Official Luthiers Forum!

On numerous spruce-top classicals I've heard, string 1/fret 6 and string 1/fret 11 die out quickly and produce a cluster of discordant pitches. The overall effect is more of a noise than a note.

I've also noticed on the same guitars that open string 1, string 1/fret 7 and string 1/fret 12 produce a twanging sound rather than a clean, single pitch.

1. What causes these problems?
2. Can they be significantly minimized or eliminated by repair work?
3. Will they diminish as the guitar is played over the years?

Paul

It has been my experinece that most such problems are linked to the notes in question being near in pitch, or having partials, at frequencies that are strong resonances of the top or air. For example, there is on many classical guitars a strong resonance near A-440 that seems to be the result of coupling between the top and the air. The top motion can 'feed back' on the string, and cause it to vibrate in odd ways.

Sometimes these things can be remedied, and sometimes not. As usual, you have to understand the nature of the specific problem before you can solve it. This sort of thing is generally more common on 'better' guitars; a little 'nervousness' seems to go along with high performance in many fields. Often simply switching to a different weight of strings can help.

Sometimes these things do diminish as the guitar ages. The wood resonances tend to drop a bit in pitch over time, and this alters the way they couple with both the strings and the air. OTOH, sometimes these things get worse.

My work re-voicing guitars has convinced me that problems such as this are due to the interaction of the ring any particular frequency finds resonance with on the top, and the place on the braces or bridge this ring passes through.

Of course, one must believe that Sound is Round, and that it doesn't like corners or ridges.

On classical guitars, I believe that the high frequencies are generated near the center of the lower bout, which the bridge passes through, and can cause havoc, such as dampening and weird tone. The center of all rings generated by any frequency is in the middle of the bridge on classicals.

If your observation of this consistent problem occurs in differently braced classicals, then it would seem to be related to the traditional bridge design. In general, it causes a weak third or fourth string, with the bevel at the end of the wings, as well as dampening the first string with the tie block corners. I discuss my work on a classical bridge on my web site at the bottom of my re-voicing page.

I would have to examine your guitar to see if I could eliminate the problems you describe, but I have certainly done such corrections on many steel string guitars.

Scott

Thanks for your helpful reply.

I am a professional player looking for a guitar without these problems for a recording project. Do your guitars have them?

I've noticed that most classical guitars cost less than $2000 or more than $4000. I and many other classical guitarists would appreciate more guitars priced in the middle of this range. Could you build a guitar for $3000, and if so, what compromises would you have to make relative to your $4500 basic model?

[QUOTE=Alan Carruth]
...
For example, there is on many classical guitars a strong resonance near A-440 that seems to be the result of coupling between the top and the air. The top motion can 'feed back' on the string, and cause it to vibrate in odd ways.
...
[/QUOTE]

Sorry to stray off the thread but that brings up a question I have been meaning to ask Alan; since guitars have a natural ring or resonance that can couple with a note plucked by the string, why do some luthiers try to brace-shave-tap-tune their tops to a particular note? It seems that would be building-in a strong natural resonance. Why not discourage any particular resonance during the build? In other words how do you make a guitar that responds the same to all frequencies?

Marc,

You raise a basic question that separates my approach from traditional approaches. I have read many articles describing how certain braces are "tuned" to certain notes.

It is my opinion, based on re-voicing existing guitars with countless variations in bracing, that the braces should not be viewed in this manner. I believe that braces have the task of balancing the string forces on the top, spreading their energy throughout the soundboard, and then getting out of the way.

Braces that are heavier than they need to be to balance forces acting at any point along their length, or are shaped in a way that disturbs the flow of energy, will dampen the surface to which they are glued, or inhibit the release of energy into the soundboard.

From an engineering standpoint, bracing patterns on both steel string and classical guitars are considered "statically indeterminate", and do not submit to easy analysis of force resolution. But we all learned from playing on teeter totters that the farther away from the fulcrum, the less force it takes to balance the other side.

