FFts

[John Bertotti]

I'd like to start tracking these from day one before any top or back I make is glued up. The question is where do I start? I run a mac and that won't change. Software title, the company that makes it, other components needed and if you know an estimated cost. That may be prohibitive and I certainly wont hold off putting my first couple together just for this. Thanks all for any info. John http://www.mandolincafe.net/iB_html/non-cgi/emoticons/cool.gif

[WireBoy]

I don't know know the term FFTs. Please explain
thanks

[Chris Baird]

FFT is Fast Fourier Transform. It is an algorithm that breaks down a complex wave into its independant sine forms. You can find a free/shareware FFT and a graphic wave anylizer from various sites. You just have to surf around to find them. I use a fair film trasducer as a pickup and plug it into the mic input on my sound card. The better your card and transducer the higher the quality of the result. I find it helpful to "tap" the body with sound, i.e. sine and sawtooth both static and sweeping. You can obtain lots of information; hard part is trying to make sense of any of it.

[Dave Cohen]

The significant part of the FFT is what it transforms from & to. When you take any data, regardless of whether the transducer is a mic, piezo film, etc, you are acquiring a signal vs time. Usually it is a voltage (occasionally a current) proportional to the response from the transducer. Your computer A/D acquires values from the transducer sequentially at regular time intervals, as fast as every 1/100,000 to 1/1000,000 (on some boards) of a second. This is called time domain data. As the information contains periodic (i.e., repeating at regular time intervals) components, as all musical information does, it is more conveniently understood if the signal or something related to it is plotted vs frequency. And that is what the FFT (or Ft) does; it is a mathematical algorithm which transforms from the time domain to the frequency domain. So you acquire time domain data and the FFT converts it into a related value plotted vs frequency.

There is fft shareware out there, and also some low cost software. One of the most readily available programs was "Cooledit", which has been swallowed up by a software company (I forget which) and is now in the low cost category. You will also find FFT functions in plotting & data analysis packages.

[onlyagibsonisgoodenuff]

Wonder what software Mr. Loar used?

[Scotti Adams]

I don't know know the term FFTs. #Please explain
thanks
..and I still dont know what it does http://www.mandolincafe.net/iB_html/non-cgi/emoticons/rock.gif

[Dave Cohen]

For the technophobes:

(1) Frequency equates to "pitch". So if you wnat to know what the harmonic content of a sound is, you need to know what the frequency components are, and their relative amplitudes (read "strengths"). That is where the FFT comes in. The simplest generalization is as follows: a sound which is weak in the higer frequency components will sound muddy and indistinct, while a sound which is lacking in lower frequency components will sound nasal and harsh.

(2) Obviously, there was no such thing as software in 1922. At that time, the technique for obtaining a frequency domain spectrum was clumsy at best. With "Mr. Loar"'s interests, howwever, it is a safe bet that he would have used FFT software had it been available.

[sunburst]

I too have surfed looking for FFT software for Mac. If you find any please let me know.

[John Bertotti]

No luck yet sunburst but I'm checking with the guys at my mac store. John http://www.mandolincafe.net/iB_html/non-cgi/emoticons/sad.gif

[Chris Baird]

Have a look thru some of this (if you haven't already). If you can find your way to some of the actual home pages you may find some mac versions.

http://www.hitsquad.com/smm/win95/SPECTRUM_ANALYZERS/

[Chris Baird]

All right here is mac side of that site.

http://www.hitsquad.com/smm/mac/SPECTRUM_ANALYZERS/

[sunburst]

Thanks Chris!

[Chris Baird]

I know it is terrible but I download these programs one at a time and when the free demo time runs out I download a different one. So don't download em all at once or they will all run out at the same time. I suppose one day I'll make it big and be able to afford the $20.

[John Bertotti]

Spending a lot of time traveling but hey thanks for the link. Chris do you have a favorite out of the bunch of demos yet? John http://www.mandolincafe.net/iB_html/non-cgi/emoticons/smile.gif

[Chris Baird]

I've got a pc so most of the programs are different. The features you want are:
The ability to record and selectively isolate very specific time periods.
A way of telling exactly the frequency and amplitude of any given point on the spectra.
A corresponding relative amplitude graph which is coupled to some definte time data, which can be used to both identify tap attack and decay and also to record body response as relative amplitude plotted against the input of a stable decibel sweep, either sawtooth or pure sine.
A tone generator that is easy to use and has the capability of being changed by some kind of user interaction in real time.
And just generally easy to read and figure out graphs and fft data points.

Hope that helps.

[Flowerpot]

It seems to me that to get any useful data from an FFT, you would need not only several transducers mounted in various places yielding many FFT's, you would also need to take a lot of care how and where you apply a stimulus (a calibrated felt hammer where treble and bass bridge feet sit?). Just measuring the deflection, for instance, in the very center of the top could detect the effect of the "trampoline mode" (is that the (0,0)?) quite apparently but may miss other important vibrational modes for which the center point is a zero-movement point (a node). Maybe I'm wrong, but it seems to me there's much too much motion going on in a vibrating top to effectively capture a snapshot in just one spectral analysis. But I would guess that the use of a ribbon-type transducer would help average the motion of several square inches of the top.

[Chris Baird]

Flowerpot, you are right in insinuating that the variables are overwhelming. They are! As the mandolin is a complete system that is coupled to itself in many ways you will get similiar results from various transducer positions, but of course there are always discrepencies esp. if you put the transducer smack on a node. If you do get into ffts you will undoubtly take many many spectras, and the more you take the less you will KNOW.

