Chapter 6 Transfer Functions 6.1. DEFINITION OF A CONTINUOUS SYSTEM TRANSFER FUNCTION [As shown in Chapters 3 and 4, the response of a time-invariant linear system can be separated into two parts: the forced response and the ffee response. This is true for both continuous and discrete systems. We consider continuous transfer functions first, and for single-input, single-output systems ‘only. Equation (4.8) clay illustrates this division for the most general constant-coefficent, lin ordinary differential equation. The forced response includes terms due to inital values uf ofthe input, land the free response depends only on initial conditions 3 on the output If terms due to all initial values, that i, uf and yf, are lumped together, Equation (48) can be written as roe Ene E ale) (ems det a inal alse.) im transform notation, a8 ¥s) (Ean /% 4.8} U8) + (sm eo a ital vaes wf,» ‘The transfer function P(s) of a continuous system is defined as that factor in the equation for ¥(s) ‘multiplying the transform ofthe input U(s). For the system described ahove, the transfer function is dys gfe es ty asa, Pisy= So s'/ Sas Fo Fay {he denominator i the characterise polynomial, and the transform ofthe response may be rewsiten as Y(s) = P(s)U(s) + (Lerms due to a initial values wb, 8) 1 the quantity (Jerms due to all intl values u, yf) is zero the Laplace transform of the output 1a) in response to an input U(3) sen by ¥(5) = P(s)U(s) If the system is at rest prior to application of the input, that is, d4y/dt* = 0, k= 0,1,.... = 1, for <0, then (teems due to al iil values uf, 9 and the output as a function of time (4) is simply the inverse transform of P(s)U(s). Its emphasized that not all transfer functions are rational algebraic expressions. For example, the ‘transfer function of a continuous system including time delays contains terms of the form e "7 (¢, Problem 5.28). The transfer function ofan element representing a pure time delay is P(s) =e", where T is the time delay in units of time Since the formation ofthe output transform Y(s) is purely an algebraic molipicaion of P(s) and UCs) when (terms due to all initial values uf, »f)= 0, the multiplication is commutative; that i, ¥(s) = U(s)P(s) = P(s)U(s) (6.1) CHAP 6) ‘TRANSFER FUNCTIONS, 19 62. PROPERTIES OF A CONTINUOUS SYSTEM TRANSFER FUNCTION ‘The transfer function of a continuous system has several useful properties: 1. It isthe Laplace transform of its impulse response y4(1), 12 0. That is, ifthe input to a system with transfer function P(s) isan impulse and all inital values are zero the transform of the ‘output is PCs), 2. The system transfer function can be determined from the system differential equation by taking the Laplace transform and ignoring all erm arising {rom initial values, The transfer function ‘P(s) is then given by Ys) we) 3. The system differential equation can be obtained from the transfer function by replacing the s variable with the diflerential operator D defined by D = d/dh. 4. ‘The stability of a time-invariant linear system can be determined from the characteristic ‘equation (see Chapter §). The denominator ofthe system transfer function is the characterise ‘polynomial. Consequently, for continuous systems, if all the roots of the denominator have negative real pats, the sytem is stable. '5. The roots of the denominator are the system poles and the roots of the numerator are the system zeros (see Chapter 4). The system transfer function can then be specified to within a constant by specifying the system poles and zeros. This constant, usually denoted by K, isthe system galn factor. As was described in Chapter 4, Section 4.11, the system poles and zeros can be represented schematically by a polezero map in the s-plane. 6. If the system transfer function has no poles or zeros with postive real parts, the system is a ‘minimum phase system, Ps EXAMPLE 6.1. Consider the system with the diferent equation b/41+ 2 = du/d + ‘The Laplace uansform version of ts equation with all inital values tt equal to ero is (e+ 2)¥(8) = (or DUG) ‘The system transfer function is thus given by P(s)= ¥(8)/U(8)=(8 +13 +2), EXAMPLE 62. Given P(s) = (25+ 1)/(s? +24 1), the system dierent equation is eye Be yt waa [23 , EXAMPLE 6. The transfer function P(s) = K(s + a)/(s+ \s+ c) can be specified by sving the zero location ~a, the pole locations ~B and ~c and the gain factor K: 63. TRANSFER FUNCTIONS OF CONTINUOUS CONTROL SYSTEM COMPENSATORS AND CONTROLLERS ‘The transfer functions of four common contol system components are presented below. Typical ‘mechanizations of thre of these transfer functions, using R-C networks, ae presented in the solved problems. EXAMPLE 6.4. ‘The general transfer function of «conto system lead compensator boa (62) ‘This compensator has a zero ats and a pole ats =~130 "TRANSFER FUNCTIONS CHAP 6 EXAMPLE 6.5. The general ansfer function ofa contimous system lag compensator is as) Peal) SE poe 6 wal Faray (63) However, in this case the zero at ¢= ~B and the pole i at 1 = the way is usualy mechanized (Problem 613) 4. The gin factor 2/6 is included because of (6) ‘This compensator has two 205 and two poles For mechanization considerations, the resticton abs = has is sully imposed (Problem 614) EXAMPLE 6.7. The ransfer function ofthe PID comtaller of Example 214 js (aye bel gga Be Kot Re Ke roe gE) Ket Ker 7 “This controler has two zeros ad one pole Is similar othe lag dead compensator ofthe previous example exept that ie sale ple ate oka an neta) ant des not have the cond pole It ply mechan) in an analog odigtl computer 9) 64 CONTINUOUS SYSTEM TIME RESPONSE ‘The Laplace transform of the response of a continuous system to a specific input is given by Ys) = P(s\U(s) when all intial conditions are zero. The inverse transform y(1) = "{P(s)U(s)] is then the time response and (1) may be determined by finding the poles of P(=)Li(s) and evaluating the residues at these poles (when there are no multiple poles). Therefore y(+) depends on both the poles and zeros of the transfer function and the poles and zeros of the input. The residues can be determined graphically from a pole-zero map of Y(s), constructed from the polezero map of P(s) by simply adding the poles and zeros of U(s). Graphical evaluation of the residues may then be performed as described in Chapter 4, Section 412. 65. CONTINUOUS SYSTEM FREQUENCY RESPONSE, ‘The steady state response of a continuous system to sinusoidal inputs can be determined from the system transfer function. For the special case ofa step function input of amplitude A. often called a de. input, the Laplace transform of the system output is given by Ve) = ry If the system is stable, the steady state response isa step function of amplitude AP (0), since this is the residue atthe input pole. The amplitude of the input signal is thus multiplied by P(O) to determine the amplitude of the output. P(0) is therefore the dc. gain of the system. "Note that for an unstable system such a5 an integrator (P(s) = 1/s), steady state response does not always exist If the input to an integrator is a step function, the output is a ramp, which is ‘unbounded (see Problems 5.7 and 5.8). For this reason, integrators are sometimes said to have infinite de. gain. ‘The steady state response of a stable system to an input w= A sinar is given by = A|PJ))sin(ot +9) magnitude of P(jo), $= arg P( jo), and the complex number P( js) is determined where [PC a a CHAP. 6) "TRANSFER FUNCTIONS a from P(3) by replacing s by ju (see Problem 6.20). The system output has the same frequency as the input and can be obtained by multiplying the magnitude of the input by |P( a) and shifting the phase angle ofthe input by arg P( js), The magnitude |P(j)] and angle arg Ps) forall w together define the system frequency response. The magnitude |P( jo) isthe gain of the system for sinusoidal inputs with frequency «, “The system frequency response can be determined graphically in the splane from a pole-zero map cof P(s) in the same manner as the graphieal calculation of residues. In this instance, however, the ‘magnitude and phase angle of P(s) are computed at a point on the jw axis by measuring the ‘magnitudes and angles ofthe vectors drawn from the poles and zeros of P(s) tothe point on the jw [EXAMPLE 6.8. Consider the system withthe transfer function 1 Gey Referring to Fig. 61, the magnitude and angle of P(e) for w= 1 are computed inthe plane as follows. The magnitude of PC) i Pa) =0n6 1 ODI ey Fig. 61 ag Pil) = 266° 48° = 16° The system frequency response is usally represented by two gzaphs (sce Fig, 6-2): one of P(je)] asa function of wand one of arg PC jo) a6 a funedon of w. For the wansfer function of Example 6%, P(s)= 1/(s+ e+ 2, these graphs are easly deternined by posting the values of |P(ju)] and arg PCyw) for several valves of was shown Blom, @ a os 10 7 wo Puen | os oa 0316 oust 0st args) | 0 | we | ne [tas [oie Pie) 7 - . te) Fig, 62rey "TRANSFER FUNCTIONS [cur 6 66 DISCRETE-TIME SYSTEM TRANSFER FUNCTIONS, COMPENSATORS AND TIME RESPONSES. “The transfer function (2) fora discrete system is defined as that fcr in the equation fot the ansform ofthe output Y() hat tulips te transform ofthe input Us) I al terms de inal conditions ae re, then the stem response to an input U2) gen by: Ye) = PCe}UCa) the somain and (0) = =-"1P( CEC] ate tne dma “Te transfer uncon ofa drt System has the flloming propre: 1. P(2)is he z-ansform of is Kronecker dea pons (k= 0,1 2. The system dfleene equation can be obtained fom (2) by replacing the = variable wth he {hi operator 2 defined for any integers kd n By Z[y(K)) 9k +) (60) 3. The denominator of P(2) she system characteristic pobomil, Consequential the oot af the denominator are within the unit cle ofthe plane, te System sale 4. The roots ofthe denominaloe are sytem poles and the oo of the numerator are the system zzres P(:) cn be spose by Specying the system poles and ers and te gain factor K(etaylets)- (ee play Reade tte) (2455 (Crp e) FP) The sytem poles and ats canbe represented schematically bya pole ro map inthe zplane “he pole eo tap ofthe cup response canbe conatd rom the ple ero map of PC) by including the poles and zeros of the input U(z). 