CIRCUI'IS SYSTEMS SIGNAL PROCESS

VOL. 19, NO. 6, 2000, PIa. 4 8 7 - 4 9 9

MILLER' S THEOREM
REVISITED*
S. C. Dutta Roy l

Abstract. Miller's theorem is generally used for approximate analysis of high-frequency

amplifier circuits. That it can be used without much difficulty for the exact analysis of
such circuits as well as others, including passive networks, does not appear to be widely
recognized in the literature. Although formulated and proved in 1972, the dual of Miller's
theorem has hardly ever been used. In this paper, we present the exact analysis of a number
of circuits of practical interest with the help of either Miller's theorem or its dual or both,
almost by inspection.
Key words: Miller's theorem, dual of Milker's theorem, general network analysis.

1. Introduction

Miller's theorem and its dual generally do not even find a mention in textbooks
on network analysis. Textbooks on analog electronic circuits mention the theorem
only in the context of dealing with the effect of the feedback capacitance Cb,cor
Ct, in a Bipolar Junction Transistor (BJT), Cga in a Field Effect Transistor (FET),
and Cga in a vacuum tube [1], [2], [6]-[8], [11]-[I3], [15], where the symbols
have their usual meanings. Further, an approximation is made on the gain so as
to completely decouple the input and output circuits, thereby greatly simplifying
the analysis. Notwithstanding the fact that the results so obtained are not far from
the actual ones, one wonders why the exact analysis is discarded midway. The
reason, it appears to us, is the occurrence of the yet undetermined gain parameter
in the equivalent Miller impedances. We shall conclusively demonstrate in this
paper that this feature should not act as a deterrent, and that exact analysis can
still be carried out, almost by inspection, of active as well as passive networks.
The dual of Miller's theorem was discussed in [12] in 1972, and much later,

* Received April 4, 2000.

1 Department of Electrical Engineering, indian Institute of Technology, Hauz Khas, New
Delhi 110 016, India. E-mail: [email protected]
488 DbTTA RoY

in 1988, in [9]; however, except in ~:hese two references, no serious attempt was
made to apply it in network analysis. We give here an example of its application
in a passive circuit of practical interest. We also include an example of an active
circuit that can be analyzed with ease by using Miller's theorem as well as its
dual.
The paper is organized as follows. In Section 2, we present the essence of
Miller's theorem and its dual and show how the former can be applied to analyze,
exactly, the conventional high-frequency transistor amplifier. Section 3 presents
three examples of passive null networks that are analyzed by using Miller's theo-
rem. Section 4 contains another example of a passive null network whose analysis
is facilitated by the dual of Miller's theorem. In Section 5, we consider an example
of a high-frequency transistor amplifier, which can be analyzed with ease by
applying Miller's theorem and its dual.
In our experience, the efforts involved in network analysis using Miller's the-
orem or its dual are generally no more than, and in some cases less than, those
required in other alternative methods, including mesh or node analysis. Further,
the analysis can be carried out almost by inspection, coupled with some amount
of algebra.
Note that Miller's theorem has been generalized in [ 14] for all possible two-port
interconnections and illustrated by several examples of analysis and synthesis of
active networks,

2. Miller's theorem, its dual, and an application

Miller's theorem refers to the bridged network shown in Figure la, and replaces
the effect of ZB by two impedances Z1 and Z2 connected across ports 1 and 2,
respectively, as shown in Figure lb. If the two currents Ii and /2 are to be the
same in both the circuits, then we need

V1/ZI = (VI - V2)/ZB and V2/Z2 = ( V 2 - V1)/ZB (1)

or
Z1 = Z B / ( I -- H ) and Z2 = ZB/ ( 1 -- H-~), :,2)
where H = V2/V1 is the voltage transfer function of the network.
The dual of Miller's theorem refers to the network of Figure 2a and its equip.a-
lent shown in Figure 2b. The voltage drop across Zs is (I1 +/:2) Zs, and if V1 and
V2 in the two networks are to be identical, then we require

