The Biquad Circuit

A biquad circuit (biquadratic circuit) is an analog filter circuit that

realizes a second order transfer function, meaning its output is
determined by a

A biquad circuit can be configured to implement various filter types,

including:

Low-pass filter
High-pass filter
Band-pass filter
Band-stop filter
All-pass filter
QU The biquad circuit is one of the most useful circuits to the
electrical engineer because it is a universal filter. It is widely
available as a module from industrial sources. It serves as our
introduction to second-order filters.

Q In our previous study, we considered transfer functions with

poles and zeros on the real axis of the s plane.

U Beginning with this chapter, poles will be complex conjugates

in the left half of the s plane, and zeros will also be complex
conjugates, but in either the left or the right half of the s
plane.
 T(s) = Vi)
Vis)
__Z ___ 1GCs
Z,+2, Ls+R+1/Cs
5.1)
’
Dividing numerator and denominator by L and multiplying by s, we obtain
1/LC
1) = ST R/L)s + 1/LC G2
This result may be put into a standard form by defining two new quantities. First

we observe that when the circuit is lossless with R = 0, then the denominator re-
(b) duces to the simple form from which the pole positions may be determined:

1
Fgu/;/’ s"+L—C—0 5.3)

or

Sy, 1 S 92 =+j\/L=+jw
- LC - 0} 5.4)
B

This means that the poles are on the imaginary axis and are conjugates. The
 other parameter which we require originated in studies of lossy coils for which a
quality factor Q was defined by Johnson* as

Wl 1L
=
Q= R _R‘/C (5.5)

which is the ratio of reactance at the frequency w, to resistance. The historical

identification of Q with a lossy coil is no longer appropriate, of course, since we
will identify many kinds of circuits with the parameter Q. The last equation may
be solved for the ratio R/L in Eq. (5.2):

R_w
I-0 (5.6)
Substituting this equation and Eq. (5.4) for 1/LC into Eq. (5.2), we obtain
2

T(s) = £+ (wo/Q)s
=] + w2= D(s)
X0 .7)
This is the desired standard form. Before studying T(jw), we turn our attention to
the s-plane location for the poles of T(s).
 w,’ _ N(@s)
T(s) =
£+ (wo/Q)s + wi? D(s)

The poles of 7(s) are the values for which D(s) = 0 in Eq. (5.7). Let their s-

plane location be —a # j8 so that

D(s) = (s + a+jB)(s + a = jB)
|
=9+ 2as+ (a + ) (5.8) |
Y =cos™ (1/20) || B = wyVT— (17409
Equating like terms of this equation and D(s) in Eq. (5.7), we find that

a= 2”72 (5.9)
I
|
or
I ‘//

=&
0=52 (5.10) _)l .o wO/ZQ . o

Similarly, equating the constant terms in D(s) in Egs. (5.7) and (5.8) FIGURE 5.2
wi=at+ p 6.11)
Combining this with Eq. (5.9) and solving for B, gives

1
B=w 1—4~Q’ (5.12)

All of these relationships are shown in Fig. 5.2. In this figure we also define the
angle y with respect to the negative real axis as

w-cos"(fn)=cos"
2 (5.13)
 Lines of constant & = w/2Q
Figure 5.3 shows important contours in the s plane. Contours of constant w,
are circles of radius w, with their centers at the origin, as shown in Fig. 5.3a.
From Eq. (5.13) we see that lines of constant Q are lines of constant angle v, as
shown in Fig. 5.3b. Finally, lines of constant ratio w,/2Q are lines parallel to the | Contours of 0,<0,
imaginary axis, as shown in Fig. 5.3c. constant w
In circuit design we will ordinarily deal with Q values greater than 1. This Woy< wyy < Wy
has implications with respect to pole positions. From Eq. (5.13) we make the fol-
lowing tabulation:

12 ()
Q (degrees)
FIGURE 5.3

0.707 45
1 60 Sectors of usual
2 75.52 pole locations
5 843
20 88.5
100 89.7

Hence we conclude that we will be interested in a small sector of the s plane,

which is shaded in Fig. 5.4. Observe that when Q is greater than 5, then Eq. (5.12) FIGURE 5.4

for B simplifies to B = w, with an error less than 1%.

