14

The Network Functions and Feedback1

Wai-Kai Chen
University of Illinois, Chicago

14.1 Blackmans Formula......................................................... 14-1 14.2 The Sensitivity Function.................................................. 14-6

We now study the effects of feedback on amplier impedance and gain and obtain some useful relations among the return difference, the null return difference, and impedance functions in general. Refer to the general feedback conguration of Figure 13.6. Let w be a transfer function. As before, to emphasize the importance of the feedback element x, we write w = w(x). To be denite, let w (x) for the time being be the current gain between the output and input ports. Then, from (13.24) we obtain

w ( x) =
yielding

I pq Is

Y2Vpq
Is

Yrp,sq ( x ) Yuv ( x )

Y2

(14.1)

w ( x ) Yrp,sq ( x ) Yuv (0) F ( x ) = = w (0) Yuv ( x ) Yrp,sq (0) F ( x ) provided that w(0) 0. This gives a very useful formula for computing the current gain:

(14.2)

w ( x ) = w (0)

F (x ) F (x )

(14.3)

Equation (14.3) remains valid if w(x) represents the transfer impedance zrp,sq = Vpq /IS instead of the current gain.

14.1 Blackmans Formula

In particular, when r = p and s = q, w(x) represents the driving-point impedance zrr,ss(x) looking into the terminals r and s, and we have a somewhat different interpretation. In this case, F(x) is the return difference with respect to the element x under the condition Is = 0. Thus, F(x) is the return difference for the situation when the port where the input impedance is dened is left open without a source and

References for this chapter can be found on page 16-17.

14-1
2006 by Taylor & Francis Group, LLC

14-2

Circuit Analysis and Feedback Amplier Theory

we write F(x) = F(input open circuited). Likewise, from Figure 13.6, F(x) is the return difference with respect to x for the input excitation Is and output response Vrs under the condition Is is adjusted so that Vrs is identically zero. Thus, F(x) is the return difference for the situation when the port where the input impedance is dened is short circuited, and we write F(x) = F(input short-circuited). Consequently, the input impedance Z(x) looking into a terminal pair can be conveniently expressed as Z ( x ) = Z (0) F (input short circuited) F (input open circuited)

(14.4)

This is the well-known Blackmans formula for computing an active impedance. The formula is extremely useful because the right-hand side can usually be determined rather easily. If x represents the controlling parameter of a controlled source in a single-loop feedback amplier, then setting x = 0 opens the feedback loop and Z(0) is simply a passive impedance. The return difference for x when the input port is short circuited or open circuited is relatively simple to compute because shorting out or opening a terminal pair frequently breaks the feedback loop. In addition, Blackmans formula can be used to determine the return difference by measurements. Because it involves two return differences, only one of them can be identied and the other must be known in advance. In the case of a single-loop feedback amplier, it is usually possible to choose a terminal pair so that either the numerator or the denominator on the righthand side of (14.4) is unity. If F(input short circuited) = 1, F(input open circuited) becomes the return difference under normal operating condition and we have F (x ) = Z (0) Z (x )

(14.5)

On the other hand, if F(input open-circuited) = 1, F(input short-circuited) becomes the return difference under normal operating condition and we obtain F (x ) = Z (x ) Z (0)

(14.6)

Example 1. The network of Figure 14.1 is a general active RC one-port realization of a rational impedance. We use Blackmans formula to verify that its input admittance is given by Y =1+ Z3 Z 4 Z1 Z 2 (14.7)

Appealing to (14.4), the input admittance written as Y = Y(x) can be written as Y ( x ) = Y (0) F (input open circuited) F (input short circuited)

(14.8)

where x = 2/Z3. By setting x to zero, the network used to compute Y(0) is shown in Figure 14.2. Its input admittance is: Y (0) = Z1 + Z 2 + Z 3 + Z 4 + 2 Z1 + Z 2 (14.9)

When the input port is open-circuited, the network of Figure 14.1 degenerates to that depicted in Figure 14.3. The return difference with respect to x is:

2006 by Taylor & Francis Group, LLC

The Network Functions and Feedback

14-3

1 Z2 2V3 Z3

Z4 I

Z1 1

V3 +

Z3

FIGURE 14.1 A general active RC one-port realization of a rational function.

