Official Luthiers Forum!

I have a recording engineer's background. I also have access to a decent amount of microphones, DAWs, etc... I was thinking how nice it would be for my next couple of builds to document the tap tones and record how they change as I remove more wood, add braces, etc... I'm definitely not one of those guys that can gather info from Chladni patterns (more power to you if you can!!) but I certainly would like to at least document the pitches and it's changes. I think it would help to get to "know" the wood a bit better.

Has anyone done this? Can anyone suggest a good microphone setup, etc....? There might be some complications.... For instance, I have one piece of wood that absolutely rings like a bell when I hold it right and find the sweet spot. I noticed of course the farther I held it from my ear the "tone" seemed to change a little. In the recording example the microphones would be my "ears". Any thoughts?
-John

John, I recorded the tap tone of my current project. I think you'll be more equipped to get a good recording than I am, but I set up my large condenser (can't even remember what it is, but I think ADK TC) where my ear would be and tapped as usual. The clip is right on through my headphones.

I've done that on my first guitar using a Zoom H4 portable recorder. I think the key is some sort of consistency in mic/top/room placement from guitar to guitar. I don't think any method is going to be perfect because of the complex dynamics of acoustics.

If you're primarily using your ears to tune the tops, I think all you'll need is a decent recording to remind you of the sounds the last top or top several instruments ago was making.

_________________
http://www.birkonium.com CNC Products for Luthiers
http://banduramaker.blogspot.com

You could purchase the Peterson Stobosoft software and record your tap tones.......and have the ability to graph the frequencies produced by the tap. That would be the ultimate comparison from top to top wouldn't it?

_________________
Formerly known as Adaboy.......

You can also do some frequency analysis in Audacity (which is free). I'm not sure how its capabilities differ from Strobosoft.

_________________
Evan McCartney

It would be good comparison if and only if you do a very good job of maintaining consistency in mic placement and room acoustics.

_________________
http://www.birkonium.com CNC Products for Luthiers
http://banduramaker.blogspot.com

I don't want to paint myself as an expert (since I'm not) but as I understand it, the tap is very short in duration and most software won't be able to capture the short duration "thump" and give you the dominant frequency. That's my understanding at least......if someone has more insight I'm all ears. I'm not familiar with Audacity unfortunately.

_________________
Formerly known as Adaboy.......

That's my understanding too. I've got Strobosoft and the tone holds a note for a very brief period of time.

Well I guess this brings up the question of what significance the decay of the struck note has in the finished product. Although once again, the force with which you strike the top and positioning of all components would have to remain as constant as possible.

Maybe it's just me, but I have always had the feeling that the sound I hear slightly after striking (as soon as my brain can segregate out the sound of the initial "whump") gives the clearest picture of the sound of the plate itself vibrating. Perhaps the closest approximation of how the plate reacts to a vibrating string is through the constant drive of a signal generator in Chladni patterning?

_________________
Evan McCartney

a few tap tone videos on youtube:
http://www.youtube.com/watch?v=6OP84aJU4Xg
http://www.youtube.com/watch?v=RFMqKKigdzs
http://www.youtube.com/watch?v=HniEVGCv ... re=related
http://www.youtube.com/watch?v=5HChRXyg ... re=related

This is easy to do with Audacity. It doesn't require any great percision because you're looking for the frequency of the modes, not exact relative amplitudes. So if your mic or room gives some equalization it will effect the hight of some peaks but it's not changing the frequency.

The duration of a tap is certainly long enough for a basic spectrum analysis. Like glitter patterns, I've only used this to look at the low frequencies of plates. I don't know what the higher frequeny info would tell you.

I've also done this with taps of completed guitars. It really tells you a lot about the guitar.

Here are a few that I've recorded:

http://sculptorjones.com/section/19244.html

I agree that recording tap tones is ultimately just an empirical collection of sounds relative to one another. This is of course only meaningful if the process of capturing the sounds is as close to the same as possible across the group. That said, I've strayed a bit in my attempts, using a couple of different mics, and more significantly, varied the "hold point", which completely changes the pitch and resonance. Going forward, I plan to keep these things constant. My current recording setup is:

Guitar top held one inch from edge of plate on the bass side upper bout between UTB and X brace, approx 8" from mic. I lightly but firmly strike the top in a clockwise direction with a small felt covered wooden mallet --> SE Electronics H3500 LDC --> Toft ATC2 preamp/comp/eq (eq flat, light comp) --> Tascam FW1884 interface --> Cubase SX3 (24 bit 48K samplerate). The weak link in the chain is that I converted the wav to a 160kbps mp3, so the online samples lack the sustain of the original wav's.

I use Audacity to analyze the spectrum, using the logarithmic setting instead of linear. This show the low end frequency peaks that I'm interested in. I've also toyed with a tuner plugin in Cubase to look at fundamental "pitch". Again, these samples aren't really meaningful individually, but when taken collectively, they can tell me something about the progression of my top tuning efforts.

_________________
Mountain Song Guitars www.mountainsongguitars.com

Another free program is 'Wavesurfer', downloadable from:
http://www.speech.kth.se/software

It has the recording function built in, and lots of options, including a spectrogram, but won't do waterfall plots.

As I understand it, the sample size establishes the amount of information you have to work with, and this translates into restrictions on the upper frequency limit and resolution of the FFT. In the 'old days', twenty years or so ago, all papers that discussed results of FFT analysis spent a lot of time talking about these tradeoffs, since processor speeds limited the amount of data they could use. Now that's not a problem, but it's nice to have some idea of how it works.

