I'm trying to model a transfer function in Python and thought I could do it by simply plotting the transfer function at many frequencies. This seemed to work for a 2nd order LPF. See the below figure.

Magnitude response of 2nd Order LPF.

A bit of sample code would be like:

s = np.arange(10 * 2 * np.pi, 1000 * 2 * np.pi, 10) # Creates vector of frequencies in rad/s
lpf = np.zeros(len(s))  # Initializes lowpass filter vector

for k in range(len(s)):
   lpf[k] = lpfilt_func(s[k])  # Calls LPF function with TF = cutoff_freq ** 2 / (s + cutoff_freq) ** 2

To get the frequency response, instead of using s as a real number, I changed s to a complex number using the built-in function complex(0, s) and calculated the magnitude and phase as follows:

for k in range(len(s)):
   lpf[k] = lpfilt_func(complex(0, s[k]))
   gain_lpf[k] = np.absolute(lpfilt_func(complex(0, s[k])))
   phase_lpf[k] = np.angle(lpfilt_func(complex(0, s[k]))) * 180 / np.pi  # Phase in deg

I think this is essentially like setting $s = j\omega$ and then calculating the magnitude as $|LPF| = \sqrt{A^2 + B^2}$ and the phase response as $\angle{LPF} = \angle{(B/A)}$.

Is this a valid method to calculate the frequency response of a transfer function? Is this conversion of s from a real number to a complex number valid?


If the response is real you need not convert it to complex in order to compute the magnitude, but the process is correct for complex and real number since the abs function will use complex conjugate product prior to taking the square root. If it was real, there would be a trivial phase response since the phase would always be 0 degrees. This would not be a causal system so not a realistic example to have a real frequency response. For complex cases the approach used is correct for magnitude and phase, just not the most efficient.

To confirm results (or to compute directly), use the freqs and freqz commands that are part of scipy.signal for continuous-domain (s) and discrete-domain (z) transfer functions.

For example the transfer function

$$H(s) = \frac{s+1}{s+4s+4}$$ could be plotted using:

w,h = scipy.signal.freqs([1, 1], [1, 4, 4])

Returning the frequency (w) and complex response (h)

Read the help for both of those functions (also available under same name in MATLAB and Octave) for further info.

For MIMO systems the control package has this all worked out as part of the bode function.

  • $\begingroup$ I'm currently working with a MIMO system and was wondering if my method above could work. I simplified the question for ease of asking. Could those functions you mentioned be used for MIMO systems? $\endgroup$ – Quyed Sep 30 '20 at 21:41
  • $\begingroup$ @Quyed I am not sure but the control package that is available for Python certainly does that with the bode function $\endgroup$ – Dan Boschen Sep 30 '20 at 22:12

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.