What Is a Signal Flow Graph?

A signal flow graph (SFG) is a graphical representation of the relationships among variables in a dynamic system. It originated in the 1950s as an alternative to block diagrams, offering a more compact and mathematically intuitive way to model linear systems. Each SFG uses nodes to represent system variables and directed branches to indicate causal relationships, with branch weights acting as transfer functions or gains. Because the graph directly mirrors algebraic equations, it simplifies both manual analysis and computational simulation of control systems, filters, and mechanical or electrical networks.

Engineers and students rely on SFGs to visualize complex interdependencies without getting lost in cumbersome algebra. By focusing on signal flow rather than physical components, the diagram highlights how disturbances, feedback loops, and multiple inputs affect overall system behavior. The power of an SFG becomes especially clear when applying Mason’s gain formula, which yields a system’s transfer function directly from the graph topology.

Key Components of a Signal Flow Graph

Every signal flow graph is built from two fundamental elements:

  • Nodes – Points that represent system variables, such as voltages, forces, temperatures, or error signals. Each node typically corresponds to a single variable, and the value of that variable is the sum of all incoming signals multiplied by their respective branch gains.
  • Branches – Directed line segments connecting one node to another. Each branch has a direction (from cause to effect) and a weight (gain or transfer function) that defines how the source node influences the destination node. A branch without a weight implies a unity gain.

In contrast to block diagrams, where blocks represent functional operations and summing junctions exist as separate elements, SFGs integrate these operations into branch weights. This compactness makes SFGs particularly useful for analyzing multi-loop feedback systems, cascaded subsystems, and systems with multiple inputs and outputs.

Advantages of Using Signal Flow Graphs

Signal flow graphs offer several practical benefits over traditional block diagrams and algebraic methods:

  • Simplicity in handling feedback loops – Loops can be identified visually, and their effect can be computed using straightforward graph reduction or Mason’s rule without redrawing the diagram.
  • Direct algebraic representation – Each node’s value equals the sum of the incoming signals, which corresponds to a set of simultaneous linear equations. This property enables direct conversion between the graph and the system’s state-space representation.
  • Easy identification of paths and loops – Forward paths, feedback loops, and non-touching loops become visible at a glance, simplifying the application of Mason’s gain formula.
  • Scalability – Adding more variables or connections does not clutter the graph as quickly as a block diagram would; the node-and-branch structure remains readable.
  • Useful for both analysis and design – Engineers can modify branch gains to examine sensitivity, stability margins, and transient response without re-deriving equations.

These advantages make SFGs a standard tool in control system design, filter synthesis, and even certain areas of signal processing and economics.

Step-by-Step Construction of a Signal Flow Graph

Building an accurate SFG requires a systematic approach. The following steps guide you from the system equations to a complete, analyzable graph.

Step 1: Identify All System Variables

Begin by listing every independent and dependent variable in the system. In a control context, these include inputs (references, disturbances), outputs (controlled variables), and intermediate states (e.g., error signals, actuating signals, feedback signals). For a mechanical system, variables might include positions, velocities, and forces. For an electrical network, voltages at key nodes and currents in each branch.

Assign each variable a unique symbol or label. It is helpful to write the set of system equations first and then extract the variables. For example, for a simple RC circuit with input voltage \(V_{in}\) and output voltage \(V_{out}\) across the capacitor, the variables are \(V_{in}\), \(V_{out}\), and the current \(I\) flowing through the resistor (or the voltage drop across the resistor). In general, the number of nodes in the SFG will equal the number of distinct variables.

Tip: Include intermediate variables even if they are not final outputs; they are essential for expressing interdependencies and will later appear as internal nodes.

Step 2: Determine the Relationships (Transfer Functions)

From the system equations, derive the functional dependence of each variable on the others. In linear time-invariant systems, these relationships are linear algebraic or differential equations that can be transformed into transfer functions using the Laplace transform (with zero initial conditions). Write each variable as a sum of contributions from other variables, each multiplied by a gain.

For example, if a variable \(x_2\) depends on \(x_1\) through a gain \(G\) and also on itself via a feedback loop, the equation becomes:

\[ x_2 = G \cdot x_1 + H \cdot x_2 \]

In the SFG, this translates to a branch from \(x_1\) to \(x_2\) with weight \(G\), and a self-loop on \(x_2\) with weight \(H\). Similarly, summing junctions in block diagrams become multiple incoming branches to a single node.