 we’ _ N
T(s)= 5.7)
£+ (W/Q)s + w2 D(s)

The transfer function for the low-pass filter derived as Eq. (5.7) was written in a
normalized form such that 7(;0) = 1. A more general form for 7{(s) will recognize
the possibility of gain and also that the associated circuit may be inverting or
noninverting. Such a transfer function is
+Hw?
T(s) = 52+ (wo/Q)s + w2 /In normalized form:

+H +H
T(s) = — ~
T2+ (1/Q)s, +1
 T(s) = “H _ W (5.18)
S+1/Qs+1 V,
We wish to manipulate this equation until it has a form that can be identified
with simple circuits which have been studied in past chapters. We rewrite Eq.
(5.18) as

(sz+ls+ 1) V,=-HV, (5.19)

Q
If we divide this equation by the factor s(s + 1/0Q), it becomes

1 -H (5.20)
[' Tt l/Q)J =G0
We may now manipulate this equation to the form

-1 -H I (5.21)
V2=[s+ AT l/QV'} ‘(‘?)'(_1)
 -1 —-H 1
i i Re e
Starting at the right-hand side of this equation, we recognize that the (—1) term
may be realized by an inverting circuit of gain 1. Similarly, the factor (—1/s) is
realized by an inverting integrator. Two operations are indicated by the remain-
ing factor. The circuit realization must produce a sum of voltages, and it must
have a transfer function of the form 1/(s + 1/Q). The three circuits that provide
for these three operations are shown in Fig. 5.7. The circuit marked T, sums volt-
ages V, and V, with appropriate multiplication, and also realizes the first-order
transfer function with a circuit that is sometimes called a lossy integrator. The
circuit marked 7T, is the standard inverting integrator circuit, and the circuit
marked T, is an inverting circuit of unity gain. If we connect the three circuits to-
gether, including a feedback connection of the output ¥, to the input, the result is
the circuit shown in Fig. 5.8. This is a scaled version of the circuit called the bi-
quad circuit or the ring of 3 circuit, or sometimes the Tow—Thomas biquad.
1o

1/H

=T 2

FIGURE 5.7
FIGURE 5.8
We have found that the biquad circuit shown in Fig. 5.9 is described by two
transfer functions, depending on our selection of the output. To begin, let the out-
put be ¥, so that the transfer function that applies is that given by Eq. (5.18). We
are interested in the magnitude and phase of this T(jw). For simplicity, let H = 1

and also frequency scale so that w, = 1. Then

1
TUo) =Ty
o+ 7070
0l0 (5.32)

From this complex quantity we find that the magnitude is

1
TG
T(jw) = s
= ——/—0———u (5.33)
5.33

and the phase is

9 Ep—
tan 74
T (5.39)
 -H A
T(s) = (5.18)
S+ (1/Qs+1 V,

MV
Cl
—
CZ
Il
R R ]

FIGURE 5.9
The magnitude and phase functions are plotted in Fig. 5.13 for a given value of
Q. For the magnitude function we see from Eq. (5.33) that

TGO =1, |T4HI =@, [T(o)| =0 (535

and that for large w|T(jw)| = 1/’ Similarly for the phase
6j0) =0°, 6(j1)=—90°, (joo) — —180° (5.36)
The magnitude plot on Bode coordinates is shown in Fig. 5.14 for a range of val-
ues of Q from 0.707 to 10. The asymptotic Bode plot decreases at the rate of —12
dB per octave, and this is sometimes described as two-pole rolloff.
These responses can be visualized in terms of the pole locations of the trans-
fer function. Starting with Eq. (5.18),
1
TO= mos+1
v (5.37)

The poles of this function are located on a circle of radius 1 and at an angle with
respect to the negative real axis of

Y =cos™' (ZLQ (5.38)

This equation can be combined with the previous one to give an alternative rep-
resentation:

(5.39)
T(S)=x’+2(:os Ys+1
 +20

oo =VT (11209~ 1 +15

+10

7Vil im
|
! 1T~ 1/w?
+5

} (or —12 dB/octave)

I
{ ~—.
O0 1 w

All Q 0.5 1.0 2.0 5.0 10.0

0
0, degrees

L2417 ]]
400 =1.67

|
| —180° asymptote —80

6. degrees
| for all Q
| — —120)