[sunburst]

Flowerpot,

I hope Dave Cohen sheds some light on this soon, but here's a quote from an Email he sent me in January:

You don't need a fancy gizmo like a Conn or a Peterson to observe the body
modes. All you need is your PC, a microphone into the sound card, and an
FFT program to analyze the data. I think that Gavin Baird has been doing
some of that, and guitar luthier Alan Carruth is a big proponent of it. So
all you have to do is stuff a small piece of soft foam between the strings
and the fingerboard and bonk the body of your instrument with your knuckle,
with the microphone preferably about 1 meter away. Then you get the FFT
program to fourier transform the resulting sound file or wav file, and
voila, the frequencies of the peaks in the transformed spectrum are the
frequencies of the various modes (and you get 'em all from that one bonk;
transforms are powerful stuff!) There are a few such "SPL (for "Sound
Pressure Level") spectra in our first paper. If you do that on a few
instruments, you will start to see some similarities, but also some
differences between the instruments. Contrary to what the fuzzy and artsy
types say, the spectra are quite sensitive. The difficulty, of course, is
in the interpretation. But that will be another lesson, after you have done
a few of those SPL spectra and I have given you a few quizzes <g>.

Yours,
Dave Cohen


Used without permission...hope you don't mind Dave.

[Flowerpot]

Ah, silly me. The talk about transducers had me thinking down that line, and I forgot about simply using a microphone. That would work better, as you are measuring the actual sound waves vs top deflection. OK, maybe you CAN get a decent snapshot of an instrument's response with a single FFT. The only effects you wouldn't get would be non-linearities. Of course, amplitude --sheer volume -- might be hard to compare between instruments without calibrating your knuckle rap, but you'd have the tonal profile right there in visual form. I'd love to see some different instruments' spectra to compare.

[sunburst]

I'd love to see some different instruments' spectra to compare.
Me too!
But so far I'm not computer savy enough to figure out how to download an FFT, get it to work, or even figure out if I've got a usable sound card in this pieced-together relic of a computer I'm running.
Besides, how do you tell if something like that is working if you don't even know what you're looking at?
Lots to learn in this mando biz....

[Chris Baird]

The only time I use a transducer over a mic is when I am exciting the body with a tone generator rather than a piano hammer. The transducer is much less affected by the actual vibrations in the air and picks up on what the body is doing in response to the tone generator. You can give the mandolin a sawtooth sweep from 0 - 10k hz and get both a map of amplitudes at various sawtooth frequencies and exact decibel comparisons of each partial for each point in time/sawtooth sweep tone.

[John Bertotti]

So I checked out the link above and there were several mac offerings. 1 called a an xa1 claims to be a band pass analyzer with better performance than an fft. Anyone care to comment on that. I have an understanding of bandpass, actually I've forgotten more about it than I remember but have no clue how the fft actually works or performs. Thanks John

[Dave Cohen]

I have been away for the past 11 days, else I would have responded to Flowerpot's & Reesaber's & John's queries earlier.

First, re "transducers":
A 'transducer' is any device which converts a signal consisting of one form of energy into another form of energy, as for example, from mechanical vibrations into an electrical signal. A mic and a piezo accelerometer or film both do that. The difference between a mic and a piezo device is that the mic responds to pressure variations in a fluid (i.e., air), whereas the piezo device produces an electric current in response to its own motion. Thus, the piezo has to be in physical contact with the vibrating (mechanical) object.

Second, re multiple transducers vs a single transducer:
The normal modes of motion of a vibrating object are the same regardless of where on the object the excitation is applied, unless, as Chris pointed out, the excitation is applied right on the location of a node for a particular normal mode, in which case that particular mode will not be excited. All of the remaining modes, however, will be excited. In doing holography, I usually use a single source of excitation; a small magnet driven by a coil positioned nearby w/ a/c current running through it. If one mode is conspicuously absent, I just re-position the magnet and coil until I find the mode. The only reason for using two or more transducers would be if you want to find a phase relationship, e.g., between a top plate mode and the corresponding back plate mode. You sometimes need that when the plate modes are multiplets, though for what most of you will be doing, that will not be of much (if any) concern.

Third, re which transducer to use, and how to use it:
If you use a mic in response to plucked strings, the mic will respond primarily to string modes. So if you pluck the open G, you will see the 196 Hz fundamental, as well as the 2nd, 3rd, 4th, & etc., harmonics. If you happen to pluck the string at the location of the node of a particular harmonic or harmonics, that (or those) harmonic(s) will be weak or absent. If you want to "see" the body modes and air modes, you will need to damp the strings and excite the body, e.g., with a tap or "bonk". With a mic, you will see peaks for all of the body modes plus the air modes (principally the helmholtz resonance). With an accelerometer, the air modes will be absent, i.e., you will see a spectrum of the body (and neck)motion.

4th, re the knuckle spectrum. Alan Carruth seems to be greatly concerned about that. He uses an acrylic ball on a pendulum arm as his "bonker". Rossing and his lab don't think that there are significant frequency limits from a knuckle tap, unless you are awfully fat, er, uh, "circumferentially challenged". I have been using my knuckle as my "bonker", and I have not seen any apparent frequency limits in the spectra. You can see a few of my knuckle spectra in our first (Catgut) paper, and a few accelerance spectra in the 2nd (Acoustical Science & Technology) paper. We used braodband excitation for the accelerance apectra, and if anything, there are probably more artifacts in those than in the sound spectra.