5. The order of the denominator polynomial of the transfer function of a causal (physically. relcabe)dcetetime stm mas Be peter than or equa wo the oder of the umeaor polynomial. 6 The steady sate rexpone of dcetetime syle 1 unit step np is called the gala ‘and is given by the Final Value Theorem (Section 4.9): (67) tim »( Feo a]--m (68) EXAMPLE 6.10. Consider a discrete-time system characterized by the difference equation 42) + LAy( 1) +039(K) = (K+ 2) +0240 K +1) “The sransform version ofthis equation with ll inti conditons Set equal to 20s (41424039 1(2) = (2440229002) “The system tafe function i given by (2402) 32402) Peiaeso3 ” rosy roa ‘This system has a zero at —0.2 and evo poles, at ~0$ and —06. Since the poles are inside the unit crt, the system is stable The de. gin is >) 142) OT EXAMPLE 6.11, ‘The general transfer function of 3 dig lead compensator Kuea(?4) ® “This compensator has aro at 22, and a poe at 2 =. Its steady state gains aR (69) Pamil 2) (6.10) cHar. 6) "TRANSFER FUNCTIONS 13 ‘The gain factor Kio is included inthe transfer fasction to ad robles 1213, for example, Kay is chosen to render the steady analog counterpart. its prin ata given w to 2 deste value tn tga of Pgs (at = 0) equal to that of [EXAMPLE 6.12. The general ransfer function ofa digital lg compensator i @-G-2) G=s)G=2) “Tis compensator has a zero at 2~ 5 anda pole at zp. The gain factor (1 p.)/l 2) inched so tat the low fegueney or steady state gain PI) ~1, analogous tothe continuous-time lag compensator. Pugl2) 25R (1) EXAMPLE 6.13. Digit! lag and lead compensators can be designed directly from domain specifications by ‘sing the transform between the - and 2omains defined by 2 =, Thats the pokes and zeros of tha ater) Pale Pa Fray an be mapped according fo 2~6"T. For the lead compenstor, the 2er0 at $= —a maps into the x0 at Fez,met and the pleat 2~ —b maps ino the pole at == p, = ¢ "7. This gives (6.2) Similaly, (619) ‘This transformation is only one of many possible fo digital lead and lag compensators, or any type of compensators for that matte. Another variant of the lead compensator is illustrated in Problems 12.13 through 12.15, ‘An example of how Equation (613) can be used in applications is given in Example 127. 67 DISCRETE-TIME SYSTEM FREQUENCY RESPONSE ‘The steady state response to an input sequence (u(k)~ A sin okT) of a stable discrete-time system with transfer function P(z) is given by syn AlP(e)bin(wkT +9) k= 041, (6.14) where [P(e’T)] is the magnitude of P(e"), $= arg P(e"), and the complex function P(e) is determined from P(z) by replacing 2 by e”” (se Problem 6.40). The system output isa sequence of samples of a sinusoid with the same frequency asthe input sinusoid. The output sequence is obtained by multiplying the magnitude 4 ofthe input by [P(e*T)] and shifting the phase angle of the input by arg P(e"). The magnitude [P(e/")| and phase angle arg P(e’), for all w, together define the Aiserte-time system frequency response function. The magnitude |P(e"")] is the gain of the system for sinusoidal inputs with angular frequency «. AA discrete-time system frequency response function can be determined in the zplane from a ppole-zero map of P(z)in the same manner asthe graphical calculation of residues (Section 412). In this instance, however, the magnitude and phase angle are computed on the e/ circle (the unit circle), by ‘measuring the magnitude and angle of the vectors drawn from the poles and zeros of P to the point on the unit circle, Since P(e”) is periodi in w, with period 2"/T, the frequency response function need only be determined over the angular frequency range ~2/T (ee Defion 43). The stant of this sequences Y=) = EF. gH(KTYE (ef ton a) wth assum an oem Galcanbe sites Le pesne. 1 edn ag ft me forthe region of convergence >, This relinsip between the Laplace wansform and the Santa can bc cand pplton of Cac ape i) Homie oes ik Ssuay mo neces tose the compl ams approach “The contnsourne uncon Nina {Yt} ca be determined fom Ys) and a ble of Lape: wasfoma andthe ime vale hn Fepacel bY AT prong the th men ofthe Sele seen a (as) VOT) =2 (VN ocr ‘Then the stransform of the sequence (AT), = 0.1.2... is generated by referring to a tale of transforms, which Yields the desired result (kT) == (2 MYM eer) (6.16) ‘Thus, in Equation (416), the symbolic operations ' and = represent straightforward table lookups. and [jou generates the sequence to be z-transformed. ‘A common combination of discrete-time and continuous-time elements and signals is shown in Fig 63. “0 ney Toy ey Tle) Fig. 63 If the hold cireit is zero-order hold. then as shown in Problem 6.42 the discrete-time transfer function from U*(:) to ¥*(2) i given by sPis){ | oy eur In practice, the sampler at the output. generating #*(1) in Fig. 6:3. may not enist. However it is Sometimes convenient to asume one exists at that point, for purposes of analysis (se, eg. Problem 10.13), When this is done, the sampler is often called a fttios sampler. [If both the input and output of a system lke the one shown in Fig, 6:3 are continuous-time signal, and the input is subsequently sampled, then Equation (6.17) generates a discrete-time transfer function (6.17) CHAP. 6} ‘TRANSFER FUNCTIONS. 135 which relates the input at the sampling times 7,27... to the output at the same sampling times. However, this discrete-time system transfer function does not relate input and output signals at times + ‘enween sampling times, thats, for KT <1 <(k+ 1)7, k=0,1,2, EXAMPLE 6.14. In Fig. 63, ifthe bold circuit isa zroonder bold and P(s) = 1/(4 +1), the from Equation (617), the discrete-time eager function of the miaed-leent subsystem i (etesnll..} bleu) 20-292 (1) -e Mhar) 2 (Mar) -e") ole Mary) 2 fe] sor yefe(t Solved Problems ‘TRANSFER FUNCTION DEFINITIONS 6.1. What isthe transfer function ofa system whose input and output are related by the following diferential equation? ay oe ty SesSeayeusS “Tang the Laplace wansform ofthis equation, ignoring terms duet intial conditions, we obtain Ys) +36Y(s) + 2¥(s) = U(s) +501) ‘Tis equation can be writen as 19-[poG}oo ‘The waster function ofthis system i eer en by a MO" Tied (62. A particular system containing atime delay has the dferential equation (€/at) (0) + 0) = u(t 7), Find the wansfer function of thi system. The Laplace yasfory of the dierent eqn igsig ems doe 1 nial condtons. 1g) Yobme"U) To) aod Ul) a eed ye flown anon a, ch be Sem wet ica Xe) et UG)" T ms)136 ‘TRANSFER FUNCTIONS, [CHAP 6 63. The positon » of a moving object of constant mass M is related to the total force f applied to the objet by the differential equation M(d*y/dr®) = f. Determine the transfer function flating the postion to the applied force. Taking the Laplace wansform of the differental equation, we obiaia Me?¥(s)= Fla. The wanser function relating ¥(5) 0 FCs) is therefore Ps) = ¥(s)/F(5) = 1/ 64. A motor connected to a load with inertia J and viscous friction B produces a torque proportional to the input current i. If the differential equation for the motor and load is (a8 /di°) + B(d8/dt) = Ki determine the transfer function between the input current ' and ‘the shaft position & ‘The Laplcetransorm version ofthe diferent equation i (Jt? + 82) transfer function is Pls) = 8(8)/1(s) = K/s(Js +B) = Ks), and the requieed PROPERTIES OF TRANSFER FUNCTIONS, ‘65. An impulse is appli a the input of «continuous system and the output is observed t be the time function ¢~*, Find the transfer function of this system, ‘The transfer function is P(s) = Y(0)/U(s) and Us) 1 for w(t) = 8(2). Theretore L oy V9= 5 66, The impulse response of a certain continuous sytem isthe sinusoidal signal sin. Determine the system transfer function and diferentil equation The system anser funtion i the Laplace tapsform of ts impulse response, P(s)~1/(s? + 1) Then P(D)= y/wm AD +1), Diy +y mu or diy/de+ ym 6.7. The step response of a given system is»: function of this system? Since the derivative of a step isan nspule(ee Definition 3.17), the impulse response for this system is Dit)mdh/drm jee ge 4 “The Laplace transform of p() isthe desired transfer fneton, Thus boot id oe ReaD ‘Note that an alternative solution would be to compute the Laplace transform of y and then multiply by + to determine P(r), since a malipcation by + im the ndomain is equivalent to diferentiaton in the ‘oe domain, Jett Je te, What is the transfer Ps) nitniti 68. Determine ifthe transfer function P(s) = 2s-+ 1)/(s?