7
//
—180
0 1 w ~160|
FIGURE 5.13 —200
0.1 0.2 0.5 1.0 5.0 100
w
(b)
FIGURE 5.14
In terms of the poles shown in Fig. 5.15,

1
16 = oo+ ) (5.40)

where g, is the conjugate of p,. With s = jw, the two factors in this equation be-
come

Jw + p, = mle, and Jw + p, = mylo, (5.41)

\ e

vA1 _— Unit circle

,h——\—(

\\ // o
—y %
\//[’1

/ FIGURE 5.15
In terms of these quantities, the magnitude and phase are

ITGw)] = mm,
—— (5.42)
and

0=—(¢ +¢) (5-43)

Phasors representing Eq. (5.41) are shown in Fig. 5.16, as was done in Fig. 4.31.
The figure shows the values computed using these last two equations for three
different values of frequency—one below w,, one at w,, and one above w, From
this construction we see that the short length m, near the frequency w, is the rea-
son why the magnitude function reaches a peak near w,. These plots are useful in

visualizing the behavior of the circuit. In solving problems, Egs. (5.33) and (5.34)
may be evaluated using a hand-held calculator.
 For a lowpass filter the usual specifications will be the half-power frequency
and the value of |T,..,; these quantities are identified in Fig. 5.17. The term half-
power comes from the equation for power P = I’R, from which we see that if P is
to be reduced by one-half, then it is necessary that / be reduced by 1/2. We
are not dealing with current, but with |T(jw)|; however, the name half-power is
now applied to most response curves. So we see that the half-power frequency w,,
corresponds to the value of |T| of 0.707. The relationship between w, and w,, can
be estimated from Fig. 5.17. Since |T',.., is approximately equal to Q, the specifi-
cation of a relatively flat response in the pass band implies a low value of Q.
 hp
FIGURE 5.17
In a second-order transfer function, the denominator controls the
location of the poles and thus the critical frequencies. The
numerator controls the zeros and thus determines if a transfer
function is a low-pass, high-pass, band-pass, or band-reject filter.
TABLE 5.1

Case k, k, ky Name

a 0 0 1 Lowpass
b 1 0 0 Highpass
¢ 0 1 0 Bandpass
d 1 0 1 Bandstop
e 1 | 1 Allpass
TABLE 5.2 Standard forms of second-order responses

Frequency response Poles/zeros Name

; jw
il 1

Tp= :"’ Lowpass

£+ 6" s+ wg?

il LP
o,
Q
T o\ —moA o —
.
1
HP
Tep= Bandpass - ~ ~
P s | N AP
4 1
|
|
IT|

T, £+ w? Bandstop
BE = « »
0
2+ Dt%o noteh

IT|

Tup = £ Highpass
£+ % s+ wg

IT}
- % 5+ wo?
Tap= Allpass
£+ % s+ w’
 Let us use the quadratic formula to determine the poles:

w
52+<—0>s+w =0
Q 0

Wo wo\?
P12 =_Ei (%) - wj

= (~14T-47)
The key is the square root. If the argument under the square root is
Positive, there will be two real roots. If the argument is negative, the roots
will be a complex conjugate. The dividing line is: @, = 0.5
Op <05 Two real, distinct, negative roots.

There will be two distinct real roots. The

step function response would be an
overdamped transient. = (0
0r=05 Two real, repeated, negative roots.

There are two identical (repeated) real

roots. The step function response would = O
be a critically damped transient. P

@,
J=bagiis
20p
 Qr>05 Complex conjugate roots.
P I
The roots will be a complex-conjugate pair. Xl 1
The step response would be an
underdamped transient. .G

X
P,=P3*

Note that as the “quality factor” of the poles, Q, increases, the real part decreases while the
imaginary part approaches +wg
In other words, for high Q’s, the poles look “very imaginary,” thereby bringing the circuit closer
to instability.
 wo/ )s + kw3 w} :
T(s) =
Q T(s s
(5= 2 +(w°/Q)s+w0
sz +
 QEN
| 72| = \/( 3 m5)3+(w., m)z
6, p = — arctan (m")
Op
,m ~
o, —a-
G or
a] — R —

At low frequencies, |Tze| = 1, as expected. At high frequencies, the

magnitude varies inversely with the square of the frequency. The phase
ranges from 0° at low frequencies to —180° at high frequencies.
Note that for @ = we, |Tze| = Or and Azr =—90°. Here is an interesting
observation: If Op > 1, then at w = w,, |Tjp| > 1 !! This requires some
further exploration.
The corner frequency is defined in exactly the same manner as with
first-order filters,

|TLP (mc)
v
? 1

Unfortunately, the math isn’t as simple as in the first-order case. After

some tedious algebra that includes choosing the correct root when
applying the quadratic equation:
A plot of we/ wo vs. Q. (Expression from the previous slide.)

