Signal Flow Graphs and Mason's Gain Formula

When analyzing complex control systems or electronic networks, the traditional block diagram reduction method can become a tedious exercise in algebraic manipulation. Signal flow graphs (SFGs) offer a powerful graphical alternative, and when paired with Mason's Gain Formula, they provide a systematic, almost algorithmic method for finding the overall transfer function of a system without redrawing or simplifying the diagram. This technique is indispensable for engineers dealing with multi-loop, multi-input systems where intuition fails and algebraic complexity soars.

The Building Blocks of a Signal Flow Graph

A signal flow graph represents a system of linear algebraic equations. It is a network of nodes and directed branches that visually maps the flow of signals and their interactions. Understanding its core components is foundational.

Nodes represent the system's variables, typically signals like voltage, current, or system states. Each node has an associated value. A source node has only outgoing branches (an input), while a sink node has only incoming branches (an output). Most nodes are mixed, both receiving and sending signals.

Branches are directed lines connecting nodes. They show the direction of signal flow and are labeled with a branch gain or transmittance. This gain, often a constant, a Laplace transfer function like $G(s)$, or a simple multiplier, defines the relationship between the node it comes from and the node it goes to. The signal at the end of a branch is the signal at the start multiplied by the branch gain.

The fundamental operation within an SFG is the node equation: the value of any node is the algebraic sum of all signals entering it. For example, if signals $x_1$, $x_2$, and $x_3$ arrive at node $y$ with gains $a$, $b$, and $c$ respectively, then $y=a{x}_1+b{x}_2+c{x}_3$. This simple rule governs all signal propagation through the graph.

Key Concepts: Paths, Loops, and Non-Touching Elements

Before applying Mason's formula, you must be able to identify specific features within the tangled web of branches. These features are the direct inputs to the formula.

A forward path is any consecutive sequence of branches from the input (source) node to the output (sink) node, following the direction of signal flow, where no node is encountered more than once. Each forward path has an associated forward path gain, $P_k$, which is the product of all branch gains along that path. A system can have multiple, distinct forward paths.

A loop is any closed path where starting from a node and following the direction of the branches returns you to the same starting node, without passing through any other node more than once. The loop gain, $L$, is the product of the branch gains in that loop. A graph can contain many loops.

Two parts of a graph (like two loops, or a loop and a forward path) are said to be non-touching if they share no common nodes. Identifying sets of non-touching loops is a critical step. For instance, you might find two loops that don't touch each other, forming a pair of non-touching loops. You might also find three loops that are all mutually non-touching.

Mason's Gain Formula: The Systematic Solution

Mason's Gain Formula provides a direct calculation for the overall transfer function $T=\frac{Y(s)}{R(s)}$ of a system represented by an SFG. It elegantly combines all the paths, loops, and their interactions into one equation:

$$T=\frac{1}{\Delta }\sum _k{P}_k{\Delta }_k$$

Where:

• $P_k$ = Gain of the $k^{th}$ forward path.
• $\Delta$ = The graph determinant or characteristic function of the system.
• ${\Delta }_k$ = The cofactor for the $k^{th}$ forward path.

The real work lies in calculating $\Delta$ and ${\Delta }_k$. The graph determinant $\Delta$ is given by:

$$\Delta =1-\sum L_1+\sum L_2-\sum L_3+\cdots$$

• $\sum L_1$ = Sum of the gains of all individual loops in the graph.
• $\sum L_2$ = Sum of the gain products of all possible pairs of non-touching loops.
• $\sum L_3$ = Sum of the gain products of all possible triplets of non-touching loops.
• This alternating sum continues for higher combinations.

The cofactor ${\Delta }_k$ is calculated *identically to $\Delta$, but for the sub-graph that does not touch* the $k^{th}$ forward path. In practice, you find $\Delta$ for the entire system, then you mentally remove all nodes and branches that touch the $k^{th}$ path and recalculate $\Delta$ for the remaining graph—that is ${\Delta }_k$. If a forward path touches every loop in the system, then ${\Delta }_k=1$.

