You are on page 1of 51

Musical Instruments BSc

Module: CP6017

The Cass Dissertation

Jonathan Smith

03000425

An Investigation into the Theory and Practice of Guitar Intonation

April 2016
Contents

Introduction ................................................................................................................................... 1
Chapter 1 ....................................................................................................................................... 2
12 Tone Equal Temperament ...................................................................................................... 2
The Rule of 18 ............................................................................................................................. 3
Compensation .............................................................................................................................. 5
Tuning Error ................................................................................................................................ 6
Chapter 2 ....................................................................................................................................... 8
Calculation vs. Measurement ...................................................................................................... 8
Modelling the System.................................................................................................................. 9
Mass Per Unit Length ............................................................................................................ 10
Frequency .............................................................................................................................. 12
Tension .................................................................................................................................. 13
String Length ......................................................................................................................... 13
Elasticity ................................................................................................................................ 16
Fret Alignment....................................................................................................................... 23
Inharmonicity ........................................................................................................................ 27
Bypassing the Model: Empirical Techniques............................................................................ 31
Chapter 3 ..................................................................................................................................... 35
String Properties ........................................................................................................................ 35
Setup Specification.................................................................................................................... 37
Playing Technique ..................................................................................................................... 38
Building the Spreadsheet........................................................................................................... 39
Evaluation.................................................................................................................................. 43
Conclusions .................................................................................................................................. 45
Bibliography ................................................................................................................................ 46
Appendix: Using the Spreadsheet.............................................................................................. 49
1

Introduction

The pursuit of accurate intonation presents a unique problem to makers and players of
guitars and similar fretted instruments. At first glance, the matter of placing the nut, saddle and
frets in the correct position relative to one another appears simple, but in practice it is beset with
complex confounding factors. So apparently intractable is the problem that the vast majority of
luthiers either overlook it entirely, or accept it as an inherent imperfection of guitars as a whole,
resulting in almost all guitars being impossible to play accurately in-tune, in all positions. The
aim of this paper is to unravel the factors at play, to examine modern and historical approaches to
the problem, and to explore new ways in which the theory can be applied by the working luthier
to improve the intonation of his instruments.

Chapter one introduces 12 Tone Equal Temperament, and discusses historical means by
which this tuning system has been applied to fretted instruments, most notably by the 'rule of 18'
fretting scheme and its modern equivalent, the ‘rule of 17.817’. It explains the necessity for
string length compensation, the normal method by which this is achieved, and the shortcomings
of this system.

Chapter two examines the system in greater detail, explaining the physical principles at
work. With reference to modern methods, most notably those of Byers, Gore and Magliari, it
explores ways in which these principles can be applied to mathematical models to improve
intonation from the design stage. Attention is also paid to more empirical approaches to the same
problem, such as stretch tuning and the ‘True Temperament’ system.

The third and final chapter details the development of a spreadsheet-based tool, which
can be used to design a new instrument with improved intonation. The tool uses the theory set
out in chapter 2 to model the strings, fingerboard and playing technique, resulting in a set of
compensation values which can be used in the design of a new instrument.
2

Chapter 1

12 Tone Equal Temperament

All tuning systems derive their intervals from the harmonic series. ‘Pure’ intervals are
those between frequencies related by small integer ratios, hence directly matching those of the
harmonic series. 2:1 for example, is the ratio between the fundamental and the first harmonic and
represents a pure octave, while 3:2 is a pure fifth; 4:3 a pure fourth, and so on. Tuning systems
constructed from pure intervals are known collectively as just intonation, and produce a sweet,
consonant sound, as notes played harmonically share the frequencies of their partials.
Unfortunately, to maintain these pure intervals just intonation requires continuously variable
pitch control in relation to the key being played, a condition suited to choirs or string ensembles,
but not to keyboards or guitars.

12 Tone Equal Temperament (12TET) is the tuning system used ubiquitously in modern
western music, wherein the pure (2:1) octave is divided into twelve equal semitones, producing
tempered intervals which approximate but do not precisely match their pure counterparts. Major
thirds and minor sixths in particular are considerably wider than pure. 1 12TET is in essence a
system of blunt convenience, sacrificing the pure intervals of just intonation for the more
practical utility of playing in any and all keys without the need to re-tune or otherwise modulate
pitch. This arrangement is especially useful to an instrument with fixed frets such as a guitar,
where the intonation cannot be readily adjusted by playing technique, or even by re-tuning the
strings. Its intonation is an inherent aspect of its form, which is determined by design and cannot
be significantly affected without some mechanical intervention.

Unlike modern guitars, lutes and viols typically featured gut frets tied around the neck,
which could be adjusted by the player. This facility may have been intended to permit the use of
unequally tempered scales, wherein adjustments to intonation would be required to play in
different keys. Lindley exhaustively compiles the various versions of Pythagorean and meantone
temperaments in use for these instruments between the early 16 th and mid-18th centuries.2

1
Ross W. Duffin, How Equal Temperament Ruined Harmony (And Why You Should Care) (New York, W. W. Norton
& Company, Inc., 2007), pp.32-33: Chapter 2: How Temperament Started
2
Mark Lindley, Lutes, Viols & Temperaments (Cambridge, Cambridge University Press, 1984)
3

Despite this variety of tuning practice, he asserts that by around 1550 equal temperament was
well established, as evidenced by the contemporary emergence of equally tempered fretting
schemes, most notably the ‘rule of 18’.

The Rule of 18

Lindley and Meffen3 agree that the ‘rule of 18’ was first suggested by Vincenzo Galilei in
1582, in his Dialogo della Musica.4 In this scheme, the total scale length is divided by 18 to find
the position of the first fret. The remaining length is divided by 18 again to find the second, and
so on. In general terms:

(1)

Where Xn is the distance from fret n to the bridge.

When the rule of 18 is used to position the frets on a fingerboard, one might assume that
the 12th fret ends up at a point exactly equidistant between the nut and the saddle, sounding the
pure, 2:1 octave. In fact it falls slightly closer to the nut, an apparent flaw in the rule which was
not lost on its proponents.

Galilei himself notes the “...slight discrepancy that there is between the measure [rule of
18] and the measured [theoretically correct divisor]”.5 Mersenne also implicitly recommends the
rule of 18 for tuning arch lutes to equal temperament, a result which “...is found by dividing the
open strings into 100,000 parts... 50,363 represents half of the string, yet this number is too large
by 363, since 50,000 is half of 100,000”6. A spreadsheet set up to perform the rule of 18 and with

3
John Meffen, A Guide to Tuning Musical Instruments (Newton Abbot: David & Charles, 1982) p.127
4
Vincenzo Galilei, Dialogo della Musica Antica et della Moderna, (Venice, 1584)
5
Vincenzo Galilei, Dialogo della Musica Antica et della Moderna, (Venice, 1584), quote translated by Lindley,
1984.
6
Marin Mersenne, Harmonie Universelle, (Paris, 1637), iii, p.48
4

a total string length of 100,000 reproduces this figure, giving a 12 th fret value of 50,363.627:

Fret
Xn-1 - Xn Xn
n

10 3321.354573 56463.02774
11 3136.834874 53326.19286
12 2962.56627 50363.62659
13 2797.979255 47565.64734
14 2642.535963 44923.11137

Figure 1: ‘Rule of 18’ Spreadsheet

The exact formula for the equally tempered semitone was first proposed by William
Brouncker, in his 1653 translation of Descartes' Musicae Compendium.7 Rather than attempting
to define it with integer ratios, he instead used an irrational number, i.e. one which cannot be
defined precisely as either a fraction or a decimal, namely the 12 th root of 2. This number can be
expressed as √2 or 21/12, and is approximately equal to 1.05946... To apply this figure to a
fretting scheme, the total scale length is divided by 21/12, to give the distance from the 1st fret to
the bridge. This new distance is divided by 21/12 to give the distance from the 2nd fret to the
bridge, and so on. In general terms:

(2)

and by extension:

7
William Brounker, Animadversions upon the Musick Compendium of Renat. Descartes (London, 1653), p.91
5

(3)

Where X0 is the distance from the bridge to fret 0 i.e. the nut, which is to say the total
scale length.