-+5+ 1) represents a stable or an unsta- bie system, ‘The characteristic equation ofthe system is obtained by setting the denominator polynomial to zero, that i. «2 11-410, The characterise equation may then be tested ung one ofthe sabity entra met wma © = a Fig. 613 “The diferent equations of the motor mate cic and he inet on are and Ky ae & inte a “Taking the Laplace transform of each equation, ignoring inal conditions, (esky V=Ks8 and K I= (984 BO Solving these eauatons simultaneously forthe ester function between V and 8, we have o- K Kyl Gs = (BT RIL) TaN Is RoR RIL KK itue “TRANSFER FUNCTIONS (char. 629, The back em generated by the armatute circuit ofa dc. machine is proportional tothe angular velocity of its shaft, as noted in the problem above. This principle is utilized in the dc. {tachometer shown schematically in Fig. 6-14, where oy i the voltage generated by the armature, Lis the armature inductance, R, isthe armature resistance, and isthe output voltage. If K, is the proportionality constant between v5 and shaft velocity a6 /dr, that is, v= K,(d8/at), m Show that the steady state response of a stable discrete-time system to an input sequence UK) = ASiMGAT, R= O1,2,.00 i given by Y=A[P(e*T)fin(wkT +9) k=O, where P(2) is the system transfer function. Since te system is Kea, if this results te for A= 1 then is ve for arbiwary values of A. To simplify the arguments, an input w(K) =e”, k= 01,2... wed. By noting that 2, (6.18) ky = eT cotakT + jsinwkT the response of the system to (u(K)) isa complex combination of the sesponses to (coswkT)) and CHAP. 6) 6a, "TRANSFER FUNCTIONS 19 (sinwkT), where the imaginary part tthe response to (sinukT). From Table 42 the e-ransform of renyy ‘Thus the transform ofthe system output Ys) is Y(2) =P) ‘To invert ¥(2), we form the partial esetion expansion of re 1 Ont “This expansion consis of ems de othe ples of P(e) and erm to the pole t+ = e™. Therefore en (2) =2[¥ tems due w potest #(2) + at (yD) = 27 tes de topes of #(2)] + (Ce )e4) ‘ice the syste is sable ef em vanes a become re and Jan Meme =| emp enearen = P(e loos wk +6) + snk +69) where 6 arg Pe), The steady te pone tthe input sink i te imaginary pat of J nolo TbiowkT #8) = 041.2, 30.1.2. Show that, if continuous-time function y(+) with Laplace transform ¥(s) is sampled uniformly with period T, the z-transform of the resulting sequence of samples ¥*(2) is related to ¥(3) by Equation (6.15) From Detaition 43: 1 prot yeayge WO~ 35 [Leder ‘where ¢> a Uniformly sampling ys) generates the samples y(AT). k=0, WT) [Me b=... ‘Therefore ‘The transform ofthis sequence i and ater interchanging summation and integration, 1 pesmyeyy $ gates ah Ee m2) Better Eeerrs is pomei series, which converges it e711 <1 thi ct 1 zr ‘The inequality [oF] <1 opie that [s|> Je'7| On the integration contour, fe] = Jel"° = ty130 ‘TRANSFER FUNCTIONS. [cHAP.6 ‘Thus the series converges for |z|> 7. Therefore 1 esr rey [ome for el> €7, mich is Equation (615) Show that if the hol ict in Fig. 63 sa zero-order bod, the discrete-time teansfr function is Biven by Equation (6.17). Let p(s) =¥°"{P(s). Then, wing the consolation integral (Definition 3.2), the output of P(e) can be wetes a Hed = [9 Dayal) de Since s(t is he outpu of zero-order hol, tis constant over each sampling interval. Thus (1) ean be Hd [Fate ar fle nyaiyars fe Myce 2ITLde+ fp AUNT ee hee = 1 Naw sry 8 (fr -n ae) By lg # T= an een for rn den fe" peayan where 1=0,1,2.3.., 1 Now, defining Me) = f:p(8)48 and k= j~1 or jk +1 yields LO" poate [Pay an fe ay anm [pay ad [0 a =A 46 DT] AT] Therefore we em mite staan] = Eatiento yrs) = E ako scr) Using the elaonsipBermen he conolton sm and the poet of sranforms in Seton 49, he Sit Theorem reper 6 Son 49), an he defaiton ofthe sanfrm, he ransom of feast quan 2¥%(2) = H(z) (2) — (2) (2) were (2) 6 the saat of the stems (KT). £041.25 He) i he eaorm of Wiptesde, en Orr and Ce) te tant of (KT), K=O. Rearangg ea els rey *G) (=a) CHAP. 6 649. 650. "TRANSFER FUNCTIONS 131 Ld} Compare the solution in Problem 6.42 with that in Problem 6.37, What is fondamentally different about Problem 6.42, thereby permitting the use of linea frequency domain methods on this problem? ‘The presence of sampler atthe output of P(x) permits the use of + domi transfer functions forthe ‘combination of he sample, zero-order bed, and PCs). Thesis A) = rf), TAO} = Pe and 7 ye[e(2 ‘Supplementary Problems Determine the transfer function of the R-C network shown in Fig 618 Fig. 618 Fig. 619 ‘An equivalent cuit of am electronic amplier is shown in Fig 6-19. What sits wanser function? Find the transfer function ofa system having the impulse response p(t) =e" ~ sn), ‘A sinusoidal input x= 2sin2r is applied to a system with the wanser function P(s)=2/s(5-+ 2). Determine the steady state output J, Find the step response of system having the transfer function P(e) = 4/(22—1Ke? +1 Determine which ofthe following transfer functions represent table systems and which represent unstable stems: we» __42)0=2) PO Caayergy OPO” enue) Gy Oro EeH OME anP e roye ye GFE =I) Use the Final Vale Theorem (Chapter 4) to show thatthe steady sate valve of the output ofa stable system in response toa unit step input i equal to the dc. gain ofthe system.132 651 654, "TRANSFER FUNCTIONS [cuar.6 Determine the transfer function of two of the networks shown in Problem G44 connected in cascade (cere) Examine the iterate forthe transfer functions of two- and thee depre-of-fresdom gyros and compare them ith the oe degree of feedom gyeo of Problem 6:31 Determine the ramp response of system having the taster function P(s) = (2+ 1)/(2 +2) ‘Sow that if system described by for m-s mis tres peor w application of the input, hat is, d!y/d =0, k= 0,1... Io foe F< 0, then (ers de 0 al ail values wy 96) = 0- (Hint: Integrate the diferent equation times fom 0° tim, gat F and then let 1+") Determine the frequency response ofthe ideal zroouder bald (ZOH) device, with tansfer funtion given in Problem 6.6, and sketch the gai and phase characteristics, ‘A zeroonde hold was defined ia Defaition 213 and Example 29. festorder hold maintains the ape of the fuetion defined by the las two values ofthe sampler outpa, unt the next sample time. Determine the “Sseretetime transfer function trom U"=) 19 (2) forthe subiystem in Fig 6, with st-order hold clement, Answers to Supplementary Problems =an, R)RGSt (a+ TR RR, RF P= Gear ea = 0707siagr— 138%) yo dee tees toot (2) and (4) represent stable systems; (a), (e), and (e) represent unstable systems Ao V7 Fe G7RO STE yodnte "ete cHaP.6} Sp uy= 656) oe” ‘TRANSFER FUNCTIONS. Fig. POSS Feu oA Vr e a-ryja(e(22 280 IL. 153Chapter 7 Block Diagram Algebra and Transfer Functions of Systems 74. INTRODUCTION Tis pointed out in Chapters 1 and 2 that the block diagram i a shorthand, graphical representation of a physical system, illustrating the functional relationships among its component. This latter feature permits evaluation of the contributions of the individual elements to the overall performance of the system, In this chapter we fist investigate these relationships in more detail, uiliing the Frequency domain and transfer function concepts developed in preceding chapters, Then we develop methods for reducing complicated block diagrams to manageable forms so that they may he used to predict the overall performance of a system 7.2. REVIEW OF FUNDAMENTALS. In general, a block diagram consists of a specific configuration of four types of elements: blocks, summing points, takeof points, and arrows representing unidirectional signal fo Fig 7 ‘The meaning of each element should be clear from Fig. 7-1, “Time-domain quantities are represented by lowercase lees. EXAMPLE 73 1) for continous signals, and 114.) 0 rk), k= 1,2, for discrete goals Capital leters in this chapter ace used for Laplace transforms, of s-transforms. The argumeat # or: is often suppressed, to simplify the notation, ifthe context i clear, oF ifthe results presented are the ‘same for both Laplace (continuous-time system) and 24 discrete-time system)transfer function domains EXAMPLE 7.2. R= R(s)or R= R(2) ‘The basic feedback contol system configuration presented in Chapter 2is reproduced in Fig. 7-2 with all quantities in abbreviated tansform notation CHAP. 7} BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS 185 N o Fie 72 ‘The quantities G,, Gand H are the transfer functions of the components inthe blocks. They may bbe either Laplace of :-ttansform (aster functions, EXAMPLE 7.3, G,-U/Eor GE. I is important to note that these results apply either to Laplace transform or to z-transform transfer functions, but not necessarily 10 mixed continuous discrete block diagrams that include samplers. Samplers are linear devices, but they are not time-invariant. Therefore they cannot be characterized by an ordinary sdomain transfer function, a defined in Chapter 6, See Problem 7.38 for ‘some exceptions, and Section 68 for a more extensive discussion of mixed continuous discrete systems. 73 BLOCKS IN CASCADE Any finite number of blocks in series may be algebraically combined by multiplication of transfer functions. That is, n components or blocks with wansfer functions G,,Gi...