Skip to content
Tutorials Center
My learning

Topics

Lesson 17 of 37 ยท Stability

Routh Special Cases

Handle the two special cases of the Routh array, a zero in the first column and an entire row of zeros, with the epsilon trick and auxiliary polynomial in Python.

The Routh array from the previous lesson runs smoothly as long as every first-column entry is non-zero. Two situations break the procedure, and both have standard fixes. They are not rare curiosities: a row of zeros appears exactly when the system is on the boundary of stability, which is the case you most want to analyse when you search for a critical gain.

Special case 1: a zero in the first column

If the first entry of a row is zero but other entries in that row are not, the next row would need a division by zero. The fix is the epsilon method: replace the zero by a small positive number epsilon, finish the array, and then let epsilon tend to zero when you read the signs.

Take s^5 + 2 s^4 + 3 s^3 + 6 s^2 + 5 s + 3:

Text
s^5 | 1    3    5
s^4 | 2    6    3
s^3 | 0    3.5  0      (first entry is zero: use epsilon)
s^2 | (6 eps - 7)/eps   3   (about -7/eps, large negative)
s^1 | about 3.5   0
s^0 | 3

With a tiny positive epsilon, the entry (6 eps - 7)/eps is a large negative number. The first column reads +, +, +eps, -, +, +: two sign changes, so two right-half-plane roots. (If you instead replaced the zero by a tiny negative epsilon, the count of sign changes comes out the same.)

An alternative: multiply the polynomial by (s + 1) to move the zero, or substitute s = 1/z and reverse the coefficients, then run the standard test on the new polynomial.

Special case 1

Zero pivot First entry zero, row not all zero.
Use epsilon Replace with a tiny positive number.
Read the signs Take the limit as epsilon goes to 0.

Count sign changes in the first column.

1 / 2

Special case 2: a whole row of zeros

If an entire row is zero, the polynomial has roots that are symmetric about the origin: pairs +/- a on the real axis, or +/- j w on the imaginary axis, or complex quadruples. The recipe:

  1. Form the auxiliary polynomial A(s) from the row just above the zero row. Its coefficients are the entries of that row, with powers that drop by two.
  2. Replace the zero row by the coefficients of dA/ds.
  3. Continue the array as usual.
  4. The roots of A(s) are also roots of the original polynomial. Solve it to see whether they are on the imaginary axis.

Take s^3 + 2 s^2 + s + 2:

Text
s^3 | 1    1
s^2 | 2    2
s^1 | 0    0     (row of zeros)
s^0 | 2

Auxiliary polynomial (from the s^2 row): A(s) = 2 s^2 + 2. Its derivative is 4 s, so the s^1 row becomes 4, 0. Then the s^0 entry is 2. First column: 1, 2, 4, 2: no sign changes, so no right-half-plane roots. The roots of 2 s^2 + 2 = 0 are +/- j, which are on the imaginary axis: the system is marginally stable. Indeed s^3 + 2s^2 + s + 2 = (s + 2)(s^2 + 1).

Implementing both fixes

Python Routh with both special cases handled

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The first polynomial uses the epsilon method; the printed first column shows a huge negative entry in place of -7/eps, and two sign changes, matching the two right-half-plane roots from np.roots. The second hits the zero row; the auxiliary polynomial 2 s^2 + 2 has roots +/- 1j, and the sign-change count is zero.

Finding the critical gain and oscillation frequency

A zero row at a particular gain K marks the stability boundary. Return to s^3 + 6 s^2 + 8 s + K from the previous lesson. At K = 48 the s^1 row is (48 - K)/6 = 0. The auxiliary polynomial from the s^2 row 6, K is 6 s^2 + 48, with roots s = +/- j sqrt(8) = +/- 2.83 j. So at K = 48 the closed loop oscillates at 2.83 rad/s. This is how you find the ultimate gain and ultimate period used by the Ziegler-Nichols tuning method in Chapter 6.

Python Critical gain and oscillation frequency

Press Run (or Ctrl + Enter) to see the result here.

Program input ยท what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

Remember these points

  • A zero in the first column only: use epsilon, count sign changes.
  • A zero row: build A(s) from the row above, differentiate, continue; the roots of A(s) lie in symmetric pairs.
  • If the zero row appears and the first column then shows no sign change, there are no right-half-plane roots, but there are roots on the imaginary axis (the system is marginal, not stable).
  • If the zero row appears and there are sign changes, the right-half-plane count is still the number of sign changes, but some roots of A(s) may also lie in the right half-plane.

Check your understanding

Quick check ยท 4 questions

of 4 answered You scored / 4 ยท perfect!

  1. 1. What does an entire row of zeros in a Routh array indicate?

    A zero row appears when the polynomial has factors like (s^2 + a), so roots come in symmetric pairs such as +/- j w or +/- a.
  2. 2. How do you continue after a zero row?

    Build A(s) from the row above, differentiate, and use the derivative's coefficients as the new row.
  3. 3. For s^3 + 2s^2 + s + 2, what are the imaginary-axis roots?

    The auxiliary polynomial is 2s^2 + 2, with roots s = +/- j. The remaining root is s = -2.
  4. 4. When using the epsilon method for a zero pivot, you

    Epsilon keeps the algebra valid; the signs of the first-column entries in the limit give the count.

Summary

  • A zero pivot is fixed with a small positive epsilon; read the first-column signs as epsilon -> 0.
  • A zero row signals symmetric roots; use the auxiliary polynomial from the row above and its derivative.
  • Solving the auxiliary polynomial at the critical gain gives the oscillation frequency.
  • These special cases are how you find stability boundaries and the ultimate gain for tuning.

Next: Gain Ranges and Relative Stability