You are on page 1of 6

Block Diagram Models, Signal Flow Graphs and Simplification Methods

Dynamic Systems and Control Lavi Shpigelman

Block Diagram Manipulation Rules
X1
G1

X2

G2

X3 =

X1

G2 G1

X3

X1 X1 + ± X2 X3 =

G

+ ±
G

X3

G

X2

Block Diagram Models
• Visualize input output relations • Useful in design and realization of (linear) components • Helps understand flow of information between internal
variables. X1

Block Diagram Manipulation Rules
G

X2 = X2

X1
G

G

X2

X2 X1 X1
G

• Are equivalent to a set of linear algebraic equations
(of rational functions ).

X2

X1 =

G G-1

X2

• Mainly relevant where there is a cascade of information
flow

X1

Block Diagram Manipulation Rules X1 G Block Diagram Reduction (with SISO components) H2 ! G4 G2 + X3 ± X2 X1 + = ± G-1 G X3 U(s) + X2 + G H - G1 + - + + G3 G4 H1 Y(s) X2 = X1 (I+GH)-1G X2 H3 Block Diagram Reduction (with SISO components) H2 U(s) + G1 Block Diagram Reduction (with SISO components) H2 ! G4 G2 + - - G2 + + G3 G4 H1 Y(s) U(s) + - G1 + - + + G3 G4 H1 Y(s) H3 H3 .

Block Diagram Reduction (with SISO components) H2 ! G4 G2 G3 G4 !!!!!! 1-G3 G4H1 Y(s) Block Diagram Reduction (with SISO components) U(s) + - G1 + - U(s) + - G1 G2 G3 G4 !!!!!!!!!!!! 1-G3 G4H1+ G2 G3H2 Y(s) H3 H3 Block Diagram Reduction (with SISO components) H2 ! G4 G2 G3 G4 !!!!!! 1-G3 G4H1 Y(s) Block Diagram Reduction (with SISO components) U(s) + - G1 + - U(s) + - G1G2 G3 G4 !!!!!!!!!!!! 1-G3 G4H1+ G2 G3H2 Y(s) H3 H3 .

SFG ± G H x2 x1 1 x1 G(s) x2 1 x2 ±H(s) U(s) G1 G2 G3 G4 !!!!!!!!!!!!!!!!!!!!!!! 1-G3 G4H1+ G2 G3H2+ G1 G2 G3 G4H3 Y(s) • Blocks • Edges + junctions Edges (aka branches) Vertices (aka nodes) (representing transfer functions) (representing variables) Signal Flow Graphs • Alternative to block diagrams • Do not require iterative reduction to find • Can be used to find the transfer function • Look familiar to computer scientists (?) Algebraic Eq representation • x = Ax + r x1 = a11x1+a12x2+r1 x2 = a21x1+a22x2+r2 G11(s) u1(s) G21(s) G12(s) u2(s) G22(s) y1(s) transfer functions (using Mason’s gain rule) between any two variables (not just the input and output). • y(s) = G(s)u(s) y2(s) .Block Diagram Reduction (with SISO components) x1 + Block Diagram Vs.

. a source and a sink. no edge.!(all loop gains) + !(products of non-touching-loop gain pairs) . !k = ! of the SFG after removal of the kth forward path • Output (sink) has only incoming edges • any variable can be y2 1 a12 y1 y2 a32 a23 y3 1 y3 made into an output by adding a sink with “1” edge .Loops that have no vertex in common (and.!(products of non-touching-loop gain triplets) + . • Forward-path: path connecting a source to a sink • Loop: A simple graph cycle..Another SFG Example a32 Definitions y5 y2 = a12y1 + a32y3 y3 = a23y2 + a43y4 y4 = a24y2 + a34y3+a44y4 y5 = a25y2 + a45y4 a12 y1 y2 a32 a12 y1 y2 a23 y3 y4 y5 y3 a43 y4 a32 a12 y1 y2 a23 y3 a43 a34 y4 a24 a44 y5 a32 a12 y1 y2 a23 y3 a43 a34 y4 a24 a25 a44 a45 y5 • Input: (source) has only outgoing branches • Output: (sink) has only incoming branches • Path: (from node i to node j) has no loops. • Path Gain: Product of gains on path edges • Loop Gain: Product of gains on loop Loops: • Non-touchingtherefore.) Input / Output • Input (source) has only outgoing edges a32 a12 y1 y2 a23 y3 Mason’s Gain Rule (1956) Given an SFG. N forward paths between them and K loops. the gain (transfer function) between the source-sink pair is !Pk!k Tij = !!!! ! Pk is the gain of path k. ! is the “graph determinant”: ! = 1.

!2 = 1-L5 = 1 .01 G=20000 y(s)/u(s)=20000/(1+20000*0.G4H4 y(s) P1+P2!2+P3 T(s) = !! = !!!!!` x(s) ! .01)=99.01 Forward paths: Feedback loops: L1 = -G2G3G4G5H2 L4 = -G7H2G2 L7 = -G1G2G7G6H3 P1 = G1G2G3G4G5G6 P2 = G1G2G7G6 P3 = G1G2G3G4G8 L2 = -G5G6H1 L3 = -G8H1 L5 = -G4H4 L6 = -G1G2G3G4G5G6H3 L8 = -G1G2G3G4G8H3 Loops {3.01)=99.Mason’s Rule for Simple Feedback loop P1 = G(s) L1 = -G(s)H(s) ! = 1 .{4.7} don’t touch ! = 1-(L1+L2+L3+L4+L5+L6+L7+L8)+(L3L4+L4L5+L5L7) !1 = !3 = 1 .5} and {5.4}.(-G(s)H(s)) !1 = 1 P1 !1 G(s) G(s) T(s) = !!! = !! = !!!!! ! ! 1+G(s)H(s) G7 G8 x(s) 1 G1 G2 G3 G4 -H4 -H2 -H3 G5 -H1 G6 y(s) A Feedback Loop Reduces Sensitivity To Plant Variations x2 x1 1 x1 G(s) x2 1 y(s)/u(s)=G/(1+GH) u(s) 1 G(s) 1 y(s) -H(s) G=10000 y(s)/u(s)=10000/(1+10000*0.50 -H(s) = 0.