Five lectures on

THE ACOUSTICS OF THE PIANO

Anders Askenfelt, editor 1990 Royal Swedish Academy of Music

Contents
Preface............................................................................................................................................................................... 5 Introduction .................................................................................................................................................................... 7 Piano design factors .................................................................................................................................................. 17 From touch to string vibration ............................................................................................................................. 39 The hammer and the string.................................................................................................................................... 63 The coupled motion of piano strings.................................................................................................................. 85 The strings and the soundboard .......................................................................................................................... 97

Lectures: Harold A. Conklin Jr.: Piano design factors - their influence on tone and acoustical performance Anders Askenfelt & Erik Janson From touch to string vibrations Donald E. Hall: The hammer and the string Gabriel Weinreich The coupled motion of piano strings Klaus Wogram The strings and the soundboard

Preface
This volume contains five lectures given at a public seminar at the Royal Institute of Technology in Stockholm, May 27, 1988. The lectures are based on accumulated experience in piano design as well as recent experimental and theoretical studies -all presented in a popular style. The seminar day was preceded by two days of discussions between the lecturers and invited representatives from piano manufacturers. Two representatives from Steinway & Sons, Daniel T. Koenig, Vice President of Manufacturing, and William Y. Strong, Director of Research and Development, joined the speakers in a closing panel session at the seminar, answering questions from the audience and pondering future improvements in piano design. The seminar day was closed by a remarkable concert -"From harpsichord to concert grand" -in which the development of the piano was illustrated. The stage featured six instruments representing piano design from 1813 to 1980, and a harpsichord as a reference to the keyboard instruments before the piano epoch. Three pianists performed on the instruments playing music contemporary to each instrument. Excerpts from this concert are included on two gramophone records accompanying this book. The concert was recorded by The Swedish Radio Company and later broadcasted. The seminar was initially proposed by the Music Acoustics Committee of the Royal Swedish Academy of Music. Later a Keyboard Committee of the same academy was founded, which ran the seminar and additional events in cooperation with the Department of Speech Communication and Music Acoustics at The Royal Institute of Technology and the Swedish Radio Company. The editing of this volume was considerably facilitated by the continuous and thoughtful support of my colleague Erik Jansson. Due thanks are given to Si Felicetti, Gudrun WeinerRispe and sa Wallner for patient assistance in the processing of the manuscripts and figures. Stockholm in January, 1990 Anders Askenfelt, editor

Introduction
Background
The scientific study of the acoustics of the piano goes back to Hermann von Helmholtz (1821 - 1894), a German physician and scientist, active in both neurology, optics, electricity and acoustics. He compiled much of his thinking about sound, musical instruments and hearing in a book "On the Sensations of Tone", which still is very much worth reading.(*) Helmholtz's interest in musical instruments was strongly coupled to the perception of their sound. In view of his limited measurement equipment - in which his ears played a central role - he made remarkable contributions to the understanding of the tonal characteristics of several musical instruments, among them the piano. In a series of Appendices, of which some have become more famous than the text itself, he also presented theoretical analyses, including the case of a string struck by a hammer. Helmholtz was followed by occasional studies during the decades around the turn of the century. These early investigators dealt in particular with the interaction between the hammer and the string, a question which in fact still not has been completely settled. After important pioneering works on almost every aspect of the piano in the 40's and 50's, by the use of what we would call rather modern equipment, the study of the acoustics of the piano has gained a renewed interest during the last decade. Although many, many questions remain to be answered, a deeper understanding of the sound generation in the piano now seems less remote than for several other instruments, in particular the bowed instruments. The piano was invented in the 18th century, developed to its present design during the 19th century - a period during which the bulk of classical piano music was written - and produced on a large scale and frequently used in all kinds of music during the 20th century. However, a complete understanding of the acoustics of the instrument will probably not be reached until the next century. This may sound a little discouraging from a scientific point of view, but the same statement holds true for almost all traditional instruments. The situation is nothing but a result of man's incredible ingenuity in developing sound sources which not only produce a pleasant sound, but which can also be intimately controlled by the player. This evolution has resulted in musical instruments for which the acoustical function turns out to be extremely complex, despite the fact that the instruments are based on seemingly simple principles and made of common materials. The piano is a representative example among the string instruments. The principle of its function is indeed simple; a felt hammer strikes a metal string which is connected to a large wooden plate. The string is set in vibration by the impact, and the vibrations are transferred to the plate which radiates the sound. However, for none of the steps in this process - the collision of the hammer with the string, the transmission of the string vibrations to the wooden plate, and the radiation of sound from the plate into the air - the physics is well enough understood to permit a detailed description of what actually happens in the real instrument. In addition, "simple" materials like felt and wood turn out to have very complex properties - different from sample to sample! -which further increases the difficulty of describing the phenomena.

All this would have been enough, but the most cumbersome step is yet to come. The quality of a traditional instrument is rated using our hearing as the ultimate test instrument. This means that results of acoustical measurements should always be viewed in the light of how they relate to the perceived sound. But this may not even be possible, because the perception of sound, especially musical sounds, is a field which unfortunately is very poorly explored. There are still many gaps in our knowledge of the relationship between physical and perceptual properties of sounds. For this reason, many interpretations of experimental results must remain on the level of advanced guesses. With these difficulties in mind it is not surprising that it was possible to put a man on the moon before the acoustics of a traditional instrument like the piano had been thoroughly explained.

Landmarks in piano history

In contrast to most other traditional instruments like the violin or the trumpet, whose origins vanish in the haze of the past, a specific year and name can be attributed to birth of the piano. In 1709 the Italian harpsichord maker Bartolomeo Cristofori replaced the plucking pegs in a harpsichord by small leather hammers which he let strike the strings. Since this new design allowed the notes to be played either soft or loud depending on how the key was struck(**), he called his new instrument gravicembalo col piano e forte ("a large harpsichord with soft and loud"). Soon the grandiose name was shortened to pianoforte or fortepiano and eventually to piano. Cristofori's piano was developed from the harpsichord and consequently rather small and made entirely out of wood. As time passed, however, the development of larger instruments with more and heavier strings at higher tensions - all in order to increase the volume of sound - necessitated a more rigid construction. The wooden frame was successively reinforced with more and more pieces of iron, and in 1825 the complete cast iron plate was introduced by the American piano maker Babcock. The iron plate could withstand the increased string tension, and prevented the instrument from gradually changing shape as the wooden instruments did. Also, it now became possible to keep the tuning stable over longer periods of time. The hammers of the early pianos were tiny, light pieces made out of leather. However, the introduction of coarser strings at higher tensions demanded larger and heavier hammers. In 1826, felt hammers were tried for the first time by an ingenious piano maker in Paris named Pape. The success was immediate and lasting. An incredible amount of work was devoted to the development and refinement of the actions. A prominent name in this connection is the French piano manufacturer Erard who invented the so-called double repetition action in 1821, which is the type of action still used in the grand piano. The construction was refined by another French manufacturer named Herz around 1840. Smaller improvements were made during the following decades, but since then no essential changes have been made. A simpler type of action, the Viennese action, lived a parallel life before it eventually vanished during the first decades of this century. The compass of the piano has increased successively during its history. Cristofori's piano had only four octaves. Today a piano with a standard setup of 88 keys will cover more than

seven octaves (A0 = 27.5 Hz to C8 = 4186 Hz), no less than the pitch span of the modern symphony orchestra. Furthermore, the acoustic output at fortissimo - small as it might seem (of the order of 0.1 W) - surpasses all other string instruments. This power is enough to fight even the largest ensemble (although brute force not always is the best way of making a solo instrument heard above the orchestra). The early pianos were of the type we now call a grand piano. During the 19th century the manufacturers discovered a market for smaller and cheaper models, and squares and uprights were constructed, both instruments being economy versions of the "real" piano and filled with compromises. Both the grand and upright pianos as we know them today developed during the 19th century, which saw a wealth of patent applications during its latter half. The period of development declined shortly before the turn of the century, indicating that the construction was perfected, at least for the time being. Several of the recognized piano makers have had a long tradition including connections with famous composers. Mozart played a Stein piano from Austria, Beeethoven preferred an English Broadwood, and Chopin's piano was made by Pleyel in France - instruments from eminent makers which today, however, are out of business or operating on a very low level. Liszt and Wagner, on the other hand, used grands from Steinway & Sons (New York, Hamburg) which were very close to the instruments we still are used to hearing 100 years later. Other old, recognized piano manufacturers still in operation are Bsendorfer (Vienna), Bechstein (Berlin), Baldwin (USA) and Yamaha (Japan). The 20th century has been rather quiet as regards the development of the piano, but a dramatically increased production has manifested itself in an undesirable way. The beginnings of a lack of suitable wood and felt for piano purposes can be discerned. This will successively put pressure on the manufactures to search for new materials which can replace the traditional ones. This could, or probably will, demand changes in the design of several major parts in the piano, and the possibility of an active period of development like the one a century ago cannot be ruled out.

Thinking about the future

Today, the piano is challenged by synthesizers, especially so the economy versions of upright pianos. These pianos do not perform particularly favorably either in price or in tone quality compared to dedicated piano synthesizers ("digital pianos, samplers"). Still, the production of traditional pianos is large, estimated at 900 000 instruments a year worldwide (1988). In particular, the grand piano seems to continue to attract professional keyboard players of all genres, apparently for a number of reasons. Although the quality of the sound probably is the main cause of its fascination, the mechanical response from the instrument via the keys and the vibrating structure also seems to be very important. In view of the rapid development of new instruments based on digital sound generation, it is tempting to speculate about the future for the piano and the other traditional instruments. It seems reasonable to suppose that the singing voice will be recognized as a musical instrument as long as we use speech in communication. The vowels in speech and singing will familiarize us with harmonic sounds, i. e. sounds which are associated with a distinctive pitch. As long as pitch is used as a mean of communication in music, string and wind

instruments will take an exclusive position, because strings and pipes are the only tools available for generating such sounds mechano-acoustically. A piano-like instrument with struck strings could thus be assumed to be a natural member also of a future instrument inventory, should the traditional way of generating sounds survive. However, it is also possible that in the future most music will be performed on electronic devices. This technique gives a much wider freedom in designing the sounds, including imitation of the traditional instruments. Such imitations could also include extrapolations to new pitches and dynamic levels, not accessible by the original instruments. It is hard to deduce a priori if the piano sounds belong to the group of traditional musical sounds which will survive in the long run, when transferred to a family of new instruments. However, in view of the present popularity of the piano and recognizing the slow change in taste of musical sounds hitherto, it is an advanced guess that pianolike sounds will be used and enjoyed for at least another century.

Basics of piano acoustics

In this section, a survey of basic piano acoustics is given for those of the readers who want an introduction to the lectures. The fundamental principles which govern the acoustics of the piano are presented in a somewhat simplified form. A detailed and more realistic story of the sound generation in real pianos follows in the lectures.

Construction
A schematic view of the piano is shown in Fig. 1. A steel string is suspended under high tension between two supports (the agraffe or capo d'astro bar and the hitch pin) fastened in the metal frame (the plate). Close to the hitch pin end, the string runs across a wooden bar, the bridge, which is glued to a large and thin wooden plate, the soundboard. The level of the bridge is slightly higher than the string terminations, thus causing a downbearing force on the bridge and the soundboard. The soundboard is reinforced by a number of ribs glued to the underside, one reason being to make the soundboard withstand the downbearing force. The string is struck by a felt hammer, which gains its motion from the key via a complicated system of levers, the action.

Fig. 1. Principal sketch of the piano, designating the main components.

String motion
Physically, the string motion can be described in the following way. As the hammer strikes the string, the string is deformed at the point of collision (see Fig. 2). The result is two waves on the string, travelling out in both directions from the striking point. The wavefronts enclose a pulse, or hump, which gradually gets broader.

Fig. 2. The evolution of the propagating pulse on the string after hammer impact.

However, as the string is struck close to its termination at the agraffe, one of the wavefronts (the one travelling to the left in the figure) soon reaches this end and is reflected. The reflection at a rigid support makes the wave turn upside down. This inverted wave starts out to the right and restores the string displacement to its equilibrium level. The surprising situation has now developed that the wavefront initially travelling to the left in the figure, has turned into the trailing end of a pulse of fixed width, propagating to the right towards the bridge. At the bridge, the entire pulse is reflected, the effect being that the pulse starts out in the opposite direction upside down. A new reflection at the agraffe turns it right side up again, and soon the pulse has completed one round trip and continues out on the next lap. If the key struck happens to be A4 = 440 Hz ("concert A"), the pulse completes 440 such round trips per second.

Pitch, partials and inharmonicity

The propagation velocity of the pulse on the string is determined by the tension and mass per unit length of the string, a higher velocity the tauter and lighter the string. The number of round trips per second, the fundamental frequency (closely related to the perceived pitch), also depends on the distance to be covered - the longer the string the longer the round trip time (fundamental period), and hence, the lower the pitch. The pitch of a string is thus determined by a combination of its length, tension, and mass per unit length. In particular, string length can be traded off against mass per unit length in order to reduce the size of the instrument. This can be seen in the bass section, where the strings are wrapped with one or two layers of copper in order to make them heavy and thus relatively short. The

advantage of a wrapped string over a plain string is that the mass can be increased without reducing the flexibility drastically. A piano string need not be perfectly flexible, but a too stiff a string would have a detrimental influence on the tone quality as will explained below A piano string, like all other strings, has a set of preferred states of vibration, the resonances, or modes of vibration (see Fig. 3). When a string is vibrating at one of its resonances, a condition which usually only can be reached in the laboratory, the motion of the string is of a type called sinusoidal. The corresponding sound is a musically uninteresting sine wave. In normal use, however, where the string is either struck, plucked or bowed, all resonances are excited, and the result is a set of simultaneously sounding sine waves, partials, forming a complex tone.

Fig 3. The four lowest modes (resonances) of a rigidly supported string. Sometimes these elementary states of vibration are referred to as standing waves, because the amplitude contour does not change with time.

Such a tone is conveniently described by its spectrum, which shows the frequencies and strengths (amplitudes) of the partials (see Fig. 4, bottom). As mentioned, the pitch of the tone is related to the frequency of the lowest member in the spectrum, the fundamental. To be more specific, it is the frequency spacing between the partials - which for a piano tone is closely the same as the fundamental frequency - which is the closest physical correlate to the perceived pitch. The relations between the amplitudes of the partials and their evolution in time contribute to our perception of tone quality. The pulse running back and forth on the piano string has a most surprising connection to the string modes (resonances). It can be shown mathematically that the travelling pulse is made up of a sum of all the string modes! The shuttling pulse and an (infinite) sum of string modes of appropriate amplitudes are equivalent; they are just two ways of representing the same phenomenon (cf. Fig. 4). So while our eyes will detect the pulse motion (if slowed down enough by the use of a stroboscope) our ears prefer to analyse the string motion in terms of its partials or Fourier components, so named after the French mathematician who first described this equivalence. Fourier also stated that if the motion is periodic, that is, the same events will repeat indefinitely with regular intervals, the frequencies of the corresponding partials will be harmonic. This means that the frequency ratios between the partials will be exactly 1 : 2 : 3 : 4 . . . , which will be perceived as a sound with a clearly defined pitch and steady tone quality. The statement can also be turned the other way around; if the resonance frequencies of a string are strictly harmonic, the resulting motion of the string will always be periodic.

Fig. 4. Schematic illustration of the equivalence of the pulse motion on the string (top) and a sum of the string modes (resonances) (middle). The properties of the tone are conveniently summarized by its spectrum (bottom), showing the frequencies and amplitudes of the components (partials).

In real pianos, the resonance frequencies of the strings are not exactly harmonic. The frequency ratios are slightly larger than 1 : 2 : 3 : 4 . . . , more like 1 : 2.001 : 3.005 : 4.012 . . . , which is referred to as inharmonicity. The inharmonicity in piano strings, which is caused by the bending stiffness of the steel wire, is a desirable property as long as it is kept within limits. According to Fourier, the string motion will now not repeat exactly periodically as the note decays, but change slowly which gives a "live" quality to the note. Returning to the excitation of the string by the hammer impact, not only the amplitude of the initial pulse on the string changes with the strength of the blow, but also its shape. This is due to a remarkable property of the felt hammer, more specifically the characteristics of its stiffness. The stiffness increases (the hammer becomes progressively harder to compress) the more the hammer already has been compressed, a phenomenon referred to as nonlinear stiffness. This means that a harder blow not only will give a larger amplitude but also sharper corners of the pulse on the string. Again, according to Fourier, sharper wiggles in the waveform correspond to more prominent high frequency partials in the spectrum. Consequently, the piano tone will attain a different ("more brilliant") tone quality at forte (loud) compared to piano (soft).

Sound radiation and impedance mismatch

The vibrating string contains all the partials we would like to hear, but unfortunately the string is in effect unable to radiate sound. The difficulty is well described by the English saying: "You can't fan a fire with a knitting needle!" The reader may easily verify this statement by making the experiment, but can also notice that by means of a large object like a tray instead of the needle, it is quite possible to fan a fire even from a distance. The point is that a certain flow of air must be pumped back and forth per second in order to radiate a "fan wave."

