MODELING PRACTICE 7 MIN READ

Mason's formula or block reduction: which is faster?

Both give the same transfer function. Reduction wins when loops nest or sit in a row; Mason's formula wins when loops overlap or several paths reach the output.

The two methods are often taught as if one were a shortcut for the other. They compute the same thing, and each turns into tedious bookkeeping on the wrong diagram. The question to ask before starting is how the loops sit relative to one another.

01 / FIELD NOTE

A diagram whose loops overlap

Follow the signal. r enters a junction that subtracts the output y and H₁·x₂, giving e₁; G₁ turns e₁ into x₁; a second junction subtracts H₂·y, giving e₂; G₂ turns e₂ into x₂; G₃ turns x₂ into y. There are three loops: G₁G₂ fed back through H₁, G₂G₃ fed back through H₂, and the outer unity loop through G₁G₂G₃.

Mason's formula takes three lines. One forward path, P₁ = G₁G₂G₃. Three loop gains, −G₁G₂H₁, −G₂G₃H₂ and −G₁G₂G₃. Every pair of loops shares G₂, so there are no non-touching pairs, and the path touches every loop, so Δ₁ = 1. Hence T = G₁G₂G₃/(1 + G₁G₂H₁ + G₂G₃H₂ + G₁G₂G₃).

Block reduction cannot start until something moves, because the H₁ loop picks off x₂ inside the H₂ loop while the H₂ loop's junction sits inside the H₁ loop. Moving the H₁ pick-off point forward past G₃ turns H₁ into H₁/G₃; then the G₂G₃ loop closes, G₁ joins in series, the two feedback paths from y add to 1 + H₁/G₃, and the outer loop closes: five steps. Moving the second junction back past G₁ instead turns H₂ into H₂/G₁ and also takes five steps. Pick the block that is easy to invert: if G₃ = 1/s, the first route puts a differentiator H₁·s into an intermediate diagram, while a static G₁ costs nothing.

02 / FIELD NOTE

The same answer three ways

With static gains G₁ = 2, G₂ = 4, G₃ = 0.5 and H₁ = H₂ = 0.5, Mason's denominator is 1 + 4 + 1 + 4 = 10 and T = 4/10 = 0.4. Solving the five signal equations directly gives the same y = 0.4, with e₁ = 0.2, x₁ = 0.4, e₂ = 0.2 and x₂ = 0.8. The first reduction route passes through G₂G₃/(1 + G₂G₃H₂) = 1 and 1 + H₁/G₃ = 2, then 2/(1 + 2·2) = 0.4.

With dynamics, G₁ = 2, G₂ = 4/(s+2), G₃ = 1/s and H₁ = H₂ = 0.5, the same formula gives T(s) = 8/(s² + 6s + 10): poles at −3 ± j1, a final value of 0.8 for a unit step, a 10–90% rise time of 0.98 s and a 2% settling time of 1.66 s.

03 / FIELD NOTE

Where reduction wins

Now put two loops in a row: G₁ with its own feedback H₁, followed by G₂ with its own feedback H₂. Reduction closes each loop and multiplies, T = G₁/(1 + G₁H₁) · G₂/(1 + G₂H₂). Mason's formula must notice that the two loops do not touch and include their product, Δ = 1 + G₁H₁ + G₂H₂ + G₁H₁G₂H₂, which you then factor back into the same product. With three loops in a row Δ has eight terms, one for every combination of loops. With G₁ = 4, H₁ = 0.5, G₂ = 2 and H₂ = 1.5 both routes give 8/12 ≈ 0.667.

DiagramFasterWhy
Loops nested inside one anotherReductionClose from the inside out; nothing has to move.
Separate loops in a rowReductionEach loop closes on its own; Mason's Δ needs every non-touching combination.
Loops that overlapMason's formulaReduction must first move a pick-off point or a junction.
Several forward paths, such as a feedforwardMason's formulaEach path gets its own Δₖ; reduction has to merge the paths first.
You need an internal signal, such as the errorMason's formulaTreat that node as the output; reduction usually erases it.
04 / FIELD NOTE

Try it

On the block-diagram bench, choose Feedback and set G₁ = 4, G₂ = 0.5, H = 0.5: y/r = 1, the closed G₂G₃ loop of the first route. Then set G₁ = 2, G₂ = 1, H = 2: y/r = 0.4, the whole diagram. The bench handles one loop with static gains, so it checks the steps of a reduction, not the overlap itself.

In the step response calculator, enter Numerator 8 and Denominator s^2+6s+10 and choose K·G(s) rather than the closed loop: the final value 0.8, rise time 0.98 s and settling time 1.66 s match the dynamic case above.

05 / FIELD NOTE

What neither method tells you

Both methods are bookkeeping for linear, time-invariant blocks. They say nothing about saturation, about a cancelled pole that still lives inside the loop, or about whether the resulting T is stable; check its poles separately. Mason's formula is also easy to misapply when a loop is counted twice or a non-touching pair is missed, so on a diagram you did not draw yourself, solving the node equations numerically is a cheap cross-check.

TAKEAWAY

Look at how the loops sit before choosing: nested or in a row, reduce; overlapping or with several paths, use Mason's formula. Either way, check the result against the signal equations.

CHECK YOUR UNDERSTANDING: Three loops: L₁ and L₂ touch each other, and L₃ touches neither. One forward path touches all three. Write Δ.

Δ = 1 − (L₁ + L₂ + L₃) + (L₁L₃ + L₂L₃). The product L₁L₂ is left out because those loops touch, and Δ₁ = 1 because the path touches every loop.