Active Circuits Lectures 1 2 3 4 5 6 DR - Abodina

Active Circuits
Lecture 0: Introduction
ELC 302 – Spring 2015
Dr. Mohamed M. Aboudina
maboudina@gmail.com
Department of Electronics and Communications Engineering
Faculty of Engineering – Cairo University
2
This Course
Information
• Teaching Assistants:
– Eng. Karim Megawer
– Eng. Ahmed Naguib
• Office hours:
– Mondays and Tuesdays: 12-1 pm
– Or by appointment (maboudina@gmail.com)
• Course Website:
– http://www.eece.cu.edu.eg/~maboudina/teaching.html
© Mohamed M. Aboudina, 2015

3
This Course
Course Objective
• Acquire a thorough understanding of the basic principles of

filter design;
– Focus on concepts, architectures and design challenges;
– Preparation for further studies.
• Strategy
– Acquire breadth understanding via a survey of existing architectures
and techniques;
– Acquire depth through a midterm projects that entails design and
thorough simulations of a specific circuit example in modern
technology.

4
This Course
Assessment
• Assignments: 5%
• Midterm: 10%
• Midterm Projects: 15%
– Mini-Project 1
– Mini-Project 2
• Final Exam: 70%

5
This Course
Why Filters?
• Example 1: Communication Receiver

6
This Course
Why Filters?
• Example 2: Hard Disk Drive (HDD)
Why LPF?
• Limit signal and noise bandwidth.
• Provide anti-aliasing prior to sampling.
• Provide significant contribution to overall
equalization.

7
This Course
Course Contents
Lecture # Topic Assessment

1 Introduction
2 Poles/Zeros, Filter Approximation
3 Filter Approximation and Frequency Transformation
4 Introduction to Gm-C filters
5 Circuit Implementation of Transconductors (Gm)
6 Filter Implementation using Gm-C circuits Mini-Porject 1
7 Switched Capacitor (SC) Circuits (Introduction)
8 Filter Implementation using SC circuits
9 Single Opamp 1st and 2nd Order Opamp-RC Filters
10 Two and three Opamp-RC filters
11 Higher-Order Filters (Cascaded and Leapfrog)
12 Active Inductors, Negative Resistors

8
This Course
References
• B. Razavi, “Fundamentals of Microelectronics”. Chapter 14.
• T.L.Deliyannis, Y.Sun, J.K.Fidler, “Continuous-time active filter

design”.

Active Circuits
Lecture 1: Introduction To Active
Circuits/Filters
maboudina@gmail.com
2
History
Filters
• Early 1920s to Late 1960s: Voice filters were realized in RLC

networks.
• In the 1950s: a goal to reduce the size and cost of inductors by

replacing them by active circuits was launched.
• First book on Active RC Circuits “Theory and Design of Active

RC Circuits,” Mc Graw-Hill co., NY 1968.
• Early 1970s, Thin-film hybrid integrated circuits were

developed.
2/9/2015 © Mohamed M. Aboudina, 2015

3
History
Filters
• Integrating filters on chip is one of the main demands in most

systems.
• Active RC filters, Gm-C filters, Switched-Capacitor Filters are all

types of filters that can be integrated on chip.

4
Filters
Filter Applications
Biomedical Hard disk drive

Wireless communication
Hearing aid Space
Video
Sensors Wireless communication
Sigma delta converters
Loop filter in PLL Military
Hz - KHz MHz GHz

5
Filter Realizations
Passive Filters - RLC
• Use Passive components (R, L, C)

• Do not provide gain (at DC)
• Bulky inductors for low frequencies (not suitable for integrated
circuits)
• Filters parameters are coupled (changing one component can
change different filter parameters)
1
• Cannot realize ideal integrator (H s = )
𝑠
• Suitable for high-Q filters (high selectivity)
• Typical 2nd order filter transfer function is given by:
𝐺𝐷𝐶
𝐻 𝑠 =
𝑠/𝜔
1 + 𝑄 0 + (𝑠/𝜔0 )2

6
Filter Realizations
Passive Filters - RC
• Filters constructed with Resistors and Capacitors only result

only in real poles.
• Question: What’s the maximum achievable “Q” from a
passive RC filter?
– To answer this question, we need to know what type of poles does a
passive RC filter create? Real or Complex?
– Selectivity is always poor.
• Bulky dimensions for low frequencies.

7
Filter Realizations
Active RC Filters
• Use of Rs and Cs along with active components (mainly

opamps) to provide high-Q (high selectivity) filters.
• Effectively, inductors are replaced with Active Circuits. (A

circuit that emulates the inductor functionality).

8
Filters
Types of Filters
• Low Pass Filter (LPF): Practical - Approximation
Ideal
Impulse
Response
• Ideal response is called: “Brick Wall”

• Ideal response is not feasible because its impulse response
has non-zero values for t<0  Non-causal system

9
Filters
Types of Filters
• High-Pass Filter (HPF):
• Band-Pass Filter (BPF):
• Band-Stop Filter (BSF):

10
Filters
Types of Filters
• All-Pass Filter (APF):

– All-pass filters do not change the magnitude of the input, rather they
only change the phase.
– Can be used in phase equalization.
• Amplitude Equalizers:
– The amplitude equalizer has an amplitude response that does not
belong to any of the filter responses considered above.
– It is used to compensate for the distortion of the frequency spectrum
that the signal suffers when passing through a system (cable, wire,
optical fiber … etc).
– Its amplitude response is therefore drawn as complementary to the
signal spectrum.

11
Filters
Continuous-Time Filters
• In time domain:
𝑡
– 𝑟 𝑡 = 0
ℎ 𝑡 − 𝜏 𝑒 𝜏 𝑑𝜏
• In frequency domain:
– 𝑅 𝑠 =𝐻 𝑠 𝐸 𝑠
– H(s) : Filter transfer function
𝑛
𝑁 𝑠 𝑎𝑛 𝑠 𝑛 +𝑎𝑛−1 𝑠 𝑛−1 +⋯+𝑎1 𝑠+𝑎0 𝑎𝑛 𝑖=1(𝑠−𝑧𝑖 )
• 𝐻 𝑠 = = = 𝑚
𝐷 𝑠 𝑠 𝑚 +𝑏𝑚−1 𝑠 𝑚−1 +⋯+𝑏1 𝑠+𝑏0 𝑖=1(𝑠−𝑝𝑖 )
– 𝑚≥𝑛
– 𝑎𝑖 𝑎𝑛𝑑 𝑏𝑖 are real for all i.
– All 𝑏𝑖′ 𝑠 are positive (stability)

Continuous-Time Filters 12
Filters
Continuous-Time Filters
• If the signal is sinusoidal with frequency ω , Replace s with 𝑗ω

• 𝐻 𝑠 → 𝐻 𝑗𝜔
• 𝐻 𝑗𝜔 = 𝐻(𝑗𝜔) 𝑒 𝑗𝜙(𝜔)
• In most cases, we represent the magnitude as :
– 𝐴 𝜔 = 20log( 𝐻(𝑗𝜔) ) (dB) [Even Function : 𝐴 𝜔 = 𝐴(−𝜔)]
– 𝜙 𝜔 = arg 𝐻 𝑗𝜔 (degree) [Odd function: 𝜙 𝜔 = −𝜙 −𝜔 ]
• Group delay: Sometimes it is useful to look at the group delay
𝑑𝜙 𝜔
of a filter 𝜏𝐺 𝜔 = − (sec.)
𝑑𝜔
• Units of group delay is time. Group delay represents the time
delay that each frequency component will experience due to
the filter characteristics.