Important: The direction of the branch must always point from the independent (cause) variable to the dependent (effect) variable. This matches the flow of signals in the system.

Step 3: Draw Nodes and Connect Branches

Place a node for each variable on the page. Usually, the input node is placed on the left and output node on the right, but any layout that avoids crossing branches is acceptable. Draw directed branches (arrows) from each source node to its dependent node(s). Label each branch with its corresponding transfer function or gain symbol.

If the same variable influences multiple others, that source node will have multiple outgoing branches. If a variable receives contributions from several variables, it will have multiple incoming branches. There is no summing junction node separate from the variable node; the addition happens implicitly at the destination node.

For a standard feedback control system with plant \(G(s)\) and controller \(C(s)\) in the forward path, and sensor \(H(s)\) in the feedback, the SFG would include:

  • Input node \(R(s)\)
  • Error node \(E(s)\) with branches from \(R(s)\) (gain 1) and from the output \(Y(s)\) through \(-H(s)\)
  • Actuating node \(U(s)\) from \(E(s)\) through \(C(s)\)
  • Output node \(Y(s)\) from \(U(s)\) through \(G(s)\)

Plotting these nodes and branches creates a clear visual map of signal propagation.

Step 4: Assign Transfer Functions to Every Branch

Every branch must have a weight defined, even if it is unity (in which case the weight is simply omitted or written as 1). The transfer functions can be constants (gains), Laplace-domain expressions like \(1/(s+1)\), or any rational function. Ensure that the gain is appropriate for the direction: a feedback path might have a negative sign included in the gain (e.g., \(-H(s)\)).

If multiple parallel paths connect two variables, combine them into a single branch with the sum of the individual gains for simplicity. Similarly, series branches can be merged by multiplying their gains together, but only if there are no nodes along the path that are used elsewhere. Graph reduction should be done only after the full graph is drawn, to avoid losing loop information needed for Mason’s rule.

Once the complete graph is drawn, you may simplify it by eliminating unnecessary nodes (those that are not outputs and have only one incoming and one outgoing branch) using the series rule. Self-loops can be absorbed by modifying the gains of incoming branches. However, simplification may inadvertently remove paths or loops required for accurate analysis using Mason’s formula. Therefore, it is safer to keep the original graph for systematic analysis and only simplify for a quick manual check.

Common simplifications include:

  • Series branches: Replace two or more branches in cascade by a single branch whose gain is the product of the individual gains.
  • Parallel branches: Replace two or more branches from the same source to the same destination by one branch with gain equal to the sum of the individual gains.
  • Removal of self-loops: A self-loop with gain \(L\) on a node can be removed by dividing all incoming branch gains to that node by \(1-L\).

These operations preserve the overall input-output relationship but may alter internal node values. Use them with caution.

Practical Example: RC Low-Pass Filter

Consider the simple RC low-pass filter with input voltage \(v_i(t)\) and output voltage \(v_o(t)\) across the capacitor. The equations in the time domain are:

\[ v_i(t) = R\,i(t) + v_o(t) \]
\[ i(t) = C \frac{dv_o(t)}{dt} \]

Taking Laplace transforms with zero initial conditions:

\[ V_i(s) = R\,I(s) + V_o(s) \]
\[ I(s) = sC\,V_o(s) \]

From these we can write the variable relationships:

\[ V_o(s) = \frac{1}{sC}\,I(s) \]
\[ I(s) = \frac{1}{R} ( V_i(s) - V_o(s) ) \]

Nodes: \(V_i\), \(V_o\), and \(I\).
Branches:

  • From \(V_i\) to \(I\) with gain \(1/R\)
  • From \(V_o\) to \(I\) with gain \(-1/R\) (since \(V_o\) appears with a negative sign in the second equation)
  • From \(I\) to \(V_o\) with gain \(1/(sC)\)

The resulting SFG has three nodes. The input node \(V_i\) has one outgoing branch to \(I\). Node \(I\) has one incoming branch from \(V_i\) and one from \(V_o\), and one outgoing branch to \(V_o\). Node \(V_o\) has one incoming branch from \(I\) and one outgoing branch back to \(I\). This graph clearly shows the feedback path through the resistor.

