Chapter Active Filters and Oscillators ectrical clements have bean used to make frequency-selective filters since the early part of ‘the 20th century. These early filters, which utilized only passive inductors, capacitors, and resistors, helped foster the development of the first radio transmitters and receivers by providing circuits with frequency-selective capabilities. The limited range and selectivity of passive RLC circuits were improved somewhat by the invention of the vacuum tube, which permitted the desi of filter circuits with feedback. Moder filter design really began with the arival of high-quality integrated-circuit operational amplifies in the early 1960s. Modem filters utilize op-amps in combination with RC feedback networks to provide countless filter functions with a wide range of frequency-selective properties. 'As demonstrated in Chapter 10, te frequency response of an op-amp feedback circuit can ‘be dramatically changed by the addition of capacitor to its feedback network, This property can bbe exploited to produce op-amp circuits with well-defined and controllable frequency-response ‘characteristics, Such circuits are par ofa family of stable analog feedback circuits called active {ftes. An analog feedback circuit tha is intentionally operated outside its stability mits called an oscillator. In this chapter, the characteristics and properties of several active filter and oscillator circuits are examined in detail. The functions performed by these circuits are important in many signal- and information-processing applications. As we shall see, an active op-amp filter can achieve all ofits desired properties without the use of inductors. This result is fortunate, because the inductors needed for filter circuits below about 1 MHz tend to be large, difficult to produce in ideal form, and unsuitable for fabrication on an integrated circuit, Filter circuits made solely from ‘op-amps, resistors, and capacitors are realy fabricated in an integrated-citcuit environment. 13.1 A SIMPLE FIRST-ORDER ACTIVE FILTER [As a prelude to a general discussion of active filters, we first illustrate the basic concepts of active filtering using the circuit of Fig. 13.1(a). This simple filter is a low-pass variety that passes all frequency components below its cutoff frequency and attenuates all frequency components above. (We also recognize this circuit as the modified op-amp integrator of Chapter.) Because ‘we are interested in the behavior of the circuit under sinusoidal steady-state, rather than transient, ‘conditions, the circuit is best analyzed in the frequency domain. oy808 © Chapter13 © Active Filters and Oscillators Figure 13.1 A simple active filter example, (@) Inventing amplifier with feedback “element” Zy= RiZe; (©) equivalent topology ofthe inverting-amplifier configuration Figure 13.2 Magnitude Bode plot ofthe active her of Fig. 13.100). The filters “cuot?” requency is designated wo, a zy =e) ‘The circuit has the same basic topology asthe invertin the parallel combination RoiiC is used as a feedbac capacitor behaves as an impedance element of value Fig. 13.1(@) cam be found by first expressing Ro Ze 'g amplifier of Fig. 13.1(b), but in this case, *K element. In the frequency domain, the Zc = 1/joC. The output of the filter of 8.4 single feedback impedance element of value 1 Re %= RJ =—_% _ 13.1) Ble 1+ joR,C ee By analogy tothe inverting-amplifier topology of Fig. 13.1(b, the ‘Output of the filter becomes ——- (13.2) + oRC ° here 21 = Ri, and where vour and vy have been represented in sinusoidal phasor form as Vey and Vir, “© (og scale) ‘The transfer function (13.2) has a single pole at a, below a. ‘The magnitude Bode plot ofthis transfer fant ‘Sem ofa single-poe low-pass filter. As the frequency ofthe input signal is increased above the “culofl” frequency c, the filter output decreases at the rate of 20dS per decade, he steauency dependency deseribedby the transfer function (13 year ans be synthesized using the passive