Tap tuning as a rough guide when shaving the braces

Construction and repair of Classical Guitar and related instruments
amezcua
Posts: 593
Joined: Tue Oct 04, 2016 9:50 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by amezcua » Thu Jan 18, 2018 9:14 pm

printer 2 the coloured diagrams at the end of each write up all looked much the same to me. I must admit I learned nothing at all from those pictures . In the picture with divided areas the bit that struck me was how dead the bridge area was. Surely that could be improved if guitarists want more sound . It reminded me how incredibly heavy a cat can be when it decides to sleep on your bed .

Alan Carruth
Luthier
Posts: 2987
Joined: Mon Dec 24, 2007 6:56 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by Alan Carruth » Fri Jan 19, 2018 6:37 pm

Most of the models I've seen have left out the very important element of the air. This can totally change the way the guitar works, even in the low to mid range, and it's not very well understood.

I'm convinced that the guitar is an irreducibly complex system, and deliberately so. All of that complexity is what gives us the tone color that we rely on. It also means that it may well be impossible to make 'tonal copies' that are exact mimics. 'Different', of course, may not be 'better' or 'worse', it's just 'different'. Even if we can't exactly duplicate a particular great guitar, we may be able to make one that is equally great in a slightly different way. That's the hope, anyway.

User avatar
prawnheed
Posts: 705
Joined: Wed Dec 27, 2017 9:07 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by prawnheed » Fri Jan 19, 2018 7:18 pm

Alan Carruth wrote:
Fri Jan 19, 2018 6:37 pm
Most of the models I've seen have left out the very important element of the air. This can totally change the way the guitar works, even in the low to mid range, and it's not very well understood.

I'm convinced that the guitar is an irreducibly complex system, and deliberately so. All of that complexity is what gives us the tone color that we rely on. It also means that it may well be impossible to make 'tonal copies' that are exact mimics. 'Different', of course, may not be 'better' or 'worse', it's just 'different'. Even if we can't exactly duplicate a particular great guitar, we may be able to make one that is equally great in a slightly different way. That's the hope, anyway.
They also miss the way the top is "pre-stressed" by the strings, imperfections in the materials and many other significant factors. I, like you, don't believe there is much hope for a FEA model to reproduce the complexity of the real guitar to a level required to understand how it would sound. Not even to the level of whether it would sound like a guitar or, for example, a banjo.

What I think might be feasible is an AI system that could be trained to predict how a guitar would sound based on the way the top sounds before fitting to the rest of the guitar. This is a fairly classic, if somewhat complex, application of a supervised learning system and, with some careful consideration of the data fed to it, it might get better rsults than a human at the tap test. Alternatively, it might show that tap testing is actually no use at all as a predictor. Either outcome would be interesting.

amezcua
Posts: 593
Joined: Tue Oct 04, 2016 9:50 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by amezcua » Fri Jan 19, 2018 10:33 pm

In the Acoustic Masters site they advise the Back of a guitar to be one octave higher than the Top . That`s after it`s put together , but for the original question they say some makers (then) shave the braces and sand the inner surface to achieve the right tuning. So braces and top inner surface are both involved in tuning ( they say ). That contradicts a previous post but it`s good to sort out contradictions (sometimes). I don`t think any specific frequency is as important as their relationship of an octave. So compare the back and top before you assemble the guitar and see what happens afterwards.

Alan Carruth
Luthier
Posts: 2987
Joined: Mon Dec 24, 2007 6:56 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by Alan Carruth » Sun Jan 21, 2018 5:40 pm

In fact, there are some things about the way the guitar works that differentiate it from other similar instruments, and I'm pretty sure that at least one of them has to do with the air in the box.