13
Filters
Pole-Zero Locations
𝑛
𝑁 𝑠 𝑎𝑛 𝑖=1(𝑠−𝑧𝑖 )
• 𝐻 𝑠 = = 𝑚 (𝑠−𝑝 )
𝐷 𝑠 𝑖=1 𝑖
• Since all coefficients of N(s) are real, Zeros of H(s) can be real or
complex conjugate.
• Similarly for D(s), Poles of H(s) can be real or complex conjugate.
For stability purposes, all real part of poles must be negative (< 0).

14
Filters
Stability
• Mathematically, for a system to be stable:

∞
– 0
ℎ 𝑡 𝑑𝑡 = 𝑀 < ∞
– Only expressions that are allowed: 𝐴 𝑡 𝑛 𝑒 −𝜎𝑡 sin(𝜔𝑡)
• Where 𝜎 is a positive  Decaying exponential.
1
• Pole must be on the form of 𝑠+𝐴 , left-half-plan pole (A is positive).
• What about the impulse response of a right-half plan pole?

– ℎ 𝑡 ~ 𝑒 𝑡 ∶ Positive exponential  Unstable

15
Poles and Zeros
Insight View
𝑗𝜔
|𝑠 − 𝑠𝑥 | 𝑠 ∠(𝑠 − 𝑠𝑥 )
• 𝑠 − 𝑠𝑥
– 𝑠 represents any point on the s-plane. 𝑠𝑥 = 𝜎𝑥 + 𝑗𝜔𝑥
– 𝑠𝑥 is a specific value, can be real or 𝜎

complex.
– 𝑠 − 𝑠𝑥 : is a vector between them which s-plane
can also be defined as a distance
(magnitude) and a phase.
• In sinusoidal steady state (normal
case), 𝑠 = 𝑗𝜔. It moves only on the
𝑗𝜔 − axis.

16
Poles and Zeros
In case of sinusoidal steady state
𝑗𝜔
• Moving only on the 𝑗𝜔 axis. 𝑠𝑥 = −𝜎𝑥 + 𝑗𝜔𝑥
• |𝑠 − 𝑠𝑥 | |𝑠 − 𝑠𝑥 |
– Maximum distance between 𝑠 and 𝑠𝑥 = ∞, 𝜎
it happens at 𝜔 = ∞.
– Minimum distance between 𝑠 and 𝑠𝑥 = 𝜎𝑥 ,
this value happens at 𝜔 = 𝜔𝑥 .
• ∠ 𝑠 − 𝑠𝑥
−𝜔𝑥
– @ 𝜔 = 0, ∠ = tan−1
𝜎𝑥
– @ 𝜔 = ∞, ∠ = 90𝑜
– ∠ = 0, when 𝜔 = 𝜔𝑥

17
Poles and Zeros
Real Poles
𝑗𝜔
1
• 𝐻 𝑠 = 𝑠 + 𝑠𝑝
𝑠+𝑠𝑝
– 𝑠𝑝 : Is called a Pole.
−𝑠𝑝 𝜎
– Minimum distance exists at 𝜔 = 0 → 𝐻 𝑗𝜔
Maximum 𝐻 𝑗𝜔 happens at 𝜔 = 0. 𝐴𝑚𝑎𝑥
– Distance increases as 𝜔 increases → 𝐴𝑚𝑎𝑥
𝐻 𝑗𝜔 decreases. 2
– Phase at DC = 0
𝜔
– Phase gradually increases till it reaches ∠𝐻 𝑗𝜔
−90𝑜 at 𝜔 = ∞.
– @𝜔 = 𝑠𝑝 → ∠𝐻 𝑗𝜔 = −45𝑜 𝜔
𝑜 −45
−90𝑜

18
Poles and Zeros
Real Zeros
𝑗𝜔
• 𝐻 𝑠 = 𝑠 + 𝑠𝑧 𝑠 + 𝑠𝑧
– 𝑠𝑝 : Is called a Pole.
– Minimum distance exists at 𝜔 = 0 → −𝑠𝑧 𝜎
Minimum 𝐻 𝑗𝜔 happens at 𝜔 = 0. 𝐻 𝑗𝜔
– Distance increases as 𝜔 increases →
𝐻 𝑗𝜔 increases. 2𝐴𝑚𝑖𝑛
𝐴𝑚𝑖𝑛
– Phase at DC = 0
– Phase gradually increases till it reaches 𝜔
90𝑜 at 𝜔 = ∞. ∠𝐻 𝑗𝜔
– @𝜔 = 𝑠𝑧 → ∠𝐻 𝑗𝜔 = +45𝑜 90𝑜
45𝑜

19
Poles and Zeros
Complex Poles
𝜔02 𝑗𝜔
• 𝑇 𝑠 = 𝜔
𝑠 2 + 0 𝑠+𝜔02
𝑄
– Two complex poles: conjugate
𝜔 𝜔0 2 𝑗𝜔𝑝
– 𝑠𝑝1,2 = − 2𝑄0 ± 2𝑄
− 𝜔02
1
– 𝑄 < 2 → Real Poles
1 𝜎𝑝 𝜎
– 𝑄 > 2 → Complex Conjugate poles
𝜔 1 −𝑗𝜔𝑝
– 𝑠𝑝1,2 = − 2𝑄0 ± 𝑗𝜔0 1 −
4𝑄2 1
𝜔0 𝑄=
• 𝜎𝑝 = 2
2𝑄
1
• 𝜔𝑝 = 𝜔0 1 − 4𝑄2

20
Poles and Zeros
Complex Poles
𝜔02
• 𝑇 𝑠 = 𝜔
𝑠 2 + 𝑄0 𝑠+𝜔02 −𝜎𝑝 ± 𝑗𝜔𝑝 𝑗𝜔
1 𝜔0
– 𝑄 = →Double pole at .
2 2𝑄
– The higher the Q, the smaller the 𝜎
real part is and the closer the
imaginary part is to 𝜔0 .
• The closer 𝑗𝜔 is to one of the
poles, the higher 𝑇 𝑗𝜔 will be
1
(𝑄 > ).
2

21
Poles and Zeros
SecondSecond order Low-pass Frequency Response
Order System

22
Poles and Zeros
SecondSecond order Low-pass Frequency Response
Order System
1
For 𝑄 = ,magnitude response is
2
maximally flat (Butterworth Filter:
Maximum bandwidth without
peaking)
1
For 𝑄 > , response shows
2
undesired peaking.
1
For 𝑄 < : Filter’s bandwidth
2
capability is wasted.
At w<<wo, filter has unity gain.

At w>>wo response exhibits two-pole
roll-off at -40dB/decade.
At w=wo, gain of filter =Q.

