What to do first to improve the sound of a recently complete classical guitar

[Simon Magennis]

I want to "revisit" some guitars which don't sound quite the way I would like. Two were using the Brune GAL 1912 Ramirez plan. I followed it as closely as I ever do.

The first one I want to tackle sounds pretty balanced and overall I am fairly pleased but the trebles, and in particular the high e string, don't sing as I would like. It is not as responsive as I would like. I am going to partially refret it in any case as it is a bit dead at a couple of frets and I am pretty sure that part of the issue is just the way the frets are seated.

The things I know I could do:

Shave the bridge wings a bit.
Go in and adjust some top braces.
Removed the top finish and thin the top a little more.
Add some weight to the sides.
Possibly adjust something on the back braces.

Methods for determining next step:

stick some additional weigh on the top and test the sound by ear
tap the top and record the patterns with audacity or some such
set up to test chadlni patterns

Given that the instrument is not as responsive as I would like, then it seems to me that a slight reduction in the stiffness of the system is what I need. But as it seems to me that the sound is balanced, the question is how to get response without loosing balance. Willy nilly sanding the top could well ruin that balance. So what would be the recommended first step here?

In particular using the various scientific methods, what would indicate more response?

I have both the Somogyi and Goreand Gillet books - so if there is a miracle answer in either those, a page number would be great.

[Alan Carruth]

I'd look first for some way to reduce the mass of the bridge, without necessarily reducing the stiffness. This makes it easier for the strings to move, and particularly at high frequencies. What is the bridge material, and do you know what it weighs?

Do you know the pitch of the 'cross tripole' resonant mode? That's usually in the 500 Hz range, and can be a good radiator of sound. If it's too high pitched or not very active reducing the height of the bridge wings at the tieblock end can lower it and make it more active. That could also drop the dipole pitch, but it's less likely. It will probably drop the monopole pitch, and add a bit of power there as well. Since the monopole has such a long 'tail' in the spectrum that might actually get you some more treble. I would not do this until I'd reduced the mass of the central section first.

I'd avoid reducing the top stiffness for the most part. In some cases, however, lowering the pitch of the 'cross dipole' frequency can add a little 'cut' to the sound. You could try shaving the outside fans a little, or reducing the stiffness of the top in the 'wings' between the bridge ends and the edge. Don't reduce the top thickness at the lower block if you can help it; that tends to make the bass 'fuller' in my experience.

Iirc, the '12 Ramirez had a pretty thin top. He may well have gotten a stiffer and denser piece of spruce than you did, in which case copying his thicknesses would leave you with a top that's not stiff enough. Does anybody know if a CAT scan has been done on that instrument? That would give a reading of the top density, which would at least point to a likely number for the Young's modulus along the grain. Anything we can do to tighten up on the variables on these copies would be a help.

[Trevor Gore]

Wise and familiar words from Alan, as ever!

First thing, though, is to get some decent measurements. Measure monopole mobility (an objective measure of responsiveness with a good correlation to perceived loudness) and get a tap response (use VA rather than Audacity). Figure out which peaks correspond to which modes, then start down the path as suggested by Al, measuring as you go. Then you'll know which changes made the differences you wanted. You know where the detailed instructions are!

[Clay Schaeffer]

One of the first things I would do is try different (tension) strings. Perhaps you have already tried that. With the classical guitars I have had having the right strings made a big difference.
Depending on how recently they have been completed, they may just need more time to "play in".

[Alan Carruth]

Putting on heavier strings would have some of the same effects as reducing the bridge mass, since they have higher impedance. OTOH, it also tends to change some other things, so it's not as clear as you might like. Adding a little mass to the bridge might give you some indication of what would happen if you removed some: if it gets worse in a particular way when you add a few grams, maybe removing mass would improve it. If it doesn't make a lot of difference, perhaps some other avenue of approach would be better. It's always nice when you can start by making reversible changes.

[Simon Magennis]

Thank for all the comments so far.

I have finally ordered myself a microphone (shure pg 57 as Trevor mentioned at some point) and a variety of strings from low to high tension. I usually use boring old Daddario pro-arte ej45 but got a few more "flavours" with a total difference about 6kg between and highest tensions.

Next up some blu-tack to add weigh easily and reversibly. I started another thread asking about speakers for a chladni set-up. http://www.mimf.com/phpbb/viewtopic.php?f=10&t=2663 - I am completely ignorant in all things to do with sound systems so there will be a learning curve.
As of about half an hour ago I am officially on vacation for two weeks, so I hope to make a bit of progress. On US time in Europe. What my colleagues call "jet-lag in the comfort of your own home." And I guess half the world calls it shift work.

[Chuck Morrison]

On the string side of things, I'd suggest trying some different materials as well. The various carbon composites on the trebles for example since it's the trebles that concern you. I wouldn't remove any wood until I'd done that.

Also, I'd make sure I had a pretty complete set of metrics on the guitars before doing anything to them. By metrics I mean (at least) plots of the tap tones, chladni patterns to 1000Hz, and whatever other measurements/calculations you can do. I'd include deflection measurements as well.