Of course, when I have re-voiced a guitar to maximize its full sound production, there are still frequencies any given brace, or the whole soundbox produces when tapped, but once the braces are "invisible" to the soundboard by virtue of being balanced to forces and shaped to allow the energy to flow through them, those frequencies are not reinforced. Close to the entire soundboard can then vibrate.    

It is in this way that I feel the guitar can best respond to all frequencies.

Since the back has little tension from the strings, I believe those braces can be much weaker than traditional to allow the back to reinforce the frequencies generated on the top. There was a recent thread discussing the "tight" back vs. the "loose" back.

I have been fussing with a Martin OOO for days trying to get the open low E to be as loud and punchy as the drop D note and the low G. The culprit turned out to be the ends of the third back brace. Stock, those (and only those) ends dropped lower than their height at the kerfing about one inch in, before starting the scallop up to the original full height. I've never seen this on a guitar before, and it seemed the last possibility causing the E's loss of volume. Plucking the E and feeling over them showed no vibration, whereas the rest of the perimeter of the lower bout vibrated well. It took chiseling the ends down to about half their original height at the kerfing before the parabolic curve I had shaped into the whole brace flowed down evenly to the kerfing. Problem solved.

It is still hard for me to believe that such subtle irregulaties (and even more subtle ones) can influence sound, but time and time again, I have found that they do.

Scott

Classical asked:
"I am a professional player looking for a guitar without these problems for a recording project. Do your guitars have them? "

Y'know, I really couldn't say. I'm not trying to be 'flip' here: your hearing is different from mine, and no doubt much better. The only way for you to know whether my guitars would fill the bill for you is to try one.

Also: this is _not_ the place to be talking business.

Marc asked:
"In other words how do you make a guitar that responds the same to all frequencies?"

The short answer to that is, you can't. At least, it's not possible, IMO, to make a guitar with 'flat' response that would be usable, and if you did, it would probably sound like a tall glass of warm spit. As usual, we're trying for some sort of balance; a tone that is even enough, or that seems even enough, to be useful, with enough variation in detail to be interesting.

A truely 'flat response' system would be one that has very high losses. Since we've got only a limited amount of power to use we can't afford that. Loudspeaker systems are designed for flat response, and are generally terribly inefficient: down in the range of tenths of a percent. You make up for it by using a bigger amp. A vibrating string can only deliver a few hundredths of a watt, iirc, and you need a much more efficient 'speaker' if you're going to get anywhere with that little power. The usual figure cited for the efficiency of the guitar is somewhere in the neighborhood of 5%. That sounds low, but it's actually greater than that of most instruments. The violin is generally taken to be about 2% efficient, for example, and horns are much lower. They get volume by dumping in much more power.

The problem with a higher-efficiency system is that it will have more sharply defined resonant peaks. There isn't a lot you can do about that: it goes with the turf, so to speak. But there are ways to work around it.

One is to place the inevitable resonant pitches in such a relationship to each other that they 'couple'. In the guitar the two lowest resonant pitches are the 'Helmholtz' air mode and the 'main top' mode. The 'Helmholtz' mode is what you get from blowing on an ocarina or a soda bottle. The 'main top' resonance has the top vibrating like a loudspeaker cone. Because the air mooving in and out of the body pushes on the top, and the top pushes on the air, the two modes are strongly coupled, even though they are nearly an octave apart in isolation. The result is that they work together, which has the effect of pushing them further apart in pitch, and broadening the response peaks. All guitars with soundholes do this.

It's also possible to get the 'main back' resonant pitch to couple with the 'main top', by 'tuning' it to be close enough, but not too close, in pitch. This enhances the 'main air' peak in the spectrum, since the back can help move air through the soundhole, and also broadens the 'main top' peak. This tactic, of enhancing coupling between resonant modes, is a general way to even out the response of the guitar. But, there are only so many of these modes, particularly at low frequencies, so small discrepancies can matter a lot.   