23
Poles and Zeros
Second Order System

24
Poles and Zeros
Second Order System

25
Poles and Zeros
Examples (Sections)

26
Poles and Zeros
Examples (Sections)

27
Poles and Zeros
Examples (Sections)

28
Poles and Zeros
Examples (Sections)

Active Circuits
Lecture 2: Poles/Zeros and Filter
Approximations
maboudina@gmail.com
2
Outline
• What are the poles of zeros of a system?
• Filter Approximation:
– Butterworth
– Chebyshev
– Elliptic

3
Filter Design Procedure
• Determine filter specification: System dictates that.
• Approximate the required specifications with a feasible

circuit design: Results in a transfer function H(s). Today’s
Lecture
• Synthesize the filter transfer function: Create a circuit that

implements the transfer function H(s).

4
Approximation
Why?
|H(f)|
• Why can’t we implement
an ideal filter?
• Ideal filter:
– 𝐻 𝑓 = 1 𝑓 < 𝑓𝑐
= 0 𝑓 > 𝑓𝑐
– ℎ 𝑡 = 2𝑓𝑐 𝑠𝑖𝑛𝑐 2𝜋𝑓𝑐 𝑡
• h(t) is not a causal

system  Cannot be
implemented.

5
LTI Filters
Specifications
• Causal
𝑁 𝑠 polynomial
• Rational functions = =
𝐷 𝑠 polynomial
• Stable
– Strictly Stable : All poles in the LHP (excluding the 𝑗𝜔 axis)

6
Approximation
Formulation of the Problem
𝐴𝑝
𝐴𝑠
• 𝜔𝑐 = 𝜔𝑝 : Cut-off frequency = Pass-band frequency

• 𝐴𝑝 : maximum allowable deviation before 𝜔𝑐
• 𝜔𝑠 : Stop-band frequency
• 𝐴𝑠 : minimum attenuation after 𝜔𝑠

7
Approximation
Formulation of the Problem
• Let’s start with studying Low Pass Filters.

• Let’s study the case of 𝜔𝑐 = 1.
• We focus more on the magnitude response:
1
– 𝑀(𝜔) = |𝐻 𝜔 | =
1+𝜀2 𝑤(𝜔2 )
– 𝜀 is constant: 0 < 𝜀 < 1
– 𝑤(𝜔2 ): 0 < 𝑤(𝜔2 ) < 1 for 0 < 𝜔 < 1
– 𝑤 𝜔2 increases rapidly for 𝜔 > 1

8
Approximation
Butterworth (Maximally Flat) Approximation
• 𝑤 𝜔2 = 𝜔2𝑛 (n is positive)
1
• 𝑀 𝜔 =
1+𝜀 2 𝜔2𝑛
• n : Order of the filter.
• Important points:
20 log
– 𝑀 0 =1 0𝑑𝐵
1 20 log
– 𝑀 𝜔𝑐 = 1 = −10 log 1 + 𝜀 2 𝑑𝐵
1+𝜀2

9
Approximation
Butterworth (Maximally Flat) Approximation
• Error in the pass-band increases monotonically from 𝜔 = 0 𝑡𝑜 𝜔 = 1

1 20𝑙𝑜𝑔
• For 𝜔 ≫ 1 𝑀 𝜔 ≈ −20𝑛 𝑙𝑜𝑔 𝜔 − 20𝑙𝑜𝑔𝜀
𝜀𝜔𝑛
• Falling Slope:
– −20 × 𝑛 (dB /decade) Max. Allowed Attenuation
– −6 × 𝑛 (dB /octave) 𝑀(𝜔) (dB) in the passband
– Faster roll-off means higher order. 𝐴𝑝 Min. Allowed
Attenuation in
the stopband
• 𝐴𝑝 = 10 log 1 + 𝜀 2 𝜔𝑝2𝑛 → Find ε n1
• 𝐴𝑠 = 10 log(1 + 𝜀 2 𝜔𝑠 2𝑛 ) n2
𝐴𝑠
n2>n1
1 100.1𝐴𝑠 −1 1
• 𝑛≥ log × 𝜔 passband stopband
2 100.1𝐴𝑝 −1 log𝜔 𝑠
𝑝
𝜔𝑝 =1 𝜔𝑠
10
Approximation
Butterworth – How to find 𝐻(𝑠)
• We already have 𝐻 𝜔
1
• 𝑀 𝜔 = 𝐻 𝜔 =
1+𝜀 2 𝜔2𝑛
• Remember from Signal Processing:
– 𝐻 𝑠 = 𝐻(𝑗𝜔)
1
– 𝑀2 𝜔 = 𝐻 𝑗𝜔 𝐻 −𝑗𝜔 =
1+𝜀2 𝜔2𝑛
• Let’s define: 𝑃 𝑠 2 = 𝐻 𝑠 𝐻 −𝑠 ⇒ 𝑀2 𝜔 = 𝑃 −𝜔2
– Simply: Replace 𝜔2 𝑏𝑦 − 𝑠 2
• Poles of 𝑃 𝑠 2 are the roots of : 1 + −1 𝑛 𝜀 2 𝑠 2𝑛 = 0
1 1/𝑛 2𝑘−1 2𝑘−1
– Roots are : 𝑠𝑘 = 𝜎𝑘 + 𝑗𝜔𝑘 = [sin 𝜋 + 𝑗 cos 𝜋 ],
𝜀 2𝑛 2𝑛
where k=1 to 2𝑛

11
Approximation
Butterworth – How to find 𝐻(𝑠)
• Example: n = 4 and 𝜀 = 1
– 𝑠1 = – 0.3827 + 𝑗0.9239 𝒔𝟏
– 𝑠2 = – 0.9239 + 𝑗0.3827
– 𝑠3 = – 0.9239 − 𝑗0.3827
𝒔𝟐
– 𝑠4 = – 0.3827 − 𝑗0.9239
• Choose only the poles in the LHP.
1
• 𝐻 𝑠 = =
(𝑠−𝑠1 )(𝑠−𝑠2 )(𝑠−𝑠3 )(𝑠−𝑠4 )
1 𝒔𝟑
=
(𝑠 2 +0.7654𝑠+1)(𝑠 2 +1.8478𝑠+1)
1
𝑠 4 +2.613𝑠 3 +3.414𝑠 2 +2.613𝑠+1 𝒔𝟒
• Polynomial in the denominator is
called: Butterworth polynomial.

12
Approximation
Butterworth - Example
• Design a Butterworth filter such that:

– Ω𝑝 = 1 and maximum attenuation in the passband is 1 dB (𝐴𝑝 ≤ 1𝑑𝐵)
– Ω𝑠 = 2 and minimum attenuation in the stopband is 20 dB (𝐴𝑠 ≥
20𝑑𝐵)
• What does it mean to design a Butterworth (MFM) filter?