-,0, connected in cascade are equivalent to single element G with a transfer function given by 6=6,-6,-6,--- 6,= 116, (a) “The symbol for multiplication "somite when no confusion resus. EXAMPLE 74, E utope -« 78 Mutilation of transfer functions commmuaticesthat i 66-66, (72) for any 0 j EXAMPLE 75 £ Ga, |“. £ GG, |S Pig 14 Loading effects (interaction of one transfer function upon its neighbor) must be accounted for in the derivation ofthe individual transfer functions before blocks can be cascaded. (See Problem 7.4.)136 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS [CHAP.7 74 CANONICAL FORM OF A FEEDBACK CONTROL SYSTEM ‘The 10 blocks inthe forward path of the feedback system of Fig. 7-2 may be combined. Letting G = G,G,, the resulting configuration is called the canonical form of feedback control system. G and Hare not necessarily unique for a particular system, ‘The following definitions refer to Fig. 7-5. Rie te o # Fig 7-5 Definition 7.1: G = direct transfer functor forward transfer function Definition 7.2: H= feedback transfer function Definition 7.3: GH = loop transfer function = open-loop transfer function Definition 7.4: C/R = closed-loop transer function = control ratio Definition 75: E/R ctuatng signal ratio= eror ratio Definition 76: B/R= primary feedback ratio In the following equations, the ~ sign refers to @ positive feedback system, and the + sign refers to a negative feedback system: c6 cas 2 Ro 1iGH (73) zo R1nGH ial 2 oH ae ns Ro1tGH @s) ‘The denominator of C/R determines the characteristic equation of the system, which is usually determined from 1 + GH = 0 or, equivalently, Paw Now (7) where Dey is the denominator and Noy is the numerator of GH, unless a pole of G cancels a zero of Ht {Gee Problem 7.8), Relations (7) through (7.6) ate valid for both continuous (-domain) and disrete (2-domain) systems. 75. BLOCK DIAGRAM TRANSFORMATION THEOREMS Block diagrams of complicated control systems may be simplified using easily derivable transform: tions. The first important transformation, combining Block in cascade, ha already been presented in Section 7.3. It is epeated for completeness in the chat illustrating the transformation theorems (Fig. 7-6). The letter P is used to represent any transfer function, and W, X,¥,Z denote any transformed signals CHAP.7| BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS 157 ‘rantormation Rewaton ‘Block Diagram ecient Best % in Cascade. y= @pox | 2p ns rar, Pe Combining Blacks | rranldier |» — pxeral x . cal 2) Briminating mi "2 a pi Py Forved Leo z Removing 4 Blk ofp 2O% | “tome Forerd | y = Px 2x B Ey 4) Peedbeck Loop |¥ = Pik Par] % a ceiial Removing a Block kEO-fea) | troma Feedack |¥ = Puce P,x) % Lop P40 tO-E 240-404 [2] summing Points |2 = W*X=¥| % x [| summing Points = Wexey ‘Moving a Summing ‘i 7| "in Abead ot 2 Ss fie : i, ‘Moving a Summing a | "ont Reynts | 2 = eax 7 ‘oe : i Fig. 7-6158 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS [CHAP.7 vol guvalent Block ‘Transformation ‘Block Diag et Moving « Takeott es el —r1t 9| Point Abeadota | ¥ = Px rH Moving a Takeott TT 10| Point Beyond a | ¥ = PX ‘Bec x 2 Moving a Takeott 20: G ii] Pot Aneadeta | 2=x2¥ | x Summing Point Moving a Takeott 1a| PointBeynda | 2 = X2¥ Summing Fo Fig 76 Commued 716 UNITY FEEDBACK SYSTEMS Definition 7.7: A unity feedback system is one in which the primary feedback b is identically equal to the controlled output EXAMPLE 7.8. 171 for Boer, unity feedback system (Fa, 7-7) z z ¢. tI Fig 77 ‘Any feedback system with only linear time-invariant elements can be put into the form of a unity feedback system by using Transformation 5 EXAMPLE 77 Rs, co. ek ba tr LJ Fig 78 (CHAP.7| BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS 159 ‘The characteristic equation forthe unity feedback system, determined from 1 4 G= 0. is DgtNe=0 77) where Dy is the denominator and Nj the numerator of 6. 7.7 SUPERPOSITION OF MULTIPLE INPUTS. Sometimes itis necessary to evaluate system performance when several inputs are simultaneously applied at different points of the system, When multiple inputs are present in a linear sytem, each is treated independently of the others. The output duc to all stimul acting together is found inthe following manner. We assume zero initial conditions, as we sek the system response only 10 inputs. ‘Step 1: Set all inputs except one equal to zero. Step 2: ‘Transform the block diagram to canonical form, using the transformations of Section 7S. ‘Step 3: Calculate the response due to the chosen input acting alone Step 4: Repeat Steps 1 to 3 foreach of the remaining inputs. ‘Step 5: Algebraically add all of the responses (outputs) determined in total output of the system with all inputs acting simeltaneously We reemphasie here thatthe above superposition process is dependent on the system bein linear. ips Ito 4, This sum isthe EXAMPLE 7.8. We determine the output C due to inputs U and R fr Fig. 7-9. Rt alt a c Fig 1-9 Pu vw0 ‘The sytem reduces to ‘Step By Equation (7.3, the output Cy det input R is Cy ~1G,6,/11 + 6,64) Step der Pur R=0. Step at: Put ~ I into block, representing the negative Teedback eet: vy o | Rearrange the block diagram: vs co a160 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS {CHAP.7 Let the ~1 block be absorbed ino the samaing point v4 a G Cm} _ Step 4e: By Equation (7), the output C,, due to input U is Cy=[6,/11 + G0) Step S The total ouput [ee aca © GRU) 18 REDUCTION OF COMPLICATED BLOCK DIAGRAMS, ‘The block diagram ofa practical feedback control system is often quite complicated. It may include several fedback or feedforward loops, and multiple inputs. By means of systematic block diagram reduction, every multiple loop linear feedback system may be reduced to canonical form. The techniques developed in the preceding paragraphs provide the necessary tools. The following general steps may be used asa basic approach in the reduction of complicated block diagrams. Each step refers to specific transformations listed in Fig. 7-6. ‘Step 1: Combine allcaseade blocks using Transformation 1 ‘Step 2 Combine all parallel blocks using Transformation 2. ‘Step 3: Eliminate all minor feedback loops using Transformation 4 ‘Step 4: Shift summing points tothe eft and takeof™ points tothe right of the major loop, using ‘Transformations 7, 10, and 12. ‘Step & Repeat Steps to 4 until the canonical form has been achieved for @ particular input. ‘Step 6: Repeat Steps 1 to 5 foreach input, as required. Transformations 3, 5, 6,8, 9, and 11 ate sometimes useful, and experience with the reduction technique will determine their application, EXAMPLE 7.9, Let us reduce the block diagram (Fig 710) 0 canonical form, a 4 Fig. 740 a G04 CHAP.7| BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS 16 Step: | a a} * G+, Step: . Gas . irae, 2 Re [aor] +O | 1 G6, | my Step 6 Does net apply ‘An occasional requirement of block diagram reduction isthe isolation of a particular block in a feedback or feedforward loop. This may be desirable to more easily examine the effect of a particular block on the overall system. Teolation of a block gen system, but usull others Rearranging Summing Points (Transformation 6) and Transformations 8, 9, and 11 are especially ‘useful for isolating, blocks. lly may be accomplished by applying the same reduction steps to the in a different order. Als, the block to be isolated cannot be combined with any EXAMPLE 7.10. Let us reduce the block diagram of Example 79, isolating block H ‘Steps Land 2 6, —-—{ a, + 616 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS [CHAP.7 ‘We donot apply Sep 3a his tine, but go ict to Step 4, moving takeo point beyond black G+ 6 6,46, | ph 2 A are % ‘We may now rearrange summing points and 2 and combine the cascade blocks in the forward loop using ‘Transformation 6 then Teasformation fas 6,646,+69 | 254 © ty % step: es B09 c reir % ae Fall, we apply Transformation § to remove 1/(G + 6) fom the Fedbck oop ° Ga ©. Gt G TEGO HAGE OS % [Note tha the same result could have been obtained ater apslying Sip 2 by moving takcof point 2 ahead of G+ Gy, instead of theo pois 1 beyond G, + Gy, Block Gy * Gy bas the sae effect om the eotroleatio C/R sthether it directly follows R or diretly precedes C. CHAP. 7] BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS 163 Solved Problems BLOCKS IN CASCADE 7.1. Prove Equation (7.1) for blocks in cascade ‘The block diagram form anser functions 6,,0;,..0, in cascade sven in Fig 7-1. x x x a x x an G - a Fig 7 “The output transform fr any blocks gual tothe inp transform mule by the transfe function (oce Section G1) Thteore My = ¥Gy, My = Nae Me Xe Oe-veyes~ Nib Combining thee uations we ave My KG =X AG Gy = NGG. GG, Dividing both sides by, we obtain My. /X = GG GG 72. Prove the commutativity of blocks in cascade, Equation (72) ‘Consider tmo blocks in cascade (Fig, 7-12) x Kio a Xp 6 Fig 722 From Equation (4) we have X= XG," GX, and X,., XG, G,X,.,. Therefore X= (X66, XG, Dividing both idee by Ny X.4/% = G6, "Aho, X01 "G(G,4) = GG %, Dividing again by X, X,../X,= 6,6, Thus G6, = 6.6, ‘This resi i eXtended by mathematical induction to any fete numberof wanset functions (blocks) in cavade 7A. Find X,/%; foreach of the systems in Fig. 7-13, x [ae] ao x fs fae. @ ati = 1 ost x [0% © re 73 (2), One way to work this problem iso fst write Xin terms of X X-( x ‘Then write X, ia terms of Xy x-(Zls-(Aallzals “Multiplying out an dividing both sides by, e have X,/%, = 10/402 — 1)164 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS (CHAP.7 ‘A shorter method i s fll, We know from Equation (7!) that tmo blocks canbe reduced to cone by simply multiplying their wanserfuncons Also, the taster faction of ingle block i it futputioiaput ransom. Hence x E-(=)laa)- (8) This system has the same transfer function determined in pat (2) because multiplication of transfer allall)-a5 (6), By Equation (7.1), we have Fig 7106 No. If to networks are connected in series (Fig. 7-18) the second loads the ft by drawing curret ftom it Therefore Equation (7-1) canna be dizely applied to the combined sytem. The correc transfer function forthe connected networks is wi/(x?+ Says a8) (se Problem 616), and this is nt equal 10 fasta? z Lk LoL Network 1 meine Network? Fig. 715 CANONICAL FEEDBACK CONTROL SYSTEMS, 7S. Prove Equation (7.3), C/R=G/Q + GH), ‘The equations describing the canonical feedback system are taken desl from Fig, 7-16. They are tiven by E= R¥B, B= HC, and C= GE, Substituting ov into he other, we have C= G(REB) ~6( RE HC) GRE GHC= GR+(FGHC) Subtracting (FHC) fom bot sides, we obtain C+ GHC ™ GR or C/R= G/L + GH). CHAP.7| BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS 165 16. 