This can be achieved with a limited motion of the tray having a large cross section, while the needle would have to make unreasonably large movements to reach the same effect. The acoustic engineer would "explain" the situation by saying that the radiation resistance of the tray is much higher than that of the needle. In other words, because of its larger area, the tray is much better than the needle as a transmission link between the motion of the arms and the motion of the air. Returning to the piano, we now realize that as the thin string cannot radiate a sound wave itself, its motion has to be transferred to a much larger object which can serve as a more efficient radiator of sound. This is readily done by incorporating a soundboard in the design, including a bridge as a connecting element to the string(s). But now the piano designer meets with a new difficulty. The soundboard is much heavier than the string, which means that the string will not be able to vibrate the soundboard efficiently and the vibrational energy will still be trapped in the string. Only slowly the energy will leak into the soundboard during repeated reflections of the string pulse at the bridge. In engineering terms, there is a mismatch between the mechanical impedance of the string and that of the soundboard. The mechanical impedance is a property that tells us to what degree an object resists (impedes) motion. From the point of view of the string, the soundboard has a very high (input) impedance; it can be thought of as a very heavy stone, or a very stiff spring, which must be vibrated vigorously. The experienced reader will certainly agree that this is a most uncomfortable task with little chance of success.

Loudness versus "sustain"

However, conditions can be improved, or in other words, the impedance mismatch can be diminished, by increasing the (characteristic) impedance of the string. This is easily done by making it heavier and by increasing its tension. But a heavier string usually means a thicker string, which automatically gives a higher stiffness and hence more inharmonicity, which soon spoils the desired piano timbre. Piano designers circumvent this problem in two ways, either by wrapping a rather thin steel core with copper (which also influences the pitch as mentioned), or by "splitting" a thick plain string into two or three strings, tuned to (almost) the same frequency, a technique called multiple stringing. Now the vibration energy is transmitted more efficiently from the string(s) into the soundboard and the note sounds louder, perhaps "too" loud. Because here the next difficulty appears; the gain in loudness does not come for free. It stands to reason that the pianist cannot feed energy continuously to the string like the violinist via the bow. Consequently the piano tone is condemned to decay and die. The question is then how to spend the energy quantum delivered at the key stroke in the best way. If a loud and thus necessarily shorter note is desired, the impedance mismatch between string and soundboard should be decreased by making the strings heavier and tightening them even harder. On the other hand, the note can be made longer by using lighter and less tense strings, but at the expense of loudness. The trade-off between loudness and duration, or "sustain," of the tone is a difficult problem in piano design, especially as the impedance of the soundboard

can vary wildly from note to note, due to its inherent resonances. It is easy to get a piano in which some notes are loud and short while adjacent notes are much softer and longer, a musically most unsatisfying situation. Fortunately, such fluctuations between notes as well as the basic conflict between loudness and sustain can be alleviated in an almost miraculous way by multiple stringing, a phenomenon which is covered in detail in one of the lectures.

The imperfect soundboard

The soundboard radiates sound much better than the strings do, as mentioned, but nevertheless it has several severe shortcomings. One occurs at very low frequencies and is due to the fact that both sides of the soundboard are directly exposed to the surrounding air. The reason is the following. Let the soundboard be moving upwards, pushing the air above its upper surface together. This causes a temporary excess of air molecules in a region above the soundboard, a compression, corresponding to an increased pressure. The underside of the soundboard is also moving upwards, so there is at the same moment a temporary loss of air molecules beneath the soundboard, a rarefaction, corresponding to a reduced pressure. As nothing prevents the compressed air on the upper side from flowing into the lower region, this pressure difference will soon be neutralized. Half a period later, when the soundboard is moving downwards, the process repeats but now the air flows from the lower to the upper side. So, at low enough frequencies - as long as the motion of the soundboard is slow enough to allow the exchange of air to take place before the direction of its motion has reversed -the soundboard will uselessly pump air from its upper side to its lower side and back again instead of radiating sound. The phenomenon is called acoustic short-circuiting, and can be avoided by separating the two radiating sides of the soundboard by an (almost) closed sound box, as in the guitar or in most harpsichords. A similar phenomenon can be observed also at higher frequencies. Now the soundboard no longer vibrates as a unit but spontaneously divides into smaller vibrating areas separated by thin regions of no motion (nodal lines). Depending on frequency, the vibrating areas form different patterns; the higher the frequency, the smaller and so the more numerous are the areas. These preferred states of vibration are called the eigenmodes (modes), or often, the resonances, of the soundboard. Adjacent vibrating areas vibrate in what is called opposite phase, which means that while one area is moving upwards its neighbour is moving downwards and vice versa. Also in this case, it is easy to imagine that a useless exchange of air between adjacent areas can occur instead of the desired sound radiation.

That's all!
This closes the short survey of basic piano acoustics. Once again, it is to be understood that the explanations are simplified, dealing only with the basic aspects of the phenomena. Against this background, the lectures that follow will illustrate the wealth of complications which arise in real instruments.

A note on units
In this volume, the use of metric (SI) units is encouraged. While the use of meters and kilograms probably will cause English and American readers only minor problems, the force unit Newton (N) might be less familiar. As a rule of thumb, 1 N corresponds to the weight of an apple (mass 100 g)!(***) Likewise, 10 N corresponds approximately to the weight of a mass of 1 kg, for example 1 litre (1 US quart) of milk. The naming of octaves and pitches follows the straightforward nomenclature given by American standards. In this notation the "middle octave" is indicated by number four (middle C = C4). The lowest note on full size piano is A0 and the highest C8.

Departure
After these introductory passages, it is time for a detailed voyage into the world of the acoustics of the piano, guided by experts in the different areas. The lectures follow in the same (logical) order as they were given on the seminar day, but as the contributions are essentially independent the readers may feel free to follow their own paths. In the first lecture, Harold Conklin, an experienced piano design engineer, outlines the design principles of the parts of the piano, and makes comparisons between the early and the modern instruments. Secondly, Anders Askenfelt and Erik Jansson, researchers in music acoustics with a focus on string instruments, present measurements from the initial steps in the tone production, from the moment when the pianist touches the key up to and including the string vibrations. Then follows a theoretical study by Donald Hall, a physics professor with a strong personal interest in keyboard instruments, who describes a computer model of what actually happens during the collision between the hammer and the string, and the implications for the string vibrations. The decay of the piano tone, and in particular the influence of multiple stringing is covered next by Gabriel Weinreich, also a physics professor with a strong interest in music acoustics. Finally, the sound radiation and its connection to the properties of the soundboard are described by Klaus Wogram, a researcher with many years of experience in investigating musical instruments, in particular brass instruments and the piano.

Notes (*) Hermann von Helmholtz: Die Lehre von Tonempfindungen als physiologische Grundlage fr die Theorie der Musik, first edition 1862, English translation of the fourth edition in 1885 by A. J. Ellis: On the Sensations of Tone as a Physiological Basis for the Theory of Music, reprinted (paperback) by Dover Publications Inc., New York 1954. (**) It is true that also the harpsichord can be played at somewhat different dynamics depending on how the key is depressed. Compared to the piano, however, the dynamic range is narrow, and dynamics are usually not prescribed in harpsichord music. (***) This useful remark was given by one of the lecturers (G. Weinreich) on an earlier occasion.

Piano design factors

- their influence on tone and acoustical performance Harold A. Conklin Jr

Introduction
My presentation will be an overview of some of the ways in which the design of a piano affects its tone and acoustical performance. It is not possible in a short lecture to mention all the important factors, because so many things in a piano affect its sound. Fig. 1 shows the oldest existing piano, the Cristofori instrument of 1720, which is in the Metropolitan Museum at New York City. In a recorded excerpt we can hear this historic instrument followed by the familiar sound of a contemporary concert grand (sound example 1).

Fig. 1. Cristofori piano of 1720.* (By permission of the Metropolitan Museum of Art, New York: The Crosby Brown Collection of Musical Instruments, 1889. Piano Forte (89.4.1219): compass 4 octaves and one quarter (C - F), Italian, Florence, 18th C., 1720. Maker: Bartolomeo Cristofori.).

It is obvious that the extreme differences in design between these two instruments produce extreme differences in tone quality. From time to time we will refer again to the oldest piano, but in order to concentrate on acoustical factors we will ignore or mention only briefly some of the important mechanical differences. For example, it is evident that the action of the 1720 piano (see Fig. 2, Pollins 1984) is much simpler and less controllable than that of a modern grand. After listening to the recording of the old piano one can say almost with certainty that the music that was played on the modern instrument could not be played properly with the action of Cristofori.

Fig. 2. Action of 1720 Cristofori piano. (By permission of the Journal of the American Musical Instrument Society).

The hammers
The hammers of a piano not only define the instrument: they also are among the most important factors in determining its tone quality. The hammers in the 1720 piano have wooden heads which are covered with leather (Fig. 3). Modern piano hammers are covered with wool felt that is compressed and stretched over a wooden molding (Fig. 4). Often two layers of felt are used. In Fig. 5 a modern hammer with the outer felt loosened is shown.

Fig. 3. Hammers of 1720 Cristofori piano (By permission of the Metropolitan Museum of Art).

Fig. 4 and 5. Modern grand piano hammers as normal and with outer felt loosened.

A view of the action from the 1720 piano is seen in Fig. 6. Cristofori glued his hammers onto wooden shanks as we are still doing today. In a modern grand piano the mechanical system of the hammer head on its somewhat flexible shank exhibits a major vibrational mode around 260 Hz as installed in the piano. This mode is not normally audible in the lower half of the piano's compass, but it can be heard in the treble register, as part of the "knock" component of the tone, beginning somewhere around A4 (key 49), and can be shown to affect noticeably the tone of the instrument. An improvement in the tone can sometimes be

obtained by shifting the frequency of this resonance. The motion of the hammer as it contacts the strings is very complex, and is only recently becoming clear (Hall 1986, Hall 1987a, 1987b, Hall & Clark 1987, Boutillon 1988, Hall & Askenfelt 1988).

Fig. 6. Cristofori action being played (By permission of the Metropolitan Museum of Art)

The hammers of the Cristofori piano are all about the same size. In a modern piano, the size and weight of the hammers increase from treble to bass in order to achieve the best compromise between tone quality, loudness, and playability. Fig. 7 shows typical hammer head weights for a modern grand piano.

Fig. 7. Typical weight curve for modern grand hammer heads.

The largest bass hammers may weigh around 11 grams. The smallest treble hammers may weigh as little as 3.5 grams each. Somewhat more output could be obtained at the extreme treble end of the scale if the hammers were somewhat lighter, but this would increase manufacturing problems. In the bass, tones having somewhat more fundamental energy could be obtained by using heavier hammers but then the piano would become harder to play. An increase in hammer weight can be counterbalanced statically by installing additional lead weights near the fronts of the key levers so that the force required to depress a key very slowly will remain at its nominal value (usually around 50 grams). However, this cannot compensate dynamically for increased hammer mass. Key velocities

corresponding to higher musical dynamic levels can require a finger force corresponding to several kilograms, and the value of this force increases noticeably with an increase in hammer weight. The heavier a piano hammer is, the longer it will stay in contact with the string(s). There is a critical region of the piano's compass, between about G4 and G6 on the keyboard. Within this range the contact time of the hammer against the strings becomes equal to the roundtrip travel time for the initial pulse on the strings (Benade 1976). Below this range the hammer leaves the strings before the arrival of the first major reflection from the far end; above this range the hammer normally is still in contact. If the hammer is still in contact at the time of the first reflection, losses occur that decrease the output of the piano and may cause an undesirable quality of tone. In order to produce a graceful tone within this critical range, it is important to have an optimum hammer striking position along the strings, to have the hammer strike all of the strings of a note equally, and to keep the hammers from being too heavy. These factors are interdependent. If the hammers in the treble are too heavy, the tone will not be as loud. If the hammers of the 1720 piano were to be used in a modern instrument, the tone of the bass and middle registers would sound too thin and bright, and the treble tone probably would be harsh. The hardness of a piano hammer directly affects the loudness, the brightness, and the overall tone quality of the instrument. In order to produce the best tone, each hammer must have its hardness within a certain range. Also, the hardness should have a gradient such that the string-contacting surface is softer than the inner material. If there is no gradient, the result can be poor tone or undesirable noise components. In Fig. 8 a special tool called a durometer is shown in use to measure the hardness and indicate the gradient of a hammer. This measurement can indicate whether the hammers have the right hardness to make a good piano tone. You could also find this out just by listening to the piano, if the hammers were already in place. But by measuring the hardness first it can be determined in advance whether the hammers can sound good, and it will be indicated how much work will be required to voice them.

Fig. 8. Durometer in use to measure hammer hardness.

Fig. 9 shows the measured hardness for three different hammers of similar size and weight. To demonstrate the relation between hardness and tone quality I have made a recording of

the tone produced by each of these hammers when installed at G5 (key 59) in the same piano (sound example 2). First you will hear the softest hammer played six times, then the harder hammer, and finally the hardest hammer. (The amount of difference you hear in the tone may depend on where you are sitting in relation to the loudspeakers.) As I hope you can hear, the softest hammer produces a pleasant tone that is perhaps a bit too soft (dark); the middle hammer produces a significantly brighter and louder tone, and the hardest hammer produces a still brighter but somewhat harsh tone that contains excessive noise components. Fig. 9. Shore A hardness for three hammers.

The optimum hardness for a hammer varies widely with its keyboard position. In order to produce tones of uniform loudness all across the scale, the extreme treble hammers must be much harder than the middle or bass hammers. The need to make the hammers harder in the treble usually begins, probably not by coincidence, in the critical region where the roundtrip time becomes equal to the hammer-string contact time. Fig. 10 indicates approximately how the relative hardness of hammers should vary across the scale in order to produce tones of equal loudness for an equal key effort. Of course the optimum value for hardness also depends on how bright a tone the listener prefers, so this graph gives only a general indication.

Fig. 10. Approximate relative hardness of piano hammers for equal loudness.

Hammers can be "voiced" by a skilled piano technician to make them harder or softer, in order to produce the best tone and smooth response from note to note. In voicing, the felt may either be softened by piercing it with needles at certain carefully chosen locations, or it may be hardened, either by filing away the soft outer felt with sandpaper, or by applying a chemical hardening agent. Voicing has little measurable effect on the lower partials of bass tones. In the treble, all of the partials are affected. Good hammers, properly voiced, are necessary to make a fine piano, but they are not sufficient. The other parts of the instrument are at least equally important. The hammers merely provide the exciting force for the strings. A bad piano equipped with even the best hammers will still be judged a bad piano.

Where should the hammer hit the string

The hammer striking ratio (d/L) for the 1720 Cristofori piano and for two representative modern pianos is shown in Fig. 11. Here L stands for the speaking length of the string and d is the distance from the closest string support (the agraffe) to the point where the hammer strikes. The values for the 1720 piano seem to wander over a wide range to no apparent purpose. Early makers did not fully appreciate the effect of varying d/L but by the late 18th century, piano makers began to know what values work best (Harding 1933). Many books about pianos state that the best place for the hammer to strike the strings is between 1/7 and 1/9 of their speaking length. (Good 1982, Marcuse 1975, Mc Ferrin 1972, Briggs 1951, White 1946, Wood 1944, Vant 1927, Ortman 1925, Wolfenden 1916, White 1906, Hansing 1888, Brinsmead 1879, Helmholtz 1863).

Fig. 11. Striking ratio (d/L) for two contemporary pianos and for 1720 Cristofori.

This is certainly not true for all the notes of modern pianos. In the best modern grand pianos the smallest treble hammer (C8) is always positioned at the factory for each piano individually and is set to produce the loudest tone. This normally occurs for a d/L-value much smaller than 1/9, usually in the range between 1/12 and 1/17. As you can see from

the curves labeled "contemporary" in Fig. 11, d/L in the bass is a little less than 1/8, and it decreases gradually up to around A4 (note 49), and then decreases rapidly. How d/L should vary across the compass depends on a number of factors and is decided by the designer of the piano. In the mid-treble, the best striking ratio often is a compromise between maximum first partial energy and the most graceful tone. Reducing the striking distance in this region generally makes the tone sound thinner because less fundamental energy is present. Increasing the striking distance makes the tone fatter, but may produce an unclear, muddy quality. Here, hammer weight is also an important factor. In the lower part of the scale, hammer contact time is small in comparison with the roundtrip time for the pulse on the string - from the striking point to the bridge and back again. Consequently, damping due to the hammers is small. Moving the striking points of the hammers changes the tone quality primarily by rearranging the relative amplitude of the partials. If the hammer should strike the string at a nodal point, or near, where the string motion is small, then the amplitude of the corresponding partial will also be small. Fig. 12 shows how the measured output of one particular string varied as the hammer striking ratio was changed. The graph shows partials 5 through 9. The amplitude of each partial passes through a distinct minimum point as the striking ratio is increased. If you were listening to the tone of the string you would hear obvious differences in timbre as the hammer striking distance was changed, and I am sure you would like the tone at certain d/L-values better than at others. Fig. 12 Output vs. striking ratio (d/L) for partials 5, 6, 7, 8, and 9.

Fig. 13 shows the instantaneous peak output spectrum for two different values of the hammer striking ratio. For d/L = 0.019 (1/53), the lower partials all have very small amplitude. This is because the hammer is striking almost at the very end of the string. For such a small d/L the tone sounds thin and weak. For a longer striking distance, d/L = 0.143 (1/7), the lower partials have gained in amplitude and the 7th partial is almost completely missing. At one time it was believed that the 7th and 9th partials were dissonant and ought to be eliminated by a proper choice of the striking distance. Personally, I do not believe that any string partial should be deliberately minimized.

Fig. 13. String spectra for short and long striking ratio (d/L = 0.019 and 0.143).