Y(0)

Z1+Z2

1+Z3+Z4

FIGURE 14.2 The network used to compute Y(0).

1 Z2 2 Z3 Z1 1 V 3 + Z3 Z4

FIGURE 14.3 The network used to compute F(input open-circuited).

2006 by Taylor & Francis Group, LLC

14-4

Circuit Analysis and Feedback Amplier Theory

1 Z2 2/Z3 Z4 I Z1 1 V 3 + Z3

FIGURE 14.4 The network used to compute F(input short-circuited).

F (input open-circuited) = 1 V3 =

Z1 + Z 3 Z 2 Z 4 2 + Z1 + Z 2 + Z 3 + Z 4

(14.10)

where the returned voltage V3 at the controlling branch is given by V3 = 2(1 + Z 2 + Z 4 ) (14.11)

2 + Z1 + Z 2 + Z 3 + Z 4

To compute the return difference when the input port is short circuited, we use the network of Figure 14.4 and obtain F (input short-circuited) = 1 V3 = Z1 Z 2 Z1 + Z 2 (14.12)

where the return voltage V at the controlling branch is found to be 3 V3 = 2Z 2 Z1 + Z 2 (14.13)

Substituting (14.9), (14.10), and (14.12) in (14.8) yields the desired result. Y =1+ Z3 Z 4 Z1 Z 2 (14.14)

To determine the effect of feedback on the input and output impedances, we choose the series-parallel feedback conguration of Figure 14.5. By shorting the terminals of Y2, we interrupt the feedback loop, therefore, formula (14.5) applies and the output impedance across the load admittance Y2 becomes Z out ( x ) = Z out (0) F (x ) (14.15)

2006 by Taylor & Francis Group, LLC

The Network Functions and Feedback

14-5

Na Z1 Y2 Vs Nf

FIGURE 14.5 The series-parallel feedback conguration.

demonstrating that the impedance measured across the path of the feedback is reduced by the factor that is the normal value of the return difference with respect to the element x, where x is an arbitrary element of interest. For the input impedance of the amplier looking into the voltage source Vs of Figure 14.5, by open circuiting or removing the voltage source Vs, we break the feedback loop. Thus, formula (14.6) applies and the input impedance becomes Z in ( x ) = F ( x ) Z in (0) (14.16)

meaning that the impedance measured in series lines is increased by the same factor F(x). Similar conclusions can be reached for other types of congurations discussed in Chapter 12 by applying Blackmans formula. Again, refer to the general feedback conguration of Figure 13.6 If w(x) represents the voltage gain Vpq /Vrs or the transfer admittance Ipq /Vrs . Then, from (13.27) we can write w ( x ) Yrp,sq ( x ) Yrr ,ss (0) = w (0) Yrp,sq (0) Yrr ,ss ( x ) (14.17)

The rst term in the product on the right-hand side is the null return difference F(x) with respect to x for the input terminals r and s and output terminals p and q. The second term is the reciprocal of the null return difference with respect to x for the same input and output port at terminals r and s. This reciprocal can then be interpreted as the return difference with respect to x when the input port of the amplier is short circuited. Thus, the voltage gain or the transfer admittance can be expressed as w ( x ) = w (0) F (x ) F (input short-circuted)

(14.18)

Finally, if w (x)denotes the short circuit current gain Ipq /Is as Y2 approaches innity, we obtain w ( x ) Yrp,sq ( x ) Ypp,qq (0) = w (0) Yrp,sq (0) Ypp,qq ( x ) (14.19)

The second term in the product on the right-hand side is the reciprocal of the return difference with respect to x when the output port of the amplier is short-circuited, giving a formula for the short circuit current gain as w ( x ) = w (0) F (x ) F (output short-circuited) (14.20)

2006 by Taylor & Francis Group, LLC

14-6

Circuit Analysis and Feedback Amplier Theory

Again, consider the voltage-series or series-parallel feedback amplier of Figure 13.9 an equivalent network of which is given in Figure 13.10. The return differences F( k ), the null return ) and the voltage gain w were computed earlier in (13.45), (13.52), and (13.44), and are k differences F( repeated next: F (1 ) = 93.70, F ( 2 ) = 18.26 (14.21a) (14.21b) (14.21c)

