Lect08 PDF
Lect08 PDF
Feedback
Ching-Yuan Yang
Outline
The General Feedback Structure
Some Properties of Negative Feedback
The Four Basic Feedback Topologies
The Series-Shunt Feedback Amplifier
The Series-Series Feedback Amplifier
The Shunt-Shunt and Shunt-Series Feedback Amplifier
Determining the Loop Gain
The Stability Problem
Effect of Feedback on the Amplifier Poles
Stability study Using Bode Plots
Frequency Compensation
xo A
xo = Axi , xf = βxo , xi = xs − xf Af ≡ =
x s 1 + Aβ
β : feedback factor
Aβ : loop gain
1 + Aβ : amount of feedback
If Aβ >> 1, then Af ≈ 1/β.
The gain of the feedback amplifier is almost entirely determined by the
feedback network.
Aβ
Feedback signal: x f = xs
1 + Aβ
For Aβ >> 1 we see that xf ≈ xs , which implies that xi 0. (xi : error signal)
Microelectrics (III) 8-3 Ching-Yuan Yang / EE, NCHU
Some Properties of Negative Feedback
dA f 1
=
dA (1 + Aβ )2
dA
dA f =
(1 + Aβ )2
dA f 1 dA
=
Af (1 + Aβ ) A
Bandwidth Extension
Consider an amplifier whose high-frequency response is characterized by a
single pole.
AM AM : the midband gain
A(s ) =
s
1+ ωH : the upper 3-dB frequency
ωH
Apply the negative feedback with a frequency-independent factor β :
AM
A(s ) 1 + AM β
Close - loop gain A f (s ) = =
1 + βA(s ) 1 + s
ω H (1 + AM β )
The feedback amplifier: (high-frequency response)
AM
Midband gain AMf =
1 + AM β
Upper 3-dB frequency ωHf = ωH (1 + AM β )
The amplifier band width is increased by the same factor, (1 + AM β ), by which
its midband gain is decreased, maintain the gain-bandwidth product constant.
Similarly, the low-frequency response of the feedback amplifier has a lower 3-dB
ωL
frequency ω Lf =
1 + AM β
(clean amp.)
A1A2 A1
Vo = Vs + Vn
1 + A1A2 β 1 + A1A2 β
S Vs
∴ = A2
N Vn
Example
v i = V+ − V−
Vs u vi u Vo u V− u vi v enegative feedback
Microelectrics (III) 8-13 Ching-Yuan Yang / EE, NCHU
Current Amplifiers
Characteristics:
Input signal: current output signal: current
Current-controlled current source
Low input impedance
High output impedance
Feedback topology
Current-sampling shunt-mixing (shunt-series) topology
The feedback network samples the output current, and the feedback signal
xf is a current that can be mixed in shunt with the source current.
“Shunt” refers to the connection at the input and “series” refers to the
connection at the output.
Characteristics:
⌦ Stabilize the current gain
⌦ Lower input resistance
(shunt connection at the input)
⌦ Higher output resistance
(series connection at the output)
Example
Is u ii (Ib1) u Ic1 u Vc1 v Ic2 (Io) v Ie2 (Io /α) v If u ii (Ib1) v enegative feedback
Example
Example
Norton’s
theorem
Is u ii u Vo v If u ii v enegative feedback
Microelectrics (III) 8-19 Ching-Yuan Yang / EE, NCHU
Two-Port Network Parameters
y parameters
Equivalent circuit:
z parameters
Equivalent circuit:
Equivalent circuit:
g parameters
Equivalent circuit:
Vo A
The close-loop gain A f ≡ = (A and β have reciprocal units)
Vs 1 + Aβ
Input resistance
Vs Vs V V + βVo V + βAVi
Rif ≡ = = R i s = Ri i = Ri i = Ri (1 + Aβ )
I i Vi /Ri Vi Vi Vi
Vt
Output resistance Rof ≡
I
Vt − AVi
where I = and Vi = −V f = − βVo = − βVt (for Vs = 0)
Ro
V + AβVt
I = t
Ro
Vt Ro
Rof ≡ =
I 1 + Aβ
The voltage-sampling feedback reduces the output resistance by (1 + Aβ).
Z o (s )
General form Z of (s ) =
1 + A(s )β (s )
Rif and Rof are the input and output resistances, respectively, of the feedback
amplifier, including Rs and RL.
