Experiments: kick off

Welcome to Falasol Experiments series.

Here we’ll try to take the application to the limits and try to answer some intriguing questions related to the nature of sound.

This is the first – introducing post and the goals are:

  • get familiar with selected spectral analysis applications
  • compare the Falasol’s graph of tutorial sound with it’s output spectrograms

Preparation

If the word spectrogram doesn’t tell you much, please check out this link: http://en.wikipedia.org/wiki/Spectrogram.

We’ll also need some tools to perform the spectral analysis of sound samples. I’m giving out Falasol for free and I’m not planning to pay for anything neither. The best free stuff I was able to find:

OK, we’ve got the equipment. The only part missing is the sample. I encourage you to generate it from tutorial steps available in help file (distributed with Falasol). You can also download it here:

Execution

Let’s start with the visualization given by Falasol. Here’s what you can see on the Spectrum Graph (with Envelope influence on and log frequency scale):

Tutorial graph

Tutorial sound - Falasol graph

Now we’re going to find out whether it matches the spectograms obtained from already generated sound. I used the sampling window of 2048:

Tutorial sound - SPEAR spectrogram

Tutorial sound - SPEAR spectrogram

Besides different xy scaling and color saturation (which is probably result of log scale used) the shape looks similar. Lowest harmonic looks a bit flattered because linear frequency scale was used for x axis. SPEAR also detected some additional resonating phantom frequencies (pale lines around the highest harmonic) which are just noise in this case.

Tutorial sound - Sonic Visualiser spectrogram

Tutorial sound - Sonic Visualiser spectrogram

Here you can see the linear color scale and log frequency – same as in Falasol’s graph. The lowest harmonic is fuzzy – this is because lower frequencies would require bigger sampling window to be captured sharply (however in that case the sinusoidal change would become more blurry).

What I really like about Sonic Visualiser is the ability to customize the visualizations. Here is same example with dBV^2 color scale, default palette and linear frequency axis:

tutorial-sonicvisualiser-cool

Tutorial sound - Sonic Visualiser cool spectrogram

As we can see the spectrograms more or less matches the expected sound graph given by Falasol. We have also observed the characteristic noise artifacts of spectral analysis and noticed the role of color and frequency scales.

I realize this article was not much of a big discovery, but every journey has to start somewhere :)

Tags:

Reader's Comments »

  1. By Margaret on June 18, 2009 at 8:35 pm

    Thanx for detailed explanations:P

    By koshik on June 18th, 2009 at 11:05 pm

    I hope this was not sarcastic :]
    Currently I’m in the middle of writing next post: “slurring vs. crossfading”.
    The spectrograms already look quite cool, all should be ready some time next week.

    Regards from the Land of Fire & Ice *<:)