ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Continuous-Time Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Introduction Distortionless Transmission Nonideal Filters Filter Design Irma Zakia School of Electrical Engineering and Informatics, Institut Teknologi Bandung *Untuk kalangan terbatas Butterworth Filter Chebyshev Fil. March 28, 2019 Frequency Transformation Equalization 1 / 89 Presentation Outline ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 1 Introduction 2 Distortionless Transmission 3 Nonideal Filters 4 Filter Design 5 Butterworth Filter 6 Chebyshev Fil. 7 Frequency Transformation Chebyshev Fil. Frequency Transformation 8 Equalization Equalization 2 / 89 Presentation Outline ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 1 Introduction 2 Distortionless Transmission 3 Nonideal Filters 4 Filter Design 5 Butterworth Filter 6 Chebyshev Fil. 7 Frequency Transformation Chebyshev Fil. Frequency Transformation 8 Equalization Equalization 3 / 89 Filtering ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter The convolution property of LTI system is the whole basis of filtering, for which the input-output of LTI system is related by the filter H(jω) The system performs filtering on the input signal by presenting a different response to components of the input that are at different frequencies This also means that by filtering, the frequency components of the input are affected in terms of magnitude and phase Chebyshev Fil. Frequency Transformation Equalization 4 / 89 Typical Parameters in Filtering ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization Passband : range of frequencies that are passed Stopband : range of frequencies that are attenuated Transition band : range of frequencies for which the transition from passband to stopband occurs, and vice versa The magnitude squared response |H(jω)|2 is commonly described in units of dB as |H(jω)|dB = 20 log10 |H(jω)| dB Cut-off frequency ωc frequency for which the magnitude squared response |H(jω)|2 decreases to 21 its maximum value, or |H(jω)| decreases to √12 of its maximum value Cut off frequency is also called the −3 dB point At cut off frequency, the filter passes half of the input power Majority of filtering involves real impulse response h(t), which yields |H(jω)| to be an even function. Thus plotting |H(jω)| for ω ≥ 0 suffices 5 / 89 Presentation Outline ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 1 Introduction 2 Distortionless Transmission 3 Nonideal Filters 4 Filter Design 5 Butterworth Filter 6 Chebyshev Fil. 7 Frequency Transformation Chebyshev Fil. Frequency Transformation 8 Equalization Equalization 6 / 89 Definition ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Block diagram of a continuous time LTI system F x(t) − → X (jω) Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization F y (t) − → Y (jω) Distortionless transmission : output is an exact replica of the input, except, for a possible modifications 1 Introduction F h(t) − → H(jω) 2 scaling of amplitude C time delay t0 In time-domain, distortionless transmission yields y (t) = C x(t − to ) The system impulse response for distortionless transmission becomes h(t) = C δ(t − to ) whereas the frequency response H(jω) = C e −jωt0 7 / 89 Frequency Response for Distortionless Transmission ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Two requirements for distortionless transmission For the range frequencies of interest: 1 Magnitude response must be constant |H(jω)| = C C Irma Zakia Introduction ω Distortionless Transmission Nonideal Filters Filter Design 2 0 Phase response must be linear and intercept zero arg(H(jω)) = −ωt0 arg(H(jω)) Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization ω 0 slope = −t0 8 / 89 Frequency Response of Ideal LPF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters The frequency response of an ideal LPF arg(H(jω)) |H(jω)| 1 ωc Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization −ωc 0 ω ωc −ωc (a) Magnitude Response H(jω) = e −jωt0 , 0, 0 ω slope = −t0 (b) Phase Response |ω| ≤ ωc |ω| > ωc where C = 1 Distortionless transmission is achieved by ideal filters It requires abrupt transition from passband to stopband In this chapter, we put emphasis on the frequency characterization of system (frequency response), although we know that the step response is also significant in the design of some applications, like automobile suspension. 