Official Luthiers Forum!

Is there an "Audacity for Dummies" (me!) outlining the procedure for capturing and analyzing tap tones? I'm thicknessing my plates for my first build challenge and thought I'd experiment with Trevor Gore's thickness targeting procedure. My first attempts at this don't make any sense, and before I question my data, I need to know if I'm going about it right...

Thanks!
Dan

Here's some info I copied out of a thread a couple of years ago from David Malicky:=====================================================

Post subject: Re: Looking for an FFT software package

One more: Audacity. It's not nearly as flexible as Wavesurfer (which I also like) but if you're mainly looking for the peak frequencies, I find Audacity is less quirky and easier to use.

My main problem with Wavesurfer is that I can't get it to do the FFT on the selected signal. That is, if I click and drag to create a dotted yellow selection, the Spectrum Section plot doesn't account for the full selected region, even if I click the "Average of selection" box. It only analyzes from where my selection starts (the vertical yellow line on left) to the ~0.5 seconds afterwards to get enough data for the 8192 FFT points (what I use, with 16k samples/sec).

A quick-start on Audacity:
- Install as usual. http://audacity.sourceforge.net
- In Audacity: Edit > Preferences > Audio I/O > Select your recording device > OK.
- Record the taps.
- Select the tap(s) signal with the mouse
- Analyze > Plot Spectrum
-- Change 512 to 16384
-- Change Linear Frequency to Log Frequency
-- Drag the right window edge to make it a *lot* wider
-- Hover the cursor near the peaks... Audacity will automatically lock onto the peak frequencies! Numbers are in the lower left. I've found it's within 1 Hz of Wavesurfer for the air, top, and back modes.

_________________
David Malicky (david82282)
==================================================

Here's the whole thread: http://www.luthiersforum.com/forum/viewtopic.php?f=10101&t=24630

and here's another interesting associated thread: http://luthiersforum.com/forum/viewtopic.php?f=10101&t=31330

_________________
Gene

Politicians and diapers must be changed often, and for the same reason- Mark Twain

I'm with you Dan. A tutorial would be nice

_________________
Joe Beaver
Maker of Sawdust

Are you looking for info on what you are seeing or just getting the FFT. I do as what David says and have no problems, do a series of 6 to 10 taps and select them all and plot the spectrum/fft graph. The highest peak is the one you want the frequency of. From this point what you deduce from what you are looking at is experience.

I made an Excel spread sheet of Trevor's formula and proved it out using the table of results in the book. It works great but is easy to mess up with bad input due to inaccurate measurements or tap frequencies.

_________________
Fred Tellier
http://www.fetellierguitars.com
Facebook page http://www.facebook.com/pages/FE-Tellier-Guitars/163451547003866

Whichever spectrum analyser program you use, it has to be set up properly. If you want to use Audacity and get plots reminiscent of those in the book, when in Audacity you will need to go to Edit\Preferences... and change the default sample rate to 11025. Then when you plot the spectrum (having selected whatever you want to analyse) you need to select 16384 in the buffer size box (or whatever the top right box is called in your version). Then change to "linear frequency" in the box below the previous one. As you do all these things you will see how the chart changes for the same input data. If you export that data and save the file, the frequency increments in the text file should be 0.67Hz. That's the check that you got it right.

One of the down sides of using Audacity is that you can't zoom the spectrum window. Or at least I haven't found a way of doing that. That's one of the reasons I mainly use VA. Instruction for setting up VA are here.

If you just want to measure tap tones to collect material properties data, you only need a frequency meter. There's one in VA, but Audacity will do, if you want to use it. I use G-tune, which is still available as a demonstration version, downloadable from this page.

_________________
Trevor Gore, Luthier. Australian hand made acoustic guitars, classical guitars; custom guitar design and build; guitar design instruction.

http://www.goreguitars.com.au

Thanks all! That was, I think, exactly what I needed. To summarize, here is what I did:

In Audacity/Preferences, I changed:

Quality -> Sampling Rate -> 11025 hz
Spectrograms -> Window Size -> 16384

Then I recorded a series of taps using built-in microphone on my MacBook Pro, tapping with my finger.

