Signal Flow Graphs

An alternative graphical model of dynamic systems, instead of block diagrams is the signal flow graph.
Although the Block Diagram approach and signal flow graph approach are functionally the same, the
Block diagrams are generally used more for representing systems in the Laplace domain, while signal
flow graph are more intuitively used for modelling systems in the time domain (or State-Space
equations).

A signal flow graph is a pictorial representation of a set of simultaneous linear algebraic equations
describing a system. Signal flow graphs typically use nodes and curved lines to represent signals and
systems, instead of using lines at right-angles and boxes, respectively. A systems is represented by a
lines with an arrow showing the direction of signal flow through the system, while signal is represented
by a small dot called a node. This interconnections of nodes with directed lines depicts the
transmission of signals from one point (value, form, type, nature) to another, in the system.

Signal Flow graphs are particularly desirable in control systems due to certain analysis that can be
made easy using tools such as the Mason’s Gain Formula.

Definitions
The terminologies frequently used in signal flow graph theory are defined as follows, with the help of
a sample signal flow graph shown in figure 1.

Node: A node is a point representing a signal. For example, X1, X2, X3, and X4, are nodes.

Branch: A branch is a directed line segment connecting two nodes. Branches are always unidirectional.
For example, A21, A42, A23 and A33 are branches.

Input Node or Source: This is a node with only outgoing branches. This corresponds to an independent
variable. For example X1 is an input node.

Output node or Sink: This is a node with only incoming branches. This corresponds to a dependent
variable. For example X4 is an output node.

Mixed node: A mixed node is one that has both at least one incoming and one outgoing branch. For
example X2 and X3 are mixed nodes.

Transmittance or gain: This the transfer function of the subsystem represented by the branch
between two nodes. These can be real (constants) or complex (s-domain) gain values. The gain of a
branch is the transmission function of the branch when the transmission function is a multiplicative
operator.
1

Figure 1: Sample signal flow graph

Page
Path: A path is a continuous, unidirectional transversal of branches along which no node is passed
more than once. X1 to X2 to X3 to X4, X2 to X3 and back to X2, and X1 to X2 to X4 are paths.

Forward Path: This is a path from the input node to the output node. For example, X1 to X2 to X3, to
X4, and X1 to X2 to X4 are forward paths.

Loop: This is a path which originates and terminates on the same node. For example, X2 to X3 and back
to X2 is a feedback loop.

Self-loop: This is a feedback loop consisting of a single branch. For example, A33 is a self-loop.

Non-touching loops: These are loops that are do not share any common nodes. There are no non-
touching loops in Figure 1.

Path gain: This is the product of the branch gains encountered in traversing a path. For example, the
path gain of the forward path from X1 to X2 to X3, to X4 is A21A32A43.

Loop gain: This is the product of the branch gains of a loop as you traverse from one point, around the
loop, back to the starting point. For example, the loop gain of the feedback loop from X2 to X3, and
back to X2 is A32A23.

Fundamentals Properties of Signal Flow Graphs

• A conveys the functional relationship of one signal on another. This mathematical operator
mapping the two signals is considered multiplicative in the direction of the arrow.
• An additive operation acts at nodes with all incoming branches [Addition Rule] and the value
designated to nodes with outgoing branches are transmitted to the successive nodes
[Transmission Rule]. For example X4=A42X2 + A43X3.
• A cascaded A cascaded (series) connection of n - 1 branches with transmission functions A21,
A32, A43,. . . , An(n- 1) can be replaced by a single branch with a new transmission function equal
to the product of the old ones.

Mason’s Gain Formula

Mason’s Gain rule presents a simplified method to compute the transfer function of a system without
carrying out reduction of a signal flow graph. The Transfer function of a signal flow graph is given by:

𝐶(𝑠) ∑𝑖 𝑃𝑖 ∆𝑖
𝐺(𝑠) = =
𝑅(𝑠) ∆
Where,

𝑃𝑖 = the ith forward path gain

𝐿𝑗𝑘 = jth possible product of k non-touching gains

∆ = determinant of the graph

= 1 − (−)𝑘+1 ∑𝑘 ∑𝑗 𝐿𝑗𝑘
2
Page

= 1 − ∑𝑗 𝐿𝑗1 + ∑𝑗 𝐿𝑗2 − ∑𝑗 𝐿𝑗3 + ⋯ + ∞

 = 1 – (Sum of all loop gains) + (Sum of gains of all possible combinations of two
non-touching loops) – (Sum of gains of all possible combinations of three non-
touching loops) + ⋯

∆𝑖 = Cofactors of the kth forward path. i.e. ∆ evaluated with all loops touching the 𝑖 𝑡ℎ
forward path eliminated.

Examples
1. Convert the block diagram of the canonical feedback form in Figure 2 to a signal flow graph
and find the transfer function of the system.

Figure 2

Block diagram conversion:

By Graph reduction:
Tip: 𝐶(𝑆) = 𝐺𝑅(𝑠) ∓ 𝐺𝐻𝐶(𝑆)
≡
≡
By Mason’s Rule:

There is only one forward path,

𝑃1 = 𝐺
𝑃2 = 𝑃3 = ⋯ = 0
There is also only one loop

𝐿11 = ∓𝐺𝐻
𝐿𝑗𝑘 = 0 for 𝑗 ≠ 1 , 𝑘 ≠ 1

Thus,

∆ = 1 − 𝐿11 = 1 ± 𝐺𝐻
∆1 = 1 − 0 = 1
𝐶(𝑠) 𝑃1 ∆1 𝐺
∴ 𝐺(𝑠) = = =
3

𝑅(𝑠) ∆ 1 ± 𝐺𝐻
Page
 2. Convert the block diagram in Figure 3 to a signal flow graph and find the transfer function of
the system.

Figure 3

The signal flow graph is given in figure 4. There two forward paths:

𝑃1 = 𝐺1 𝐺2 𝐺4 𝑃2 = 𝐺1 𝐺3 𝐺4

Figure 4

There are three feedback loops:

𝑃11 = 𝐺1 𝐺4 𝐻1 𝑃21 = −𝐺1 𝐺2 𝐺4 𝐻2 𝑃31 = −𝐺1 𝐺3 𝐺4 𝐻2

There are no nontouching loops, and all loops touch both forward path, thus

∆1 = 1 ∆2 = 1
Therefore the transfer function is
𝐶(𝑠) 𝑃1 ∆1 + 𝑃2 ∆2 𝐺1 𝐺2 𝐺4 + 𝐺1 𝐺3 𝐺4
∴ 𝑇(𝑠) = = =
𝑅(𝑠) ∆ 1 − 𝐺1 𝐺4 𝐻1 + 𝐺1 𝐺2 𝐺4 𝐻2 + 𝐺1 𝐺3 𝐺4 𝐻2
𝐺1 𝐺4 (𝐺2 + 𝐺3 )
=
1 − 𝐺1 𝐺4 𝐻1 + 𝐺1 𝐺2 𝐺4 𝐻2 + 𝐺1 𝐺3 𝐺4 𝐻2
4
Page
Quick Practice Exercises
i. Find ∆𝑖 in Figure 5.

Figure 5

ii.

5
Page