9 / 89 Presentation Outline ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 1 Introduction 2 Distortionless Transmission 3 Nonideal Filters 4 Filter Design 5 Butterworth Filter 6 Chebyshev Fil. 7 Frequency Transformation Chebyshev Fil. Frequency Transformation 8 Equalization Equalization 10 / 89 Practical Filters ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Ideal filters is impractical, since the discontinuous transition band is not realizable. It is actually noncausal as well. On the other hand, nonideal filters with gradual transition from passband to stopband is generally preferable. Distortionless Transmission Nonideal Filters X (jω) X2 (jω) X1 (jω) Filter Design Butterworth Filter Chebyshev Fil. ω Frequency Transformation Equalization 11 / 89 Characteristic of RC LPF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction First order RC LPF circuit R x(t) + − C 20 Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization − y (t) 3 dB apprx. |H(jω)|dB Nonideal Filters + 1 The transfer function H(s) = 1+sRC Plot of the magnitude response in Bode diagram 0 dB Distortionless Transmission − + −20 −40 0.1 RC 1 RC 10 RC 100 RC ω 1 The cutoff frequency is RC 1 The pole is real at s = − RC 1 The roll-off (asymptotic slope ω >> RC ) is −20 dB/decade. This roll-off is related to the transition band Higher roll-off can be achieved by cascading these first-order RC LPFs, e.g. cascade of N filters yields roll-off −20N dB/decade 12 / 89 Fourth Order Cascaded RC LPF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Cascade interconnection of N = 4 first-order RC LPF Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization Op-amp is used to avoid loading effect, thus, separate the individual filter stages The transfer function becomes H(s) = 1 (1 + sRC )4 1 Yet, real poles (4 poles) at s = − RC Cutoff of each individual filter is cascaded interconnection 1 RC times higher than the 13 / 89 Frequency Response of Fourth Order Cascaded RC LPF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Roll-off becomes −80 dB/decade Distortionless Transmission |H(jω)|dB Introduction arg(H(jω)) Irma Zakia Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation normalized frequency ω̃ (a) magnitude response normalized frequency ω̃ (b) phase response Reference: Texas Instruments, Chapter 16: Active Filter Design Techniques, Literature Number SLOA088 Equalization 14 / 89 Cascaded First-order Filters ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Compare to ideal fourth order filter, the cascaded RC has : Not flat passband gain Wider transition band (the 80 dB roll-off is shifted 1.5 octaves above cut-off) Nonlinear phase response Consider transfer function of first-order H(s) = Cascading two first-order filters yields Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation 1 s+ω0 H(s) = 1 1 1 = 2 s + ω0 s + ω0 s + 2ω0 s + ω02 This results in quality factor Q = 0.5 which means that: gradual transition from passband to stopband significant attenuation in the passband Equalization 15 / 89 Optimizing the Frequency Response ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission using real poles are not enough, so use complex conjugate poles complex conjugate poles allow designer to optimize one of the following filter criteria : Nonideal Filters 1 2 Filter Design 3 Maximum flatness in passband Smaller transition band Linear phase response (at least in the passband) Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 16 / 89 Optimizing the Frequency Response continues ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation the transfer function becomes H(s) = b̃ 1 (s 2 + c1 s + d1 )(s 2 + c2 s + d2 ) · · · (s 2 + cn s + dn ) the denominator is cascade of second-order LPF, where the filter coefficients ck and dk , k = 1, 2, · · · n are real the filter coefficients determine the complex pole locations the filter coefficients are determined based on specific criteria the Butterworth filter optimize maximum flatness in passband the Chebyshev filter sharpening the transition band the Bessel filter linearize the phase response (in passband) Equalization 17 / 89 Presentation Outline ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 1 Introduction 2 Distortionless Transmission 3 Nonideal Filters 4 Filter Design 5 Butterworth Filter 6 Chebyshev Fil. 7 Frequency Transformation Chebyshev Fil. Frequency Transformation 8 Equalization Equalization 18 / 89 Background ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Although signal processing is done mainly in digital domain, the physical signal is analog. In digital processing of analog signals, LPFs are used. Front-end: restrict bandwidth for sampling Far-end: smoothing / remove higher frequency components Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Analog filters are used as prototype for the design of the so-called Infinite Impulse Response (IIR) digital filters by using transformation from the s−plane to the z−plane (bilinear transformation) Focus is on the design of LPF filters, since other types of filters are obtained simply by frequency transformation. Equalization 19 / 89 Filter Design Specifications ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Nonideal/practical filters involve an acceptable values of ’distortion’ if compared to ideal filters Those values are the filter characteristics in frequency domain, which translates to filter specifications from a design perspective |H(jω)| 1 (1 − δp ) δp Distortionless Transmission Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization stopband passband Nonideal Filters transition band δs δs 0 ωp ωs ω Filter specifications: tolerance tolerance passband stopband in passband : δp in stopband : δs edge : ωp edge : ωs 20 / 89 Design Rule of Thumb ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation The design can’t always meet the specifications exactly. From financial perspective, the design has to be met with minimum cost. If with a filter order