F (1 ) = 103.07 103 , F ( 2 ) = 2018.70 w= V25 = w (1 ) = w ( 2 ) = 45.39 Vs

We apply (14.18) to calculate the voltage gain w, as follows: w (1 ) = w (0) where w (0) = Y12,55 (1 ) Y11,55 (1 ) =
1 =0

F (input short-circuited)

F (1 )

= 0.04126

103.07 103 = 45.39 93.699

(14.22)

205.24 10 12 = 0.04126 497.41 10 11 = 466.07 10 9 = 93.699 4.9741 10 9

(14.23a)

F (input short-circuited) = and w ( 2 ) = w (0) where w (0) = Y12,55 ( 2 ) Y11,55 ( 2 )

Y11,55 (1 ) Y11,55 (0)

(14.23b)

F (input short-circuited)

F ( 2 )

= 0.41058

2018.70 = 45.39 18.26

(14.24)

=
2 =0

104.79 10 10 = 0.41058 255.22 10 10 466.07 10 9 = 18.26 25.52 10 9

(14.25a)

F (input short-circuited) =

Y11,55 ( 2 ) Y11,55 (0)

(14.25b)

14.2 The Sensitivity Function

One of the most important effects of negative feedback is its ability to make an amplier less sensitive to the variations of its parameters because of aging, temperature variations, or other environment changes. A useful quantitative measure for the degree of dependence of an amplier on a particular parameter is known as the sensitivity. The sensitivity function, written as (x), for a given transfer function with respect to an element x is dened as the ratio of the fractional change in a transfer function to the fractional change in x for the situation when all changes concerned are differentially small. Thus, if w(x) is the transfer function, the sensitivity function can be written as x (x ) = lim0 w w = w x x x w 1n w =x x x (14.26)

2006 by Taylor & Francis Group, LLC

The Network Functions and Feedback

14-7

Refer to the general feedback conguration of Figure 13.6, and let w(x) represent either the current gain Ipq /Is or the transfer impedance Vpq /Is for the time being. Then, we obtain from (13.24) w ( x ) = Y2 As before, we write Y ( x ) Yuv ( x ) = uv x Y ( x ) Yrp,sq ( x ) = rp,sq x obtaining Yuv ( x ) = Yuv (0) + xYuv ( x ) Yrp,sq ( x ) = Yrp,sq (0) + xYrp,sq ( x ) Substituting (14.27) in (14.26), in conjunction with (14.29), yields (14.29a) (14.29b) (14.28a) Yrp,sq ( x ) Yuv ( x ) or Yrp, sq ( x ) Yuv ( x ) (14.27)

(14.28b)

(x ) = x Yrp,sq

(x ) x Yuv (x ) = Yrp,sq (x ) Yrp,sq (0) Yuv (x ) Yuv (0) Yuv ( x ) Yrp,sq ( x ) Yuv ( x ) rp ,sq ( x )

Y (0) Y (0) 1 1 = uv rp,sq = Yuv ( x ) Yrp,sq ( x ) F (x ) F (x ) Combining this with (14.3), we obtain

(14.30)

(x ) = F 1x ()
Observe that if w(0) = 0, (14.31) becomes

w (0) 1 w (x )

(14.31)

(x ) = F 1x ()

(14.32)

meaning that sensitivity is equal to the reciprocal of the return difference. For the ideal feedback model, the feedback path is unilateral. Hence, w(0) = 0 and = 1 1 1 = = F 1 + T 1 (14.33)

For a practical amplier, w(0) is usually very much smaller than w(x) in the passband, and F 1/ may be used as a good estimate of the reciprocal of the sensitivity in the same frequency band. A singleloop feedback amplier composed of a cascade of common-emitter stages with a passive network providing the desired feedback fullls this requirements. If in such a structure any one of the transistors fails, the forward transmission is nearly zero and w(0) is practically zero. Our conclusion is that if the failure of any element will interrupt the transmission through the amplier as a whole to nearly zero, the sensitivity is approximately equal to the reciprocal of the return difference with respect to that element.

