You are on page 1of 24

University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

Amplifier Frequency Response

Amplifier Frequency Response ............................................................................................................ 1


1 Review: s-Domain Analysis: Poles, Zeroes, Bode Plots ............................................................. 2
Exercise 1: .................................................................................................................................... 4
2 The Amplifier Transfer Function ................................................................................................. 5
2.1 The Gain Function ............................................................................................................... 5
3 Determining the Low-Frequency Response ................................................................................. 6
3.1 Using short-circuit time constants (SCTC) to approximate L ........................................... 6
3.2 Analysis of the Common-Source Amplifier at Low Frequency .......................................... 7
Exercise 2: .................................................................................................................................... 8
3.3 Analysis of the Common-Emitter Amplifier at Low Frequency ......................................... 8
Example 1: ................................................................................................................................... 9
Exercise 3 ................................................................................................................................... 10
4. High-Frequency Response & Miller’s Theorem........................................................................ 11
4.1 High-Frequency Models for Devices ................................................................................. 11
1 The BJT High-Frequency Hybrid- Model ....................................................................... 11
2 The FET High-Frequency Model....................................................................................... 12
4.2 Miller’s Theorem ............................................................................................................... 13
5 Determining the High-Frequency Response .............................................................................. 14
5.1 Using open-circuit time constants (OCTC) to approximate H ......................................... 14
5.2 High-Frequency Response Calculation Examples ............................................................. 15
Example 2: (Common-Source FET Amplifier) ..................................................................... 15
Example 3: (Common-Emitter BJT Amplifier) .................................................................... 18
Exercise 4: .................................................................................................................................. 19
Example 4: (Common-Base BJT Amplifier)......................................................................... 20
Exercise 5: .................................................................................................................................. 21
Example 5: (Common-Collector or Emitter-Follower BJT Amplifier) ................................ 22
Exercise 6: .................................................................................................................................. 23
5.3 High-Frequency Response: Multistage Amplifier ............................................................. 24

Text References:

Sedra & Smith: Microelectronic Circuits


4th Ed. Chapter 7; 5th Ed. Appendix E, Chapters 4, 5 & 6; 6th Ed. Chapter 8.
or
Jaeger & Blalock: Microelectronic Circuit Design
2nd Ed. Chapters 10 & 17; 3rd Ed. Chapters 10 & 16; 4th Ed. Chapters 10 & 17.

V M Srivastava Page 1
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

1 Review: s-Domain Analysis: Poles, Zeroes, Bode Plots


Here we will be concerned with the amplifier voltage gain as a transfer function of complex
frequency s: T(s)  Vo (s) Vi (s) . This can be evaluated for physical frequencies (sinusoidal
excitation) by replacing s by j and obtaining the magnitude and phase responses.
For the circuits dealt with here T(s) can be expressed in the form:

Vo (s) a m s m  a m 1s m 1  .....  a 0


T(s)  
Vi (s) s n  b n 1s n 1  .....  b 0 (1)

where a & b are real numbers, n = order of the network, m < n, and for a stable circuit the roots of
the denominator all have negative real parts.

An alternative form of expressing T(s) is:

(s  Z1 )(s  Z 2 )....(s  Z m )
T (s )  a m 
(s  P1 )(s  P2 )....(s  Pn ) (2)

where Z1, Z2, .…, Zm are the roots of the numerator polynomial and called the zeros
and P1, P2, …., Pn are the roots of the denominator polynomial and called the poles of the network.

The number of poles in a transfer function = number of independent energy-storing elements


(capacitors or inductors) in the network. The number of finite zeros is (nm) less than the number
of poles.

Single Time Constant (STC) networks have been studied previously1 e.g. first order low-pass and
high-pass networks.

For a first order low-pass network:

a0
a0 0 1 1 (3)
T (s )    K  T(j )  K 
s  0 s s 
1 1 1 j
0 0 0

where K = dc gain and 0 = pole frequency = 1/ for the STC network
This network has a real pole at s = 0 and a zero at s = 

This results in the familiar Bode plots of magnitude and phase response for a first order low-pass
network: (normalizing for K=1)
1 
T  Av A v dB  20 log10 ; A v   arctan
2 0
  
1   
 0 

1
Sedra & Smith: Section 1.6 and Appendix F (4th Ed.) or D (5th Ed.); Jaeger & Blalock: Section 10.7

V M Srivastava Page 2
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

Gain

A V dB

-20 dB/dec

 (log scale)
0
Phase
Shift

Oo  (log scale)
 AV
-45o

-90o

Note how the (s  0 ) or 1  s 0  term in the denominator (i.e. the pole at s = 0) gives the
familiar low-pass response.

Similarly, a (s  0 ) or 1  s 0  term in the numerator (i.e. a zero at s = 0) will result in a


magnitude response with a corner frequency at 0 followed by a response rising at + 20 dB/decade.

Gain

A V dB +20 dB/dec

 (log scale)
0

For a pole or zero at s = 0, the Bode magnitude plot is simply a straight line with a slope of
−20 dB/decade (for a pole) or +20 dB/decade (for a zero), intersecting the 0 dB line at ω = 1 rad/s.