N, the specifications are fulfilled, don’t increase the order to N + 1. The passband specifications are a priority If the passband specifications are satisfied, then the stopband will be oversatisfied in terms of a lower ωs value or narrower transition width The design delivers the filter frequency response H(jω) or alternatively the transfer function H(s) After obtaining the transfer function, the filter is implemented in a physical system Equalization 21 / 89 Analog Filter Types ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Analog filter prototypes: Bessel filter, Butterworth, Chebyshev, elliptic Bessel Introduction Distortionless Transmission Nonideal Filters |H(jω)|dB Irma Zakia Butterworth Chebyshev I Chebyshev II elliptic Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation normalized frequency ω̃ Figure: Fifth order filter Equalization Reference: inst.eecs.berkeley.edu/ ee247/fa04/fa04/lectures/L2 f04.pdf 22 / 89 Group Delay Comparison ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 23 / 89 Comparison of Chebyshev Type 1 and Type 2 ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission (a) s-plane Type 1 Nonideal Filters Filter Design Butterworth Filter Figure: magnitude and phase response Chebyshev Fil. Frequency Transformation Equalization (b) Reference:2.161 Signal Processing: Continuous and Discrete Fall 2008, MIT OpenCourseWare, ocw.mit.edu s-plane Type 2 24 / 89 Which filter type? ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Trade-off between transition width and phase response Filters with high signal attenuation per pole has poor phase response For a given signal attenuation requirement of preserving constant group delay, we require higher order filter, but more cost and area In cases where analog filter is followed by digital processing, it is possible to digitally correct for phase non-linearities incurred by the analog circuitry by using phase equalizers Elliptic : optimum but requires numerical computation. Optimum in the sense that no other filter of the same order has narrower transition width for a given passband and stopband tolerance This chapter considers the design of Butterworth and Chebyshev Equalization 25 / 89 Presentation Outline ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 1 Introduction 2 Distortionless Transmission 3 Nonideal Filters 4 Filter Design 5 Butterworth Filter 6 Chebyshev Fil. 7 Frequency Transformation Chebyshev Fil. Frequency Transformation 8 Equalization Equalization 26 / 89 Description of a Butterworth Filter ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Butterworth filter of order N is described by the magnitude squared of its frequency response 1 |H(jω)|2 = 2N 1 + ωωc where ωc is the cutoff frequency Distortionless Transmission N=∞ 1 Nonideal Filters Butterworth Filter N = 25 |H(jω)|2 Filter Design N=2 N=1 0.5 Chebyshev Fil. Frequency Transformation 0 ωc ω Equalization 27 / 89 Properties of Butterworth Filter ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization For any order N, |H(jω)|2 |H(jω)| =1 ω=0 For any order N 2 |H(jω)| ω=ωc ω=ωc 1 = 2 = , −3.0103 dB 0.707 2 |H(jω)| is a monotonic function (no maxima or minima) As the order N increases, |H(jω)|2 approaches an ideal LPF |H(jω)| is maximally flat at the origin By using Binomial series, the magnitude response |H(jω)| = 1+ ω ωc 2N !− 1 2 =1− 1 2 ω ωc 2N + higher order of ω ωc 2N Taking the derivative d k |H(jω)| dω k ω=0 , k = 1, 2, .., 2N − 1 Butterworth filter has 2N − 1 (a maximum possible number) derivatives which vanish at ω = 0 (flat curve at ω = 0) 28 / 89 Asymptotic Property of All-Pole Filters ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Asymptotic property of |H(jω)| for ω → ∞ This property is valid for any all-pole filters, not only Butterworth −N ω |H(jω)|ω→∞ ≈ ωc ω |H(jω)|dB ≈ −20N log ωc Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Roll-off/asymptotic slope −20N dB/decade Equalization 29 / 89 Example of Butterworth LPF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia First order RC circuit + Vr (t) − R + Vs (t) − + Vc (t) − C Introduction Distortionless Transmission Nonideal Filters RC circuit acts as LPF for the capacitor voltage Vc (t) The frequency response H(jω) = 1 1 + jωRC Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization The magnitude squared response yields |H(jω)|2 = 1+ 1 ω 1 RC 2 which shows the response of a first order Butterworth filter with 1 ωc = RC 30 / 89 Design the Butterworth Filter ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Need to define H(s) What we get from the filter specification is |H(jω)|2 Assume h(t) real, thus H(jω)∗ = H(−jω) |H(jω)|2 = H(s)H(−s) = Irma Zakia Introduction Distortionless Transmission Nonideal Filters 1+ 1+ Poles for H(s)H(−s) are derived from −s 2 ωc2 N 1 ω ωc 1 2N −s 2 ωc2 N = −1 !N = e j(2k−1)π , k = 1, 2, · · · , N Frequency Transformation sk2 = −ωc2 e j Equalization sk = ±jωc e j Filter Design Butterworth Filter Chebyshev Fil. −sk2 ωc2 (2k−1)π N (2k−1)π 2N = ωc e j (2k−1)π ±π 2N 2 31 / 89 Poles Characteristic ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Poles of H(s)H(−s) are at sk = ωc e ℑ{s} N=1 j (2k−1)π ±π 2N 2 , k = 1, 2, · · · , N ℑ{s} N=2 π 2 