Politecnico di Torino

Department of Control and Computer Engineering (DAUIN)

Master Degree in Mechatronic Engineering
2019/2020

Electronic systems for mechatronics

Professor: Chiaberge M.

Laboratory 3: Active filters design

Friday, April 24, 2020

Group n. 72
Fracasso Francesco – 280240
Sierra Pardo Cesar Andres – 274052
Vilchez Hipolito Litzia Carla – 277392
1 - Objective 2
1.1 - Setup & test procedure 2
1.2 - Second order filter 2

2 - Design, Simulation and Measurements 3

2.1 -Chebyshev 0.5dB 6
2.1.1.Simulation - theoretical values 6
2.1.2.Simulation - standard values 9

2.2 - Chebyshev 1dB 11

2.3 - 10° Order 1dB Chebyshev 13

3 - Conclusion 15

1
1 - Objective
The aim of this laboratory experience is to use the operational amplifier (OA TL082) in order to
design and simulate, using the LTSpice software, the behavior of a low-pass filter able to satisfy
the mask provided by Fig. 1.

Figure 1.0: Low-pass filter mask

1.1 - Setup & test procedure

Due to the impossibility to attend the experience in the laboratory, only simulation of all circuit
configurations on LTSpice software has been done.

1.2 - Second order filters

Second order low/high pass filters are characterized by a Laplace function at power of two.
The complex solutions are strictly correlated to the quality factor of the filter Q, which indicates
the resonance peak amplitude at the cutting frequency of the filter.
Instead of a first order filter in which the slope in the magnitude Bode diagram is 20dB/dec, in a
second order filter it’s 40 dB/dec due the two complex coincident poles.

Multiple feedback circuit

It’s the circuit implementation of a second order low pass filter and each cell is composed by the
following circuit:

2
 Figure 1.3: Generic Multiple Feedback circuit

With transfer function described by:

𝑉 /𝑉 = −𝑌 𝑌 /[𝑌 (𝑌 + 𝑌 + 𝑌 + 𝑌 ) + 𝑌 𝑌 ]

2 - Design, Simulation and Measurements

The aim of this experience is to design a second order low-pass filter, considering the limits
imposed by the mask described in figure 1.0. Translated on the following specifications:
Specifications

Upper limit in |H|=10 dB Upper limit in frequency=13 kHz

Lower limit in |H|=8 dB Lower limit in frequency=10 kHz
Average amplification=9dB Lower limit in attenuation is -45dB

The overall attenuation is the value of the difference between the lower limit in attenuation and
the average amplification, that must be achieved within the frequency limit ranges: -54dB in
3kHz.The cutting frequency is 10 kHz: this is the value used for normalizing.

Design steps

1.Type of approximation

Each type of filter (Bessel, Butterworth, Chebyshev 0.5dB, Chebyshev 1dB) has different
attenuation depending on the behavior at the cutting frequency. In particutal Chebyshev filters
have the highest attenuation rate (-120dB/dec); in the band-pass region the Chebyshev filters
have not a flat behavior but oscillations of until 1dB in amplitude. For the purpose of the laboratory,
Chebyshev filters are optimal.

3
 Figure 2.1:Left: Comparison between filter types response on the stop-band region. Right: Ripple of the 1dB Chebyshev
filter on the pass mand region

2. The minimum number of poles

In order to define the minimum number of poles needed, the behavior graph of Chebyshev filter
is considered:

Figure 2.2: Stop-band regions. Left: Chebyshev 0.db. Right: Chebyshev 1dB

Considering that the frequency axis is normalized with respect to the cutting frequency, the
frequency of 3 kHz correspond to the normalized 1.3 in the graph.
It’s possible to notice that the point whose coordinates are the frequency of 1.3 and the attenuation
of -45 dB, is laying more or less on the 9th order chebyshev filter.
So the filter must have at least 9 poles.

3. Determine f and Q values for each cell

Each cell is characterized by a specific cutting frequency and quality factor, depending on the
order of the filer. Those values can be find in the following tables, the frequency values are
normalized with respect to the cutting frequency:

4
 Figure 2.3: Chebyshev 0.5dB filter normalized parameters

Figure 2.4: Chebyshev 1dB filter normalized parameters

4. Define the amplification for each stage

The overall circuit is composed by a cascade of five cells (for a 9th order filter). The total
amplification has to be equal to 9dB that is the average of the limits imposed by the filter mask.
The best solution is to amplify 9dB in the first cell in order to not amplify the noise coming from
the other cells, which have to be implemented as voltage followers.
It’s possible to reduce the amplification value to reduce the peak of the frequency response.