Nyquist's theorum says that you need at least two data points per wave period to define a sine wave, so the sample rate sets the upper frequency limit for the analysis. That is, if you record at 16,000 samples/second, the highest frequency the FFT can give you will be 8000 Hz. You have to keep in mind that the mechanics of the tap itself will also impose an upper limit: as long as the 'hammer' is in contact with the 'anvil' the thing can't vibrate freely. A soft tapper, like your finger tip or a super ball, might stay in contact with the top for 1/1000 of a second or longer. The sound will start to fade out at 500 Hz, and there will be very little energy above 1000 Hz. In this case a high sample rate, like 96,000/sec, would be nonsensical.

The length of the sample imposes a limit on the frequency resolution. The lowest frequency the FFT can break out will have one full wave length in the sample, so a 2 second sample will contain valid information down to 1/2 Hz, and a 1 second sample down to 1 Hz. The FFT sorts the information into 'bins' that are mutiples of that lowest resolvable frequency: for a 2 second sample the bins will be 1/2 Hz wide, and things that are closer together than that won't be resolved.

We can get by with short samples in looking at tap tones because we're not really interested in either high resolution or high frequency limits. There are so many possible vibration modes in a top above 1000 Hz that they're closer together than their band widths; you're in a 'resonance continuum'. At that point all you can do is get a rough idea of how many modes there are for every 1/3 octave band, say, and find the average damping by looking at the ratio of height from the peaks to the dips in the spectrum. Mode spacing down in the controlable area, below, say, 500 Hz, is usually something like one every fifty Hz or so, so .1Hz resolution won't do you much good. All in all, I'd suggest a low sample rate; for my sound card the lowest is 6000 samples/sec, but , for various reasons, I often use 8000 or 16,000. Multiple taps can stretch out the sample length, and give higher resolution, if you take some care to make sure they're all comparable.

I often use an ancient (in PC terms; it was written for 286 processors) freeware FFT program called FFT4WAV3. One of it's many quirks is that it can only use a sample of 34768 data points: one less and it crashes. To use a short sample I 'zero pad' the data: adding in a period of silence before and after the actual tap. This artificially inflates the resolution, of course. This is not a problem so far as I can find out, as long as you recognize that you're doing it. With FFT4WAV3 you can export the 'real' and 'imaginary' parts of the FFT to a spreadsheet as comma delimited text, and this allows you to average over the levels in the spectrum plot to reduce the resolution to it's proper limit of you want.

You will sometimes see references to 'windowing' the data. Ideally the data sample should start and end with vibration amplitudes near zero. There are various ways to accomplish this; zero padding is one, but others such as the 'Hamming' and 'Hanning' window use a mathematical process to, essentially, change the gain so that the signal starts off low, builds up, and then fades out. This is fine if the signal spectrum itself doesn't change during the sample time: for a tap or other 'noisy' signal it's better to use a 'rectanguilar' window (no change in gain) and zero padding if needed, or so I understand.

In case anyone is interested, here is a pic of a frequency analysis from Strobosoft of a tap so you can compare with to what is available with other software.

_________________
Formerly known as Adaboy.......

Based upon an earlier recommendation from Alan to utilize WaveSurfer I have done so for the past few years with very good results.

I import the WaveSurfer file into Excel and then truncate the data table down to utilize only up to 1000 hz. I then insert a graph on the same sheet as the data and it is then fairly straighforward to see the primary modes. I usually run at least three tests (see the tabs on the spreadsheet file at the bottom) to see if they indeed correlate.

Here is a screenshot of an Excel file showing tests of a recent OM:



You could also use Audacity to produce much the same FFT file and export over to Excel but I've had the best results using Wavesurfer. It's a bit more user friendly.

_________________
Peter

How do tap tones of the free plate correlate with the sound of a finished guitar?

_________________
Ken

Ken, The correlation I mentioned was to run at least three tests and see if the results were, in fact, repeatable on the finished guitar. I have yet to establish enough of an empirical database to provide guidelines for progressive tap tone to completed instrument on an OM type body.

I do utilize this set of progressive resonance development targets for my classical builds. It was developed by Greg Byers and I have found that it is useful in getting very close to the final desired results on a completed build. If nothing else it allows some form of progressive system discipline in attempting to get repeatable results. It's not a magic bullet although because there is so much variability in the materials we use. It serves as a good guideline.......

_________________
Peter

Peter,
Good chart. I use something similar in my violins. But I get the weight and calculate effective stiffness (ES = frequency squared X weight). There is a bunch of Stradivari plate data on this so I can try for something similar. As a target this can help get a consistent low frequency resonance of the finished instrument.

I should add that this represents the long grain stiffness of the plate.

_________________
Ken

There in lies the million dollar question, and that is where Chladni patterns of an attached top pays out. I just wished I was set-up to do them

Do high end guitar makers use these tools to get consistency? Im thinking Goodall and Collings? Collings guitars are "consistently" my favorite.

I guess what your doing is building a database, and if you get a guitar you like (recorded),-then you can use that as a model, is that correct?

_________________
Here is what a Parlor Guitar is for!
http://www.youtube.com/watch?v=oEa8PkjO6_I

I think the most basic and also useful thing that can be done with measuring free plate taps is to set the stiffness to weight ratio where you want it. And I think this is mostly useful for not overbuilding guitars. Which is mostly useful for beginning guitar builders because once you build a few you get a feel for what's overbuilt.

If you're into Chladni patterns, measuring the tap is a quick way to find what frequencies to look for the patterns at.

Measuring the tap on a completed or attached guitar is another story and it can tell you a lot, maybe everything about the behavior of the guitar. Chladni patterns can't really do that because they show steady-state, not transient behavior.