Here is the spectrogram for the long grain vibration of the rough top plate (spruce, presumably sitka) 3.8mm thick. If I am reading it properly I get 57hz to be fed into the equations.

I'd appreciate any feed back on the appearance of the spectrogram (i.e. does it look like I did it right!) and my interpretation.

Thanks!
Dan

PS Thanks Trevor for the other software recommendations - appears though they are Windows programs, and I use Mac. And really glad about it as my wife has spent the entire morning trying to get rid of the Blue Screen of Death on her work PC...

That's what you should see, now you need to apply the information seen in the display to what you are doing.

Fred

_________________
Fred Tellier
http://www.fetellierguitars.com
Facebook page http://www.facebook.com/pages/FE-Tellier-Guitars/163451547003866

Looks better if you select Linear frequency in the Axis box of the spectrum window. In my version the linear plot ends at about 5kHz. Most of the stuff you can manipulate happens below 1kHz (which is why a zoom option would be useful!)

_________________
Trevor Gore, Luthier. Australian hand made acoustic guitars, classical guitars; custom guitar design and build; guitar design instruction.

http://www.goreguitars.com.au

That is indeed the goal! Took me the better part of the day to get the spreadsheet worked out. (Note to self... 0.02 is a whole lot different than 0.2 )

I don't feel competently confident interpretation the data I'm getting, so I'm resurrecting this thread.

I did ten taps before plotting this spectrum to measure the Cross Grain Vibration Mode. As far as what you can't see in the picture goes, microphone is an Apogee Mic, and I set the sampling rate quality is set to 11025 Hz. I penciled in X's to note my nodal points. The cedar soundboard is 514mm long x 206mm wide x 5mm thick.

When I look at the graph I get, I can't tell which of the three peaks I should think my tap tone is, meaning, I can't determine which value I should plug into Trevor's equation. Again, this one was trying to measure the Cross Grain Vibration Mode. Even though I wrote a value down, the Twisting Grain Vibration Mode also seemed unclear.

I'd hoped certain peaks would jump out as the obvious values, but I'm in need of a tutor, I'm afraid.

As a Mac-user, I was quite intrigued when this thread originally appeared and was inspired enough to do fair amount of experimenting with the software. However, based on a more recent thread, it seems I was chasing a chimera and that Audacity now really is only for dummies. You can read all about it here:

viewtopic.php?f=10101&t=45925&hilit=audacity

_________________
George :-)

George, in that thread, keep in mind that if someone finds Audacity is "inaccurate", it's not because of Audacity. As Trevor has advised, every spectrum analysis program needs to be setup reasonably well (e.g., the FFT size needs to be set fairly high, and the sampling frequency not too high). The tapping method can also influence results, and possibly the mic and computer's audio chip (if low quality). In my recent testing, Audacity matches other FFT programs and frequency meters within 1 Hz or less. This is not surprising, since the math behind FFT analysis is universal and publicly available.

For tap testing, I still prefer Audacity because:
- One can record a track and easily choose which time segment to analyze. For example, I do a sequence of taps -- first on the bridge, then the back (then more specifically if desired) -- all in 1 recording, and analyze each separately or together.
- One can make a series of tracks and easily choose which one to analyze. For example, when I progressively shave braces, it's simple to monitor the effect of each step, and review prior steps at any time.
- In the Frequency Analysis window, the mouse cursor automatically snaps to the nearest peak and displays that frequency. Fast, accurate, and no fine-hovering required.
- The amplitude graph in each track shows the decay shape of each tap.
- It's very easy to use and very stable (while VA is powerful and I like it for other uses, it has been buggy for me).
- It's open-source, so I can install it on any computer without licensing hassles.

I don't know of any other software with those features. Like Trevor, my main complaint with Audacity is that we can't (yet) zoom in to show a frequency range. But with it on log scale and a sampling rate (project rate) around 11 kHz, the relevant tap-testing range from 80 to 500 Hz is clearly visible.

