Official Luthiers Forum!

Hi all,

I've been building for a few years but still have almost no reference point when it comes to tap tones and whether I'm hearing something good or not. With that in mind, I decided to record some of my guitars for comparison. The links below will take you to some audio files I recorded of different tops. My recording technique was pretty crude (2 condensor mics plugged into a laptop), however they're good enough to get an idea. Feel free to post your comments and how these tap tones compare to what you hear with your own guitars.

This first tone is my benchmark, a David Schramm build Hauser model with cedar top: Guitar 1

This second tone is one of my spruce versions of a Hauser guitar: Guitar 2

This is the same guitar, however I cranked the mics way up--the initial taps are distorted, but it lets you hear the ringing a little better. I found it very hard to record some of the nuances of tap tones: Guitar 2, louder

Here is a cedar guitar braced exactly like guitar 1, however the back isn't glued on yet: Guitar 3

This last guitar is a spruce top, braced like a Hauser, with quilted maple sides and no back glued on yet: Guitar 4

I'm not sure if these files are helpful or not, but I thought it might be useful to gather a library of audio files together for comparison. I know it would take me a long time to develop a good ear for taptones building only 3 guitars per year. Feel free to post some audio files of your guitar tops. The more files posted, the better perspective we'll all have on this elusive thing.

Cheers!

John

Thanks for the sound clips, John. That's pretty cool.

I find it's hard to really get a good idea on what's going on exactly... Are you tapping at the same spot all the time or moving around.

If I were to compare to mine, I would think that my sound seems to be of a lower pitch, like your guitar no.2...

I think it's a good idea to have a little archive of these kinds of sound bites...

I'll try to get some recorded too!

_________________
I'd like to be able to prove, just for once, that money wouldn't make me happy...

Good Idea!
I believe we all use tap tones in different ways and for different reasons.
Someone more experienced will hopefully list them for us.

Maybe you could call out where you are tapping and where the microphone is as you record the clips?

Question is, how do we use this information?
What are we looking for?Wade S.39016.5769097222

_________________
Wade
Nashua, NH
http://www.wadefx.com

All the recordings have the tapping in the same spot, right behind the bridge. Also, the micophone was located about 1-2 feet in front of the sound hole.

As far as what I hope to accomplish...I guess I'm not sure. At a minimum I would like to hear other tap tones for comparison. I know many people report hearing a bell-like ring, or a dull thud--I would like to hear other tones for comparison so I know if mine are closer to the bell tone, or the thud tone. I haven't heard enough different tops to know what an average top migt sound like, and which ones are better and which ones are worse.

Are you sure you weren't hitting an I beam with a hammer on that third one there. Just kidding. All sounds like a bunch of thin wood to me.

No seriously. I'm trying to learn just this very thing and am pretty lost. I have about 6 top sets downstairs and was up till 3 last night tapping them and trying to evaluate. I noticed one of them had a more for lack of better terms, brick-like sound, that was higher and seemingly tighter--less boomy I'd imagine, and wondered whether that would be the best of them all. I also have a very aged one that I bet is adirondack. It was more bassy. The one I'm working on now is bearclaw sitka, and I'm not sure if it's nice or not. It was closer to the adirondack actually.

Do you listen more to the after rings, or the actual tone range (bassy or trebly)? Obviously one would like to hear a ringing bell, but this may be subtle on a spruce top, which is more sensitive or responsive than a back or side set.

By the way. How do you get those sound snippets uploaded like that?

Oh yes. One more thing. Another topic discussion mentioned cooking tops. I was ironing out a few little dings on my top when I decided to see what would happen after completely ironing the spruce for an hour or so (without a wet towel). Interesting. Little pitch streaks eventually appeared, and I was able to reverse the direction of cupping that had almost motivated me to scrap the whole thing at one point. Also I seemed to notice a sound difference. The dryer top had more ring it seemed, which of course would make sense (to me at least). Moisture would tend I imagine to dampen tone. Maybe I was just imagining it though.

I did all this of course before remembering a stern warning about how wood does not like sudden drastic changes in humidity or it will risk cracking. But no cracks. I guess that means I have a top that will at least move well.

nickton39016.6040972222

_________________
"Preoccupation with an effect gives it power and enhances the error"
from "Your Owner's Manual" by Burt Hotchkiss.

Wade S asked:
<<Question is, how do we use this information?
What are we looking for?>>

That's the $64,000 question, isn't it?

I've been recording tap tones for years, and comparing the spectra, and I'm seeing some consistent trends. But, and this is a biggie, you _have_ to standardize your testing to get anywhere. Even small changes can alter the sounds enough to throw you off; mics just aren't as good at generalizing as your ear and brain are.

