Dismiss Notice
Join Physics Forums Today!
The friendliest, high quality science and math community on the planet! Everyone who loves science is here!

MATLAB: Fast Fourier Transform

  1. Apr 11, 2010 #1
    I have run the following command:

    c = wavread('sample.wav');
    amplitude = log (abs(fft(c)));

    and obtained the following plot:

    [PLAIN]http://img179.imageshack.us/img179/8733/withoutplusone.jpg [Broken]

    however, i was told to use this instead:

    amplitude = log (1+abs(fft(c)));
    and obtained the following plot:

    [PLAIN]http://img146.imageshack.us/img146/4594/withplusone.jpg [Broken]

    my question is: why do we have to take the log 1 plus amplitude, instead of just the log of amplitude.

    Last edited by a moderator: May 4, 2017
  2. jcsd
  3. Apr 11, 2010 #2
    do you mean that the reason is because we cant log (0)?

    is that the only reason we are adding 1 to the log?
  4. Apr 20, 2010 #3


    User Avatar
    Science Advisor
    Gold Member

    I don't know why you would add 1. I never add 1, and plot log(abs(fft)) all the time (usually I do log10 because I want to get to decibels). If you are worried about zeros, you can always add eps or some other small number. By looking at your plot it is clear the adding 1 is covering up information you may care about. You should ask the person who suggested it to you.

  5. Apr 21, 2010 #4
    May I ask what you are trying to learn from the data? It would help us to answer your question. Looking for peaks in the spectrum? Total noise?
  6. May 1, 2010 #5
    hi guys, thanks for the reply.

    the purpose is to observe the symmetry property of Fourier transform and also to explain the spikes that is occurring.
  7. Jul 21, 2010 #6
    It's symmetrical because the inputs are real. If the sample.wav contained complex numbers, the output would not (necessarily) be symmetrical.
Share this great discussion with others via Reddit, Google+, Twitter, or Facebook