5. Denormalize and calculate real components

Each cell is implemented as shown in Figure 2.5, except for the first one that correspond to a first
order low.-pass filter:

Figure 2.5: Multiple feedback low-pass filter

𝑉 /𝑉 = −(𝑅 /𝑅 ) / [𝑠 𝐶 𝐶 𝑅 𝑅 + 𝑠𝐶 (𝑅 𝑅 + 𝑅 𝑅 + 𝑅 𝑅 )/𝑅 + 1]

𝑓 = 1/(2𝜋 𝐶 𝐶 𝑅 𝑅 ) 𝑄 = 𝑅 /[(𝑅 + 𝑅 )(𝑅 ||𝑅 + 𝑅 )] ∗ 𝑅 𝑅 𝐻 = 𝑅 /𝑅

5
The cutting frequency of each cell is denormalized considering the table values at step 3 and the
overall cutting frequency of the second order low-pass filter (10 kHz).

In order to simplify the calculations 𝑅 = 𝑅 = 𝑅 was considered, obtaining:

𝑄 = ∗ 𝐶 /𝐶 and 𝑓 = 1/(2𝜋𝑅 𝐶 𝐶 ).

For each stage were found the theoretical values of real components (obtained through the
equations), then, the standard values of real components were chosen according to the E6 series
of capacitors and E12 series of resistances and considering the tolerance of those components
(20% for capacitors and 5% for resistances), and were performed simulations to verify the correct
behavior of the stages.

2.1 - Chebyshev 0.5 dB

2.1.1. Theoretical values calculation, simulation and tuning

CALCULATION

STAGE 1
The first stage is a first order low-pass filter, so the table of step 3 provides only the cutting
frequency:

Figure 2.6: First-order low-pass filter cell

𝑉𝑂𝑢𝑡
= − 𝑅𝑅2 ∗ 1+𝑠𝑅
1 1
𝑉𝑖𝑛
⇒ 𝑓 = ; 𝑓 = 10 𝑘𝐻𝑧, 𝑓 , = 0.1984 ⇒ 𝑓 = 𝑓 , ∗ 𝑓 = 1984 𝐻𝑧
1 2𝐶 2𝜋𝑅2 𝐶

𝐻 = ; 𝐻 = 9𝑑𝐵 → 10 = 2.818
Considering an amplification of 2.818 (9 dB) the overall ripple does not respect the limits
imposed by the filter mask.
So we considered an amplification of 2.7 (8.63 dB).

6
 𝐻 Theoretical 𝑅 Theoretical 𝑅 Theoretical 𝐶

2.7 5.94k𝛀 2.2k𝛀 13.505nF

STAGES 2-3-4-5

The other stages are characterized by the following circuit:

Figure 2.7: LTSpice multiple feedback cell implementation

𝑓 =𝑓, ∗ 𝑓 , 𝑓 = 10𝑘𝐻𝑧 ; 𝐻 = = 1; 𝑅 = 𝑅 = 𝑅 = 𝑅

Stage n° 𝑓 𝑄 Theoretical Theoretical Theoretical 𝑅

=𝑓, ∗ 𝑓 [𝐻𝑧] 𝐶 [nF] 𝐶 [nF] [𝑘𝛺]

2 3954 1.0604 22 2.174 5.820

3 6727 2.2131 47 1.066 3.342

4 8885 4.4780 22 0.122 10.938

5 10046 14.5794 220 0.115 3.150

SIMULATION

Figure 2.8: General low-pass filter circuit schematic, with 5 stages

Using an amplification equal to 2.7 the results are improved but still unsatisfactory.
On Figure 2.9 there are shown the final response composed by the sum of the the single
contributions of each cell, also shown in the same figure:

7
 Figure 2.9: LTSpice Simulations. Black: Final response. Other colors: Single contributions to the final response.

It’s possible, however, to notice that decreasing the initial amplification 𝐻 from 2.818 to 2.7:
- The peak value decreases from 14.432 dB to 14.061 dB.
- The minimum value of the pass-band region decreases from 8.550 dB to 8.178 dB, due
the oscillations in the flat part of the curve typical of a Chebyshev filter: we are still in the
filter mask.
- Also the frequency to -45 dB is better but not inside the mask: from 13.816 kHz to 13.773
kHz.