2006 by Taylor & Francis Group, LLC

14-8

Circuit Analysis and Feedback Amplier Theory

In the case of driving-point impedance, w(0) is not usually smaller than w(x), and the reciprocity relation is not generally valid. Now assume that w(x) represents the voltage gain. Substituting (14.27) in (14.26) results in

(x ) = x Yrp,sq

(x ) x Yrr ,ss (x ) = Yrp,sq (x ) Yrp,sq (0) Yrr ,ss (x ) Yrr ,ss (0) Yrr ,ss ( x ) Yrp,sq ( x ) Yrr ,ss ( x ) rp ,sq ( x )

Y (0) Yrp,sq (0) 1 1 = rr ,ss = Yrr ,ss ( x ) Yrp,sq ( x ) F (input short-circuited) F ( x ) Combining this with (14.18) gives w0 1 (x ) = F input short-circuited 1 w (x) ( ) ( )

(14.34)

(14.35)

Finally, if w(x) denotes the short circuit current gain Ipq /Is as Y2 approaches innity, the sensitivity function can be written as Y (0) Y (0) 1 (x ) = Y pp,qq x Yrp,sq x = F output short-circuited 1 ( ) F (x ) pp ,qq ( ) rp ,sq ( ) which when combined with (14.20) yields w 0 1 (x ) = F output short-circuited 1 w (x) ( ) ( )

(14.36)

(14.37)

We remark that formulas (14.31), (14.35), and (14.39) are quite similar. If the return difference F(x) is interpreted properly, they can all be represented by the single relation (14.31). As before, if w(0) = 0, the sensitivity for the voltage gain function is equal to the reciprocal of the return difference under the situation that the input port of the amplier is short-circuited, whereas the sensitivity for the short circuit current gain is the reciprocal of the return difference when the output port is short-circuited. Example 2. The network of Figure 14.6 is a common-emitter transistor amplier. After removing the biasing circuit and using the common-emitter hybrid model for the transistor at low frequencies, an equivalent network of the amplier is presented in Figure 14.7 with

VCC RB1 C1 Rc C2

R1 + Vs RB2 RE CE R2

FIGURE 14.6 A common-emitter transistor feedback amplier.

2006 by Taylor & Francis Group, LLC

The Network Functions and Feedback

14-9

C
+

I23
+

R1 = (R1 + rx)||r

Is

gmV

V23

FIGURE 14.7 An equivalent network of the feedback amplier of Figure 14.6.

I s = G1 =

Vs R1 + rx

R2 = R2 || Rc

(14.38a)

1 1 1 = + R1 R1 + rx r 1 1 1 = + R2 R2 Rc

(14.38b)

G2 =

(14.38c)

The indenite admittance matrix of the amplier is: G1 + sC + sC sC G1 sC Y = g m sC G2 g m G2 + sC G1 sC g m G2 G1 + G2 + sC + g m

(14.39)

Assume that the controlling parameter gm is the element of interest. The return difference and the null return difference with respect to gm in Figure 14.7 with I as the input port and R , as the output port, are: s 2 F(gm ) = Y33 ( g m ) Y33

(G + sC )(G + sC ) + sC (G + g ) (0) (G + sC )(G + sC ) + sC G

=
1 2 2 m 1 2 2

(14.40)

Y ( g ) sC g m g F ( g m ) = 12,33 m = =1 m sC Y12,33 (0) sC The current gain I23 /I s as dened in Figure 14.7, is computed as w( g m ) = R2Y33 ( g m ) Y12,33 ( g m ) =

(14.41)

R2 (G1 + sC ) G2 + sC + sC (G2 + g m )

sC g m

(14.42)

Substituting these in (14.30) or (14.31) gives

(gm ) =

(sC

g m (G1 + sC ) G2 + sC + sC (G2 + g m )

g m G1 + sC + sC G2 + sC

)[

)(

(14.43)

2006 by Taylor & Francis Group, LLC

14-10

Circuit Analysis and Feedback Amplier Theory

Finally, we compute the sensitivity for the driving-point impedance facing the current source I. From s (14.31), we obtain

( g m ) = F (1 ) gm
where Z(gm ) =

sC g m Z (0) 1 = (G1 + sC ) G2 + sC + sC (G2 + g m ) Z(gm )

(14.44)

Y11,33 ( g m ) Y33 ( g m )

(G + sC )(G + sC ) + sC (G + g )
1 2 2 m

G2 + sC

(14.45)

2006 by Taylor & Francis Group, LLC