Skip to main content

All-pass in delay equalization; 4th order Bessel-Thomson

Short answer

All-pass filters add delay where the original filter delay is low, flattening total group delay without changing magnitude, crucial for data and video. The fourth-order Bessel-Thomson low-pass transfer function is T(s) = 105/(s^4 + 10s^3 + 45s^2 + 105s + 105).

Explain the importance of all pass filters in delay equalization. Find the transfer function of fourth order Bessel-Thomson low pass filter.

Importance of all-pass filters in delay equalization

An all-pass filter has ∣T(jω)∣=1|T(j\omega)| = 1 at all frequencies but a frequency-dependent phase. A first-order section is T(s)=s−as+aT(s) = \dfrac{s-a}{s+a} and a second-order section is T(s)=s2−ω0Qs+ω02s2+ω0Qs+ω02T(s) = \dfrac{s^2 - \frac{\omega_0}{Q}s + \omega_0^2}{s^2 + \frac{\omega_0}{Q}s + \omega_0^2}: zeros are mirror images of the poles.

  • Sharp filters (Chebyshev, elliptic) and transmission channels have non-constant group delay, largest near the band edge. This causes phase distortion of pulses and data.
  • An all-pass filter cascaded with such a filter adds delay without changing the magnitude.
  • Its delay is designed to be large where the original delay is small, so that the total delay is nearly flat: τtotal(ω)=τfilter(ω)+τAP(ω)≈\tau_{total}(\omega) = \tau_{filter}(\omega) + \tau_{AP}(\omega) \approx constant.
  • Used in telephone/modem lines, video and data channels, and after sharp anti-aliasing filters.
 tau    filter delay   +  all-pass delay  = total
  |      _/\              \_    _/         ______
  |   __/   \               \__/
  +--------- w        +-------- w    +-------- w

Fourth-order Bessel-Thomson low-pass filter

Ideal 1 s delay: T(s)=e−s=1/(cosh⁡s+sinh⁡s)T(s) = e^{-s} = 1/(\cosh s + \sinh s). Expand coth⁡s\coth s as a continued fraction and truncate after four terms:

coth⁡s≈1s+13s+15s+17s\coth s \approx \frac{1}{s} + \cfrac{1}{\cfrac{3}{s} + \cfrac{1}{\cfrac{5}{s} + \cfrac{1}{\cfrac{7}{s}}}}

Working from the bottom:

5s+s7=s2+357s3s+7ss2+35=10s2+105s(s2+35)1s+s(s2+35)10s2+105=s4+45s2+10510s3+105s\begin{aligned} \frac{5}{s} + \frac{s}{7} &= \frac{s^2+35}{7s} \\ \frac{3}{s} + \frac{7s}{s^2+35} &= \frac{10s^2 + 105}{s(s^2+35)} \\ \frac{1}{s} + \frac{s(s^2+35)}{10s^2+105} &= \frac{s^4 + 45s^2 + 105}{10s^3 + 105s} \end{aligned}

So m(s)=s4+45s2+105m(s) = s^4 + 45s^2 + 105 (approximates cosh⁡s\cosh s) and n(s)=10s3+105sn(s) = 10s^3 + 105s (approximates sinh⁡s\sinh s):

B4(s)=s4+10s3+45s2+105s+105B_4(s) = s^4 + 10s^3 + 45s^2 + 105s + 105

(Check by recursion: B4=7B3+s2B2=7(s3+6s2+15s+15)+s2(s2+3s+3)B_4 = 7B_3 + s^2B_2 = 7(s^3+6s^2+15s+15) + s^2(s^2+3s+3), which gives the same polynomial.)

T(s)=105s4+10s3+45s2+105s+105T(s) = \frac{105}{s^4 + 10s^3 + 45s^2 + 105s + 105}

This gives unity DC gain and a delay of 105/105=1105/105 = 1 s at low frequency; replace ss by sτ0s\tau_0 for a delay of τ0\tau_0.

✓ Answer checked by Claude Sonnet 5.5

All Filter Design chapter 2 questions →