To apply this rule in the same way as the rule of 18, it is necessary to find the divisor d,
which gives the distance between frets. Adopting the form of the ‘rule of 18’ equation:

(4)

Cancelling Xn and solving for d gives another irrational number, 17.81715… Just shy of
the earlier approximation to 18. This is the figure which is used by the majority of modern
luthiers, to the extent that today the term 'rule of 18' is often used as shorthand for the 'rule of
17.817'. To avoid confusion, the terms are used literally throughout this paper.

Compensation

The rule of 17.817 is a precise mathematical formula for modifying string length to
produce equally tempered intervals. Where it falls short is in taking no account of the sharpening
effect of fretting strings. When a string is pressed down to a fret, it stretches, increasing in
tension and therefore in pitch. Therefore if a guitar is built with accurate 'rule of 17.817'
positioning of the nut, saddle and frets, every fretted note will play sharp. This may explain why
Galilei and Mersenne recommended the rule of 18, despite clearly understanding its
imperfection; since the fret spacing is slightly narrower than that of the rule of 17.817, its notes
sound slightly flatter, compensating for the effect of fretting action. In the absence of any such
built-in compensation, modern instruments require further intervention.
6

The typical method of compensation is to set the saddle back further from the
fingerboard, increasing the string length. The final intonation is usually achieved by tuning each
string to match the 12th fretted note with the 1st natural harmonic. As different strings require
different amounts of compensation, the saddle must also be shaped and/or slanted in some way to
produce a staggered profile; this effect is most obvious on electric guitars with individually
adjustable saddles.

This standard method of setup represents an uncomfortable marriage of a mathematical


model (rule of 17.817) to position the frets, with an empirical method (tuning the octave by
comparison) to approximate the correct intonation. The result of this compromised system is that
all frets below the 12th play slightly sharp, while all frets above the 12th play slightly flat. The
discrepancy between the notes sounded and the desired notes of the equally tempered scale is
commensurate with the degree of saddle compensation applied, and with the distance (in
semitones) from the correctly tuned 12th fret octave. This means that the greatest pitch
inaccuracies are to be found on the lower frets of most guitars, which typically see the most use,
when playing chords for example.

Tuning Error

Before moving on to some of the ways modern luthiers have attempted to improve upon
traditional guitar intonation, one might reasonably ask the question: how much does this matter?
In other words, how significant is the tuning error caused by saddle compensation, and by
extension, how significant an improvement can be made to an instrument by its correction? The
answers to these questions are subtle, and depend largely on the pitch perception of the
individual listener.

Objectively, the tuning error can be demonstrated very easily by anyone with an accurate
chromatic tuner. In the introduction to his methods for intonation, Trevor Gore plots a graph of
fret number against pitch error from 12TET, showing a range of around 6-8 cents for the lower
frets of a steel string guitar.8 Gary Magliari has done similar work, also on steel string guitars,

8
Trevor Gore & Gerard Gilet, Contemporary Acoustic Guitar: Vol.1: Design, p.4.98
7

and presents an almost identical graph in his American Lutherie article with Don Macrostie.9
Gore compares this 'tuning error' with the 'tuning gap' which exists between 12TET and just
intonation, noting that ‘The 6-8 cents itself is not too much of an issue, but when it is stuck on
top of the thirds of the equally tempered scale which are already 14 cents sharp compared to the
Just intonated scale, yielding a third that is now 20-odd cents sharp, it is way too far from the
comfort and consonance of just intonation’, contrasting this with a description of a well intonated
guitar thus: ‘All open chords can be played adequately in tune, complex chords sound true and
sparkly, and because more harmonics are more accurately aligned the guitar sounds truer with
longer sustain.’10

Appealing though these descriptions may be, objective bases are rather harder to come
by. Boullosa et al's 1999 paper recorded, among other factors, the pitch accuracy of a range of
nylon-strung guitars, positively correlating good tuning characteristics with the 'subjective
quality' of an instrument, as judged by a listening panel. 11 This study is instructive as it
demonstrates that, even to those listeners who do not perceive an instrument to be 'out of tune',
an improvement to its intonation will tend to have a positive effect on its perceived sound.

The following chapter examines the problem in greater detail, and explores the various
ways in which contemporary luthiers have attempted to achieve this kind of improvement.

9
Gary Magliari and Don Macrostie, 'Beyond the Rule of 18: Intonation for the 21 st Century' American Lutherie #116
(Guild of American Luthiers, 2013), p.8
10
Gore, pp.3.21-3.23
11
Ricardo R. Boullosa, Felipe Orduna-Bustamante, A. Perz-Lopez, 'Tuning Characteristics, Radiation Efficiency
and Subjective Quality of a Set of Classical Guitars' Applied Acoustics #56, issue 3 (Elsevier, 1999), pp.183-197
8

Chapter 2

Calculation vs. Measurement

In the previous chapter, saddle compensation was described as a compromise between


empirical and mathematical methods. This example will be discussed in more detail here, as a
means to evaluate the strengths and weaknesses of these two apparently opposite approaches,
and to consider how they might complement one another in developing a more accurate and
reliable method for setting intonation.

The mathematical model employed is the 'rule of 17.817'. The great advantage to this rule
is that it can be applied to any guitar, with any scale length, with approximately equal success.
The formula requires only a single specific measurement, the nominal scale length of the
instrument, to calculate the position of every fret with a reasonable degree of precision. This
makes it a powerful tool, which is simple to understand and apply. This very simplicity is also its
weakness however, as it is a very incomplete model, which does not take account of numerous
measurable components of the system which affect pitch.

This incompleteness necessitates the empirical aspect of the method, saddle


compensation. The open string is tuned to the desired pitch, and the 1 st natural harmonic
compared to the note sounded at the 12th fret. If the fretted note sounds sharp, as we have reason
to suppose it will, the saddle is moved back by a small amount; if it sounds flat, the saddle is
moved forwards. This process of trial and error is repeated until the notes are in tune with one
another. This allows the intonation to be fine-tuned for each string, but is limited to a single
crude adjustment to the string length at one end. As we have seen, this results in more-or-less out
of tune notes at every fret other than the 12th.