Gain Gain
-20 dB/dec +20 dB/dec
A V dB A V dB

0  0 
(log scale) (log scale)
 

pole at s = 0 zero at s = 0

V M Srivastava Page 3
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

For a first order high-pass network:



j
s j 0
T (s )  K   T(j )  K   K (4)
s  0 j   0 
1 j
0

where K = gain when s or   


0 = 1/ for the STC network

Gain

A V dB +20 dB/dec

 (log scale)
0
Phase
Shift

+90o  (log scale)


 AV
+45o

Oo

Exercise 1:
10 s
An amplifier has the following voltage transfer function: T(s) 
 s  s 
1  2 1  5 
 10  10 
Find the poles and zeros and sketch the magnitude of the gain versus frequency and the phase of the
transfer function versus frequency (i.e. the Bode magnitude and phase plots).

(To check the solution: see Sedra & Smith: 4th Ed. p.587-589 or 5th Ed. App. E p.E4-E6)

V M Srivastava Page 4
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

2 The Amplifier Transfer Function


Most amplifiers have a bandpass frequency response similar to that shown below. If the amplifier
is direct (or dc) coupled, then the gain will not fall off at low frequencies. The amplifier gain is
almost constant over a wide frequency range called the midband. In this frequency range all
capacitances (coupling, bypass, and transistor internal capacitances) have negligible effects and can
be ignored in the gain calculations. At low frequencies the coupling and bypass capacitances no
longer acts as ideal short circuits and cause the gain to drop (unless dc coupling is used). At high
frequencies the transistor internal capacitances cause the gain to drop.

A V (dB) 3 dB

AM dc coupled

or f
L H (log scale)
fL fH

The extent of the midband is defined by the two frequencies fL and fH and the amplifier bandwidth
is usually defined as BW = fH − fL . Since usually fL << fH , BW ≈ fH

A figure of merit for the amplifier is the gain-bandwidth product defined as GBP  A M  BW
It is generally possible to trade off gain for bandwidth.

2.1 The Gain Function

The overall gain of the amplifier can be expressed in the general form A (s)  A M  FL (s)  FH (s)
where FL(s) and FH (s) are functions that account for the dependence of gain on frequency in the
low and high frequency regions respectively.

The midband gain is determined by analyzing the amplifier equivalent circuit assuming the
coupling and bypass capacitors are acting as perfect short circuits and that the transistor internal
capacitances (and stray wiring capacitances etc.) are perfect open circuits.

The low frequency transfer function A L (s)  A M  FL (s) is determined by analyzing the equivalent
circuit including the coupling and bypass capacitors, but assuming the transistor internal
capacitances are perfect open circuits i.e. FH (s)  1

The high frequency transfer function A H (s)  A M  FH (s) is determined by analyzing the
equivalent circuit including the transistor internal capacitances, but assuming the coupling and
bypass capacitors are perfect short circuits i.e. FL (s)  1

V M Srivastava Page 5
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

3 Determining the Low-Frequency Response


The function FL(s) which characterizes the low-frequency response of the amplifier, takes the
general form
(s  Z1 )(s  Z2 )    (s  ZnL )
FL (s)  (5)
(s  P1 )(s  P2 )    (s  PnL )

where P1, P2, …, PnL are positive numbers representing the frequencies of the nL low-frequency
poles and Z1, Z2, …, ZnL are positive, negative or zero numbers representing the nL zeros.

We are usually interested in the low-frequency response close to the midband so that we can
determine or modify the lower 3 dB frequency L. In many cases the zeros are at frequencies
much lower than L and hence are of little importance on determining L.

Usually one of the poles – say P1  has a much higher frequency than all the other poles. Thus for
s
frequencies close to the midband, FL (s)  which is the equation of a first-order high-pass
s   P1
network. In this case the low-frequency response is dominated by the pole at s = P1 and the
lower 3 dB frequency is approximately equal to P1 i.e. L  P1 ; f L  P1 2

This dominant-pole approximation can be made if the highest-frequency pole is at least two
octaves (i.e. a factor of 4) higher in frequency than the nearest pole or zero.

If a dominant low-frequency pole does not exist, an approximate formula for L is:

L  2 P1  2 P2  ....  22 Z1  22 Z2  .... ; f L  L 2

If the poles and zeros of the amplifier transfer function are known or can be determined easily, then
the technique above can be used to determine L. In many cases, however, it is not a simple matter
to determine the poles and zeros. In such cases, an approximate value for L can be found using the
following method.

3.1 Using short-circuit time constants (SCTC) to approximate L

The amplifier low-frequency equivalent circuit is analyzed considering one capacitor (Ci) at a time
while replacing all other capacitors with short circuits. The resistance (Ris) seen by this capacitor
and hence the associated time constant is then determined. This process is repeated for all
capacitors.
n
1 L 1
If one of the poles is dominant, an approximation for fL is: f L  
2 1 Ci R is
(6)

In a complex circuit it may be difficult to establish whether or not a dominant low-frequency pole
exists, but this method usually provides a reasonable estimate of fL, which is usually sufficient for a
“hand analysis”. It also provides the designer with considerable insight into how the various
capacitors affect the low-frequency response.