I hang the guitar up with two loops of string under the E string machines. I put small strips of foam rubber under the strings at the 4th and 11th frets, which pretty well deadens them. The mic is a meter out in front, on the center line, between the bridge and soundhole. I use a hard wood or plastic bead, about 5/8" diameter and swinging freely, to tap the saddle between the 5th and 6th strings. I record the sound on my computer, and use an FFT program (freeware; FFT4WAV3, that runs under DOS) to deconvolve the frequency structure. The 'real' and 'imaginary' parts of the transform are saved as comma delimited files, which are then opened in a spreadsheet program, and the spectrum computed. This is a little cumbersome, but it allows me a lot of flexibility in looking at the results.

_Everything_ you do differently will alter the spectrum you see from a given guitar. Small changes in the mic position, for example, will 'hear' different modes more or less strongly. The room itself will tend to emphasize or suppress various frequencies, and moving the gutiar to a different location within the room will alter these. There is no one such test that will tell you everything you want to know. Rather, you have to look at them as 'snapshots', which give a lot of information quickly, but also leave out things that you'd like to know.

In my tests I've found that there are several things that correlate well with guitar sounds that I like. Features like the relative height of the 'main air' and 'main wood' peaks in the spectrum, the width and conformation of the 'main wood' peak, the number and height of peaks between the 'main wood' and 1000 Hz, and the average height and 'peakiness' of the spectrum in the 2-4 kHz range, all seem to predict certain features of the tone.

But your results will most likely vary, simply because you've got a different setup from mine. There's no 'royal road' here: you're just going to have to do a bunch of testing, and see what you come up with on your own.

Finally, and most importantly: believe your ears! None of these tests means a darn thing by itself, but they take on meaning when you can correlate the results with things you like or don't like. For example, one of my customers brought in his guitar for a pickup installation, after he'd had it a year or so. It had played in really well. When I did the 'bonk' test on it I noticed that the 'main wood' peak in the spectrum was particularly tall and broad, and I got out all of the other charts I had to see what that meant. In sorting them out according to the shape and height of that peak, it soon became apparent that the ones I didn't like tended to have tall, narrow 'main wood' peaks, while the better instruments had braoder ones. The 'prize' for tall and narrow went to a Hohner 00 in plywood, with a couple of Ovations not far behind, and they all sounded 'thin' and 'harsh' to me. Guitars with a 'full' and 'solid' tone tended to have the tall, broad 'main wood' peak. In subsequent reading I found that at least two researchers in the classical guitar world had noted the same thing.

The important thing is that nobody came up with any fancy theory that predicted this in advance; it's all based on corellating measurements with the judgements of tone. In retrospect it makes some sense, but the data has to come first. So, get the best data you can, and get a lot of it. The data will tell you what you're looking for.   

Al, that is some great info. I was able to load the FFT4wav program and can get it running to recognize my wav files, however there isn't much in the help section about how to run the analysis. Were you able to find a help file or user manual that gives a little more guidence? Also, did you write an Excel spreadsheet, or were you able to find one on the net to help with the analysis?

Thanks!

John

Ah yes, 'Help' in FFT4WAV3 is, shall we say, minimal.

In theory FFT4WAV3 can record files. In practice, it was written back in the days when 386 processors were new, and soundcard options were limited, and does not seem to be able to find most sound cards these days. I use an old version of 'Coolwave' for recording, but you can also use 'Wavesurfer'; another free program for that. you may find that with 'Wavesurfer' you won't need FFT4WAV3, but I still use it. Anyway, once the file is recorded, save in the same directory/folder as FFT4WAV3: it has no way of looking elsewhere for sound files. The file must be saved as a .wav file, of course.

One peculiarity of this program is that it looks at the first 32,768 data points. If you give it one fewer it will crash, and any more than that will be ignored. If you need to 'zero pad' your data to 'fluff it out' to the right size, do that. It will mean that some of the resolution you seem to be getting is not really there, if I understand correctly, but you'll get some results.

Open the file by highlighting the menu entry and clicking. Once it's open you'll see the wave form on screen. Click again to get back to the menu, and highlight the selection 'Fourier Transform'. Click there; you'll get a query asking whether you really want to do it, with the selections "j/n". That's 'Ja oder Nien'; Lars Otte, who wrote the program, was not a native english speaker. Hit the j key.

When the menu comes back up, click on 'Visualize'. This will bring up the 'real' and 'imaginary' parts of the transform as graphs. If you want to see the amplitude, hit the 'a' key. You can window segments of the plot, and there should be numbers up the corner of the screen that show the cursor location by frequency and amplitude.

I typically window the first thousand Hz worth of the spectrum, and then hit the 'backspace' key. This discards all the data that is not in the window, keeping the size of the spreadsheet down a bit. It takes you back to the menu. Otherwise, 'Escape' will get you back to the menu too.