Two potential avenues for improvement immediately spring to mind. The first is to
increase the adjustment facility of the guitar itself, to allow not only the saddle but also the nut
and perhaps even the frets to be repositioned in relation to the string, and then fixed in place
when a satisfactory result is achieved. This presents a technological challenge, which may be
costly and difficult to overcome, but one which should result in improved intonation without
9

requiring any in-depth theoretical knowledge. The second is to employ a more accurate model of
the system, such that the instrument can be designed to intonate well without making
adjustments. This approach requires a number of variables to be recorded and/or calculated, but
avoids the need to significantly re-engineer the instrument. Any attempt to rely solely on either
approach will tend to encounter prohibitive limitations. All of the methods described therefore
necessarily employ a combination of empirical and mathematical techniques to achieve optimal
results.

Modelling the System

The following section focuses on developing an accurate mathematical model of the


system, which can be applied at the design stage. This requires a detailed analysis of the forces at
work when the instrument is played, and a means of applying the physical principles to a usable
plan for an instrument.

The natural starting point for any model of a guitar is the strings. Specifically, to model
intonation it is necessary to determine the factors which contribute to the string's fundamental
frequency. Mersenne's Law12 states that:

(5)

Where:

f = fundamental frequency (Hz)

L = vibrating string length (M)

T = tension (N)

μ = mass per unit length (KgM-1)

12
Marin Mersenne, Marin Mersenne: An Edited Translation of the Fourth Treatise of the Harmonie Universelle, by
Robert Fortson Williams (Ann Arbor, MI: UMI Dissertation Services, 1995).
10

By simple substitution, this formula can be arranged to calculate any one of the four
variables, if the other three are known:

(6)

(7)

(8)

As these equations are so central to the behaviour of strings, the following section will
briefly explain each parameter, its units and the means by which it can be measured or derived.

Mass Per Unit Length

Mass per unit length, or µ, is a basic physical property of the string which cannot
reasonably be altered by the instrument designer or player, other than by selecting a different
gauge or composition of string. It is also, manufacturing defects notwithstanding, constant along
the length of the string, unlike string length and frequency which are intentionally modified by
fretting.

µ is measured in KgM-1, or Kilograms per linear Metre. µ can be determined in a number


of ways. It is sometimes expressed as ρA, the string's material density ρ multiplied by its cross-
11

sectional area A. The cross-sectional area of plain strings is relatively easy to calculate, on the
assumption that they are uniformly cylindrical. The diameter of the string is commonly given by
the manufacturer in inches, and referred to as the string gauge. This is converted to the area of a
circle by:

(9)

The material density (KgM-3) can be estimated with reasonable reliability from published
data, and multiplied by A to arrive at a value for µ. This method requires no special equipment
and is a quick way to estimate a µ value for hypothetical strings. It does rely on published data
for material density which may not be correct for the string in question, and is less reliable when
dealing with wound strings, which do not have a simple geometric relationship between their
diameter and cross-sectional area.

An even simpler method would be to cut sections of the string in question to known
lengths, and weigh them to obtain a direct reading. The weight in Kg, divided by the length in M
gives a value for µ. Provided that the length and weight measurements are sufficiently precise,
this method is entirely feasible, and works equally well for both plain and wound strings. The
only downside is that it is destructive, and therefore expensive and impractical for studying large
numbers of strings.

Mass-per-unit-length can also be determined experimentally. A test rig can be made


which allows the tension, string length and frequency to be accurately measured, and the values
are substituted into the Mersenne equation (8).
12

Frequency

The fundamental frequency required of a note is a defined number, measured in Hertz,


which is determined by temperament and pitch standard. For the purposes of this paper, 12TET
and the modern pitch standard A=440 are used without exception. The target frequencies for
each string at each fret position, in standard tuning and with 24 frets, are as follows:

String
E A D G B e
0 82.407 110.000 146.832 195.998 261.626 349.228
1 87.307 116.541 155.563 207.652 277.183 369.994
2 92.499 123.471 164.814 220.000 293.665 391.995
3 97.999 130.813 174.614 233.082 311.127 415.305
4 103.826 138.591 184.997 246.942 329.628 440.000
5 110.000 146.832 195.998 261.626 349.228 466.164
6 116.541 155.563 207.652 277.183 369.994 493.883
7 123.471 164.814 220.000 293.665 391.995 523.251
8 130.813 174.614 233.082 311.127 415.305 554.365
9 138.591 184.997 246.942 329.628 440.000 587.330
10 146.832 195.998 261.626 349.228 466.164 622.254
11 155.563 207.652 277.183 369.994 493.883 659.255
12 164.814 220.000 293.665 391.995 523.251 698.456
Fret

13 174.614 233.082 311.127 415.305 554.365 739.989


14 184.997 246.942 329.628 440.000 587.330 783.991
15 195.998 261.626 349.228 466.164 622.254 830.609
16 207.652 277.183 369.994 493.883 659.255 880.000
17 220.000 293.665 391.995 523.251 698.456 932.328
18 233.082 311.127 415.305 554.365 739.989 987.767
19 246.942 329.628 440.000 587.330 783.991 1046.502
20 261.626 349.228 466.164 622.254 830.609 1108.731
21 277.183 369.994 493.883 659.255 880.000 1174.659
22 293.665 391.995 523.251 698.456 932.328 1244.508
23 311.127 415.305 554.365 739.989 987.767 1318.510
24 329.628 440.000 587.330 783.991 1046.502 1396.913

Figure 2: Table of Target Frequencies in 12TET

Frequency can be accurately measured with electronic devices or computer software,


which count the number of cycles per second, expressed as Hertz (Hz).
13

Tension

As tension is a force, it is measured in Newtons (N). Often however it is expressed in


units of mass such as Kilograms (Kg). To convert from Kg to N, the value is multiplied by the
gravitational constant g, approximately 9.81 Ms-2. This figure is equivalent to the acceleration
due to gravity on Earth, meaning that a mass of 1Kg exerts a force of ~9.81N. This is useful in
an experimental situation, where tension can be accurately applied to a string by hanging
weights. For simplicity, tension is expressed in Newtons throughout this paper.

Tension is usually applied to the string by means of a mechanical tuning peg, which is
continuously variable and therefore allows any given string at any given fret position (i.e. known
values for µ and L) to be precisely tuned to the target value for f. In practice, it is usually the
open string which is tuned to its target frequency. Like µ, T is not intentionally modified by the
player during normal performance; once the string is tensioned it is expected to remain in tune
without adjustment to the tuning pegs.

String Length

String length is the inherent aspect of an instrument which can be altered, by design or
adjustment, to adjust its intonation. Specifically, the relative position of the nut, saddle and frets
should regulate the string length in such a way that the frequency at each position conforms to
that shown in the target frequency table in figure 2.

The target frequency at fret n on a given string is given by:

(10)
14

By comparing equations (5) and (6), it can be said that frequency is inversely
proportional to length, or:

(11)

Substituting into the target frequency equation (10) yields:

(12)

Which is the familiar standard fret placement formula, referred to previously as the 'rule
of 17.817'.

This calculation holds true on the basis that the values of both T and µ are constants,
which are impervious to changes in length intended to alter frequency. In practice however,
tension does not remain constant. The action of pressing the string down to the fret necessarily
increases its length slightly, which due to the string's longitudinal stiffness causes it to resist the
stretch, increasing its tension. This in turn causes the frequency to increase, deviating from its
target value.