Flattening the response also comes at a cost. Our senses are attuned to changes. Things that don't change much, like rocks and trees, are generally neither deadly nor edible, so we don't waste processor power on them if we can help it. We're much more interested in whether that noise in the bush is a chicken or a tiger than we are in the bush itself. If you could make a guitar that had 'flat' response every note would have the same spectral 'recipe': the same balance of fundamental and upper partials. The player could vary that somewhat, of course, but the sound would still be boring. Think of a solidbody electric played 'clean' with the amp at very low power. So we're really looking for a spectrum that gives a different recipe for each note, and allows the player to control that recipe, but, at the same time, makes the tone sound 'even' enough in loudness and timbre that it's not too confusing.

This is _not_ easy. Fortunately, a lot of the work has been done for us by the people who devised the designs we tend to copy. But, 'the Devil's in the details', and, as Scott points out, that gets to be more and more true as the instruments get better. As Dante said, the closer you get to perfection, the more the imperfections matter. Partly the problem is that we're much more sensitive to higher frequencies, and that sensitivity varies a lot from one individual to another. What seems 'balanced' to me, with my particular set of ears, might have all sorts of strong and weak notes to somebody else. We also _learn_ to hear, and get better and better at picking out little things that we used to miss. And, of course, we learn to listen for different things: I'd say that most luthiers really do hear guitiars differently than most players, and sometimes it's hard to communicate those differences.

So, in some sense, we're not trying to make a guitar that responds 'evenly', but we're trying to fool poeple into thinking it does. 'You can't fool all the people all of the time' was never more truely spoken.

So Alan your saying that a flat response guitar would sound pitch A-440 plain in charterer somewhat like a generic midi generates A-440. On pitch but no emotion at allMichaelP39058.637650463

Al, you have a way of explaining your thoughts on these things that is quite unique; you're simply a great teacher. Thanks!

_________________
Rian Gitar og Mandolin

Alan, I’m ignorant when it comes to guitar response curves so please bear with me. Suppose you input a perfect white noise signal (some sharply peaked short duration sound like tapping) into a guitar and then measured a frequency response with some electronics (BTW I haven’t done this but I’m betting you have). Presumably, you would have some band-limited response over some range that falls off at the ends at so many db/octave and some characteristic peaks and valleys throughout the spectrum that result from geometry, material makeup of the guitar, etc. etc..    From what you said it sounds like there are three dominant peaks; air, top, back. This being the case I would think that tap-tuning by ear would favor the top’s peak within that spectrum causing it to become more narrow, in effect creating a resonance at a particular note.

As you say, it’s impossible to prevent a guitar from imparting color to the spectrum, and in fact the color is what gives the guitar its characteristic sound. But, instead of tuning the guitar top to a single particular bell-like frequency, why not deemphasize narrow peaks and attempt to spread more of their energy into broadness? If it’s possible why not merge the peaks to form a sort of bumpy-hill over the range of input frequencies?   Or does this give the bland no volume sound you were referring to?
Marc39058.7296759259

Thanks Michael, that workedMarc39058.7317708333

Most likely the length of the subjuct line has grown too long you can remove some of the re:'s and it will let the edit go through.MichaelP39058.718125

Suppose you input a perfect white noise signal (some sharply peaked short duration sound like tapping) into a guitar and then measured a frequency response with some electronics (BTW I haven’t done this but I’m betting you have). Presumably, you would have some band-limited response over some range that falls off at the ends at so many db/octave and some characteristic peaks and valleys throughout the spectrum that result from geometry, material makeup of the guitar, etc. etc..

Marc, to satisfy your curiosity, here is the kind of response you get with a tap as you describe. I don't recall the exact state of the guitar (it may not have had the bridge yet), but overall it shows the kind of thing you see.

Thanks, Pete: I haven't figured out yet how to post pics on this forum.

The peak in Pete's chart at 104 Hz is usually called the 'main air' peak, and the pair at 205 and 233 are called the 'main top' and 'main back' peaks. This is not strictly a true way to talk about them: acutally, the activity of all of these things is coupled together, and what you're seeing in the chart is a sort of sum of all of them. The actual resonant peaks for the seperate parts (if you could seperate them) would not be at those pitches.