Once there I select 'Save only real', which brings up the choice of format. Select 'Text'. A screen comes up to enter the name of the file. Remember, this is DOS, so keep it to a few characters! It will save the file in the FFT folder with the name you gave it, and the extension ".(T)". Yeah, I know, but that's how it works. go back and save the 'Imaginary' file the same way.

I use Microsoft 'Works' for my spreadsheet: it's a crippled but cheap version of Excel. Open a blank sheet, open the file you want, say the 'real' part, and paste it into the first column. do the same for the 'imaginary' part. These are files that give the magnitude of two orthogonal vectors for each 'bin' of the FFT, as comma delimited text. The square root of the sum of the squares of the two vectors is the amplitude, and the ratio between them allows you to calculate the phase angle. So, if the real and imaginary parts are colunms A and B, then space C1=A1^2, D1=B1^2.

The data runs from 0 Hz to an upper limit of 1/2 the sample rate, so if you made your recording at 16000 samples/sec, the top limit of resolution is 8000 Hz. Since the program used 34,768 data points, there will be 16,284 data points in the output, representing frequencies from zero on up. Thus, in the case above, you'd have one data point approximately every 1/2 Hz, with the first one representing the amplitude at zero Hz, and the last one the amplitude at about 7999.5. you want to convert from ordinal data points to frequency, so in space E1 put "0", and then fill the series down from there by 1. So F1=E1*(8000/16383); that is, half the sample rate over one less than the number of data pairs in the sample.

Finally, you need to get the amplitudes: G1=sqrt(C1+D1).
'Fill Down' all the columns except E, which is 'fill series', of course.

Highlight the last two columns, 'Frequency' and 'Amplitude', and select the option that will turn that data into a scatter graph. I make the actual data points invisible, and connect them with a solid line.

All of this assumes that your spreadsheet will open comma delimited files correctly. If it won't I'm sorry to have wasted so much band width.      

[QUOTE=Alan Carruth]
...
One peculiarity of this program is that it looks at the first 32,768 data points.
...
[/QUOTE]
Some early Fast Fourier Transforms algorithms as opposed to Discrete Fourier Transforms were constrained by design to work only on powers of 2, apparently the one you use works on 2^15 data points.
[QUOTE=Alan Carruth]
...
If you need to 'zero pad' your data to 'fluff it out' to the right size, do that. It will mean that some of the resolution you seem to be getting is not really there, if I understand correctly, but you'll get some results.
...
[/QUOTE]
Padding your data with zeros is okay but make sure the endpoints of your aquired data are already zero or very close to zero. In other words, your no-sound baseline should be represented by a zero in your data. If not, by padding with zeros you introduce a step function into your data which gives a 'ringing' artifact in the frequency domain.

Okay. I am sufficiently confused and intimidated. Thanks.

_________________
"Preoccupation with an effect gives it power and enhances the error"
from "Your Owner's Manual" by Burt Hotchkiss.

John
Interesting post, may I ask, what are you tapping with.
Some type of mallet? Thanks for sharing.

Tom

_________________
http://www.moriciguitar.com

Man, i need to re-enter University!

I'm just using my thumb for tapping and hitting the top as if I were playing a tambora while playing. I need to develop a method like Al describes so it's more consistent. That will be the first thing I do once I figure out the software.

[QUOTE=Serge Poirier] Man, i need to re-enter University! [/QUOTE]

Forget University, kids are learning this stuff in High School

This is some seriously wild and cool stuff ya'll. Thanks for sharing. My head hurts

_________________
My Facebook Guitar Page

"There's really no wrong way, as long as the results are what's desired." Charles Fox

"We have to constantly remind ourselves what we're doing....No Luthier is putting a man on the moon!" Harry Fleishman

"Generosity is always different in the eye of the person who didn't receive anything, but who wanted some." Waddy Thomson

Thanks Marc: there's always something to learn in this business. I use zero padding all the time; almost all of the information from a 'thump' is in about 1/4 second worth of signal, and I've got some real 60-cycle hum problems, so zero padding gets rid of some spurious peaks and dips.

Yeah, this stuff is complicated when you first start to get into it. So was basic guitar making, if you remember. Compared to what things were like thirty or more years ago, we've got powerful tools that are so easy to use... Carleen Hutchins had $20,000 worth of dedicated B&K analog equipment 20 years ago that didn't do any more than my laptop can, and I can still play 'Solitare' when I want to, which you certainly could not do with her chart recorder. My father in law spent a week at MIT in the 30's doing a Fourier transform on 350 data points, using a pencil, slide rule, and adding machine. Sometimes we forget how cool all this technology really is.