## = -t/q

Instead, here, equations are used that are independent of the t1v, by simply finding t2j from T = [I + PG]-1PGF. Thus,

(e) ][ =

(d) 1

1 (c)

(b) 1

)1( (a)

1 2

p

L

) - (pfg = c

pp

pp =

L

Lqg = qg = L

L

c + Lf = t

P

+−

+−

+

+

γ γ

γ

γ

or the MISO structures of Fig. 7.2. This design is done after the design of Eq. (6.17a) has been completed by means of Eqs. (7.3a-d), so that L1, f11, and f12 are known (Use L1 not L1o -- see Appendix F). It is then necessary to find g2, f21, and

f22 so that in Fig. 5.2 the outputs y21 and y22 are stable and satisfy the tolerances on |y21| and |t22|, respectively. These are single-loop problems similar to Fig. 7.l, except that only the uncertainty P∈ P need be considered, as the c2j in Eqs. (7.4a-e) are not functions of the elements of τij, which they are in Eqs. (7.3a-d). At each step, design execution is that of a MISO single-loop system -- which is what makes this design procedure so tractable.