Frequency response of guitar hardwoods

[erikbojerik]

Hi everyone

What we (my son and I) would like to do is to set ourselves up to measure the frequency response curves for different kinds of hardwoods used in making solid-body electric guitars. The idea is to see if different hardwoods have characteristic frequency response spectra; we know that different wood species have different tonal characteristics, what we'd like to do is to quantify it a bit.

Here's how we're thinking of approaching it; feedback would be appreciated (no pun intended...).

1) constant dimensions for all body blanks (14" x 24" x 1-3/4" or similar)

2) either 1-piece (more pure measurement) or 2-piece (more common) body blank

3) white noise or sine-wave sweep for sound source

4) some kind of piezo system to record the vibration response of the board

We're struggling with how to mount the board in front of the sound source (prob. my amp), where on the board to mount the piezo transducer (center? or node?), and how to record a "baseline" response of the amp+piezo that we could subtract from the data obtained on the boards.

Oh yeah, and we'd love the software equivalent of Cool Edit Pro (now Adobe Audition) for the Mac....anything to do an frequency analysis of the recorded output of the piezo.

This will become my son's science fair project for the coming school year.

TIA

Erik

[Ford]

Wow, I studied stuff like this in my accoustics class in College and loved it. I still have my text book, so if theres anything I can do to help, let me know. I'll see if I cant dig through the book and find any answers to your questions.

I remember one day the teacher putting a big rubber drum head over a speaker and blasting loud tones. He then turned out the lights, and turned on a strobe light set to the same frequency of the tone being played. You could see in slow motion the movement of the drum head, and the way it oscilated around its nodes. Wierd stuff. Loved it.

[gwm]

sounds like a cool project!!

in one of my eng classes we determined the speed of sound in a steel bar. the details of the experiment escape me at the moment, but an accelerometer was attached to one end of the bar. the bar was then tapped very very very lightly.

i think if you did something similar, but maybe tapped the wood harder to get a whole range of frequencies you'd get some neat results.

i think the mount for the piece of wood would be key though. depending where you clamped it you'd get different responses.

keep us posted.

gwm

[guitar_ed]

Hi Erik,

I think that the key would be consistency. If most of the boards are two piece, then they should all be two piece, clamp in the same place every time, all of the woods the same size, stuff like that. Because I can guarantee you, no matter what you do, somebody will complain that they have a better and more accurate way of doing the tests.

I would think that a 1 piece vs. 2 piece body test would also be of interest to the people on this forum.

Guitar Ed

After inflation, my $.02 ain't worth much

[frank falbo]

Rather than pink or white noise, I would think it's more beneficial to use some sound that is more concentrated within the range that a guitar's strings produce. As a matter of fact, it may be more accurate to produce a mechanical picker and string them up with a surrogate "neck". The noise won't tell you how it would respond to string vibrations. Also having a soundsource blasting air at it is entirely different than transfering string vibrations into it. For example, a more porous wood will absorb more of the sound into it acoustically, and a tightly grained wood will reflect more, regardless of their density. So acoustically the mahoganies and ashes will absorb, fracture, and dissapate higher frequencies, even though those woods still do have good highs frequency transference. So you have to think along the lines of transference rather than absorbtion/reflection. So I'd try to find a "soundsource" that was contact based.

As far as what's "flat" or what to use as a reference, there can't be one. You should just be looking at the results as variance from eachother, unless you want to pick a wood type and call it the middle ground, or use the one that comes closest to the mean average.

[srt]

here is a semi-related site, interesting:

Modal Analysis of an Electric Guitar

might give some ideas for mounting fixtures, measurement methods, etc. ...

srt

[erikbojerik]

Frank, thanks for those great comments. I hadn't thought of reflection! It is those kinds of things that are giving us pause as to how to transfer the vibrations to the body blank. We've thought about all sorts of ideas, from screwing a speaker directly into the blank, to suspending the blank in front of the speaker in some way.

As far as subtracting the frequency response of the speaker+piezo, I think you're right in that it may not be possible. We're going to perform the experiment with a laminated plywood body and a piece of plexiglass (amorphous, no grain!) as well, that might be the closest we can get!

At the beginning we also thought about using a "mechanical strummer" that would be reproducible in its attack (a heavy pendulum with a pick on the end), and I have a spare neck I could use. Then you could use a piezo bridge for the output. The problem I see is that this approach adds the neck variables (woods, tuners, truss rod) and strings (age) into the equation, not to mention where you fret the neck (if you want to get all the notes on the neck). If we get to keep the body blanks we use, we'll probably try that as a second experiment.