RC circuit of Fig. 13.3, for which T/C and a gain of —Ra/R, well ion, shown in Fig. 13.2, has the basic Vox Ze a Va” 4% “Th jome cu ‘The Bode plot of the later circuit's ‘The advantage of the active filter version ofFigure 13.3 Passive RC circuit having the same ‘peneral frequency response as the active filter of| Fig. 13.100. Section 13.2 © Ideal Filter Functions © 809 of the active filter of Fig. 13.1(a) is negligibly small; the op-amp functions as a voltage source that drives the output terminal. This feature allows the active filter to drive a load impedance or another stage in a multistage filter cascade without changing the filer characteristics. In contrast, the output impedance of the passive circuit of Fig. 13.3 is equal to Ry[(1/jwC). This relatively high impedance causes the circuit's output voltage and frequency response to be affected by the characteristics of its load, R In principle, the transfer function of any passive filter can be synthesized in active form to realize the advantages stated above. Additionally, passive filter circuits that would normally require inductors can be made in active form without the use of inductors. As discussed previously, high-quality inductors are difficult to make in both discrete and integrated environments and are usually avoided in modern active circuit design. 13.2 IDEAL FILTER FUNCTIONS Figure 13.4 ‘deal “bick-wall” sponses of a) low-pass; Dyhigh-pass; ©) bandpass; and a) band-reject iter. ‘The low-pass filter function described in Section 13.1 is butone ofa class of analog filter functions that also includes high-pass, band-pass, and band-reject filters. As its name implies, the high-pass filter passes only those frequency components that lie above some designated cutoff frequency. Wout! Nell 4: aitand ] _ Fasbad Stopband Stopband o o ® ° (Low pass (High pass Waal Yl Nou Pastband Pastband Pasband Stopband Stopband ‘Stopband 2 ar ‘o ar Or o (e) Bandpass (@Band-xject810 © Chapter 13 * Active Filters and Oscillators Figure 13.5. Filter function efiniions shown, for (a) low-pass fier and () band-pas iter. “The band-pass filter transmits only those frequency components lying within arange specified by Upper and lower cutoff limits. The band-reject itr is the inverse of the band-pass filter it passen ‘only those frequency components lying outside some specified frequency range. Mou! Vit A ‘Transition region . Af Sophand e, @ - Sa Patek Lover sas g Be pertain exon 7 CLM Fete el Uinta & >» “The basic forms of the transfer function for each of the various filter types are depicted in Fig, 134. ‘These perfect, boxlike plots are sometimes called brick-wall responses. Each one represents an ideal case in which the filter gain remains constant over frequency regions where signal transmission is desired and falls to zero otherwise. Much of filter design is concerned with approaching these ideal responses as closely as possible while remaining within. the practical Constraints of part count, cost limitations, and filter complexity. As an example of this concept, ‘consider the simple low-pass op-amp filer of Fig. 13.1(a). Its ~20-dB per decade rolloff above «, ‘provides only a very crude approximation tothe ideal brick-wall low-pass response of Fig. 13.4(a) ‘The filter is inexpensive and easy to build, however, and is adequate for many applications. A more complex op-amip circuit involving many more components could be constructed to provide ‘a response more closely approaching the ideal, but this choice would result in a larger number of parts, and hence a greater cost per filter. In order to quantify the degree to which any given filter approaches the ideal “brick-wall response, it is helpful to define several quantities related to the filter's response curve, These {quantities are summarized in Fig. 13.5 using the low-pass and band-pass fiters as examples Similar definitions exist forthe high-pass and band-reject filters. The filter's passband is defined asthe frequency region over which signal transmission is desired. The largest response occurring, ‘anywhere within the passband is designated A,. In the ideal case, the filter gain would be equal to A, throughout the passband, A real filter will always have a gain that changes with frequency, hhence the parameter A is used to define the lowest value to which the passband gain can fallFigure 13.5 Filter function definitions shown {or (a) low-pass fier and (band-pass filter. 