V M Srivastava Page 6
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

3.2 Analysis of the Common-Source Amplifier at Low Frequency

A typical FET CS amplifier circuit is shown below:

+ VDD

RG1 RD
C2
vo
Rsig C1

vsig RG2
RL
RS CS

The low frequency equivalent circuit including the coupling and bypass capacitors (ignoring ro) is
shown below:
Rsig C1 C2
g d
vo
vgs
gm.vgs
vsig s
Rin RD RL

RS CS

Rin = RG1 // RG2

The input loop is a simple STC high-pass network determined by C1 with a zero at dc and a low
1
frequency pole at P1  .
C1 (R sig  R in )

CS introduces a low frequency pole and a lower frequency zero. The resistance ROS looking back
into the source of the FET is 1/gm and thus the resistance seen by CS is RS//(1/gm). Hence the low
1 1
frequency pole set by CS is:  P 2  . The zero set by CS is given by Z 
 1  CS R S
CS  R S 
 g m 

In the output portion of the equivalent circuit, we again have a simple STC high-pass network
1
determined by C2 with a zero at dc and a low frequency pole at P3 
C 2 (R D  R L )

Having determined the poles and zeros for the amplifier, the low frequency 3 dB corner (cutoff)
frequency fL may be found as on page 6.

V M Srivastava Page 7
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

In design, since (1/gm) is usually a much smaller resistance than those determining the other poles
and zero, it would be wise to first choose CS to set  P 2   L  2  f L (so that P2 dominates) and
then set the other poles much lower in frequency (typically about 10 times lower).

Exercise 2:

See: Exercises: Chap. 1: Frequency Response. Do Exercise 2.

3.3 Analysis of the Common-Emitter Amplifier at Low Frequency

A typical RC-coupled CE amplifier circuit is shown below.

+ Vcc

RC
R1 C2
vo
Rsig C1

RL
vsig R2
RE CE

The low-frequency equivalent circuit of this CE amplifier including the coupling and bypass
capacitors (ignoring ro) is shown below.

Rsig C1 C2
b c
vo
vbe r
gm.vbe
e
vsig RB=
R1//R2 RC RL

RE CE

The “impedance reflection” property of the BJT indicates that at low frequencies the input
impedance of the amplifier includes the effect of CE, and thus C1 and CE interact. Trying to
determine the exact poles and zeros of the transfer function is thus not practical, and the method of
short-circuit time constants (SCTC) is the most appropriate to estimate fL.

V M Srivastava Page 8
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

First, set Vsig to zero (i.e. short-circuit) then set CE and C2 to infinity (i.e. short-circuits) and find the
resistance RC1 seen by C1. From the equivalent circuit:
R C1  R sig  R B r 

Next, set C1 and C2 to  (short-circuit) and find the resistance RE seen by CE. Using the resistance
reflection rule (neglecting ro) we obtain:

r  (R B R sig )
R E  R E R OE  R E
 1

Finally, set C1 and CE to  (short-circuit) and find the resistance RC2 seen by C2:

R C2  R L  R C

Hence using Eq. (6) (SCTC):


1  1 1 1 
fL      (7)
2  C1R C1 C E R E C 2 R C2 

1 
The frequency of the zero introduced by CE is Z  or f Z  Z which is usually much
CE R E 2
lower than L or fL justifying the approximation involved in using the method of short-circuit time
constants (SCTC).

In design, Eq. (7) can be used as follows: Since RE is usually the smallest resistance, the total
1
capacitance is minimized by selecting CE to make the term the dominant term in Eq. (7),
C E R E
1
e.g.  (say) 80% of L  80% of 2  f L and selecting the other capacitors (C1 and C2) so
C E R E
that the other two terms in Eq. (7) contribute 10% each to the value of fL.

The next greater standard capacitor value would be chosen in each case.

Example 1:

Assuming  = 100, VT = 25 mV, IC(q) = 1 mA, and ignoring ro, determine the low frequency 3 dB
corner frequency fL for this amplifier circuit using the method of short-circuit time constants
(SCTC).

Comment on the validity of this method in this case.

V M Srivastava Page 9
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

+ 12V

4k7
82k 4,7F
vo
50 10F

4k7
vsig 27k 100F
2k2

Solution:
  VT 100 25 103
r     2,5 k ; R B  27 k 82 k  20,3 k
IC 1 10 3
1 1
 R C1  50  (20,3 k 2,5 k )  2,28 k ;    43,9 rad / s
C1R C1 2,28k 10

2,5 k  (20,3 k 50)


R OE   25,3 
100  1
1 1
 R E  2,2 k 25,3   25,0  ;    400rad / s
C E R E 25,0 100

1 1
R C 2  4,7 k  4,7 k  9,4 k ;    22,6 rad / s
C 2 R C 2 9,4 k  4,7 

1 1 1  467
 L     43,9  400  22,6  467rad / s  f L  L   74,3 Hz
C1R C1 C E R E C 2 R C 2 2 2

Since the 1 C E R E term in the L expression is much larger than the others, this means that the pole
frequency set by CE dominates the low frequency response.
1 1
The frequency of the zero set by CE is Z    4,55 rad / s which is much
C E R E 100  2,2 k
lower than all the other frequencies. Hence the approximations involved in the short-circuit time
constant method are valid.