To compensate for the deviation in frequency, it is first necessary to quantify the increase
in length caused by fretting action. This is most easily visualised with a geometric model of the
neck.
15

W. and P.A. Bartolini produced such a model in their 1982 paper on classical guitar
acoustics13, reproduced below.

Figure 3: Bartolini Geometric Model

The string is divided into four sections: D1, D2, D3 and D4, simulating the increased
length of the fretted string as compared to the unfretted string length L2. The model
approximates the fretting action to a single point of contact at the midpoint between two frets,
approximating the more complex shape of a fingertip, to keep the calculations to a sensible
minimum.

Greg Byers' model builds on the Bartolinis', adding the variables 'm' and 'j', representing
the distance from the fret where the string is depressed, and the height above the fingerboard
where the string is suspended.14 The intention is to accommodate differences in playing styles,
by varying these parameters according to the player. The fretted string length is otherwise found

13
W. and P.A. Bartolini, 'Experimental Studies of the Acoustics of Classic and Flamenco Guitars' Journal of Guitar
Acoustics #6 (1982), pp.75-103: 'Fret Position Calculations for Nylon Strings Including String Stretch'.
14
Greg Byers, 'Classic Guitar Intonation' American Lutherie #47 (Guild of American Luthiers, 1996) pp.34-45
16

in an identical manner, by summing four sections of the string, relabelled here as l n1, ln2, ln3 and
ln4.

Figure 4: Byers Geometric Model

Trevor Gore takes the geometric model to an even greater degree of sophistication,
modelling the neck surface as a summation of a straight line, an arc of an ellipse and an arc of a
circle, allowing for the physical deformation of the neck, or 'relief', caused by string tension. 15

While the details of the models differ in complexity, the intention is identical: to
accurately predict the change in string length which occurs during fretting. Once the model has
been used to establish the relative spatial positions of the various contact points, calculation of
the fretted string length is straightforward; diagonal straight-line segments are evaluated using
Pythagoras' theorem, and summed. With a working value for the fretted string length at each fret,
it is necessary to equate this to the attendant increase in tension.

Elasticity

Elasticity is the measure of a material's resistance to deformation. It is defined as the ratio


of stress/strain experienced by an object when a tensile force is applied. Stress is defined as the

15
Trevor Gore & Gerard Gilet, Contemporary Acoustic Guitar: Vol.1: Design, pp.4-110 - 4-115
17

force applied to an object per unit of cross sectional area, measured in megapascals (MPa),
equivalent to 106 newtons per square metre (106 NM-2).16

(13)

where:

σ = Stress (MPa)

F= Force (N)

A = Cross-sectional area (M2)

Strain is defined as elongation experienced by the object divided by its original length
with no load applied; since both are measures of length, strain itself is unitless: 17

(14)

where:

ε = Strain

ΔL = Change in length (M)

L0 = No-load length (M)

16
William D. Callister, Jr., Materials Science and Engineering: An Introduction (New York, John Wiley & Sons,
Inc., 2006), pp.134-135
17
Callister, 2006, pp.134-135
18

Hooke’s law states that stress and strain are proportional to one another in elastic
deformation, and the constant of proportionality is known as the Elastic Modulus E. As strain is
unitless, elastic modulus is expressed in units of stress, namely gigapascals (GPa), equivalent to
109 NM-2.18

(15)

where:

σ = Stress (MPa)

ε = Strain

E = Elastic Modulus (GPa)

In laboratory conditions, elastic modulus can be determined using tensile test equipment,
which returns highly accurate data for elongation and load, however this kind of equipment is not
practical for the average luthier wishing to test strings. Published material data is available,
provided the string material is correctly identified.

Magliari19 describes an experimental method for obtaining E, requiring neither published


data nor expensive apparatus. For the purposes of this test, it is assumed that the core of the
string is the sole contributor to its modulus of elasticity; any windings are purely intended to add
mass to the string, increasing the value of µ.

18
Callister, 2006, p.137
19
Gary Magliari, 'Method for Positioning Musical Instrument Frets that Compensate for Fretting-Induced String
Tension' <http://www.luth.org/images/web_extras/al116/MagliariFretComp.pdf> Accessed 1st December 2015,
pp.16-18, 'String Evaluation’
19

A test rig is described with a 'nut' and 'saddle', one of which is moveable to increase the
string length by a measurable amount. Tension is applied by hanging weights at one end, and the
frequency is measured with an electronic tuner. Thus L, T and f are known, and μ can be found
using the relevant Mersenne equation (8). The string is clamped down at the end at which tension
has been applied, and the length is increased by a small amount, stretching the string and
increasing the tension:

(16)

As these are the 'stretched' parameters, each symbol receives the subscript 's'.

The next variable calculated is the length of the 'original' stretch, from zero tension up to
the original tension T. This length is designated ΔLor and calculated thus:

(17)

This stands to reason, as ΔTs/ΔLs is simply a ratio of measured changes in tension and
length.

With a value for the stretch, the original length Lor is easily calculated:

(18)
20

Finally, the value EA, the product of the modulus of elasticity E and the cross sectional
area of the core A is calculated:

(19)

This calculation is in essence the reverse of that which will be applied to the geometric
model. The test rig uses a known change in tension to calculate elastic modulus, while the model
requires a known value of E to calculate tension.

(20)

where:

ΔTn = Change in tension caused by fretting the string at fret n (N)

ΔLn = Change in length caused by fretting the string at fret n (M)

L0 = The original (unfretted) length of the string (M)

The desired frequencies are known for each position. This gives us all of the necessary
parameters to calculate compensated fret positions for each string:
21

(21)

Clearly, this is simply the second Mersenne equation (6), with the T value adjusted for
the effect of fretting.

This step-by-step method of determining the modulus of elasticity in order to equate an


increase in length to one in tension is shared closely by Magliari and Gore. The earlier Bartolini
model, adopted by Byers, neatly sidesteps part of this reasoning, by introducing a dimensionless
quantity R, which directly correlates the increase in length to that in frequency.

Bartolini defines R as the “fractional pitch change to length change ratio”, 20 and it is
determined experimentally with apparatus similar to that used by Magliari and Gore to determine
µ and E.

Two strings are tuned in unison, one of which is then stretched by a small amount, by
turning a wing nut on a lead screw. The change in length can be calculated by the pitch of the
screw, and the change in frequency by counting the 'beats' occurring between the two strings. R
can then be calculated:

(22)

Having determined a value for R, the geometric model is used to determine the fretted
and unfretted length of the string at each fret, and defines the fractional difference as T:

20
Bartolini, 1982, p.98
22

(23)

This substituted into the equation for R gives a value to the fractional pitch change which
occurs:

(24)

Rather than using specific target frequencies, the '12th root of 2' fret spacing is taken as a
starting point, but with the exponent n denoting the fret number modified to compensate for the
fractional pitch change:

(25)

which gives compensated fret positions for a given string's R value.

