2
\$\begingroup\$

If we have a transfer function that shows no peaking in the magnitude bode plot (Starting from a flatline and then rolling off). Does this mean that there is no resonant frequency? Or do we consider the point at which the curve begins to roll off the resonant frequency?

I understand that resonant frequency is the location at which we have the maximum value so I'm assuming that there isn't a resonant frequency in this case but I wanted to be sure.

\$\endgroup\$
7
\$\begingroup\$

My answer applies to higher-than-1st-order systems.

There will always be a resonant point even if you can't see it. You need to understand how "poles" work. Take a look at this: -

enter image description here

Even if there doesn't appear to be a resonance in the bode plot there will be a "pole" that is present and this pole represents the resonant frequency even though the "dampening" is causing it not to appear in the bode plot. Here is what a 2nd order low pass filter looks like with varying degrees of Q (where Q = \$\dfrac{1}{2\zeta}\$): -

enter image description here

If you could determine the phase angle where the output shifted by 90 degrees from the input you would find the resonant frequency even if it doesn't appear to have a "peak" in the bode plot.

\$\endgroup\$
  • \$\begingroup\$ Where are these images from? \$\endgroup\$ – Matt Young May 30 '14 at 18:05
  • \$\begingroup\$ @MattYoung I drew the top one for a presentation I did a few years ago - it's a bit scruffy and could do with a redraw. The lower image I did get from the web sometime last year but I can't remember. \$\endgroup\$ – Andy aka May 30 '14 at 18:14
  • \$\begingroup\$ Nice pictures. Good explanation - however, I wouldn´t call this a "resonant point". This could be misleading because in each textbook and all other relevant papers this effect is explained using the pole location in the complex frequency plane. This is clearly shown in Andy´s comment. However, one important point: The peaking in the amplitude response occurs NOT at the pole fequency, unless it is a bandpass response. But for lowpass and highpass functions the peaking occurs in the vicinity of the pole frequency. However, with rising Q values the maximum gets closer to the pole frequency. \$\endgroup\$ – LvW May 30 '14 at 19:37
  • \$\begingroup\$ @LvW - the "This is clearly shown in Andy´s comment" is confusing - did you mean some of this comment for me and some for someone else dude (there goes that word again!!) \$\endgroup\$ – Andy aka May 30 '14 at 20:19
  • \$\begingroup\$ @Andy aka: I was referring to your contribution showing relevant pictures which explain the relation between the poles in the complex s-plane and the corresponding Bode plots. \$\endgroup\$ – LvW May 31 '14 at 7:56
2
\$\begingroup\$

Peaks in the frequency response can only exist in systems with conjugate complex poles. For an underdamped (\$\zeta<1\$ or \$Q > 0.5\$) second-order system, the peak appears specifically for \$\zeta<1/\sqrt{2}=0.707\$.

$$H(s)=\frac{\omega_n^2}{s^2+2\zeta\omega_ns+\omega_n^2}$$

where \$\omega_n\$ is the natural frequency (also called corner frequency when considering assymptotes), the peak

$$M_p=\frac{1}{2\zeta\sqrt{1-\zeta^2}}$$

occurs at resonant frequency

$$\omega_p=\omega_n\sqrt{1-2\zeta^2}$$

Note on figure below: When varying the damping ratio \$\zeta\$, the peak follows a specific curve. In filter theory, that special value for \$\zeta=0.707\$ corresponds to a Butterworth response. The magnitude curve is sais to be maximally flat (no peak). The meaning of \$w_n\$ for the Butterworth response is the same as for the first-order case, that is, \$w_n\$ represents the -3 dB frequency, also called cuttoff frequency. Only in this case. Also, \$w_n=w_p\$, causes an infinite response (undamped system - oscillator).

SecOrderPeak

\$\endgroup\$
  • \$\begingroup\$ Coming back to the original question ("resonant frequency") the above contribution clearly shows the difference between "resonant frequency" and "pole frequency". For my opinion, we can speak about a "resonant" effect only in case a frequency dependent network shows a phase shift of zero at one frequency only (RC bandpass or LC tank). However, the transfer functions as discussed above exhibit at the pole frequency a phase of -90 deg. \$\endgroup\$ – LvW May 31 '14 at 8:02
  • \$\begingroup\$ @LvW: I agree with your comment:**The peaking in the amplitude response occurs NOT at the pole fequency, unless it is a bandpass response. But for lowpass and highpass functions the peaking occurs in the vicinity of the pole frequency. However, with rising Q values the maximum gets closer to the pole frequency.** \$\endgroup\$ – Dirceu Rodrigues Jr May 31 '14 at 21:37
0
\$\begingroup\$

It is absolutely correct to say that there is no resonant frequency defined if the system is sufficiently underdamped.

Is there a mathematical threshold for damping constant below which the system will just have no resonance? Yes. D = 2^(-0.5) = 0.707 approx. ( D = damping constant)

For a general second order system with transfer function as : Wn^2/{s^2 + 2*sDWn + Wn^2} Wn = undamped natural frequency.

You can compute the resonance frequency Wr by differentiating w.r.t Wn and equating the result to 0. The result will be : Wr = Wn*sqrt{1-2D^2} which can only be real if D > 1/sqrt{2}

Hope I have answered your query to satisfaction.

\$\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.