I've made a bunch of youtube screencasts for my regular job -- I can also make one for Audacity / tap-testing, if helpful.

David

_________________
David Malicky

Can't speak for anyone else, but I think a YouTube piece covering this process would be pretty awesome.

_________________
George :-)

I use Audacity and like it. I use a Tascam USB interface and a decent mic. I didn't get good results with my computer sound board - depends on the computer though.

_________________
Steve Smith
"Music is what feelings sound like"

Agreed!

I actually started the thread George linked to. The most practical advice led me to this one. The data I got from testing just isn't clear to me. My results don't seem to reflect the typical range noted in "Build," and I don't have one peak that's clearly a winner.

I have been using VA. You can zoom in to the 0 to 600hz range an get a much better picture of that is happening. Tervor talks about a way to export the data but I have not figured out how to do that with the current VA version. Trevor, if your still around, can you give a quick statement of how to export the data to a file so it can be imported into excel? Thanks,

_________________
_____________

EddieLee

I'm still using Version 9, Eddie. I tried some of the later versions but found them clunky for what I want to do.

In Version 9, from the pop-up spectrum window that appears after your ten taps, select File/Save as Text. Two files get saved. One with an extension .txt, which is the text file full of numbers. Double click on that file and it will open in Notepad or whatever you have your defaults set to for that extension. The text file can be imported into Excel using commands off the Excel Data/Import menu. The other file has a double extension, .txt.wmf. Only the last extension is relevant. It is a Windows Media File which will open in a photo viewer of your choice. It is a "dead" picture, in that it can't be manipulated in VA and is basically a screen shot of whatever was in the pop-up window when you hit save (so make sure you have the pic framed/zoomed as you want it when you save it). The above is relevant for Win XP and Win 7. I don't know what happens in Win 8 or 10.

The later versions of VA save the files in a format that can be re-loaded into VA and so can be manipulated like a "live" VA file (so you can isolate peaks etc. in the image). That's about all I remember about the newer versions other than they took a lot more clicks to get what I wanted. If there is a Save as Text function, I would expect it to be in much the same place.

Let us know how you go.

_________________
Trevor Gore, Luthier. Australian hand made acoustic guitars, classical guitars; custom guitar design and build; guitar design instruction.

http://www.goreguitars.com.au

Thanks Trevor. Much appreciated.

_________________
_____________

EddieLee

Here is part 1 of a screencast on tap testing acoustic guitar bodies with Audacity: https://www.youtube.com/watch?v=y28ztJoIqhQ
I don't claim to be an expert like Alan or Trevor -- I'm just passing on what I've learned (much of it from them!) Feedback is welcome, of course.
I haven't yet tried Trevor's method of thicknessing plates, so I can't comment on using it for that, but I expect the process would be pretty similar.

Audacity does have a "wishlist" of requested features: http://wiki.audacityteam.org/wiki/Featu ... t_Spectrum
I think most of us want this one: "User-defined frequency bands: (11 votes) CoolEdit/Audition allow this. Use case: the most relevant region for music is 20 to 8000 Hz."
And this would be helpful: "Remember window height/width across sessions (3 votes)"

They act on a feature request when it gets enough votes; you can vote by emailing them: feedback@audacityteam.org
(http://web.audacityteam.org/contact/#feedback)
Stand up and be counted.

_________________
David Malicky

Thanks, David!

_________________
George :-)

I agree. Great tutorial, David. Your presentation was fantastic.

I received a thank you from the Audacity technical team for casting my votes. They also noted that they had received additional votes for the same features yesterday. Your participation matters!

_________________
George :-)

Great video David. Looking forward to #2.

Very nice David, there's some good information there. I'll be sure to cast a vote to the audacity team.

_________________
Jim Watts
http://jameswattsguitars.com

I use the export function after I analyze the bonks and bring the string of numbers into Excel. That way I can plot several graphs or guitars and also zoom into a frequency range.


Sent from my iPhone using Tapatalk
Ken

_________________
Ken