1. 78. 19. Rit Se © # Fig. 7.416 Prove Equation (74), E/R = 1/(1 + GH). From the preceding problem, we have E= RF B, B= HC, and C= GE. Then E=REHC™ RE HGE, Es GHE~ R,and E/R= 1/0 = Gt). Prove Equation (7.5), B/R= GH/Q + GH), From E= RF B, B= HC, ad C= GE, we obuin B= HGE = HG(R B) = GHR ¥ GHB. ‘Then Bs GHB GHR, B= GHR/L + GH), aod B/R= GH/Al + GI, Prove Equation (76), Day Now= 0. ‘The characteristic equation is vsuly obtained by setting 1 GH =0, (See Problem 79 for an exception) Puting GI = Noy/Dey, We Obtain Dey + Noy =. Determine (a) the loop transfer function, (b) the control ratio, (c) the error ratio, (d) the Drimary feedback rato, (e) the characteristic equation, for the feedback control system in which Ky and K, are constants (Fig. 7-17. (2) Te loop wansfer function is equal to GH | & | Kk Hence GH Kee Sore) ee (B) Tee contol ratio, or losed-lop transfer function is given by Equation (7.3) witha minus sgn for positive feedback): C6 kK Gli” etp— KK) (©) The error ratio aetating sgl rai, is gen by Equation (7.4 eo. ne R-InGH "TR Astp) pK (4) The primary fsdback ratios given by Equation (7. BH KK ROT GH p-KR (€) The characteristic equation i given by the denominator of C/R howe, s+ p~ KK) =O. Ta this ease, 1= GH 54 p~K,K,=0, which is nr the caracterise equation, because the poles of G ‘ances the nro # of166 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS (CHAP.7 BLOCK DIAGRAM TRANSFORMATIONS, 7140, Prove the equi lence of the block diagrams for Transformation 2 (Section 7:5) The equation in the second eoluma, ¥ = Py X's: PX, governs the construction of the block diagram in ‘he third column, as shown. Rewrite this equation as ¥~(P; = P,)X. Th equivalent Bock diagram in the Tas coum is clery the representation of hs form ofthe equation (Fig. 7-18) x Pah r, 7.11. Repeat Problem 7.10 for Rewrite ¥~ PX 4 PX as ¥~(P,/P,)P,X + PX. The block diagram for this form othe equation is clearly piven in Fig 7.18, 7 x Ty Pa Pig. 79 AZ Repeat Problem 7.10 for Transformation 5 We have P= RIX AY] =P P(1/P) NY The bloc digram forthe ater form is given in Fig, 720, Fig. 7.20 7A3,_ Repeat Problem 7.10 for Transformation 7 We have 2= PX £ Y= PLY + (1/P)Y | which yields the block diagram given in Fig, 7-21 z, Fig 721 7.14, Repeat Problem 7.10 for Transformation 8 Werbave Z= P(X + ¥) = PX PY, whos block digram ie clewy given in Fig, 722, CHAP.7| BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS. 16 x pL z x P Fig, 722 UNITY FEEDBACK SYSTEMS TAS. Reduce the block diagram given in Fig. 723 to unity feedback form and find the system Zig characteristic equation. = a z Fig 123 ‘Combining the blocks inthe forward path, we obtain Fig, 724. B+ c me i me Fig. 724 Applying Transformation 5, we have Fig. 725 INF Fig 725, By Equation (77), the characteristic equation for this system is s(s+ IMs +2) +1 = 0 oF 1) 4357+ aesied fab MULTIPLE INPUTS AND OUTPUTS. 6. Determine the output C due to U, Us, and R for Fig. 7-26. 2 Ie Rt G o < a ty ie, Fig. 7.26168 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS [cHar.7 Let Y= U0, Afr combining the cascaded blocks, we obtsin Fig 7-27, where Cis the carp due to R acting alone. Appliee Equation 78) 0 this ssstem, Cy~ [6 6/0 ~ 6, 6G, HH). Billy Fig. 727 [Now lt R= U,~ 0. The block dagram is now given in Fig. 7-28, where C; ithe sponse due 1 Us acting, alone. Rearanging the blocks, we have Fg. 7-29 From Equation (73), we get C\= 16/0 = 6,6, 4,400; A o. uy Hay Fig 728 co & Gt, Fig. 7-29 Finally, let R= U; =0. The bloc diagram ie gven in Fig. 730, where Cis the response Qe to Us acting alos Reaeangog the backs, we et Fig. 7-3, Hence C= (GG, 24/11 ~ 6,0; FH, )1Us ce Us Fig 730 o Gath By Fig 731 CHAP.7] BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS 169 By superposition, the total output is G.G,R+ GL + GEIL CN GG Ht 17. Figure 7-32 is an example of a multinput-multioutput sytem, Determine C, and C; due to, and R, m+ a a @ o & + ee G. Fig. 732 Fin pt the block diagram in the form of Fi, 733, ignoring he output C, og c. points, we get Fig. 7-34 as a cu GGG, Fer Hence the sip 6, do, alos Gy = 6.8L ~ 66,40) Fr Ry = 0, we bv Fi ns * ou a =6.060, & Fens Hence Gy = ~©,6,6.8/0~ 66,6) the ua oC, dR, soe Ths = Gy + i (GR, ~ GGG.R)/~ G66)170 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS (CHAP. 7 Now we reduce the original block diagram, ignoring output C, Fis we obaia Fig 7.36. a oe Cor oe! ® Fig. 726 ‘Then we obtain the block diagram given ia Fig. 7.37. Hence Cy = G Rz/tl ~ 6,6:6)6.). Nex letting, y= 0, we obtain Fig, 7.38. Hence Ce, =~ G,6:64R,/C~ 6,6,6,6,) Finally, G = Cn Gy = (GR ~ GGG I/0~ GO a a Fig 738 BLOCK DIAGRAM REDUCTION 78. Reduce the block diagram given in Fig. 739 to canonical form, and find the output transform C. Kis a constant B40 + 1 ©. a mH a Fig. 7129 Fist we combine the eascade blocks ofthe forward path an apply Transformation 4 to the innermost Feeback oop to cbain Fig. 7-40 Rs ©, Fig 7.40 Equation (7.3) or the reapplication of Transformation 4 yields C= KR/{C + K)s + (1 +014) CHAP. 7] BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS m 749, Reduce the block diagram of Fig. 7-39 to canonical form, isolating block K in the forward loop. My Transformation 9 we can move the takeoff point abead of the 1/(++ 1) block (Fig. 7 oa Fig. 241 Applying Transformations I and 6h we get Fig, 7-42 e+ ae mH #1 Fig 1-42 Now we can apply Transformation 2 to the feedback loops, resting in the final form given Fig 78. B+ 1 6, Fig 743 7.20. Reduce the block diagram given in Fig. 7-44 o open loop form, a Rip 746m BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS [CHAP.7 Fist, moving the leftmost summing point beyond 6, (Transformation §), we obtain Fig. 745 Pee Fig 745 Next, moving takeoff point «beyond G,. we get Fig. 7-46. @}+O—+C @ a t Gut, Fig 146 Now, using Tansformation 6b, and then Transformation 2, to combine the two lower feedhack oops (vom GH, )eatering d and e, we obain Fig. 7-47, & \ 1 ea m a]}+O- a : Caen | Fig. 747 ‘Applying Transformation 4 1 this inner loop, the stems becomes CHAP 7) BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS we ‘Aza, spplvng Transformation 4 1 the remaining feedback loop yilss Finally, Transformation {and 2 give the ope-lop block diagram 26+ 6- oom, oH, MISCELLANEOUS PROBLEMS. aa. 12. 12. Show that simple block diagram Transformation 1 of Section 7.5 (combining blocks in cascade) isnot valid ifthe first block is (or includes) 8 sampler ‘The output transform U*(s) of an ideal sampler was determined in Problem 439 a5 wry~ Eewuery “Taking U°(s) a8 the inp of block F, of Transformation 1 ofthe table the output wansform ¥(s of block plyure) =A) Be ‘Clearly the input vansform X(2)= U(s) cannot be factored from the righthand sde of Y(0), that is Y(0)# Fls)OUe). The same problem occurs iP, incldes other elements, a5 well 353 sampler W)= ‘AT Why isthe characteristic equation invariant under block diagram transformation? ‘Block diagram transformations ate determined by rearranging the input-output equations of one of ‘more ofthe subeystems that makeup the fol syste, Therefor the final transformed systems governod by the same equations, probably arranged n a diferent manner than those forthe ongial ste "Nom, the sharacteite equation t determined from the denominator of the overall system transfer function st equal to aero. Factoring or other rearrangement of the merator and denominator of the system transfer fonction leary does not binge it aor does it ler the denominator Set equa 1020 Prove thatthe transfer function represented by C/R in Equation (7.3) can be approximated by £1/H when |G] or [GH] are very large. Dividing the numerator and denominator of G/( + Git) by G, we get aele L dm | 1 +H Dividing by GH and aking she Limit, we ebtain ela.4 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS (CHAP.7 724. Assume that the characteristics of G change radically or unpredictably during system operation. Using the results of the previous problem, show how the system should be designed so that the ‘output C ean always be predicted reasonably well 1m problem 723 we found that im (S]e2 vole) a ‘Thus C-+ 48/1 as GH) 2, oF € i independent of or Ire GH, Hence te stem sould be designed so that [1 725. Determine the transfer function of the sytem in Fig, 7-48, Then let Hy = 1/6) and Hy= 1/6, B+ + | - a 7 i Fig 148 Reducing te inne lops we hav ig 749. —_ a, © i Fig 78 ‘Aplying Tasfomaton again, we obtain ig. 7-0 » ae © Tar oar Fie 7-50 [Now put, =1/6, and = 1/6, Tis yds c 66, 1 a7 0-0) Gam 726, Show that ig, 7-5 is valid Rs 7 © Fig 751 CHAP.7]__ BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS 175 and Cm From the open-loop diagram, we have C= R/(s +p.) Rearrangng. (5 +p) (1/98 B.C) Te closed oop diner fll fom this equation 121. Prove Fig. 7-82. a [eta]. z Fig 7-52 ‘This problem illustrates how a finite zero may be removed from a block rom the fortardoop diagram, C= B+ (2, — p)R/(s+ pi) Reseraoging. Sona (ETE ae (a 1H =m “This mathematical equivalence clearly proves the equivalence othe block diagrams c= 7.28, Assume that linear approximations inthe form of transfer functions are available foreach block ‘of the Supply and Demand System of Problem 2.13, and that the system can he represented by Fig. 7-53, Hy Rs ie a © Hy Fig, 759 Determine the overall transfer function of the system. Block diagram Transfoemaion 4 applied twice to thi system, gives Fig 756 Tra 7a Fig 754 Hence the wanser fenton foe the nearized Supply and Demand models Cee _ aly TF Gig nT)16 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS (CHAP. 7 Supplementary Problems 729, Determine C/R foreach system in Fig 755 720. va. @ ” © Consider the Blood pressre regulator deserted in Problem 214 Asume the vasomotor enter (YMC) ca be deserbe by Hear transfer faction G(3), abd the baroreceptors bythe transfer function ks ky (Gee Problem 6 3), Transform the blk diagram into is simplest. unity feedback form Rese Fig, 7.56 to canonical frm ty a a Fig. 756 Har. 7} BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS ca 132, Determine C fr the syrem represented by Fig 757 i By +h +0+A— a ©. “Th, Fig. 757 733. Give an example of two feedback systems in canonical frm having identical contol ratios C/R but dierent G and 1 components 734. Determine C/R; fo the system given in Fig 7.88 Bt Fig 758 735, Determine the complete output C, with both inputs R, and R; acting simultaneously forthe system given in the preseding problem, 736 Determine C/R forthe system represented by Fi. 7-59 % a a aby & a £. I m | Fig 759 ‘TST. Determine the characteristic equation foreach ofthe systems of Problems (a) 7.32, (8) 7.3, (c) 736, “138 What block diagram transformation mules inthe table of Seton 75 permit he inlision ofa sampler?178 730. 