Some years back one of my students rigged up a small accelerometer for me. I stuck it on the bridge of a guitar, hung it up with the strings damped, and used a 'sweep' signal from my computer to drive the bridge from about 50-1000 Hz. I recorded the output of the accelerometer on the computer, and ran it through a spectrum analysis program. The result was a chart with clear peaks and dips that showed how active tat spot on the top was at every frequency. I then used Chladni patterns to find all the resonances of the top and back that I could, to see how they corresponded with the peaks and dips on the chart. As you'd expect, all the lower top resonances that were active at the bridge showed up as peaks. The back resonance tended to show up as dips, since the back gets all of it's energy from the top. I was able to account for all of the peaks pretty well, except for one very close to A=440; fifth fret on the high E string. There was no 'wood' resonance near that pitch, and the closest 'air' resonance in the literature was at about 350Hz. Other tests showed that the peak was really there in the played sound, and the sound was coming out of the hole. A microphone inserted into the hole showed that the resonance was, in fact, an 'air' mode; in fact, the one that was supposed to be at 350 Hz. This is sometimes called a 'lengthwise bathtub' mode, as the air 'sloshes' the length of the body. The sound level is high at either end of the box, and there's a 'null' in the middle (you can think of this as analogous to a string that you pluck while touching it in the middle, at the 12th fret. A pickup in the middle would hear no sound). In fact, there was a resonance like this at 351 Hz and another at 438Hz !

One of the rules about this stuff is that you won't see the same resonance shape at two different frequencies unless there's another resonance coupling in and influencing it. In fact, the two 'air' resonances, at 351Hz and 436Hz, had slightly different shapes: the pressure nulls were in a somewhat different place at the two frequencies. I thought this might have to do with the sound hole and the waist.

I put together a simplified test using a cardboard mailing tube the length of the guitar body. I made up a 'waist' out of foam coffee cups that could be inserted in the right place, a little above the center of the length. I drove the air through a tube hooked into a loudspeaker at one end, and put a microphone in the other end to measure the sound level along the length of the tube. I then looked at the mode under four different conditions. One was the closed tube. Anther was the closed tube with the 'waist' inserted. The third was the tube with no waist, but a sound hole in the 'right' position, just above the waist. Finally, I looked at the mode with both the waist and the hole.

The plain tube had the mode near 350 Hz with the null exactly in the middle. Adding the waist changed the pitch a little, and moved the null upward, due to the speeding up of the air flow through the restriction of the waist. The hole by itself moved the null downward, and dropped the pitch a little. There is some sound produced at the hole, and that moves air in and out, which, apparently, increases the effective mass of the air in the upper end of the tube. Finally, with both the waist and the hole I got both modes. I wrote up a short paper on this, and submitted it to the Catgut Acoustical Society journal.

A few weeks later the subject of 'air' modes came up on a violin maker's list, and I mentioned this. I got an e-mail from a physics professor in England expressing some doubt, so I sent him a copy of the paper. A few weeks later I got an e-mail back from him saying that he had checked my result using a ceramic drain pipe and a roll of lead for the waist, and had not seen what I saw. "Check your apparatus". I withdrew the paper, and got back to work.

The long and short of it is that the split into two modes only happens when the walls of the box can vibrate. With rigid walls the hole and waist produce a single 'compromise' mode, but when the walls can move you see two, but only if there is a pronounced waist above the center of the length of the box, AND a sound hole just above that. 'Dreadnought' guitars without a well defined waist don't show this, and have a different characteristic timbre. Similarly, when the sound hole is moved up into the corner of the box to drop the 'main air' pitch the effect goes away.

This was not, of course, something that was planned when the shape of the guitar was first devised. Rather, they made the shape, and found that it had a 'characteristic' timbre that was pleasant. This is one of the reasons we keep making them this way: if you don't they don't sound as much like guitars.

A few years after that a group in Spain published a finite element model of the guitar that included both the wood and the internal air. This 'added' air resonance, which had been missed in tests in which the box was embedded in sand to immobilize the walls, showed up clearly. It took a pretty complicated model to catch it.

