Physics 173 / BGGN 266 Primer on Filters and Modularity
David Kleinfeld, Spring 2008
S TEA DY STA TE A NA L YS IS
Our circuit analysis is much simplified if we make use of the concept of complex impedance, which
is appropriate for steady-state signals in terms of the spectral, or frequency, content.
For a capacitor we know that I = C
"#
dV
. For V of the form V ( t ) = $ d! !V (! )!e"i! t , the relation
dt
"#
between I(ω) and V(ω) is I (! ) = i! C!V (! ) and we identify Z C (! ) =
1
as the impedance – an
i! C
effective resistance that affects the phase as well as the amplitude of the signal in a frequency
dependent manner – of the capacitor.
t
1
Similarly, for an inductor we know that I = $ d! !V and we identify Z L (! ) = i! L as the
L "#
impedance of the inductor.
The approach of complex impedances allows us to solve algebraic equations in the frequency domain
(again, for steady state) rather than differential equations (convolution integrals) in the time
domain.
As an example, consider a RC low-pass filter, i.e.,
Kirchoff’s Law gives
V2 =
1 ! i" RC
1 + (" RC )
2
V2 ! V1
1
+ V2 ( i" C ) = 0 or V2 =
V1 . A little rearrangement leads to
R
1 + i! RC
V1 ! , or in vector (phasor to the “j” crowd) form, V2 =
1
e!i tan
!1
(" RC )
1 + (" RC )
2
V1 ! .
CA S E OF A TW O- P OL E L OW -P AS S FI L TER WI TH OUT BUFFERI NG
Consider a circuit that has a sequence of two RC filters with node voltages V1(ω), V2(ω), and V3(ω),
i.e.,
Kirchoff’s Law gives
V2 ! V1
V ! V3
+ V2 ( i" C ) + 2
=0
R
R
and
V3 ! V2
+ V3 ( i" C ) = 0 .
R
Thus
V2 ( 2 + i! RC ) " V3 = V1
and
V2 = V3 (1 + i! RC ) .
Combining gives:
V3 #$(1 + i! C ) ( 2 + i! RC ) " 1%& = V1
so that
V3 =
1 ! (" RC ) ! i3" RC
2
1
1 ! (" RC ) + i3" RC
2
V1 =
#1 ! (" RC ) % + ( 3" RC )
$
&
2
2
2
V1 =
e
# 3" RC &
!i !tan !1 %
(
%$ 1! (" RC )2 ('
)1 ! (" RC )2 + + ( 3" RC )2
*
,
2
V1
or
V3 =
e
# 3" RC &
!i !tan !1 %
(
%$ 1! (" RC )2 ('
1 + 7 (" RC ) + (" RC )
2
4
e
V1 =
# 3" RC &
!i !tan !1 %
(
%$ 1! (" RC )2 ('
) 6.854 + (" RC )2 + ) 0.146 + (" RC )2 +
*
,*
,
V1
from which we see that the break frequencies have considerably moved from (! RC ) = 1. By the
way, the asymptotic behavior is:
! "#
V3 $$$
"
e%i&
(! RC )
2
V1 = !
1
(" RC )2
V1
2
CA S E OF A TW O- P OL E L OW -P AS S FI L TER WI TH BUFFERI NG
We now consider the rather different situation that occurs when an OpAmp is used to isolate the
two stages of the RC filter i.e.,
The node voltages are V1(ω) and V2(ω) for the first stage and V2’(w) and V3(w) for the second
stage, where from our model for the OpAmp,
V2! (" ) =
A (" )
Low
V2 (" ) #Frequencies
###
$ V2 (" )
1 + A (" )
Kirchoff’s Law gives
V2 ! V1
+ V2 ( i" C ) = 0
R
or
V2 (1 + i! RC ) = V1
and
V3 ! V2
+ V3 ( i" C ) = 0
R
or
V3 (1 + i! RC ) = V2
so that
1
V3 =
V =
(1 + i! RC ) 2
1
(1 + i! RC )2
V1 =
1 ! (" RC ) ! i2" RC
2
1
1 ! (" RC ) + i2" RC
2
V1 =
#1 ! (" RC )2 % + ( 2" RC )2
$
&
2
V1
or
e
V3 =
# 2 " RC &
!i !tan !1 %
(
%$ 1! (" RC )2 ('
)1 ! (" RC )2 + + ( 2" RC )2
*
,
2
V1 =
e
# 2 " RC &
!i !tan !1 %
(
%$ 1! (" RC )2 ('
1 + (" RC )
2
V1
which shows that the break frequencies are both at ωRC = 1.
above:
! "#
V3 $$$
"
e%i&
(! RC )
2
V1 = !
1
(" RC )2
V1 .
3
The asymptotic behavior is, as
An important issue is to realize that the relationship between V3 and V1 is just the product of the
response of two single pole filter stages.