Soundboards
Fig. 14 shows the top of the soundboard of the 1720 Cristofori piano. The original soundboard was made of cypress wood, 3.5 mm thick, which may have come from Crete (Pollins 1984). The original soundboard was replaced in 1938 with what is said to be an accurate copy. The bottom of the soundboard can be seen in Fig. 15. In contrast, Fig. 16 shows a contemporary concert grand. Note that much of the contemporary soundboard is covered by the cast iron string plate. The soundboards of modern pianos usually range in thickness between 6.5 and 9.5 mm approximately. In the U.S.A., spruce, and particularly Sitka spruce, has been the favored soundboard material for high quality pianos.

Fig. 14 (left) Plan view of 1720 Cristofori (By permission of the Metropolitan Museum of Art). Fig. 15 (right) Bottom of 1720 Cristofori (By permission of the Metropolitan Museum of Art).

Fig. 16. Plan view of contemporary concert grand.

How does a soundboard vibrate?

Soundboards vibrate more readily at their modal or resonance frequencies than at other frequencies. The photos in Figs. 17 - 20 show how a piano soundboard vibrates at some of its modes (resonances). The lowest frequency at which a soundboard can vibrate strongly is called the first mode. In Fig. 17 we see an experiment in which a concert grand piano soundboard has been vibrated at its first mode. The vibration generator, the circular object that can be seen to the left in the photos, has been connected mechanically to the soundboard at a point near its edge. For such a test the procedure is the following: Before being vibrated the soundboard is covered uniformly all over its surface with a mixture of fine particles (in this case sand and "glitter"). Then the vibration generator is turned on and tuned slowly until its frequency coincides approximately with that of a soundboard mode, as will be indicated by a noticeable increase in sound level from the soundboard. Then the generator level is increased until the acceleration of the particles exceeds "1 g" (the acceleration of gravity, 9.8 m/s2) and the particles begin to dance on the soundboard. As they dance, the particles gradually collect in those areas that are not moving at all or are moving with minimum velocity. This produces a pattern called a Chladni figure, so named after the famous German physicist. The first mode of this soundboard occurred at 49 Hz. In this mode, it is the center of the soundboard that moves most violently; the edges, where you see most of the particles, nearly stand still. A piano soundboard rapidly loses its effectiveness as a sound radiator at frequencies below that of the first mode, so notes below the first modal frequency usually do not have very much energy in the first partial. In Figs. 18, 19, and 20, you can see how the soundboard moves at some of its other modes. Remember that the particles collect where the soundboard is moving least.

Fig. 17. First (lowest) soundboard mode at 49 Hz.

Fig. 18. Second mode at 67 Hz.

Fig. 19. Third mode at 89 Hz.

Fig. 20. Eighth mode at 184 Hz.

The modal frequencies are determined by many factors, the primary ones being the material, size and shape of the soundboard, its thickness and grain direction, and also the material, dimensions, and placement of its ribs. Secondary factors include the characteristics of the rim or case to which the soundboard is attached. In general, the thicker the soundboard, the louder the piano but the less the duration of its tone. Soundboard design is often a compromise. Today there is a better way, called modal analysis, to study the vibration of piano soundboards. Using this method, the soundboard is tapped with a special hammer that is fitted with a force transducer. An accelerometer attached to the soundboard responds to vibrations caused by the hammer and the force and acceleration signals are stored digitally. The tapping is repeated at a number of different preselected points on the soundboard, and after all the data has been taken, a computer analyzes it and identifies the modes (Suzuki 1986). With modern equipment it is possible to see an animated display of the modal motion of the soundboard on a TV-screen, a technique which will be described in more detail in the lecture by Klaus Wogram. Fig. 21 presents modal information in another way: it is a graph giving the velocity of motion of the soundboard at one particular point (for a constant driving force), as a function of frequency. This plotted quantity is called mobility, and is the reciprocal of mechanical impedance. Each of the large peaks you see in Fig. 21 corresponds to a particular soundboard mode like those earlier shown in the Chladni patterns.

Fig. 21. Driving point velocity vs. frequency for concert grand soundboard with no strings or plate (top), and with the piano fully assembled and tuned (bottom).

The frequency and shape of soundboard modes are affected by the strings and the cast-iron plate. For the graph at the top in Fig. 21, the plate and strings were removed from the piano. The lower graph in the same Fig. shows the mobility at the same point on the same soundboard with the strings and plate in place and the piano fully tuned. Notice how much the picture has changed: the first mode has shifted upward in frequency from 48 Hz to around 60 Hz, and the modal peaks are broader than before and not so high. In pianos of this size (concert grand) you can often identify the first mode by playing single notes up and down the scale. You may feel a slight increase in the vibration level of the case, usually around C2 - D2 (keys 16 - 18), and you may hear an increase in the sound level of the first partial. In order to do the analyses just mentioned we have to test an actual soundboard. So we need first to build a piano before we can measure it. But now, by still another new technique called finite element analysis (FEA), we can construct a model of the proposed soundboard with computer software. Then, using a computer, we can find out how the soundboard will move before we build the piano!

The varnish
The varnish on a piano soundboard does not have a significant effect on the tone, as far as I have been able to discover. However, the varnish has a very significant effect on the tuning stability of pianos. Without varnish on its soundboard a piano can go rather quickly out of tune if the humidity should change. This is because the size and weight of any piece of wood depends on the relative humidity and temperature of the air around it. Wood absorbs moisture until the amount in the wood is in equilibrium with the surrounding environment.

Fig. 22. Equilibrium moisture content of wood vs. relative humidity at 24o C.

Fig. 22 shows how much moisture will be in a piece of wood at equilibrium for different values of relative humidity at a temperature of 24o C (75o F) (see Wood Handbook 1987). The amount will be about the same for all species of wood. As you can see, at 50% relative humidity about 9% of the weight of the wood will be moisture. The varnish on a soundboard slows down the rate at which moisture can enter and leave the soundboard, and so lets the piano stay in tune longer. I once gave a copy of this graph to a friend who had just bought a new grand piano. A couple of days later he telephoned, sounding somewhat upset: it seems he had calculated that his new $30,000 piano contained 7.2 gallons (27 litres) of water and that each gallon had cost $229!

New materials
Why can't we use a material for soundboards that is not affected by humidity? Some early research on soundboard materials other than wood apparently was done here in Sweden by a man named Fridolf Frankel. Fig. 23 shows the cover of a booklet, written in English and dated 1923, describing a soundboard that is said to have been made of steel, 0.65 mm thick. I found the booklet in some old files of an American piano company. Despite the favoured testimonials for the performance of his pianos (Fig. 24), few instruments seem to have survived. Many years later, in 1961, the American harpsichord builder, John Challis, constructed a piano having a metal soundboard and bridge (Challis 1963). Such an instrument was demonstrated by the pianist Arthur Loesser at a concert in New York City in 1967 (Henahan 1967). An excerpt from a recording of this concert can be heard in sound example 3. Apparently, the Challis piano was not suitable for playing a wide range of standard piano literature, for even at this concert it was used by the performer only for a few 18th century pieces.

Fig. 23. Cover of Frankel booklet describing steel soundboard (1923).

Fig. 24. Testimonials for Frankel's pianos.

In 1969, a U.S. patent was issued to P.A. Bert describing a soundboard for pianos employing sandwich construction with a cellular core and plastic facings. I have been told that at least one such instrument was built, but so far as I know, they were not marketed. I personally believe strongly that researchers today have available better materials for piano soundboards than ever before, and that only diligent applied research is needed in order to produce the next significant improvement in the piano. When we try to "improve" the piano we must remember that we dare not change its essential character. If we do, it is almost certain to be rejected. Pianists who have spent years in learning to deal artfully with existing instruments quite naturally do not want to have to relearn their skills. They are incredibly sensitive to changes! I was present at one occasion on which an instrument of rather exotic construction was being tested. It looked like an ordinary piano and its sound was extremely pleasant, though slightly unusual. An excellent pianist was called in to give an opinion. He played at some length. To us in the audience the instrument sounded quite beautiful. Finally our pianist said it had a "nice sunny sound" and would be very good for Spanish piano music. After this encouraging initial response some minor changes were made and a rather more famous pianist was called in. This pianist reported that the instrument was good only for French piano music! Still further changes were made but this particular instrument never moved across the border and into Germany, musically speaking.

The piano case

As you can see from Figs. 25 and 26, a modern grand case is very substantially made. The rims of the best modern grand pianos are usually made from heavy hardwoods such as maple or beech, and may have a total thickness between 80 - 90 mm. The case for a piano of this size may weigh 150 - 200 kg. The acoustical benefit of this is that it provides a massive termination for the edges of the soundboard. This means that the vibrational energy will stay as much as possible in the soundboard instead of dissipating uselessly in the case parts, which are inefficient radiators of sound.

Fig. 25. Grand rim with keybed attached.

Fig. 26. Grand rim nearly completed.

Cristofori had a totally opposite idea about the soundboard as the sketch in Fig. 27 shows (Pollins 1984). His soundboard, 3.5 mm thick, was glued to an extra inner vertical case wall, only about 4 mm thick. This was mechanically decoupled from the main outer walls of the case. Cristofori must have felt that connecting the soundboard directly to the outer case would impede its vibration. Fig. 28 is a view of the underside of the 1720 piano with its bottom board removed. The large cross members are not ribs but rather stiffening members that are connected to the sides of the outer case.

Fig. 27. Cross section of Cristofori case. (By permission of the Journal of the American Musical Instrument Society).

Fig. 28. Close-up of underside of 1720 Cristofori piano. (By permission of the Journal of the American Musical Instrument Society).

The cast-iron plate

Piano makers gradually learned that pianos could be made louder by increasing the weight and tension of the strings with the result that wooden frames soon became inadequate to support the increased stresses. Until around the beginning of the 19th century the loadbearing structure of pianos was made entirely of wood. By the end of that same century almost all pianos had cast-iron string plates. The metal plate (Fig. 29) brought improved tuning stability and, at least to most modern ears, better tone.

Fig. 29. View of contemporary concert grand plate.

The need for a stronger supporting structure for the strings is clearly indicated by Fig. 30, which shows the pulling force per string (tension) and the total string pull, calculated for the 1720 Cristofori and for a contemporary concert grand. The average pull of the strings of the Cristofori is only about 70 N (16 lbf, "pound force"), versus about 830 N (190 lbf) for the contemporary piano. The total string load is roughly 7500 N (1700 lbf) for the Cristofori, compared with about 210 000 N (47 000 lbf) for the modern concert grand.

Fig. 30. String pull for 9-ft (274 cm) contemporary grand and for 1720 Cristofori.

The plate must be strong enough not to break under the load of the strings, and it should also be stiff enough to provide good tuning stability. Beyond this, the design of the plate affects the tone of the instrument in many less obvious ways, of which only their general direction will be indicated. In any stringed instrument the speaking length of each string has two ends. In a piano, one end is connected, via the bridge, to the soundboard, which is expected to radiate sound efficiently. The other end is always connected in some way to the frame of the instrument, which is invariably an inefficient radiator of sound. In modern pianos the forward string termination (the agraffe) is located on the iron string plate, as shown in Fig. 31 This part of the plate should be designed so as not to steal energy away from the strings and the soundboard. The plate should not vibrate appreciably at string frequencies as the piano is being played. Generally, this requires that the plate be rather massive. Plates of concert grand size may weigh 160 - 180 kg.

Fig. 31. Close-up of forward termination (agraffe panel) of modern grand.

Strings
The fundamental frequency of a stretched string is given by the expression below, known since the beginning of the 17th century, in which L is the length of the string, T is its tension or pull, and M is its mass per unit length.

The fundamental frequencies of a modern piano are known in advance because A4 has a standard frequency (440 Hz) and because the frequencies of adjacent notes all across the equally tempered scale have a ratio equal to the twelfth root of 2, about 1.05946. The most common calculation in designing a piano scale is the tension, not the frequency of the string. (The scale is the distribution of the string lengths and gauges over the compass of the instrument). The above expression can easily be rearranged to give the tension. Also, a factor (F) can be added to the formula above to allow for the use of wrapped strings, and a

changeable constant (k) may be employed in order to permit calculation for any string material in any system of units. Then the tension of a string may be written as:

In this formula, dc is the string diameter (or the core wire diameter, in the case of a wrapped string). F is unity (1) for a plain string but has some larger positive value for a wrapped string. For a string of steel music wire for which the length and diameter are given in centimeters, the tension will be given in Newtons (N) for k = 4096. I show this formula to make a certain point: there's a lot of multiplying here! In the old days it could take a long time to calculate just one string. What if you had to do this for 88 different strings with only pencil and paper? Much in early piano design obviously was empirical. Empiricism seems to have persisted for longer than one might expect. Would you believe that even in the 20th century, piano designers still didn't know how to calculate the tension of a wrapped string, and had to find it by actual measurement. (Wrapping the core of a string helically with turns of another wire is a very old method, still in use, to make a bass string heavy without having its core too thick and thereby too stiff to produce a good tone.) The following sentence appears in a supplement, dated 1927, to Wolfenden's well-known book about piano design: "It is remarkable that, at this date, after spun strings have been in use for, say, a matter of two centuries, neither in this country nor any other, as far as many enquiries have shown, is there in trade use, a method by which the tensional stress upon a spun string, tuned to a given pitch, can be approximately ascertained" (Wolfenden 1927). Equations for calculating the pull of wrapped piano strings are now well known to at least some piano manufacturers and also to many piano technicians. Also, with computers we can calculate piano strings and scales very accurately and much more quickly than ever before.

Longitudinal string modes

Of course there is more to designing good piano scales than merely calculating the tension of the strings. In 1967, I applied for a patent under the heading, "Longitudinal Mode Tuning of Stringed Instruments" (Conklin 1970). I found the technique outlined in this patent to be such a powerful tool in scale design work, especially in the design of wrapped strings, that today I would not consider designing a piano without it. In longitudinal modes of vibration, energy propagates lengthwise along the string (as periodic compressions of the string material) without sidewise (transverse) motion of the string. Longitudinal and transverse vibrations of a piano string can occur simultaneously. However, the lowest-frequency longitudinal mode of a piano string is always more than ten times the frequency of the lowest-frequency transverse mode.

It has long been known that the strings of pianos and other musical instruments can have longitudinal modes of vibration (Rayleigh 1877, Knoblaugh 1944, Leipp 1969). My patent simply teaches what the designer should do about it in order to make the best sounding instrument. I learned the importance of the longitudinal mode by accident; one day, while I was installing a new string, I noticed that the string sounded better when it was tuned to the wrong frequency! After some study it became apparent that the reason had to do with the longitudinal mode. The first longitudinal mode of a piano string normally occurs at a frequency somewhere in the range between 3 octaves plus a fifth and 4 octaves plus a third above the "normal" fundamental transverse frequency of a string. This range is determined by certain design constraints related to the properties of piano wire that are common to all present-day pianos. A piano tuner tunes the transverse, or flexural, modes of the strings by changing the tension of the strings as he turns the tuning pins. A piano tuner can do nothing to affect the frequency of the longitudinal mode because turning the tuning pins doesn't change it. The longitudinal frequency of a plain steel string in a piano can be changed only by altering its speaking length. In the case of wrapped piano strings, the longitudinal mode can be tuned only in two ways: either by changing the speaking length or by changing the weight of the wrapping wire in relation to the weight of the core wire. So, the tuning of the longitudinal mode is established, either deliberately or accidentally, by the designer of the piano; and, as a practical matter, it cannot be changed after the piano has been built. In designing a piano nowadays, it is possible to tune the longitudinal modes of its strings to those frequencies that will make the piano sound best. In sound example 4, you can hear what kinds of changes in the timbre of piano tones are produced by changing the tuning of the longitudinal mode. I think you will agree that each string sounds different from the others. However, all those six strings were tuned to the same transverse frequency by a piano tuner: all the notes were low G1 (key 11) on the piano! They sounded different because each had a different tuning of the longitudinal mode. In sound example 5, you can hear a little tune which is known as "Yankee Doodle". The tune is played in two different keys. I am sure you will recognize that a tune is being played. However, the tune was played on strings that were all tuned to the same transverse frequency! The tunes could be heard because each string was designed so that its longitudinal mode differed in frequency by a semi-tone (100 cents) from that of the preceding string. (The common transverse frequency was not the same for the two versions in different keys.) Next, listen to some chords, each chord followed by a bass note having a different tuning of the longitudinal mode but the same tuning of the transverse mode (sound example 6). As I hope you can hear, the longitudinal mode is important in determining the tone color of the bass and tenor regions of the piano. The longitudinal mode creates a formant-like emphasis in the tone at its own frequency, with the result that some tunings of the longitudinal mode sound much better than others. In particular, it is desirable to have the longitudinal mode tuned so that it blends harmoniously with the tone from the transverse

modes. This can be achieved by careful and deliberate choices in the design of the strings and scale of the instrument. In the examples you have heard so far the longitudinal mode was deliberately tuned at intervals of a certain number of semi-tones with reference to the fundamental transverse mode. Strange and undesirable things can happen to the tone if the longitudinal mode is ignored or left to chance by the designer. Next you will hear some scales played on two different pianos. The first piano has the longitudinal mode tuned by design, the second one does not. As you can hear, the piano having deliberately tuned longitudinal modes has a much more uniform and pleasing voice through the scale (sound example 7). Physicists may want to know if it is possible to measure what we are hearing. Fig. 32 is an acoustical spectrogram of piano note E1, with a fundamental frequency of about 41 Hz. The "normal" transverse partials are identified by small dots near each peak. The longitudinal mode can be seen between the 14th and 15th partials and is about 20 decibels lower in level than the neighboring partials (Podlesack & Lee 1988).

Fig. 32. Spectrum of piano note E1 (41 Hz) showing longitudinal mode (indicated by the vertical line at about 600 Hz).