The actual input and output resistances of the feedback amplifier usually
exclude Rs and RL :
1
Rin = Rif − Rs and Rout =
1 1
−
Rof R L
(h parameters)
The source and load resistances should be lumped with the basic amplifier.
In addition, the resistances in the feedback should be also lumped with the
basic amplifier.
h-parameter two-port feedback network:
Neglect h21, the forward transmission effect of the feedback network, and
thus omit the controlled source h21I1.
h21I1 is neglected.
V1
Determine β : β = h12 ≡ (port 1 open-circuited)
V2 I1 = 0
A circuit:
β circuit:
V 'f R1
β ≡ = ≈ 10 −3 V/V
V 'o R1 + R2
Rout ≈ 100 Ω
Equivalent circuit
Io A
The close-loop gain A f ≡ = (A and β have reciprocal units)
V s 1 + Aβ
Input resistance
Vs Vs V V + βVo V + βAVi
Rif ≡ = = R i s = Ri i = Ri i = Ri (1 + Aβ )
I i Vi /Ri Vi Vi Vi
V
Output resistance Rof ≡
It
In this case, Vi = −V f = − βI o = − βI t (for Vs = 0)
V
Rof ≡ = Ro (1 + Aβ )
It
The voltage-sampling feedback increases the output resistance by (1 + Aβ ).
Rif and Rof are the input and output resistances, respectively, of the feedback
amplifier, including Rs and RL.
The actual input and output resistances of the feedback amplifier usually
exclude Rs and RL :
Rin = Rif − Rs and Rout = Rof − R L
(z parameters)
The source and load resistances should be lumped with the basic amplifier.
In addition, the resistances in the feedback should be also lumped with the
basic amplifier.
z-parameter two-port feedback network:
Neglect z21, the forward transmission effect of the feedback network, and
thus omit the controlled source z21I1.
z21I1 is neglected.
V1
Determine β : β = z12 ≡ (port 1 open-circuited)
I2 I1 = 0
Determine A:
Vc1 − α (RC1 rπ 2 )
'
= = −14.92 V/V
Vi re1 + [RE1 (RF + RE 2 )]
Vc 2
Vc1
{ }
= −gm 2 RC 2 (h fe + 1)[re 3 + (R E 2 (RF + RE1 ))] = −131.2 V/V
I o' I 1
= e3 = = 10.6 mA/V
Vc 2 Vb 3 re 3 + (RE 2 (RF + RE1 )
I o' Vc1 Vc 2 I o'
⇒ A≡ = × × = 20.7 A/V
Vi ' Vi ' Vc1 Vi '
β circuit:
Determine β :
V f' RE 2
β= = × RE1 = 11.9 Ω
I o' RE 2 + RF + RE1
Io A 20.7
Closed-loop gain: A f ≡ = = = 83.7 mA/V
Vs 1 + Aβ 1 + 20.7 × 11.9
Vo − I c RC 3
Voltage gain: = ≈ − A f RC 3 = −83.7 × 10 −3 × 600 = −50.2 V/V
Vs Vs
Input resistance of the A circuit: Ri = (h fe + 1)[re1 + (RE1 (RF + RE 2 ))] = 13.65 kΩ
Output resistance Ro of the A circuit: The resistance looking between Y and Y’.
RC 2
Ro = [RE 2 (RF + RE1 )] + re 3 + = 143.9 Ω
h fe + 1
Output resistance of the feedback amplifier:
Rof = Ro (1 + Aβ ) = 35.6 kΩ
Homework #1
Problems 6, 14, 24, 25, 27, 37
Vo A
The close-loop gain A f ≡ = (A and β have reciprocal units)
I s 1 + Aβ
Input resistance
Vi I i Ri I Ii Ii Ii Ri
Rif ≡ = = Ri i = Ri = Ri = Ri =
Is Is Is Ii + I f I i + βVo I i + βAI i 1 + Aβ
Vt
Output resistance Rof ≡
I
Vt − AI i
where I = and I i = − I f = − βVo = − βVt (for I s = 0)
Ro
V + AβVt
I = t
Ro
Vt Ro
Rof ≡ =
I 1 + Aβ
The voltage-sampling feedback reduces the output resistance by (1 + Aβ).
Z o (s )
General form Z of (s ) =
1 + A(s )β (s )
Rif and Rof are the input and output resistances, respectively, of the feedback
amplifier, including Rs and RL.
