This question is a part of a more general question the answer of which I don't know - How to apply a filter in the freq domain and then convert the filtered signal back to the time domain? Well, I partially googled the answer that I need to

  • convert the signal in FFT
  • multiply by the filter
  • convert back to the time domain

but I'm not entirely sure if I've applied this idea correctly to atmospheric absorption filtering (see below).

I'm also not sure whether I can do the filtering entirely in the time domain (as suggested here by convolution?) somehow alleviating the need to switch back and forth to the freq domain.

Input: sound wave (gunshot sound pressure subtracted atmospheric pressure) and atmospheric conditions. Output: the same sound wave attenuated with an atmospheric absorption filter.

I'm using python-acoustics to find the atmospheric attenuation coefficient as described in Engineering Acoustics/Outdoor Sound Propagation.

I've came up with the following code:

def atmosphericAttenuation(signal, distance, Fs, **kwargs):
    Apply atmospheric absorption to the `signal` for all its FFT frequencies.
    It does not account for the geometrical attenuation.

    signal - a pressure waveform (time domain)
    distance - the travelled distance, m
    Fs - sampling frequency of the `signal`, Hz
    kwargs - passed to `Atmosphere` class

    signal_attenuated - attenuated signal in the original time domain
    # pip install acoustics
    from acoustics.atmosphere import Atmosphere

    atm = Atmosphere(**kwargs)
    signal_rfft = np.fft.rfft(signal)
    freq = np.fft.rfftfreq(n=len(signal), d=1. / Fs)
    # compute the attenuation coefficient for each frequency
    a_coef = atm.attenuation_coefficient(freq)
    # (option 2) signal_rfft *= 10 ** (-a_coef * distance / 20)
    signal_rfft *= np.exp(-a_coef * distance)
    signal_attenuated = np.fft.irfft(signal_rfft)
    return signal_attenuated

Am I doing it right? Which one is correct:

  • signal_rfft *= np.exp(-a_coef * distance) <- $P(r) = P(0) \exp (-\alpha r)$
  • signal_rfft *= 10 ** (-a_coef * distance / 20) <- $A_a = -20 \log_{10} \frac{P(r)}{P(0)} = \alpha r$

If neither, please describe how it should be done. Thank you.


1 Answer 1


FFT has a fixed frequency resolution and I don't recommend to do such a frequency modification with FFT. See more at this question.

You may use Atmosphere.impulse_response to obtain the impulse response and then apply a time-domain convolution, which gives a more reasonable result.

Which one is correct:

  • signal_rfft *= np.exp(-a_coef * distance)
  • signal_rfft *= 10 ** (-a_coef * distance / 20)

According to the documentation, attenuation coefficient $\alpha$ describing atmospheric absorption in dB/m as function of frequency. So the second one is correct.

  • $\begingroup$ Thank you! Do you mean ir = atm.impulse_response(distance=distance, fs=Fs, ntaps=len(signal)); signal_attenuated = np.convolve(signal, ir, mode='same')? Indeed, this looks better. $\endgroup$
    – dizcza
    Jun 28, 2021 at 12:14
  • $\begingroup$ @dizcza yes that’s right $\endgroup$
    – ZR Han
    Jun 28, 2021 at 12:25

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.