Making such a model that would behave exactly like a specific guitar would be very difficult. All the very small inconsistencies in the wood properties from point to point would need to be modeled, and you'd need t measure them first. This would pretty much require chopping up the wood into little pieces and measuring them all, and then, of course, you won't have the guitar any more. We can learn a lot from these models using 'average' values of the materials, but there are limits.

Given the normal sort of variation that you get in wood, I strongly doubt that exact 'tonal copies' can be made. Tight control, which IMO includes things like 'free' plate tuning, can get you 'arbitrarily close' though.

gjo
Posts: 321
Joined: Wed Sep 16, 2009 8:48 am

Re: Tap tuning as a rough guide when shaving the braces

Post by gjo » Sun Jan 21, 2018 6:23 pm

Alan, that is very interesting to read. Thank you very much for your detailed description.

But ... how did the guitar you analyzed sound in your ears?

User avatar
Chris.Conery
Amateur luthier
Posts: 125
Joined: Sun Jan 17, 2010 10:29 pm
Location: Boulder, Colo.

Re: Tap tuning as a rough guide when shaving the braces

Post by Chris.Conery » Sun Jan 21, 2018 7:31 pm

My two cents on this interesting discussion (from someone whose background is in physics and who has built almost 20 classical guitars in the last ten years):

I think it is safe to say that there are two areas of classical physics that have not been "solved" - fluid flow and acoustics, there being SO many variables that consistent solutions are difficult to find. Classical guitar construction is a perfect example of this. That is not to say that we shouldn't make the measurements and do and share the research - how else can knowledge advance?

It took me several guitars to develop a pretty consistent approach to tap tuning the top as it is thinned and the braces shaved to recognize the sound that would result in consistently good sounding guitars. I can recognize the "sound" but can't really describe it in a way that would be useful to others. Then I took a very enlightening weekend class from Robbie O'Brien the week after he returned from Jose Romanillos' course where Romanillos had taught his method of flexing the top until it felt "right". At the end of the weekend with Robbie, with the tops of the five guys in the class flexing properly (ala Romanillos) I held the top to my ear, tapped it, and there was the "sound" that I am always searching for.

The moral to this story is that there are different ways to get to the same destination. And that our fingers and ears are excellent instruments. And that there is still more art than science in this craft.
Conery Guitars
Chris Conery, luthier

amezcua
Posts: 593
Joined: Tue Oct 04, 2016 9:50 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by amezcua » Mon Jan 22, 2018 1:06 pm

There is a thesis about the effect of adjusting violin soundposts and negatively refers to subjective reactions to the sounds. These are dismissed as "wrong" in the final write up , although descriptive words such as harsh and nasal and dull sound easily understandable . The objective attitude using "finite element analysis " gives a haughty self important impression . I don`t like long dissertations that use unfamiliar words . Admittance is one of those ( to me ) and looking it up it is nothing more impressive than the opposite of Resistance . Rejecting musician`s reactions for using familiar subjective words is a very irritating habit among "experts". Experts is another word I might have to look up .
But the posts above have done well to clarify some guitar mysteries . Good topic .

User avatar
prawnheed
Posts: 705
Joined: Wed Dec 27, 2017 9:07 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by prawnheed » Mon Jan 22, 2018 1:26 pm

Alan Carruth wrote:
Sun Jan 21, 2018 5:40 pm
In fact, there are some things about the way the guitar works that differentiate it from other similar instruments, and I'm pretty sure that at least one of them has to do with the air in the box.