ωc ℜ{s} Irma Zakia ωc ℜ{s} Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter ℑ{s} N=3 N=6 π 3 Chebyshev Fil. Frequency Transformation ℑ{s} ωc π 6 ℜ{s} ωc ℜ{s} Equalization 32 / 89 Poles of Butterworth Filter ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Focusing on poles at left-half of s-plane (LHP) → take the + sign sk = = j (2k−1)π +π 2N 2 , k = 1, 2, · · · , N ωc e (2k − 1)π (2k − 1)π ωc − sin + j cos 2N 2N Poles are at radius ωc and equally spaced in a circle, with 2π spacing 2N A pole never lies on the ℑ{s} axis, and occur at ℜ{s} axis for N odd Further work on the transfer function considers the normalized Butterworth filter, such that ωc = 1 rad/s. If not normalized, the parameter s is replaced by transfer function s ωc in the Equalization 33 / 89 Determining the Transfer Function ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Order N = 1, pole at s = −1, thus 1 H(s) = s +1 Order N = 2, poles at s = H(s) = Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization −1 √ 2 s2 ± j √12 , thus 1 √ + 2s + 1 Order N = 3, poles at s = −1, s = H(s) = 1 −1 2 ±j √ 3 2 , thus (s + 1)(s 2 + s + 1) If order N odd, there exists a single real pole at s = −1, besides N−1 pairs of complex poles 2 If order N even, all poles are complex conjugate Denominator of H(s) is the Butterworth polynomial 34 / 89 Determining the Transfer Function continues.. ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters For N > 1, complex conjugate poles are present Rewrite sk (assume normalized filter ωc = 1 rad/s) sk = − sin Irma Zakia (2k − 1)π 2N + j cos (2k − 1)π 2N Each complex pole results in the denominator of H(s) as Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter = s + sin 2 s + 2 sin | (2k − 1)π 2N (2k − 1)π 2N {z bk Equalization (2k − 1)π 2N Define a constant bk = 2 sin k= s + sin (2k − 1)π 2N − j cos (2k − 1)π 2N s+1 } Chebyshev Fil. Frequency Transformation + j cos 1, 2, · · · , 1, 2, · · · , (2k−1)π 2N , where N 2 for N even N−1 2 for N odd 35 / 89 Determining the Transfer Function continues.. ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization General transfer function 1 , N even N 2 Q (s 2 +bk s+1) k=1 H(s) = 1 , N odd N−1 2 Q 2 (s+1) (s +bk s+1) k=1 If not normalized (ωc 6= 1), replace s → ωsc ωcN , N even N 2 Q 2 2 (s +bk ωc s+ωc ) k=1 H(s) = ωcN , N odd N−1 2 Q 2 2 (s+ωc ) (s +bk ωc s+ωc ) k=1 Butterworth polynomial only requires to determine bk , no need to know pole locations 36 / 89 Transfer Function for N = 4 ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters The constants π 8 = 3π b2 = 2 sin 8 = b1 = 2 sin avoid rounding numbers The transfer function Distortionless Transmission H(s) = 1.84776 | {z } avoid rounding numbers Irma Zakia Introduction 0.76537 | {z } 1 (s 2 + 0.76537s + 1)(s 2 + 1.84776s + 1) Nonideal Filters Filter Design Butterworth Filter If filter is not normalized (ωc 6= 1) H(s) = Chebyshev Fil. Frequency Transformation = s ωc 2 + 0.76537 s ωc 1 2 s s +1 + 1.84776 + 1 ω ω c c ωc4 (s 2 + 0.76537ωc s + ωc2 )(s 2 + 1.84776ωc s + ωc2 ) Equalization 37 / 89 Transfer Function for N = 5 ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters The constants π 10 3π b2 = 2 sin 10 b1 = 2 sin Irma Zakia Introduction Filter Design Butterworth Filter H(s) = = 1.61803 1 (s + 1)(s 2 + 0.61803s + 1)(s 2 + 1.61803s + 1) If filter is not normalized (ωc 6= 1) H(s) = Chebyshev Fil. Frequency Transformation 0.61803 The transfer function Distortionless Transmission Nonideal Filters = = s ωc +1 s ωc 2 + 0.61803 ωc5 s ωc 1 +1 s ωc 2 + 1.61803 s ωc +1 (s + ωc )(s 2 + 0.61803ωc s + ωc2 )(s 2 + 1.61803ωc s + ωc2 ) Equalization 38 / 89 Determining Filter Order ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia From specifications of δp , δs , ωp , ωs → filter order N, and next ωc |H(jω)| 1 (1 − δp ) Introduction δs Distortionless Transmission ω Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization ωp 0 Nonideal Filters Recall : |H(jω)|2 = 1/ 1+( ωωc )2N Inserting the specifications 1 2N ω 1+( ωpc ) ωp ωc 2N ωs 1 ωs 1+( ω c = (1 − δp )2 , = 1 (1−δp )2 − 1, ωs ωc ) 2N 2N = = δs2 1 δs2 −1 39 / 89 Determining Filter Order continues ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction ωs ωp 2N = Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 1 −1 δs2 1 − (1−δp )2 log10 N ≥ s log10 1 1 −1 δs2 1 −1 (1−δp )2 ωs ωp Chebyshev Fil. Frequency Transformation Equalization 40 / 89 After Filter Order... what next? ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Passband requirement must be fulfilled exactly Two alternatives in determining ωc Take the one which satisfies the passband requirement exactly ωc = ωp 1 (1−δp )2 Determine ωs from equation 1 2N −1 1+ 1 ωs ωc 2N = δs2 Frequency for which |H(jω)| = δs is to the left of ωs . Let’s call this frequency ωs ′ . Do we perform better or worse? Equalization 41 / 89 Butterworth LPF Filter Design Example ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Design an analog Butterworth filter that has passband attenuation 2 dB at 20 rad/s and at least 35 dB attenuation at 40 rad/s (1 − δp )2 = 10−2/10 = 0.63096, ωp = 20 rad/s δs2 = 10−35/10 = 0.00031622, ωs = 40 rad/s r filter order 1 log10 ωc = Butterworth Filter Frequency Transformation Equalization ≥ N = log10 7 Cut-off frequency Filter Design Chebyshev Fil. N 20 1 0.63096 −1 −1 0.00031622 1 −1 0.63096 40 20 = 6.2 1/14 = 20.78106 rad/s The frequency which corresponds to δs 1/14 ′ ωs = 20.78106 1 −1 0.00031622 = 36.95393 rad/s < ωs 42 / 89 LPF Butterworth Design Example continues ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Transfer function for N = 7 H(s) = h ωc7 The constants Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Equalization b1 = b2 = b3 = π = 0.44504 14 3π = 1.24698 2 sin 14 5π 2 sin = 1.80194 14 2 sin Inserting those values, the transfer function yields Chebyshev Fil. Frequency Transformation i (s+ωc )(s 2 +b1 ωc s+ωc2 ) (s 2 +b2 ωc s+ωc2 )(s 2 +b3 ωc s+ωc2 ) H(s) = h 1.67369 109 i (s+20.78106)(s 2 +9.24840s+431.85245) (s 2 +25.91357s+431.85245)(s 2 +37.44622s+431.85245) 43 / 89 LPF Butterworth Design Example continues ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Dari hasil fungsi transfer H(s) H(s) = h 1.67369 109 (s+20.78106)(s 2 +9.24840s+431.85245) (s 2 +25.91357s+431.85245)(s 2 +37.44622s+431.85245) (a) (b) i Berapa nilai magnituda |H(jω) saat ω = ωp = 20 rad/s? Berapa nilai magnituda |H(jω) saat ω = ωs = 40 rad/s? Misal Anda mengambil orde filter N ≤ 6.2 = 6, (a) (b) Berapa nilai magnituda |H(jω) saat ω = ωp = 20 rad/s? Berapa nilai magnituda |H(jω) saat ω = ωs = 40 rad/s? Mengapa demikian? Frequency Transformation Equalization 44 / 89 Presentation Outline ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 1 Introduction 2 Distortionless Transmission 3 Nonideal Filters 4 Filter Design 5 Butterworth Filter 6 Chebyshev Fil. 7 Frequency Transformation Chebyshev Fil. Frequency Transformation 8 Equalization Equalization 45 / 89 Characteristics ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Performance of filter is increased by allowing ripple → equiripple For a given δp , δs , ωp , N, transition width of Chebyshev is narrower than Butterworth For a given δp , ωp , ωs , N, stopband attenuation of Chebyshev is larger than Butterworth Two types of Chebyshev : Type I and Type II Nonideal Filters |H(jω)| |H(jω)| Distortionless Transmission √1 1 1 √1 1+ǫ2 1+ǫ2 Filter Design δs Butterworth Filter Chebyshev Fil. Frequency Transformation δs ω ω 0 ωp ωs (a) Type I 0 ωp ωs (b) Type II Preferred: Type I (Chebyshev Type II has complex zeros, poor roll-off) Parameter ǫ is chosen such that the peak-to-peak passband ripple 1 − √ 1 1+ǫ2 is acceptable Equalization 46 / 89 Description of Chebyshev Filter ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Magnitude frequency response of order N |H(jω)| = q A 1 + ǫ2 CN2 ( ωωp ) A need not be 1, and CN (x) ≡ cos(Ncos−1 x), cosh(Ncosh−1 x), |x| < 1 |x| > 1 |x| < 1 translates to |ω| < ωp , since cosine is oscillatory → equiripple in passband |x| > 1 translates to |ω| > ωp , since cosh is monotonic → monotonic in stopband Without loss of generality, assume A = 1 Equalization 47 / 89 Chebyshev Polynomials ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters A recursive formula in determining the polynomial CN (x) = 2xCN−1 (x) − CN−2 (x), N ≥ 2 Irma Zakia N CN (x) 0 1 2 3 4 5 6 7 8 9 10 1 x 2x 2 − 1 4x 3 − 3x 8x 4 − 8x 2 + 1 16x 5 − 20x 3 + 5x 32x 6 − 48x 4 + 18x 2 − 1 64x 7 − 112x 5 + 56x 3 − 7x 128x 8 − 256x 6 + 160x 4 − 32x 2 + 1 256x 9 − 576x 7 + 432x 5 − 120x 3 + 9x 512x 10 − 1280x 8 + 1120x 6 − 400x 4 + 50x 4 − 1 Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 48 / 89 Polynomial Values at Origin and Passband Frequency ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Take a look at |H(jω)| = A q 1+ǫ2 CN2 ( ωωp ) What is |H(jω)| at ω = 0 and ω = ωp ? N = 0, CN ( ωωp ) ω=0 N = 1, CN ( ωωp ) ω=0 N = 2, CN ( ωωp ) ω=0 N = 3, CN ( ωωp ) ω=0 N = 4, CN ( ωωp ) ω=0 = 1, CN ( ωωp ) = 0, CN ( ωωp ) = −1, ω=ωp ω=ωp CN ( ωωp ) = 0, CN ( ωωp ) = 1, CN ( ωωp ) =1 =1 ω=ωp ω=ωp ω=ωp =1 =1 =1 and so on Equalization 49 / 89 Ripple Characteristic ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Recall : |H(jω)| = q 1 1+ǫ2 CN2 ( ωωp ) Filter magnitude at the origin |H(jω)| ω=0 = ( 1, √1 1+ǫ , 2 N odd N even Filter magnitude at passband frequency |H(jω)| ω=ωp = √ 1 1+ǫ2 Total number of maximum and minimum in passband is N |H(jω)| |H(jω)| 1 1 √ 1 1+ǫ2 √ 1 1+ǫ2 Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 0 ωp (a) Even (N = 4) ω 0 ωp ω (b) Odd (N = 5) 50 / 89 Cut-off Frequency ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Cut-off frequency is determined from 1 1+ ǫ2 CN2 ( ωωp ) ω=ωc = ωc CN = ωp −1 ωc = cosh Ncosh ωp ωc 1 2 1 >1 ǫ 1 ǫ = ωp cosh 1 cosh−1 N 1 ǫ Chebyshev Fil. Frequency Transformation Cut-off frequency in Chebyshev is not as significant as in Butterworth from a design perspective. Why? Equalization 51 / 89 Poles of Chebyshev ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Need to define H(s) No need to know poles location H(s)H(−s) 1 1 + ǫ2 CN2 ( jωs ) = p Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Poles for H(s)H(−s) are derived from ǫ2 CN2 s jωp = −1 Take the LHP poles Poles are located in an ellipse (compare to Butterworth!) Figure: Fourth order Chebyshev filter Equalization 52 / 89 Transfer Function ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization Denominator