Using Mason’s gain formula, the overall transfer function becomes:

\[ \frac{V_o(s)}{V_i(s)} = \frac{ \frac{1}{RCs} }{ 1 + \frac{1}{RCs} } = \frac{1}{1 + RCs} \]

which matches the expected low-pass filter characteristic.

Analyzing Signal Flow Graphs with Mason’s Gain Formula

Mason’s gain formula provides a direct method to compute the transfer function from any input node to any output node without reducing the graph. The formula is:

\[ T = \frac{ \sum_k P_k \, \Delta_k }{ \Delta } \]

where:

  • \(P_k\) is the gain of the \(k\)-th forward path from input to output.
  • \(\Delta\) is the determinant of the graph, given by \(1 - \sum L_i + \sum L_i L_j - \sum L_i L_j L_k + \cdots\), where \(L_i\) are loop gains (products of branch gains around each closed loop).
  • \(\Delta_k\) is the cofactor of the \(k\)-th forward path, obtained by removing from \(\Delta\) all terms that involve loops touching that forward path.

Before applying Mason’s rule, identify all forward paths and all individual loops (including self-loops). Then compute the determinant and cofactors. This method is especially powerful for systems with multiple feedback loops and multiple forward paths, where manual reduction would be tedious.

For more detailed derivations and examples, consult standard control textbooks such as Modern Control Engineering by Ogata or MIT OpenCourseWare on Feedback Control Systems. Online resources like Control Systems Tutorials also provide interactive examples.

Common Mistakes When Constructing Signal Flow Graphs

Even experienced engineers can fall into certain traps. Here are frequent errors and how to avoid them:

  • Reversing branch direction – Always check the cause-effect relationship. A common error is to draw feedback branches in the same direction as forward branches. Remember that feedback paths go from the output back toward an earlier node.
  • Omitting negative signs – In feedback loops, the subtraction at the summing junction should appear as a negative branch gain. If you omit the sign, the graph will imply positive feedback, leading to incorrect analysis.
  • Creating unnecessary nodes – While intermediate variables are needed, avoid splitting a variable into two nodes if they represent the same signal. This would create a false zero-gain path that changes the graph topology.
  • Ignoring self-loops – If a variable depends on itself through a direct term in the equations, represent it as a self-loop. Forgetting it will eliminate that feedback and alter the transfer function.
  • Premature simplification – Reducing the graph before identifying all loops can hide loops needed for Mason’s rule. Always draw the full graph first, then apply the formula, and simplify only if necessary for a sanity check.

A helpful habit is to verify the graph by manually writing node equations from the drawn graph and comparing them to the original system equations. Any mismatch indicates a mistake in the graph structure or branch gains.

Applications Beyond Control Systems

While signal flow graphs are most commonly taught in control engineering, they have broader utility:

  • Digital filter design – SFGs directly represent difference equations and are used to implement filters in hardware or software. Each branch delay element corresponds to a unit delay \(z^{-1}\).
  • Economic and biological models – Any system described by linear equations, such as input-output models in economics or population dynamics, can be analyzed using SFGs.
  • State-space analysis – An SFG can be transformed into a state-variable diagram where integrators become branches with gain \(1/s\) and states are the integrator outputs.

Because of this versatility, mastering SFG construction and analysis is a valuable skill for any engineer or scientist dealing with dynamic systems.

Further Reading and Resources

To deepen your understanding, explore the following authoritative sources:

These resources, combined with practice on real systems (electrical circuits, mechanical mass-spring-damper setups, thermal processes), will solidify your ability to construct and analyze signal flow graphs efficiently.

Conclusion

Constructing signal flow graphs is a systematic process that transforms complex system equations into an intuitive visual model. By identifying all variables, capturing their cause-effect relationships through directed branches, and assigning proper transfer functions, you create a diagram that is both analytically powerful and easy to interpret. Adding Mason’s gain formula to your toolkit enables direct computation of system transfer functions without graph reduction. With careful attention to direction, sign, and loop identification, signal flow graphs become an indispensable method for analyzing dynamic systems in control engineering, signal processing, and beyond.