I think we've settled on piezo placement; it will be placed near where the bridge would be located, since this is where the string vibrations are transmitted to the wood. It is neither at the center (resonance!) nor the nodes (depending on body shape of course...).

Amadeus II appears to be a good spectral analysis package for Mac OSX that we'll use for analyzing the piezo output (and it's shareware!)

Amadeus II

Forging ahead

[nollock]

Hi Erik, a place you might find helpfull is the music dsp mailing list. Alot of software developers measure the frequency response of tings they are trying to model. I am sure if you asked there they would tell you the best way to go about it. I suspect you would do best to measure the impulse response. Which means a very short blast of noise (hiting the wood with somthing very quickly might be perfecty good enough) and then you measure the response after that. The quick blast will get some wideband noise into the wood and then you measure the response, ie just record what sound it makes. That recorded sound is the impulse response, a spectrum analasis of that will tell you the frequency response of the wood. You could even use the impulse response to filter audio to see how the differant woods sound aplied to an common audio sample.

I am not sure a constant sine wave or noise will give you a very good measurement because inevitably any affect on the sound will be very hard to measure against the sound source. Its like reverb, you dont actaly notice the reverberation in most rooms untill you make a short sharp noise and then listen for the decay afterwards. Alot of rooms the decay is to short to even notice.

Likewise the way software developers model reverbs is by making a short sharp sound and then measuring the response afterwards. Ie measuring the "impulse response".

Personaly i would attach a piezo to the wood to record the sound, and try hitting the wood with somthing quite quickly. I would use the same object to hit all the test woods as that may affect the sound aswell. Chop the actual sound of the hit off the recorded sample and then do an spectrum analasis of the remainder. Mabey even do it 5 or 6 times and mix the sounds together so you get an average.

Anyway here is the music dsp site.

http://www.musicdsp.org/

If you decide to try there I advise you to join the mailing list rather than the web forum cause the forum is very quiet compared to the mailing list.

chris

[Jimm]

Erik,

I'm gonna add a couple things that might help.

First, I'm nowhere near a sound egineer or acoustical scientist.

But I have a couple ideas for you.

Your tests shoud use the most common "solid" wood used for guitar bodies as your base model. That would be Basswood.

You will need to use actual bodies in your tests to get accurate information. Along the lines of what Frank described about vibration transference, different body shapes and routings will yield different results as they will vary the amount of wood to transfer vibration.

I think you will need only a few bodies, but you will need these bodies for each wood you test.

1. Double Cutaway - Strat Style - Top Route.

2. Double Cutaway - Strat Style - Back Route.

3. Single Cutaway - LP Style.

4. Single Cutaway - PRS Style.

You may need to take into consideration that some bodies have exotic tops and that depending in thickness of these tops, results will vary.

It may be easier to mount a tuning fork into the body, strike the tuning fork and then measure the rate of tonal decay from the fork. You will be able to measure the decay over a wide range of tones depending on the tuning fork used.

Sorry, forgot one thing.

The tuning fork will need to be mounted in the EXACT center of the body for accurate measurement.

[erikbojerik]

Thanks nollock. Someone on the MIMF suggested exactly what you have proposed, and it does sound like a very interesting test (maybe we have 3 to do now...). Our thought was that different wood types don't actually enhance certain frequencies; rather, they damp the ones you don't hear. So we thought the simplest approach would be to put all frequencies into the wood at the same level (e.g. white noise, or alternatively a sine-wave sweep from 20Hz to 20kHz) and see what frequencies end up being damped by different woods.

Ed, your tuning fork approach would be better at sending specific vibration frequencies into the wood, compared with simply placing it in front of a speaker, for sure! You would need to have a very reproducible way of hitting the fork, otherwise the attack and amplitudes would be too variable. Then if you go with different forks, their amplitudes might also be variable compared with, say, A440. I think with that approach it is too difficult to control input levels. White noise, on the other hand, has equal levels at all frequencies (but then you bias the results via the speaker frequency response...sigh...we probably can't get away from that).

What you'd want ideally is one big strong piezo to vibrate the board at all frequencies (or sweep through frequencies) and another to measure what is damped and what is not. I don't know if such a thing exists...

For sure, body shape and top woods will change things (as will neck woods, etc). But we need to keep it simple. If we get to keep the body blanks we get, we'll definitely check to see how shaping the body changes things.

Great comments everyone; this forum is so cool!