[IZI = (11 + I2)Zs and 12Z2 = (I1 + I2)Zs (3)

or
Zj -- (i ~- T)Z~, and Z2 = (1 -~- T-~)Zs, (4)
where T = I2/I 1 is the current transfer function of the network.
 MILLER'S THEOREM REVISITED 489

N I1
Vl+ + ~ Vt + +

1'~ ,, 2 ' 1 '~ - - - ~2 '

(a) (b)

Figure 1. (a) Original network. (b) Equivalent network obtained by Miller's theorem.

lc > < o2 1 2

1'o o 2 ' !'o ~_ Q 2'

(~) (b)

Figure 2. (a) Original network. (b) Equivalent network obtained by the dual of Miller's theorem.

As mentioned earlier, the occurrences of H in Z1 and Z 2 of Figure 1b and T

in Zl and Z2 of Figure 2b, which have yet to be determined, act as a deterrent
in proceeding with the exact analysis. As an example of the fact that this should
not be so, we consider the conventional application of Miller's theorem, which
concerns the common-emitter amplifier at high frequencies. The equivalent circuit
is shown in Figure 3a, where the notation has the usual meanings. Denoting V o / V
by H, the effect of C~ can be replaced by an impedance Z ~ / ( I - H) across C~
and another impedance Z y ( 1 - H -~) across RL, as shown in Figure 3b, where
Z~ = l / ( s C ~ ) , Z~r = l/(g~r + sGr), and Rts = Rs + rx. Very promptly, at
this point, one generally assumes that H ~- --gmRL and that IHI >> 1. Then
Zn / (1 - H) is a capacitor of value Cn (1 + gin RL), and Zu / (1 - H - 1) is another
capacitor approximately equal to Cn. The time constant RL C~ of the output
circuit is usually much smaller than that of the input circuit, which is (r~ IIR~)
[Czr + Cu (1 + g m RC ) ], so that the contribution due to Zu / (1 - H - 1) at the output
circuit can be completely ignored. It is not that these assumptions are far from
truth in practice, but rather than the end justifying the means, one could go ahead
with the exact analysis and then make the appropriate approximations. In this
particular case, the exact value of H can be found solely from the output circuit.
We have
Vo = --gin V { R L II[Z~/(1 -- H-~)I} (5)
490 DUTTA ROY

or

H = --gm/[GL + (I -- H-~)YIz]. (6)

H occurs on both sides of (6). By cross-multiplication and simplification (for
brevity, we shall refer to this process as CMAS), we get

H = ( G - g,,~)/(G + Go). (7)

From the input circuit,

V = V,{Z~ [l[Z,/(l - H)I}/[{Z~ [l[Z#/(l - H)]} + R',] (8)

so that
V l
~ss = I + R~s{Y~ + Y~(I - H)] (9)
The overall gain of the circuit is, therefore,

A--
"6 k V / k V, ] 1+G[G + I%(1 - H)] "
Substituting for H from (7) and simplifying, we obtain

,4 = gu-gm . (ii)
fp + GL + R'~[Y~(L + GL) q- Y.(G6 +gin)]
Finally, substituting for Ysr and Y~ and simplifying, we get the desired gain as

A : G. s - (g.,/C~n (i2)
C~ s 2 4-s{(GL/Cp) + [(G(~ + g~ 4-gm 4- GL)/Crc]} 4- [GL (G's 4- grr)/(C~Ciz)]
For a typical practical situation,

R's = 0.3 K, r~r = 2 K, C~r = 19.5 pF,

C~ = 0.5 pF, g,z = 0.05 S2-I , and R L = 0.6 K. (t3)

With these numerical values, (12) becomes

A = 17.! x 108(s - 1011)/[(s + 11 x 107)(s + 6.07 x 109)1. (14)

Equation (14) clearly shows that there is a dominant pole at - 11 x 107 radians/sec,
which therefore determines the - 3 dB bandwidth of the circuit as 1752 MHz.
Note that the approximate calculation gives the - 3 dB bandwidth as
!
= 17.44 MHz, (15)
27r(r~llR's)[C~ + C~(I + gmRL)]
which is close enough to the actual value and justifies the approximation made in
the conventional analysis.
Note that in this particular example, Miller's theorem for exact analysis offers
little advantage because it is a two-node problem, and eliminating V from the out-
put node equation and substituting in the other requires almost the same amount
of algebra.
 MILLER'S THEOREMREVISITED 49l

R +r C~

I
+ +
+
C~ ) RL
i v~
------o

gm V

-O
(a)

O
+

) 1 2
9
RL Vo
+

(b)

Figure 3. (a) High-frequencyequivalentcircuit of a common-emittertransistoramplifier.(b) Simpli-

fied formobtainedby applicationof Miller's theorem.