If we back-up a bit, and consider a single stage,
Kirchoff’s Law gives:
V2 ! V1
+ V2 ( i" C ) = 0
R
so that
1 ! i" RC
1
=
V2 =
V1 =
2 V1
(1 + i! RC )
1 + (" RC )
e!i !tan
!1
( 2" RC )
1 + (" RC )
2
V1
The relation between V3 and V2 in the original problem (Fig. 2) is given by the square of this, i.e.,
2
# e!i !tan!1 ( 2" RC )
V3 = %
% 1 + (" RC )2
$
!1
&
e!i ! 2 tan (" RC )
( V1 =
2 V1
(
1 + (" RC )
'
Recall that
e!i ! 2 tan
!1
(" RC )
= cos #$ 2 tan
= cos #$ tan
2
!1
!1
(" RC )%& ! sin 2 #$ tan !1 (" RC )%& ! i2 sin #$ tan !1 (" RC )%& cos #$ tan !1 (" RC )%&
1
=
1 + (! RC )
= e
(" RC )%& ! i sin #$ 2 tan !1 (" RC )%&
2
"
(! RC )2 " i2
2
1 + (! RC )
! RC
1 + (! RC )
2
1 + (! RC )
1 ! (" RC ) ! i2" RC
2
1
2
=
1 + (" RC )
2
# 2 " RC &
!i !tan !1 %
(
%$ 1! (" RC )2 ('
and we recover the previous result of
V3 =
e
# 2 " RC &
!i !tan !1 %
(
%$ 1! (" RC )2 ('
1 + (" RC )
2
V1 .
In general, as we employ N RC stages (poles) that are separated by OpAmp buffers, the amplitude
2
falls off as "1 + (! RC ) $
#
%
&N/2
'(
)!))
' (! RC ) . The phase is a complicated, albeit smooth function
-N
that varies from 0 at ω = 0 to Nπ/2 as ω → ∞. Various designs exists for filters that maintain
the same asymptotic behavior but have different fall-off and phase behavior, particularly near
ωRC ≈ 1.
4
RES ONA NC E A ND TH E TW O-P OLE L C FI LTER
Our next example uses a single LC pair, rather than 2 RC pairs, i.e.,
Kirchoff’s Law gives
V2 ! V1
+ V2 ( i" C ) = 0
i" L
or
V2 =
1
V1
1 ! " 2 LC
! "#
"
This shows the previous frequency-dependent fall-off, i.e., V2 $$$
resonance at ω2LC = 1, which is not desirable.
e%i&
V1 , but also shows
! 2 LC
This can be tempered by the addition of a small
resistance in series with the inductor, and/or a large resistance in parallel with the capacitor.
We will consider a special case of equal resistors and see under what conditions we can dampen
the resonance. The circuit diagram is:
By working in terms of equivalent impedance, the analysis is straight forward.
The inductor/resistor pair has impedance iωL + R = R(1 + iωL/R).
The capacitor/resistor pair as impedance (R/iωC)/[R + (1/ iωC)] = R/(1 + iωRC)
Kirchoff’s Law gives
V2 ! V1
# 1 + i" RC &
+ V2 %
(' = 0
$
R (1 + i" L R )
R
or
1
V2 =
V
(1 + i! RC ) (1 + i! L R ) + 1 1
=
2 ! " 2 LC ! i" ( RC + L R )
1
V
V1 =
2
2 1
2 ! " 2 LC + i" ( RC + L R )
2 ! " 2 LC + " 2 ( RC + L R )
(
or
5
)
e
V2 =
(2 ! "
2
# " ( RC + L R ) &
!i tan !1 %
(
%$ 2 ! " 2 LC ('
LC
)
2
+"
2
( RC + L R )
2
V1 =
e
(
2 1 + " 2 LC
)
2
# " ( RC + L R ) &
!i tan !1 %
(
%$ 2 ! " 2 LC ('
2
#
) RC + L R , &
! " 2 % LC ! +
.- ( !
*
2
$
'
V1
from which we immediately see that the amplitude is finite at resonance when ω2LC = 1.
! RC + L R $
A considerable simplification occurs for the choice LC = #
&% , which factors to RC = L R
"
2
2
or ( RC ) = LC , and drives the third term in the denominator to zero, i.e.,
2
V2 !=!
e
# 2 " RC &
(
!i tan !1 %
%$ 2 ! (" RC )2 ('
2 1 + (" RC ) !
4
V1 .
Note the complete loss of resonance; the factor of 2 in the denominator reflects voltage division
by the resistors.
The above form must be contrast with that for the two-pole RC filter with buffering, for which
V2 =
e
# 2 " RC &
!i !tan !1 %
(
%$ 1! (" RC )2 ('
1 + (" RC )
2
V1
They both have the same asymptotic limit, but the fall-off near ω = (RC)-1 is sharper for the twopole RC filter than with the LCR filter. On the other hand, the LCR circuit does not require an
Op-Amp.
6