A Relation between Filter Parameters of the Tow-Thomas-Biquad

In this paper a relation between several filter parameters of the often empolyed Tow-Thomas-
Biquad will be derived, with the aim to make usually necessary compromises between these
filter parameters comprehensible. Filters at higher frequency and/or high quality factors are
of special interest here.

Suppose the biquad is configured as a bandpass according to the following signal flow:

Fig. 1

The filter therefore possesses the transfer function

VBP τ 2s
= H ( s) =
Vin 1 + τ 2s + τ 1τ 2 s 2

s
ω mQ
=
s s2
1+ + 2
ω mQ ω m

With the center frequency ωm and the quality factor Q and the bandwidth B= fm / Q, as well as
the passband gain A0 =1. Now imagine n poles inserted into the feedback path at ωpn ,
representing the frequency deviation of the used operational amplifiers. In a strict sense, the
poles should have been distributed to the feedback and forward path. Poles in the forward
path become poles of the total transfer function. But if the pole frequencies lie magnitudes
above the center frequency ωm, poles not in the feedback loop influence the frequency
deviation of the filter only marginally in practice.

At frequencies far below the bandwidth of the used operational amplifiers one can, as an
approximation of first order, replace all poles by one dominant pole:

© Copyright Quintenz Hybridtechnik. All Rights reserved.

 1 1
K
1 + jΩ / Ω p1 1 + j Ω / Ω pn
1 1
≈ =
1 1 1 + jΩ / Ω p
1 + jΩ ( +K + )
Ω p1 Ω pn

with the normalized frequency Ω = ω/ωm and the normalized pole angular frequencies Ω pl,
…Ω pn , as well as the normalized substitute angular frequency Ω p (all referenced to ωm ).
The total transfer function in jΩ is then:

jΩ / Q
Hˆ ( j Ω) =
1
+ jΩ / Q − Ω 2
1 + jΩ / Ω p

1
For Ω p >> Ω: ≈1 − jΩ / Ω p the following approximation is valid:
1 + jΩ / Ω p

jΩ / Q
Hˆ ( j Ω) =
1 1
1 + jΩ ( − ) − Ω2
Q Ωp
jΩ / Qˆ
= Aˆ 0
1 + j Ω / Qˆ − Ω 2

1 1 1
with = − , and therefore
Qˆ Q Ω p

Qˆ 1
= ‘Q enhancement’
Q ω
1− m
ωp

Qˆ
and Aˆ 0 = A0 ‘Amplitude enhancement’
Q

The transfer function of the filter with additional poles in the feedback is therefore, under the
given assumptions for approximation, again a bandpass function of second order with the
same center frequency, though with a higher Q and a higher passband gain. (A more exact
analysis would show a shift of the center frequency also.)

The Tow-Thomas-Biquad generally consists of three operational amplifiers. If we assume for

all operational amplifiers a frequency response of

© Copyright Quintenz Hybridtechnik. All Rights reserved.

 AAmp
H Amp ( jΩ ) =
jΩ
1+
Ω Amp

we get
ω Amp
ωp ≈
3

Should we need a deviation of the passband gain of one being smaller than 1 dB, and
respectively a tole rance of Q of maximum ca. 10% , we would have to require

1
≤ 1,12
3ω m
1−Q
ω Amp

leading to the following demand on the operational amplifier bandwidth

f Amp ≥ 30 ⋅ Q ⋅ f m

( f m is centerfrequency )

Requiring a maximum deviation of bandpass gain of

Aˆ 0
≤ 1 + δ one gets:
A0

1
f Amp ≥ 3 ⋅ (1 + ) ⋅ Q ⋅ fm
δ

For very small δ though, the decrease in passband gain by the poles in the forward path has to
be considered.
One percieves how the OpAmp bandwidth has to increase with Q to maintain filter precision!

Now we want to consider dynamic also. One can see from fig. 1, that VTP / Vin represents a
lowpass transfer function with ω m , Q . So one has

VTP Q
max = ≈Q
Vin 1
1−
4Q 2

the approximation being valid for higher quality factors Q ≥ 2 .