Machines similar to those shown in Fig. 33 have been used for a long time to wind the copper covering wire onto wrapped piano strings. With these machines the characteristics of the finished strings are strongly dependent upon the technique of the operator. Operator technique varies, not only from person to person, but also from string to string. It was found to be impossible with such machines to control the tuning of the longitudinal mode precisely enough so that successive strings could be accurately tuned and alike in tone. For this reason, it was necessary to design a new type of machine, also patented, with which the characteristics of the finished string would be independent of the operator. With this machine, optimum settings are predetermined for each type of string.

Fig. 33. Old string machines.

The tuning pins

One thing about pianos has hardly changed at all in the 268 years since the 1720 Cristofori was built: the tuning pins! They are still small metal cylinders that are driven into holes bored in a slab of wood. As all piano technicians know, tuning pins can have various problems that interfere with accurate tuning of the instrument. I am very pleased to announce - and this is the first public announcement - that I have devised a new type of tuning pin that seems to eliminate problems encountered with conventional tuning pins. My tests indicate that the new pin will make tuning easier, faster, and more accurate. Because of patent considerations, I cannot yet describe it to you but before long I hope to be able to convince piano manufacturers to use it!

References
Benade, A.H. (1976): Fundamentals of Music Acoustics (Oxford University Press, London) pp. 339-343. Boutillon, X. (1988): "Model for piano hammers: Experimental determination and digital simulation," J. Acoust. Soc. Am. 83, pp. 746-754. Briggs, G.A. (1951): Pianos, Pianists and Sonics (Wharfdale Wireless Works, Bradford Rd., Bradford, Yorks) p. 37. Brinsmead, Edgar (1879): The History of the Pianoforte (reissued by Singing Tree Press, Detroit, orig. pub. by Novello, Ewer and Co., London, 1879) p. 47. Challis, John (1963): "New: A 20th Century Piano," American Music Teacher, Jul-Aug 1963, p. 20. Conklin, Harold A. Jr. (1970): U.S. Pat. 3,523,480, "Longitudinal Mode Tuning of Stringed Instruments," Aug. 11, 1970. Good, Edwin M. (1982): Giraffes, Black Dragons, and other Pianos (Stanford University Press) p. 9. Hall, D.E. (1986): "Piano string excitation in the case of small hammer mass," J. Acoust. Soc. Am. 79, pp. 141-147.

Hall, D.E. (1987a): "Piano string excitation II: General solution for a hard narrow hammer," J. Acoust. Soc. Am. 81, pp. 535-546. Hall, D.E. (1987b): "Piano string excitation III: General solution for a soft narrow hammer," J. Acoust. Soc. Am. 81, pp. 547-555. Hall, D.E. & Clark, P.J. (1987): "Piano string excitation IV: The question of missing modes," J. Acoust. Soc. Am. 82, pp. 1913-1918. Hall, D.E. and Askenfelt, A. (1988): "Piano string excitation V: Spectra for real hammers and strings," J. Acoust. Soc. Am. 83, pp. 1627-1637. Hansing, Siegfried (1888): Das Pianoforte in seinen akustischen Anlagen (New York), p. 92. Harding, Rosamond E.M. (1933): The Piano-forte - Its History Traced to the Great Exhibition of 1851 (Da Capo Press, New York 1973, reprint of 1933 edition), pp. 64-66. Helmholtz, Hermann L.F. (1863): On the Sensations of Tone as a Physiological Basis for the Theory of Music (Dover Publications Inc., New York 1954, first ed. Friedr. Vieweg & Sohn, Braunschweig 1863) p. 77. Henahan, Donal (1967): "Loesser, Pianist, Exhumes 'Ghosts' to Mark Halloween," (a review of the concert), New York Times, Oct. 3, 1967. Knoblaugh, Armond F. (1944): "The clang tone of the pianoforte," J. Acoust. Soc. Am., 19, p. 102. Leipp, Emile (1969): The Violin (University of Toronto Press), pp. 97-99. Marcuse, Sibyl (1975): Musical Instruments, A Comprehensive Dictionary (W.W. Norton & Co., New York), p. 405. McFerrin, W.V. (1972): The Piano - Its Acoustics (Tuners Supply Co., Boston), p. 84. Ortman, Otto (1925): The Physical Basis of Piano Touch and Tone (E.P. Dutton & Co., New York), p. 96. Podlesak, M. & Lee, A.R. (1988): "Dispersion of waves in piano strings," J. Acoust. Soc. Am. 83, pp. 305-317. Pollins, Stewart (1984): "The Pianos of Bartolomeo Cristofori," J. Amer. Musical Instrument Soc. 10, pp. 32-68. Rayleigh Lord (Strutt, John William, 1877): The Theory of Sound (MacMillan & Co., Ltd., London, reprinted by Dover Publications, Inc., New York) Vol. I, p. 252 (a quotation referring to longitudinal mode of piano and violin strings taken from p. 154 of Donkin's Acoustics). Suzuki, H. (1986): "Vibration and sound radiation of a piano soundboard," J. Acoust. Soc. Am. 80, pp. 1573-1588. Vant, Albert (1927): Piano Scale Making (pub. by Albert V. Vant, 543 Academy St., New York), p. 30. White, William Braid (1906): Theory and Practice of Pianoforte Building (Edward Lyman Bill, New York, reprinted by Dover Publications, Inc., New York 1975) p. 34. White, William Braid (1946): Piano Tuning and Allied Arts (Tuners Supply Co., Boston) 5th ed., p. 45. Wood, Alexander (1944): The Physics of Music (Metheun & Co., Ltd., London) p. 94. Wolfenden, Samuel (1916): A Treatise on the Art of Pianoforte Construction (original ed. 1916, reprinted by Unwin Bros., Ltd, Old Working, Surrey, U.K., 1975), pp. 50-51. Wolfenden, Samuel (1927): Supplement to A Treatise on the Art of Pianoforte Construction (orig. ed. 1927, reprinted by Unwin Bros., Ltd., Old Working, Surrey, U.K., 1975), p. 208. Wood Handbook (1987): Wood as an Engineering Material (Forest Products Laboratory, United States Department of Agriculture, Washington D.C.), Chap. 3, pp. 9-11.

From touch to string vibration

Anders Askenfelt & Erik Jansson

Introduction
This lecture will present a series of experiments exploring the initial stages of the sound production in the piano - beginning with the motion of the key and ending with the string vibrations. This chain of events is closely connected with the performance of the pianist, who, by depressing the key, sets the parts of the action in motion, which eventually causes the hammer to strike the strings. In contrast to performers on other string instruments, like the violinist or the guitar player, the pianist could be said to only have an indirect control of the string excitation. Using the computer biased terminology of today, it is tempting to call the action an "interface" between the pianist and the string. This interface is an interconnecting device, which at the input end (the keys) is particularly adapted to the soft and sensitive fingers of the pianist, while the output end is equipped with hard felt hammers, capable of exciting even the thickest of the tense piano strings vigorously. The function of this "interface" in playing is by no means simple. We will illustrate some important properties of the action by presenting measurements of the timing in the action under different conditions, and also show how the motions of the key and hammer change, depending on how the key is depressed. Furthermore, the resulting string vibrations will be closely examined and the manufacturer's, the piano technician's and the pianist's influence on the spectrum of the piano tone will be compared. During the presentation it will successively become clear that the successful piano performer is accompanied by two mostly anonymous artists, the piano technician and the tuner, sometimes but far from always combined in one and the same person. In contrast to a widespread belief, the fact is that it is not sufficient to have the piano tuned prior to the performance (in such a way that it stays in tune during the entire concert); the piano must also be properly regulated in order to play well. Despite the remote control of the actual string excitation by the hammer impact, pianists pay great attention to the way the key is depressed. Often the term "touch" is used to denote this process. Physicists and piano players have had contrasting views regarding the importance of this point for a long period of time, and later on we will try to add some material regarding this question. However, we must hasten to add that at the moment we will not be able to resolve this conflict, but perhaps we can indicate in which directions the answers can be sought.

Timing in the action

First of all, let us present a rather detailed description of the function of the grand piano action. The actions of all grand pianos of today are in principle identical, and the small differences which do remain are limited to the design of the individual parts.

Fig. 1. View of the action of a modern grand piano (Steinway & Sons). The shaded areas indicate felt and the broad lines indicate leather.

Principally, the action consists of four major parts: the key, the lever body with appurtenant parts, the hammer and the damper (see Fig. 1). The successive steps in the operation of the action during a blow is illustrated in Fig. 2.

2(a) Rest position. The hammer rests via the hammer roller on the spring-supported repetition lever, a part of the lever body. The lever body stands on the key, supported by the capstan screw. The weight of the hammer and the lever body holds the playing end of the key in its upper position. The damper rests on the string, pulled down by lead weights.

2(b) Acceleration. When the pianist depresses the key, the lever body is rotated upwards. The jack, mounted on the lever body, pushes the roller and accelerates the hammer. The damper is lifted off the string by the inner end of the key.

2(c) Let-off. The tail end of the jack is stopped by the escapement dolly, and the top of the jack is rotated away from the hammer roller. The hammer, which now is free, continues towards the string. The repetition lever is stopped in waiting position by the drop screw.

2(d) Check. The rebounding hammer falls with the hammer roller on the repetition lever, in front of the tripped jack, before it is captured at the tail of the hammer head by the check. The stroke may now be repeated, either by releasing the key as usual, or by using the double-repetition feature (see text).

The action of the grand piano features a special construction for fast repetitions, the doublerepetition mechanism, not incorporated in the action of the upright piano. In order to use the double-repetition feature, the key is let up only about a third of its travel after a stroke. At this stage, the hammer has been released from the check and lifted slightly by the springsupported repetition lever (cf. Fig. 2 d). This allows the spring-loaded jack to slip back into its initial position under the roller, and the action is set for a second blow. The doublerepetition mechanism enables very fast repetitions on the same key, without the damper touching the string between notes. A correct function of the action requires a careful regulation. Of crucial importance is the distance between the top of the hammer at rest and the string, in the following hammerstring distance (piano technicians term: "blow level"). This distance is adjusted with the capstan screw (typical value 45 - 47 mm). Of equal importance is the setting of the release of the jack ("let-off"). This is adjusted with the escapement dolly. The adjustment is made by observing the distance between the string and the top of the hammer at the highest point of its travel (let-off distance), when the key is depressed slowly. The let-off distance is typically set between 1 and 3 mm, the actual value depending on such factors as the diameter of the string, interval between regulations, and sometimes, the personal taste of the pianist. In all contact points between moving parts, one of the surfaces is covered with felt or leather in order to ensure a smooth and silent motion, free from backlash. In particular, thin shafts with close tolerances, for example the shaft for the hammer shank in the flange, are mounted in bushings of high-quality felt. The combination of wood and felt parts means that the action will change condition not only because of wear, but also due to changes in temperature and humidity. Periodic regulation is thus necessary in order to keep the instrument in optimum condition.

Measuring the timing

We measured the timing in the action electrically by a network of "micro-switches," integrated with the action. These switches consisted of copper foils and thin copper wires glued and sewn to the contact surfaces. During a blow, the different switches turned off and on as the parts moved, and a stepwise signal was generated at the output of the network. In order to obtain blows which could be repeated with a high degree of reproducibility in the timing experiments, we also had to develop "a mechanical pianist." It turned out that for this particular purpose a long pendulum was ideal as a substitute for the player.

Fig. 3. Overview timing diagram of the grand action for a "staccato-touch," forte, C4. The string vibrations are included for reference. Note the time scale (20 ms/div), and the comparison with the duration of the sixteenth-note in andante (M.M. = 107 beats/min). A typical example of a timing diagram for a "staccato-note" - a short note played with the finger striking the key from some distance above the key - is shown in Fig. 3. First of all, it can be noted that the processes which take place in the action are very fast. While the duration of the rather short note in this example is about a 1/10th of a second (100 ms), the contact between the hammer and string lasts only a few thousandths of a second (1/1000 s = 1 millisecond = 1 ms). The contact with the string is preceded by a long history, during which different parts are engaged in turn, but it is only during these few milliseconds energy is transferred into the string. Listing the most important steps chronologically, starting with the player's finger on the key, we see that the damper is lifted off the string plenty of time (15 ms) before the hammer strikes the string. Much later, only a few milliseconds before hammer-string contact, the tail end of the jack reaches the escapement dolly, and the top of the jack starts to withdraw from the hammer roller. Just before (less than 1 ms) the impact on the string, the jack loses contact with the roller, and the hammer swings freely during the final part of its travel. At about the same moment the key happens to reach its bottom position. The height of this chain of events, the hammer-string contact, occupies only about 2 ms. Almost immediately after the hammer-string contact has ceased, the hammer returns, first making contact with the repetition lever before it is checked. If the key were held down after the stroke, the action would remain in this state while the note decayed undisturbed. In this example, however, the key is immediately released ("staccato"), and the parts of the action revert to their initial positions. The note is terminated when the damper falls down on the string approximately 80 ms after hammer-string contact and brings the string to rest after some bouncing.

Key bottom and hammerstring contact

A close-up of the events at string contact at three dynamic levels shows two interesting facts (see Fig. 4). One observation is of special interest for the pianist. It turns out that at a certain dynamic level, about mezzo forte, the key reaches its bottom position at the same moment as the hammer strikes the string, while at other dynamic levels this synchrony is dissolved. For those of us who don't play the instrument, this observation may not seem very interesting, but every piano player can tell that the bottom feeling of the key is an important feedback from instrument.

Fig. 4. Influence of dynamic level (p - mf - f) on the timing in the action. The dashed line shows how the moment of key bottom contact is advanced relative to the moment of hammerstring contact, as the dynamic level is raised. The dotted line indicates the shortening of hammerstring contact duration with rising level.

The relation between the moments of key bottom contact and hammer-string contact shows a smooth progression with dynamic level (see Fig. 5). At a soft level (piano) the bottom contact is delayed compared to the hammer-string contact, and at louder levels, the contact is advanced - at forte it occurs even before the hammer strikes the string(s). Interestingly, the magnitude of this shift in timing with level was about the same for an untrained subject as for a professional pianist. This indicates that the phenomenon probably is a combination of the inherent properties of the action and general differences between the ways the key is depressed in forte and piano, respectively ("stroke" vs. "press").

Fig. 5. Influence of dynamic level on the timing relation between key bottom contact and hammer-string contact. An untrained subject (unfilled symbols) and a professional pianist (filled symbols) are seen to perform almost identically. The dots indicate notes played with very unusual types of touch. The dynamic span ranges from pp to ff, corresponding to 33 dB.

One may wonder if the pianist is aware of this variable "delay" between the onset of the note (the hammer strikes the string) and the response in the finger (the key reaches its bottom position). Probably not, even though the delay can reach an order of magnitude (about 20 ms) that it likely affects playing at an artistic level. Presumably, the "bottom feeling" and a compensation for the delay are unconsciously developed during the years of study. If the piano is not properly regulated, this timing property will be upset. Here, the relevant parameter is the hammer-string distance. A longer hammer-string distance than the nominal value - a normal result of wear - will advance the key bottom contact. An increase in hammer-string distance from a normal 47 mm to 50 mm will result in a shift in the delay comparable to a change in level from mezzo forte to piano (cf. Fig. 5). Most probably a pianist will feel uncomfortable when facing a piano with such an "out-of-tune" regulation, especially if she is accustomed to proper conditions. And, even worse, if the regulation and hence the corresponding delays happens to vary between keys, the artist's performance is likely to be disturbed. While we are discussing the piano technician's regulation we must also shortly mention the influence of changing the let-off distance. This adjustment determines mainly how long the free flight of the hammer before string contact will last. A very short setting of the let-off (< 1 mm in the midrange and treble) gives the pianist close control of the hammer motion almost up to string level, and there is only a very short interval (< 1 ms) between the moment when the hammer leaves the jack and the moment when it hits the string. Such a close setting is preferred by pianists as it enables a delicate pianissimo-playing under strict control, and it is also claimed to facilitate a powerful fortissimo. Unfortunately, the instrument cannot be left without inspection for any longer period of time in this condition, so the average piano player (and owner) will have to resort to a somewhat more moderate regulation.

String contact duration and dynamic level

A second interesting observation that can made from Fig. 4 is that the duration of the hammer-string contact decreases as the dynamic level is raised. This is due to a peculiar property of the piano hammer, called nonlinear stiffness. In effect, this nonlinearity makes the hammer harder, the higher the impact velocity. The details of this effect which is of extreme importance for a proper tone generation, is explained in the lecture by Donald Hall. At the moment it may suffice to know that the duration of the hammer-string contact time is one of the major parameters in determining the spectral content. It can be shown that a shorter contact duration increases the amount of high frequency partials in the spectrum. The graph in Fig. 6 illustrates how the contact duration decreases with rising level for a note in the midrange, indicating a boost of the high-frequency partials when the dynamic level grows from piano to forte. The variation in contact duration within a comfortably accessible dynamic range (p - ff) is about +/- 20 % compared to mezzo forte. Later in this presentation we will return to measurements of the spectral changes which follow from changes in dynamic level.

Fig. 6. Illustration of the how string contact duration decreases with rising dynamic level (C4). A touch by a pendulum (mf) is represented by an unfilled square. The dashed lines represent the range in contact duration covered in a comfortably accessible dynamic span from p to ff.

Given a fixed dynamic level, the contact duration is determined by such design factors as the ratio between the mass of the hammer and the string(s) and the striking point along the string. As these parameters vary over the compass of the piano, the contact durations will change slowly from bass to treble. In Fig. 7, the contact durations are sampled by one note per octave, and as we can see the contact durations decrease from about 4 ms in the bass to less than 1 ms in the highest treble. Not surprisingly, the treble could thus be assumed to contain spectral components with higher frequencies than the bass.