of H(s) contains poles in the form (s + ωp c0 ) and(s 2 + bk ωp s + ck ωp2 ) But bk is not the same as Butterworth Transfer function N 2 Q 1 N √ ck ω p 1+ǫ2 k=1 , N even N 2 Q (s 2 +bk ωp s+ck ωp2 ) k=1 H(s) = N−1 2 Q ωpN c0 ck k=1 , N odd N−1 2 2 2 (s+ωp c0 ) Q (s +bk ωp s+ck ωp ) k=1 53 / 89 Constants in the Transfer Function ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Let’s define "r #1/N "r #−1/N 1 1 1 1 1 1+ 2 + 1+ 2 + − dN = 2 ǫ ǫ ǫ ǫ The constants Distortionless Transmission c 0 = dN Nonideal Filters ck Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization bk (2k − 1)π 2N (2k − 1)π = 2dN sin 2N = dN2 + cos2 Compare to Butterworth : c0 = 1, ck = 1 and dN = 1 Compare coefficient bk to Butterworth. For Chebyshev dN 6= 1, hence poles location of Chebyshev traces an ellipse 54 / 89 What to do given filter specifications? ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. replacements |H(jω)| 1 (1 − δp ) δs ω 0 ωp ωs Determine ǫ and N Frequency Transformation Equalization 55 / 89 Determining ǫ and Filter Order ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Evaluating |H(jω)| at ω = ωp 1 √ = (1 − δp ) → ǫ = 1 + ǫ2 Nonideal Filters Filter Design Butterworth Filter q 1 1 + ǫ2 CN2 ( ωωps ) cosh Ncosh−1 ≤ δs ωs ωp → ≥ Chebyshev Fil. Frequency Transformation Equalization 1 −1 (1 − δp )2 Evaluating |H(jω)| at ω = ωs , and inserting ǫ, remember that we perform better at stopband Introduction Distortionless Transmission s v u u δ12 − 1 ωs CN ≥ t 1s ωp −1 (1−δp )2 | {z } v u u t >1 1 −1 δs2 1 − (1−δp )2 cosh−1 N ≥ s cosh−1 1 1 −1 δs2 1 −1 (1−δp )2 ωs ωp 56 / 89 Alternative Way in Determining Filter Order ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Defining N of Chebyshev requires cosh−1 , which may not be available Alternative way: 1 trial and error: increment N until the inequality is fulfilled, drawback? Irma Zakia q Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 2 Equalization 1 + ǫ2 CN2 ( ωωps ) ωs CN ωp ≤ δs ≥ 1 ǫ r 1 −1 δs2 p Use the relation cosh−1 y = ln y + y 2 − 1 , which results in log10 Chebyshev Fil. Frequency Transformation 1 N≥ s 1 −1 δs2 1 −1 (1−δp )2 log10 ωs ωp + + s r ωs ωp 1 −1 δs2 1 −1 (1−δp )2 2 −1 ! −1 ! 57 / 89 Filter Order Formulas of Butterworth and Chebyshev ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. N Butterworth N Chebyshev v u u u log10 t 1 −1 δs2 1 −1 (1−δp )2 ω log10 ω s p v u u u −1 cosh t 1 −1 δs2 1 −1 (1−δp )2 cosh−1 ωωps Frequency Transformation Equalization 58 / 89 LPF Chebyshev Design Example ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization Design analog Chebyshev LPF with acceptable passband ripple of 2 dB at 20 rad/s, stopband attenuation of 35 dB or more at 40 rad/s (1 − δp )2 = 10−2/10 , ωp = 20 rad/s δs2 = 10−35/10 , ωs = 40 rad/s Find ǫ 1 = (1 − δp )2 , ǫ = 0.76478 1 + ǫ2 Find N with trial and error CN ωs ωp ≥ Initial guess, try N = 3, ωs ωp C3 (2) = C4 (2) = 1 ǫ r 1 − 1 = 73.51819 δs2 =2=x 4(2)3 − 3(2) = 26 8(2)4 − 8(2)2 + 1 = 97 So N = 4 → compare to N = 7 Butterworth 59 / 89 LPF Chebyshev Design Example continues ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Transfer function for N = 4 H(s) = (s 2 + b1 ωp s √ 1 ω4 c c 1+ǫ2 p 1 2 + c1 ωp2 )(s 2 + b2 ωp s + c2 ωp2 ) The constants dN = 1 2 ( "r 1+ 1 1 + ǫ2 ǫ #1/N π = 0.92867, 8 π b1 = 2dN sin = 0.20977, 8 c1 = dN2 + cos2 − "r 1+ 1 1 + ǫ2 ǫ #−1/N ) = 0.27408 3π = 0.22157 8 3π b2 = 2dN sin = 0.50643 8 c2 = dN2 + cos2 Butterworth Filter Chebyshev Fil. Frequency Transformation Inserting those values, the transfer function yields H(s) = 2.61513 (s 2 + 4.19540s + 371.46800)(s 2 + 10.12860s + 88.62800) Equalization 60 / 89 Exercise ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia 1 Design Chebyshev LPF that has passband from 0 to 200 Hz with acceptable ripple of 1 dB, and a monotonic stopband that is down at least 40 dB at 250 Hz 2 Repeat for Butterworth 3 Sketch both filters Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 61 / 89 Kuis 5 Sem I 2016/2017 ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters 1 Seorang user menghendaki filter analog band-pass (BPF) dengan spesifikasi: Melewatkan sinyal dengan redaman passband 2 dB pada frekuensi antara 3 KHz dan 5 KHz Menghalangi sinyal di bawah 2 KHz dan di atas 6 KHz dengan redaman setidaknya 35 dB Tidak mengizinkan adanya ripple Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter (a) (b) Tentukan fungsi transfer H(s) Apa filter yang dihasilkan memenuhi spesifikasi user? Jelaskan! Chebyshev Fil. Frequency Transformation Equalization 62 / 89 Presentation Outline ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 1 Introduction 2 Distortionless Transmission 3 Nonideal Filters 4 Filter Design 5 Butterworth Filter 6 Chebyshev Fil. 7 Frequency Transformation Chebyshev Fil. Frequency Transformation 8 Equalization Equalization 63 / 89 What is Frequency Transformation? ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Any analog filters, i.e. LPF, HPF, BPF, BSF is designed by applying freq.transformation from its corresponding Normalized LPF (NLPF) form Filter specs, LPF/HPF: δp , δs , ωp , ωs BPF/BSF: δp , δs , ωp1 , ωp2 , ωs1 , ωs2 Conversion to NLPF specifications NLPF specifications Design of NLPF NLPF transfer function H(S) Frequency transformation Chebyshev Fil. Frequency Transformation Equalization Desired filter transfer function H(s) Note : symbol S is used to distinguish NLPF with the desired analog filter 64 / 89 What is Normalized LPF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters NLPF is LPF filter with normalized passband edge frequency, i.e. ωp of the NLPF is 1 rad/s |H(jω)| 1 (1 − δp ) δs ω Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation 0 1 ωs Finding ωs of the NLPF are simple for LPF and HPF, but tedious for BPF and BSF Equalization 65 / 89 Frequency Transformation for LPF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Replace S→ Here, ωp belongs to LPF Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization s ωp S s 0 1 ωs ∞ 0 ωp ωs ωp ∞ |H(jω)| |H(jω)| 1 (1 − δp ) 1 (1 − δp ) δs δs ω 0 1 ωs (a) NLPF 0 ω p ωp ω s ω (b) LPF 66 / 89 Frequency Transformation for HPF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Replace S→ Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization ωp s S s 0 ∞ j1 jωs ∞ 0 jωp ω j ωp s Remember : |H(jω)| is even |H(jω)| |H(jω)| 1 (1 − δp ) 1 (1 − δp ) δs δs ω 0 1 ωs (a) NLPF 0 ωp ωs ωp ω (b) HPF 67 / 89 Determining ωs of the NLPF Prototype in Designing LPF and HPF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters LPF design S = inserting s = jωs S= S= Thus in LPF design ωs of NLPF = HPF design S = ωp s inserting s = jωs jωs ωp Filter Design Butterworth Filter s ωp ωs ωp ωp jωs Thus in HPF design ωs of NLPF = ωp ωs Chebyshev Fil. Frequency Transformation Equalization 68 / 89 Frequency Transformation for BPF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Passband edges (ωp1 , ωp2 ) and stopband edges (e ωs1 , ω es2 ) are geometric symmetric w.r.t. ω0 (compare to arithmetic symmetric!) Replace S→ s 2 + ω02 Ws |H(jω)| Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation S s 0 ∞ j1 jωs jω0 0, ∞ jωp1 , jωp2 jω es1 , j ω es2 1 (1−δp ) ω0 : BPF center freq. ω02 = ωp1 ωp2 = ω es1 ω es2 W = ωp2 − ωp1 δs 0 ω es1 ωp1 ω0 ωp2 ω ω es2 Any two frequency with identical value of |H(jω)|, satisfy the geometric symmetry property Equalization 69 / 89 Frequency Transformation for BSF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Passband edges and stopband edges are also geometric symmetric Replace S→ Ws s 2 + ω02 |H(jω)| Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation S s 0 ∞ j1 jωs 0, ∞ jω0 jωp1 , jωp2 jω es1 , j ω es2 1 (1−δp ) ω0 : BSF center freq. ω02 = ωp1 ωp2 = ω es1 ω es2 W = ωp2 − ωp1 δs 0 ωp1 ω es1 ω0 ω es2 ωp2 ω Any two frequency with identical value of |H(jω)|, satisfy the geometric symmetry property Equalization 70 / 89 Stopband Edges Adjustment in Designing BPF and BSF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission User specifications: δp , δs , ωp1 , ωp2 , ωs1 , ωs2 In the design process : the passband and stopband edges must be geometric symmetric Need to make adjustment in stopband Since passband is fix ω02 = ωp1 ωp2 6= ωs1 ωs2 There are two cases 1 Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 2 ωp1 ωp2 < ωs1 ωs2 BPF : ω es2 < ωs2 , ω es1 = ωs1 BSF : ω es1 < ωs1 , ω es2 = ωs2 ωp1 ωp2 > ωs1 ωs2 BPF : ω es1 > ωs1 , ω es2 = ωs2 BSF : ω es2 > ωs2 , ω es1 = ωs1 After adjustment ωp1 ωp2 = ω es1 ω es2 = ω02 71 / 89 Determining ωs of the NLPF Prototype in Designing BPF and BSF ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia BPF design S = inserting s = j ω es1 jωs = Introduction Distortionless Transmission Nonideal Filters Filter Design Frequency Transformation Equalization 2 + ω2 −e ωs1 0 Wj ω es1 inserting s = j ω es2 jωs = Butterworth Filter Chebyshev Fil. s 2 +ω02 Ws 2 + ω2 −e ωs2 0 Wj ω es2 BSF design S = (1) ω es2 − ω es1 ωp2 − ωp1 inserting s = j ω es1 jωs = Wj ω es1 2 −e ωs1 + ω02 (3) inserting s = j ω̃s2 (2) Either (1) or (2) leads to ωs = Ws s 2 +ω02 jωs = Wj ω es2 2 −e ωs2 + ω02 (4) Either (3) or (4) leads to ωs = ωp2 − ωp1 ω es2 − ω es1 72 / 89 Example on HPF Design ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Design a Butterworth analog HPF that will pass all signals greater than 27 kHz with 6 dB attenuation, and have a stopband attenuation at least 20 dB for all frequencies below 18 kHz Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 73 / 89 Example on BSF Design ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Design a Chebyshev analog BSF that will reject signals between 4.6 kHz and 7.4 kHz with at least 10 dB attenuation, and have a acceptable passband ripple of 3 dB for frequencies below 2 kHz and above 10 kHz Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 74 / 89 Presentation Outline ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter 1 Introduction 2 Distortionless Transmission 3 Nonideal Filters 4 Filter Design 5 Butterworth Filter 6 Chebyshev Fil. 7 Frequency Transformation Chebyshev Fil. Frequency Transformation 8 Equalization Equalization 75 / 89 Why equalize? ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Recall : Distortionless transmission requires that the system has constant magnitude and linear phase response If not: use equalizer Heq (jω) = 1 1 = e −jarg(H(jω)) H(jω) |H(jω)| Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 76 / 89 Delay equalizer ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter If Htot = H(jω)Heq (jω), then we wish to have Htot = 1 ? It is impractical to construct an equalizer with inverse phase response of the channel (Htot = 1) Instead, make the total phase response to be linear or phase proportional to ω, i.e. Htot ≈ e −jωt0 Delay-equalizer can be achieved by series interconnection with all-pass networks When designing Butterworth/Chebyshev analog filters : are their phases linear? Chebyshev Fil. Frequency Transformation Equalization 77 / 89 Phase Response of Butterworth and Chebyshev Filters ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization (a) 5th order Butterworth (b) 5th order Chebyshev Reference: inst.eecs.berkeley.edu/ ee247/fa04/fa04/lectures/L2 f04.pdf 78 / 89 Further Phase Response of Chebyshev Filters ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction (a) LPF Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization (b) BPF Reference: Y. Su, Group Delay Variations in Microwave Filters and Equalization Methodologies, Master ’s Thesis in Microtechnology and Nanoscience, 2012 79 / 89 Frequency Response of Several Filters ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission (a) magnitude response Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization (b) group delay 80 / 89 Intersymbol Interference (ISI) due to Channel Impairments ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters (a) linear phase response channel (b) nonlinear phase response channel Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Reference: Y. Su, Group Delay Variations in Microwave Filters and Equalization Methodologies, Master ’s Thesis in Microtechnology and Nanoscience, 2012 Equalization 81 / 89 ISI Effect on Data Transmission ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 82 / 89 First-order All-pass Systems ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia s − α0 s + α0 |H(jω)| = 1 H(s) = argH(jω) = −2tan−1 ω α0 Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 83 / 89 Second-order All-pass Systems ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction H(s) = |H(jω)| = 1 argH(jω) = −2tan Distortionless Transmission Nonideal Filters Filter Design s2 − (s − α1 )2 + β12 = s2 + (s + α1 )2 + β12 where ωr = −1 ωr ω Q ωr2 − ω 2 q α12 + β12 , and Q = ωr Qs ωr Qs + ωr2 + ωr2 ωr 2α1 Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization 84 / 89 Before and After Equalization ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design (a) magnitude response (b) group delay Butterworth Filter Chebyshev Fil. Frequency Transformation Reference: Y. Su, Group Delay Variations in Microwave Filters and Equalization Methodologies, Master ’s Thesis in Microtechnology and Nanoscience, 2012 Equalization 85 / 89 After Delay Equalization ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Figure: Group delay of (a)BPF, (b)second-order all-pass network, (c)overall circuit Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Reference: Y. Su, Group Delay Variations in Microwave Filters and Equalization Methodologies, Master ’s Thesis in Microtechnology and Nanoscience, 2012 Equalization 86 / 89 Chebyshev with Allpass ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission (a) magnitude response (b) phase response Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization (c) group delay Reference: inst.eecs.berkeley.edu/ ee247/fa04/fa04/lectures/L2 f04.pdf 87 / 89 UAS Sem 2 2015/2016 ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Chebyshev Fil. Frequency Transformation Equalization Seorang user menghendaki filter analog low-pass (LPF) dengan spesifikasi: Mengizinkan adanya ripple di passband sebesar 3 dB pada frekuensi di bawah 2 kHz Respons monotonic di stopband dengan redaman sedikitnya 10 dB untuk frekuensi di atas 3 kHz (a) Tentukan fungsi transfer H(s) untuk memenuhi spesifikasi tersebut. (b) Apakah filter rancangan Anda memenuhi spesifikasi user ? Jelaskan. (c) Sketsakan respons magnitude (skala linear) LPF dengan memperhatikan bentuk/nilai ripple yang tepat serta titik-titik signifikan pada passband dan stopband. (d) Sketsakan respons fasa (skala linear) dari LPF tersebut. (e) Perkirakan gambar group delay LPF dengan menggunakan hasil sketsa respons fasa serta formulasi group delay = − d arg(H(jω)) dω (f) Jika LPF diinterkoneksi secara seri/cascade dengan filter all-pass, perkirakan gambar group delay filter all-pass yang dibutuhkan sehingga respons fasa sistem setelah diinterkoneksi adalah linear. 88 / 89 Reference ContinuousTime Signal Processing (ET 2004) Chapter 6 : Introduction to Analog Filters Alan V. Oppenheim, Alan S. Willsky, with S. Hamid, Signals and Systems, 2nd edition, Prentice-Hall, 1996. Irma Zakia Introduction Distortionless Transmission Nonideal Filters Filter Design Butterworth Filter Simon Haykin, Barry Van Veen, Signals and Systems, 2nd edition, John Wiley & Sons, Inc., 2004 inst.eecs.berkeley.edu/ ee247/fa04/fa04/lectures/L2 f04.pdf Y. Su, Group Delay Variations in Microwave Filters and Equalization Methodologies, Master ’s Thesis in Microtechnology and Nanoscience, 2012 Chebyshev Fil. Frequency Transformation Equalization 89 / 89