3. Application of Miller's theorem to passive networks

As the first example of a passive network, consider the bridged-T network shown
in Figure 4a, which is known to be a null network useful for the measurement
of inductance (L) and its associated resistance (r). To find H = V2/VI, we use
Miller's theorem to get the equivalent network shown in Figure 4b, where Z1 and
Z2 are given by (2) with Z e = r + sL. Because Zl appears across a voltage
source, it does not affect the transfer function. Also, let

YL = GL + Y2 = GL + [(1 -- H - 1 ) / ( r + sL)]. (16)

Then the network to be analyzed simplifies to that shown in Figure 4c. Application
of Thevenin's theorem to Figure 4c reduces the latter to Figure 4d, where p =
sCR. It is now easy to see that

v2 1 (I/YL)
V1 p + 1 (1/YL) + [ R / ( p + 1)] + ( R / p ) " (17)
492 DUTTAROY

C C

-- [_o
+

(-) (~)

1t
C p+l C

+
+
v~

o
(c) (a)

Figure 4. (a) The bridged-Y network. (b) Equivalent network obtained by application of Miiier's
theorem. (c) Simplified form of (b). (d) Network obtained by applyingThevenin's theorem to (c).

Substituting for Yc from (16) and simplifying gives

1
H = it8)
1 47 p 47 (2p 47 1){GL 47 [(1 -- H-l)/(r + sL)]}"
By CMAS, we finally obtain
[pZ(r + sL)/R] 47 2p + 1
H = (~9)
[p2(r + sL)/R] + 2p + 1 + (r + sL)[(p/R) + (2p + 1 ) G L ]
This agrees with the result in [4] obtained by a different method. For the special
case L = 2CRr, (19) reduces to
(p2r/R) + 1
H = (pZr/R) + (pr/R) + 1 + rGL(2p 47 1)' (20)
Clearly, the network behaves as a notch filter with null transmission at coo =
l / ( C ,/-RT).
In this example also, nodal analysis requires two node equations, and the use
of Miller's theorem offers no special advantage.
 MILLER'S THEOREM REVISITED 493

cl

+ + §
v2 v2

o
(~) (b)

R R(~+2)
-- p2+ap+l R

#h u c" i

(~) ] (a)

Figure 5. (a) The bridge&ladder network. (b) Simplified equivalent circuit obtained by using Miller's
theorem. (c) Result of applying Thevenin's theorem to (b). (d) Result of applying Thevenin'stheorem
to (c).

As the second example, we consider the bridged-ladder network of Figure 5a,

whose Miller equivalent is shown in Figure 5b, where YL = 112 + GL, Y2 =
sCj (1 - H-1), and the Miller impedance across the source has been omitted. This
is a ladder network and can be simply analyzed by the technique given in [10].
If instead, we apply Thevenin's theorem twice as shown in Figures 5b and c, we
obtain the circuit of Figure 5d, from which, by inspection, we get
1 (I/YL)
H = V2/VT -- p Z + 3 p + l (I/YL)+R+[R(p+2)/(p2+3p+I)]" (21)