731 ma. 74. 138, 138 BLOCK DIAGRAM ALGEBRA AND TRANSFER FUNCTIONS OF SYSTEMS |CHAP.7 Answers to Supplementary Problems See Problem 815 Tima] ER Sethe th coe Tear g + on c Toa o GG, +R; 6:8) ~ G61 Re © G+ GH) Be TSG + I+ OGM, GOGR,+ O14 GH)R, Ts Gli 76H, 60.G,H, © aes; R” OF G MIU = GG) = GGA, (@) 1+6,1,+6,6,H,=0 @) 16) (0) + 6.6, + 66) + 66,450. + GH, + GGG", =0 ‘The rests of Problem 721 indicate that any transformation that volves any prod of two of more transform is aot valid i «sampler is isd But all those that simply involve the som or diference of signals aze valid, that &, Transformations 6, 11, and 12. Each eepresens 4 simple earangement of Signals as a Lnearsum, and addition is 4 commutative operation, even for sampled signals that's Zaks yn Ven Chapter 8 Signal Flow Graphs 841. INTRODUCTION The most extensively used graphical representation of a feedback control system is the block diagram, presented in Chapters 2 and 7. In this chapter we consider another model, the signal flow ‘raph. ‘A signal flow graph is a pictorial representation of the simultaneous equations describing a system. 1k graphically displays the transmission of signals through the system, as does the block diagram. But it is easier to draw and therefore eater to manipulate than the block diagram, “The properties of signal low graphs are presented in the next few sections. The remainder of the chapter treats applications. 82. FUNDAMENTALS OF SIGNAL FLOW GRAPHS Let us fist consider the simple equation Kn Ayt, (a1) ‘The variables X, and X, can be functions of time, complex frequency, of any other quanti ‘even be constants. which are “ variables” in the mathematical sense. For signal flow graphs, 4,, is a mathematical operator mapping X, into X,, and is called the transmission funetion, For example, A,, may be a constant, in which case X, is a constant times X, in Equation (81); if X, and X, are functions of s or 2, 4,, may be a transfer function 4,,(s) of 4,2). ‘The signal low graph for Equation (8./) is given in Fig. 8-1. This isthe simplest form of a signal flow graph. Note that the variables X, and X, ae represented by a small dot called a node, and the transmission function 4,, i represented by aline with an arrow, called a branch ‘They may Nae Ay Ngte x Branch x Fig 81 Every variable ina signal low graph is designated by a node, and every transmission function by & branch, Branches are always unidirectional. The arrow denotes the direction of signal flow. EXAMPLE 8.1. Ohm’ law states that E~ RU, where Eis a voltage, I curtent, and R a resistance The flow graph fortis equation i given ia Fig 82. ——____ 4, 7 Fig. 82180 SIGNAL FLOW GRAPHS [cHAP.& 83. SIGNAL FLOW GRAPH ALGEBRA 1. The Addition Rule ‘The value ofthe variable designated by 2 node is equal to the sum of all signals entering the node, In other words, the equation = LAX fs represented by Fig. 83 x Ay 7 Aa Aa 7 x Ay, xy Fig 83 [EXAMPLE 8.2. The signal fw graph forthe equation o line in rectangular coordinates, Y= mX'+ 6, given in Fig 84 Sine 8 he Y-axis intercept, sa constant i may represent abode (arable) o a ransmsson Funston, F x 2 The Transmission Rule ‘The value ofthe variable designated by a node is transmitted on every branch leaving that node. In other words, the equation Xa AyXe DD fed is represented by Fig. 8. Ay 7 Ay ty x A, Ay i, Fig 65 CHAP.) SIGNAL FLOW GRAPHS 18 EXAMPLE 2.3, “Te signal low graph ofthe smltaneous equations Y~ 3X, Z = i ven in Fig 86 3. The Multiplication Rule ‘A cascaded (series) connection of #~1 branches with transmission functions Aay Anny Ausen-y Ages) a8 be teplaced by a single branch with a new transmission function equal to the product ofthe old ones. That is, Kye day Asa Aas Agenay Xs “The signal flow graph equivalence is represented by Fig. 8-7. Fig. 67 EXAMPLE 8.4. The signal flow graph of the simllancous equations Y= 10X, Z= ~20Y is ive ia Fig 88. eaten reduces tog = F Fig 88 ‘84 DEFINITIONS ‘The following terminology is frequently used in signal flow graph theory. The examples associated swith each definition refer to Fig 69. Fig 89 Definivion £1: A paths a continuous, unidirectional suecession of branches along which no node is passed more than once. For example, X, to X; to Xj to Xy, X; to X; and back to Xz,and X; to X10 X, are paths Definition 8.2: An input node or source is 2 node with only outgoing branches. For example, Xj is an input node.182 SIGNAL FLOW GRAPHS [cHars Definition 8.3: An output nade or sink is node with only incoming branches. For example, X, is an output node Definition 84: A forward path isa path from the input node to the output node. For example, X, to X; to X; to X,, and X, 1X, wo Ny ate forward paths, Definition 85: A feedback path or feedback loop is path which originates and terminates on the same node, For example, X; to X, and back to X, is a feedback path Definition £6: seloop is feedback loop consisting of a single branch, For example, ysis a selfloep. Definition 87: The gain of branch is the transmission function of that branch when the ansmission funetion is « multiplicative operator. For example, Ay, isthe gain of the selleop if 44, 18 a constant oF transfer function, Definition 88: The path gain isthe product of the branch gains encountered in traversing a path For example, the path gain of the forward path trom X; 10 X; to %; 10 Xe is AnsAssdas Definition 89: The loop gain is the product of the branch gains ofthe loop. For example, the loop sain ofthe feedback loop from X, to X, and back to Xy 18 AseA3y, Very often, a variable in system is a function of the output variable. The canonical feedback system is an obvious example. In this case, if the signal low graph were to be drawn directly from the equations, the “output node” would require an outgoing branch, contrary 10 the definition. ‘This problem may be remedied by adding a braach with transmission function of unity entering a “dummy” node. For example, the two graphs in Fig. 8-10 are equivalent, and Y, is an output node, Note that ¥,= Ys i vw rs y vk re % Fig 810 85 CONSTRUCTION OF SIGNAL FLOW GRAPHS. ‘The signal flow graph of a linear feedback control system whose components are specified by noninteracting transfer functions can be constructed by direct reference to the Block diagram of the system. Fach variable ofthe block diagram becomes a node and each block becomes & branch, EXAMPLE 85. The block diagram ofthe canonical feedback conto sytem is given in Fig. ay A c CHAP. §} SIGNAL FLOW GRAPHS 183 “The signal Now graph is easly constructed from Fig £12. Note that the ~ or + sign of the summing pot is associated with Fig 812 ‘The signal flow graph ofa system described by a set of simultaneous equations can be constructed in the following general manner. 1, Write the system equations in the form Mad X + Aa Xt HAM Nye AyXitAnXat Asay Nae bea Xt Aga Xe oo tA My ‘An equation for X, is not required if, isan input node. 2. Arrange the mor » (whichever is large) nodes from left o right. The nodes may be rearranged, If the required loops later appear to0 cumbersome. 