Eq. (7.4), it follows that A, ~ 1/B, which is a very interesting result: The gain feedback amp! Lieris almost entirely determined by the feedback network. Since the feedback network usu- ally consists of Passive components, which usually can be chosen to be as accurate as one Wishes, the advantage of negative feedback in obtaining accurate, predictable, and stable ' . 5 'neatier chapters, we used the subscript “sig” for quantities associated with the signal source (e.g., tieand Ray). We did tha to avoid confusion with the subscript “s,” which is usually used with pee Tote quantities associated with the source terminal of the transistor. At this point, however, itis sere that readers have become sufficiently familiar with the subject that the possibility of confu- Sion is minimal, Therefore, we will revert to using the simpler subscript s for signal-source quantities.ra & CHAPTER 7 FEEDBACK gain should be apparent. In other words, the overall gain will have very little dep the gain of the basic amplifier, A, a desirable property because the isusd tion of many manufacturing and application parameters, some of which migh tolerances, We have seen a dramatic illustration of all of these effects in o where the closed-loop gain (which is another name for the gain-with-| entirely determined by the feedback elements. Equations (7.1) through (7.3) can be combined to obtain the following exp feedback signal x,: {xp et Xs [95 Teap'7 Thus for AB > 1 we see that x/~ x,, which implies that the signal x; at the input oft amplifier is reduced to almost zero. Thus if a large amount of negative feedback is the feedback signal x; becomes an almost identical replica of the input signal x,. ; of this property is the tracking of the two_input terminals of an op-amp. Th between x, and x, whic! "is sometimes referred to as the “error signal.” ea = ‘Whe input differencing circui mixer.) An expres ion for x, can be easily determined as ~~~ from which we can verify that for A, x; becomes very small. Observe feedback reduces the signal that appears at the input terminals of the basic a amount of feedback, (1 + AB). ‘op-amp configuration shown in Fig. E7.1 provides ad7.2 SOME PROPERTIES OF NEGATIVE FEEDBACK € 7.2, SOME PROPERTIES OF NEGATIVE FEEDBACK - properties of negative feedback were mentioned in the Introduction. In the following, Tw fall consider some of these properties in more detail, wes! 72,4 Gain Desensitivity the effect of negative feedback on desensitizing the closed-loop gain was demonstrated in Brerse 7-1, where we saw that & 20% reduction in the gain of the basie amplifier gave rise twonly 20.02% reduction in the gain of the closed-loop amplified This sensitivity-reduction property can be analytically established as follows: Assume that Bis constant. Taking differentials Of both sides/ofEg: C/A) esate dA, = —A aa (1 + Apy Dividing Eq. (7.7) by Eq. (7.4) yields eo 1 dA | (1+ AB) A_) 8) of feedback, 1 + AB, is also Known as the desensitivity factor. JO Bandwidth Extension Consider an amplifier whose high-frequen cy response is characterized by Y a single pole. Its gain at mid and high frequencies can be expressed as = Au (Oa ae where Ay denotes the midband negative feedback, with a frequel Closed-loop gain A,(s) given by eG) aE 1+BA(s) Substituting for A(s) from Eq. (7.9) results, after a little manipulation, in — _Aw/( + AuB) ae 5S 1+ 8/@y(1 + AyB) oo) pas the feedback amplifier will have a midband gain of Ay/(1 +A yB) and an upper 3-dB “aUENCY Oy given by Duy = Oy(1 + AyB) aay Tt fot, es fedbacy. that the upper 3-dB frequency is increased by a factor equal to the amount of ing can be shown that if the open-loop gain is characterized by a dominant low- Pol © Biving rise to a lower 3-dB frequency a, then the feedback amplifier willS&B ciarters reeosack have a lower 3-dB frequency Oy, 72 Consider the noninverting op-amp circuit of Exercise 7.1. Let the open-loop gain A have a low-f value of 10° and a uniform ~6-dB/octave rolloff at high frequencies with a 3-dB frequency of Find the low-frequency gain and the upper 3-dB frequency of a closed-loop amplifier with and R)=9 kQ. ‘Ans. 9.99 V/V; 100.1 kHz 7,2,3° Noise Reduction Negative feedback can be employed to reduce the noise or interference in an amp! more precisely, to increase the ratio of signal to noise. However, as we shall now this noise-reduction process is possible only under certain conditions. Consider the illustrated in Fig. 7.2. Figure 7.2(a) shows an amplifier with gain A;, an input signal _hoise, or interference, V,,. It is assumed that for some reason this amplifier suffe noise and that the noise cai be assumed to be introduced at the input of the amp Yn‘N pace GB 717 7.2. SOME PROPERTIES OF NEGATIVE FEED! = jee ratio for this amplifier is noise F3 salt ” SIN = V,IV,) (7.13) a next the circuit in Fig. 7.2(b). Here we assume that it is possible to build another er Hijge with gain A> that does not suffer from the noise problem. If this is the