Upon simplification, we get

1 . (22)
H= p2+3p+l+R(p2+4p+3)[GL+sCl(l_H_l)]
CMAS results in the following:
p(p2 + 3) + 4[p 2 + (o~/4)]
H = , (23)
p(p2 + 3) + 4[p 2 + (oe/4)] + ce(p + 3)[p + RGc(p + 1)]
where o~ = C/CI. Note that for ~ = 12, there exists a pair of transmission
zeros at p = l j ~ , f 3 , that is, there will be null transmission at coo = ,/3/(CR),
irrespective of the value of RL. The final transfer function is
(p2 + 3)(p + 4)
H -----(p2 + 3)(p + 4 ) + (p + 3)[p + RGL(p + 1)]' (24)
which agrees with the result of [4].
Note that nodal analysis in this case requires the inversion of a 3 x 3 matrix.
494 DUTTA ROY

R R

c g c

R
I1 .I2
tt R
~-O
(b)
R
+ +
2p+l R
2C - --
v~

(a) ~2p+l l -o

(e)

Figure 6. (a) The para!lel-T network. (b) Simplified equivalent circuit obtained from Miller-type
argument. (c) Result of applying Thevenin's theorem to (b).

It will therefore be appreciated that Miller's method, combined with Thevenin s

theorem, as used here, requires less effort for obtaining the final result.
Finally, in this section, we take the example of the well-known parallel-T net-
work shown in Figure 6a. Although this is not a bridged network like the two
previous examples, we can still use the essence of Miller's theorem by replacing
the upper-T network, for example, by two impedances Zl and Z2 so that the
currents that they draw are the same as I1 and 12, respectively. Because Z~ does
not affect the transfer function V2/V1, we need to find Z2 only. Clearly, from
Figure 6a,
I2 = (V2 - V)sC, (25)
where V is the potential across R/2. V can be related to V] and V2 by the node
equation
(V - V[)sC + (V - Vz)sC + [V/(R/2)] = O. (26)
Thus
V = [(VI + Vz)/2][p/(p + ~.)], (27)
where p = sCR, as used earlier. Combining (25) and (27) and simplifying, we
get
Y2 = (12/V2) = l - [(1 -? H - l ) / 2 ] [ p / ( p + !)]. (28)
The network to be analyzed then becomes that shown in Figure 6b, where YL =
Y2 + GL. Once again, a single application of Thevenin's theorem gives the equiv-
 MILLER'S THEOREM REVISITED 495

alent circuit of Figure 6c, from which we get, after simplification,

1
H = (29)
2p + 1 + 2R(p + 1){GL + sC(1 -- [(1 + H - l ) / 2 ] [ p / ( p -4- 1)])}"
By the usual technique of CMAS, we finally get
p2+l
H = (30)
p2 + 4 p + l + 2R(p + I)GL"
This result agrees with that of [4], and the network is seen to produce a null at
coO = 1/(CR).
As in the bridged-ladder network of the previous example, node analysis of
the parallel-T network will require solving three simultaneous equations, and it
should be obvious that our method requires less effort.

4. An application of the dual of Miller's theorem

Consider the network of Figure 7a, which is also known to be a null network [3].
Using (4), we get the equivalent network of Figure 7b. This is also a ladder
network and is most conveniently analyzed by the technique of [10]. To this end,
we define some intermediate variables in Figure 7b. Let ZL = Z2 + RL, and write,
by inspection,
v2= --I2RL (31)
v3= --I2ZL (32)
I3= V3sC - / 2 = -/2(1 + sCZL) (33)
V4= I3R + V3 = - I 2 [ R + ZL(1 + p)] (34)
t4= V4sC + 13 = - I 2 [ p + 1 + sCZL(2 + p)] (35)
vs= 14R -4- V4 = - I 2 [ R ( p + 2) + ZL(p 2 + 3p + 1)] (36)
II = VssC q- 14 = -12[p 2 q- 3p + 1 + s C Z L ( p 2 q- 4p + 3)] (37)
v~= I1Zt + V5 = -I2[(ZL + Z1)(p 2 + 3p + 1)
+sCZLZ1 (p2 q_ 4p + 3) + R(p + 2)]. (38)
Equation (37) gives
/2 -1
T _
I1 (p2+3p+l)+sC(p 2+4p+3)[RL+(R/12)(I+T-1)] ' (39)
which, by CMAS, results in
(pC + 3)(p + 4)
T = , (40)
(p2 + 3)(p + 4 ) + 12(p + 3)[1 + fi(p + 1)]
where fl = RL/R. Hence
R Rp(p + 3)[1 + fl(p + 1)]
Z1 = -i~(1 q- T) = ( p Z + 3 ) ( p § (41)
496 DUTTA ROY