I'd also do recordings of the sound of the notes you want to affect the sound of. This way you can see if those notes have changed due to the the changes you've made (including the different strings). This is a way of checking the inherent wish fulfillment needs we seem to color our endeavors with. Given that this is work on classical guitars, be sure to take multiple samples using different playing techniques. You should be able to see (perhaps subtle) differences in the spectral analysis of these samples. Make sure you identify the samples accurately so you can compare apples to apples. A rest stroke with the index (i) finger 10cm from the bridge should not be compared to a free stroke with the 3rd (a) finger 5cm from the bridge. I'm aware that there are more rigorous methods for doing this, but we do what we can with what we have.

[Simon Magennis]

Got my mic and pre-amp. Still experimenting with the settings and with how to tap best. Here is a spectrum for the first guitar I am thinking about. It is a Cypress and Euro-spruce classical on the Gal 1912 Ramirez plan. To recap the main issue is that some of the notes on the high e string are not all that responsive although improving with playing.


The peaks seem to be in the following positions:

1) F2 (88Hz)
2) B2 (125 Hz)
3) D3 (148 Hz)
4) G3 (191 Hz)
5) F4 (342 Hz)
6) G#4 (407Hz)
7) c5 (530Hz)

Tapping in different ways yielded slightly different results but this is pretty typical. I mainly tapped with my finger on the bridge wing holding the guitar on my knees. Strung with Daddario Pro Arte HT and tuned up.

All comments welcome.

[Chuck Morrison]

Simon,
I'm not seeing anything that I would expect to be able to identify as the back resonance. Could you turn the guitar around and do the same with a few taps on the center of the back ? For what it's worth, I've come to prefer using a non-log view of the X axis, not that it matters here. On some instruments with softer back wood it is possible for the back and top to be very close (depending on bracing) but I doubt that is the case here. 342 seems a bit high for a cypress back and 148 seems too low. Generally though, the air resonance is pretty low for this style of instrument, which makes me think the top and back are quite flexible. Another reason I doubt 342Hz or higher for the back.

Also, How about a few notes ? I like to see the layout of the harmonics on at least the Low E (82.4Hz). This helps me equate what's going on with the taps and what I hear from the actual notes. More importantly, to me at least, I can equate what I see with what I hear. A view of a few of the notes you think are too weak could be helpful as a reference as you go along.

[Simon Magennis]

Hi Chuck. Thank you. Here is a low E and the back.

The back peaks are are:
F2 (90Hz)
F#3 (184Hz)
C#4 (283Hz)
D#4 (304Hz)
E4 (335Hz)
G#4 (415Hz)
A#4 (462Hz)
C5 (527Hz)

I will put in shots of the "offending" notes in another post below.
Simon

[Simon Magennis]

The notes that stick out as being a bit "dead" are the B-flat and C on the high e string. Here are the spectrums.

B-flat:

and

C:

Thanks.

[Chuck Morrison]

Hi Simon,
Lots of info there. I see an oops from my original reply. When I do a non-log plot I limit the range. Audacity doesn't let you do that so I save the plot to a file (x,y plot file, not a graphic) and then look at it in a graphing program that lets me limit the range (usually 0-1000Hz for low notes, to 5k for higher notes). My apologies for not mentioning that. Those non-log plots aren't useless, just harder to read than would be nice. It is easier to see the harmonics (and holes) when they are equidistant.

When looking at note plots, it's important to note that what you're looking at is a plot of the harmonics of the note. Each harmonic will be the fundamental times an integer (1,2,3,4...). The first six or so being important. I prefer having a relatively flat response over these first 6 harmonics. Any deep holes or very dominant peaks are problematic (others are free to disagree). The 7th harmonic is the flat 7th (+2 octaves), the 9th harmonic is the 2nd (+3 octaves). These are dissonant sounds and being too strong can cause some nasty sounding notes. Above the 9th harmonic you end up with some truly nasty dissonances. In order these harmonics and their relationship to the major scale are:
1 - The fundamental
2 - The Octave above the fundamental
3 - The fifth (+ an octave)
4 - The second octave
5 - The major 3rd (+ 2 octaves)
6 - The fifth (+ 2 octaves)
7 - The flat 7 (+2 octaves)
8 - The third octave
9 - The 2nd or 11th (+ 3 octaves)

It's interesting to see that from the first to 6th harmonics you are actually hearing a major chord in that one note. OK, first thing I notice from the Low E plot is that the fundamental is lower than the 2nd and 3rd harmonics. This is pretty standard with traditional guitars. Often this changes once the note is higher than the main air resonance (in this case 88-90Hz) and the fundamental regains dominance or at least equality with the next few harmonics. I also note a "hole" in the Low E plot is the second octave, 329Hz (E4) and a strong falling off above 500 Hz or so. The falling off is fairly normal, but it's always interesting to see (and compare) how fast it does so and if there are any strong peaks past this point.

About the back tap frequency, The emergence of the 283-304 Hz peak is pretty conclusive but still seems high to me. I guess I've been concentrating on building lower pitched backs too long. Anyway, I'd look at the modes (chladni tests) for the back and see what it's doing (top too).