Fig. 7. Hammer-string contact durations expressed in milliseconds, illustrating how the contact durations decrease from bass to treble. The bars indicate the range in contact duration between a blow in ff (left end) and pp (right end). The vertical line in each bar represents a blow by a pendulum at mezzo forte. The solid curve marked To/2 represents half a period of the fundamental.

Fig. 8. The same hammer-string contact durations as in Fig. 7, but now expressed in percent of half a period of the fundamental (To/2). Note that in this relative measure, the durations are short in the bass and long in the treble.

An entirely different picture develops if the contact durations are related to the period time of the fundamental (the inverse of the fundamental frequency) of the corresponding note (see Fig. 8). Now it can be seen that the contact durations are short in the bass in comparison with the fundamental period, while the opposite situation prevails in the treble. A general theorem in string physics helps us to interpret these differences. The theorem tells us that string resonances with period times essentially shorter than the contact duration will be only weakly excited by the blow. The implication of this theorem for the sound quality of the piano is that the bass notes will be rich in partials compared to the treble notes. This is so, because in the bass the contact duration is only about 10 % of the fundamental period; it is not until the 10th partial that we reach a period time that is short in comparison with the contact duration. In the treble on the other hand, even the fundamental period is much too short to be optimally excited, and the situation for the higher partials becomes successively worse. An alternative design of the piano could be imagined, such that the string contact durations were a constant fraction of the fundamental period over the entire compass. This would demand a rescaling of the striking points, much heavier bass hammers than "normal," and also lighter treble hammers. However, the sound would probably be rated as second to the present design, being too muddy in the bass and too brilliant in the treble.

In this case the customers are likely to have good acoustical reasons to reject the new design. But even in the case of smaller changes which would be welcomed by the manufacturer, the customers are notoriously hard to convince about the benefit of news. Tradition bears a strong power on music and musical instruments. Imagine a traditional piano concerto being played on anything but a black grand piano!

The motions of the key and the hammer

Leaving the timing aspect of the processes in the action, we now turn to the actual motions of the key and hammer. These motions were measured by means of an inexpensive optical system, developed with an ordinary reflex camera as a central part. A position sensitive photodetector which replaced the film in the camera registered the positions of small light sources on the key and hammer.

Key motion
The motion of the key turned out to be very different for different types of touch. In a touch with the finger striking the key from some distance above ("staccato"), the key normally made a temporary "stop" during its descent (see Fig. 9, left). This is seen as a plateau in the key position curve at about a third of the key travel, and more clearly in the key velocity curve which oscillates up-and-down due to the temporary retardation. This strange motion of the key was found to be connected with a resonance in the hammer. The reason is that during a "staccato-touch" a strong, initial impulse is delivered to the base of the hammer shank via the jack and roller. This impulse sets the hammer in vibration, the relatively heavy hammer head oscillating up-and-down on the flexible hammer shank. The bending motion of the hammer shank is partly transferred back via the action to key, where it can be measured and probably also perceived by the pianist as some kind of "response" from the instrument. In a touch with the finger initially resting on the key ("legato"), the motion always followed a smooth path. A typical example is shown in Fig. 9 (right). The key velocities are generally rather low. At mezzo forte (cf. Fig. 9), the maximum velocities are approximately 0.3 - 0.5 m/s. Even in forte the peak velocity does seldom exceed 1 m/s (about 4 km/h).

Fig. 9. Influence of different types of touch on the key motion (mezzo forte, C4). The two cases represent a staccato-touch with the finger starting its motion from some distance above the key (left), and a legato-touch with the finger initially resting on the key (right).

Hammer motion
The hammer must travel a distance which is approximately five times longer than the travel of the key in essentially the same time. Consequently, the hammer velocities are about five times higher than the key velocities. Two typical examples are given in Fig. 10 which shows a mezzo forte and forte touch, respectively. In the forte example, the maximum hammer velocity is about 5 m/s (18 km/h), not very far from the highest velocity observed during the experiments. Observe the short moment of hammer-string contact indicated in the figure, and also that the hammer is checked at a lower level in forte because of the higher return velocity when rebounding from the string.

Fig. 10. Typical registrations of hammer position and velocity at mezzo forte and forte (C4). The horizontal line indicates the level of the string. Note the short moment of hammer-string contact and the differences in check level.

Depending on the touch, the motion of the hammer can be quite different, even when the dynamic level is kept constant. The examples in Fig. 11 show the hammer velocity and acceleration for three different types of touch as measured at the hammer head. Indeed, the motion of the hammer before string contact is different in the three examples, despite the fact that the final velocities were approximately the same. At the moment it may suffice to note that the hammer motion seems to involve two components of oscillation, which are excited differently depending on the touch. One is a slow "backwash" motion at about 50 Hz and the other is a "ripple" type motion at about 400 Hz. While the slow motion is more prominent in the gentle types of touch, the "ripple" motion seems to be a characteristic of the vigorous, impulse-like types of touch. The perceptual differences between these notes were only rated informally. The pianist described the differences in tone character as large, but the experimenters on the other hand, had the impression that the differences were rather subtle. A formal evaluation would need a reliable listening test according to one of the recognized (and elaborate) methods prescribed by music psychologists, including a close control of the dynamic level.

Fig. 11. Influence of different types of touch on the hammer motion (acceleration and velocity) mezzo forte, B3. The pianist described the types as: "middle finger only" (top), "heavy arm with relaxed finger" (middle), "heavy arm with strained finger" (bottom).

Hammer resonances
In order to verify that the observed hammer vibrations were not an isolated characteristic of the particular piano used in the experiment, we also investigated hammers from an other instrument. Indeed, the same types of resonances were observed, and in fact, the "ripple" mode was more developed in the piano rated as superior of the two instruments (see Fig. 12). The vibrations were observed both before and after string contact, naturally more vividly so after the violent collision.

Fig. 12. Comparison of the motions (accelerations) of the hammer head for two grand pianos, showing two components of vibration; one slow ("backwash"), and a faster ("ripple"). Note that the "ripple" does not develop for a hammer with a slightly loose hammer head (bottom). Piano A was rated superior to piano B. "Staccato-touch" in forte, C4.

Having observed these differences between the two hammers from the eminent vs. the medium quality piano, our interest in the hammer properties and their influence on tone quality was aroused. The next observation, that a loose hammer head, which always is connected with a bad tone quality, prevented the ripple mode vibrations from developing (Fig. 12, bottom), gave further indications that these vibrations are in some way connected with the tone generation in the piano. The same conclusion could have been reached by asking an experienced piano technician, who would tell you that in traditional manufacturing, the hammer shanks are assorted according to their tap tone - the resonance frequency as heard when the shank is dropped on the bench - before the hammer heads are glued to the shanks. Shanks with high tap tones are used in the treble and vice versa. Quite correctly, of the hammers of the experimental piano, a bass hammer exhibited a lower "ripple" mode frequency (-15%) than the middle C hammer, while the corresponding frequency for a treble hammer was much higher (+80%).

Fig. 13. Approximate shapes and frequencies of two hammer resonances during acceleration towards the string; "backwash" (top) and "ripple" (bottom).

The hammer resonances were further investigated by a technique called modal analysis, which calculates the motion of the hammer at the resonance frequencies. These measurements indicated that during the upward acceleration while the hammer is supported by the jack under the roller, the hammer exhibits a series of resonances, of which two are shown in Fig. 13. At the lowest resonance at approximately 50 Hz, the hammer shank flexes over its entire length with the hammer head as a mass load at the end. This is the resonance corresponding to the "backwash" earlier discussed in connection with the key motion. At the next resonance at 250 Hz - corresponding to the "ripple" mode - the hammer shank shows a higher order bending resonance. Interestingly, at this resonance the top of the hammer head vibrates back and forth in the string direction.

The pianist and the touch

In trying to connect the observations of the hammer vibrations with the tone generation, we automatically return to the question of the pianist's "touch" and the conflicting opinions of physicists and pianists hinted at in the introduction. Before trying to answer, we must first clarify the question. Physicists usually claim that as the mechanical contact between the key and the hammer is broken before the hammer strikes the string, the pianist can only influence the final velocity of the hammer and thereby the loudness, and nothing more. Many pianists on the other hand, claim that important shadings in the character of the notes can be achieved by applying different types of touch, and that such a skill is an important component of the art of piano playing. First of all it is important to realize that the term "touch" probably is used with several different meanings. If a pianist is said to have a "beautiful touch," it may refer to the way a melody part is lifted above an accompaniment, or how certain notes in a chord are emphasized. This seems mainly to be a question of the timing and strength of certain notes relative to other notes, factors which have been shown to separate the artist from the amateur. But in certain connections, a "beautiful" or a "bad" touch can refer to the character of a single note at a given dynamic level, a topic which has interested prominent pedagogues and even created tensions between pianists of different schools. Surprisingly, a strong candidate for part of the answer has nothing to do with the normal string motion. It is probably so that a characteristic percussive component ("thump") at the onset of the note plays a decisive role for the character of the piano tone. This "thump" is generated by the key as it hits the stop rail on the key frame. The impact shock excites the keybed (the supporting surface under the action), and partly also the soundboard and iron frame. The significance of the "thump" sound is illustrated in sound example 8, in which the normal airborne sound of a grand piano and the sound of the string vibrations in isolation can be compared. The listener will probably agree that the normal piano sound as recorded in the room has a certain resemblance with the sounds in a blacksmith's shop. The string sound component on the other hand, lacks something of the interesting piano character, resembling a plucked string more closely than a struck one. Once these components have been identified, they are usually easy to distinguish in all piano tones.

The "thump-component" is undoubtedly excited differently depending on the touch and could be assumed to be characteristic of a pianist's way of playing. The importance of this component of the piano sound is further illustrated by knowing that the recognized piano manufacturers select the wood for the keybed with great care in order to achieve the right "thump" quality. Nothing definite can be said yet about the pianist's ability to influence the tone quality by controlling the motion of the hammer It is true that the entire history of the hammer motion during acceleration can be very different depending on the way the key is depressed (cf. Fig. 11). But this observation does not automatically imply that the interaction between hammer and string is influenced in some way. As mentioned, the hammer is flying freely the final distance before string contact. However, we have recently observed that the motion of the freely flying hammer can be somewhat different for slightly different types of touch (see Fig. 14). This is due to the "ripple" resonance, which is seen to influence the hammer motion during the last milliseconds before string contact, i.e. after "let-off." In this way, it is possible that a hammer resonance could serve as a "memory" of the history of the hammer motion. This would allow the pianist to have at least an indirect influence on the very final part of the hammer motion, even after mechanical contact between key and hammer has ceased.

Fig. 14. Comparison of the vertical motion of the hammer head before string contact for slightly different types of touch and regulations of the let-off distance ("staccatotouch," mezzo forte, C4). The dashed lines indicate the moment of contact between jack and escapement dolly ("let-off" begins). Note that the motion of the hammer head can be influenced by the "ripple" mode during the short interval between this line and string contact. The curve at the top corresponds to a longer setting of the letoff distance (3 mm) than normal.

However, we still have to be very careful in the interpretation of these results. Although the observations suggest that the hammer can make a rubbing motion against the string during contact, it has not yet been shown that such a motion takes place. Even if it does, it remains to be shown that a motion of this kind could influence the string excitation and hence the tone quality. In summarizing the present results, we cannot rule out the possibility that the pianist's touch may have an influence on the character of the piano tone. As regards the initial "thump" it seems rather clear that it contributes to the character of the note, while an actual influence on the tone character via the hammer motion has not been verified yet.

String vibrations
We now leave the "interfacing" steps of the action and hammer and focus our attention on the result, the vibrating string. The detection of the string motion was accomplished by utilizing the law of induction - "when a conducting wire is moving in a magnetic field a voltage is generated across the wire." For this purpose we applied a concentrated magnetic field at a desired point along the string by the use of a small, strong magnet. The induced voltage over the string - unfortunately very low due to the short-circuiting iron frame - was proportional to the string velocity at the point of the magnet.

String motion
The string motion on each side of a hammer in the middle section of the piano is illustrated in Fig. 15. On the side facing the bridge (upper panel) one sees the following. First the initial pulse, or hump (I) passes on its way to the bridge. Then nothing happens for a period of time, while the string is at rest a little displaced relative to its equilibrium position. After some delay, corresponding to the travelling time to the bridge and back again, the pulse returns (II), now turned upside down (inverted) on reflection at the bridge. The pulse continues to the agraffe where it is reflected once more and turned right side up. Shortly after this reflection, the pulse returns to the observation point (III). (Because of the short distance between the hammer and the agraffe, the travelling time from the hammer and back again is very short, and the incoming pulse (II) and reflected pulse (III) partly merge.) The first period of the string motion is now completed, and the pulse continues towards the bridge for the next round trip, and the process repeats. The curve displaying the string velocity may be somewhat more difficult to interpret, but is in fact more informative on the very details of the process. A hump passing the magnet, which is observed as a single pulse in the displacement curve, corresponds to a positive and a negative peak in the velocity curve. This is so since the string moves in the opposite direction during the latter half of hump when the string is restored to its initial position. Remember also that the velocity is high where the slope of the displacement curve is steep. On the other side of the hammer, towards the agraffe (Fig. 15, lower panel), the picture is entirely different during the initial moment when the hammer is still in contact with the string. During that period, the hammer acts as a temporary string termination and the initial pulse is reflected back and forth on the short string segment between the hammer and the agraffe. This causes repeated impulses on the hammer, and after about four or five such impulses the hammer is released from the string. In fact, this motion of the trapped pulse on the short string segment is the major mechanism of hammer release for most notes on the piano.

Fig. 15. String motions close to the hammer; bridge side, observation point B (upper panel), and agraffe side, observation point A (lower panel) fro a C4 note atfortelevel.In the displacement curve (I) denotes the initial outgoing pulse, (II) the same pulse after the first reflection (at the bridge), and (III) the same pulse after the second reflection (at the agraffe). The corresponding pulses in the velocity curve are denoted by 1, 2 and 3. Note that each passing displacement pulse corresponds to a positive velocity wave (up) as well as a negative (down). The round-trip time for a pulse on the string (period time) is indicated by T. Observe that the string motion on the agraffe side is entirely different from the motion on the bridge side during the hammer-string contact.

Bass - mid - treble

We can now proceed and compare the string motion and corresponding spectra in the bass, mid and treble sections, respectively (see Fig. 16). The pulse pattern on the string, described above for a midrange note (Fig. 15), is in fact even more pronounced for the bass notes. Due to the relatively short excitation pulse in comparison to the fundamental period (cf. Fig. 8), a bass string remains at rest for a long period of time before the first reflected pulse returns. Surprisingly, the returning pulse is foreboded by a ripple-like motion with growing amplitude. This behavior is caused by a strange phenomenon called dispersion, which is a manifestation of the stiffness of the string. As discussed in the Introduction to this book, the propagating pulse is made up of many Fourier components (partials), each with its own frequency. Due to the stiffness, those Fourier components with high frequencies propagate with slightly higher velocity and thus returns a little earlier, which causes the spread of the pulse.

Fig. 16. Comparison of the string waveforms (velocities) for treble (C7, top), mid (C4, middle) and bass notes (C2, bottom) played mezzo forte. The pulse character of the string motion is clearly seen for the bass note, while completely obscured in the treble.

The string motion in the treble is very different from the pulses observed in the mid and bass sections. This is due to the width of the initial pulse, which is so long in relation to the round trip time on the string, that it is no longer possible to observe individual pulses. The outgoing and reflected pulses get mixed up and form a pattern resembling a standing wave. The corresponding spectra for the bass, mid and treble notes also show interesting differences (see Fig. 17). As predicted in the discussion of the hammer-string contact durations, the treble contains only few partials, but these reach high frequencies (about 10 kHz for the treble note in Fig. 17). The middle register note contains more partials, up to approximately 7 kHz. The bass is very rich in partials, but on the other hand they do not

extend beyond 4 kHz. These dramatic spectral differences reflect the present preferences of the piano designer.

Fig. 17. The corresponding spectra for the notes in Fig. 16. The treble note (top) shows only a few partials but these reach high frequencies. The bass note on the other hand (bottom), is rich in partials, but they do not extend very high in frequency.

By changing the hammer properties, remarkable differences in spectra can be obtained (see Fig. 18). The three spectra in this figure all refer to middle C (C4), struck with different hammers. By means of a treble hammer (C7) the spectrum can be extended beyond 10 kHz, compared to the normal 7 kHz. With a bass hammer (C2) on the other hand, the spectrum is cut off at approximately 4 kHz. Apparently none of these extremes are appreciated by the manufacturers of today (or the customers).

Fig. 18. Spectrum of the note C4 (mf), illustrating the influence of hammer properties. The original hammer (middle) is replaced by a treble hammer (C7, top) and a bass hammer (C2, bottom). Note that the spectral changes are substantial, resembling the spectral differences between bass - mid treble in Fig. 17.

Voicing
The spectral properties can also be influenced by the piano technician. By a procedure called voicing, in which the hammer felt is treated with needles, the technician gradually works out the optimum stiffness of the felt. Not surprisingly, this procedure gives smaller changes than the preceding experiment with exchanged hammers (see Fig. 19).

Fig. 19. Effects of voicing (C4, mf), showing the changes in waveform and spectra when a hammer which initially is much too hard (full line), is needled to normal stiffness (dashed line), and eventually "ruined" (thin line).