20

15 — e =

g ©

05

00

Op
As Or become large, we = 1.55w..
Note thatwe = wo when Qp = ll\/2- . So in this once instance the corner
frequency is easy to find. We will see that this is an important case.
Bode versions of magnitude plots from previous slide. Note that for
frequencies sufficiently high into the cut-off band, the magnitudes decreases at
rate of =40 dB/dec — twice the slope for first-order. This is dueto IT| = w2
forw >> w,.
10 T

1]

-10

-20
=
=
E %
- QP=0.125 LR
- QP=05 =
- QP=125 )
L -~ first-order
-« 3dB
50 1 .

-60
10 100 1000 10000 100000
Phase angle frequency responses for the various cases from the previous
slides.

45
phase (degrees)

-90

-135

Frequency (Hz)
T(jw) =
(w)?
(w)? + (wO/Q)jw + w?
 z

oo+ (50
|THP| -
= -
; = Oyp = 180"el — arctan fl)fl
(—Q ) (_m?-—
w
m?)
2 212 a, P (]

The 180° comes from the negative sign in the numerator of the T.F.

These expressions are very similar to the corresponding low-pass

functions. The resulting plots are mirror images of the low-pass plots.
The expressions for the cut-off frequency and the properties of the
“bump” in the high-0p plots are similarly symmetric to those from the
low-pass case.
 il [

Tgp U“’ ) = Q:, = 2

—m2 +j (_ )m+m;} (mgmwz)'%j(%)a)
O

Extracting the magnitude and phase.

|TBP| = (Qr ) : Ogp = 90° — arctan (&) %

\/(mj-a)g)g—i-(m” -m)z Or Ben®
2 O

The 90° comes from the imaginary value in the numerator of the T.E

Again, very similar expressions to the low-pass and high-pass cases.

The peak of the bandpass magnitude occurs at the characteristic
frequency and falls off at both higher and lower frequencies. Thus there
are two corner frequencies, defined in the usual manner. The difference
between the two frequencies is the bandwidth of the filter. To calculate
the two corners and the bandwidth, we start in the usual manner:

(Z_) e 1
\/(a)g - m})z + (—Zf . tur)2 ‘/E

After a fair amount of tedious — and sometimes tricky — algebra, we find

the high and low corner frequencies, which we denote as we+ and wc-.

‘"fi-‘"[\/ ‘G [\/ az]

The bandwidth (BW) is the difference between these corners. It turns
out to have a surprisingly simple relationship to w. and Qe:
: : £ il @, The quality factor directly
batdwidth (BW):. Aw =@, =0,
= 0, controls the bandwidth.
» Passive filters suffer from a number of drawbacks; e.g., they constrain the
type of transfer function that can be implemented, and they may require
bulky inductors.

» Most active filters employ op amps to allow simplifying idealizations and

hence a systematic procedure for the design of the circuit.

» There are many ways of constructing active filters. One general-purpose

circuit that is widely used is that of Sallen and Key. We refer to the Sallen
and Key circuit as a VCVS because it uses an op-amp and two resistors
connected so as to constitute a voltage-controlled voltage source (VCVS).
Such a configuration offers good stability, requires a minimum number of
elements, and has low impedance, which is important for cascading filters
with four or more poles.
4.5.1 Sallen-Key Circuits
 KG,G,/C? Hw?
T() = — 1G2/ = “o
s+ 5(G, + G222 — K))/C + GG,/ C? s—z.;{-.vwg/Q+(u(2,

Gain Adjustment in Sallen-Key Filters

The Effect of A(s) on the Sallen-Key Circuit

Changing a Lowpass Filter to a Highpass Filter:

The RC-CR Transformation

Bandstop filter and Allpass filter: Self-study