MATLAB Answers


Problem with plotting a phase response of a time-series data FFT

Gustavo da Silva Lima さんによって質問されました 2019 年 6 月 17 日
最新アクティビティ Star Strider
さんによって コメントされました 2019 年 6 月 17 日
I'm trying to compare a time-series FFT to a transfer function frequency response. For this, I'm using the commands "abs" and "phase" to colect magnitude and phase, respectively, from the data and from the transfer function (using freqresp( ) command). However, when I plot the both graphs, the magnitude is ok, but phase response is not. It's returning to zero phase at the Nyquist Frequency (this frequency is the last one of my frequency vector).
See my code bellow :
Ryu_fft = fft(Ryu)/Ryu_length; %Obtainnig the FFT of my time-series data
Ryu_fft_dB = 20*log10(abs(Ryu_fft(1:NFFT/2+1))); %Evaluating the magnitude response
Ryu_fft_phase = phase(Ryu_fft(1:NFFT/2+1))*180/pi; %Evaluating Phase response in degrees
f = linspace(0,1,NFFT/2+1)
Now, see my semilogx() plot of this:
How could I fix this problem?
I suspect it is only a code problem once the magnitude is ok.

  0 件のコメント

サインイン to comment.

1 件の回答

Star Strider
回答者: Star Strider
2019 年 6 月 17 日

See if using the unwrap function on the phase vector (in radians, before transforming it to degrees) has the desired effect.
I do not recognise the phase function, since it appears only in:
C:\Program Files\MATLAB\R2019a\toolbox\ident\idobsolete\phase.m

  2 件のコメント

Gustavo da Silva Lima 2019 年 6 月 17 日
Thank you for help me. Unfortunately the command "unwrap", implemented as you suggested, is not solving my problem.
About the command "phase"...yes, I'm using this to identify a dynamic system. I've tried to use angle() and atan2(), but "phase" looks better.
Star Strider
2019 年 6 月 17 日
My pleasure.
Oh, well.
It was just an hypothesis that phase-wrapping was the problem. Not enough information for any other Answer.

サインイン to comment.

Translated by