Skip to main content

Chebyshev order and elliptical pole locus

Short answer

n ≥ cosh⁻¹√((10^{0.1α_min}–1)/(10^{0.1α_max}–1)) / cosh⁻¹(ω_s/ω_p). Normalized poles satisfy σ²/sinh²a + ω²/cosh²a = 1, with a=(1/n)sinh⁻¹(1/ε), an ellipse centered at origin; semi-minor axis sinh a on real axis, semi-major cosh a on jω axis.

Derive the expression of the order of a low pass Chebyshev approximation and then prove that locus of its pole is an ellipse centered at origin.

Order of a Chebyshev low-pass filter

The Chebyshev low-pass magnitude response (normalized so that the passband edge is Ω=ω/ωp=1\Omega = \omega/\omega_p = 1) is

∣T(jΩ)∣2=11+ε2Cn2(Ω)|T(j\Omega)|^2 = \frac{1}{1+\varepsilon^2 C_n^2(\Omega)}

where Cn(Ω)C_n(\Omega) is the Chebyshev polynomial of order nn:

Cn(Ω)={cos⁡(ncos⁡−1Ω),∣Ω∣≤1cosh⁡(ncosh⁡−1Ω),∣Ω∣>1C_n(\Omega) = \begin{cases} \cos(n\cos^{-1}\Omega), & |\Omega| \le 1 \\ \cosh(n\cosh^{-1}\Omega), & |\Omega| > 1 \end{cases}

The attenuation in dB is α(Ω)=10log⁡10[1+ε2Cn2(Ω)]\alpha(\Omega) = 10\log_{10}\left[1+\varepsilon^2 C_n^2(\Omega)\right].

Step 1: ripple factor from the passband edge. At Ω=1\Omega = 1, Cn(1)=1C_n(1) = 1 for every nn, so the attenuation is exactly αmax\alpha_{max}:

αmax=10log⁡10(1+ε2)  ⇒  ε2=100.1αmax−1\alpha_{max} = 10\log_{10}(1+\varepsilon^2) \;\Rightarrow\; \varepsilon^2 = 10^{0.1\alpha_{max}} - 1

Step 2: stopband condition. At Ωs=ωs/ωp (>1)\Omega_s = \omega_s/\omega_p\ (>1) the attenuation must be at least αmin\alpha_{min}:

10log⁡10[1+ε2Cn2(Ωs)]≥αmin  ⇒  Cn2(Ωs)≥100.1αmin−1ε210\log_{10}\left[1+\varepsilon^2 C_n^2(\Omega_s)\right] \ge \alpha_{min} \;\Rightarrow\; C_n^2(\Omega_s) \ge \frac{10^{0.1\alpha_{min}}-1}{\varepsilon^2}

Step 3: solve for n. Since Ωs>1\Omega_s > 1, Cn(Ωs)=cosh⁡(ncosh⁡−1Ωs)C_n(\Omega_s) = \cosh(n\cosh^{-1}\Omega_s):

cosh⁡(ncosh⁡−1Ωs)≥100.1αmin−1100.1αmax−1\cosh(n\cosh^{-1}\Omega_s) \ge \sqrt{\frac{10^{0.1\alpha_{min}}-1}{10^{0.1\alpha_{max}}-1}} n≥cosh⁡−1100.1αmin−1100.1αmax−1cosh⁡−1(ωsωp)n \ge \frac{\cosh^{-1}\sqrt{\dfrac{10^{0.1\alpha_{min}}-1}{10^{0.1\alpha_{max}}-1}}}{\cosh^{-1}\left(\dfrac{\omega_s}{\omega_p}\right)}

The order is the next whole number above this value. (Useful identity: cosh⁡−1x=ln⁡(x+x2−1)\cosh^{-1}x = \ln\left(x+\sqrt{x^2-1}\right).)

Locus of the poles is an ellipse

The poles are the left-half-plane roots of 1+ε2Cn2(s/j)=01+\varepsilon^2 C_n^2(s/j) = 0, i.e. Cn=±j/εC_n = \pm j/\varepsilon. Putting cos⁡−1(s/j)=u+jv\cos^{-1}(s/j) = u + jv:

cos⁡(nu)cosh⁡(nv)−jsin⁡(nu)sinh⁡(nv)=±jε\cos(nu)\cosh(nv) - j\sin(nu)\sinh(nv) = \pm \frac{j}{\varepsilon}

Equating real parts: cos⁡(nu)=0⇒uk=(2k−1)π2n\cos(nu)=0 \Rightarrow u_k = \dfrac{(2k-1)\pi}{2n}. Equating imaginary parts: sin⁡(nuk)=±1⇒v=a=1nsinh⁡−11ε\sin(nu_k)=\pm1 \Rightarrow v = a = \dfrac{1}{n}\sinh^{-1}\dfrac{1}{\varepsilon}. Then s=jcos⁡(uk+ja)s = j\cos(u_k + ja) gives (taking the left-half-plane roots)

sk=σk+jωk,σk=−sinh⁡a sin⁡(2k−1)π2n,ωk=cosh⁡a cos⁡(2k−1)π2ns_k = \sigma_k + j\omega_k,\quad \sigma_k = -\sinh a\,\sin\frac{(2k-1)\pi}{2n},\quad \omega_k = \cosh a\,\cos\frac{(2k-1)\pi}{2n}

for k=1,2,…,nk = 1, 2, \dots, n (normalized to ωp=1\omega_p = 1; multiply by ωp\omega_p to denormalize).

Now eliminate θk=(2k−1)π/2n\theta_k = (2k-1)\pi/2n. From the pole expressions:

σksinh⁡a=−sin⁡θk,ωkcosh⁡a=cos⁡θk\frac{\sigma_k}{\sinh a} = -\sin\theta_k, \qquad \frac{\omega_k}{\cosh a} = \cos\theta_k

Squaring and adding, and using sin⁡2θk+cos⁡2θk=1\sin^2\theta_k + \cos^2\theta_k = 1:

σk2sinh⁡2a+ωk2cosh⁡2a=1\frac{\sigma_k^2}{\sinh^2 a} + \frac{\omega_k^2}{\cosh^2 a} = 1

This is the equation of an ellipse centred at the origin of the s-plane. Since cosh⁡a>sinh⁡a\cosh a > \sinh a:

  • semi-major axis =cosh⁡a= \cosh a, along the jωj\omega axis;
  • semi-minor axis =sinh⁡a= \sinh a, along the σ\sigma axis;
  • foci at ±j1\pm j1 (because cosh⁡2a−sinh⁡2a=1\cosh^2 a - \sinh^2 a = 1), i.e. at the passband edge.
            jw
      .-----+-----.   cosh a
     /  x   |   .  \
    |       |       |
 ---x-------+-------+--- sigma
    |  sinh a       |
     \  x   |   .  /
      '-----+-----'
  (only left-half poles x are used)

As ε→0\varepsilon \to 0 (smaller ripple), aa grows and sinh⁡a≈cosh⁡a\sinh a \approx \cosh a, so the ellipse approaches a circle. Relative to its size the response then approaches the Butterworth response.

✓ Answer checked by Claude Sonnet 5.5

All Filter Design chapter 2 questions →