Some years back one of my students rigged up a small accelerometer for me. I stuck it on the bridge of a guitar, hung it up with the strings damped, and used a 'sweep' signal from my computer to drive the bridge from about 50-1000 Hz. I recorded the output of the accelerometer on the computer, and ran it through a spectrum analysis program. The result was a chart with clear peaks and dips that showed how active tat spot on the top was at every frequency. I then used Chladni patterns to find all the resonances of the top and back that I could, to see how they corresponded with the peaks and dips on the chart. As you'd expect, all the lower top resonances that were active at the bridge showed up as peaks. The back resonance tended to show up as dips, since the back gets all of it's energy from the top. I was able to account for all of the peaks pretty well, except for one very close to A=440; fifth fret on the high E string. There was no 'wood' resonance near that pitch, and the closest 'air' resonance in the literature was at about 350Hz. Other tests showed that the peak was really there in the played sound, and the sound was coming out of the hole. A microphone inserted into the hole showed that the resonance was, in fact, an 'air' mode; in fact, the one that was supposed to be at 350 Hz. This is sometimes called a 'lengthwise bathtub' mode, as the air 'sloshes' the length of the body. The sound level is high at either end of the box, and there's a 'null' in the middle (you can think of this as analogous to a string that you pluck while touching it in the middle, at the 12th fret. A pickup in the middle would hear no sound). In fact, there was a resonance like this at 351 Hz and another at 438Hz !

One of the rules about this stuff is that you won't see the same resonance shape at two different frequencies unless there's another resonance coupling in and influencing it. In fact, the two 'air' resonances, at 351Hz and 436Hz, had slightly different shapes: the pressure nulls were in a somewhat different place at the two frequencies. I thought this might have to do with the sound hole and the waist.

I put together a simplified test using a cardboard mailing tube the length of the guitar body. I made up a 'waist' out of foam coffee cups that could be inserted in the right place, a little above the center of the length. I drove the air through a tube hooked into a loudspeaker at one end, and put a microphone in the other end to measure the sound level along the length of the tube. I then looked at the mode under four different conditions. One was the closed tube. Anther was the closed tube with the 'waist' inserted. The third was the tube with no waist, but a sound hole in the 'right' position, just above the waist. Finally, I looked at the mode with both the waist and the hole.

The plain tube had the mode near 350 Hz with the null exactly in the middle. Adding the waist changed the pitch a little, and moved the null upward, due to the speeding up of the air flow through the restriction of the waist. The hole by itself moved the null downward, and dropped the pitch a little. There is some sound produced at the hole, and that moves air in and out, which, apparently, increases the effective mass of the air in the upper end of the tube. Finally, with both the waist and the hole I got both modes. I wrote up a short paper on this, and submitted it to the Catgut Acoustical Society journal.

A few weeks later the subject of 'air' modes came up on a violin maker's list, and I mentioned this. I got an e-mail from a physics professor in England expressing some doubt, so I sent him a copy of the paper. A few weeks later I got an e-mail back from him saying that he had checked my result using a ceramic drain pipe and a roll of lead for the waist, and had not seen what I saw. "Check your apparatus". I withdrew the paper, and got back to work.

The long and short of it is that the split into two modes only happens when the walls of the box can vibrate. With rigid walls the hole and waist produce a single 'compromise' mode, but when the walls can move you see two, but only if there is a pronounced waist above the center of the length of the box, AND a sound hole just above that. 'Dreadnought' guitars without a well defined waist don't show this, and have a different characteristic timbre. Similarly, when the sound hole is moved up into the corner of the box to drop the 'main air' pitch the effect goes away.

This was not, of course, something that was planned when the shape of the guitar was first devised. Rather, they made the shape, and found that it had a 'characteristic' timbre that was pleasant. This is one of the reasons we keep making them this way: if you don't they don't sound as much like guitars.

A few years after that a group in Spain published a finite element model of the guitar that included both the wood and the internal air. This 'added' air resonance, which had been missed in tests in which the box was embedded in sand to immobilize the walls, showed up clearly. It took a pretty complicated model to catch it.

Making such a model that would behave exactly like a specific guitar would be very difficult. All the very small inconsistencies in the wood properties from point to point would need to be modeled, and you'd need t measure them first. This would pretty much require chopping up the wood into little pieces and measuring them all, and then, of course, you won't have the guitar any more. We can learn a lot from these models using 'average' values of the materials, but there are limits.