Exercise 3

See: Exercises: Chap. 1: Frequency Response. Do Exercise 3.

V M Srivastava Page 10
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

4. High-Frequency Response & Miller’s Theorem


4.1 High-Frequency Models for Devices

1 The BJT High-Frequency Hybrid- Model2

At high frequencies, the BJT model used needs to be extended to include the internal capacitance
effects and other resistances that become important in high frequency analysis. A popular model
for this is the hybrid- model as outlined below.
r (rbc)
rx (rbb')
b b' c

C(Cbc)

r (rbe) C ro (rce)
(Cbe)
gm.Vb'e
e e

rx = rbb' = base-spreading resistance. This is the ohmic resistance of the relatively lightly doped high
resistivity path between the external base connection and the internal active base region,
typically 25 - 200 Ω.

rπ = rb'e = dynamic slope resistance of forward-biased b-e junction = /gm = VT/ICq  hie

rµ = rb’c = feedback resistance. With r this gives the reverse voltage feedback effect produced by
base-width modulation or "Early" effect, the same effect as the hre term. Typically MΩ's,
and usually neglected in a simplified model.

ro = rce = output resistance = (VA+VCEq)/ICq  VA/ICq = 1/hoe. Usually neglected for low resistance
loads, but may be significant for high-resistance loads (such as active loads).

gm.Vb’e = voltage-controlled current source representing the main BJT current control action.
Transconductance (or mutual conductance) gm = Ic/Vb'e = ICq/VT = hfe / hie

Cµ = Cb'c = depletion capacitance of reverse-biased b-c junction - typically a few pF in a small-signal


BJT, strongly dependent on Vce  Vcb. It is essentially equal to Cob usually given on the
data sheet.

Cπ = Cb'e = input diffusion capacitance of forward-biased b-e junction, representing the effect of
stored charge in the base region. It is related to the unity-gain (or transition) frequency
(fT) of the BJT. Cπ = gm /(2..fT)  Cµ ≈ gm /(2..fT).
At very low collector currents it may be dominated by the residual depletion capacitance
of the b-e junction usually given on the data sheet as CEB or Cib (at ICq = 0).

2
Sedra & Smith 4th Ed. p.321-326, 5th Ed. p.485-490;
Jaeger & Blalock 2nd Ed. p.1297-1302, 1305-1308, 3rd Ed. p.1005-1009, 1012-1015.

V M Srivastava Page 11
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

The unity-gain (or transition) frequency (fT) of a BJT is the frequency at which the
(projected) short circuit current gain magnitude has dropped to unity i.e. where (f)= 1.

2 The FET High-Frequency Model3

Both MOSFETs and JFETs may be modelled by a simplified high-frequency model as follows:

Cgd
g d

Cgs ro Cds
gm.vgs
s s

Cgs, Cgd, & Cds represent the respective terminal capacitances.


Cds is usually much smaller than the other capacitances and is often neglected in a simple model.

These terminal capacitances are not usually specified on data sheets. Often the manufacturer
specifies more conveniently measured capacitances as follows:

d
Ciss = input capacitance in CS with output (d) shorted
Cgd
g
Crss = reverse capacitance in CS with input (g) shorted
Ciss Cgs
s
Coss = output capacitance in CS with input (g) shorted

Where:
d
C rss  C gd ; C iss  C gs  C gd ; C oss  C ds  C gd Cgd
and hence g
Coss
C gd  C rss ; C gs  C iss  C rss ; C ds  C oss  C rss Cds

The capacitances Cgs and Cgd are related to the unity gain frequency (fT) of the FET (in similar fashion
gm
to a BJT). f T 
2(C gs  C gd )

3
Sedra & Smith 4th Ed. p.441-447, 5th Ed. p.320-325; Jaeger & Blalock 2nd Ed. p.1302-1305, 3rd Ed. p.1009-1012.

V M Srivastava Page 12
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

4.2 Miller’s Theorem

This is a theorem applicable to many electrical circuits and of particular usefulness in analysing the
response of amplifiers at high frequencies.

It is applicable to a two-port network where an impedance Z (or admittance Y) connects between


(bridges) input and output, and where the voltage gain from input to output is known or can be
determined independently. In this case Miller’s theorem provides the means for replacing the
“bridging” impedance (or admittance) with two impedances (or admittances): Z1 (or Y1) across the
input port and Z2 (or Y2) across the output port.

I1 Z (Y) I2 I1 I2

V1 V2 V1 V2
V2 = Av.V1 Z1 Z2
(Y1) (Y2)

(a) (b)

With the gain Av known, the values of Z1 (Y1) and Z2 (Y2) may be determined as follows:

In diagram (a), the current drawn by Z (Y) at the input port is I1  ( V1  V2 ) Y  V1 (1  A v ) Y


For the circuit in (b) to be equivalent to that in (a) the current drawn by Z1 (Y1) must be equal to I1.

I1  V1Y1  V1 (1  A v ) Y Hence Z
Y1  Y(1  A v ) and Z1 
(1  A v )

In diagram (a), the current drawn by Z (Y) at the output port is I 2  ( V2  V1 )Y  V2 (1  1 A v )Y


For the circuit in (b) to be equivalent to that in (a) the current drawn by Z2 (Y2) must be equal to I2.

