i got a system with transfer function given by:

$$H(\omega)=1-e^{-j\omega}$$ I already plot it, and that seems to be a periodic function with $H(0)=0$, $H(\pi)=2$, , is that enough to show that this is a FIR bandpass filter. thank you very much

  • 1
    $\begingroup$ That is a highpass filter. How did you calculate $H(\infty)$? Since a digital filter has a response that is periodic, it doesn't really make sense to try to calculate its response at infinite frequency. $\endgroup$ – Jason R Apr 3 '13 at 15:20
  • $\begingroup$ You need to consider what the definition is for a LP/BP/HP filter in the case of a discrete time system. Consider that the useful/meaningful range of frequencies maps to w being in the range 0-pi. $\endgroup$ – user2718 Apr 3 '13 at 15:45
  • $\begingroup$ @BZ For complex filters (which this is) the range is $0$ to $2\pi$ or, equivalently, $-\pi$ to $\pi$. $\endgroup$ – Jim Clay Apr 3 '13 at 16:13
  • $\begingroup$ @JimClay - Actually this is not a complex filter. All the coefficients are real. The term $$e^{-j\omega}$$ represents a real delay element. Certainly the frequency response is defined by a complex function, but most filter defintions are based only on the non-negative portion of the response. Not a big deal though, I think the proper point has been passed on to the OP. $\endgroup$ – user2718 Apr 3 '13 at 17:04

As was pointed out in the comments above, this is a highpass filter. There is no strict definition for what constitutes a lowpass, bandpass, or highpass filter. Instead, they possess certain qualities:

  • Lowpass filters attenuate high frequencies while passing low frequencies.

  • Bandpass filters attenuate low and high frequencies, while passing some band in the middle.

  • Highpass filters attenuate low frequencies while passing high frequencies.

Note the lack of precise terms in the descriptions above. In order to classify a filter into a category like those above, one evaluates the character of its magnitude transfer function. That requires a little manipulation:

$$ H(\omega) = 1 - e^{-j\omega} $$

$$ \begin{align} |H(\omega)| &= |1 - e^{-j\omega}| \\ &= \left|e^{-j\frac{\omega}{2}} \left(e^{j\frac{\omega}{2}} - e^{j\frac{\omega}{2}}\right)\right| \end{align} $$

Using Euler's identity, we can simplify the expression inside the parentheses:

$$ e^{j\frac{\omega}{2}} - e^{j\frac{\omega}{2}} = j2\sin\left(\frac{\omega}{2}\right) $$

$$ |H(\omega)| = \left|e^{-j\frac{\omega}{2}} j2\sin\left(\frac{\omega}{2}\right)\right| $$

We can drop the $j$ and $e^{-j\frac{\omega}{2}}$ terms here, because each of them have unity magnitude and therefore don't affect $|H(\omega)|$. Therefore:

$$ |H(\omega)| = \left|2\sin\left(\frac{\omega}{2}\right)\right| $$

We can check this result by comparing it with a plot from MATLAB's Filter Visualization Tool (fvtool):

enter image description here

Which looks about right. This is plotted over the region $\omega \in [0, \pi]$, and shows that the filter has zero response at DC (attenuating low frequencies) and a magnitude response of $2$ at $\omega = \pi$ (passing high frequencies). That looks like a highpass filter.

| improve this answer | |

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.