After completing my 3rd classical build, I feel like I've got a handle on the woodworking challenges. I still have a lot to learn, and it will be several more builds before I feel confident with every step of the process, but for now I feel like I know what I don't know.
Right now I'm becoming interested in just how we can control the acoustics or "voice" of our instruments in a reasonably predictable manner. After attending a talk by luthier Brian Burns of Fort Bragg, CA on using spectral analysis to test tone woods and voice tops, I want to duplicate his processes in as reliable way as possible. That is, using off-the-shelf components and open-source software.
To that end I'm starting to build his apparatus for driving the tops of instruments using a speaker coil and rare earth magnet, and then recording the resulting sound using an open source spectral analysis program. This thread will document my progress.
I have very little background in electrical engineering or acoustics - everything I know I've been picking up from the luthier community. So I'd like to give back by documenting what I'm doing in as clear a way as possible.
For those interested in Brian's procedures I'd refer you to this page on his web site for starters. The basic idea is this. We fix a rare earth magnet to our guitar top. Over that we position a coil from a speaker driven by an amplifier. Through that we send a basic tone (sine wave) at known frequencies. Brian uses a program to generate a constantly ascending tone from around 60 to 1450 Hz, as I recall. In theory, the energy or amplitude of the signal driving the top is constant, or at least a smooth curve. But if we capture the resultant sound with a microphone (essentially, letting the top amplify the signal just like it does a plucked string), we find that some frequencies are much louder than others, while others are quite diminished. Figuring out where these peaks and valleys occur can help in shaping the overall response of the instrument.
Stay tuned!
Building an acoustic driver to test tops and tune instrument
-
Dave Bagwill
- Posts: 5951
- Joined: Tue Dec 13, 2011 7:44 pm
Re: Building an acoustic driver to test tops and tune instru
I've studied with Brian a few times and he is very good luthier. I too have thought of copying his top tester - I already copied his jig for initial 'Q' testing - but the spectrum analyzer he uses is pretty expensive - $700 or so - an amazing program, but out of my reach for now.
Keep us posted on your progress.
Keep us posted on your progress.
-Under permanent construction
Re: Building an acoustic driver to test tops and tune instru
I will be interested in seeing this.
Jim.
Jim.
-
Dave Bagwill
- Posts: 5951
- Joined: Tue Dec 13, 2011 7:44 pm
Re: Building an acoustic driver to test tops and tune instru
That freebie is a good one for a lot of things.
I does not do the elusive 'Q' thing that is key for some luthiers, though.
But if you don't need the Q, it works fine.
Thanks Ken.
I does not do the elusive 'Q' thing that is key for some luthiers, though.
But if you don't need the Q, it works fine.
Thanks Ken.
-Under permanent construction
Re: Building an acoustic driver to test tops and tune instru
I've managed to run a first test of my setup, and recorded a reasonable sounding (and looking) track from my most recent classical build. I'm just trying to get the audio setup roughly calibrated, so I'm using a completely built instrument (strung up to tension) to test the apparatus. I'll post pictures at some point.
Regarding the spectral analyzer - there's an open-source, cross-platform program called Audacity I've been playing with. Of the free/open-source packages out there, this seems the most robust and flexible. Critically, it will take a sound file spanning a couple of minutes and produce the spectral profile of the entire sample. In other packages, this display is dynamic (produces the spectrum of very small slices of time, dynamically). Even Spectra-Plus appears to do this - I set it to track the maximum peaks as it analyzed the 2 minute sound sample, and got something that looks like Brian's diagrams.
I'm at this point less worried about the "Q" measure, here's why. I've found a great source in David Hurd's book Left-Brain Lutherie. I can't recommend this volume highly enough for any engineering-oriented builders. He hits a sweet spot of requiring a relative minimum of math and physics knowledge (high school level) to follow along, while also gathering up models, procedures, and ideas from a variety of more technical sources.
From reading Hurd's volume, Brian's site, plus other sources, I'm gathering there are some relatively sound (no pun intended) principles of construction and plate tuning that will narrow the random variability of sound from build-to-build, as well as raise the probability of a producing a well-balanced instrument. It's not like we can just set some parameters and automatically dial in a Hauser sound or a Ramirez sound, but finding consistent ways in which instruments we like vary (and compare to instruments we don't like) does shed some light on where things can go right or wrong.
So for now I'm going to focus on tuning the plates (back and top) using electronic wizardry, and paying attention to the stiffness of the top and its resonances. If I can build my next guitar while staying reasonably within some recommended guidelines derived by Hurd, I'll call it a success.
Back to the Q factor - it seems to me this is a useful quantity for sorting wood at the very beginning of a build. It's a numeric measure of the "liveliness" of a piece, as far as I can tell. Using Audacity, I believe it should be possible to export the spectra as a table of numbers and use a spreadsheet to calculate Q for certain peaks. I'll post whatever progress I can make on that.
My goal is to do everything as open-source/freely as possible. A $700 piece of software is a big risk for just fiddling around - Audacity seems well-supported and could be a reasonable substitute.
Regarding the spectral analyzer - there's an open-source, cross-platform program called Audacity I've been playing with. Of the free/open-source packages out there, this seems the most robust and flexible. Critically, it will take a sound file spanning a couple of minutes and produce the spectral profile of the entire sample. In other packages, this display is dynamic (produces the spectrum of very small slices of time, dynamically). Even Spectra-Plus appears to do this - I set it to track the maximum peaks as it analyzed the 2 minute sound sample, and got something that looks like Brian's diagrams.
I'm at this point less worried about the "Q" measure, here's why. I've found a great source in David Hurd's book Left-Brain Lutherie. I can't recommend this volume highly enough for any engineering-oriented builders. He hits a sweet spot of requiring a relative minimum of math and physics knowledge (high school level) to follow along, while also gathering up models, procedures, and ideas from a variety of more technical sources.
From reading Hurd's volume, Brian's site, plus other sources, I'm gathering there are some relatively sound (no pun intended) principles of construction and plate tuning that will narrow the random variability of sound from build-to-build, as well as raise the probability of a producing a well-balanced instrument. It's not like we can just set some parameters and automatically dial in a Hauser sound or a Ramirez sound, but finding consistent ways in which instruments we like vary (and compare to instruments we don't like) does shed some light on where things can go right or wrong.
So for now I'm going to focus on tuning the plates (back and top) using electronic wizardry, and paying attention to the stiffness of the top and its resonances. If I can build my next guitar while staying reasonably within some recommended guidelines derived by Hurd, I'll call it a success.
Back to the Q factor - it seems to me this is a useful quantity for sorting wood at the very beginning of a build. It's a numeric measure of the "liveliness" of a piece, as far as I can tell. Using Audacity, I believe it should be possible to export the spectra as a table of numbers and use a spreadsheet to calculate Q for certain peaks. I'll post whatever progress I can make on that.
My goal is to do everything as open-source/freely as possible. A $700 piece of software is a big risk for just fiddling around - Audacity seems well-supported and could be a reasonable substitute.
-
Dave Bagwill
- Posts: 5951
- Joined: Tue Dec 13, 2011 7:44 pm
Re: Building an acoustic driver to test tops and tune instru
Audacity is an excellent program. I agree on the $700 issue. :-)
-Under permanent construction