Given the normal sort of variation that you get in wood, I strongly doubt that exact 'tonal copies' can be made. Tight control, which IMO includes things like 'free' plate tuning, can get you 'arbitrarily close' though.
Indeed fluid mechanics and the related subject of acoustics are complex. The mathematical models are very approximate and one of the challenges in making physical models is the scaling issue. An effect seen on one scale, does not necessarily replicate easily on another scale. That's possibly why your experiements with tubes and waists did not replicated with different sized tubes and waists. I'd be similiarly wary of extrapolating from your result to the scale of the guitar.

I also remain dubious of the detiled accuracy of any FEA model that includes air resonances - the movement of air is going to be turbulent in places, laminar in other places and the way the air and wood conduct sound in and around something with waists, braces, holes, edges, etc. is just not feasible to model in that way.

User avatar
prawnheed
Posts: 705
Joined: Wed Dec 27, 2017 9:07 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by prawnheed » Mon Jan 22, 2018 1:35 pm

amezcua wrote:
Mon Jan 22, 2018 1:06 pm
There is a thesis about the effect of adjusting violin soundposts and negatively refers to subjective reactions to the sounds. These are dismissed as "wrong" in the final write up , although descriptive words such as harsh and nasal and dull sound easily understandable . The objective attitude using "finite element analysis " gives a haughty self important impression . I don`t like long dissertations that use unfamiliar words . Admittance is one of those ( to me ) and looking it up it is nothing more impressive than the opposite of Resistance . Rejecting musician`s reactions for using familiar subjective words is a very irritating habit among "experts". Experts is another word I might have to look up .
But the posts above have done well to clarify some guitar mysteries . Good topic .
The use of language specific to a domain is natural and necessary. The language of music is full of terms which have no or different meaning to non-musicians.

The terms "harsh", "nasal" and "dull" might be well understood by you, but essentially useless to a scientist or mathematician because they are subjective and ill-defined: two people will use and understand the same terms to describe different frequency spectrums.

Alan Carruth
Luthier
Posts: 2987
Joined: Mon Dec 24, 2007 6:56 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by Alan Carruth » Mon Jan 22, 2018 5:31 pm

First: the guitar in question was a good one. Generally speaking the ones I have checked that have not shown this particular mode couple have not been as good. The exception to that being, of course, Dreadnoughts, in which that timbre is 'characteristic', and even then, the 'better' ones do tend to show this.

My understanding, from back when I messed around with model airplanes, is that much of the scale issue relates to Reynolds number: for dynamic similarity you need to have similar Reynolds numbers. The Reynolds number is a function of length in the direction of flow and velocity, for the most part, iirc. In the case of the guitar most of the flow involved is at fairly low velocities, and all guitars are similar enough in size that it's not an issue. I could, of course, be wrong about that. At any rate, the prof who duplicated my experiment used a similar size rig that simply eliminated the wall flexibility, and that's what made the outcome different.

Certainly you have to take the output of any model with a generous dose of seasoning. On the other hand, with appropriate caution, we can learn a lot from good models. We also need to be cautious about over simplified experiments. In the case I cited the model showed something that I had seen in real guitars that had not been seen when experimenters simplified the system too much.

Several years ago one of the denizens of a violin maker's list I'm on talked about a study that had been done on tone descriptors. People, probably at a conservatory, were asked to list terms they would use to describe violin tone. The huge list ("chocolaty") was pared down to a smaller number of terms. Careful recordings were made of the same violinist playing the same short passage on a number of instruments. These were played back in random pairs through headphones, and the listeners were asked to rate them in terms of one of the chosen descriptors: "Which is more 'open', A or B?" They found three things.

The first was that each listener used the terms in a different way: you're 'bright' might be my 'nasal'.

The second was that each listener used the terms consistently: if 'A' was 'darker' than 'B', and 'B' was 'darker' than 'C' to a particular person, then they would always hear 'A' as 'darker' than 'C'.

