Signal Flow Graphs: Complete Beginner's Guide For students.

A signal flow graph draws linear equations as circles (nodes) and arrows (branches): each node is a variable, each branch a transmittance (gain), and a node's value is the sum of incoming branch signals. Mason's gain formula then gives the transfer function from any input to any output directly, with no step-by-step reduction.

Flow graphs shine on multi-loop systems that are painful to reduce block by block: list forward paths, list loops, combine in one formula. They are also the natural language of digital filters, where the graph maps almost one-to-one onto code.

Key Concepts

• Node: a variable. Source nodes have only outgoing branches (inputs); sink nodes only incoming (outputs).
• Forward path: input-to-output route never repeating a node; gain P_k is the product of its branches.
• Loop: closed route returning to its start; a node-to-itself branch is a self-loop.
• Non-touching loops: loops sharing no nodes.
• Determinant Delta: 1 - (sum of loop gains) + (sum of non-touching pair products) - (triplets) and so on.
• Cofactor Delta_k: Delta recomputed after deleting all loops touching path k.

Formulas

• Mason's rule: T = (1/Delta) x sum_k (P_k x Delta_k).
• Delta: 1 - sum(L1) + sum(L2) - sum(L3) + …, alternating signs over loop products.
• Assumptions: linear system, unilateral branches (signals flow one way).

Worked Example

One forward path with branch gains 2, 3, 4; one feedback loop of gain -0.5 touching the first two branches:

1. P1 = 2 x 3 x 4 = 24.
2. L1 = -(2 x 3 x 0.5) = -3; Delta = 1 - (-3) = 4.
3. The loop touches the path, so Delta_1 = 1.
4. T = 24 x 1 / 4 = 6. The forward gain of 24 is cut to 6 by negative feedback, in five mechanical steps.

Common Misconceptions

• "Delta_k always equals Delta." The top Mason's-rule error: Delta_k must exclude loops touching path k.
• "Self-loops don't count." A self-loop is a valid first-order loop and belongs in the sums.
• "It works for nonlinear systems." Branches multiply and add, so the method is strictly linear.

Real-World Applications

• Multi-loop control: cascade and nested feedback in one pass.
• Digital filters: direct-form and lattice structures drawn as graphs before coding.
• Sampled-data systems: discrete-time loops map cleanly onto flow graphs.
• Analog simulation: the method's historical origin: integrators wired as a graph.

Signal flow graphs turn equations into a picture of nodes and gains; Mason's formula turns the picture into the transfer function: sum each forward path times its cofactor, over the determinant. List paths, list loops, watch for touching loops, and multi-loop systems surrender in lines of arithmetic.