From this paper (slide 11), the insertion loss (or attenuation) is defined as

$$ L(\omega^2)=\frac{\lvert V_i(j\omega)\rvert^2}{\lvert V_o(j\omega)\rvert^2}=\frac{1}{\lvert H(j\omega)\rvert^2}=10 \log\left(\frac{1}{H(j\omega)H(-j\omega)}\right) $$

I'm interested specifically in, where does the last inequality follow?

  • $\begingroup$ The last equality does not follow. The first three terms are in linear scale, the last one is logarithmic. $\endgroup$
    – Peter K.
    Jun 27, 2016 at 11:17
  • $\begingroup$ @PeterK. This is given here d-filter.ece.uvic.ca/SupMaterials/Slides/DSP-Ch10-S1-7.pdf p. 11. $\endgroup$
    – mavavilj
    Jun 27, 2016 at 11:39
  • $\begingroup$ Well then it's complete rubbish. $\endgroup$
    – Peter K.
    Jun 27, 2016 at 12:02
  • $\begingroup$ Looks like a typo. The 10 log() crept in one line to early $\endgroup$
    – Hilmar
    Jun 28, 2016 at 20:02

1 Answer 1


As pointed out in PeterK.'s comment, the way it is written this does not make sense. What is meant here is that the insertion loss can be measured on a linear scale (i.e., as a factor) as


or it can be given in decibels:


The remaining question could be why $|H(j\omega)|^2$ can be replaced by $H(j\omega)H(-j\omega)$. The reason is the symmetry of the frequency response $H(j\omega)$ for real-valued filters:


From this it follows that



Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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