Because the output voltage of the operational amplifiers sets an upper limit to the dynamic
range and the highest signal voltage (or highest signal current, depending on the

© Copyright Quintenz Hybridtechnik. All Rights reserved.

implemetation) is to be found at VTP , it makes sense to indroduce some quantity D (the name
should remind of dynamic range), by

Vin
D := min
f Vmax

| Vmax | should be the smallest output voltage limit of the amplifiers used. Obviously D will
give no information whatsoever about the lower limit of the dynamic range because it doesn't
consider noise! But D will indicate directly to what extend input and output voltage range of
the operational amplifiers can be utilized. Over the whole frequency range it has to be at least
(rail to rail operation) in fig. 1:

Vin ≤ D ⋅ VB / 2

with Vin being the effective voltage of a sinus signal at the input and VB the minimum of
positive and negative supply voltage.

This leads to a second general condition (approx. for high Q)

Q ⋅ D ≤ const

wich provides a relation between quality factor and dynamic range in the sense of a maximum
allowable (input or output ) level. "const" depends on the chosen implementation, as will be
shown in the following example.

Let us now take a look at a circuit implementation of a Tow-Thomas-bandpass :

VTP

V2

V1

Abb.
2

For the transfer function one gets

© Copyright Quintenz Hybridtechnik. All Rights reserved.

 R2 jωC 2 R3 R4 / R2
V2 = − ⋅ ⋅V
R1 1 + j ωC2 R3 R4 / R2 − ω 2C1C2 R3 R4 in

and for the output voltage of the lowpass contained in the feedback path:

R3 1
VTP = − ⋅ ⋅V
R1 1 + jωC 2 R3 R4 / R2 − ω 2 C1C2 R3 R4 in

The formulas yield a passband gain of A0 = −R2 / R1 and therefore

VTP R3
max ≈ Q.
Vin R1

(Again for higher quality factors, meaning about Q≥2)

1 R
Consequently a useful definition of D would be D = ⋅ 1 .
Q R3

The damped integrator provides DC-gain also. The feedback factor is given by

1
k=
R2
1+
R1 || R3

So for higher gain>>1 the bandwidth of the amplifier in the forward path of the filter would
have to be enhanced by the factor 1/k to keep the phase deviation from worsening. This is at
least true for voltage feedback amplifiers. But even current feedback amplifiers won't help
much. Because of the capacitor present at the inverting input, a resistor would have to be
inserted between the node R1 , R3 , R2 , C1 and the inverting input of the amplifier – which
again would make the bandwidth depend on k respectively R1 and R3 !
R
Further follows with - 2 = A0 as passband gain in this realization:
R1
1 R2
= 1+
k R1 || R3
R R
= 1 + 2 (1 + 1 )
R1 R3
= 1+ | A0 | ⋅(1 + D ⋅ Q)

Inserted into the relation found earlier for the OpAmp bandwidth

f Amp ≥ 3 ⋅ (1 + ) ⋅ [1 + A0 (1 + Q ⋅ D )] ⋅ Q ⋅ f m
1
δ
R1
respectively for Q ⋅ D = >> 1, δ << 1, the estimation
R3
1
f Amp,int ≥ 3 ⋅ ⋅ A0 ⋅ D ⋅ Q 2 ⋅ f m
δ

© Copyright Quintenz Hybridtechnik. All Rights reserved.

may give at least some qualitative relation between the different parameters. Being – after so
many steps of simplification – some way from truth, it is recommended to simulate the circuit
for given amplifiers in any case after having done the first iterations and estimates by hand
calculation !

Conclusion:

It should have become clear by now, what compromises between center frequency, quality
factor, precision, passband gain and bandwidth have to be made. Practical experience indeed
will show what above fo rmulas lead to suspect: For high Q and high center frequencies the
Tow-Thomas-Biquad may be no longer the right choice of circuit to realize a high quality
bandpass. That is true especially for the use as filterblock in more complex filters, e.g. a FLF-
structure. Either one uses another resonator topology or partitions the filter further by using
filter blocks with lower pole Q.
Similar relations can – by the way – be derived for other filter types like Sallen-Key.

© Copyright Quintenz Hybridtechnik. All Rights reserved.