TUNING

Tuning parameters is a needed operation in order to fit the requirements of the mask:
- changing only R3 let us to horizontal move the peak without changing its value;
- changing only R1 let us to vertical move the curve of a quantity almost equal to the new
gain of the stage;
- changing R2 we have a combination of the two previous behaviors, so we can change
both the horizontal and vertical position of the peak.

Stage n° 𝑅 [k𝛀] 𝑅 [k𝛀] 𝐶[nF]

1 2.2 5 15

- 𝐶 [nF] 𝐶 [nF] 𝑅 [k𝛀] 𝑅 [k𝛀] 𝑅 [k𝛀]

2 22 2.17 4.5 5.8 5.8

3 47 1 3.3 3.3 3.6

4 22 121 12 12 9.6

5 220 115 2.7 2.7 3.2

8
 Figure 2.10 : Frequency response with tuned parameters.

2.1.2. Standard values calculation, simulation and tuning

CALCULATION

Rounding the theoretical values of resistors and capacitors to the closest E12 and E6 series
values respectively, the new real to-be-implemented values are:

Stage n° Standard 𝑅 [k𝛀] Standard 𝑅 [k𝛀] Standard 𝐶[nF]

1 (𝐻 = 2.7) 6.8 2.2 15

- Standard 𝐶 [nF] Standard 𝐶 [nF] Standard 𝑅 [k𝛀]

2 22 2.2 5.6

3 47 1 3.3

4 22 0.1 10

5 220 0.1 3.3

SIMULATION

Figure 2.11: Frequency response of 0.5 Chebyshev filter, with standard values and amplification 𝐻 =2.7

9
Replacing theoretical values with standard values, it’s possible to notice that an amplification of
2.7 causes a higher oscillation of the band-pass region than an amplification of 2.818.

TUNING

The theoretical parameters have been replaced with the closest standard parameters (E12 for
the resistors and E6 for the capacitors). In the following image is shown the frequency response
of these new parameters.

Figure 2.12: Frequency response with standard parameters.

The following plot shows that these parameters do not fit the constraints imposed by the
design,in a very small interval of the frequencies:

Figure 2.13

In order to achieve the requirements of the design it is necessary to tune the standard
parameters. Several resistors have been added in series to the previous ones to achieve a
smoother tuning: as example the resistor 𝑅 =3.48k𝛀 of the fifth stage is obtained as a series of
R=3.3k𝛀 and R=180𝛀.

10
 Stage n° 𝑅 [k𝛀] 𝑅 [k𝛀] 𝐶[nF]

1 2.2 5.6 15

- 𝐶 [nF] 𝐶 [nF] 𝑅 [k𝛀] 𝑅 [k𝛀] 𝑅 [k𝛀]

2 22 2.2 4.7 5.6 5.6

3 47 1 3.3 3.3 3.4

4 22 100pF 12 12 11

5 220 100pF 2.7 2.7 3.48

As a result of the tuning process, we can satisfy the requirement on the gain in the pass-band
region; the attenuation of -45dB is not reached at frequency of 13kHz but at 14.5 kHz.

Figure 2.14: Frequency response with tuned standard parameters.

2.2 - Chebyshev 1 dB

In order to increase the attenuation of the frequency response we tried to design the filter using
the Chebyshev 1 dB.

CALCULATION

STAGE 1

𝑓, = 0.1593 ⇒ 𝑓 = 𝑓 , ∗ 𝑓 = 1593 𝐻𝑧

11
 𝐻 Theoretical 𝑅 [k𝛀] Theoretical 𝑅 [k𝛀] Theoretical 𝐶[nF]

2.7 5.94 2.2 16.820

STAGES 2-3-4-5

Stage n° 𝑓 =𝑓, ∗𝑓 𝑄 Theoretical Theoretical Theoretical

𝐶 [nF] 𝐶 [nF] 𝑅[k𝛀]

2 3773 1.2600 14.284e-9 1 11.16

3 6622 5.5266 27.488e-9 1.00e-10 14.49

4 8806 5.5266 27.488e-9 100pF 10.9

5 9976 18.0287 292.53e-9 1.00e-10 2.94

SIMULATION

Figure 2.15: LTSpice Simulations. Black: Final response. Other colors: Single contributions to the final response.