– Choose 𝜀 and n .
• First Find 𝜀: 𝐴𝑝 ≤ 1𝑑𝐵 → 10 log 1 + 𝜀 2 ≤ 1 → 𝜀 = 0.509

• Then find n:
1 100.1𝐴𝑠 −1 1
– 𝑛≥ log × = 4.3 → Pick n=5
2 100.1𝐴𝑝 −1 Ω
logΩ 𝑠
𝑝

Active Circuits
Lecture 3: Filter Approximations 2
maboudina@gmail.com
2
Outline
• Approximation:
– Butterworth
– Chebyshev
– Elliptic

3
Approximation
Chebyshev (Equiripple) Approximation
1
• 𝑀2 𝜔 =
1+𝜀 2 𝐶𝑛2 (𝜔)
• 𝐶𝑛2 𝜔 is a Chebyshev polynomial of degree n.
cos 𝑛 cos −1 𝜔 for 0 ≤ 𝜔 ≤ 1

– 𝐶𝑛 𝜔 =
cosh 𝑛 cosh−1 𝜔 for 𝜔 ≥ 1
– In the passband, 0 ≤ 𝜔 ≤ 1 ⇒ −1 ≤ 𝐶𝑛 (𝜔) ≤ 1
1
• Meaning that: <𝑀 𝜔 <1
1+𝜀2
1
• This leads to : 𝐴𝑝 = 0 − 20 log = 10log(1 + 𝜀 2 ) (dB)
1+𝜀2
– 𝜔 ≫ 1 ⇒ 𝐶𝑛 𝜔 ≫ 1
• From Special Functions Books:
– 𝐶𝑛+1 𝜔 = 2𝜔𝐶𝑛 𝜔 − 𝐶𝑛−1 𝜔
• 𝐶0 𝜔 = 1 and 𝐶1 𝜔 = 𝜔

4
Approximation
1, n odd
• 𝑀 0 = 1
2
, n even
1+𝜀
1
• 𝑀 𝜔𝑝 = 𝜔𝑐 = 1 = (End of the passband)
1+𝜀 2
1
• @ 𝜔≫1 𝑀 𝜔 ≈ 𝜀 2𝑛−1𝜔𝑛
• 𝐴𝑠 = −20 log 𝑀 𝜔 = 20𝑙𝑜𝑔𝜀 + 6 𝑛 − 1 + 20𝑛𝑙𝑜𝑔(𝜔) dB
• For a fixed 𝜔𝑠  𝐴𝑠 = 20𝑙𝑜𝑔𝜀 + 6 𝑛 − 1 + 20𝑛𝑙𝑜𝑔 𝜔𝑠
• Need to increase 𝜀, 𝑛 to increase 𝐴𝑠 .
• Trade-off between 𝜀, 𝑛, 𝜔𝑠 𝑎𝑛𝑑 𝐴𝑠 .
𝜔 𝜔 𝜔
5
Approximation
• Poles:
– 𝑆𝑘 = 𝜎𝑘 + 𝑗𝜔𝑘
2k−1
– 𝜎𝑘 = sinh 𝛽 sin 𝜋
2𝑛
2k−1
– 𝜔𝑘 = cosh 𝛽 cos 𝜋
2𝑛
1 1
σ
– 𝛽 = sinh−1 and k=1,2 … n
𝑛 𝜀
𝜎𝑘2 2
𝜔𝑘
– Note: + = 1 (Ellipse)
𝑠𝑖𝑛ℎ2 𝛽 cosh2 𝛽
𝜔 𝜔 𝜔
7
Approximation
• Design Procedure:
– Passband: Ap (Amax)  𝜀
– Stopband: As (Amin) and 𝜔𝑠  n
– Find the poles sk
– Write down H(s):
𝐶𝑜𝑛𝑠𝑡𝑎𝑛𝑡
– 𝐻 𝑠 =
𝑠−𝑠1 𝑠−𝑠2 … 𝑠−𝑠𝑛
– Constant is chosen to give the required DC gain.
• Example:
n
n
n

8
Approximation
• Comparison:
– Butterworth:
Monotonic
– Chebyshev: Ripples in
passband and
monotonic in stop
band.

9
Approximation
Inverse Chebyshev Approximation
1
𝜀 2 𝐶𝑛2 (𝜔)
• 𝑀2 𝜔 = 1
1+𝜀 2 𝐶𝑛2 (𝜔)

10
Approximation
Inverse Chebyshev Approximation
𝐴𝑠 = −20log(1 − 𝛽 2 )
1
𝜀 2 𝐶𝑛2 (𝜔)
𝑀2 𝜔 =
1
1 + 𝜀 2 𝐶𝑛2 (𝜔)

11
Outline
• Approximation:
– Butterworth
– Chebyshev
– Elliptic

12
Approximation
Elliptic (Cauer) Approximation
• All previous filter approximations (except inverse Chebyshev)

do not have zeros except at infinity.
• In some cases, higher fall-off is required in the stopband.
• In other words, a higher attenuation is required very near the

passband corner frequency.
• Ripples in both the passband and stopbands are needed to

obtain higher attenuation.

13
Approximation
𝐾(𝑠 2 +𝜔02 )
• Example: 𝐻 𝑠 =
(𝑠+𝛼)(𝑠 2 +𝛽𝑠+𝛾)
• Equiripple in both passband and stopbands.
• Same filter requirements must be satisfied: 𝐴𝑝 in passband,

𝜔𝑠 , 𝐴𝑠 in stopband.

14
Approximation
1
• 𝑀2 𝜔 =
1+𝜀 2 𝑅2 (𝜔)
1
• Note : 𝑅𝑛 𝜔 = 𝑅𝑛 ( )
𝜔
– Notice the relation between poles and zeros of 𝑅𝑛 𝜔 .

16
Approximation
Typical Magnitude Response for a

3rd-order Elliptic Filter.
|𝐻 𝑗𝜔 |
𝐴𝑝 = 𝐴𝑚𝑎𝑥
𝐾(𝑠 2 + 𝜔02 )
𝐻 𝑠 =
(𝑠 + 𝛼)(𝑠 2 + 𝛽𝑠 + 𝛾)
𝐴𝑠 = 𝐴𝑚𝑖𝑛
• Elliptic Filters produce the most attenuation for a given filter order
 Lowest cost possible  Most Common

17
Approximation
Elliptic (Cauer) Approximation - Tables

18
Outline
• Approximation:
– Butterworth
– Chebyshev
– Elliptic
• Frequency Transformation

19
Frequency Transformation
Introduction
|𝐻(jΩ)|
• So far, we can design a normalized low-pass filter
with some given specifications.
• How do we design for different de-normalized

filters and/or different types of filters? Answer:
1 Ω
• Example: Having the 𝐻(𝑠𝑛) of a normalized low-
pass filter, to find the 𝐻(𝑠) of a de-normalized |𝐻(𝑗𝜔)|
low-pass filter with a cut-off frequency 𝜔𝑐 , we
𝑠
simple substitute every 𝑠𝑛 with .
𝜔𝑐
• Note: Frequency transformation does not change

amplitude specifications. It only
𝜔𝑐 𝜔

20
What Happens to L, C and R values in a Filter after Freq. Trans.?
|𝐻(jΩ)|
𝑠
• 𝑠𝑛 →
𝜔𝑐
• 𝑅𝑛𝑒𝑤 = 𝑅𝑜𝑙𝑑
1 Ω
𝐿𝑛
• 𝐿=
𝜔𝑐 |𝐻(𝑗𝜔)|
𝐶𝑛
• 𝐶=
𝜔𝑐
𝜔𝑐 𝜔

21
LP  HP
|𝐻(jΩ)|
1
• 𝑠𝑛 →
𝑠2
𝜔𝑐
𝑠𝑛 →
𝑠 𝑠
• 𝑠2 → |𝐻(jΩ2 )|
1 Ω
𝜔𝑐
1 Ω2
• 𝐿𝑛 : Changes into a capacitor:
|𝐻(𝑗𝜔)|
– 𝐶 = 1/𝜔𝑐 𝐿𝑛
• 𝐶𝑛 : Changes into an inductor:
– 𝐿 = 1/𝜔𝑐 𝐶𝑛
𝜔𝑐 𝜔
22
LP  BP
𝜔0 𝑠 𝜔0
• 𝑠𝑛 → ( + ) |𝐻(jΩ)|
𝐵 𝜔0 𝑠
• 𝐿𝑛 : Changes into L and C in series:

1 Ω
𝐿
– 𝐿 = 𝐵𝑛
𝐵
– 𝐶 = 𝜔2 𝐿 |𝐻(j𝜔)|
0 𝑛
• 𝐶𝑛 : Changes into C and L in parallel:
𝐶
𝐵
– 𝐶 = 𝐵𝑛
𝐵
– 𝐿 = 𝜔2 𝐶 𝜔0 𝜔
0 𝑛

23
LP  BS
𝐵 1
• 𝑠𝑛 → × 𝑠 𝜔0
𝜔0 +
𝜔0 𝑠
|𝐻(jΩ)|
• 𝐿𝑛 : Changes into L and C in parallel: 1 Ω

𝐵𝐿𝑛
– 𝐿=
𝜔02
1
– 𝐶= |𝐻(j𝜔)|
𝐵𝐿𝑛
• 𝐶𝑛 : Changes into C and L in series:
– 𝐶=
𝐵𝐶𝑛 𝐵
𝜔02
1
– 𝐿=
𝐵𝐶𝑛 𝜔0 𝜔

24
Summary

25
Summary

Active Circuits
Lecture 4: Gm-C Filters Introduction
maboudina@gmail.com
2
Circuit Implementation of Filter
• How to implement the following expressions:

1 1 1 2 + 𝑎𝑠 + 𝑏)
𝑠 + 𝑎 (𝑠
𝑠 𝑠 + 𝑎 𝑠 2 + 𝑎𝑠 + 𝑏
• Passive Implementation
• Active Implementations
– Gm-C
– Swiched Capacitor
– Opamp-RC

3
Introduction to Gm-C Filters
• A type of filters that rely on 2

components: Gm and C.
• Gm: Transconductance Amplifier (OTA)

– Input voltage
– Output current
– Gain (A/V): Gm (Voltage Controlled Current
Source)
• C: Capacitor

4
Basic Cells
+
• Amplifiers: Vin Gm1 Vout
-
𝑉𝑜𝑢𝑡 𝐺𝑚1
– =
𝑉𝑖𝑛 𝐺𝑚2
Gm2
• Integrator:
𝑉𝑜𝑢𝑡 𝐺𝑚
– = +
Vin Gm
𝑉𝑖𝑛 𝑠𝐶 Vout
-

5
Gm Introduction
• Transconductor Amplifier (OTA) Design

– Bipolar-based
– MOSFET-based
– Linearity of an OTA

6
What Do we look for in a Gm cell?
• The Value of Gm.
• Linearity of the circuit.

– (Linear) Input Range
– How constant is the value of Gm over the whole input range.
𝑆𝑎𝑚𝑒 𝑆𝑙𝑜𝑝𝑒 𝐺𝑚 𝑜𝑣𝑒𝑟 𝑖𝑜𝑑 𝐼𝑠𝑠 𝑖𝑜𝑑 𝐼𝑠𝑠

𝑡ℎ𝑒 𝑤ℎ𝑜𝑙𝑒 𝑖𝑛𝑝𝑢𝑡 𝑟𝑎𝑛𝑔𝑒 𝑆𝑙𝑜𝑝𝑒 = 𝑔𝑚0
𝑣𝑖𝑑 𝑣𝑖𝑑
−𝐼𝑠𝑠 −𝐼𝑠𝑠 𝐷𝑖𝑓𝑓𝑒𝑟𝑒𝑛𝑡 𝑆𝑙𝑜𝑝𝑒 𝐺𝑚 𝑜𝑣𝑒𝑟
𝑡ℎ𝑒 𝑤ℎ𝑜𝑙𝑒 𝑖𝑛𝑝𝑢𝑡 𝑟𝑎𝑛𝑔𝑒

7
Bipolar-Transistor-Based Transconductance
Note: 𝑔𝑚 = 𝐼𝐶 /𝑉𝑇 (For a BJT
transistor)

8
Bipolar-Transistor-Based Transconductance with Degeneration
Assignment: Prove
This equation.

9
Transconductor Amplifier
MOSFET-based OTA
• 𝑔𝑚0 is the 𝑔𝑚 of the MOS

transistor in the balanced state
(Vid = 0)
𝑖𝑜𝑑 𝐼𝑠𝑠
𝑆𝑙𝑜𝑝𝑒 = 𝑔𝑚0
𝑣𝑖𝑑
−𝐼𝑠𝑠

10
MOSFET-Based: Small Signal at Zero Bias
• 𝑖𝑜𝑑 = 𝑖𝑜𝑢𝑡 = 𝑖𝑑1 − 𝑖𝑑2
• 𝑔𝑚 𝑣𝑥 (𝑖𝑑1 ) = 𝑔𝑚 𝑣𝑦 (−𝑖𝑑2 )
𝑔𝑚 𝑣𝑖𝑑 𝑔𝑚 𝑣𝑖𝑑
• 𝑖𝑜𝑢𝑡 = + = 𝑔𝑚 𝑣𝑖𝑑
2 2
1
2𝐼𝐷 2 2𝐼𝑆𝑆 𝐼𝑆𝑆
• 𝑔𝑚0 = = =
𝑉𝑒𝑓𝑓 𝑉𝑒𝑓𝑓 𝑉𝑒𝑓𝑓
𝑊 𝑊
• Also, 𝑔𝑚0 = 2𝜇𝑛 𝐶𝑜𝑥 𝐼 = 𝜇𝑛 𝐶𝑜𝑥 𝐼
𝐿 𝐷 𝐿 𝑆𝑆
𝑖𝑜𝑑 𝐼𝑠𝑠
𝑆𝑙𝑜𝑝𝑒 = 𝑔𝑚0
𝑣𝑖𝑑
−𝐼𝑠𝑠

11
MOSFET-Based : Large Signal
• To increase linearity: Increase 𝑉𝑒𝑓𝑓0

1 𝑊 2
• 𝑀1 : 𝐼𝐷1 = 𝜇𝑛 𝐶𝑜𝑥 𝑉 1
𝐼
2 𝐿 𝑒𝑓𝑓1 • 𝑉𝑒𝑓𝑓0 = 1
2 𝑆𝑆
𝑊
𝜇𝑛 𝐶𝑜𝑥
1 𝑊 2 2 𝐿
• 𝑀2 : 𝐼𝐷2 = 𝜇 𝐶 𝑉
2 𝑛 𝑜𝑥
𝑊
𝐿 𝑒𝑓𝑓2 • Either Increase 𝐼𝑆𝑆 or reduce 𝐿
• 𝐼𝐷1 + 𝐼𝐷2 = 𝐼𝑆𝑆 • MOS diff pair is self degenerated.
1 𝑊 2𝐼𝑆𝑆
• 𝐼𝐷1 − 𝐼𝐷2 = 𝐼𝑜𝑢𝑡 = 𝜇𝑛 𝐶𝑜𝑥 𝑣 − 𝑣𝑖𝑑 2 =
2 𝐿 𝑖𝑑 1
𝑊
2 𝐿
1 𝑊 2 2 𝐼𝑜𝑑
𝜇𝑛 𝐶𝑜𝑥 𝑣𝑖𝑑 2𝑉𝑒𝑓𝑓0 − 𝑣𝑖𝑑 𝐼𝑠𝑠
2 𝐿
• 𝑉𝑒𝑓𝑓0 : Effective voltage at zero bias.
• Note: When current steers:
= 2𝑉𝑒𝑓𝑓0 𝑣𝑖𝑑
– 𝑣𝑖𝑑 𝑐𝑙𝑖𝑝𝑝𝑖𝑛𝑔 = 𝑉𝑒𝑓𝑓 = 2𝑉𝑒𝑓𝑓0

