Sax on the Web Forum banner

Spectrum Analysis

2K views 9 replies 6 participants last post by  Gange 
#1 ·
I'm working on a capstone project for college where I am comparing three instruments of the same make and model to determine how similar their spectra end up being. I have multiple samples of each instrument on a variety of sustained notes, and want to eventually have a set of graphs comparing their spectra. Does anyone here know of a program I could use to export a graph as a text file that contains the average values of a spectrum over time? Essenially I'm looking for the data of a spectrogram but as an averaged graph as opposed to a heatmap looking image. Spectrograms don't look precise enough and if I can export as a graph or text file of amplitude v frequency then I can do further analysis and comparisons from there.
 
#3 ·
Almost any DAW program will have an EQ plugin that will display what you want. A frequency spectrum as a line graph, with loudness in dB on the vertical axis and frequency on the horizontal. I don't think it will export the values though, so maybe you might have to manually write down the values at various points if you need to do an analysis of the values.
 
#4 ·
Finding programs that have the one spectrum graph is not the main problem. It's the fact that those are instants in time, as opposed to the spectrogram's time axis. In an attempt to minimize the influences of the player, I would want as accurate an average as possible for the graphs, hence my desire to export and compare them.
 
#5 ·
The variations introduced by players is a huge problem here. The human factor.

Beyond the human factor is the change of mouthpiece from horn to horn....any difference and another big variable is introduced.

I would suggest same neck on each horn. Then the mp and reed stay the same.

Remember the scientific method. Only one thing can be changed at a time for a valid experiment. Unfortunately, with a human blowing, that variable of him or her going from sax to sax really mucks things up, as would change of neck and the accompanying mp repositioning on the different necks if you don't use the same setup moving from sax to sax.

I have worked projects such as this without the human factor involved, and that can be challenging. With the human factor...well, I'd be picking a more controllable experiment unless the interest of a particular and influential person involved precludes that.

A robot blowing the saxophones would be a much more controlled experiment and could have some real validity.

 
#6 ·
This is for trombone, luckily. I would never try to test this kind of thing on sax, too many variables. Trombones locked in first position with the same mouthpiece are much simpler and less prone to error. I've tried to build an artificial embouchure to remove the human element entirely, but it's not functional for this. The data has been collected and the way the experiment was done is set in ink. Nevertheless, the human element is present and is mitigated by averaging as accurately as possible. That's my current hiccup.
 
#7 ·
Why not compile the spectrum for each snapshot in time (in my line of work, we would call this as sample but I don't want to confuse this with the recordings themselves which could also be correctly referred to as samples) from each sample then average the results after the fact? Matlab was my go-to tool for anything computationally intensive and I did a lot of signal processing labs using it, but I don't know if it has a function (either as a part of the basic software or a special toolbox) that can run a fast Fourier transform on an imported audio file. I would be surprised if it didn't though.

One problem that you may run into is that I doubt that the fundamental pitch will be held perfectly steady if a human is doing the playing so the exact frequencies of the harmonic content will change with it but you probably already figured this out.
 
This is an older thread, you may not receive a response, and could be reviving an old thread. Please consider creating a new thread.
Top