Byers borrows the concept of R from Bartolini, but takes a different approach to deriving
compensated fret positions. He helpfully renames the value T, which does not denote tension, Qn:
23

(26)

Instead of modifying the exponent n, Byers calculates his fret positions using the
unmodified, 12th root of 2 derived fret positions Xn, the experimentally determined R value, and
the Qn value from the geometric model:

(27)

Fret Alignment

At this point in the modelling process, each of the methods described has arrived at a
common juncture; each string has been reduced to a set of Xn values corresponding to a unique
fret layout. Due to differences in the mass-per-unit-length, cross-sectional area and longitudinal
stiffness, the calculated fret spacing is different for each string. If the frets used are to be straight
and parallel, a means must be found to bring these sets into alignment, simplifying the model yet
further into a single one-dimensional layout for the fret board.

Magliari's method is characteristically visual in nature. The six data sets are drawn
individually using CAD software, then aligned at the third fret, which is designated as the 'zero-
out' fret. The distance is taken from the aligned third fret position to each set's 14 th fret, or 'even-
out' fret, and this distance is averaged. The scale length of each set is then adjusted so that the
14th frets all coincide at the averaged distance, while maintaining their proportional spacing:
24

Figure 5: Magliari Fret Alignment System21

The remaining frets will not align perfectly, so the fret positions are established by taking
average distances from the third fret. It is immediately apparent that the major deviations
between the adjusted sets occur at the nut and the saddle, rather than at any of the frets. This
occurs because the sharpening effect occurs in a more-or-less progressive manner across the
frets, but is entirely absent at the nut. Unlike the frets, which remain straight and parallel, the nut
can be shaped to satisfy this requirement, exactly as the saddles are in the normal compensation
system.

Nut compensation is common to all systems of intonation which seek to maintain straight
and parallel frets. The resultant appearance of the modified nut, as produced by Gore, is shown
below:

21
Magliari, 2006, p.4
25

Figure 6: Gore Compensated Nut22

Byers' alignment approach starts by expressing each string set in what he calls his
'canonical' equation for fret placement:

(28)

Where X'n is the compensated position of fret n, a is the string's nominal scale length, and
b is the amount of saddle compensation added per fret.

The aim is to match the calculated fret positions from 2.25 to a set which conforms to
'rule of 17.817' fret spacing, with a modified scale length and including saddle compensation.
This is done with statistical software.

a then is equivalent to X''0, or, 'scale length adjusted to rule of 17.817', and b to ΔS, or
'change in saddle position from its nominal placement':

22
Gore, 2011, p.4-118
26

(29)

The ≈ ('almost equal to') symbol indicates that the calculated values X'n are fitted to the
canonical form by statistical means. With known values for the adjusted scale length and saddle
compensation, the nut position can also be calculated:

(30)

X0 is the original scale length, from which the compensated positions for each string X'n
were calculated. The fret positions were all effectively moved back towards the nut to
compensate for the fretting effect, so the ΔN value is negative, and the nut is therefore moved
forwards.

Because X''n is in 'canonical' form (i.e. compatible with the rule of 17.817), each string
can be scaled up to a chosen common scale length, so that the same fret slots can be used for all
strings. The ΔS and ΔN values are also scaled, resulting in precise compensation distances for
both nut and saddle, on every string.

Aside from being more mathematically rigorous than Magliari's system, the great
strength of this method is that it allows frets to be cut to a predetermined spacing, but with
calculated compensation. This is especially useful to a luthier using a fretting template to cut
slots using a table saw for example, as it does not require a different template for every player or
set of strings etc., only a differently prepared nut and saddle. Furthermore the layout is much
more easily expressed to the knowledgeable player or luthier. It consists of a scale length, from
27

which fret slots are cut in the normal way, and a nut compensation and saddle compensation
value for each string, as opposed to Magliari's fret positions which must be presented as a table
containing the position of each individual fret.

Inharmonicity

The intonation models described so far have taken the principles which apply to ideal,
infinitely flexible strings and modified them to account for real strings' longitudinal stiffness, or
resistance to an increase in length. A second factor to consider is a string's bending stiffness,
which causes the phenomenon of inharmonicity.

Inharmonicity is the tendency for a string's upper partials to vibrate at progressively


higher frequencies than those of a perfect harmonic series. For an ideal string with no bending
stiffness, the 1st Mersenne equation (5) applies equally to every harmonic:

(31)

Where υp is the frequency of a partial p, so that υ1 is the fundamental frequency.

Thus for a given note, where length, tension and µ are fixed, the partials conform to an
harmonic series:

(32)
28

To take bending stiffness into account requires the modulus of elasticity E, the cross
sectional area S, and also the radius of gyration κ, which is equal to half the radius of the string.
Inharmonicity can be calculated with the following equation, given by Morse: 23

(33)

To simplify the notation, Byers24 defines the terms α and β:

(34)

and

(35)

so that:

(36)

23
Philip M. Morse, Vibration and Sound (New York: McGraw-Hill, 1948)
24
Byers, 1996, pp.39-40
29

The exercise then becomes one of solving the equation for, l'n the adjusted position of
each fret for which frequency corresponds to the above equation:

(37)

Finally, Byers combines the two models, for stretch and stiffness, into a single equation
by which he will generate fret positions:

(38)

The weakness of Byers' approach here is that he considers only the effect on the
fundamental, which is minor compared to that on the upper harmonics. Altering the length at any
given position will not alter this basic aspect of the string's behaviour. Gore, in addressing the
issue of inharmonicity concludes that ‘the bending stiffness of a string [has] negligible effect on
its frequency of vibration as the string [is] reduced in length’. 25 In other words, any sharpening of
the fundamental due to bending stiffness is effectively independent of changes in length, so it
would not be significantly affected by repositioning frets.

25
Gore, 2011, p.4-118
30

Inharmonicity is most apparent when low strings with strong upper harmonic content are
played together with higher strings. The fundamentals may be in tune with one another, albeit in
different registers, but the harmonics clash, causing an unpleasant sensation of roughness. One
solution is to flatten the low strings slightly, so that the harmonics are more consonant, a method
known as 'stretch tuning'.

This system is more commonly associated with piano tuning where, due to the greater
diversity of string diameters, inharmonicity is more apparent. Giordano describes an effect
known as the 'Railsback stretch', which shows the practice of professional piano tuners to tune
the fundamental frequencies of notes progressively lower than dictated by equal temperament in
the low registers, and higher in the upper registers, by a deviation of as much as 40 cents. 26 The
sigmoid curve of frequency deviation is even built into modern electronic piano tuning meters, to
reproduce the effect.

Howard “Buzz” Feiten recommends a version of stretch tuning for the guitar, whereby
the nut is set forward by a set percentage of the scale length, depending on the type of strings
(nylon, steel string acoustic or electric), and the open string tuning and saddle compensation are
tempered by small amounts as dictated by the 'Feiten Temper Tuning Tables'. 27 This system is
intended primarily for after-market application to guitars, with 'rule of 17.817' fret spacing, to
achieve a general improvement to intonation, and lacks the finesse available using a more
rigorous modelling approach. As such it is unlikely to provide reliable compensation for
inharmonicity.