In this experiment, a hammer which initially was much too hard was softened in steps until it was destroyed. While the changes may appear very small as viewed in the string waveform, the spectral changes are in fact substantial, reducing the upper limit of the spectra from about 8 kHz to approximately 4 kHz. By normal playing, the felt gradually compresses and the strings introduce grooves in the hammer surface. Once in a while - the keener the pianist, the more frequently - the instrument is ready for a reshaping of the hammers (in which the outer layers of felt are removed by filing), followed by a voicing. Since a proper voicing procedure is not reversible - the hammer cannot be made harder in a satisfactory way once it has been needled too much - it is strongly recommended to always consult an authorized piano technician.

The nonlinear hammer

The most remarkable feature of the tone generation in the piano still remains to be presented. In almost all traditional instruments, the tone quality changes with the dynamic level. This means that a note in forte not only is an amplified version of the same note in piano, but also contains many more high frequency partials. Apparently this has been a desired property in the design of the traditional musical instruments - which by the way has caused the designers of synthesizers much trouble - but the reasons for it can only be speculated on. One answer could be that the traditional instruments simulate the properties

of the human voice, for which such a spectral sharpening with level is normal, and in fact almost impossible to suppress. Anyway, the boosting of the higher partials with rising level can be achieved by a variety of mechanisms, different for different types of instruments. They have only one property in common; the generated waveform - be it a variation in air pressure as in the wind instruments or the motion of a string as in the violin or in the piano - will exhibit more abrupt twists and turns and hence more high frequency Fourier components (partials), the higher the dynamic level. For instance, in the clarinet the design of the reed and the closing surface of the mouthpiece are responsible for the increasing "distortion" of the waveform with rising level. In the violin, a strange friction characteristic between the rosined bow hair and the string serves the same purpose. A corresponding feature is also incorporated in the piano. The secret is hidden in the felt hammers which become progressively harder the more they are compressed (nonlinear stiffness). The result is remarkably indeed; the hammer "feels" soft to the string when striking at a low velocity (piano), but "transforms" to a much stiffer piece of felt when striking at a high velocity (forte). This increase in effective stiffness sharpens the initial pulse created at hammer contact, and the high-frequency part of the spectrum is enhanced at loud dynamics, as illustrated in Fig. 20. The changes are large, in fact much larger than the changes reached in the experiments with exchanged hammers or voicing. It is very likely that without this design feature of "an automatic treble control connected to the volume knob," the piano would never have reached its present high appreciation.

Fig. 20. Comparison of the spectra for a piano tone at three dynamic levels, p - mf - f, showing the boost of the higher partials with rising dynamic level (C4).

The properties of the hammers are thus most critical to the performance of a piano. In paraphrasing certain advertisements of automobile manufacturers it could be said: "Use only original spare parts!" and indeed, poor hammers will certainly ruin the sound of even the best of pianos. Also, service should be carried out by professionals only. An experienced

piano tuner/technician should be welcomed periodically into every home which appreciates a well-playing piano.

Coda
In closing this exploration of the initial stages of the tone production in the piano, it is tempting to conclude that the simpler a design may look, the more sophisticated its function appears to be. For instance, it was no great surprise to observe that a rather complicated process takes place in the elaborate action during a stroke. However, the flexing hammer shank and the nonlinear hammer felt - both seemingly very simple parts - showed a much more complex function than we had imagined. The "simple" piano hammer, which is decisive for the tone quality, still hide some secrets, and more research awaits before they can be completely revealed. In the meantime, we may continue to play and enjoy the piano unconcerned, even though we do not know exactly how the tone is produced, not even in the initial stages.

Acknowledgments
The authors are indebted to the piano technicians at the Swedish Radio Company, in particular Hans Norn and Conny Carlsson, for their patient sharing of expertise concerning pianos and piano regulation. The kind participation of pianists Elisabeth von Waldstein and Ove Lundin in the experiments is gratefully acknowledged. Special thanks are given to The Swedish Radio Company for generously making one of their grand pianos available for the experiments.

Recommended reading
The interested reader can now and then find rewarding articles in the international journals on music and acoustics. We especially recommend the following articles, which cover different aspects of the material we have presented. Boutillon, X. (1988): "Model for piano hammers: Experimental determination and digital simulation," Journal of the Acoustical Society of America 83, pp. 746-754. Hall, D. (1986): "Piano string excitation in the case of a small hammer mass," J. Acoust. Soc. Am. 79, pp. 141-147. Hall, D. (1987): "Piano string excitation II: General solution for a hard narrow hammer," J. Acoust. Soc. Am. 81, pp. 535-546. Hall, D. (1987): "Piano string excitation III: General solution for a soft narrow hammer," J. Acoust. Soc. Am. 81, pp. 547-555. Hall, D. & Clark, P. (1987): "Piano string excitation IV: The question of missing modes," J. Acoust. Soc. Am. 82, pp. 1913-1918.. Hall, D. & Askenfelt, A. (1988): "Piano string excitation V: Spectra for real hammers and

strings," J. Acoust. Soc. Am. 83, pp. 1627-1638. Hart, H., Fuller, M. & Lusby, W. (1934): "A precision study of piano touch and tone," J. Acoust. Soc. Am. 6, pp. 80-94. MacKenzie, C.L., Vaneerd, D.L., Graham E.D., Huron, D.B. & Wills B.L. (1986). "The effect of tonal structure on rhythm in piano performance," Music Perception 4(2), pp. 215-225. Podlesack, M. & Lee R. (1988): "Dispersion of waves in piano strings," J. Acoust. Soc. Am. 83, pp. 305-317. Sloboda, J. (1983): "The communication of musical metre in piano performance," Quarterly Journal of Experimental Psychology 35A, pp. 377-396. Suzuki, H. (1987): "Model analysis of a hammer-string interaction," J. Acoust. Soc. Am. 82, pp. 1145-1151.

About the authors

Anders Askenfelt and Erik Jansson received their basic training from the School of Electrical Engineering at the Royal Institute of Technology in Stockholm. After finishing their studies they joined the Music Acoustics Group at the Department of Speech Communication at the same institute; Jansson as a founding member in 1967 and Askenfelt in 1975. Since then they have been active in research in music acoustics, receiving their doctorates in 1973 and 1988, respectively. While Jansson has concentrated on the acoustics of the violin and the guitar, Askenfelt has been active in computer transcription of melodies, acoustical analysis of voice quality, and, more recently, research on the acoustics of the bowed instruments and the piano. They both enjoy music as amateur string players.

The hammer and the string

Donald E. Hall

Introduction
If we want to have a full understanding of how the best piano music is heard, we have many things to consider, as I have outlined in Fig. 1. Many important and complicated things happen in our fingers, ears, and brains; I can recommend a recent book for some enjoyable reading about those things (Wilson 1986). Even the physical processes through which the piano generates sound (the motions of the hammer, strings, and the soundboard) involve several stages which all require careful work to understand. I want to start by supposing that the hammer is already moving in the proper way, and studying how it gives its energy to the string. If we could understand that striking process completely, then we could know what kind of string vibration reaches the bridge and serves as input for the more difficult question of how the soundboard works. We can see better what kind of problem we have by comparing some other instruments. When we pluck a guitar, harp, or harpsichord, the process is well described by saying that the string is first held steadily in a triangular shape; at the moment of release, the finger or plectrum "suddenly" disappears and leaves the string to vibrate freely. This motion of the string by itself is a simple enough physical problem that we can summarize it in Fig. 2. When we show the spectrum, which tells the strength of each harmonic component in the string vibrations, the triangular initial shape of the string always provides an envelope curve requiring that the vibration level decreases by 6 dB (meaning 1/4 as much energy) for every octave higher in frequency. The particular place where the plucking was done acts further to suppress any harmonics which have nodes at that point; for example, plucking a 50 cm long string 10 cm from one end gives a spectrum in which the fifth, tenth, and fifteenth etc harmonics are absent.

Fig. 1. Steps in the process of playing the piano.

Fig. 2. Theoretical spectrum of vibration mode energies for a string plucked onefifth of its length from one end.

If we try to think of a similar simple picture for a piano string, we might consider two possibilities. First, imagine that the hammer pushes the string into a triangular shape, as for the harpsichord; then the string must push the hammer back down in a similar way as an arrow is moved by a bow-string (Fig. 3). If that motion is slow enough that the string always has the shape of a triangle, merely getting more and more shallow, we can predict a harmonic spectrum much like that for plucking. The only difference is that it falls off at a faster rate, 12 dB per octave. Unfortunately, this prediction is very different from what we measure in real pianos. That is not surprising when we say that we should not expect that picture to be correct unless the hammer is much heavier than the string, which is true only for the extreme treble end of a piano.

Fig. 3. Naive prediction of possible vibration spectrum for a thin string struck by a very massive hammer.

A second simple picture would be that the hammer bounces away so quickly that most of the string does not even begin to move while the hammer is touching. This picture leads again to an incorrect prediction, namely that the harmonic strengths do not fall off at all toward higher frequencies (0 dB/oct), apart from the "holes" related to the striking position (Fig. 4). Again, we realize that we should not expect this picture to be correct unless the hammer is much lighter than the string, as in the extreme bass range - and yet, even there, the predictions do not agree well with measurements. You will notice that these two incorrect pictures are quite contradictory as to whether the piano's sound would have a brighter or a darker spectrum (0 dB/oct or -12 dB/oct) than a harpsichord (-6 dB/oct).

Fig. 4. Naive prediction of possible vibration spectrum for a string struck a very brief blow by a hammer with very little mass.

We see that we can only come to a correct understanding by admitting that the hammer remains in contact with the string for a finite time, during which the force exerted by the hammer upon the string may change in a complex way. In Fig. 5 we are reminded of how the effective hammer mass may vary only from about 12 g in the bass to 6 g in the treble, while the string masses cover a much wider range. (Note that where there is double or triple stringing, it is the total string mass for each note that is important here.) Since the hammer mass makes it about an equal partner with the string through much of the important middle range of the piano, we are not surprised to see in Fig. 6 that measured contact times in the middle range are quite similar to half of the period with which the string would vibrate alone - and longer than the vibration periods of most of the upper harmonics.

Fig. 5. Comparison of hammer mass (m) with total string mass (M) on a typical grand piano.

Fig. 6. Comparison of string mode frequencies with typical hammer contact durations.

So if we want to understand what is happening during this process we must be able to measure the string and hammer motion on time scales shorter than a millisecond. And, to construct a correct theory explaining this process, we must study how the hammer and string each control what the other is doing during that very brief time of contact. Appropriate things to try to understand would include:

Duration of hammer-string contact - how long? Energy transfer to the string - how efficient? String motion after contact ends - what spectrum? Parameters - How do each of those depend upon: a. b. c. d. Striking position (fraction of string length) Hammer mass (compared to string mass) Hammer stiffness ("hard" vs. "soft") Hammer velocity (pp vs. ff)

Other possible complications (less important): a. b. c. d. Hammer width Hammer resistance (energy dissipation) String stiffness Loss of energy to the soundboard

Idealized theory - a brief summary

Let me describe what we find if we try to calculate what happens with the simplest possible model of the hammer-string interaction. (Details may be found, along with a survey of the history of this problem, in Hall (1986, 1987a, 1987b).) There is a fairly simple differential equation that describes what must happen if the hammer is pictured as only an extra mass temporarily attached to the string (Fig. 7). We can write solutions to this equation that represent first a simple wave travelling away from the hammer in both directions (the curve labelled 0), and then the reflection of this wave from the ends of the string. When a reflected wave tries to pass the hammer, it pushes downward on it, and this interaction generates a new outgoing wave (labelled 1). This wave in turn, after reflection from the ends, gives rise to grandchildren (labelled 2), and so on. Thus, if the hammer remains in contact long enough, the total disturbance will include a large number of waves travelling in both directions. Such a theory, with perfectly hard hammers, will surely not predict what happens in a normal piano. But it could be applied to the case where a very "tinny" sound is achieved by putting thumbtacks (an American term; you may know them as "drawing pins") in the hammers.

Fig. 7. Shapes of the successive generations of waves created on a flexible string by the impact of a pointmass hammer; (0) initial wave, (1) daughter wave, (2) granddaughter wave, etc. The curves are normalized so that they may represent any key. Time is given in multiples ("units") of the typical hammer deceleration time (mc/2T) and the displacement in multiples of (mcV/2T), where m is the hammer mass, c is the propagation velocity on the string, T is the string tension, and V is the initial hammer velocity. For a blow in the middle register at mf the "units" would typically correspond to 1 ms and 1 mm, respectively.

Fig. 8. Representation of a soft hammer as a mass with a spring between it and the string. The compression of the hammer is equal to the difference between the displacement of the hammer after string contact (n), and the displacement of the string (y).

If we admit that the normal felt hammers are soft, we must go on to do our calculations based on a picture where a spring cushions the contact of the hammer mass with the string (Fig. 8). Now the exact shape of the wave travelling along the string depends on the strength of the spring (Fig. 9a). Here 0.1 denotes small compliance, meaning a strong spring and a stiff hammer; 1.5 means large compliance, a weak spring and a soft hammer. Figs. 9(b) and (c) show the effect of the compliance on the daughter and granddaughter waves. It is not surprising that greater compliance makes a smoother, more gentle wave.

Fig 9. Initial wave (a) and its descendants (b) and (c) created by impact of an idealized soft hammer on a string. The same "units" as in Fig. 7.

Now doing a complete calculation of the string motion, including all these waves, would be very impractical by hand. But it is an ideal problem for a small computer, which can produce pictures like Fig. 10 showing the total effect of all the reflected waves on the shape of the string.

Fig. 10. Example of successive shapes taken by a string after hammer impact. The vertical lines represent how fast each part of the string is moving.

The computer program can be responsible not only for the arithmetic, but also for deciding at each moment how many waves are present, and for determining when the hammer finally loses contact with the string. A typical result is shown in Fig. 11, expressed in terms of the history of the force applied to the string. See how a soft hammer pushes rather steadily on the string, but a very hard hammer gives a series of sharp pulses of force. Once this information is known, it can be used to calculate how much energy has been delivered to the string, and how that energy is divided among the different harmonic modes.

Fig. 11. Force history for an idealized hammer with mass equal to that of the string, striking at 1/8 its length from one end. Examples of hammer compliance range from 0 (perfectly hard) to 0.8 (rather soft). This plot is normalized, using the same "unit" for time as in Fig. 7. Force is given in multiples of (2TV/c). For a mf blow in the middle register this "unit" would correspond to approximately 10 N.

What is most interesting is to ask the computer to do many of these calculations, in order to see how the results depend on the properties of the hammer. Fig. 12 shows some sample predicted spectra, with very hard hammers at the back and softer ones toward the front; as we would expect, this says that the softer hammers would not have as much of the higher harmonics in their sound.

Fig. 12. Sample predicted spectra for a simplified (linear) hammer as a function of its compliance (softness).

Comparison with measurements

We should not trust those calculations too much without making some measurements on real pianos. I had the opportunity to make such measurements together with Anders Askenfelt in 1985 at the Royal Institute of Technology in Stockholm (which is reported in more detail in Hall & Askenfelt, 1988). One of the things we did (Fig. 13) was to measure the string motion on a 2 meter GrotrianSteinweg piano. For this we placed a small magnet so that the string would move between its poles; this motion generated an electrical voltage which we could record and analyze.

Fig. 13. Placement of detectors for measuring string velocity and hammer acceleration.

In sample results for middle C (Fig. 14), we can observe a repeated pattern of string motion when it is vibrating freely. But during the time the hammer is in contact with the string, we see evidence of the action of individual reflected waves travelling between the hammer and the agraffe. (The upper oscilloscope trace tells directly about the force acting between hammer and string.)

Fig. 14. Oscilloscope traces of hammer acceleration (top), which is proportional to force, and of string velocity (bottom) for middle C played mezzo forte.

A spectrum analysis of this data shows (Fig. 15) rapidly declining strength for higher mode numbers. If we compare a 12 dB/oct curve, we can see that the measured spectra are much steeper, especially for pianissimo playing, so that our simple "first guess" (cf. Fig. 3) is not nearly good enough in explaining what happens. Another interesting result came when we measured a prepared set of hammers with different degrees of "pricking" (voicing) of the felt (Fig. 16).

Fig. 15. Measured string vibration spectra for middle C, struck with different degrees of force.

Fig. 16. The effect upon string vibration spectra when the hammer is treated by "pricking" (voicing); (1) "too hard", (3) "normal", and (4) "too soft."

Cases (1) and (4) were too hard and too soft, respectively, to be satisfactory to the piano technician, and they produce easily measurable differences in strengths of the high harmonics. Treatment of hammers by a skilled piano technician will control their effective stiffness within about 5 or 10%, or even less. Now we must compare such measurements with predictions made by the computer (Fig. 17). We find that there is resemblance in some of the general features of the spectra, but definitely not good agreement in detail, no matter which value of hammer stiffness we try. To see what is probably the most important reason for this, look back again at Fig. 15, and notice that the pianissimo spectrum is much steeper than the fortissimo. This means, as all musicians realize, that the sound of fortissimo playing is not the same as if you merely play pianissimo and use an electronic amplifier. The first case is not merely louder, it has a brighter tone quality as well.

Fig. 17. Predictions from linear theory, to be compared with Figs. 15 and 16. Compliance varies from 0.1 (hard) to 0.9 (soft).

But the theory I have described so far is a "linear" theory. This means one in which a doubling of the original hammer velocity would simply double the amplitude of string motion, double the soundboard motion, and double the air motion at your eardrum, without changing any other detail. That is, it predicts the same spectrum shape for both piano and forte playing; a single curve shifted upward and downward would cover every case. This is clearly wrong, and it will be important for us to identify which part of the system has "nonlinear" properties, and devise some way of including that in a better theory.