810 © Chapter13. * Active Filters and Oscillators ‘The band-pass filter transmits only those frequency components lying within a range specified by ‘upper and lower cutoff limits. The band-reject filter isthe inverse ofthe band-pass filter; it passes ‘only those frequency components lying outside some specified frequency range. Wau Vil a pusband |G A @ Lower region By YG rman GLA aaa eer o or © A ‘The basic forms of the transfer function foreach of the various filter types are depicted in Fig. 134. These perfect, boslike plots are sometimes called brick-wail responses. Each one represents an ideal case in which the filter gain remains constant over frequency regions where signal transmission is desired and falls to zero otherwise. Much of filter design is concerned with approaching these ideal responses as closely as possible while remaining within the practical constraints of part count, cost limitations, and filter complexity. As an example of this concept, consider the simple low-pass op-amp filter of Fig. 13.1(a). Its ~20-dB per decade ollof above o, provides only avery crude approximation tothe ideal brick-wall low-pass response of Fig 13.4(). The filter is inexpensive and easy to build, however, and is adequate for many applications. A more complex op-amp circuit involving many more components could be constructed to provide response more closely approaching the ideal, but this choice would result in a larger number of parts, and hence a greater cost per filter. In order to quantify the degre to which any given filter approaches the ideal “brick-wall” response it is helpful to define several quantities related to the filter's response curve. These quantities are summarized in Fig. 13.5 using the low-pass and band-pass filters as examples. Similar definitions exist forthe high-pass and band-reject filters. The filter’s passband is defined as the frequency region over which signal transmission is desired. The largest response occurring anywhere within the passband is designated A,. In the ideal cas, the filter gain would be equal to As throughout the passband. A real filter will always havea gain that changes with frequency, hence the parameter A; is used to define the lowest value to which the passband gain can fallFigure 13.6 ‘wo possible low-pass ter tions. (0) Gin ‘onically as Frequency is nereased;(b) gain rycles between ninimuma and ‘baximum limits ia he passband and topband. Section 13.3 * Second-Order Filter Responses © 811 and still be acceptable tothe designer. Any departure from the ideal of constant passband gain ‘may also be expressed as a maximum acceptable attenuation within the passband, defined by the factor dnax = Ay ~ At. Ian ideal filter could be constructed, its signal transmission would immediately fll to zero outside the passband. In any real filter, some signal transmission always occurs outside the passband. The quantity A2 defines the maximum signal transmission acceptable to the designer ouside of the passband. The frequency at which signal transmission firs falls to A> defines the beginning of the filter's stopband; the region between the passband and stophand is called the transition region. Note that the band-pass filter has two transition regions and two stopbands, ‘Similarly, the band-reject filter has two transition regions and two passbands. In general, the gain of a filter may lie anywhere between the limits Ay and Ay in the passband; similarly, the gain may lie anywhere below the value Aa within the stopband. The plot ‘of Fig. 13.6(a) shows a low-pass filter response that decreases monotonically from its value of Ay ‘at = 0 and reaches the value A, only once before leaving the passband. The filter response of Fig. 13.6(b) cycles between A, and Ay several times within the passband and also cycles between ero and Az within the stopband. The peak passband gain Ay is reached at some frequency other than zero in this second example, Both plots in Fig. 13.6 represent valid low-pass filter ‘responses and reasonable approximations to ideal brick-wall behavior. Each type of response can be produced by an appropriately designed filter circuit. 