[gwm]

"Our thought was that different wood types don't actually enhance certain frequencies; rather, they damp the ones you don't hear."

Yeah, that is exactly right. The wood cannot enhance frequencies, as it does not supply energy to the system. Wood is passive in this sense.

Kinda like why you need batteries for active pickups, to put energy into the system and enhance certain frequencies.

gwm

[nollock]

Our thought was that different wood types don't actually enhance certain frequencies; rather, they damp the ones you don't hear.

Yes, it is that the wood damps some frequencys more than others. The vibrations bounce about inside the material and some get attenuated quicker than others.

What you'd want ideally is one big strong piezo to vibrate the board at all frequencies (or sweep through frequencies) and another to measure what is damped and what is not.  I don't know if such a thing exists...

That would be the best way if you want to measure how the wood effects (or should that be affects?) the sound of the string. Some way of transmitting energy into the wood in the same way a string does. A small speaker attached to the wood could work, but i would attach it so the large metal chunk was in fixed very tightly to the wood. So the cone would act like the vibrating string and the metal frame of the speaker would act like the bridge. You could always butcher a set of cheapo pc speakers.

Its a facinating project

chris

[erikbojerik]

A small speaker attached to the wood could work, but i would attach it so the large metal chunk was in fixed very tightly to the wood. So the cone would act like the vibrating string and the metal frame of the speaker would act like the bridge.

That's a great analogy. The more I think about it, the more I like it better than just hanging the board in front of a cranked speaker. But I think I'd use something that responds well down in the <500 Hz range, like a 12" amp speaker.

[nollock]

A small speaker attached to the wood could work, but i would attach it so the large metal chunk was in fixed very tightly to the wood. So the cone would act like the vibrating string and the metal frame of the speaker would act like the bridge.

That's a great analogy. The more I think about it, the more I like it better than just hanging the board in front of a cranked speaker. But I think I'd use something that responds well down in the <500 Hz range, like a 12" amp speaker.

Good point, you would have to consider the frequency response of the speaker. And no doubt the frequency response of the piezo pickup. Somhow you will need to factor them out of the equation cause each of those could posibly have more effect on the sound than the wood does.

I think you could take measurements with the piezo attached directly to the back of the speaker. That would in theory give you the response of your test equipment. And then you would need to subtract that from your measurements of the wood. Not sure how well it would work though.

The other option would be to have a referance material, pick one material as a neutral and make all the comparisons against that. So as all your comparisons are relative to the referance material, the effects of the piezo and speaker are canceled out. Mabey just test a bunch of materials and use the one with the most plain or flat response as the referance. Or just forget about a referance and just use your info to make comparisons between the materials rather than absolute statments about them?

To be honest i think a 12" speaker would be overkill, the speakers in my studio monitors are only 5 inch and they are quoted as having flat frquency response from 55hz up to 12k. At least within 0.5db across the range. But certainly cheapo PC speakers are not such a good idea now i think about it

chris

[erikbojerik]

Yeah, the frequency response of the speaker sets a bias at the sound source, same for the piezo on the measurement side. I've thought hard about how to subtract it out; we'll probably just look at relative differences between woods, or maybe use plywood or OSB as a baseline.

Maybe placing the piezo on cone might work; we'll try it and see how it looks.

You've got a good set of montor speakers to have such a flat response; what are they?

[nollock]

Yeah, the frequency response of the speaker sets a bias at the sound source, same for the piezo on the measurement side.  I've thought hard about how to subtract it out; we'll probably just look at relative differences between woods, or maybe use plywood or OSB as a baseline.

Maybe placing the piezo on cone might work; we'll try it and see how it looks.

You've got a good set of montor speakers to have such a flat response; what are they?

I have a set of Yamaha MSP5's, active monitors. But i just checked and I overstated the flatness of the response, i think its 55 to 20kz, give or take 5dbs, so I was only a factor of 10 out

chris

[erikbojerik]

OK, I think we have figured out how to subtract the frequency response of the speaker+piezo from the total measurement (speaker+body blank+piezo).

1) Stick piezo on body blank

2) Run test with piezo on the speaker side

3) Flip board over and run test with piezo on opposite side

Doh! It seems simple now....... 8^B <---Brian, we need a smily for this one

Only question left...do we actually fasten the speaker to the body blank? Or just supend the blank in front at some distance? I'm leaning toward attaching the speaker (like the bridge is attached), but I worry about the resonant frequency of the speaker frame dominating the signal.