I wanted to measure the distortion of my function generator, and I discovered my oscilloscope seems to distort the signal slightly (always second and third harmonics around -40dB even after filtering). So I recorded the function generator as audio and tried two different FFT options to measure the distortion.

The first is an audio spectrum analyzer plugin which gave about 0.3% THD. This sort of thing isn't really intended for distortion analysis, but I figured FFT is pretty straight forward and there's not a lot of room for variation right?

Then I put the same audio file in Matlab and ran the FFT there, which gave me 0.06% THD. Now I wonder which to trust.

 

Hello,

What kind of window is the FFT using?
In audacity you can choose several types of FFT windows, like Bartlet, Hamming, Hanning, Blackman and Blackman-Harris :
http://www.audacityteam.org/

Bertus

 

The plugin I used was Voxengo Span, and it has no window options nor does the product info mention which window: http://www.voxengo.com/product/span/

Adjusting the settings that are available doesn't change the relative harmonic levels.

In Matlab I used a Blackman window.