13.3 SECOND-ORDER FILTER RESPONSES ‘The low-pass filter of Section 13.1 is an example of a first-order filter. Its single pole in the denominator causes the magnitude |Vox./Viq| to fll off as I/a», ot ~20dB/decade, at frequencies wellabove w.. The steep walls ofthe ideat response of Fig, 13.4(a) are only weakly approximated by the —20 dB/decade slope of a first-order filter. A better approximation can be realized by using filters of higher order. The order of a filter is formally defined as the number of poles in the denominator of the transfer function. As a general rule, filters of higher order will have steeper transition region slope(s). The transfer function of a second-order low-pass filter, for example, falls off as 1/0, or at —40.dB/decade, at frequencies well above its poles. Its slope will be twice as steep as that ofa first-order filter, making it a better approximation to the ideal brick- wall response. ‘Transfer functions of even higher order will produce steeper transition-region slopes. In this section, we examine the properties of several second-order filter configurations, In Section 13.4, these filters are used as basic building blocks to synthesize filter of higher order using the technique of cascading.812 © Chapter 13 © Active Filters and Oscillators 13.3.1 The Biquadratic Filter Function ‘The transfer function of a second-order filter can be described in terms of a ratio two of quadratic polynomials (1+ jo/oy)(1 + jojor) C+ Jojo) + jojo) In both the numerator and denominator, the quadratic polynomial has been expressed as the product of two binomials, as in Chapter 9. Ifthe filter is of order 2 or higher, the poles and zeros are generally complex numbers. A transfer function with complex poles and 2er0s is more readily described using the s-plane representation in the sinusoidal steady-state, where s = jax. The plane is defined by a set of real and imaginary axes that are used to plot the real and imaginary ‘omponents of each pole and zero in the system. In the s-plane, the sinusoidal driving frequency of the filter is equivalent to the imaginary-axis variable s = jo. If complex numbers $1, +, 5» are used to describe the poles and zeros, the biquadratic transfer function (13.4) takes on the form. HUo) (13.4) (+ s/o +s/or) _ ,(6+siN(s+5) Hs) = a, Lt ent + s/o) 135) = Arseny tajan) * “Gone os) Equation (13.5) can als be expressed inthe general form ays" + a8 +a 2O™ Ste bys +) a whetethe coefficentsay,---, a, andy, - =, include combinations ofthe poles, +++, se, This ratio of quadratic polynomials is sometimes called the biquadratic transfer function, or simply “biquad.” It can be used to describe virtually any second-order filter by appropriate selection of the @ and b coefficients. The denominator ofthe transfer function describing a second-order filter ‘must introduce a factor of 1/a at high frequencies; this criterion can be met by adjusting the coelficients bo, by, and by in (13,6) so that the s? factor in the denominator dominates at high frequencies. The filter’s overall behavior—that is, whether it will be a low-pass, high-pass, band pass, or band-reject filter—is established by adjusting the numerator coefficients ap, ay, and ap 13.3.2 Second-Order Active Low-Pass Filter Figure 13.7 Second-order active low-pass filter of the Sallen-Key ‘ype. Ifthe coefficients az and ay in Eq. (13.6) are set to zero, the transfer function