I 2  V2 Y2  V2 (1  1 A v ) Y Hence Z
Y2  Y(1  1 A v ) and Z 2 
(1  1 A v )

Thus, for example, if Z or Y is a capacitor C, then C may be replaced by a capacitor C1 across the
input and a capacitor C2 across the output, such that:

jC1  jC  (1  A v ) and hence C1  C  (1  A v ) (8)


jC2  jC  (1  1 A v ) and hence C2  C  (1  1 A v ) (9)

If Av is inverting and large (e.g. for a CE amplifier) say Av = 200, Cbc = 3 pF, then
C1  C bc  (1  A v )  3  201  603pF ; C2  C bc  (1  1 A v )  3  (1  1 200)  3 pF
Thus the small capacitance Cbc of only 3 pF appears as 603 pF across the input where it may
significantly affect the high-frequency response of the amplifier.

V M Srivastava Page 13
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

5 Determining the High-Frequency Response


The function FH(s) which characterizes the high-frequency response of the amplifier, takes the
general form
(1  s /  Z1 )(1  s /  Z2 )    (1  s /  ZnH )
FH (s)  (10)
(1  s /  P1 )(1  s /  P 2 )    (1  s /  PnH )
where P1, P2, …, PnH are positive numbers representing the frequencies of the nH
high-frequency real poles and Z1, Z2, …, ZnH are positive, negative or infinite numbers
representing the nH high-frequency zeros.

We are usually interested in the high-frequency response close to the midband so that we can
determine or modify the upper 3 dB frequency H. In many cases the zeros are either at infinity or
at such high frequencies that they are of little importance in determining H.

Usually one of the high-frequency poles – say P1  has a much lower frequency than any of the
1
other poles. Thus for frequencies close to the midband, FH (s)  which is the transfer
1  s /  P1
function of a first-order low-pass network. In this case the high-frequency response is dominated
by the pole at s = P1 and the upper 3 dB frequency H is approximately equal to P1 i.e.
H  P1 ; f H  P1 2

This dominant-pole approximation can be made if the lowest-frequency pole is at least two
octaves (i.e. a factor of 4) lower in frequency than the nearest pole or zero.

If a dominant high-frequency pole does not exist, an approximate formula for H is:
1 H
H  ; fH 
 1 1   1 1  2
   ....   2    .... 
  P1  P 2   Z1  Z2
2 2 2 2
 
If the poles and zeros of the amplifier transfer function are known or can be determined easily, then
the technique above can be used to determine fH. In many cases, however, it is not a simple matter
to determine the poles and zeros. In such cases, an approximate value for fH can be found using the
following method.

5.1 Using open-circuit time constants (OCTC) to approximate H

The amplifier high frequency equivalent circuit is analyzed considering one capacitor (Ci) at a time
while replacing all other capacitors with open circuits. The resistance (Rio) seen by this capacitor
and hence the associated time constant is then determined. This process is repeated for all
capacitors.
If one of the poles is dominant, and the zeros are not dominant, then an approximation for fH is:
1
fH 
n H  (11)
2   Ci R io 
 1 

V M Srivastava Page 14
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

In a complex circuit it may be difficult to establish whether or not a dominant high-frequency pole
exists, but this method usually provides a reasonable estimate of fH, which is usually sufficient for a
“hand analysis”. It also provides the designer with considerable insight into how the various
capacitors affect the high-frequency response.

5.2 High-Frequency Response Calculation Examples


Example 2: (Common-Source FET Amplifier)

A common-source FET amplifier has the following high-frequency equivalent circuit:

Rsig g Cgd 1 pF d vo
100 k Vgs
vsig Rin Cgs R'L
420 k 3,33 k
1 pF gm.Vgs

s
gm = 4 mS

Find the midband voltage gain AM and the high frequency 3 dB corner frequency fH.

Solution:

At midband frequencies Cgs and Cgd are open circuits, hence the midband equivalent circuit and
gain are as follows:
Rsig g d
vo Vo V Vgs
AM   o 
100 k Vgs Vsig Vgs Vsig
vsig R'L R in
Rin
 (g m R L )
420 k gm.Vgs 3,33 k R in  R sig
420k
s  (4m  3,3k )
gm = 4 mS 420k  100k
 10,8

To determine fH we shall use the method of open-circuit time constants (OCTC).

To find Rgs seen by Cgs, suppress (short) the source Vsig and let Cgd = 0.

Rsig g R
gs

100 k  R gs  R in R sig  420k 100k  80,8k


Rin
420 k Thus the open-circuit time constant is:
 gs  C gs R gs  110 12  80,8 103  80,8 ns

V M Srivastava Page 15
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

To find Rgd seen by Cgd, short Vsig and set Cgs = 0. The circuit is then as follows to which we apply
a test current Ix, calculate the corresponding voltage Vx and hence calculate Rgd = Vx / Ix.