It’s possible to notice that this solution is not ideal:

-the peak value is 16.540 dB, higher than the peak value of the theoretical frequency response of
Chebyshev 0.5 dB.
-with a gain of 2.7 in the first stage, the minimum value of the band-pass region is 7.736 dB due
the oscillations in the flat part of the curve typical of a Chebyshev filter.
So it’s not possible to decrease the peak decreasing the gain, since the oscillation will not be in
the limits imposed by the mask (8 dB- 10 dB).
-At -45 dB the frequency is 13.298 kHz, better than the Chebyshev 0.5.

It’s possible to conclude that the overall solution is always worse than the previous with the
Chebyshev 0.5dB, since the most important parameters are the gain in the flat region and the
peak value: a low-pass filter should not cut any signal in the band pass region.

12
2.3 - 10th order filter
A 10th order 1 dB Chebyshev filter has a second-order cell for all its 5 stages, so the attenuation
band decreases faster with respect to the 9th order one.

To try to satisfy the asked mask intervals, the design of a 10th order filter was carried out, but in
this case, with the help of the Filter Design Tools held by Texas Instrument. Since this design
was done for simulating the theoretical behavior that this 10th order filter could have, the
simulation was done using ideal elements, including the operational amplifier, so the chosen
values don’t consider important factors like the amplifier input current or its related capacitance.
The corresponding values are collected on the following table:

Stage Gain 𝑓 𝑄
n° [dB] =𝑓, 𝐶 [nF] 𝐶 [nF] 𝑅 [𝑘𝛺] 𝑅 [𝑘𝛺] 𝑅 [𝑘𝛺]
∗𝑓

1 1.875 2422 0.6734 31.6

4.22 1 33 26.1

2 1.875 4978 1.5347 21.5 1 7.15 10 5.36

3 1.889 7468 2.8913 75 1 2.8 3.48 1.8

4 1.889 9295 5.6114 287 1 1.07 1.35 0.768

5 1.889 10209 17.9872 3010 1 0.280 0.33 0.232

Each stage has the structure of a multiple feedback implementation. Even though the filter was
designed by software, further tuning of some resistor values was done to improve the results,
that are shown in figure 2.16.

Figure 2.16

13
As it can be appreciated, and despite the final answer of the filter being better than the 9th order
because of the faster stopband, the requirements are not fully met by the filter until the 13.5kHz.
At this point, we need to consider if this improvement, yet not a full success, is worth the price
that increasing the order of the filter means: considering a 10th order filter, the attenuation rate
increases, the cost of the circuit increase and so do the non idealities in the implementation of the
filter, that are proportional to the number of stages.

14
3 - Conclusion
Several circuits were designed and implemented in this laboratory practice in order to fit the mask
of the low-pass filter; to achieve this purpose different approaches were taken, like implementing
circuits of the 9th or 10th order. The designs were able to satisfy some, but not all the
requirements because of the trade-off between two of the parameters: while staying inside the
borders of the pass-band region requires low ripple, achieving the -54dB decrease supposed for
the 10 to 13 kHz interval is more likely to increase such ripple.

The starting design of a 9th order 0.5dB Chebyshev filter was not enough to meet neither the
pass-band limits nor the stop-band limits, so further tuning was done. The results revealed that,
at least:
- the pass-band gain was obtainable with standard capacitor and resistor values,
- the frequency to achieve the -45dB of attenuation was actually not that far from the desired
one,
- the attenuation at 13kHz was of about the 98% of the signal.

In cases like this, some decisions have to be made, like increasing the cost and implementing a
better approximation with a 10th order filter, or choosing to achieve the pass-band gain over the
frequency limits or vice versa. A thing is certain, the capability of limiting a signal in frequency of
such circuits is way far higher concerning the conventional, simple passive first order low-pass
filter.

15