3. Connect the nodes by the appropriate branches Ay. Ay. et. I the desired output node has outgoing branches, add a dummy node and a unity gain 5. Rearrange the nodes and/or loops in the graph to achieve maximum pictorial clarity. anch, EXAMPLE 8.6. Let us constuet signal flow graph forthe simple resistance getwork given in Fig. 813. There ae five variables, 0, yt, and fa, is Known. We ean write four independent equations from Krchof's ‘oltage and current laws Proceding fom let right inthe schematic, we have cfio-(ile somone uelto(e oem = Laying out the five nodes in the same order wih o, as an input sode, and connecting the nodes with the appropriate branches, we get Fig. 614 If we wish to consider sys an ouput node, we must add & unity gain84 SIGNAL FLOW GRAPHS (cuar s branch and another node, yielding Fg. 8.18. No rearrangement ofthe nodes is necessary. We have one forward ‘ath and three feedback loops leary in evidence up, 8 Nfs, SV Fig. 8s Note that signal flow graph representations of equations are not unique. For example, the addition of a unity gain branch followed by a dummy node changes the graph, but not the equations it represents, 86. THE GENERAL INPUT-OUTPUT GAIN FORMULA ‘We found in Chapter 7 that we can reduce complicated block diagrams to canonical form, from which the control ratio is easily written as ce. «¢ Ri Its possible to simplify signal low graphs in a manner similar to that of block diagram reduction But itis also possible, and much less time-consuming, to write down the input-output relationship by inspection from the original signal flow graph. This can be accomplished using the formula presented below. This formula can also be applied direcly to block diagrams, but the signal flow graph representation is easier to read—especally when the block diagram is very complicated. Let us denote the rato of the input variable o the output variable by T- For linear feedback control systems, T= C/R. For the general signal flow graph presented in preceding paragraphs T'= X,/%, where X, isthe output and %; isthe input ‘The general formula for any signal flow graph is (82) the ith forward path gain ',,=/th possible product of k nontouching lop gains ani = 1-DPat LP DP ‘= 1 (sum of all loop gains) + (sum ofall gain products of two nontouching loops) ~ (sum of all gain products of three nontouching loops) + ‘4, = A evaluated with ll loops touching P, eliminated ‘Two loops, paths, oF 2 loop and 2 path are sid to be nontouching if they have no nodes in common. 4 is called the signal flow graph determinant or characteristic function, since = 0 is the system characteristic equation ‘The application of Equation (82) is considerably more straightforward than it appears. The following examples illustrate this point, EXAMPLE 8.7. Let us fst apply Equation (#2) tothe signal ow grap ofthe canonical feedback system (Fig 416) car. SIGNAL FLOW GRAPHS ts ‘There is only one forward path Bence aoc Rene =0 ‘Ther is only ove (feedback) lop, Hence ry FH BynO jet ket Then PycleoH and Finaly, ay Rs ino EXAMPLE 0.8. The signal flow graph of the resistance network of Example R6 is shown in Fig 8:17, Let us spply Equation (82) 10 this graph and determine the voltage gain T= 0/0, of the rexstance network. ae ae un (®t Fee ‘Theres oe forward path (Fig. 18), Hence the forward path gains aR RR uy a Ry Fig 818 There are three feedback oop (Fig, 619). Hence the lop gins ae in % Re R Pam a186 SIGNAL FLOW GRAPHS (CHAP. ‘There are wo sootouching oops, loops 1 and 3. Hence Ry RR 7,.~ gaa producto the only two noatouchig loops = Py Py = ‘There are no thre loops that donot touch. Therefore RR RR RR RR Ry + RyRa+ RRs + Ry RR: a BCH t Bd aed kt ‘Since all oops touch the forward path, dy = 1. Finally, Pd Ake B78 RRP RR RRA RR RR, 8.7 TRANSFER FUNCTION COMPUTATION OF CASCADED COMPONENTS Loading effects of interacting components require litle special attention using signal flow graphs ‘Simply combine the graphs of the components at their normal joining points (output node of one to the input node of another), account for loading by adding new loops atthe joined nodes, and compute the ‘overall gain using Equation (82). This procedure is bes illustrated by example. EXAMPLE 8.8. Assume that two identical resistance networks are to be satcnded and wed as the contro ‘ements in the forward lop of s cootol system. The networks are simple vllage dividers ofthe form piven in Fig 20, ae ve(t)o-(i)o mt noms ‘The signal ow graph is easly drawa Fig 821). The ga ofthis network i, by inspection, equal to ue, x % Fig 821 I we were o ignore loading the overall ain of two cascaded networks would simply be determined by mulipying the individual pains CHAP. 8) SIGNAL FLOW GRAPHS 187 ‘This onower is inconrect We prove this a the following manner. When the to identical networks are cascaded, we note thatthe sev equivalent to the network of Example 86, seth Ry =, and Ry= Ry (Pit, 822) “The signal Now graph of his network was aluo determined in Example 86 (Fg. 8-23) mR, aR, Fig 823, ‘We obseeve that the feedback branch = Rin Fig, £23 does aot appear in the signal low graph of the cascaded signal fom graphs ofthe individual networks connected from node to. (Fig, 2). This means tha ‘Se esultof connecting the tmo network, the second one loads the ist, changing the equation fore, fom mR © eR Ri ur, aw A b \ % vas o Fig 824 “This result could ako have bese obtained by directly writing the equations for the combined networks, In this eas, only the equation foro, would have changed fora. The gain ofthe combined networks was determined in Example 88 as 5 ® 4 RP RRR, when R; is Se equal to Rand Ry is set equal to Ry, We observe that ay 8 (2) -amema,*e tis good general practice to calculate the gain of cascaded networks directly from the combined signal flow graph. Most practical control stem components load each other when connec in series 88 BLOCK DIAGRAM REDUCTION USING SIGNAL FLOW GRAPHS AND THE GENERAL INPUT-OUTPUT GAIN FORMULA (Often, the easiest way to determine the control ratio of a complicated block diagram isto translate the block diagram into a signal flow graph and apply Equation (32). Takeoft points and summing points must be separated by a unity gain branch in the signal flow graph when using Equation (82).188 SIGNAL FLOW GRAPHS (char. If the elements G and H of a canonical feedback representation are desired, Equation (82) also provides this information. The diect ts a=Era, (82) ‘The loop transfer function is Gu =3— (84) Equations (8.3) and (8.4) ace solved simultaneously for G and H, and the canonical feedback control system is drawn from the result EXAMPLE 8:10. Let us determine the control ratio C/R and the canonical block digram of the feedback contol system of Example 7.9 (Fig, 825) fa} " Fig. 825, ‘The signal ow graph is given ia Fig £26. There are to forward pas WG, = GGG Fig 826 “Thote are thre fsdback loops Py GGt Pym GG6it, ym GGG “There are no aontouching oops, and all oops touch both forward paths; then Ant aan “Therefore the control ati is C_AM+ 6.6, + 64 ROD TG + O6GH, = GGGR, 6.046, + 6) GOH + 66,0, = Har. §} SIGNAL FLOW GRAPHS 189 From Equations (8.2) and (Ad), we have G=GG(G,+G) and GH = ,0,(0, 1+ G1, ) GH (6,46) =H C7 Gta ‘The canonical block diagram i therefore given in Fig 827 Therefore r 6,0(G,+ 05) (Gt Gott, Hh ia Fig 827 The negative summing point sgn forthe Feedback loop isa result of wing 8 postive sige in the GH formula shove: It this snot obvious, efer to Equation (7.3) and is explanation ia Secon 7-4 “The block diagram above may be pt nto the Gaal form of Examples 79 or 7.10 by using the transformation theorem of Section 75 Solved Problems SIGNAL FLOW GRAPH ALGEBRA AND DEFINITIONS 8.1. Simplify the signal ow graphs given in Fig. 828, (a) (b) fe), Fig. 828, (2) Cleaty, X= AX, + BX = (A + 8) X, Therefore we have Age Ba (2) Weave X= BX; and AX, Hence X= BAX, of X (©) TA and B are mulipiative operators (eg, constants or transfer factions), we have Xp= AX, + BX, = (AAL~ B))% Hence te signal low graph becomes190 (AL FLOW GRAPHS [CHAP & 82. Draw signal flow graphs for the block diagrams in Problem 7.3 and reduce them by the ‘multiplication ule (Fig. 829). 10 w 7 =i 0 o - —, x % 7 “Og % Fig 8.29 83. Consider the signal flow graph in Fig 830. A As Aw x * % Fig 820 (2) Draw the signal flow graph for the system equivalent to that graphed in Fig, 8-30, but in which X, becomes KX, (k constant) and 4, X,,and_X, remain the same ()- Repeat part (a) forthe ease in which X, and X, become kX, and &,X;,and X, and Xy remain the same (ky and ky are constants), ‘This problem illustrates the fundamentals of a technique that can be used for scaling es (2) For the system to temin the same when a node variable is muliplied by a constant, al igaals centering the node must be mltiped by the same constant, nd al signals leaving the node divided by that constant Since, Xyand X, must en the same. the rancher are modied (Fig 31), Avft an a \ Ay x % Ba % Fig 831 (8) Sobstate kX, for Ny, and ky X for (Fig, 832) CHAP. 8) SIGNAL FLOW GRAPHS 191 1 is clear from the graph that Ay becomes ky Ay, Ayy becomes (ky/;)Ayy Ay becomes (ky/hs)ayy and Aap becomes (1/hs dey (Fit 33) sald nga (ong ean ‘84. Consider the signal low graph given in Fig. 8.34 Fig, 834 Identity the (a) input node, (b) output node, (c) forward paths, (d) feedback paths. (€) setf-loop. Determine the (J) loop gains of the feedback loops (g) path gains ofthe forward paths. @ % % (©) X 0X; 1X wX, w Xo % wo XK Hwoxwh wr XW KON WH OH OH (a) 10 Xo Xp 19K 10; Me 0 Xo Ke M10 Me XA: Hw HOH WA wx, wh; % ww: Lw% w MGW, 0% WAY Wy ¥ WH, WH WH WX to XX (© % 0x CP) AssAasi Aes Asha Aus Ases Areas AesArn days Ac Aes Ase An dirdasdaudiy: Arnderdiedes dreds (8) Aydodude dys Ari Aa Audodre SIGNAL FLOW GRAPH CONSTRUCTION 8S, Consider the following equations in which Xyxp.n.%, ate variables and 0, 85.005 8 coeficensor mathematical operators wtamts (6) x= Lanes @ x What are the minimum number of nodes and the minimum number of branches required 10 construct the signal flow graphs of these equations? Draw the graphs. (a) There are four variables in this equation: x, x3. x5, and +5. Therefore a minimum of four nodes are required. There ae thee coeficents of tansmison Functions onthe ighthand side of the equation192 SIGNAL FLOW GRAPHS (cHar.