MATLAB Why Use Log(1+Amplitude) in FFT?

[McMurry]

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

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

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

thanks.

 

[McMurry]

there are 0s :roleeyes:

do you mean that the reason is because we can't log (0)?

is that the only reason we are adding 1 to the log?
thanks.

 

[jasonRF]

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.

jason

 

[MRFMengineer]

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?

 

[McMurry]

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.

 

[Sean Ehle]

It's symmetrical because the inputs are real. If the sample.wav contained complex numbers, the output would not (necessarily) be symmetrical.