If the box were a really simple shape, and the 'main air' and 'main top' resonant modes were all that was happening, the curve would just fall off above the 'main top' peak in some simple way, like 6dB/octave. The spectrum of every note above that main top pitch would be the same, if you played the string the same way, and the timbre would be boring.

This is the basic design of a 'bass reflex' speaker cabinet. They use speakers that have as broad a frequency output as possible to begin with, and then put them into a cabinet that is tuned to have it's 'air' response at the same pitch. The cabinet is lined with foam or other damping material, to make that peak as broad as possible, too. The result is that the two peaks you see in Pete's chart become more of a low, broad, hill with two minor 'humps', and the response falls off very slowly above that. Again, the price for this is inefficiency: broad response speakers have high damping, by definition, and the damping in the box also dissipates energy. The 'flat' response requires a lot of power to drive to usable levels.

The problem, as near as I can figure it, is that most of the higher resonant modes that provide the 'interesting' features in that chart are 'losers': they tend to subtract power from the output, rather than adding it. This seems to be most true in the range between about 400-1500 Hz, I think. So you need that stuff for 'color', but you don't want too much or you lose power. One of the little conundrums of lutherie.

Another little conundrum that has bitten me here is how to talk about this stuff as truthfully as possible without offending somebody. I've been informed by one poster that my original answer was too flippant, and that he's signing off the list for good. Sigh.   

Humm Alan I have never taken anything you have posted as anything but bound in honest, truthful, scientific observation.

Best advise I could give to someone reading one of your posts would be to read the post, process the data. print a copy and attach to your notebook, absorb it regularly.

As a wise man of science once said "Science is the assimilation of observable facts. Consensus is the act of ignoring observable fact in order to please one's self or one's peers and has no place in science"MichaelP39059.6822800926

Thanks Pete the sound spectra helps my understanding a lot.

Alan thanks for the explanation, the graph Pete posted makes things much clearer. If I understand, the lower peak is created by air, a sort of 'blowing over the top of a bottle' effect and the vibration of the top and back together is coupled with this peak--a squeezing effect.

This low peak is hovering around the low-E string, in fact on this particular graph it is on a G#, so I would expect an unusually strong G# on this guitar. I would also expect that if I could remove the top and back and futz around with the tuning (brace shaving etc.) I could change the shape and strength of this peak, right? I would also expect a strong A# just below the B open 2nd string, and also relatively strong open B string, a weaker open first string then a resonance on G (392)where there is a small peak just below A 440. Would that be correct?

Another question, are the 392 and 771 peaks related to box geometry? The wavelength (in air) for the 771 peak is around the inside dimensions of the box 17 to 18 inches.

Marc wrote:
"If I understand, the lower peak is created by air, a sort of 'blowing over the top of a bottle' effect and the vibration of the top and back together is coupled with this peak--a squeezing effect. "

Again, the lower peaks in the chart are the result of the way the top, back, and air work together on that guitar. A rigid guitar-shaped box would most likely have an 'air' peak at around 125 Hz, and a 'backless' box would probably have a 'top' peak around 180. The coupling moves them apart in pitch. It would not surprise me at all of the 'main back' resonance on that guitar was actually near 210 Hz, in the 'dip' on the chart, but it also contributes to the height of the 'air' peak.

"This low peak is hovering around the low-E string, in fact on this particular graph it is on a G#, so I would expect an unusually strong G# on this guitar."

Possibly; but I don't remember it being all that strong. Hearing is logarythmic, and that really smooths things out.

"I would also expect that if I could remove the top and back and futz around with the tuning (brace shaving etc.) I could change the shape and strength of this peak, right?"

You'd get a lot more change from altering the soundhole size. We _do_ call this the 'main air' peak, and that's because most of the energy is tied up in air motion. You'd have to make some pretty big changes in the top to alter the pitch a lot.

"I would also expect a strong A# just below the B open 2nd string, and also relatively strong open B string, a weaker open first string then a resonance on G (392)where there is a small peak just below A 440. Would that be correct?"