& 4), ¢y, and FL, Hence minimum of tree branches are required. A sina) signal fw graph is stow in Fig 838(0. « o Fig. 835 (b) There are n+ 1 variables: x20: 24, and 5; and there are coefiients 0,855.25 8 1 ‘Therefore animal signal flow graph shown in Fig 8350) 86. Draw signal flow graphs for @ aa{S) 0 oe GBs Bn ae fot (@) The operations called for 19 this equation are a, and d/dt Let the equation be written as siy= 0, (d/dtX). Since there ae two operations, we may define a new vanable ds /s and weit a an intermediate node, The signal flow graph i given in Fig #36 ca Fig 836 (8). Simlay, x5 = (42/4? Wx) + (d/o) ~ Therefore we obtain Fig, 37 (©) The operation is iteration Let the operator be denoted by fd. The signal fw raph is given in Fig 4538, CHAP.) SIGNAL FLOW GRAPHS 193 8:7. Construct the signal low graph forthe following set of simultaneous equations thts Hedi t dam tds) = det + Ages ‘There are four variables sy. 4: HeDce four nodes are required. Arranging them tra lett ight and connecting them wth the appropiate branches, we obian Fig 3. Fig 839 ‘A neater way to arrange hs graph it shown in Fig £40. i 40 88. Draw a signal flow graph for the resistance network shown in Fig. 81 in which v,(0) = a’ the voltage across C, : rin [ f* dL. 7 7 : . 5 Fig 841 The fie variables are oe, and Jade the input. The four independent equations derived ftom Kizchhll's voltage and current aw are iielae sd be slale(al> 9 ee eect ety nter a198 SIGNAL FLOW GRAPHS, [cHap.s In Laplace transform notation, the gaa Bow graph is given in Fig 3 ‘THE GENERAL INPUT-OUTPUT GAIN FORMULA 89, The transformed equations forthe mechanical system given in Fig. 44 are (Ft ky X= (Ms? + fst kX, (i) yk (ae pet +a) % a x4 Fig. 44 where F is force, M is mass, k is spring constant, f is friction, and _X is displacement Determine X,/F using Equation (82). ‘There ae three variables: X;, X, and F. Therefore we need three nodes. In order to draw the signal ‘tow graph, divide Equation (i) by 4 ad Equation (i) by B, where A= Mis" fe» hy and B= Maa? + forks tks © (F)e(S)sen oi) ‘Therefor the signal flow graphs pven in Fig. 845, kya F * % Fe sas “The foward path gin bP, k,/AB, The feedback lo esi is Py ~A2/AB. then == (4a hijab and 8,1 Fins, Bk A FO aa MP aH Tht rR) car. 8) SIGNAL FLOW GRAPHS 195 810, Determine the transfer function for she block diagram in Problem 7.20 by signal flow graph techniques. ‘The signal tow graph, Fig, 646, is dean dzecly (rom Fig, 7-44. There ae two forward paths. The path gains are , = G\Gs0, and Py ~ Gy The tne feedback lonp gains ae Py = Gh, Py = GOH fad Py = G6, Hy. No loops ae nontoching, Hence 8=1— (Pi) + Py + Py) Also, dy = 1; and since ‘no lope ouch the nodes of Pi, = 8. Thus pe BO APiBs GG + 61+ GMs ~ GO,G4H + 60,0, 3 14 GH, — GiGH, + GH, Fig. 8.46 Determine the transfer function ¥,/¥, from the signal low graph of Problem 88. ‘The singe forward path gain is 1/(2°R,,C\C;). Te loop gains of the thee fedback loops re Bam =1/GRC), Py = 1/ORG), and By ='~1/(oR,C3). The ssn product of the only two non- teeing loops is y= Py -Pyy= LASR RCC). Hence PRRCIC, +S RICC, + ARGC * RRC) + RG a PREGC, (Fy # Pat Pad + Pa Since al! oops touch the forward path, 8, 1. Finally, % Ah ae V8 PR RGG TRG *RG ARG) FT 8.12, Solve Problem 7.16 with signal low graph techniques. ‘The signal Now graph i drawn dec from Fig, 7-26, a8 shown in Fig. #47 4, Fig 847196 SIGNAL FLOW GRAPHS (cuar 8 With U, = U,=0, we hive Fig. 848. Then Py = G\6, and Py = GG, HM Hence B= 1 Py = 1 GG, fy, By = 1 and oG.R GGA, 1 0 1 eX } on Hat, Fig 88 1 & L 5 \ 7 q ome, Fig. 849 Then P= Gy, Py = G1G,M My, 81 GG Hi Hy, 8, =, and Gey Now put R~ Ui =0(Fig. 50) 1 gt i u cn Fig. 850 = GG Hy y=, and PAU GG.MU GG. Finally, we have GOR + GU + GEHL, CRN OT Gott "TRANSFER FUNCTION COMPUTATION OF CASCADED COMPONENTS. ‘813. Determine the transfer function for two of the networks in cascade shown in Fig 8-81 Fig 61 CHAP. 8] SIGNAL FLOW GRAPHS it 1a Laplace transform notation the network becomes Fig 82 Fig £52 By Kirchhofs avs, we have J, = sC¥, ~sC¥, and ¥;~ Rl The signal low graph sven i Fig, 853 Fig. 883 For two networks in cascade (Fig 8-54) the V; equation is also dependent on I: V;= RY, ~ Rly, Hence ‘wo networks are joined at node 2(Fig 8.58) and a fedback loop (Rl is added between and Vs (Fit £36, ms Fig 856 Then PRC, By Py = ARC, Pa PBs T43sRC+ RIC, A 1, and ray 2 TS Be Opn) IARC198 SIGNAL FLOW GRAPHS [cHar 8 £114, Two resistance netorks inthe form of han Example 6 ae to be used for conta ments Eg inthe forward ph of a contol stem, ‘The) ate tobe cascaded and shall have ene! aR, peeve component values ar shown in Fig 857 Find 070, using Equation (2) aa libel lela ‘Only the equation fr vis dierent from those ofthe single nework of Example 86; it ha an extra term, (Rats). Therefore the signal Row diagram foreach network alone (Example 86) may be joined a node fy and an extra branch of gaia — Ry drawa from 1,100. The resulting signal Now graph forthe double ‘ncwork is given ia Fig 58 SB RR Um oe suns (i Youn Ye Vs Yes Ys fo Fig. 858 ‘The voltage gain T= 14/0, is calelated from Equation (82) as follows. One forward path yes Py (R)RA/RR,)! Te gains ofthe seven feedback loops are P= = Ry/R, = Py, Py = —Ry/R: = Py Phy — = R/Re— Pry 208 P= ~Ra/Ry “There ae 5 gin prodets of two nontoching ops, From left orght, we have Poa RR noe a RR AR oon “Tere ate 10 gin products of thee nostouching ops From let gt, we have BR, AR BR RR RR OR fore, TO, RR aR RR RR 7 = BIR pe Bite RRR RR eRe ORO Rae CHAP. SIGNAL FLOW GRAPHS 199 “Thee is one gai product of four nootouching loops: Pha Py FPP “Therefore the determinants (ReRA/RIRY Ent Eesn Etna RR Rika RRs + RR ' RR Since ll ops tush the forward path, 8, = 1 and (eka? (RR) RRR, RR RyRy OR RE) RI ARRAY R RIALS OR RRR, ORR IRI RE “1 RR, + RARe+ Ra) BLOCK DIAGRAM REDUCTION 8.15. Determine C/R for each system shown in Fig 859 using Equation (8.2). & Fig. 859 (2) The signal ow graph is given ia Big, £40. The two forward path gus are Py = Gy. P= Gy. The wo feedback loop gaint are P= Gi, y= Gt Then Bet (h 4) a1200 SIGNAL FLOW GRAPHS icHar.¢ Fig. 860 Now, &;=1 and 8; =1 beens both paths touch the fedback loops a both interior nodes. Hence CPt Rs Rs GM = Fh (0) The signal tow graph i given in Fig, 6.6, Agno, we have A= Gy and = Gy. Bus gow there is nly one feedback loop, tnd Fuy= 6th; then &=1~ Gy, The forward path trough Cy clesey touches the feedback oop at nodes and b thus 8 = 1. The forward path throuth G, touches the {eedback lop abode a; then 8, =1- Hence CAA tRA, G46, Re am, Fig B61 (6) The signal flow graph is given in Fig £42. Agia, we have F, = Gi Fy = Gy, Py = GiB, 81 — GH, and 8, =1- But the feedback path doer no touch the forward pat though G; at any node ‘Thevlore 4, A= 1~ 6, and © Rbt Rds G.>6,0.~ 6M) Re 1-6 ‘This problem strats the importance of separating summing points and takcofT points with a branch of unity gin when applying Equation (82) 8.16, Find the transfer function C/R for he system shown in Fig. 8-63 in which K is a constant CHAP. §) SIGNAL FLOW GRAPHS 201 Fig 863 ‘The sgna fow graph is given in Fig 664, The only forward path gains n() ata AP , Si le Ts eset eee ie tinea Beret ~ Cnt Pa) 7 “ RS” (seaysr+s+ 04k) 8.17. Solve Problem 7.18 using signal low graph techniques. ‘The signal ow grap pven in Fig. 868 1 1 x Frat 1 ® ee 3 Fig 865 Applying the maliplication and ndion rues, we obtain Fig, 866. Now nek yn KEE, Keto B a eT ‘ Ne, . Fig 866202 SIGNAL FLOW GRAPHS [CHAP a nak kr D7 Weis Oik c=TR 18, Find C/R for the contsol system given in Fig. 867, Fig 867 “The signal fow graph is given in Fig, 868, The oo forward path gains ae F,= 6.6.0, and A. Gy The fe feedback loop pins a2e Fy = G\GsMy Py GG, Py = G60). Py = Gully Ghd B= GG, Hence BaD (Py 4 Pg By + Pa tBu) 1+ GCG, — GOH, ~ 6.6 ~ GH + GO, Fig. 268 and 6, = A =. Final, Cat ay GOG,+ 66. ROS GGG GGM- CGI GI OG, 819, Determine C/R for the system given in Fig. 8-69. Then put G, = G\G,H CHAP. 4) SIGNAL FLOW GRAPHS 203 “The signal ow graph i given in Fig, £70, We have P,~ GG, Fy ~ Gils, Py ~~ GH, S= 1+ Gly 8,8, Tan CATS za 1 o & 1 ; 7 Fig. 8.70 Pting G, = 6:6, "h, we obtain C/R GG, and the system wanslerfonetion becomes open-loop. 820. Determine the elements for a canonical fecdback system for the system of Problem 8.10. From Problem 810, R= G\GiGy, P= Gy, B= 1+ Gall, ~ G,0:M, + GGyMy, y=, and y= From Equation (43) we have Ge EPA GGG, + G+ 664th, ~ GG.GH, + G66, and from Equation (84) we obtain a GH, GOH, + 66,4, GGG G+ CGH, OGG, + COOH, Supplementary Problems DL Find C/R for Fig, 671, sing Bqution (82) aH Fig 81 822, Determine a set of canonical feedback sytem transfer functions for the preceding problem, using Equations (83) and (84)204 SIGNAL FLOW GRAPHS IcHar 8 1823, Scale the signal flow graph in Fig, 8:72 wo that X; becomes 4/2 (se Problem 8.3), 2 i % Fig 8.72 ‘824, Draw a signal flow graph for several nodes of the lat 34 by the equation anne Bane a2, faa signal flow graph forthe sytem presented in Problem 7.3 1826, Draw a signal flow graph forthe system presented in Problem 7.32. 27. Determine C/R, fom Equation (82) forthe signal ow graph drawn in Problem 8.26 828 Draw a signal ow graph forthe electrical pework in Fig. 873 829, Determine Y/Y; from Equation (82) for the network of Problem 828 830. Determine she elements fora canonical feedback system forthe network of Problem 8 28, using Equations (and (8) 831. Draw the signal flow graph for the analog compute cet Fig 874 CHAP. §) SIGNAL FLOW GRAPHS 205 ADs eer ‘832, Seale the anslog compte cic of Problem 831 s0 that y becomes 10y,dy/dt becomes 20(d/d), and yd? becomes Sd'y/a?) aa fever ner ig. 74 Answers to Supplementary Problems BBL R= GGG, P= GGG. Py = CGM, Poy = ~G.GGHa, Py ~~ Gy G.GsG4H, + GGG,H, and By = 8, =1. Theretore C_Aa thay 6.66, 65) a7 2 M1 G6 > BM G=Pd,+ MAs= GKG.+ 6) ans. Fig 8.6Fig 6.78 026, Fig 879 © GG, 8 Raa, 208 SIGNAL FLOW GRAPHS (cHar a CHAP. SIGNAL FLOW GRAPHS 207 a2 a2, % RyRy teRaRe 8207 RARE RRP RR, RR BR 830. G= RUE +0R,) pn BURRS Ra) RoR RRO) RAR re) a3 825. aa Fig 82