acquires the form © bast + bis +55 HG) 3.)Section 13.3 © Second-Order Filter Responses ° 813 We recognize this function as that of a second-order low-pass filter, At frequencies near jw * 0, the response approaches the constant value H(3) = ao/by. At very high frequencies, the response approaches the limit H(s) = an/bys?. Because s = jo, this limit falls off as 1/oo? with a slope of —40,dB/decade, as required of a second-order low-pass filter. One filter circuit that has a transfer function of the form (13.7) is shown in Fig. 13.7. The circuit is sometimes called a Sallen-Key filter after its original inventors." Its output as a function of frequency can be found by direct analysis using KVL and KCL. Applying KCL to the v; node with all currents represented as phasors yields h=h+h (13.8) 1 the impedance of each capacitors represented by Zy = 1/jenCy, Bq (13.8) can be expressed * Va=Vi_ Vi Vi~Vou Rh Rent h Ce ‘The op-amp voltage v, canbe found in terms ofV from the complex form ofthe voltage divider: iy 2 YiaVig 13.10) ‘The op-amp output is connected directly to the v_ terminal, thereby forming a voltage follower between vs, v-, and vou. This connection forces your to have the same value as ',, s0 that Eq. (13.10) becomes ve ay Rearranging Ea (13.1) results in Ret2r Vi = Vee Pt ; 1 = Vou: Z (13.12) ‘Combining Eqs. (13.9) and (13.12) leads to an expression for Vow as a function of Via: Vin Vou Ro+Z, Vow R Lh TDs “ Cee) Equation (13.13) can be solved for Vax resulting in Rth 1, Rm) Va a Dik, Ry : Bike Vout = Ving ge 22 13.5 * Tila + TAR + RF RR ae Substation of 1/j0C; and 1/j0C, for Zy and Za in Ea (13.15) results in You 1 Go) 03.16 Vin @RIR2CiC2) + joCa(Ry + Ra) possible to factor the denominator of this frequency-dependent transfer function into the standard “product of binomials” form of Chapter9. For all but a few values of Ri, Ray Ci, T RP Sallen and EL Ke, “A Pracical Method of Designing RC Active Fiers” JRE Transactions on Circuit Theory Vol. CT-2, 74-85, March 195© Chapter 13 © Active Filters and Oscillators and C2, this factoring reveals poles in the denominator that are complex numbers. The transfer function (13.16) can also be represented as H7(3) in the s-plane ifthe following substitutions are made: a3.7) : (03.18) ©" TERE oe weRRCCe _ [EL ER) and SER _ fC (RR 3.19 o- Oa + Re) va (4% a Note thatthe parameter Q, called the “quality factor” ofthe filter, is dimensionless. By using these substitutions, Eg. (13.16) can be expressed as Vet 1 wo 4O" a" Teale tia8 FF aa, Draw On the right-hand side of Eq, (13.20), the denominator has been factored into two complex bino- rials (8 ~ s1) and (s — 2), where 3.21) and (13.22) ‘We recognize Bq. (13.20) as a biquad transfer function in which the s? and s coefficients in the ‘numerator are set to zero. This feature causes the response to be unity at de (s = 0) and to fall off as I/s? at high frequencies. As the expressions (13.21) and (13.22) indicate, sy ands, are complex conjugates with equal real parts and with imaginary parts of the same magnitude but opposite sign. ‘The factors sy and s> represent the poles ofthe transfer function (13.16). For the case where the Rand C values yield a Q les than 0.5, the factor in brackets in Eqs, (13.21) and (13.22) will be positive, so thats) and s2 will be real and Eq. (13.20) can be written in the form pC Vie (Go/any+ MGo/oy+ 1 ~ 1+ jason + Jaron (13.23) where 01 = —sy and an = ~s2, Equation (13.23) is produced by dividing the numerator and denominator onthe right-hand side of Eg. (13.20) by sys. The right-hand side of Eg (14.28) vie the standard product of binomials form of Chapter9, wherein the frequency response is deseibog by two simple, teal poles at « and a. We recognize this transfer function as that of alow pore Site of second order. At frequencies well below cand an, its gain is unity. At frequencies nell shove o1 and, its gain als off a5 1/4, tha is, at 40 dB per decade in frequen, 1Q = 0.5 exactly, the factor inside the brackets in Eqs, (13.21) and (13.22) becomes zo, For this case, the poles of the filter coincide at, reducing the transfer function vo Va 0+ jojo.figure 13.8 Jocation of poles; 8 for @ > 0.5 fie er Hho low-pass ter Inseruncion i320. Section 13.3 © Second-Order Filter Responses ° 815 If Qs larger than 0.5, the square-root terms in Eqs. (13.21) and (13.22) become imaginary, and the poles become complex-conjugate numbers s) and s2 with real part equal to sr 43.25) and imaginary parts equal to £jsy = jlo} ~ (@6/2Q)?