12
I-V Characteristics
1 𝑊 2𝐼𝑆𝑆
• 𝐼𝐷1 − 𝐼𝐷2 = 𝐼𝑜𝑢𝑡 = 𝜇𝑛 𝐶𝑜𝑥 𝑣 − 𝑣𝑖𝑑 2 =
2 𝐿 𝑖𝑑 1
𝑊
2 𝐿
1
2 2
1 𝑊 𝑣𝑖𝑑
𝑣𝑖𝑑 𝜇 𝐶 × 2𝐼𝑆𝑆 × 1 − = 𝑔𝑚0 𝑣𝑖𝑑 1 −
2 𝑛 𝑜𝑥 𝐿 2𝐼𝑆𝑆
1 𝑊
2 𝐿

13
I-V Charactereitics
2
𝑣𝑖𝑑 𝑔𝑚0 3
• 𝐼𝑜𝑢𝑡 ≅ 𝑔𝑚0 𝑣𝑖𝑑 1 − 2 = 𝑔𝑚0 𝑣𝑖𝑑 − 2 𝑣𝑖𝑑
8𝑉𝑒𝑓𝑓0 8𝑉𝑒𝑓𝑓0
• This is clearly a nonlinear I-V characteristic. It includes odd-

order harmonics ONLY.
– Remember: Differential Circuits suppress even-order harmonics.
• To measure the nonlinearity, we compute Harmonic

3 𝑊
𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 𝑜𝑓 𝑣𝑖𝑑 1 𝜇𝑛 𝐶𝑜𝑥 𝐿
Distortion 𝛼 = 2 =
𝑐𝑜𝑒𝑓𝑓𝑖𝑐𝑖𝑒𝑛𝑡 𝑜𝑓 𝑣𝑖𝑑 8𝑉𝑒𝑓𝑓0 8 𝐼𝑆𝑆

14
Distortion versus input Amplitude
𝑣𝑖𝑛 𝑣𝑜𝑢𝑡
• If 𝑣𝑖𝑛 = 𝐴 cos 𝜔1 →
𝑣𝑜𝑢𝑡 = 𝐴1 cos 𝜔1 𝑡 + 𝐴3 cos(3𝜔1 𝑡)
• 𝐴1 ≅ 𝛼1 𝐴 (Find exact)
𝛼3 𝐴3
• 𝐴3 =
4

15
Total Harmonic Distortion (THD)
𝐴3 4𝛼3
• 𝑇𝐻𝐷 ≅ =
𝐴1 𝛼1 𝐴2
• Important Factors:
– 𝛼3 to 𝛼1 ratio (the lower the better).
– Signal Amplitude 𝐴 (the smaller the better)

16
𝑉𝑜𝑢𝑡 (𝜔)
10𝑑𝐵
• If 𝑣𝑖𝑛 = 𝐴 cos(𝜔1 ) → 𝐴 = 𝐴0
𝑣𝑜𝑢𝑡 = 𝐴1 cos 𝜔1 𝑡 +
𝐴3 cos(3𝜔1 𝑡)
𝜔1 3𝜔1 𝜔
𝛼3 𝐴3 𝑉𝑜𝑢𝑡 (𝜔)
• 𝐴3 = 10𝑑𝐵
4 𝐴 = 𝐴0 /2 ?
𝜔1 3𝜔1 𝜔
17
Complete Circuit
• We need a way to subtract the two current.

VDD
ID1
Iout
ID1 ID2
ViCM+ ½ Vid ViCM - ½ Vid ZL
Iss
• Disadvantages:
– Available output range depends on input voltage DC-level (𝑉𝑖𝑐𝑚 ).
– Output Resistance. OTA is supposed to be a current source  Output resistance is ideally ∞. In this
circuit: 𝑅𝑜𝑢𝑡 = 𝑟0𝑛 //𝑟𝑜𝑝
– Unbalanced differential pair.

18
OTA
Balanced Circuit
VDD
ID1
Iout
ID1 ID2
ViCM+ ½ Vid ViCM - ½ Vid
ID2
ZL ID2
Iss
• Input differential pair have symmetric loads  Balanced.

• Output Resistance: 𝑅𝑜𝑢𝑡 = 𝑟0𝑛 //𝑟𝑜𝑝
• Available output range is decoupled from the input range and is
larger.

19
OTA
Balanced OTA with Cascode Output Stage
VDD
Iout ID1 ID2

ID2
ZL Iss
• Balanced
• Output Range is decoupled from input level
2 // 𝑔
• 𝑅𝑜𝑢𝑡 ≅ 𝑔𝑚𝑛 𝑟𝑜𝑛 2
𝑚𝑝 𝑟𝑜𝑝

20
OTA
How can we improve linearity even more?
• Degeneration
VDD
Iout ID1 ID2

R
ID2
ZL Iss/2 Iss/2

Active Circuits
Lecture 5: Linearized Gm Cells
maboudina@gmail.com
2
OTA
How can we improve linearity even more? Example # 1
• Degeneration
VDD
Iout ID1 ID2

2R
ID2
ZL Iss/2 Iss/2

3
𝑉𝑜𝑢𝑡 (𝜔)
10𝑑𝐵
• If 𝑣𝑖𝑛 = 𝐴 cos(𝜔1 ) → 𝐴 = 𝐴0
𝑣𝑜𝑢𝑡 = 𝐴1 cos 𝜔1 𝑡 +
𝐴3 cos(3𝜔1 𝑡)
𝜔1 3𝜔1 𝜔
𝛼3 𝐴3 𝑉𝑜𝑢𝑡 (𝜔)
• 𝐴3 = 10𝑑𝐵
4 𝐴 = 𝐴0 /2 ?
𝜔1 3𝜔1 𝜔
4
Looking Deep into a Gm cell
• Without Degeneration:
– Sources of nonlinearity:
• Sum of both currents is constant, Hence the I-V characteristic is saturating
(NONLINEARITY)
• Voltage is converted to current through a nonlinear device “MOS
Transistor”, Hence the output current is a nonlinear function of the input.
• With Degeneration:
– Differential input is voltage divided between:
1
• Input pair (𝑔 )[Nonlinear voltage to current conversion]
𝑚
• Resistance (𝑅)[Linear voltage to current conversion]
– Since, the voltage drop on the input pair is less, then the nonlinearity
component that is generated is also smaller leading to a more linear
overall I-V characteristic.