I believe the crucial part of the system lies in the hammers. In order to show this, Askenfelt and I made some measurements of how the hammers from the Steinweg piano act in a controlled situation outside the piano. We mounted a force measuring device in the chuck of a drill press, and placed the entire keybed below it. Then when we hit the keys in the normal way, any hammer could be thrown against what is effectively a solid wall instead of against yielding strings. We recorded how the force on the hammer changed during the brief contact by using a storage oscilloscope, with results like those in Fig. 18. Now if the hammer were an ideal "linear" device, those two curves would be equally wide. This phenomenon would be like the swinging of a pendulum, which for a given length takes the same amount of time whether it swings ten degrees or only one degree. But contact time was actually much shorter for the stronger blow; therefore the hammer felt acts as a non-ideal, nonlinear spring. To get an idea of how such a spring works, one can think of a helical spring in which the distance between the turns (pitch) gets successively smaller. If the force needed to compress the spring the first centimeter is, let's say, 10 N, then the next centimeter may take not 20 N, but 40 or even 80 N, depending on the degree of nonlinearity. Fig. 18. Force histories when a real piano hammer hits a rigid wall. The vertical scale is expanded for the soft blow (pp), during which the maximum force was less than 1/10 as much as for the hard blow (ff).

We examined the nonlinearity in the hammers by recording a series of blows for two hammers in each octave, producing data for each hammer that we analyze graphically as shown in Fig. 19. Here the lines would be horizontal if the hammer were a linear device; the greater the slopes of these lines, the greater degree of nonlinearity, and therefore the greater difference in the way the hammer will act between pianissimo and fortissimo playing.

Fig. 19. Decrease in contact time with increasing strength of blow, graphed so that greater slope indicates greater degree of nonlinearity in hammer stiffness.

Fig. 20 a,b,c,d. Variation of nonlinearity exponent from bass to treble for several different sets of hammers.

To see whether there is any definite trend in the hammer properties from bass to treble, we use the slopes from the graphs to compute a value (exponent) telling the degree of nonlinearity for the individual hammers (see Fig. 20a). Here a value of one (1) for the exponent (at the left-hand margin) would represent linear behavior. The more important result in this figure seems to be a greater degree of nonlinearity toward the treble. The decrease in contact time from bass to treble is mainly a result of the decreasing hammer mass. I have more recently measured several other sets of hammers in the same way (Figs. 20b,c,d). While the hammers on the Steinweg piano had been in use for some time, the Isaac hammers were a brand-new set just being installed on the large Steinway D in my university's Music Recital Hall, and were not yet "played-in". The Muir Wood is a Scottish piano from around 1800 - 1810, restored and located in the Department of Music at the University of California in Berkeley. There is perhaps less difference between treble and bass here, but from the variations one suspects that the hammers are presently not in very good, uniformly-voiced condition. A grand in the piano technician's workshop at my university (CSUS) was chosen as an example of an instrument with a very worn set of hammers in poor condition, producing a harsh, unpleasant sound (Fig. 20d). I do not yet feel sure that I know how to interpret these differences from one set of hammers to another. If I had guessed before measuring, I would have thought the harsher sound of the bad hammers might mean a larger value of the nonlinearity exponent, but according to the measurements it is really smaller instead. I can explain this if, when we said harsh or unpleasant, what we really meant was specifically that we did not get as much darker tone as we expected when trying to play softly. Askenfelt has suggested that this may be partly because well-used hammers develop flat grooves that cannot make as gentle a contact as the original rounded shape. Before stating any firm conclusions, I would want to make this kind of measurements on many more sets of hammers, from different makers and in different conditions.

Other complications
Before investing all our future efforts in the exact behavior of the hammers, it would be good to remind ourselves whether there are other effects not represented in our ideal theory which would also need to be taken into account. Fig. 21 (from Hall & Askenfelt 1988) tries to summarize several possibilities. One we can easily dispose of is the fact that the hammer contact is spread over as much as 5 or 10 mm of string rather than being concentrated at a single point. This is still such a small fraction of the string length that it should not matter except for frequencies above the line at the upper left - but these are so high that their strength in piano spectra is of no importance anyway.

Fig. 21. Estimates of frequencies above which hammer width, string stiffness, and soundboard motion become important in determining string spectra. (The "pulse decay" line indicates an upper limit for trusting the picture of Fig. 4, and "bow-and-arrow" a lower limit for trusting Fig. 3.)

Another problem with the simple theory is that it assumes both ends of the string are mounted on perfectly rigid supports, yet we know that the bridge actually moves. This means that some energy is lost from the string to the soundboard, so that the reflected waves are not really quite as strong as the theory supposed. This in turn changes the force that the hammer will provide later when it is interacting with those waves. I have estimated that this should be taken into account for the last octave or two at the treble end (above the line "soundboard" in Fig. 21), where it will help explain why those spectra are very steep; but otherwise it may be of only minor importance. A third effect which is more generally important comes from the finite thickness of the strings. Where it works quite well to talk of harpsichord strings as if they are perfectly flexible, piano strings are much thicker and resist bending. This stiffness, I should point out, is not a nonlinear effect; it is merely something which aids the applied tension in trying to make the string straight. I have estimated in two different ways when this might become important, shown by the two lines "stiffness" in the upper part of Fig. 21. Above these lines, the predicted spectra should really be a little steeper, as the stiffness makes it more difficult to set those high-frequency vibration modes into motion. But in the central part of this figure is a region, including something like the first ten harmonics of middleC, for instance, where none of these other details should be very important. In this region, which

covers the larger portion of the strong and interesting parts of the spectra, I believe the correct description of the hammer is the key to understanding what happens in real pianos.

Further possibilities
What else can we do that may give us a more firm understanding of the interaction of the string and hammer, and of how to shape that interaction to produce whatever spectra we find most useful musically? As for theory, there is a first step that is not very difficult. That is to calculate the wave shape, or the corresponding force (Fig. 22), for the very first wave that travels away from the hammer. The upper panel shows the case of a linear hammer, already treated in great detail in our earlier theory, with the curves peaking sooner for harder hammers and more gradually for softer ones. The lower panel shows how these force histories would change in the presence of a certain fairly large degree of nonlinearity. Perhaps there is a suggestion in these curves that the nonlinearity tends to increase their similarity, making them less dependent upon our ability to control the stiffness (K) exactly.

Fig. 22. Force histories (initial wave only) for linear hammers, exponent p=1 (upper panel) and nonlinear hammers with exponent p=4 (lower panel). In each case, larger values of K denote harder hammers. Normalized plot, the same "units" as in Fig. 11.

Unfortunately, those pictures only tell us how the hammer interacts with an infinitely long string. We cannot go on to include reflected waves by simply adding several of these together; some entirely different mathematical method must be introduced to solve the nonlinear equations of motion. Some sample calculations of this sort have recently been done by Suzuki and by Boutillon. Fig. 23, which is from Suzuki (1987), shows one calculated case of a bass hammer interacting with the A0 string of a 6-foot Steinway grand. Curve (a) shows the hammer position and curve (b) the string displacement at the contact point; we may notice that where (b) rises above (a) the hammer loses contact, but there is renewed contact when the reflected wave arrives from the agraffe. Curve (c) gives the force experienced by the string, from which the vibration spectrum can be obtained. Fig. 24, taken from Boutillon (1988), shows a calculated case for one particular hammer hitting F4; the dashed line gives the calculated history of hammer acceleration, and the solid line is a corresponding measurement.

Fig. 23. Time histories of (a) hammer displacement, (b) string displacement, and (c) interaction force for a bass hammer (A0) with nonlinear compliance (Suzuki 1987).

Fig. 24. Comparison of predicted (dashed) and measured (solid) accelerations for a nonlinear hammer (F4), (Boutillon 1988).

But we do not yet have a wide range of such cases computed in order to see how the results change if we alter the important parameters such as hammer mass, stiffness, striking point, and degree of nonlinearity. Nor can we yet compare the predictions with actual measurements for a wide range of notes and of different pianos. I have some students interested in working on this problem, and I entertain some hopes that we may write a new computer program that will make it possible to explore the action of nonlinear hammers in a systematic way. Perhaps it is not very far in the future that we may be able to claim the ability both to understand and to predict with some accuracy just what changes in sound will result when string and hammer designs are changed in various ways. We could hope that this would eventually contribute to useful diagnosis of what could be expected from any proposed change in hammer design or manufacturing technique. In the meantime, I could see the possibility of one practical application of this work in the manufacture of new piano hammers. Good quality control on a new set of hammers means both that we achieve desired values of hardness and nonlinearity, and that these properties change smoothly as we go along the row of hammers. Either before or after work by the voicing technician, this could be checked by placing the keybed in a frame where it would be moved in 88 steps. At each step a mechanical striker would hit the key, perhaps 5 or 6 times, ranging from very weak to very strong impulses. The hammer would hit a force transducer, and a computer would evaluate the height and width of the signal produced. With this data, the computer could estimate the effective stiffness and nonlinearity of each hammer head, and produce either graphs showing how these properties vary from treble to bass, or lists with warnings of which hammers are not in good agreement with their neighbors. One could perhaps also devise a quick-clamp rig so that a set of hammer heads could be tested in this way without being mounted on shanks in an action. I hasten to add that this test should only be an aid to the technician, and would never entirely replace the voicer's ear as the final judge of good hammer preparation.

Acknowledgements
Many of the figures in this paper originally appeared in the references, and are reproduced by courtesy of the American Institute of Physics.

References
Boutillon, X. (1988): "Model for piano hammers: Experimental determination and digital simulation," J. Acoust. Soc. Am. 83, pp. 746-754. Hall, D. (1986): "Piano string excitation in the case of small hammer mass," J. Acoust. Soc. Am. 79, pp. 141-147. Hall, D. (1987a): "Piano string excitation II: General solution for a hard narrow hammer," J. Acoust. Soc. Am. 81, pp. 535-546. Hall, D. (1987b): "Piano string excitation III: General solution for a soft narrow hammer," J. Acoust. Soc. Am. 81, pp. 547-555. Hall, D. & Clark, P. (1987): "Piano string excitation IV: The question of missing modes," J. Acoust. Soc. Am. 82, 1913-1918. Hall, D. & Askenfelt, A. (1988): "Piano string excitation V: Spectra for real hammers and strings," J. Acoust. Soc. Am. 83, pp. 1627-1638. Suzuki, H. (1987): "Model analysis of a hammer-string interaction," J. Acoust. Soc. Am. 82, pp. 1145-1151. Wilson, F. (1986): Tone Deaf and All Thumbs (Viking Penguin, New York)

About the author

Donald E. Hall received his doctorate in Physics from Stanford University in 1967, as well as an M.A. in Music from the University of Iowa in 1973. He has been on the faculty of California State University in Sacramento since 1974. He is the author of two textbooks, Musical Acoustics (1980) and Basic Acoustics (1987). His research has dealt with piano string and hammer interaction and with the perception of musical interval mistuning; he has also worked in relativity and astrophysics. He leads a second life as a church organist and choral singer.

The coupled motion of piano strings

Gabriel Weinreich

Introduction
As everyone knows, the word "piano" for the instrument with whose acoustics this seminar is concerned derives from the name given to it by its inventor, Bartolommeo Cristofori, who shortly after its invention in c. 1709 had his creation described as "gravicembalo (~clavicembalo) col piano e forte" because, unlike the ordinary clavicembalo (that is, harpsichord), it was capable of varying its dynamic level. The description is, in some ways, even more apt than its originator intended. In the preceding lecture, Donald Hall has described how the radical nonlinearity of the hammer produces, along with the dynamic range, a correspondingly large range of different tone colours, giving the phrase piano e forte added significance. My lecture will be concerned with an even more peculiar fact, namely that, in a certain sense, the gravicembalo piano e forte can be said to be playing piano and forte at the same time!

Loud and sustained

To understand how this phenomenon originates, let us consider a hypothetical engineer assigned to develop a musical instrument in which a string which is initially excited by a hammer blow (or, for that matter, by plucking) is to be coupled to a soundboard, and so produce a musical sound. Obviously, this engineer would have to make a decision with regard to the degree of coupling between the string and the soundboard, i.e., how efficiently the string vibrations are transferred to the soundboard. Here he would immediately encounter a difficulty, for the following reason. A strong coupling will lead to a sound which is initially loud but decays very quickly in time, since the same strong coupling must lead to a high rate of energy transfer from string to soundboard. If, on the other hand, the engineer wishes to have a long sustained sound, so as to give the instrument more of what one might call a "singing" quality, he must keep the coupling to the soundboard weak - and hence be constrained to a sound which is quiet and lacks carrying power. It is true, of course, that even without any soundboard coupling the string vibration would not last forever, since other mechanisms of energy loss - such as internal friction of the string, viscous dissipation in the air, and direct sound radiation from the string motion - do exist, but in practice this is irrelevant. To the best of my

knowledge, in all musically useful string applications the dominant mechanism of string damping is motion of the end supports, that is, coupling of the bridge to the soundboard. To a practical engineer, the competition between loudness and sustaining power would point to the need for some sort of compromise or "trade-off;" but a musical instrument is not, by its nature, an "engineering compromise." Most objects capable of producing sounds are, after all, not used as musical instruments. The ones which are selected for such a purpose always represent some type of "miracle," that is, a situation in which a seeming engineering limitation is overcome in an unexpected way. In connection with the piano, it turns out that our ear - or, more correctly, our ear plus brain - has a way of judging both loudness and sustaining power in a way that might not have been predicted. Specifically, a sound is perceived as loud if it starts out loud, even if it then decays quickly; and it is perceived as sustained if some part of it is sustained, even if that part is rather weak. Thus a sound which starts out with a high but quickly decaying amplitude, and which then, having reached a rather low level, switches to a much smaller rate of decay - so that there is a sustained but subdued "tail" or "aftersound" - is perceived as being both loud and sustained. And that is precisely the miracle of the piano tone.

Vertical and horizontal motion

The typical decay of a piano tone is shown in Fig. 1, which displays the sound pressure level as a function of time. We see that the string is struck by the hammer at about t = 2 seconds, and the damper is released, stopping the vibration, at about t = 17 seconds. The vertical scale is in decibels, so that the ordinate of the graph is proportional to the logarithm of the pressure amplitude. In such a plot, the drop in level would appear as a straight line if the decay of the sound were of a type called exponential, which is what a physicist would expect from a linear system such as the string and the soundboard. Instead, it is clear that the curve breaks into two portions of quite different decay rates. The initial portion, called "prompt sound," drops (in this case) at a rate of about 8 dB/sec; the final one, called "aftersound," at less than one-quarter that rate. As we shall see, the prompt sound is simply related to the theoretical decay rate determined by the string's coupling to the soundboard; whereas the aftersound, which gives the piano its perceived sustaining power, represents the "miracle."

Fig 1. Typical decay of a piano tone as illustrated by the sound pressure level versus time (Eb3 = 311 Hz). The decay process is divided into two parts; an initial attack part with a fast decay ("prompt sound") followed by a sustained part with slow decay ("aftersound").

The most interesting factor contributing to the existence of aftersound is the presence of more than one string for each piano note, and the consequent dynamical coupling that occurs among the strings struck by the same hammer. In the case of Fig. 1, however, we are dealing with a simpler phenomenon, since in this particular experiment only one of the three strings belonging to this note (Eb4) was allowed to sound. It should also be added that the curve in Fig. 1 was obtained by taking the signal produced by a microphone and filtering out all but the fundamental frequency; the complex decay pattern can not, in other words, be attributed to variously decaying partials (although such a factor may contribute to the way the overall note is perceived). Rather, the behavior of the decaying curve in Fig. 1 is explained by noting that even a single string vibrating at its fundamental frequency has two distinct modes of vibration corresponding to two possible directions of vibration (polarization); "vertical" and "horizontal." Here, we use the adjectives appropriate for a grand piano although the same phenomenon of course takes place also in the upright piano. What is happening here is that the vertical polarization is the primary one excited by the hammer, and so begins its life at a much higher amplitude than the horizontal one. However, since the bridge, which is attached to the soundboard, "gives" much more easily in the vertical than in the horizontal direction, the decay of the vertical mode is also much more rapid. As a result, the relatively slight amount of horizontal vibration becomes, after a while, dominant. Although this hypothesis can be verified by direct observation of the string vibration, it is interesting to approach it in a somewhat different way. In Fig. 2, we show another graph of sound level versus time (again at the fundamental frequency) which was obtained by the same procedure as Fig. 1 except that the location of the microphone is different. The data indicate clearly that we are dealing with two independent modes of vibration, which are producing sound waves through two separate radiating "antenna patterns." When one or the other mode dominates - that is, either near the beginning or near the end of the note - the two graphs show

identical behavior; but when the modes radiate with approximately equal amplitude, they can add either "constructively" or "destructively," depending on microphone placement.

Fig. 2 The decay of the same note as in Fig. 1 but recorded at a different microphone position. A comparison with Fig. 1 reveals the existence of two components in the sound field, radiated by the vertical and horizontal soundboard motions respectively. The two components are of equal strength during an interval around the intersection of the sloping lines. In this region the resulting sound level varies strongly with microphone position.

In fact, it is quite reasonable to suppose that the vibration pattern of the soundboard in response to a vertical force at the bridge is quite different from what it is for a horizontal force; and since at this frequency at about 311 Hz the wavelength of the sound is of the order of 1 meter, comparable to the size of the piano, the directional radiation pattern for the two can easily be quite different.