The third, and most interesting to me, was that if a particular instrument had some quality that a listener did not like, they would be unable to rate it on other qualities. If violin 'A' is 'nasal', and you don't like that, then you will find it hard to tell if 'A' is more 'even' than 'B'.

I'll note that it often seems to be the case that a number of people will agree on the 'best' instrument in a collection, even when they tend to describe the tone differently, and sometimes using diametrically opposed terms. At any rate, in any test of acoustic modification, it's always best to keep the evaluation to 'same' or 'different', without dragging in subjective terms. Much of what we do in studying instrument acoustics is to try to fine objective things that tend to correlate well with listener preferences.

All trades do, of course, have jargon, much of which does have a specific meaning. This can vary with context: the 'saddle' on a violin is a different part than the 'saddle' on a guitar. Much of the trouble most of us have in reading scientific articles has to do with the necessary jargon. 'Admittance' is a good example: electrically 'resistance' refers to dissipation in a direct current system that opposes the flow of current, while the analogous term in AC circuits is 'impedance'. Resistance is part of impedance, but there are other factors as well. In a mechanical system, like a guitar impedance is Force/velocity at a given frequency, and is affected by mass and stiffness as well as dissipation from friction and other mechanisms. It gets complicated, and if you had to go through the whole definition every time the subject came up you'd never get anywhere. For somebody who's not a guitar player 'rest stroke' and 'free stroke' are equally arcane, let alone 'demiquaver'.

User avatar
prawnheed
Posts: 705
Joined: Wed Dec 27, 2017 9:07 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by prawnheed » Mon Jan 22, 2018 5:45 pm

The Reynold' number helps with the airflow analysis, but it doesn't help with the other factors like the differences in speed of sound in the wood versus the air or the diffraction of airborne sound waves around the complex shapes. Sound also travels in different was in a solid - the compression waves like in air, but also transverse waves. The sound transmitted through the wood will create interference patterns with the sound transmitted through the air.

Each of the factors scale differently so one must be very careful in looking at scaled or simplified models.

Alan Carruth
Luthier
Posts: 2987
Joined: Mon Dec 24, 2007 6:56 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by Alan Carruth » Mon Jan 22, 2018 6:16 pm

As I said, one always has to take model results with copious quantities of salt. OTOH, anybody who has driven a car or flown in an airplane in the past decade or so has bet their life on the reliability of computer models. Even fairly crude models of guitar function can be quite useful to a savvy maker. Check out Howard Wright's PhD thesis: "The Acoustics and Psychoacoustics of the Guitar", given at U/Wales in Cardiff in 1996 for an example. The model is almost laughably simple by today's standards: it includes only one air resonance, for example, and the wood resonant modes are treated as flat pistons of equivalent area and mass, but for all that the results are quite useful.

I hope this thread doesn't descend into an' art vs. science' argument. For me, lutherie is a craft, striving to make better tools for the real artists; the musicians. Although there is much that is, and will always be, beyond rational or quantitative understanding in this, I'll use any tool that will help me to improve my work. Our craft has always followed this path; nobody sets out frets by eye any more. ;)

User avatar
prawnheed
Posts: 705
Joined: Wed Dec 27, 2017 9:07 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by prawnheed » Mon Jan 22, 2018 7:15 pm

My points are in now way attempting to devalue science. It is just that FEA, and other types of models, has limits and is not likely to answer the question posed by the thread. Other science, for example the AI approach I mentioned, possibly could.

Alan Carruth
Luthier
Posts: 2987
Joined: Mon Dec 24, 2007 6:56 pm

Re: Tap tuning as a rough guide when shaving the braces

Post by Alan Carruth » Tue Jan 23, 2018 4:48 pm

The AI/'machine learning' approach is, of course, an automated version of what makers have been doing for a long time. It might indeed succeed.
One issue, though, is that IMO it's the 'free' plate mode shapes rather that frequencies that matter the most, and it might take some effort to cast that into machine language. Then you need to make a lot of guitars....

Return to “Luthiers”