5
Looking Deep into a Gm cell
VDD
Iout ID1 ID2

½ Vid ½ Vid
+ 2R +
- - ID2
+ -
ZL Iss/2 Iss/2
• Total input voltage is divided between 𝑣𝑔𝑠1 , 𝑉𝑅 𝑎𝑛𝑑 𝑣𝑠𝑔2 .

1
2 2
𝑔𝑚0 𝑣𝑖𝑑 1
• 𝑖𝑑𝑖𝑓𝑓 = 𝑣
1+𝑔𝑚0 𝑅 𝑖𝑑
1− 2  𝐻𝐷3 ∝ 2
1+𝑔𝑚0 𝑅 2 2𝑉𝑒𝑓𝑓0 2 1+𝑔𝑚0 𝑅 2 2𝑉𝑒𝑓𝑓0

6
How to make this Gm cell programmable?
• Purpose: In most cases, after fabrication the Gm value is not

exactly what it was designed for, Hence we need to be able to
tune the Gm value.
• What are the variables that can be used to tune (Change) the
Gm value:
– Current (Iss) [Usually not used for tuning]
– W/L [Cannot change after fabrication]
– Degeneration resistance value ( R )

7
Tunable Gm Cell
VDD
Iout ID1 ID2

2R
ID2
ZL Iss/2 Iss/2
𝑔𝑚
• 𝐺𝑚 =
1+𝑔𝑚 𝑅
• Changing the value of R, changes the value of Gm.
1
• If 𝑔𝑚 𝑅 ≫ 1 , then 𝐺𝑚 ≅ (Highly degenerated).
𝑅

8
Example # 2: Practical Implementation of Tunable Gm Cell
• Use Transistor in “Triode” region.

1
• 𝑅𝑜𝑛 = 𝑊 −>
𝜇𝑛 𝐶𝑜𝑥 𝐿 𝑐 (𝑉𝐺𝑆 −𝑉𝑇𝐻 )
𝑐
𝑔𝑚
𝐺 =
𝑚 𝑅
1+𝑔𝑚 𝑂𝑁
2
• 𝑉𝐺𝑆 = (𝑉𝑐 − 𝑉𝑠 )
• Changing 𝑉𝑐 controls the Gm value.
• BUT, Also 𝑅𝑜𝑛 varies with 𝑉𝑠 .
• Assume 𝑉𝑖𝑑 = 0, Then 𝑉𝑠 = 𝑉𝑖𝐶𝑀 −
𝑉𝐺𝑆𝑖𝑛𝑝𝑢𝑡𝑝𝑎𝑖𝑟 .
• 𝑉𝑠 is a function of 𝑉𝑖𝐶𝑀 → 𝐺𝑚
changes with the Input CM level.
(NOT EXPECTED).

10
Example # 3:
VDD
ID1
VDD Iout
ID1 ID2
ViCM+ ½ Vid M1 M3 M4 M2 ViCM - ½ Vid ZL
Iss ViCM Iss
• What’s the effective resistance between both transistor

sources?
2
– Answer: 𝑅𝑜𝑛 =
𝑔𝑚3
𝑔𝑚1
• 𝐺𝑚 =
1+𝑔𝑚1 /𝑔𝑚3
– Question: If 𝑔𝑚1 = 𝑔𝑚3 , What is the improvement in THD in dBs?
11
Example # 4:
2
𝑖𝑜𝑑 2𝐼1
𝑅𝐸
𝑰𝑹𝑬
𝑅𝐸 𝐼1 𝑉𝑖𝑖𝑛𝑑𝑖𝑓𝑓
−2𝐼1
𝑉𝑖𝑛𝑑𝑖𝑓𝑓
• Opamps are in negative feedback  𝑉+ = 𝑉−  𝐼𝑅𝐸 =
𝑅𝐸
2
• 𝐼𝐷1 − 𝐼𝐷2 = 𝐼1 + 𝐼𝑅𝐸 − 𝐼1 − 𝐼𝑅𝐸 = 2𝐼𝑅𝐸 = 𝑉𝑖𝑛𝑑𝑖𝑓𝑓 ×
𝑅𝐸
2
• 𝐺𝑚 =
𝑅𝐸

12
Example # 5:
• Replace 𝑅𝐸 with MOSFET in Triode
• Gate Control “Vc” can be used to obtain a tunable Gm value.

13
Example # 6:
= 𝑉𝐶
• Both input transistors are biased into Triode region

𝑊 𝜕𝐼𝐷 𝑊
• 𝐼𝐷 ≅ 𝜇𝑛 𝐶𝑜𝑥 𝑉𝐺𝑆 − 𝑉𝑇𝐻 𝑉𝐷𝑆 −→ 𝐺𝑚 = = 𝜇𝑛 𝐶𝑜𝑥 𝑉
𝐿 𝜕𝑉𝐺𝑆 𝐿 𝐶

Active Circuits
Lecture 6: Gm-C Filters
maboudina@gmail.com
2
Outline
• Basic Cells
• First-Order Filters
• Second-Order Filters
– Two-loop Configurations (Biquadratic)
– Distributed Feedback
– Node Current Injection

3
Basic Cells
• What are our basic cells:

– 𝐻 𝑠 =𝐴
1
– 𝐻 𝑠 =
𝑠
1
– 𝐻 𝑠 =
𝑠+𝑎
𝑠+𝑏
– 𝐻 𝑠 =
𝑠+𝑎
1
– 𝐻 𝑠 = 𝜔
𝑠 2 + 𝑄0 𝑠+𝜔02
𝑍𝑒𝑟𝑜𝑠 (1 𝑜𝑟 2)
– 𝐻 𝑠 = 𝜔
𝑠 2 + 𝑄0 𝑠+𝜔02

4
Basic Cells
+
• Amplifiers: Vin Gm1 Vout
-
𝑉𝑜𝑢𝑡 𝐺𝑚1
– =
Gm2
• Integrator:
𝑉𝑜𝑢𝑡 𝐺𝑚
– = +
Vin Gm
𝑉𝑖𝑛 𝑠𝐶 Vout
-

5
Lossy Integrator
Vout
+
Vin Gm1 Gm2
Gm2
-
𝑉𝑜𝑢𝑡 1 Gm3
Gm Vout =
𝑉𝑖𝑛 𝐺𝑚2 𝑠𝐶
Vin +
𝐺𝑚1 𝐺𝑚1
C
Gm Vout
Vin
C
𝑉𝑜𝑢𝑡 1
=
𝑉𝑖𝑛 𝑠𝐶 𝑉𝑜𝑢𝑡
1+𝐺 =? ?
𝑚 𝑉𝑖𝑛

6
Second-Order Filters
Gm1 Gm1 Vout

Vin
C1 C2
𝐺𝑚1 𝐺𝑚2
𝑉𝑜𝑢𝑡 𝐶1 𝐶2
• = 𝑠𝐺 𝐺 𝐺
𝑉𝑖𝑛 𝑠 2 + 𝑚2 + 𝑚1 𝑚2
𝐶2 𝐶1 𝐶2
𝐺𝑚1 𝐺𝑚2
• 𝜔0 =
𝐶1 𝐶2
𝜔0 𝐺𝑚2
• =
𝑄 𝐶2

7
Two-Integrator Loop
−𝑘12 −𝑘12
−𝑘11
−𝑘22
1 1 1 1
𝑠𝜏1 𝑠𝜏2 𝑠𝜏1 𝑠𝜏2
• Note: Input and Outputs are not shown.

• The signal flow graph, shows us the poles locations but finding the loop
gain.
• Denominator:
– 𝐷 𝑠 = 𝜏1 𝜏2 𝑠 2 + 𝑘11 𝜏2 𝑠 + 𝑘12 (For the first one)
– Find D(s) for the second one.