Gore's own method is more straightforward, purely empirical and therefore specific to the
instrument and strings in question. The open first string is tuned to its target frequency using an
electronic tuner, and the remaining strings are tuned by ear, matching their natural harmonics to
fretted notes on previously tuned strings. 28 Provided that the instrument has been properly
intonated for the effects of stretch, this would seem to provide an effective means to counteract a
small measure of inharmonicity.

26
Nicholas J. Giordano, Sr., Physics of the Piano (Oxford, Oxford University Press, 2010), pp. 57-63, 'Waves on
Real Strings: The Effect of String Stiffness'
27
Howard B. Feiten et al. 'Method and Apparatus for Fully Adjusting and Intonating Stringed, Fretted Musical
Instruments, and of Making Adjustments to the Rule of 18'. US. Patent 5,814,745, 1998.
28
Gore, 2011, p.4-119
31

Bypassing the Model: Empirical Techniques

As discussed at the beginning of the chapter, empirical methods are often used to provide
data which it is difficult to model accurately. Good examples of this are measuring μ for wound
strings for which it is difficult to calculate the cross-sectional area, or using a test rig to
determine the modulus of elasticity E.

It is possible with more involved apparatus to eliminate substantial parts of the model.
The disadvantage to this approach is the requirement of high precision to achieve meaningful
results, which is expensive and difficult to attain outside of the lab. If a test is sufficiently well
designed however, it can not only negate a certain amount of complicated mathematical analysis,
but also account for factors which may be overlooked by a modelling approach, potentially
producing better results.

Byers, having explained his detailed model, describes just such a test rig. The rig consists
of a nut and saddle, arranged as they would be on the finished instrument, between which is a
short section of fingerboard. The fingerboard section is moveable, and its frets spaced so as to
acceptably substitute for every pair of adjacent frets on a complete fingerboard. The open string
is tuned to pitch, and the fingerboard section is adjusted until the fretted note sounds at its target
frequency, measured with an electronic tuner. The fingerboard is then locked in place, and its
position measured and recorded. This test provides Byers' X'n values for each string, which are
then subjected to his statistical alignment method. 29

The main drawback identified is the imprecision of measuring distances accurately using
rulers. This could be improved on using a Vernier system, but this adds yet another level of
expense and complexity to the system.

The most direct way to empirically improve intonation is to build extended adjustment
into the instrument itself. This approach negates any confounding factors which may influence

29
Byers, 1996 pp.41-43
32

the instrument differently than they do a test rig. One such system is Rick Toone's adjustable
nut.30 The device works in much the same way as a typical electric guitar bridge, allowing
continuous adjustment to the contact point for each string.

Figure 7: Toone Adjustable Nut

The accompanying literature does not include a method for setting up a guitar using such
a device without employing a model of some kind, however it would be possible by a 'trial and
error' approach to achieve improved intonation with no mathematical insight whatsoever.

The nature of the design makes it extremely limited in scope. It is compatible only with
headless instruments, and even then must be accommodated into the neck design. Also, as it is a

30
Rick Toone. 'Intonation Cantilever'. US. Patent 8,076,559 B2, 2011.
33

physical component of the instrument, its cost is extremely high per use, as opposed to a fixed
nut which has been designed based on a sound model, costing a fraction of the price.

One of the fundamental restrictions imposed on the studied models is that their frets be
straight, and that therefore it is necessary to fit the calculated fret positions for each string into a
single set of lengths, which necessarily involves a measure of compromise. The True
Temperament system use cutting edge technology to overcome this restriction, with revealing
results.

Anders Thidell's 2010 patent describes a test rig on which small, moveable frets are
adjusted individually, until the fretted note sounds at target frequency. 31 The position is
measured, and the position is fed into CAD software. The CAD file is then used to mill slots into
a fingerboard, which curve to fit the required position for each string. The frets are cast
individually to fit the appropriate slot:

Figure 8: True Temperament Frets

31
Anders Thidell. 'Device for String Instruments'. US. Patent 7,728,210 B2, 2010.
34

By employing high technology, this system not only bypasses the necessity of a
mathematical model, but also eliminates the inherent compromise necessary to make different
strings fit to a single fretting scheme. The results demonstrate graphically the shortcomings of
traditional fret and nut design. Of course this technology is proprietary, and therefore only
available to the luthier at a price. This practical objection aside, True Temperament fretting is the
most comprehensive solution to the problem of guitar intonation available to date.

The ideas and methods discussed in this chapter represent a cross section of current
thinking on the subject of improved guitar intonation. The following chapter seeks to draw
together this knowledge to devise a new fret board design method, which can be used practically
and effectively by the typical independent luthier. By way of demonstration, the method will be
used to design a solid body electric guitar with a novel specification.
35

Chapter 3

The final chapter employs the theoretical principles discussed in chapter 2 to develop a
comprehensive method for designing an instrument with tension-compensated intonation. Each
stage of the method is first described in general terms, as could be applied to practically any
fretted instrument. By way of example, values are then given for a specific design, namely a
solid body electric guitar.

First, a set of strings is selected and their physical properties established. The pertinent
instrument dimensions are specified, as are details regarding the anticipated playing technique.
These inputs are fed into a spreadsheet which calculates the tuning error which is expected for
each string at every fret position, and modifies the dimensions accordingly to return values
which compensate for the error.

String Properties

The strings selected for the example are Elixir Nanoweb Light-Heavy / .010”-.052”. The
critical properties are:

- Mass per unit length


- Cross-sectional Area
- Elastic Modulus

Mass per unit length is found by cutting each string into five specimens, each
approximately 150mm in length. Each specimen is then measured to .01mm with a digital
calliper, and the weight measured to .001g on a sealed balance. The weight is divided by the
length, and the resulting unit gmm-1 is identical to the SI equivalent KgM-1. The results for the
Elixirs are below:
36

String µ (KgM-1)
e1 0.000402
B2 0.000683
G3 0.001164
D4 0.003440
A5 0.005819
E6 0.008693

Figure 9: µ values for Elixir Guitar Strings

Cross-sectional area is also found by measurement. The plain strings are measured with a
micrometer to .005mm, and their area calculated as:

(39)

For the wound strings, only the core is considered. As the Elixirs’ cores are hexagonal,
their area is calculated as:

(40)

Where H is the height of the hexagon, i.e. the distance between opposite sides. Results
for the Elixirs are below:
37

Cross-
sectional
String area (mm2)
e1 0.053092916
B2 0.090792028
G3 0.152053084
D4 0.125054068
A5 0.160128097
E6 0.160128097

Figure 10: Cross-Sectional Area of Elixir Guitar Strings

Establishing the elastic modulus is more problematic. Attempts to calculate stress and
strain by tensile testing were unsuccessful, as the samples tended to fracture early at the ends,
producing false readings. Standard tensile tests specify ‘dog bone’ shaped specimens, where the
broad ends are clamped into the equipment, and the extension of the narrow central portion is
measured with an extensometer. Without specialised equipment it is not possible to derive the
elastic modulus of narrow, non-standard samples experimentally. Via email correspondence,
Elixir Europe informed the author that they use a high carbon steel, conforming to the ASTM-
A228 standard. The material data sheet for ASTM-A228 32 states an elastic modulus of 210 GPa,
which is used for subsequent calculations.

