I would like to do a theoritical filter with different attenuation according to the frequency. How can I do for having an attenuation different from 20 dB per decades. If for example I want to have an attenuation equal to 8.23 dB per decade ? How can I do ?

Have a nice day :)

  • 7
    \$\begingroup\$ You can approximate it with a series of shelving networks, like a staircase. First you need to know the frequency span of interest and the accuracy required. \$\endgroup\$ Jun 7, 2021 at 14:43
  • 1
    \$\begingroup\$ There is also fractional order filters. But currently they can ve realised using the method mentioned in above comments. \$\endgroup\$
    – AJN
    Jun 7, 2021 at 14:46
  • \$\begingroup\$ Use a graphic equalizer. \$\endgroup\$
    – Andy aka
    Jun 7, 2021 at 15:23
  • \$\begingroup\$ just to be sure: this is about implementing a complicated analog filter, not about a digital one, right? Because in the analog case, as user_1818839 correctly points out, realization 100% depends on the frequency range of interest. A filter with GHz bandwidths is implemented fundamentally different than one for audio, is implemented fundamentally different than for narrowband RF. \$\endgroup\$ Jun 7, 2021 at 16:30
  • \$\begingroup\$ Thank you for all your replies :) \$\endgroup\$
    – Jess
    Jun 8, 2021 at 6:23

1 Answer 1


Was looking at this not long ago, as a series of steps in the bode plot, logarithmically spaced, exactly like the first comment.

Expression below will give a slope of +/- 20\${\gamma}\$ dB/decade, between \${\omega}_0\$ and \${\omega}_f\$. N is number of pole-zero pairs.

$$\displaystyle\frac{\left(1+\frac{s}{\omega_0}\right)^\gamma}{\left(1+\frac{s}{\omega_f}\right)^\gamma} \space\space\space\space \approx \space\space\space\space \displaystyle\prod_{k=1}^{N} \frac { \left( 1+s\frac{ {q} ^ {(k-1) } }{ \omega_0 } \right) } { \left( 1+s\frac{ {q} ^ {-(k-1)} }{ \omega_f } \right) } \space\space\space\space\space\space\space\space\space\space q= \left( \frac{\omega_f}{\omega_0} \right) ^{ \frac{(N-\gamma)}{N(N-1)} } $$

I think it is equivalent to what you would get with the Oustaloup approximation for \$s^\gamma\$, which is described here for example.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.