Applying Mason's Formula: A Worked Example

Consider a system with one input $R$ and one output $C$. After drawing the SFG, you identify the following:

• Forward Paths: You find two.
• Path 1: $P_1=G_1{G}_2{G}_3$
• Path 2: $P_2=G_4$
• Loops: You identify three individual loops.
• Loop 1: $L_1=G_2{H}_1$
• Loop 2: $L_2=G_1{G}_2{H}_2$
• Loop 3: $L_3=G_4{H}_2$
• Non-Touching Loops: Check pairs. Loop 1 ($L_1$) and Loop 3 ($L_3$) share no common nodes. They are non-touching. No other pairs are non-touching. There are no triplets of non-touching loops.

Now, compute the graph determinant $\Delta$:

1. Sum of all individual loop gains: $\sum L_1=L_1+L_2+L_3=G_2{H}_1+G_1{G}_2{H}_2+G_4{H}_2$
2. Sum of products of all pairs of non-touching loops: $\sum L_2=(L_1)(L_3)=(G_2{H}_1)(G_4{H}_2)$
3. $\sum L_3=0$

Therefore: $\Delta =1-(G_2{H}_1+G_1{G}_2{H}_2+G_4{H}_2)+(G_2{G}_4{H}_1{H}_2)$

Next, find the cofactors ${\Delta }_1$ and ${\Delta }_2$:

• For $P_1$: This path touches all loops (it goes through or shares nodes with $L_1$, $L_2$, and $L_3$). Therefore, when this path is removed, no loops remain. So, ${\Delta }_1=1$.
• For $P_2$: This path touches $L_3$ and $L_2$, but does it touch $L_1$? We must check the nodes. If $L_1$ uses nodes not on Path 2, then they are non-touching. In this scenario, let's assume $L_1$ does not touch Path 2. Therefore, when we remove Path 2, loop $L_1$ remains. The determinant for the remaining graph is simply $1-L_1=1-G_2{H}_1$. So, ${\Delta }_2=1-G_2{H}_1$.

Finally, apply Mason's Gain Formula: $$T=\frac{1}{\Delta }[P_1{\Delta }_1+P_2{\Delta }_2]=\frac{G_1{G}_2{G}_3(1)+G_4(1-G_2{H}_1)}{1-G_2{H}_1-G_1{G}_2{H}_2-G_4{H}_2+G_2{G}_4{H}_1{H}_2}$$

Common Pitfalls

Misidentifying Loops and Paths: The most frequent error is incorrectly identifying a path or loop by passing through a node more than once. Remember the rule: no node can be repeated in a single forward path or loop. Always trace carefully and list each element.

Overlooking Non-Touching Combinations: It's easy to find all individual loops but then forget to check for pairs, triplets, etc., of loops that do not touch. Missing a pair of non-touching loops will lead to an incorrect $\Delta$ and thus a wrong transfer function. Systematically check each loop against every other loop.

Mistaking ${\Delta }_k$ Calculation: A common mistake is to assume ${\Delta }_k$ is always 1. ${\Delta }_k$ is not "$\Delta$ minus the loops touching the path." It is the determinant of the graph that remains after removing the entire forward path and all its nodes. You must recompute the determinant for this new, smaller graph, looking for loops that survived the removal.

Algebraic Errors in the Final Expression: After identifying all terms, the final formula is a large rational function. Errors in sign (remember the alternating signs in $\Delta$) or in combining terms in the numerator are common. Double-check the multiplication and summation of all $P_k{\Delta }_k$ products.

Summary

• Signal flow graphs provide a visual, node-and-branch representation of linear systems, where nodes are signals and branches are gains, governed by simple summation rules.
• Mason's Gain Formula computes the total transfer function directly by accounting for all forward path gains ($P_k$), all individual loop gains ($\sum L_1$), and the gains of all non-touching loop combinations ($\sum L_2$, $\sum L_3$, ...).
• The core equation is $T=\frac{1}{\Delta }\sum P_k{\Delta }_k$, where the graph determinant $\Delta =1-\sum L_1+\sum L_2-\sum L_3+...$ and ${\Delta }_k$ is $\Delta$ calculated for the sub-graph that doesn't touch the $k^{th}$ forward path.
• Successful application requires meticulous and systematic identification of all paths, all loops, and all possible non-touching combinations between them.
• This method is vastly more efficient than block diagram algebra for complex, interconnected multi-variable systems, as it avoids the need for step-by-step diagram manipulation.