Setup Specification

The instrument specified is a solid-body electric guitar, with a nominal scale length of
650mm. The relative nut and saddle heights are based on Ibanez factory specifications 33, shown
below:

32
MatWeb Material Property Data, ‘ASTM A228’
<http://www.matweb.com/search/DataSheet.aspx?MatGUID=4bcaab41d4eb43b3824d9de31c2c6849> Accessed
15th February 2016
33
Dan Erlewine, How to Make Your Electric Guitar Play Great: Second Edition (Milwaukee, Backbeat Books,
2012), p.123
38

Absolute Nut Relative Nut 12th Fret Saddle


slot height Slot Height Action Height
(mm) (mm) (mm) (mm)
1 1.6 0.150 1.600 3.050
2 1.65 0.200 1.68 3.160
3 1.7 0.250 1.76 3.270
4 1.75 0.300 1.84 3.380
5 1.8 0.350 1.92 3.490
6 1.85 0.400 2 3.600

Figure 11: Setup Specifications for Example Instrument

The fret height is determined by the wire used, in this case Jescar FW57110. For
simplicity, the model assumes the neck to be completely straight, with no allowance for relief.
Initial fret positions Xn conform to standard ‘rule of 17.817’ spacing. Tuning is set to E-standard,
but the spreadsheet contains linked frequency values for any tuning down to a perfect 5 th below
standard.

Playing Technique

The variables m and j are adopted from Byers, to denote the horizontal and vertical
components of the string path around the fretting finger. At first these were set somewhat
arbitrarily at 7mm and 0.5mm respectively for all frets, but it soon became apparent that constant
values would not give a realistic representation of fretting action. The model used is adapted
from Gore’s analogous variables f and g.34 The dimensions are calculated as a proportion of the
distance between the two active frets, Xn and Xn-1. Specifically:

34
Gore, 2011, pp. 4-114 – 4-115
39

(41)

and:

(42)

where j0 is the vertical fretting component at the first fret, and P is the fretting pressure
coefficient, set here to 0.5mm and 0.85 respectively.

Building the Spreadsheet

The raw figures obtained above are entered into separate forms. These link to a
specification sheet, containing all the relevant figures for each string. These are fed into string-
specific sheets, each of which contains 24 rows, one for each fret. For simplicity, all values in the
string-specific sheets are expressed in SI units. The unfretted string length is calculated using
Pythagoras’ theorem, including as-yet unspecified values for saddle and nut compensation, ΔS
and ΔN:

(43)

The fretted string length at each fret is calculated as the sum of four line segments, as per the
Byers/ Bartolini geometric model, also incorporating ΔS and ΔN:
40

(44)

(45)

(46)

(47)

(48)

Combining this with the string property data, both fretted and unfretted tension are
calculated:

(49)
41

(50)

And from this the modelled frequency is calculated:

(51)

The L4n segment is used as this represents the vibrating portion of the string, between fret
n and the saddle. The modelled frequency f’n is compared to the target frequency fn, and the pitch
disparity Dn is expressed in cents, so as to compare like for like:

(52)

The goal is to minimise D across all frets, by varying the values of ΔS and ΔN which are
plugged into the geometric model. This is achieved by the ‘least squares’ method; a statistical
optimisation routine performed by the Microsoft Excel add-in ‘Solver’. A column is made of Dn2
values, the sum of which is returned in Solver’s ‘objective’ cell. Solver is programmed to return
the minimum ‘objective’ result by varying the values of ΔS and ΔN.

Because the Xn column is defined by the standard fret placement equation, fret slots can
be cut in the normal way, but with calculated values for saddle and nut compensation. These are
sent to a results sheet, where they are expressed in mm to 2dp. The results for the example
instrument are as follows:
42

ΔS ΔN
String (mm) (mm)
1 0.28 -0.84
2 0.57 -1.62
3 0.95 -2.79
4 0.52 -1.36
5 0.76 -2.02
6 0.97 -2.61

Figure 12: Compensation Values for Example Instrument


43

Evaluation

Two particular unexpected trends stand out from the modelled data. Firstly, the nut
compensation value ΔN is slightly greater than saddle compensation ΔS for all strings. This is at
odds with typical values found by both Gore and Byers, while applying similar models to slightly
different instruments. Secondly, the pitch deviation is significantly greater at the first fret than at
the remaining frets. These phenomena are closely related, and it is instructive to examine their
relationship.

The calculation of L1 (‘fretted string length at the first fret’) is subtly different to that for
other Ln values. The first segment L11 equals zero, as its start and end points are one and the
same; at the nut. Consequently, the vertical component of segment L21 includes both j, which is
common to all fret positions and Hnut, the height of the nut relative to the frets. Its value is
therefore unusually large, increasing tension and accordingly pitch deviation.

When ΔN and ΔS are varied, they have different effects on pitch deviation. Increasing ΔN
lowers relative pitch at all fret positions equally, as it is effectively reducing the string’s overall
length (Ln) and hence tension. Increasing ΔS has the direct opposite of this effect, but also
increases the vibrating string length (L4n). This reduces the pitch at successive fret positions
logarithmically, so that over the space of 12 frets the effect doubles. When optimising the total
deviation squared, a large deviation at the first fret therefore biases the calculation towards
greater nut compensation.

The apparent bias may be the result of the L21 calculation, and/or the calculation for j
itself. Equation 3.4 relates j dynamically to fret number n, regulated by the constants j0 and P. To
change either constant, or tweak the relationship itself such that j decreases less rapidly relative
to n, reduces the first-fret deviation bias, altering the relationship of ΔN to ΔS.

Yet another source of inaccuracy is the approximation of the neck as being straight,
where in practice necks usually have a small amount of relief. Introducing relief to the model
would have a dynamic effect on fretted string length in opposition to that caused by equation 3.7,
which would also have the effect of reducing the first-fret deviation bias.
44

The modelled compensation results can only represent a first approximation to the
optimum specifications for a real instrument. To refine the model in a realistic and effective way,
it is necessary to test its results empirically, using either a purpose-made test rig or a complete
instrument. By setting up the instrument or rig to model specifications, using a representative
range of players and recording any differences in outcome, the model can be evaluated, and its
inputs modified to return more realistic results. By targeting specific inputs for modification, e.g.
the L21 calculation or the j constants, the model can be used in reverse, to statistically optimise
the inputs to reproduce empirically valid results. Used iteratively, this process will be used to
produce a robust and reliable model for designing future instruments.
45

Conclusions

12TET is the universally accepted tuning system in Western music, and therefore sets the
standard by which intonation can be judged. Traditional systems of compensation fail to
accurately reproduce the notes of the equally tempered scale, resulting in instruments which are
demonstrably ‘out of tune’.

To address this problem, it is necessary to model the system mathematically or to


measure the tuning error of a real instrument empirically. In practice, any effective method will
combine elements of both techniques.

The Mersenne-Taylor equation describes the behaviour of ‘ideal’ strings, with infinite
elasticity. To predict the required compensation, it is necessary also to quantify elasticity,
incorporating Hooke’s Law. Stiff strings are also subject to inharmonicity, a phenomenon which
is more readily addressed by stretch tuning, in a manner similar to that used on pianos.

