1
$\begingroup$

I just want to check that my understanding about the following paragraph from Optical Filter Design and Analysis by Christi K. Madsen, Jian H. Zhao is correct:

A filter’s magnitude response is equal to the modulus of its transfer function, $|H(z)|$, evaluated at $z = e^{j\omega}$. Based on the pole/zero representation of $H(z)$, only the distance of each pole and zero from the unit circle, i.e. $|e^{j\omega}-z_m|$ or $|e^{j\omega}-p_n|$, affects the magnitude response. Consequently, a zero that is located at the mirror image position about the unit circle, i.e. $1/z_m^*$, cannot be differentiated from $z_m$ based on the magnitude response.

Let us assume a filter with just one zero $c_1=|c_1|e^{j\phi}$ so that its transfer function is: $$ H(z)=1-c_1 z^{-1}=1-\lvert c_1\rvert e^{j\phi} z^{-1} $$ Its mirrored zero is $c_2=1/c_1^*=\lvert c_1\rvert^{-1}e^{j\phi}$ and its transfer function is $$ H_m(z)=1-c_2 z^{-1}=1-\lvert c_1\rvert^{-1}e^{j\phi} z^{-1} $$ Evaluating the transfer function at $z=e^{j\omega}$ and computing $H(e^{j\omega})H^*(e^{j\omega})$ we get the power response: $$ \big\lvert H(e^{j\omega)}\big\rvert^2=\big\lvert 1-\lvert c_1\rvert e^{j(\phi-\omega)}\big\rvert^2=1 +\lvert c_1\rvert^2-2\lvert c_1\rvert\cos(\phi-\omega) $$ and $$\big\lvert H_m(e^{j\omega})\big\rvert^2=\bigg\lvert 1-\frac{1}{\lvert c_1\rvert}e^{j(\phi-\omega)}\bigg\rvert^2=1 +\frac{1}{\lvert c_1\rvert^2}-2\frac{1}{\lvert c_1\rvert}\cos(\phi-\omega) $$

$\big\lvert H_m(e^{j\omega})\big\rvert^2$ can also be expressed as: $$ \big\lvert H_m(e^{j\omega})\big\rvert^2=\frac{1}{\lvert c_1\rvert^2}\Big[1 +\lvert c_1\rvert^2-2\lvert c_1\rvert\cos(\phi-\omega)\Big]=\frac{1}{\lvert c_1\rvert^2}\lvert H(e^{j\omega)}\rvert^2 $$ which is just the power response of the first filter scaled by a factor of $\lvert c_1\rvert^{-2}$. Is this analysis right?

$\endgroup$
1
$\begingroup$

You're right, the contribution of a zero $z_0=re^{j\phi}$ to the squared magnitude response is

$$\big|e^{j\omega}-re^{j\phi}\big|^2=1-2r\cos(\omega-\phi)+r^2\tag{1}$$

From $(1)$ it is clear that another zero resulting in the same frequency dependence as in $(1)$ must have the same phase angle $\phi$. Assuming $z_1=Re^{j\phi}$ we get

$$\big|e^{j\omega}-Re^{j\phi}\big|^2=1-2R\cos(\omega-\phi)+R^2=R^2\left(1-\frac{2}{R}\cos(\omega-\phi)+\frac{1}{R^2}\right)\tag{2}$$

Comparing $(1)$ and $(2)$ we get $R=1/r$ and a squared magnitude scaling of $R^2=1/r^2$. Consequently, the other zero $z_1$ is given by

$$z_1=\frac{e^{j\phi}}{r}=\frac{1}{z_0^*}\tag{3}$$

and

$$\big|e^{j\omega}-z_0\big|=r\left|e^{j\omega}-\frac{1}{z_0^*}\right|\tag{4}$$

$\endgroup$

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.