Let R   R sig R in

Ix Writing a node equation at node g:


Vgs
g d Vgs  Vx Ix    Vgs  I x R 
R
Vgs Vgs Vx Vgs  Vx
R RL
Writing a node equation at d:
R' gm.Vgs R'L Vgs  Vx
80,8k 3,33 k I x  g m Vgs 
s R L
gm = 4 mS I x R  Vx
 g m I x R   
R L R L
Vx  R 
 I x 1  g m R   
R L  R L 

Hence
Vx
R gd   R L  R   g m R  R L  R L  (1  g m R L ).R   3,33k  (1  4m.3,33k ).80,8k  1,16 M
Ix
 gd  C gd R gd  1 10 12 1,16 10 6  1160 ns
1 1 1
H   rad/s f H   128,3 kHz
 gs   gd (80,8  1160) 10 9 2  (80,8  1160) 10 9

From this we can see that Cgd plays the major role in determining the high-frequency response of
the amplifier.

It is also instructive to apply Miller’s theorem to this circuit:

The midband voltage gain applicable to Cgd is A v  g m R L


Hence Cgd may be replaced by CM1 across the input and CM2 across the output, where:

C M1  C gd (1  A V )  C gd (1  g m R L ) and C M2  C gd (1  1 A v )  C gd (1  1 g m R L )  C gd

R'
80,8 k g d
vo
Vgs
Cgs CM1 CM2 R'L
gm.Vgs 3,33 k
1 pF

s
gm = 4 mS
CT

V M Srivastava Page 16
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

Both the input and output sections are now STC low-pass networks for which we can calculate the
time constants.

 i / p  C T R   (C gs  C M1 )  R   C gs R   C gd (1  g m R L )  R 
 o / p  C M 2 R L  C gd R L
 total  C gs R   C gd R L  1  g m R L   R  (12)
1 1
H  f H 
 total 2   total

In equation (12), the first term is seen to be τgs and the second term τgd from the method of
open-circuit time constants (OCTC). Thus the same total time constant and hence the same ωH or
fH is obtained. Hence if CM1 and CM2 are both taken into account, using Miller’s theorem yields the
same result as the method of open-circuit time constants (OCTC).

Note that we must not treat this Miller equivalent circuit as two single pole networks since both
CM1 and CM2 are derived from the same capacitor Cµ and are thus not independent.

Often the “Miller approximation” is used to quickly estimate the high-frequency response of an
amplifier. This then effectively neglects CM2 in the above analysis and only considers CM1 in the
input circuit. Since if the gain is large C M1  C M 2 , hence the error is usually insignificant.

It was previously noted that the method of open-circuit time constants (OCTC) assumed that a
dominant-pole existed. From the analysis given in texts4 it is shown that for a CS (or CE) amplifier,
a second pole and a finite zero exists with (approximate) frequencies of:

g g g g
P2  m ; Z  m for the CS ampl. and similarly P2  m ; Z  m for the CE ampl.
C gs C gd C C

[The exact frequency of P2 is given in texts5 and it will always be higher than the above
approximation. Since we usually only wish to know if it is much higher than P1 to validate our
dominant pole assumption, the approximation is usually adequate]

Using the values from example 3, these yield:

gm 4  103 4  109
P2    4  109 rad/s  f P2   637 MHz (exact value  734 MHz)
C gs 1  1012 2
gm 4  10 3 4  109
Z    4  109 rad/s f Z   637 MHz
C gd 1  10  12 2

As these are both very much higher than the estimate for fH, the dominant pole approximation does
indeed hold.

4
Sedra & Smith: 4th Ed. p.618-619, 5th Ed. p.592-593 (In 5th Ed., load capacitance CL is included in the analysis.);
Jaeger & Blalock: 2nd Ed. p.1309-1316, 3rd Ed. p.1017-1024.
5
Sedra & Smith: 4th Ed. Eqn. (7.64) p.619, 5th Ed. Eqn. (6.67) p.593;
Jaeger & Blalock: 2nd Ed. Eqn. (17.83) p.1312, 3rd Ed. Eqn. (16.95, 16.96) p.1020.

V M Srivastava Page 17
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

Example 3: (Common-Emitter BJT Amplifier)

For the CE amplifier given in Example 1 on page 9 (without RL connected), calculate the midband
voltage gain A M  Vo Vsig and the high frequency 3 dB corner frequency fH using Miller’s
theorem. For the BJT assume (from the data sheet) Cob = 4 pF; rx (rbb) = 150 ; fT = 100 MHz.

Solution:
Given data for BJT:
+ 12V
β = 100; IC (q) = 1 mA; Cob = 4 pF;
4k7 rx (rbb) = 150 ; fT = 100 MHz.
82k 4,7F

vo Hence:
50 10F
I (q ) 1mA
gm  C   40 mS
VT 25 mV
  VT 100 25 mV
r    2,5 k
vsig 27k 100F I C (q ) 1mA
2k2
rx  150 C   C ob  4 pF
gm
C   C
2  f T
40 10 3
  4 1012 F  60 pF
2 10010 6

The high-frequency equivalent circuit is thus:

Rsig b rx b' C 4pF c


vo
50 vi 150 vb' e
C
vsig RB r
RC
2,5k
20,3k 60pF 4,7k
gm.vb' e
e
Ri RTin

At midband frequencies:

R B  82 k 27 k  20,3 k ; R i  R B (rx  r )  20,3 k (150  2,5k )  2,34 k


Vo
Av   g m  R C  40 m  4,7 k   188
Vb'e
Vo V V V r Ri 2,5 k 2,34 k
AM   o  b'e  i  188    188    174
Vsig Vb'e Vi Vsig r  rx R i  R s 2,5 k  150 2,34 k  50
AM dB
 20 log 174  44,8 dB

V M Srivastava Page 18
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

Now using Miller’s theorem, replace Cµ by CM1 and CM2 across the input and output respectively

C M1  C (1  A v )  4 pF  (1  188)  756 pF ; C M 2  C (1  1 A v )  4 pF  (1  1 188)  4 pF

Finding the Thévenin equivalent resistance and total capacitance at the input:


R Tin  R sig R B  rx   r  50 20,3k   150 2,5k   49,9  150 2,5k  185 
CTin  C  CM1  60  756pF  816pF
RT in
Finding the total time constant: b' c
vb' e vo
in  C Tin  R Tin  816pF 185  151ns ;
 out  C M 2  R C  4 pF  4,7 k  18,8 ns CM2
CT in RC
 total  151 18,8  170ns gm.vb' e
1 1
fH   Hz  936kHz e
2   total 2 17010 9

It can be seen from this example that the large voltage gain acting on Cµ to produce a large CM1 (by
the Miller effect) is the main determinant of the high frequency response in conjunction with the
Thévenin equivalent input resistance.

The resulting frequency response is thus:

A (dB)

AM 44,8 dB

3 dB

f (log scale)
fL fH
74,3 Hz 936 kHz

Exercise 4:

See: Exercises: Chap. 1: Frequency Response. Do Exercise 4.

The “sum of open-circuit time constants” (OCTC) method and the corresponding application of
Miller’s theorem may be extended to cascaded stages6 [see Section 5.3, p. 24].

6
Jaeger & Blalock: 2nd Ed. p.1341-1348, 3rd Ed. p.1043-1050; Millman & Grabel: Microelectronics 2nd Ed. p.489-495.

V M Srivastava Page 19
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

Example 4: (Common-Base BJT Amplifier)

The dc bias circuit used in Example 3 may be connected as a common-base amplifier by appropriate
use of coupling and bypass capacitors, as shown below. The BJT parameters will be the same as in
Example 3.

Determine the high frequency corner frequency fH.

+ 12V
RC
R1 4k7 CO 4,7F
82k vo
Cin Rsig
100F
50
10F CB
R2 vin
27k RE vsig
2k2

The high-frequency equivalent circuit is shown below. For simplicity, rx has been neglected.
gm.vbe
Rsig iin c
e
vo
vin
C RC
vsig RE r
C
vbe ib
b

We may simplify the analysis by substituting the T model as follows:


gm.vbe
Rsig re
e c
vo

vin= -vbe
C C RC
vsig RE
vbe
ib
b

V M Srivastava Page 20
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

This simplified equivalent circuit clearly shows the most important feature of the common-base
circuit: the absence of an internal feedback capacitance. Unlike the common-emitter circuit, here
C has one terminal grounded, and thus no Miller effect occurs. We thus expect a much higher
upper cutoff frequency fH than that of the common-emitter circuit.

The high frequency poles may be determined directly from this simplified equivalent circuit.
1 1
At the input we have: P1  ; f P1 
C  (R sig // R E // re ) 2  C  (R sig // R E // re )

1 1
At the output we have: P2  ; f P2  . (With external RL, replace RC with RL)
C R C 2   C R C

Since re is usually very small, fP1 will usually be quite high, and since C is usually very small, fP2
will usually be quite high also.

Using values from Example 3:

1 1 1
re   25 ;  f P1   160MHz ; f P2   8,47 MHz
gm 2  60pF  (50 // 2k 2 // 25) 2  4pF  4k 7

Since one pole is obviously dominant, f H  f P2  8,47 MHz

Note the considerably increased high cutoff frequency fH of 8,47 MHz compared with the same
circuit used as a common-emitter amplifier in Example 3 where fH was only 936 kHz.

Comments:

1 This analysis has neglected the effect of rx for simplification, but this will reduce the
accuracy as we cannot assume its effect is negligible. Generally the effect of rx will be to
increase the effective resistance seen by C and so reduce fP2 and hence fH predicted by the
above simplified analysis7.

2 It is instructive to consider the effect of load capacitance on the CE and CB amplifier


circuits above.

In the CE amplifier (Example 3) the input circuit and CM1 largely determined fH, and hence
adding a moderate amount of load capacitance would not alter fH much. In the CB amplifier
(Example 4) the output circuit mainly determined fH and hence adding load capacitance
would directly increase the output time constant and hence reduce fH.

Exercise 5:

See: Exercises: Chap. 1: Frequency Response. Do Exercise 5.

7
For analysis including rx see: Jaeger & Blalock: 2nd Ed. p.1327-1329.

V M Srivastava Page 21
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

Example 5: (Common-Collector or Emitter-Follower BJT Amplifier)

Consider the following emitter-follower


+ VCC
circuit.