case, amplifier A, by the clean amplifier A, and apply negative ade of such an amount as to keep the overall gain constant. " put voltage of the circuit in Fig. 7.2(b) can be found by superposition: i? V, = V; aa yee 7.14) ° V+A\AB "V4AVAB ¥ z the signal-to-noise rato at the output becomes thus the sig z (7.15) jis A> times higher than in the original case. which 1s “2 We emphasize once more that the improvement in signal-to-noise ratio by the applica~ tion of feedback is possible only if one can precede the noisy stage-by a (relatively) noise- _free stage. This situation, however, is not uncommon in practice. The best example is found wnihe output power-amplifier stage of an audio amplifier. Such a stage usually suffers from a problem known as power-supply hum The problem arises because of the large currents oe tarts stage draws from the power supply-and the difficulty in providing adequate power fe ) supply filtering inexpensively. The power-output stage is required to provide large Ppower\, /” ey cain but little or no voltage gain. We may therefore precede the power-output stage by a nce is required to be high, and the output impedance is ri at to be low.) nm» anal source is essentially a voltage source, it is convenient to repeat, ae ns of aprnevenif equivalent circuit] In a voltage amplifier the output quantity of interest Tr follows that the feedback network should sample the output voltage. eet Wy pe ourpat VOTE is pecause of the Thévenin representation of the source, fhe feedback si ay at also Peat can be mixed with the source voltage in series SE eae TO av ie : ertes) ae \ suitable feedback topology for the voltage amplifier is the voltage-mixing voltage- RU campling one shown in Fig. 7.4(a). Because of the series connection at the input and the Stale! oe shunt connection atthe output, this feedback topology is aso known as series Hee ‘hunt feedback. As will be shown, this topology not only Gtabilizes the voltage gain but aed iso results in a higher input resistance (intuitively, a result Of the series connection at the input) and a lower output resistance (intuitively, a result of the parallel connection at the “utput). which are desirable properties for a voltage amplifier The noninverting op-amp ssnfiguration of Fig. B71 is an example of series-shunt feedback. 73.2 Current Amplifiers The input signal in a current amplifier is essentially a current, and thus the signal source is most conveniently represented by its Norton equivalent. The output quantity of interest is current; hence the feedback network should sample the output current. The feedback signal ge should be in current form so that it may be mixed in shunt with the source current. Thus the Zz 2 feedback topology suitable for a current amplifier is the current-mixing current-sampling aN topology, illustrated in Fig. 7.4(b). Because of the parallel (or shunt) connection at the input, te and the series connection at the output, this feedback topology is also known as shunt— series feedback. As will be shown,¢his topology not only stabilizes the current gain but also results in a lower input resistance, and a higher output resistance, both desirable properties fora current amplifier, An example of the shunt-series feedback topology is given in Fig. 7.5. Note that the bias details are not shown. Also note that the current being sampled is not the output current, but the equal current flowing from the source of Q,. This use of a surrogate is done for circuit- design convenience and is quite usual in circuits involving current sampli . 5 The reference direction indicated in Fig. 7.5 forthe fosdack cutent is sack Se subtracts from J,. This reference notation will be followed in all circuits in this ¢ SiR itis consistent with the notation used in the general feedback structure of Fig. 7.1. Inall cicuts, therefore, for the feedback to be negative, the loop gain AB should be Ptive Te reader is urged to verify, through qualitative analysis, that in the cireuit of is. 75, Ais negative and B is negative t ; Itis of utmost ni to be able to ascertain qualitatively (and quickly) the ee ‘ck polarity (positive or negative). This can be done by “following the ae ee loop.” Ba) ne : We see that the gate voltage of P.” For instance, let the current J, in Fig. 7.5 increase. R24 the eae 1 Will increase, and thus its drain current will also increase. This will cause the dr Voltage of Q ‘i f Q,) to decrease, and thus the drain current of O3, 1, Will doorecet (and the gate voltage of Q2) 1. decreases. From the feedback network we ; tase, Thus the source current of Qn wn) will increase. THe increase in IpwillCHAPTER 7 FEEDBACK $ yzomiau __ eqpaeg S-adeijoa Surxmu-wuaimns (p) :ABojodor (s213s-s91128) Buy odor (iunys-souss) Surdures-a8eyjoa Burxrus-oBeyos (w) :sorBojodor Yu4P2} 1Sey AN} ML WL AUN -quaumn> Supwu-s8eOr (9) :KBoAOUOr (SIUTS UNS) © syomiau 4eqP224 2uomiau gorapaed soyyidwie /soueionpuoosuesy| aseg suomiu WoVgpoo4 “pyydue * Ty "1 quauin _3156q