|"? = 4 jlo} ~ 39)! (13.26) ‘These complex poles s1 and sp are located to the left of the imaginary s-axis in the s-plane, as shown in Fig. 13.8. Their placement in the left-half plane results because the s given by Eq. (13.25) is negative. The radial distance d from the origin to each of the poles s1 and s> in Fig. 13.8is given by d= (sr + jsr| = (0/20) + a} — (wo/2O) 1" = wy ‘As this equation shows, the poles sy and s, when complex, lion a circle of radius o, at an angle determined by the value of Q. The real and imaginary parts sx and sy are not independent. For a given wo, specifying sz automatically specifies Q and sy. ‘When the poles; and s2 are imaginary, the magnitude of the transfer funetion most be expressed as 327) 2 a 1328) Basal 2 IHG) = sa=- 0,220) sy=#[02-(@,207)"816 © Chapter 13 © Active Filters and Oscillators Figure 13.9 ns Pate ovthe lowpass filer of Fig 3 inte plane. The magne of VeVi proportion othe reeproeal of the proud wheres = je. Inthe s-plane representation, the magnitudes fs—si| and fs ~ sare determined by the distances between the poles s ad y and the location s= jo onthe imaginary s-axis. As depicted in Fig. 13.9, the magnitude [Vex /Val becomes Ve oe oe = [Yet] =“ __ 13.29) WOl= lO] emademal = dias (Be where dy and dy are the lengths of vectors dy and , respectively For small frequencies, such tat the diving point s = jo in Fig. 13.8 is located near the origi, the vectors dy and dy have approximately the same length and Ba, (1329) yields [Vout Vinl~ 1. This situation is depicted in Fig. 13.10¢a. Figure 13.10 Co Lonepase fier tafe function (132 inthe Slane. The Tenth of etors dyand dy ae Show at tee Siren diving frequencies @e=0 (0 0.5: @ ond; W0 ‘The transfer function (13.37) can again be represented in the s-plane by making the following substitutions: 5 (1338) = RE (13.38) and o= ee (13.40) With these substitutions, Eg, (13.37) becomes 1) = Ye 7 . asap Vin ¥8(@,/0) +03 sss) The roots of the denominator of this expression are again given by Eqs. (13.21) and (13.2, respectively, and Eq. (13.41) again has the form of a biquadratic transfer function, In this case, ‘the numerator consists ofa single factor of s*. Athigh frequencies, the denominator approaches a limit consisting of a single factor s, but this factors canceled by the factor of sin the numerator. Hence |Vou/Vial approaches a limit of unity gain at high frequencies. As the driving frequency is reduced well below ap, the denominator in (13.41) approaches the constant value w?. In this ease, the factor of s? in the numerator causes [Vou /Vin 0 fall toward zero atthe rate of 40 dB/decade nis} ® » © 41=[aR- (0,208) sy=- 0,20 ‘The magnitude of Voxx/Vig as a function of input frequency can be determined by again ‘examining the vectors dj and din the s-plane, as in Fig. 13.13. An additional vector ds, which extends from the location of the driving frequency jw to the origin, is needed to represent the factor of s* in the numerator of Eq. (13.41). This vector has a length dy = @. When jw lies near zero, asin Fig. 13.13(a) vectors dy and da have nearly the same length, and ds has approximately zero length. The magnitude of the transfer function (13.41) inthe limit jeo ~ 0 thus becomes |Neu] a fi2|-Z x0 asa)Figure 13.14 Magnitude plot of the second-order hips iter taster fueton (3a) forseveat valves of Q. When = 1v2— 0707, te pots sa be imal fa Section 13.3 © Second-Order Filter Responses ° 821 where di, d, and ds represent the lengths of d;, da, and