The BJT is biased at IC = 1 mA, and has the Rsig


following parameters: β = 100, rx = 100 Ω,
Cµ = 2 pF, and fT = 400 MHz. CO
1 k
vsig vo
Determine the midband voltage gain
A M  Vo Vsig and the frequency of the RL
I
1 k
dominant high-frequency pole.
- VEE
Solution:

Using the given dc bias condition, the relevant BJT model parameters are:
I 1 mA  100
gm  C   40 mS ; r    2,5 k ;
VT 25 mV g m 40 mS
gm 40 103
C   C   2 1012 F  13,9 pF
2  f T 2  400106

The ac equivalent circuit appropriate for high frequencies is as follows:


C
Rin b'
Rsig
b b' ie e
vo
rx ib r

vsig C
.ib
RL

c c

From this equivalent circuit, at mid-band frequencies:

R in b '  r    1 R L  2,5 k  1011k  103,5 k


Vo   1 R L 101k
Av     0,9759
Vb' R in b' 103,5 k

AM 
Vo

  1R L  101k

101k
 0,966
Vsig R sig  rx  R in b' 1k  100  103,5 k 104,6 k

Typically the emitter-follower circuit is used where the source resistance Rsig is large or significant
compared to the load resistance RL (as in this example). This will then usually result in a dominant
 
high-frequency pole determined largely by R sig  rx and the effective input capacitance.
C may be reflected to the input via the Miller theorem as CM1 where:

V M Srivastava Page 22
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

CM1  C 1  A v   13,9 1  0.9759 pF  0,335 pF [Note here CM1 << C]


The total input capacitance CT-in is thus: C T -in  C  C M1  2  0,335  2,335 pF
This gives a simplified equivalent circuit (of the input portion of the circuit) as follows:
Rsig rx
b b'
ib
CT-in 
vsig Rin b'
C  C M1
c

 
The total resistance at the input RT-in is: R T -in  R sig  rx // R in b'  1k  100 //103,5 k  1,09k
This gives an effective input time-constant of: in  C T - in  R T - in  2,335 pF 1,09 k  2,545 ns

1 1
This will usually form a dominant pole, hence: f H  f p1 (dom)    62,5 MHz
2  in 2  2,545ns

Even though this pole is usually dominant, its frequency is normally quite high, giving the
emitter-follower a very wide bandwidth, far wider than the other two BJT configurations (CE &
CB).

An alternative approach for finding an approximate value of fH is to use the method of open-circuit
time-constants (OCTC) on the high-frequency equivalent circuit given above (page 22).
(Using the same numeric values as the above example, this approach gives f H  55,3MHz 8.)

Fuller analysis of the high frequency equivalent circuit shows the circuit to have two poles and one
real zero9. The frequency of the zero is close to fT and hence usually plays a minor role in the high
frequency response. When Rsig is significant or large compared with RL then one pole usually
dominates and the analysis can be simplified as in Example 5 above. When Rsig is low, there may
not be a dominant pole, hand analysis is inappropriate, and CAD simulation should be used;
however, in such cases the need for the use of an emitter-follower is debatable.

Exercise 6:

See: Exercises: Chap. 1: Frequency Response. Do Exercise 6.

8
Sedra & Smith (4th Ed.) Exercise 7.18 p.629, (5th Ed.) Exercise 6.35 p.641
9
Sedra & Smith (4th Ed.) p.627

V M Srivastava Page 23
University of KwaZulu-Natal Analogue Electronics 2 ENEL3AE Amplifier Frequency Response

5.3 High-Frequency Response: Multistage Amplifier


+Vcc
The high-frequency response
using the OCTC method and the RE2 C3
corresponding application of RC1
RB1
Miller’s theorem may be extended
to cascaded stages (see page 19). Q2
Rsig C1 C4
Q1 VO
For example, consider the
Vsig
following 2-stage CE BJT RB2
RC2 RL
C2
amplifier: RE1

Figure 1

The small-signal equivalent circuit for high frequencies using the usual BJT model is:
Rsig RL Q1 RL Q2
b1 C c1 b2 C c2
Vo
vbe1 vbe2
Vsig r C Q1 r C Q2
RB RC1 RC2 RL
gm1.vbe1 gm2.vbe2
e e

RT in
Figure 2

Note: In applying Miller’s theorem to Fig. 2, the relevant Av must be used for each BJT,
i.e. when considering Q1 and C1 the voltage gain of stage 1 = A V1  Vbe2 Vbe1  g m1  R L Q1
must be used, and similarly for Q2.

Using Miller’s theorem and RT in


b1 c1 b2 c2 Vo
further simplification this vbe1 vbe2
yields the following RL Q1
simplified equivalent circuit: CT in
RL Q2
gm1.vbe1 Cmid gm2.vbe2 Cout
From Fig. 3 it can be seen
that the OCTC method may e e
then be very simply applied
using the time-constants as in mid out
shown.
Figure 3

Note: The OCTC method only determines the frequency of the lowest pole and hence an
estimate of fH. The frequencies of the other poles are NOT simply determined from the individual
time-constants of each stage (as is commonly mistaken!).

V M Srivastava Page 24

You might also like