The actual input and output resistances of the feedback amplifier usually
exclude Rs and RL :
1 1
Rin = and Rout =
1 1 1 1
− −
Rif Rs Rof R L
(y parameters)
The source and load resistances should be lumped with the basic amplifier.
In addition, the resistances in the feedback should be also lumped with the
basic amplifier.
y-parameter two-port feedback network:
Neglect y21, the forward transmission effect of the feedback network, and
thus omit the controlled source y21I1.
y21I1 is neglected.
I1
Determine β : β = y12 ≡ (port 1 short-circuited)
V2 V1 = 0
Small-signal analysis
A circuit:
Transfer gain:
(
Vπ = I 'i Rs R f rπ )
V 'o = −gmVπ (R f RC )
V 'o
A=
I 'i
Input resistance of the A circuit (
= −gm (R f RC ) Rs R f rπ )
Ri = Rs || Rf || rπ = 1.4 kΩ = −358.7 kΩ
Output resistance of the A circuit
Ro = RC || Rf = 4.27 kΩ
I 'f 1 1
β ≡ =− =−
V 'o Rf 47kΩ
Equivalent circuit
Io A
The close-loop gain A f ≡ = (A and β have reciprocal units)
I s 1 + Aβ
Input resistance
Vi I i Ri I Ii Ii Ii Ri
Rif ≡ = = Ri i = Ri = Ri = Ri =
Is Is Is Ii + I f I i + βI o I i + βAI i 1 + Aβ
V
Output resistance Rof ≡
It
In this case, I i = − I f = − βI o = − βI t (for I s = 0)
V
Rof ≡ = Ro (1 + Aβ )
It
The current-sampling feedback increases the output resistance by (1 + Aβ ).
Rif and Rof are the input and output resistances, respectively, of the feedback
amplifier, including Rs and RL.
The actual input and output resistances of the feedback amplifier usually
exclude Rs and RL :
1
Rin = and Rout = Rof − R L
1 1
−
Rif Rs
(g parameters)
The source and load resistances should be lumped with the basic amplifier.
In addition, the resistances in the feedback should be also lumped with the
basic amplifier.
g-parameter two-port feedback network:
Neglect g21, the forward transmission effect of the feedback network, and
thus omit the controlled source g21V1.
g21V1 is neglected.
I1
Determine β : β = g12 ≡ (port 1 short-circuited)
I2 V1 = 0
A circuit:
Determine β :
I 'f RE 2
β ≡ =− = −0.254
I 'o RE 2 + R f
A conceptual feedback loop is broken at XX’ and a test voltage xt is applied. The
impedance Zt is equal to that previously seen looking to the left of XX’. The loop gain
Aβ = −Vr /Vt , where Vr is the returned voltage.
Vr
Aβ = −
Vt
1
Aβ = −
1 1
+
Toc Tsc
X Y
Rt
X’ Y’
Break XX’ to determine the loop gain Aβ :
Rx
+ Rt
+ ∵ Vr = AxVt
Vt Vx AxVx Rt Vr R x + Rt
− − Vr Rt
Aβ = − = − Ax
Vt R x + Rt
Rx
+ + ∵ Voc = AxVt
Vt Rt Vx AxVx Voc Voc
∴ Toc ≡ = Ax
− − Vt
1 1
Thus, − =− = Aβ
1 1 1 Rx
+ +
Toc Tsc Ax Ax Rt
Microelectrics (III) 8-75 Ching-Yuan Yang / EE, NCHU
Equivalence of Circuits from a Feedback-Loop Point of View
Nyquist Plot
Nyquist criterion:
If the intersection occurs to the left of (−1, 0), the magnitude of loop gain at
this frequency is greater than unity and the amplifier will be unstable.
If the intersection occurs to the right of (−1, 0), the amplifier will be stable.
Find poles: The poles of the feedback amplifier are the zeros of 1 + A(s)β (s).
1 + A(s)β (s) = 0 (the characteristic equation of the feedback loop.)
1
ω0 =
CR
1
Q=
3−K
Frequency compensation
There exists a maximum value for β above which
the feedback amplifier becomes unstable.
Alternatively, there exists a minimum value value for
the closed-loop gain Af 0 below which the amplifier
becomes unstable.
To obtain lower values of the closed-loop gain one needs to alter the loop
transfer function L(s). This is the process of frequency compensation.
