Routh-Hurwitz Stability Criterion

Determining whether a control system will remain stable or oscillate out of control is a fundamental challenge in engineering design. While you could find the exact poles of the system's characteristic equation, solving high-order polynomials is computationally intensive and often unnecessary. The Routh-Hurwitz stability criterion provides an elegant algebraic shortcut. This powerful tool allows you to assess closed-loop stability by inspecting a simple array of coefficients, revealing not just if a system is stable, but how many unstable poles it has and what parameter values will keep it in check.

Foundations of Algebraic Stability Analysis

Every linear time-invariant (LTI) feedback control system has a characteristic equation. This equation is derived from the closed-loop transfer function's denominator set to zero. For a system to be stable, all roots (or poles) of this equation must have negative real parts, placing them in the left-half of the complex s-plane. If any root has a positive real part, the system's response will grow unbounded, leading to instability.

The Routh-Hurwitz criterion addresses this by answering a specific question: How many roots of the characteristic equation lie in the right-half plane? It does this without ever solving for the roots explicitly. The method is purely algebraic, working directly from the coefficients of the characteristic equation, which is typically written in the standard form:

$$a_n{s}^n+a_{n-1}{s}^{n-1}+...+a_{1}s+a_0=0$$

where all $a_i$ are real coefficients and $a_n>0$. The first step is a necessary (but not sufficient) condition for stability: all coefficients $a_0,a_1,...,a_n$ must be present and have the same sign. If any coefficient is zero or if there is a sign change, the system is definitely unstable. However, meeting this condition doesn't guarantee stability; for that, you must construct the Routh array.

Constructing and Interpreting the Routh Array

The core of the method is the Routh array. Its purpose is to generate a sequence of numbers whose first column reveals the stability story. You begin by writing the first two rows of the array using the coefficients of the characteristic equation.

Subsequent rows are calculated using a specific pattern of cross-multiplication. For the $s^{n-2}$ row, the elements $b_1,b_2,$ etc., are computed as follows:

$$b_1=\frac{-1}{a_{n-1}}\left|\begin{array}{cc}{a}_n& a_{n-2}\\ a_{n-1}& a_{n-3}\end{array}\right|=\frac{a_{n-1}{a}_{n-2}-a_n{a}_{n-3}}{a_{n-1}}$$

$$b_2=\frac{-1}{a_{n-1}}\left|\begin{array}{cc}{a}_n& a_{n-4}\\ a_{n-1}& a_{n-5}\end{array}\right|=\frac{a_{n-1}{a}_{n-4}-a_n{a}_{n-5}}{a_{n-1}}$$

This process continues, constructing the $s^{n-3}$ row from the two rows above it, until you reach the $s^0$ row. The key stability theorem states: The number of sign changes in the first column of the completed Routh array equals the number of closed-loop poles located in the right half-plane. Therefore, for a system to be stable, there must be zero sign changes in the first column, and all elements in that column must be positive.

Let's apply this to a simple third-order equation: $s^3+6s^2+12s+8=0$. The Routh array is constructed as:

The first column is: 1, 6, $\frac{64}{6}$, 8. All are positive with no sign changes. Thus, the system has zero right-half-plane poles and is stable.

Handling Special Cases: Zero Elements and Entirely Zero Rows

Two special cases can complicate array construction: a single zero in the first column, and an entire row of zeros. Each requires a specific remedy.

If the first element of a row is zero, but the rest of the row is not, the standard calculation fails because you would have to divide by zero to compute the next row. The solution is to replace the zero with a small positive number, $ϵ$, and then proceed with the calculations. After completing the array, examine the first column by taking the limit as $ϵ\to 0$. The signs will indicate the number of sign changes and right-half-plane poles.

An entirely zero row is a more profound indicator. It signals the presence of symmetrically located roots, such as purely imaginary pairs (e.g., $\pm j\omega$) or opposite real pairs (e.g., $\pm \sigma$). To continue, you form an auxiliary polynomial $A(s)$ from the coefficients of the row above the zero row. You then take the derivative of this polynomial with respect to $s$ and use its coefficients to replace the zero row. The roots of the auxiliary polynomial are also roots of the original characteristic equation, which you can now solve for directly.

Parametric Analysis for Controller Gain Selection

One of the most powerful engineering applications of the Routh-Hurwitz criterion is determining the range of a variable parameter (like a controller gain $K$) for which the closed-loop system remains stable. This is done by including the parameter $K$ as a symbolic variable within the characteristic equation's coefficients.

You then construct the Routh array as usual, keeping $K$ in the elements. The stability condition requires that every element in the first column be positive. By setting each element that contains $K$ greater than zero and solving the resulting inequalities, you can find the allowable range for $K$. The boundary of stability occurs when any element in the first column becomes zero, which often corresponds to the onset of sustained oscillations.

Consider a system with characteristic equation $s^3+3s^2+3s+1+K=0$. The Routh array is:

For stability:

1. $3>0$ (always true).
2. $(1+K)>0⟹K>-1$.
3. $\frac{8-K}{3}>0⟹8-K>0⟹K<8$.
4. $(1+K)>0$ again, repeating condition 2.

Thus, for all roots to be in the left-half plane, the gain must satisfy $-1<K<8$. This parametric analysis provides the designer with clear, calculable limits for tuning.

Common Pitfalls

Misapplying the Necessary Condition: Remember that the requirement for all coefficients to be positive and non-zero is only a necessary condition. A system can pass this test but still be unstable (as revealed by a sign change in the Routh array). Always complete the full array for a stability verdict.

Incorrectly Handling the Zero-First-Element Case: A common error is to simply skip a row if its first element is zero. This yields an incorrect array. You must use the $ϵ$ method, complete the array, and then analyze the signs as $ϵ$ approaches zero from the positive side. Failing to do this can lead to a miscount of sign changes.

Solving for Roots Unnecessarily: The entire point of the Routh-Hurwitz criterion is to avoid solving the characteristic polynomial. A pitfall is to abandon the array when you encounter a special case, like a zero row, and immediately try to find all roots. Instead, use the prescribed methods (auxiliary polynomial) to continue the stability analysis algebraically. You only need to solve the (usually lower-order) auxiliary polynomial to find the symmetric roots.

Ignoring Sign Changes from Negative Leading Coefficients: The method requires the first coefficient ($a_n$) to be positive. If your characteristic equation has a negative leading coefficient, multiply the entire equation by -1. Forgetting this step can lead to a confusing array where the interpretation of sign changes becomes ambiguous.

Summary

• The Routh-Hurwitz stability criterion is an algebraic test that determines the number of closed-loop poles located in the right-half of the s-plane by tracking sign changes in the first column of the Routh array.
• Construction of the Routh array follows a deterministic cross-multiplication pattern, starting with the coefficients of the system's characteristic equation.
• Special cases, such as a zero in the first column or an entire row of zeros, have defined remedies (the $ϵ$ method and the auxiliary polynomial, respectively) that allow the analysis to proceed.
• A primary engineering application is parametric analysis, which allows designers to determine the exact range of a variable (like controller gain $K$) that ensures closed-loop stability.
• The criterion provides a necessary and sufficient condition for stability: for a system to be stable, there must be zero sign changes in the first column of the Routh array, which requires all elements in that column to be positive.