The plots of the B-flat and C (466 & 523Hz) show something that I've seen quite often, which is that the air (just above 90Hz) and top (just under 200Hz) main resonances accompany the notes and show up as undertones of the (composite) note we hear. These are not always the exact same frequency as the tap tone plots shows us, but they are always pretty close. I call these "ghosts". In addition to the ghosts, and probably more importantly, we can also see that the fundamental notes are not clearly defined peaks, but are surrounded by some noise that may have some negative influence. If I was doing this, my next step would be to look at chladni patterns to see what the top was doing in that range. Also the back.

Let me throw out another idea which has bitten me in the past, especially near that location of the neck. The neck also has harmonics and it's possible that there is an active (anti)node in that area. If B is strong and b-flat and C problems, then maybe a node right on B, 7th fret ?. I've been able to get chladni patterns to show up on necks before and had necks that couldn't be tamed.

So my next steps would be to do chladni testing to map the modes of vibration to the peaks you have on the tap plot(s) and see if there are very active/inactive areas that the plots aren't showing you. This should help get a more accurate picture of what's going on and what you might want to address by adding or removing wood/weight/stiffness/... .

Sorry to be so wordy.

[Trevor Gore]

Simon, a few ideas to help you along:

1) Use Visual Analyser if you can. If you run a Mac, VA will run under Parallels (a Windows emulation program) which will let you run all sorts of other Windows software, too. Cost is ~$80. VA is free. The reason for running VA is that it will let you see much more easily the sorts of things you're looking for, because you have more control over how you record things and how you display them.

2) I checked out your speaker thread. There is no need to pay more than ~$20 for a pair of speakers. Go to an automotive store and buy a pair of car speakers. By definition they will be full range (full enough for what we need, anyway). 4" (100mm) diameter is a good size, with at least a 15 Watt continuous power rating. It helps to put it in a baffle, like the bit of PVC pipe I use. Pics in the book.

3) VA has a signal generator included, but a good one for a Mac (apparently - I run Windows) can be downloaded from here. There's lots of other good stuff on musical acoustics and Chladni patterns on Joe's pages if you have a look around.

4) Think about the amp you want to use. A guitar amp that has a socket for an external speaker can be used (provided it switches off the amp's internal speaker), otherwise just about any stereo amp from an op shop will do. The vast majority are for use with 8 ohm speakers

5) If you're tapping guitars, best to use a tapping hammer - a pencil eraser on the end of a bamboo BBQ skewer works fine.

6) There's quite a bit of stuff on the ANZLF forum. Here's a thread that is particularly pertinent

Have fun!

[Alan Carruth]

The mix of partials you get from the string will depend strongly on where you pluck it. If you pluck at the halfway point, to use an extreme example, you'll lose all of the even order partials. Basically, you're forcing the string to move at a point where all of those partials have a stationary node. Plucking at 1/3 the length of the string cuts out the 3d, 6th, and 9th partials, and on up in the same way. Try to find a spot that's not some small whole number division of the string length (good luck), and mark it with a felt tip pen on all the strings so you can hit it every time to get consistent results.

The 'family' of peaks around 500 Hz suggests that there might be a top mode there that's interacting with the string. You've got a peak in one of your tap plots at c5=520 Hz, and that's a normal range for the 'cross tripole', which is a good radiating mode and one where the bridge is moving. What can happen in a case like that is that the vertical motion of the bridge means that the end is not 'fixed' when the string is vibrating in a vertical polarity, but it is fixed for horizontal motion. Essentially the string 'sees' two different terminations in slightly different places depending on how it's vibrating: it has two different lengths and makes two different pitches. Shifting the mode pitch or adding mass to the bridge to reduce it's motion will help, but change the sound of the guitar, of course.

Or maybe it's something else....

[Simon Magennis]

Thank you all for the great responses. Lots of food for thought. Chladni patterns are next on the menu once I get the set-up together.

[Michael Lewis]

It can't hurt to take measurements for your records so you have a starting point for future comparison, but before altering anything I suggest playing the guitar a good while to let the thing develop it's voice. Flex it, vibrate it, make it produce sound, subject it to sound. The new structure is a bit stiff and needs to learn how to be a guitar. Try a Tonerite or similar device, or make up something to do that job, but get the structure moving, and for a good while.

[Michael Lewis]

The best next step is to make another one.

[Chuck Morrison]

Or you could have some loose frets... or a heavier/lighter set of tuners could eliminate the problem. You never know until you start to understand the instrument and it's foibles which is where the fun starts.

[Douglas Ingram]

I'm not overlooking all of the useful technical input here, but I'd like to add to the comments regarding string choice. EJ-45 strings are nice all-round strings. However, on these older designs I have found that suing something like Aquila Nylgut makes a rather dramatic difference in how the guitar responds. My Torres style guitars really perk up when I go from EJ-45 strings to the nylgut. The trebles suddenly "wake up" and have a shimmer that they previosly did not have.

[Waddy Thomson]

EJ-45's are very reliable strings. They are, however, also very fundamental. They tend to limit the overtones, in my opinion.