A( jω1 )β = 1 × e − jθ
where θ = 180o − phase margin
1/ β
⇒ A f ( jω1 ) =
1 + e − jθ
⌦ For a phase margin of 45o, θ = 135o,
1
A f ( jω1 ) = 1.3 Gain peaking
β
⌦ The peaking increase as the phase margin
is reduced, eventually reaching ∞ when
the phase margin is zero.
⌦ The closed-loop gain at low frequencies is
approximately 1/β .
105
A( f ) =
⎛ f ⎞⎛ f ⎞⎛ f ⎞
⎜1 + j 5 ⎟⎜1 + j 6 ⎟⎜1 + j 7 ⎟
⎝ 10 ⎠⎝ 10 ⎠⎝ 10 ⎠ 3.2×106 Hz
and
2
⎛ f ⎞
A = 100 − 20 log 1 + ⎜ 5 ⎟
⎝ 10 ⎠
2 2
⎛ f ⎞ ⎛ f ⎞
− 20 log 1 + ⎜ 6 ⎟ − 20 log 1 + ⎜ 7 ⎟
⎝ 10 ⎠ ⎝ 10 ⎠
⎡ f f f ⎤
φ = − ⎢tan−1 ⎛⎜ 5 ⎞⎟ + tan−1 ⎛⎜ 6 ⎞⎟ + tan−1 ⎛⎜ 7 ⎞⎟⎥
⎣ ⎝ 10 ⎠ ⎝ 10 ⎠ ⎝ 10 ⎠⎦
Frequency Compensation
A’ :Introduce an additional
pole at fD .
poles: fD, fP1, fP2, fP3, ⋅⋅⋅
β = 10−2
1 1
f P1 = f 'D =
2πC x R x 2π (C x + CC )R x
Microelectrics (III) 8-97 Ching-Yuan Yang / EE, NCHU
Miller Compensation and Pole Splitting
A gain stage in a multistage amplifier before compensation
Vo (sC f − gm )R1R 2
=
I i 1 + s[C1R1 + C 2R2 + C f (gm R1R2 + R1 + R2 )] + s 2[C1C 2 + C f (C1 + C 2 )]R1R2
⎛ s ⎞⎛ s ⎞ ⎛ 1 1 ⎞ s2
D (s ) = ⎜⎜1 + ⎟⎟⎜⎜1 + ⎟⎟ = 1 + s ⎜⎜ + ⎟⎟ +
⎝ ω 'P 1 ⎠⎝ ω 'P 2 ⎠ ⎝ ω 'P 1 ω 'P 2 ⎠ ω 'P 1 ω 'P 2
s s2
≈1+ + (ω’P1 << ω’P2 )
ω 'P 1 ω 'P 1 ω 'P 2
1 1
⇒ ω 'P 1 = ≈ (8.87)
C1R1 + C 2R2 + C f (gm R1R2 + R1 + R2 ) gm R2C f R1
gmC f
ω 'P 2 ≈ (8.88)
C1C 2 + C f (C1 + C 2 )
jω jω
Before Compensation After Compensation
ρ ρ
Pole splitting
Microelectrics (III) 8-99 Ching-Yuan Yang / EE, NCHU
Example 8.6: Frequency Compensation
The open-loop transfer function of the operational amplifier:
105
A( f ) =
⎛ f ⎞⎛ f ⎞⎛ f ⎞
⎜1 + j 5 ⎟⎜1 + j 6 ⎟⎜1 + j 7 ⎟
⎝ 10 ⎠⎝ 10 ⎠⎝ 10 ⎠
We wish to compensate the op amp so that the closed-loop amplifier with resistive
feedback is stable for any gain (that is, for β up to unity) and PM (phase margin) ≥ 45o.
Assume that the op amp circuit includes a stage such as that of Fig.a with C1 = 100pF,
C2 = 5pF, and gm = 40mA/V, that the pole at fP1 is caused by by the input circuit of that
stage, and that the pole is connected either between the input node B and ground or
in the feedback path of the transistor.
Fig.a:
Find the compensating capacitor CC connected across the input terminals of the
transistor stage.
1
Thus f 'D = 10Hz = ∴ CC = 1μF
2π (C1 + CC )R1
1
f 'P 1 = 100Hz = ∴ C f = 78.5pF
2πgm R2C f R1
Conclusion: Using Miller compensation not only results in a much smaller compensating
capacitor but, owing to pole splitting, also enables us to place the dominant pole a
decade higher in frequency.