ds, respectively. If jw is increased above zero, dz will increase by about the same amount that di decreases, as in the low-pass filter cas. Were it not fords, this relationship would again keep [Vou:/Viel constant for @ < sy. The length ds increases with joo, however, so that [Vow /Vinl for the high-pass filter increases as the square, of. As « approaches sy, reaches a minimum. Well above « = sy, the lengths dj, dz, and ds all approach the same value, so that the magnitude of [Vau:/Viel approaches unity. Note that if Sm 0, capacitors C, and C behave as open circuits, and vpy is effectively disconnected from the op-amp terminals. The remaining de portion of the circuit has the form of a follower with zero input. In the limit jo ~» 00, capacitors C and C2 behave as short circuits, and the circuit functions as an inverting amplifier with a feedback impedance of zeto. Because the gain of an op-amp inverter is proportional to its feedback impedance, the resulting output equals zero regardless of input, G —— m — | & ff om HK +> a Lo vee ‘A mathematical expression forthe transfer function of this circuit can be derived from KVL ‘and KCL. The op-amp, C2, and Re form an inverting amplifier between vou and vs. The voltage ‘our ean thus be expressed in the frequency domain by «a3.44) where Za = 1/oC. Both Vis and Vay, contribute to V1; hence V; ean be found using superpo- sition and the complex form of the voltage divider. Alternately setting Vou. and Vig to zero with +. assumed to be at virtual ground potential yields a value for Vi: LALA RiliZa Vi =Ve Vou Athen 1249) TON R + tilda * "2+ Ria a Substituting Bq. (13.45) for Vj into Eq. (13.44) results in Rr Zi|'Zo RiliZa Vo = FE (Ving OE 4 Vaya 13.46) te ( Ret Lilda YF Ry iz) a Equation (13.46) can be simplified by substituting appropriate expressions for each parallel com- bination of impedances and by moving the factor of Ra/Za inside the parentheses: ZR) +22) Ra /(Ri + %2) Vor = — [Ving RMS 2a) — Rael Rs ¢ ta) _ [* B+ LsLa/@i +2) * "D+ Rida Ri + ral (3.47) ~ (Yamane Nentts) F R22 + ZioSection 13.3 © Second-Order Filter Responses * 823 Solving Bq, (13.47) for Vou results, after some algebra, in Re Vou = Ri + Rida + Lil + RR" (1348) Substituting 1/je0Cy for Z and 1/JeoC2 for Zz into Eq. (13.48) produces the desired transfer function: oe =Ra/jacs Vin RifjoCr + RifioG, + (/foC VU * RR — 43. 40) =Riacs) FOR (C+) +1 + GaC FOCI RRs Dividing the numerator and denominator of Eq, (13.49) by RyRaCiC2 and expressing jo as 8 results in = JOR2C2/RiR2CiCr —w} R2Cas TET OEGIAT ~ FP 5(0,/0) +08 (13.50) 1 where a 4351) VRRCG and st e( ) (13.52) RUG +C) VR +e, ~ Equation (13.50) has the form ofa biquad transfer function in which the numerator contains a single factor of s. As the frequency is reduced well below a, the denominator approaches a constant value of «2. The factor of s in the numerator thus causes |Vou/ Vi 0 fall toward zer0 at the rate of 20 dB/decade asthe frequency is reduced. As he frequency is increased well above ‘@o, the denominator of H(s) approaches the limit s*. The factor of s in the numerator cancels one factorof sin the denominator, leaving a single factor of sin the denominator thatcauses|Vou/Vil to be reduced as 1/s i atthe rate of ~20dB/decade) as the frequency is increased. At the passband center frequency a, the magnitude [Vax /Viq| can be found by substituting «© = cy into Bq, (13.49): tel -| ~josRaCr ag 1 + C2) = oR RC Vin Lanny ~ [FaeBNCLFG)+ 1 ~ FRG eo _ RC “RC + Oa) LV RVRGCVCS. Given the expression (13:52) for 0, Fa, (1353) becomes, [Vos Ris Ven 1358) [Vin ¥ Rc? ty As Eq. (13.54) suggests, [Vosr/Vial at @ = a is dependent on Q. Analysis of Bq, (13.50) in the plane yields the magnitude plots of Fig. 13.16, shown for several values of Q and values of R ‘and C such that R2C2/ iC = 100.824 © Chapter13 © Active Filters and Oscillators Figure 13.16 ‘Magnitude plot of the band-pass filter transfer function (13.50) for several values of Q. The components Ry, Ci, Rs, and Cy have ‘been chosen so that RaCo/RiCy = 100, ‘The locations ofthe passband limits and o; re shown! forthe case Q=001 Figure 13.17 ‘The splane representation of the poles ofthe second-order band-pass filter of Fig 13.15: @e Oe: Gea and the bandwidth ofthe filter forthe specific case Q = 0.01 is illustrated in Fig. 13.16 for large O. analysis ofthe transfer function in the -plane shows the half-power frequencies to be located at «» + «p/20 on either side of wp, as we now show. ‘The denominator of Eq. (13.50) can be factored into two binomials ($ ~ s1) and (s — ) and the magnitudes ofthese binomials represented by the vector dy and dy shown in Fig. 13.17 ‘The vector ds in Fig. 13.17 represents the factor of s in the numerator of Eq. (13.50). As shown inFig. 13.17), the vector length di reaches its minimum value atthe frequency « = sy, If sy is ‘small compared to s7 (i.e, large Q), then (13.88)Section 13.4 ° Active Filter Cascading © 825 [Atthis frequency, the magnitude of Voye/Via, Which is proportional to d3/dyd, reaches its max- imum value, If @ is decreased to the value (s; — 5x), as in Fig, 13.17(b), di will increase by a factor of 2. If sg is small compared to s7, the vector lengths dy and d3 will be decreased only slightly at this new frequency. As a result of these combined effects, |Voux/Viel Will fall by a factor of 1/-V3 from its value at « = 51 * tp, The half-power frequency i, therefore, is equal to 0) = 51 — $e 13.56) "Note from the denominator of Eq. (13.50) that jsql is equal to «9/2. Given the result (13.55), the half-power frequency becomes aXe 55 «337 ‘This same reasoning can be applied to the case a2 = (s1 +5) to yield 2 O55 (13.58) ‘The resulting bandwidth of the filter for the case se < s1 is thus equal to BW =0-o1 = (13.59) 2 Note that a small sg is equivalent to a large Q, since Q = 0/2sp. EXERCISE 139 13.10 5 1311 5 13.42 1313 13.14 138 Derive the result (13.54) by substituting s = joy into Eq. (13.50). Synthesize a second-order band-pass filter with a Q of 100 and a center frequency of 10 kHz. ‘What is the magnitude of the response at «» = «,? What is the bandwidth of the filter? Design a band-pass filter using the specifications of Exercise 13.10 so that the gain is equal to +60B at «2, What is the bandwidth of your new design? ‘By analyzing the transfer function (13.50) in the s-plane, show that the band-pass filter of Fig. 13,15 produces responses with the magnitude plots of Fig. 13.16. Drawtheangle plot £ Vout/Vin a8 function of frequency for the band-pass filter transfer function (43.49) for several values of Q. Write a computer program to help with these calculations. ‘Show that Eq, (13.48) follows from Eg. (13.47). Arrive at the result (13.48) using KCL at the vy node, rather than the superimposed complex voltage-divider expression (13.45). 13.4 ACTIVE FILTER CASCADING ‘The usefulness of active filters becomes apparent when two or more are cascaded together to produce filter transfer functions of increased order or complexity. As mentioned in Section 13.1 the output of any one active op-amp filter stage will appear as a voltage source to the input of, the next stage, so that interstage loading problems are virtually nonexistent. The overall transfer function of an active filter cascade thus will be equal to the simple product of the transfer functions of each of its individual stages. A filter of order higher than 2 is easily synthesized by simply cascading several one- or two-pole filters in series. ‘The poles of each component filter are appropriately chosen such that the desired overall response is achieved. In general, the higher the order of a filter cascade, the more closely its transfer function cean be made to approach one of the ideal “brick-wall” responses of Fig. 13.4. In this section, ‘we examine the techniques required to accomplish this task. Although we shall focus primarily ‘on the low-pass filter, the concepts introduced apply equally well to high-pass, band-pass, and band-reject filters.