The mathematics can be bypassed by taking direct measurements of an instrument’s


tuning error, and making adjustments to correct it. This approach generally requires highly
accurate measuring equipment, and specialised hardware to facilitate the adjustment. As such it
is expensive and inefficient to implement, compared to a modelling approach.

A well-designed model allows the tuning error to be accurately predicted, and the design
modified to correct it. In most cases it is desirable to maintain straight, parallel frets, so for this
reason the modification is usually confined to the application of both nut and saddle
compensation to an otherwise familiar fingerboard.

The intonation spreadsheet described in chapter 3 is one such model. Its predicted tuning
error is significantly smaller than that which can be obtained by traditional ‘saddle-only’
compensation. Subject to empirical validation, it lays the groundwork for a powerful and flexible
tool for designing instruments with improved intonation.

Word Count: 8,613


46

Bibliography

Books

Angus, Jamie A. S. & Howard, David M., Acoustics and Psychoacoustics (Oxford:
Elsevier Ltd., 2009).

Backus, John, The Acoustical Foundations of Music (New York, NY: W. W. Norton &
Company, Inc., 1977).

Barbour, J. Murray, Tuning and Temperament: A Historical Survey (East Lansing,


Michigan State Press, 1951).

Callister, Jr., William D., Materials Science and Engineering: An Introduction (New
York, John Wiley & Sons, Inc., 2006).

Duffin, Ross W., How Equal Temperament Ruined Harmony (And Why You Should Care)
(New York, W. W. Norton & Company, Inc., 2007).

Erlewine, Dan, How to Make Your Electric Guitar Play Great: Second Edition
(Milwaukee, Backbeat Books, 2012).

Forster, Cris, Musical Mathematics: On the Art and Science of Acoustic Instruments (San
Francisco, Chronicle Books LLC, 2010).

Giordano, Sr., Nicholas J., Physics of the Piano (Oxford, Oxford University Press, 2010).

Gore, Trevor and Gilet, Gerard, Contemporary Acoustic Guitar: Design and Build (Terry
Hills, NSW: Trevor Gore, 2011).

Hall, Donald E., Musical Acoustics: Second Edition (Belmont, CA, Wadsworth Inc.,
1991).

Hiscock, Melvyn, Make your own Electric Guitar (London, Blandford Press, 1986).

Lindley, Mark, Lutes, Viols & Temperaments (Cambridge, Cambridge University Press,
1984).

Meffen, John, A Guide to Tuning Musical Instruments (Newton Abbot: David & Charles,
1982).

Mersenne, Marin, Marin Mersenne: An Edited Translation of the Fourth Treatise of the
Harmonie Universelle, by Robert Fortson Williams (Ann Arbor, MI: UMI Dissertation Services,
1995).

Morse, Philip M., Vibration and Sound (New York: McGraw-Hill, 1948).
47

Journal Articles

Bartolini, w., Bartolini, P.A., 'Experimental Studies of the Acoustics of Classic and
Flamenco Guitars' Journal of Guitar Acoustics #6 (1982), p.75-103

Boullosa, Ricardo R., Orduna-Bustamante, Felipe, Perz-Lopez, A., 'Tuning


Characteristics, Radiation Efficiency and Subjective Quality of a Set of Classical Guitars'
Applied Acoustics #56, issue 3 (Elsevier, 1999) p.183-197

Byers, Greg, 'Classic Guitar Intonation' American Lutherie #47 (Guild of American
Luthiers, 1996) p.34-45

Elmendorp, Sjaak, 'It's All About the Core, or: How to Estimate Compensation' American
Lutherie #104 (Guild of American Luthiers, 2010) p.56-60

Gilbert, John m., 'Intonation and Fret Placement' Soundboard Vol. XI, No. 1/Spring 1984
(Guitar Foundation of America, 1984) p.26-27

Macrostie, Don and Magliari, Gary, 'Beyond the Rule of 18: Intonation for the 21 st
Century' American Lutherie #116 (Guild of American Luthiers, 2013) p.6-17

Novak, Ralph, 'Why Build a Multiscale Guitar?' American Lutherie #121 (Guild of
American Luthiers, 2015) p.52-57

Patents

Feiten et al. 'Method and Apparatus for Fully Adjusting and Intonating Stringed, Fretted
Musical Instruments, and of Making Adjustments to the Rule of 18'. US. Patent 5,814,745, 1998.

Gimpel et al. 'Compensated Nut for a Stringed Instrument'. US. Patent 6,433,264 B1,
2002.

Novak, Ralph. 'Fingerboard for a Stringed Instrument'. US. Patent 4,852,450, 1989.

Thidell, Anders. 'Device for String Instruments'. US. Patent 7,728,210 B2, 2010.

Toone, Richard Warren. 'Intonation Cantilever'. US. Patent 8,076,559 B2, 2011.
48

Web Resources

Byers, Greg, 'Intonation Research'


<http://www.byersguitars.com/Research/Research.html> Accessed 9 th December 2015

Cambridge University Engineering Department, ‘Materials Data Book: 2003 Edition’


<http://www-mdp.eng.cam.ac.uk/web/library/enginfo/cueddatabooks/materials.pdf> Accessed
23rd January 2016

Elixir Strings, ‘Electric Guitar Strings’ http://www.elixirstrings.co.uk/guitar-


strings/electric-nanoweb-guitar-strings.html> Accessed 17th December 2015

Macrostie, Don, 'Tension Compensated Fret intervals'


<http://www.luth.org/images/web_extras/al116/01Red%20Diamond%20070322.pdf> Accessed
1st December 2015

Magliari, Gary, 'Method for Positioning Musical Instrument Frets that Compensate for
Fretting-Induced String Tension'
<http://www.luth.org/images/web_extras/al116/MagliariFretComp.pdf> Accessed 1 st December
2015

MatWeb Material Property Data, ‘ASTM A228’


<http://www.matweb.com/search/DataSheet.aspx?MatGUID=4bcaab41d4eb43b3824d9de31c2c6
849> Accessed 15th February 2016

Toone, Rick, 'Intonation Adjustable Nut' <http://www.ricktoone.com/nut.html> Accessed


10 December 2015
th

True Temperament AB 2014, 'True Temperament Fretting System'


<http://www.truetemperament.com/> Accessed 10th December 2015
49

Appendix: Using the Spreadsheet

Raw data must be entered into all cells of the input sheets: CS Area, Mass Per Unit
Length, Elastic Modulus, Tuning and Action. See Chapter 3 for examples of how to obtain these
data. The results are processed and collated in the Specification sheet.

To generate compensation values, it is necessary to load the ‘Solver’ add-in:

- Go to File > Options


- Click ‘Add-ins’
- Click ‘Go’
- Tick the ‘Solver Add-in’ box
- Click OK

Solver will now appear in the Analysis group on the Data tab.

Select one of the string-specific sheets from the menu. If all raw data has been entered
correctly, every cell should have a numerical value. Click on Solver to open the ‘Solver
Parameters’ dialogue box.

Under ‘Set Objective’ select the cell labelled ‘Sum of Squares’, in this case $U$28.

Under ‘By Changing Variable Cells’ select the nut and saddle compensation values,
$U$30:$U$31.

Set objective to Min, and solving method to GRG Nonlinear.

Click Solve to generate optimised compensation values.

You might also like