I1 R R [2

....... VIB

o
(a)

va ~ v.~ Ia-. va

T T T "1
(b)

Figure 7. (a) The network under consideration. (b) Simplified equivalent circuit obtained by using the
dual of Miller's theorem.

and
R Rp(p Jr- 3)[1 + fl(p + l)]
z2 = 89 + r -~) = (42)
(p2+3)(p+4)
Also, from (3 I) and (38), we get
V2
--~H
V1
RL
(43)
(p2 + 3p + I)(R L + Z1 + Z2) + R(p + 2) + sCZI(RL + Z2)(p 2 + 4 p + 3 )
Substituting for Z1 and Z2 from (41) and (42), and simplifying, one can obtain
the required expression for H.
A somewhat tricky situation arises when RL --> c~, i.e., the output is open-
circuited so that T = I2/I1 = 0 and Z2 = (R/12)(1 + T -1) ---> oc. The
transfer function for this case can of course be derived as the limiting case of
the general expression (43); however, it makes more sense to put Rc = oc in
Figure 7b, let Z1 = (R/12), and replace Z2 by a dependent voltage source
I1(R/12) with positive polarity on the right. The analysis then becomes much
simpler, as indicated below, where we express all variables in terms of ~ .

V2 = V3 + I t ( R / 1 2 ) (44)
[3 = V3sC (45)
1/4 = /3 R q- V3 = V3(l q- ]9) (46)
 MILLER'S THEOREM REVISITED 497

14 = V4sC + I3 = V3sC(2 + p) (47)

V5 = /4R + V4 = V3(p 2 q- 3p + 1) (48)
11 = VssC + 14 = V3sC(p2 + 4p + 3) (49)
Vl = I i ( R / 1 2 ) + 1/5 = V3[(p 2 + 3p + 1) + ( 1 / 1 2 ) p ( p 2 + 4 p + 3)]. (50)
Also, from (48),
V2 = V 3 + ( 1 ~ R / 1 2 ) = V3[1 + ( 1 / 1 2 ) p ( p 2 + 4 p + 3 ) ] . (51)
Hence
V2 (p2 + 3)(p + 4)
H -- -- (52)
V1 (p2 + 3)(p + 4) + 12p(p + 3)'
which agrees with the result derived in [3] by an alternative method. Clearly, this
is also a null network with zero transmission at coo = ,/-J/(CR).
In common with the previous two examples, the network of Figure 7a requires
the solution of three node equations; application of Miller's theorem appears to
facilitate a reduction of effort, although marginally.

5. An application of both Miller's theorem and its dual

As the last example in this paper, we consider the high-frequency response of

the common-emitter transistor amplifier with unbypassed emitter resistance. The
equivalent circuit is shown in Figure 8a. It will now be shown that this rather
involved circuit can be analyzed exactly by using Miller's theorem and its dual.
Application of the latter leads to the equivalent circuit of Figure 8b, where T =
IL/ls. Before we proceed further, it is advisable to find T. The current through
C u is
IL = --gr,, V = - (I~ - V Y~), (53)
and the voltage 1/9_is
[(Ic - g m V ) / ( s C u ) ] + V = --IL[Rc + RE(I + T - l ) ] . (54)
From (54), we can express V in terms of 1L as

V = IL [1/(sCn)] + RL + RE(1 + T - I )
[g,nl(sC~)]- l ' (55)
and when this is substituted in (53), we get

ILl1--(gm+YJr){[1/(sC~)]+RL+RE(I+T-I)}]
~ ~ i =-Is. (56)

Upon simplification, (56) gives

gin{Re + [1/(sCu)] + g ~ R e - 1}
r = (57)
1 + [Y~r/(sCtz)] + (go, + Yjr)(RL + RE)"
 498 DUTTAROY

Rx=rX+R~ Cg Z~

+ +

VL

~x RsO+r) cu N(I+T -i)

o

VL

(b)

Rx RtRl+T) RZfI+T-I)

+ +

%v vL

i + -o

(e)

Figure 8. (a) High-frequency equivalent circuit of common-emitter transistor amplifier with unb)-
passed emitter resistance. (b) Simplified equivalent circuit obtained by using the dual of Miller's
theorem. (c) Further simplificationobtained by using Miller's theorem.