8
What are the types of filters we can get?

9
Distributed-Feedback (DF)
𝐶1
• 𝜏1 = Gm5
𝐺𝑚1 Vo4
Vi5
𝐶2
• 𝜏2 = Gm6
𝐺𝑚2 Vi3
Vi6
𝐺𝑚5
• 𝑘12 = Vo3
Gm3
𝐺𝑚6
𝐺𝑚3
• 𝑘22 = Gm4
𝐺𝑚4 Vi4
Gm2 Vo2
Vo1
Gm1
Vi1 C2
C1

10
Possible Realizations of Distributed Feedback Filters

11
How to Implement zeros?
• Most of the previous techniques create zeros that are tightly

coupled to the poles.
• In order to generate flexible zeros, we need to add external

paths to the one of more of the nodes of system.
– This does not increase the total number of nodes  No extra poles.
– But only adds different paths from input to output  Zeros due to
multi-path.

12
Example of Node Current Injection
Vin Vout
Gm1 Gm2 Gm4
C1 C2
Gm3
𝐺𝑚4
• 𝑉𝑜𝑢𝑡 = 𝐺𝑚3 𝑠2 𝐶1 𝐶2
1+𝑠 𝐶1 𝐺 𝐺 +𝐺 𝐺
𝑚2 𝑚4 𝑚2 𝑚4

13
Vin Vout
Gm1 Gm2 Gm4
C1 C2
Ik Gm3
𝑉𝑖𝑛 𝐺𝑚1 𝐼𝑘 𝑠𝐶1

𝐺𝑚4
−𝐺𝑚2 𝐺𝑚4
• 𝑉𝑜𝑢𝑡 = 𝐺𝑚3 𝑠2 𝐶1 𝐶2
1+𝑠 𝐶1 𝐺 𝐺 +𝐺 𝐺
𝑚2 𝑚4 𝑚2 𝑚4

14
Vin Vout
Gm1 Gm2 Gm4
C1 C2
Ik Gm3
• Case 1: 𝐼𝑘 = 𝑔𝑚𝑘 𝑉𝑖𝑛

𝐺𝑚1 𝑠𝐶1 𝑔𝑚𝑘
1− 𝑉𝑖𝑛
𝐺𝑚4 𝑔𝑚2 𝑔𝑚1
– 𝑉𝑜𝑢𝑡 =
𝐷(𝑠)
• Case 2: 𝐼𝑘 = 𝑠𝐶𝑘 𝑉𝑖𝑛
𝐺𝑚1 𝐶1 𝐶𝑘
1−𝑠 2 𝑉𝑖𝑛
𝐺𝑚4 𝐺𝑚2 𝐺𝑚1
– 𝑉𝑜𝑢𝑡 =
𝐷(𝑠)

15
Transfer Function Zeros

1
𝑠
1+𝑠
𝑝1
𝑣𝑖𝑛  𝑣𝑜𝑢𝑡
1
𝑠
1+𝑠
𝑝2
𝑠
𝑉𝑖𝑛 𝑉𝑖𝑛 1+𝑠
• 𝑉𝑜𝑢𝑡 = 𝑠 + 𝑠 = 2. 𝑠
𝑧
𝑠 × 𝑉𝑖𝑛
1+𝑠 1+𝑠 1+𝑠 .1+𝑠
𝑝1 𝑝2 𝑝1 𝑝2
• Multipath creates zeros.

Active Circuits Lectures 1 2 3 4 5 6 DR - Abodina

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Active Circuits Lectures 1 2 3 4 5 6 DR - Abodina

Uploaded by

Copyright:

Available Formats

Active Circuits

© Mohamed M. Aboudina, 2015

• Acquire a thorough understanding of the basic principles of

© Mohamed M. Aboudina, 2015

© Mohamed M. Aboudina, 2015

• Example 1: Communication Receiver

© Mohamed M. Aboudina, 2015

• Example 2: Hard Disk Drive (HDD)

© Mohamed M. Aboudina, 2015

Lecture # Topic Assessment

© Mohamed M. Aboudina, 2015

• B. Razavi, “Fundamentals of Microelectronics”. Chapter 14.

• T.L.Deliyannis, Y.Sun, J.K.Fidler, “Continuous-time active filter

© Mohamed M. Aboudina, 2015

• Early 1920s to Late 1960s: Voice filters were realized in RLC

• In the 1950s: a goal to reduce the size and cost of inductors by

• First book on Active RC Circuits “Theory and Design of Active

• Early 1970s, Thin-film hybrid integrated circuits were

2/9/2015 © Mohamed M. Aboudina, 2015

• Integrating filters on chip is one of the main demands in most

• Active RC filters, Gm-C filters, Switched-Capacitor Filters are all

2/9/2015 © Mohamed M. Aboudina, 2015

Biomedical Hard disk drive

Hz - KHz MHz GHz

2/9/2015 © Mohamed M. Aboudina, 2015

• Use Passive components (R, L, C)

2/9/2015 © Mohamed M. Aboudina, 2015

• Filters constructed with Resistors and Capacitors only result

2/9/2015 © Mohamed M. Aboudina, 2015

• Use of Rs and Cs along with active components (mainly

• Effectively, inductors are replaced with Active Circuits. (A

2/9/2015 © Mohamed M. Aboudina, 2015

• Low Pass Filter (LPF): Practical - Approximation

• Ideal response is called: “Brick Wall”

2/9/2015 © Mohamed M. Aboudina, 2015

• High-Pass Filter (HPF):

• Band-Pass Filter (BPF):

• Band-Stop Filter (BSF):

2/9/2015 © Mohamed M. Aboudina, 2015

• All-Pass Filter (APF):

2/9/2015 © Mohamed M. Aboudina, 2015

2/9/2015 © Mohamed M. Aboudina, 2015

• If the signal is sinusoidal with frequency ω , Replace s with 𝑗ω

2/9/2015 © Mohamed M. Aboudina, 2015

2/9/2015 © Mohamed M. Aboudina, 2015

• Mathematically, for a system to be stable:

• What about the impulse response of a right-half plan pole?

2/9/2015 © Mohamed M. Aboudina, 2015

– 𝑠𝑥 is a specific value, can be real or 𝜎

© Mohamed M. Aboudina, 2015

• Moving only on the 𝑗𝜔 axis. 𝑠𝑥 = −𝜎𝑥 + 𝑗𝜔𝑥

© Mohamed M. Aboudina, 2015

© Mohamed M. Aboudina, 2015

© Mohamed M. Aboudina, 2015

© Mohamed M. Aboudina, 2015

© Mohamed M. Aboudina, 2015

© Mohamed M. Aboudina, 2015

At w<<wo, filter has unity gain.

© Mohamed M. Aboudina, 2015

© Mohamed M. Aboudina, 2015

© Mohamed M. Aboudina, 2015

© Mohamed M. Aboudina, 2015

© Mohamed M. Aboudina, 2015