88 CONTROL SYSTEM DESIGN 3.7 STATE-SPACE REPRESENTATION OF TRANSFER FUNCTIONS: CANONICAL FORMS In S€e. 3.5 we learned how to determine the transfer function of a linear, time‘invariant system, given the state-space representation. Sometimes it is necessary to go in the other direction: from the transferfunction to the state-space representation. This need may arise because the only available description of a subsystem within a larger system is the transfer function of that subsystem. In order to use state-space methods, the transfer function must be turned into a set of first-order differential equations. Another reason for converting a transfer-function representation into a state-space representation is for the purpose of transient response simulation. Many algorithms and numeri- cal integration computer programs designed for solution of systems of first- order equations are available, but there is not much software for numerical inversion of Laplace transforms. Thus, if a reliable method is needed for calculating the transient response of a system, one may be better off converti the transfer function of the system to state-space form and numerically inte- grating the resulting differential equations rather than attempting to compute the inverse Laplace transform by numerical methods. In the last section we saw that there are innumerable systems that huve the same transfer function. Hence the representation of a transfer function in state-space form is obviously not unique. In this section we shull develop several standard, or “canonical” representations of transfer functions that can always be used for single-input, multiple-output or multiple-input, singic-output systems, One canonical representation has no general advamtage over any other, and, moreover, there is no reason why a canonical representation is to be preferred over a noncanonical representation: First companion form The development starts with a transfer function of a single-input, single-output system of the form H(s) a ea (3.82) which can be written ts + ays*' +++ ++ a)y(s) = uls) (3.83) The differential equation corresponding to (3.83) is D'y+a,D''y+---+ayru (3.84) where Dy stands for d*y/di*. Solve for the highest derivative in (3.84) D*y =-a,D*"y -—a,D*“y —----aytu (3.85) Now consider a chain of k integrators as shown in Fig. 3.8(a), and suppose that the output of the last integrator is y. Then the output of the next-to-last integrator is Dy,= dv/dr, and so forth. The output from the frst integrator isSa a ee DYNAMICS OF LINEAR SYSTEMS 89 +y () Figure 38 State-space realization of transfer functions in first companion form f 1 f+ bast () Hv =———— (6) Hts = 2S Stare ee and the input to this integrator is thus D*y. From (3.85) it follows that Fig. 3.8(a) represents the given transfer function (3.82) provided that the feedback gains are chosen as shown in the figure. To get one state-space representation of the system, we identify the output of each integrator with a state variable, starting at the right and proceeding to the left. The corresponding differential equations using this identification of state variables are (3.86)90 CONTROL SYSTEM DESIGN The output equation is simply yen 87) The matrices corresponding to (3.86) and (3.87) are . ey at a 0 f°] Oat 1 9 1c A= Bal: (3.88) ‘ONO. 0 1 0 a, ay 4-2 -a, itl , Call 00 --- O) = fan bey Od = The matrix A has a very special structure: the coefficients of the denominator of the transfer function, preceded by minus signs, are strung out along the bottom row of the matrix. The rest of the matrix is zero except for the ~superdiagonal” terms which are all unity. In matrix theory, a matrix with this structure is said to be in companion form. For this reason we identify this state-space realization of the transfer function as a companion-form realization. We call this the first csmpanion form; another coinpanion form will be discussed later on. If the state variables were numbered from right to left we would have By = 04x) — Xz + Ky HE B=x Be re noe stone see cares (3.89) Seon = Xeez X= Xen and The corresponding matrices would be i i a | aon OSIM 0 i Ome o oO B=|0 i S : (3.90) I UO Us 0 0 C=[0 0 --- 0 1) = [> -- by] This representation is also called a companion form, but is less frequently used than the form (3.88). There is nothing sacred about numbering the integrators systematically from right to left or from left to right. A perfectlyren ere ne eR SPR SEAT TN RSET ORIN A NN DYNAMICS OF LINEAR SYSTEMS 92 valid, if perverse, representation would result if the integrators were numbered at random. Having developed a state-space representation of the simple transfer func- tion (3.82), we are now in a position te consider the more general transfer function H(s) G91) ‘The development is aided by the introduction of an intermediate variable 2(s) u(s) 2(s) us) s* Fas 4--- Fay We identify the first factor with the numerator and the second factor with the denominator: vis) S MS) bse bsthe cost 2) as) 7 bos ts by (3.92) and 2) u(s) s*+a,s' (3.93) Hosta The realization of the transfer function from wu to z has already been developed. And, from (3.92) y(si = (bos* + bys) + ~~ - + By )2(s) = byD*z + b, Dz +--+ + byz The inputs to the integrators in the chain are the k successive derivatives of z as shown in Fig. 3.8(4), hence we have the required state-space representation. All that remains to be done is to write the corresponding differential equations. The state equations are the same as (3.86) or (3.89) and hence the A and B matrices are the same. The output equation is found by careful examination of the block diagram of Fig. 3.8(b). Note that there are two paths from the output of each integrator to the system output: one path upward through the box labeled 0, and a second path down through the box labeled a; and thence through the box labeled by, As a consequence, when the right-to-left state variable numbering is used Y= (be ~ andor + (Bya1 — Oy bo) Xz ++ = + + (by = Aye) Xe + bot Hence C= [by = aybo, be-1 — Oe-1Bo,---, 6: Gib), =D =[bo] (3.94) If the direct path through bg is absent, then the D matrix is zero and the C matrix contains only the b, coefficients.92 CONTRO. SYSTEM DESIGN If left-to-right numbering is used, then e C = [by — aybo, bs ~ Arbo, — ado), D=[bo) (3.95) “The structure of the first canonical-form is very easy to remember (““auto- mnemonic”). The string of integrators can be visualized as the fraction bar of the transfer function (3.91) that is realized, The numerator coefficients appear above the chain of integrators in the same order as they appear above the fraction bar in (3.91) and the denominator coefficients appear below the chain of integrators in the same order as they appear below the fraction bar in (3.91). Not too much imagination is needed to “sce” the transfer function (3.91) in Fig. 3.8 ‘A generalized version of the first companion form can be used to realize a single input, multiple output system represented by / transfer functions, one from the single input to each of the / different outputs yils) _ bors + bust tt + bas uls) shin str are eer oy yas) _ bass + byjsh ts + By u(s) stays ++: - + ay The same set of stale variables serves for each transfer tunction. Each numerator, however. is realized by a ditferent set of gains, as shown in Fig, 3.9. Thus the A and B matrices are exactly as given earlier. From Fig. 3.9 it is also seen that the C and 2D mat sare yy abo Pe aa Oe bus Bol [Par Dea saseiae ah ee See ea She By = O,Py by ag ~ Aea-rbor 7 bu sabe] bor (3.96) for the right-to-left numbering, or = aby by = asbyy ~~ Far ~ asbor] bu | 24 o o . D 2 | (3.97) y= boy bay Aabq 72> Pay ~ Mabor bos he leftro-right numbering. : the first canonical form realizations cf Figs. 3.1 through 3.9 the input is d direttly to the first integrator in the chain and the output is a linear ro ‘on of the outputs of the integrators (and the input, when the D matrix t s). This form is useful not only for single-input, single-output systems, b us we have seen, for single-input, multiple-output systems, A variant of u ture of Fig. 3.8, in which the output is taken directly from the last i but the input is connected to all the integrators, is shown in Fig. 3.10. a tion of a multiple-input, single-outpvt system based on the structure of “1s, 1) is shown in Fig. 3.11.