Hence all the elements in Figure 8b are known. Now we apply Miller's theorem
to take care o f C~. The equivalent circuit becomes that shown in Figure 8c, where
Z~ = 1/(sC~),R'x = R v + R E ( I + T ) , R ' = R e ( l + T - 1 ) , a n d H V2/V.
Now

V2 = - g m V ( R ' e + RL) [Zff/(l - H-I)] (58)

[Z~/(1 - H-l)] + R's
so that
V2 -gmZ~(R~ ff RL)
H --
/ '5 ?
V Z~+(R +RL)(1 +H-I)"
From (59), H can be explicitly obtained as

(l - g,,,Z.)(R' E + RL)
H= (60)
Z~ + R~ + RL
Now
 MILLER'S THEOREM REVISITED 499

and
VL = V 2 R L / ( R E + R L ) . (62)
Hence
VL RL H
V, -- R' . (63)
e + RL [ r ~ + r ~ ( l - H)JR'~ + 1
Combining this with (60), we get an explicit expression for the gain VL/Vs in
terms of the circuit parameters.
By trying to analyze this example by the mesh, node, or two-port parameter
conversion methods [5], it will be obvious that Miller equivalence offers a good
competition.

6. Conclusions

We have, we believe, conclusively demonstrated the power and simplicity of

Miller's theorem and its dual through their application to the exact analysis of
active as well as passive networks. To the author's knowledge, Miller's theorem
has not been applied earlier to passive network analysis. It is hoped that the
examples given here will convince authors of textbooks on circuit theory that
Miller deserves a place in the future editions of their books.

References

[I] E.J. Angelo, Electronic Circuits, McGraw-Hill, New York, 1964.

[2] E.J. Angelo, Electronics: BJT's, FET's and Microcircuits, McGraw-Hill, New York, 1969.
[3] S.C. Durra Roy, On some three terminal lumped and distributed RC null networks, IEEE Trans.
Circuit Theot3', vol. CT-I 1, 98-103, March 1964.
[4] S. C. Dutta Roy, A quick method for analyzing parallel networks, lntemat. J. Elect. Engnrg.
Educ., vol. 13, 70-75, January 1976.
[5] S. C. Dutta Roy, Analysis of a high-frequency transistor stage, Students" J., Inst. Electron.
Telecommun. Engrs., vol. 29, 5-7, January 1988.
[6] E R. Gray and R. G. Meyer, Analysis and Design of Analog Integruted Circuits, 3rd ed., John
Wiley, New York, 1993.
[7] R.C. Jaeger, Microelectronic Circuit Design, McGraw-Hill, New York, 1997.
[8] D. Johns and K. Martin, Analog h~tegrated Circuit Design, John Wiley, New York, 1997.
[9] M.A. Kazimierczuk, A network theorem dual to Miller's theorem, IEEE Trans. Educ., vol. E-31,
265-269, November 1988.
[ 10] E E Kuo, Network Analysis and Synthesis, John Wiley, New York, 1966.
[1 I] J. Millman and A. Grabel, Microelectronics, 2nd ed., McGraw-Hill, New York, 1987.
[ 12] J. Millman and C. C. Halkias, Integrated Electronics: Analog and Digital Circuits and Systems,
McGraw-Hill, New York, 1972.
f13] M. H. Rashid, Microelectranic Circuits: Analysis and Design, PWS Publishing, Boston, MA,
1999.
{14] T. S. Rathore, Generalized Miller's theorem and its applications, IEEE Trans. Educ, vol. E-32,
386-390, August 1989.
[i5] A. S. Sedra and K. C. Smith, Microelectronic Circuits', 4th ed., Oxford University Press, New
York, 1998.