Topic by Jake Johnson

  • Avatar
  • Jake Johnson
  • Member
  • Offline
  • From: Atlanta
  • Registered: 04-05-2007
  • Posts: 4,491

Topic: Near frequencies and the display on analysis programs?

When I put wave files of single piano notes through different freq analysis programs, I get different results, depending on the way the program displays its results. Some programs, such as Overtone Analyzer, show a wide band of what I call "near frequencies" around each fundamental and each additional partial, indicating that many  frequencies above and below the fundamental and other partials occur when a single note is struck. Other programs, such as Spear, show only one or two of these near freqs above and below each fundamental and partial. (Goldwave, on the other hand, shows a bar chart of spikes and declines).

How does what these programs show relate to the reality? When a single note is struck, are there usually 2-4 freqs on each side of the fundamental and paritals that are also given off (but some analysis programs show a continuous color band from the main freq that extends to those on either side) or are there a large number of freqs (from transwaves, sympathetic resonances from nearby strings and their partials, the detuning of unison strings, etc) given off by playing each note (but some programs simplify the display so that only those freqs above a given amplitude are displayed, and thus have separate bands)?

Reply by Philippe Guillaume

  • Avatar
  • Philippe Guillaume
  • Admin
  • Offline
  • Registered: 25-04-2007
  • Posts: 5,250

Re: Near frequencies and the display on analysis programs?

The displayed result depends on the windowing used during the discrete Fourier analysis (FFT). Roughly speaking, the analysis program analyses short sequences, say of duration T. Then the frequency step or “resolution” (the smallest difference in frequencies that can be measured by the FFT) is 1/T. For example, with a 100 ms sequence, the frequency step is 10 Hz. For that reason, one can hardly see the tiny gaps between the 3 strings frequencies, because these gaps lie between 0 and at most a few Hz (unless the piano is badly out of tune).

Philippe Guillaume's Website

Reply by Jake Johnson

  • Avatar
  • Jake Johnson
  • Member
  • Offline
  • From: Atlanta
  • Registered: 04-05-2007
  • Posts: 4,491

Re: Near frequencies and the display on analysis programs?

I hate it when I have to admit how slow I am to understand:

1. So the reality it that , with an assumed slight detuning of unison strings, there are many, too close to differentiate, frequencies, but some programs ignore those beneath a given amplitude threshold and render only the louder frequencies as distinct bands, creating the impression that there are fewer distinct bands? (Spear, of course, lets you increase its amplitude sensitivity, but it always seems to show and render fewer near freqs than those that are implied by the graphic display of programs like Overtone Analyzer and Goldwave, which display a continuous set of frequencies that range from the fundamental and other partials to the outside range of the detuned unisons.)

2. More generally, if a single note in the bass was struck, and only one string was there, what frequencies would be present after the transients had subsided? Just the fundamental and the other partials, if the sympathetic resonance, and the nearby and thus rattled strings are removed? In other words, does the isolated single string give off a very narrow frequency band, or is there always a wider band of frequencies because of transwaves, or a refusal of a physical object to vibrate at the same rate over time, than is implied by A4=440?

These questions come partly from just a general interest, but I'm also trying to answer the simpler question--if we don't have matlab, what is the best program to use for analysis--one that renders clear bands for the strongest frequencies or one that renders a continuum of frequencies, but may make it harder to determine the most significant bands? Both ways of rendering the frequencies seem to have limitations. Better to use both?

Reply by julien

  • Avatar
  • julien
  • Admin
  • Offline
  • Registered: 19-04-2007
  • Posts: 1,000

Re: Near frequencies and the display on analysis programs?

There are some excellent open-source alternatives to Matlab: the first one is octave http://www.octave.org which tries to stay as close as possible to Matlab -- for the purpose of signal processing it provides most of the tools required. And there is scilab  http://www.scilab.org which is similar but not Matlab compatible. Scilab also has a stronger industry support than octave, and is more polished.

julien's Website

Reply by Jake Johnson

  • Avatar
  • Jake Johnson
  • Member
  • Offline
  • From: Atlanta
  • Registered: 04-05-2007
  • Posts: 4,491

Re: Near frequencies and the display on analysis programs?

Julien,

These programs, like Matlab, scare me. I don't know where to start. Has anyone created, or used them to create, relatively simple programs that can input a wave file and output a text file or graphical representation of frequencies and their amplitudes?

Or are there guides that could help me to find a way to create these?

What are your impression of the existing high-level programs? (Spear, Goldwave, Overtone Analyzer?)

Reply by julien

  • Avatar
  • julien
  • Admin
  • Offline
  • Registered: 19-04-2007
  • Posts: 1,000

Re: Near frequencies and the display on analysis programs?

Yes you can probably find many tutorials on signal processing with matlab on the net.

There are also some very simple tutorials in the book of Philippe ! Those tutorials do not assume any knowledge of matlab, and are very easy to follow

http://www.amazon.com/Music-Acoustics-I...1905209266

I can't say much about the softwares you cite, never used them. They are probably way more convenient than matlab, but not as flexible

julien's Website