Again, the differences are not all that pronounced in real life. For one thing, any note you play on a string has a whole raft of higher partials, and each of those is also coupling more or less strongly with the air in the room through the higher resonances of the box. What you percieve is the sum of all of that, weighted strongly toward the high frequency stuff. It's quite possible for a note with a fundamental that falls right in a 'dip' in the low-range spectrum to sound very loud, simply because it has a couple of higher partials that drive well. The timbre changes, of course, but that just makes it 'interesting' if the change is not too pronounced. The phenomena the OP was talking about may well have had to do with just such a too-great change between adjacent notes or stretches of notes.

"Another question, are the 392 and 771 peaks related to box geometry? The wavelength (in air) for the 771 peak is around the inside dimensions of the box 17 to 18 inches."

Possibly, but it's not that simple. The 'A-1' air mode on most guitars, which is the lowest lengthwise 'pipe' resonance, is usually around 350 Hz. It is set by the box length, mostly, but the box is rounded off, and the walls are somewhat compliant, and those effect the pitch. It also generally couples with the top 'long dipole' mode, which is often near the same pitch, so you tend to get two peaks, in the same way that the 'main top' and 'main air' pitches couple. On top of THAT, there seems to be some influence both from the waist and the soundhole location. The upshot is that I would not be at all surprised to find that the two peaks at 365 and 392 were due to that top-air couple, or even to find that one part of that couple produced a dip, say the one near 350, or the low peak at around 425.

As you get much higher than that in pitch it becomes more and more difficult to assign any peak or dip in the output to any one or two resonances. Resonances start to come in so close together up around 600-700 Hz or so that we tend to talk about a 'resonance continuum'. It's possible to move a particular peak in that range, if you want to, but difficult, at best, to predict in advance what those peaks will be, or what the result will be someplace else if you move one.

The good news is that, in some sense, it doesn't matter as much just where the peaks and dips are up there, so long as you've got them. Those high-end peaks seem to translate into 'tone color'; too few and it sounds 'uneven' too many and it can sound 'flat', too peaky and it can sound 'harsh', too flat and it can sound 'dull'. Other than not wanting them all bunched up in one place, that's about it, as far as I can tell. A different distribution will probably sound 'different', but may not sound any 'better' just so it meets the other criteria. Pete's guitar, BTW, seems to fill the bill.     

Alan, thanks for sharing what is obviously years of study and experience on this topic. You've definitly broadened my understanding of the physics of the guitar and I'm grateful.

It's clearly a complicated topic and my atempts to boiler plate it down are tending to grow the subject. It's obviosly difficult to condense to a practical issue of construction. Since I'm unable to answer the question for myself, I'll just ask:

Without aid of modern lab equipment, is there a way for me to use this information, at a the workbench, to make a better sounding guitar? Some general rules of thumb maybe? Tap tuning?

Marc asked:
"Without aid of modern lab equipment, is there a way for me to use this information, at a the workbench, to make a better sounding guitar? Some general rules of thumb maybe? Tap tuning? "

Once you've got some experience looking at the modes you can often tell a lot about what's going on from tapping and feeling the way the plates flex. I think this is the way the old boys did it, after all, and some of them got it to work pretty darn well. The problem is that, without knowing what's happening, it's hard to know what you're hearing and feeling.

The great thing is that getting 'lab equipment' is not nearly the hurdle it used to be. A junk-level PC with some freeware and shareware, a sound card and, maybe, an amplifier, can do most of what took a wall full of lab equipment to do twenty or thirty years ago. Of course, it still takes a while to learn to do this stuff, and to figure out what the results mean. Still, the actual equipment isn't the problem it once was.

I'll warn you, though, that, as addictive as lutherie is, lutherie science is worse. And _nothing_ takes more time than science. As one researcher put it, the problem with science is that you're always doing something you're not good at. By the time you do get good you've found out what it was you were looking for, and have to go on to some other new thing that you're not good at. The safe thing is just not to get started. ;)