Two or more strings

The more interesting mechanism which leads to an aftersound - and which, as we shall see, makes it possible to adjust its relative amount by careful tuning - is the one on which the title of this lecture is based, namely the dynamical coupling among the three (or two) strings struck by the same hammer. Let us imagine two identical (and identically tuned) strings attached to the bridge at the same place. It is possible for them to vibrate in the same phase, both going up and down at the same time; or in opposite phase, one going up while the other is going down. (In the present discussion we concentrate, for simplicity, on the vertical motion.) In the latter, "antisymmetric," case the forces exerted by the two strings on the bridge will cancel, so the bridge will not move at all, just as though it were infinitely rigid. Hence, the decay rate of the vibration will be very slow. Conversely, if the motion is in phase, or "symmetric," the force on the bridge (and hence its displacement) will be twice as large as it would be for one string, so the decay rate will be doubled by this coupling. In this way we see the possibility for the overall vibration to decompose into a prompt sound and an aftersound.

The experimental data of Figs. 3 and 4 illustrate this behavior. The graph in Fig. 3 shows the vibration level of a single string; it differs from our previous graphs in that this time the motion of the string is directly measured (by a capacitive probe designed to be sensitive only to the vertical motion), rather than recording the radiated sound via a microphone. As in Fig. 1, felt wedges prevent all but one string from vibrating. The general behavior is equivalent to the prompt sound part of Fig. 1, except that the time scale of the graph has been considerably expanded.

Fig. 3 Typical decay of the vertical string vibrations when only one string in a trichord is left free to vibrate.

The graph in Fig. 4 is obtained in exactly the same way, except that one felt wedge is removed so that one other string of the trichord is allowed to vibrate. It must be emphasized that the capacitive vibration probe is not disturbed; what we are looking at is the vertical motion of the same string as before, except that a second string has been "brought into the system." Clearly, a new mode of vibration has now appeared, one in which the original string can vibrate for a much longer time before it loses its energy. It is, in fact, precisely the antisymmetric mode which is allowing this to happen.

Fig. 4 The decay pattern of the same string as in Fig. 3 when another string in the same trichord is left free to vibrate (but not struck by the hammer). The decay of the first string is strongly influenced by the motion of the other string.

A closely related phenomenon is illustrated in Figs. 5 and 6. The curves in Fig. 5 show both the vibration level of one of the strings and the (simultaneous) sound level, produced when two strings are struck by a single hammer under circumstances similar to Fig. 4. In Fig. 6 the experiment is repeated, but at about t = 11 seconds a felt wedge is suddenly inserted so as to immobilize the second string. We see that the vibration of the first string, as well as the produced sound, change immediately to the higher decay rate. This is, indeed, expected, since the antisymmetric mode (in which the force exerted by one string on the bridge is cancelled by the other string) is no longer possible when only one string is vibrating. But it is especially interesting that, whereas the string vibration level simply switches to a steeper slope, the sound level simultaneously jumps up by close to 20 dB. This is, of course, because the single string exerts a much larger force on the bridge than the two strings vibrating antisymmetrically. The phenomenon may, possibly, have some musical interest, in that if it were appropriately mechanized it would enable a pianist to introduce an accent into the middle of a sustained note, which in a normal piano is impossible.

Fig. 5 Sound pressure level and string vibration level versus time for two strings struck by the same hammer. The humps reflect beats between the strings.

Fig. 6 The same curves as in Fig. 5 but now the vibrations of one of the strings is stopped at t = 12 s by a felt wedge. This terminates the interaction between the strings, as a result of which the vibrations of the other string change decay rate and the sound level momentarily increases by 20 dB.

It is interesting to note parenthetically the rather complicated beat structure in Figs. 5 and 6, which is not precisely the same for the vibration and the sound. Presumably, the discrepancy comes from the horizontal modes, which contribute to the sound but not to the vertical motion. Two questions immediately arise: (a) If the hammer hits both strings equally and at the same moment, why is the motion not completely symmetric, resulting in "purely prompt" sound? (b) Even if the antisymmetric vibration is somehow excited, why does it produce any sound at all, seeing that it exerts no force on the bridge? The first question could be answered by pointing out that a real hammer must have some geometrical imperfection so that the impact on the two strings would not be absolutely identical; the second, by the fact that the two strings do not touch the bridge at the exact same location. These answers, although correct in principle, are not very satisfying because an aftersound that depends entirely on the presence of imperfections would also tend to be erratic and unpredictable in amount. More specifically, the ratio of the amplitude of aftersound to the amplitude of prompt sound would vary randomly when one note is compared with its neighbouring ones. This would give even a very good piano a rather uneven "singing quality." In fact, the correct understanding of the string motion requires us to consider the further degree of freedom which results from the two strings not being tuned in an identical manner. "Mistuned" strings The mathematical treatment of the coupled motion of two strings which are tuned almost, but not exactly, alike is rather complicated, since the symmetry arguments which implicitly led to the identification of the "symmetric" and "antisymmetric" motions as normal modes of vibration no longer work. We shall, nonetheless, try to give some idea of the behavior of this complex system, but before we begin one thing must be made very clear, and that is that the presence of "mistuning" does not necessarily lead to beats, if by "mistuning" we mean a difference between the vibration frequencies of the two strings when they are individually excited (that is, when one string is made to sound with the other one damped). The point is, of course, that the frequencies of the two strings vibrating at the same time are affected by the coupling which occurs between them due to the bridge not being perfectly rigid. A different, and perhaps clearer, way of explaining the distinction between the frequency of an isolated string and the frequency at which it vibrates when coupled, is to note first that the frequency of vibration of a string whose end-support is not completely rigid depends on the type of resiliency the string meets at the support, or using the scientific term, the impedance of the support, see Fig. 7. If the support is "springy," that is, one which displaces sideways in the direction in which the string

applies a force to it, there will no longer be an exact node at the support. Instead, the extrapolated node will be somewhat beyond the physical end of the string; or, in other words, the string will "think" that it is longer than it really is, causing it to lower its frequency.

Fig. 7 Illustration of the influence of the end-supports on the vibration frequency of a string. A "springy" support (top) lowers the frequency without damping the motion, because it makes the string act as if it was a little bit longer than it really is. A "massy" support (middle) raises the frequency without damping its motion. The reason is that the string must pull back on the mass to reverse its direction, the result being that the string acts as if it was a little bit shorter than it really is. A resistive support (bottom) does not influence the vibration frequency but damps the motion. The friction at the sliding contact at the support means that energy is drained from the string and the vibrations decay.

The opposite happens with a "massy" support, that is, a support whose motion is inertia-limited: in this case, its acceleration is in the direction of the applied force, and its displacement in the opposite direction. The node then moves inward from the physical end of the string, with the result that the string "thinks" that it is shorter than it really is so that its frequency is raised. Finally, if the support is purely "resistive," so that its velocity is in the direction of the force, its frequency is neither raised nor lowered, but its decay rate is increased. If now a second string is attached to the bridge at the same (or almost the same) location, its presence will affect the impedance of the bridge as seen by the first string. The result is that the "frequency of the first string" will be different from its vibration frequency before the other string was introduced. How large this shift will be depends on the difference between the new "frequency of the first string" and its

original frequency. We put the phrase "frequency of the first string" in quotation marks because it is really no longer that, but rather it is the frequency of the coupled normal mode in which both strings are vibrating. To get some notion of how these coupled frequencies are affected by "mistuning" the strings, we should think about another impedance - not the impedance presented to the string by the bridge, but rather the impedance presented to the bridge by the other string. A simple demonstration experiment will illustrate this behavior. If I attach a rope (representing the string) to a fixed chair and shake the other end up and down at some arbitrary frequency of my choosing, the fact that I am holding the end of a rope will make very little difference to what I feel. But if the frequency I choose corresponds to one of the resonant frequencies of the string, the place where I hold it becomes a node and "refuses to move." It is as though the fixed point at the other end of the string were transferred to the holding point; even though I am exerting a considerable force amplitude up and down, my hand almost does not move at all. Paraphrasing this behavior in terms of impedance, we would say that the impedance presented by the string to its support is generally quite low, but becomes very high as the frequency approaches a resonant frequency of the string. It should be clear from this discussion that if we observe the motion of two coupled strings, of which one is left untouched while the other is tuned, we would observe that the untouched string will change its frequency (although it remains at constant tension) as the second string is tuned. This is under condition that the two strings are close to a unison, since under those circumstances the impedance of the second string at the frequency of the first string becomes quite high, modifying the effective impedance of the bridge as seen by the first string. What is not so clear - and, in fact, requires considerable mathematical discussion - is precisely what this frequency shift will be. Interestingly, it turns out that it can be in either direction, depending on the impedance of the bridge itself. In particular, there exists a possibility for the two frequencies to "attract" and become locked together, so that slight tuning of either string does not affect the frequency of either but only the decay rates. This is what we meant by saying that a slight "mistuning" of the strings does not necessarily lead to the appearance of beats. The last figure, Fig. 8, shows some theoretical curves of the history of the vertical force exerted on the soundboard when driven by two strings, initially excited by a perfectly symmetric hammer blow. The different curves correspond to different "mistunings," and the bridge impedance is assumed such as to allow the "locking together" of the frequencies (which is not always the case in practice). In calculating these curves, we have assumed parameters more or less typical of the middle range of a piano keyboard. For this case, there are no beats unless the "mistuning" is more than about 0.3 Hz; more correctly, for smaller "mistunings" there is just a single "beat null," followed by a beatless aftersound whose level depends on the

"mistuning." Above about 0.3 Hz beats do appear, as exemplified by the curve drawn for a "mistuning" of 0.64 Hz; even here, however, the time between beats is a bit larger than the 1.6 seconds which would be naively predicted (1/0.64 Hz = 1.6 s). The importance of Fig. 8 is that it indicates how an excellent tuner can, under some circumstances, use very fine tuning control in order to adjust the aftersound of each note to a more uniform level than if it were due entirely to imperfections in the hammer or the string mountings.

Fig. 8 Calculated vertical force on the soundboard when driven by two strings with different "mistuning" ( f). In this example beats occur only when the "mistuning" is larger than 0.3 Hz, illustrated by the curve for f = 0.64 Hz. For smaller values the strings lock to a common frequency, and the effect of the "mistuning" is to control the level of the aftersound (cf. the curves for f = 0.22 Hz and 0.06 Hz).

In fact, it was observed by Kirk in 1959 that a carefully and competently tuned piano had the strings of the trichords tuned slightly differently by an amount that appeared to vary randomly from note to note. This randomness may, however, hide an underlying regularity. If, for example, you take a sheet of paper and tear it, examination of one of the pieces will reveal an irregular and seemingly random rough edge; yet comparison with the other piece will show that one irregularity exactly matches the other. Our hypothesis here is that, in the same way, the seeming "randomness" of the tuning comes from the fact that the skilful tuner was adjusting this quantity to another randomness, namely the randomness of hammer imperfections, in such a way that the result is not random. It would be interesting to test this hypothesis by investigating, for example, whether good tuners are consistent in the "mistuning" of the individual trichords when tuning the same piano over and over again.

Acknowledgements I would like to express my thanks to the Royal Institute of Technology and the Royal Swedish Academy of Music for inviting me to participate in this seminar, and to the United States National Science Foundation for supporting my research at the University of Michigan

References Kirk, R. (1959): "Tuning preferences for piano unison groups," Journal of the Acoustical Society of America 31(12), pp. 1644-1648.

About the author

Gabriel Weinreich has been dedicated to physics since the start of his career. After receiving his doctorate in Physics from Columbia University in 1953 he joined the Bell Telephone Laboratories, where he became a member of its pioneering group in solid state physics. Since 1960 he has been professor in Physics at the University of Michigan, Ann Arbor. He has written textbooks on solid state physics and thermodynamics, as well as general physics. His research in musical acoustics, stimulated by his own playing of the piano and the cello, has dealt primarily with piano strings and bowed instruments. He is an ordained priest of the Episcopal Church.

The strings and the soundboard

Klaus Wogram

Introduction
When a piano key is depressed, the hammer is accelerated from its rest position via a series of levers, called the "action," and catapulted against the string. Shortly before the string impact, the direct contact between the action and the hammer is interrupted, and the hammer is carried the rest of the way to the string by its momentum. The shape and amplitude of the force pulse generated at the impact is determined by the final hammer velocity and by a combination of the mechanical properties of the hammer head and the strings (mass ratio, striking position, and stiffness of the hammer). The hammers and strings are designed so as to make the force pulse shorter in the treble than in the bass (see Fig. 1). Such a pulse corresponds to a spectrum, which spans a broad frequency range (up to 10 kHz for a treble hammer). From this continuous pulse spectrum, the string filters out the frequency components corresponding to its resonant frequencies. These string spectrum components transfer the vibration energy to the bridge. Both the properties of the string material and the dimensions of the string influence this energy transfer.

Fig. 1. Force pulses of the hammer impacts for a bass note (No: 2 = Bb0) and a treble note (No: 88 = C8).

The bridge serves as the connecting element between the string and the soundboard, transferring the vibrations of the string into the soundboard. The strings are distributed over the entire length of the bridge which extends approximately diagonally across the soundboard. As a consequence the strings will deliver their vibration energy at different points on the soundboard, which means that they will all encounter different matching conditions. The soundboard transforms the mechanical vibrations into radiated sound. As a first approximation, the soundboard acts like a large diaphragm clamped around its edge. Like all diaphragms, the soundboard exhibits a series of resonances, the individual intensities being determined by the point of excitation. If we ignore the resonances for a moment, a diaphragm exhibits a radiation efficiency that increases as a function of frequency and reaches a theoretical maximum in the upper treble region. This would mean an optimal match between the soundboard and the surrounding medium in the upper treble, provided no losses occurred in the wood. Losses do occur, however, and these result in a decrease in the sound radiation in the upper treble. With increasing frequency, the dimensions of the soundboard also become increasingly large in relation to the bending wavelength. This means that the soundboard no longer vibrates as a simple stiff diaphragm, but tends to divide into a number of individual vibrating areas. The result is a pronounced peaking of the directional radiation, as well as a decrease in the radiated acoustic energy - air is uselessly pumped between areas which vibrate in opposite phases. In the lowest frequency region, the lack of separation between the upper and lower surface of the soundboard reduces the sound radiation (acoustic short-circuiting). Altogether these phenomena result in a favored region of acoustic radiation at frequencies between approximately 100 Hz and 2000 Hz. Returning to the soundboard resonances, we note that it is particularly easy to vibrate the soundboard at the resonant frequencies. Using a constant driving force, we therefore obtain a high vibration amplitude and velocity at these frequencies. The quotient of force and velocity gives us a highly informative quantity called mechanical impedance, which can be easily measured as a function of frequency (see box below). The resonance frequencies of the soundboard are readily observed in such an impedance curve, where we find them at the valleys, or minimum values. The contribution to the sound radiation from each resonance must, however, be determined from another measurement, that is, sound level versus frequency. This quantity is also readily measured.

In addition, by means of a modern technique called modal analysis, the vibration pattern of the soundboard at the resonances can be computed from simple measurements and visualized on a terminal screen. This technique also allows computer simulations of changes in the shape and thickness of the soundboard, which thus can be evaluated before a new prototype instrument is built.

Input impedance and soundboard properties

The vibration energy of the string is transferred to the soundboard, transformed from mechanical to acoustic energy, and radiated as airborne sound. The rate at which this energy flow takes place is determined by the properties of the soundboard (the "consumer") in relation to the properties of the string (the source). In engineering terms this relation is referred to as "impedance matching." The loading exerted by the soundboard on the string can be expressed by means of its input impedance (Z), which is defined as the excitation force (F) divided by the resulting velocity (v) at the point of excitation. Z = F/v A high value of the input impedance means that a large force must be expended in order to achieve a certain vibration velocity, whereas a low impedance means that a smaller force is sufficient to achieve the same velocity. Actually the impedance also includes a phase angle (), which tells us about the character of the impedance (massy or springy). Measurement technique A measuring setup for input impedance of piano soundboards is shown in Fig. 2 (Wogram 1984). An electrodynamic shaker is mounted on a carriage, which is moveable to any position of the soundboard. On the shaker an impedance head is mounted, which consists of two vibration pickups. The impedance head makes it possible to measure the excitation force and vibration velocity simultaneously, in other words, the two quantities needed to obtain the input impedance. A block diagram of the measuring setup and a description of the measurement procedure are given in the box below.

Fig. 2a. Measuring setup for impedance measurements. Stand (vertical beams) and crossbeam with movable carriage.

Fig 2b. Detailed view of bridge with shaker, impedance head and metal pin connecting to the bridge.

In order to obtain a representative set of measurements, no less than fourteen measuring points were analyzed for each instrument, spaced about 12 cm apart along the bridges (see Fig. 3). The strings were muted with strips of felt in order to prevent sympathetic vibrations of the strings from influencing the measurements.

Fig. 3. Soundboard of an upright piano with the positions of the measuring points (MP1 - MP14) marked on the bridges.

Measurement of input impedance and sound level The technique used for an impedance measurement is briefly the following. The soundboard is driven by a shaker at a selected point on the bridge, the force being transmitted via the impedance head. The shaker is fed by a swept sine wave in the frequency range from 20 Hz to 10 kHz, supplied by a tuneable oscillator and an amplitude controlled amplifier (AVC). The output signal of the acceleration pickup (A) in the impedance head is integrated and fed to the AVC-amplifier as a velocity signal (v). This amplifier regulates the input signal to the shaker so that the velocity at the bridge remains constant over the entire frequency range. Since the velocity remains constant, the force (F) is proportional to the input impedance (Z), which means that with proper calibration, the obtained force curve can be interpreted as an impedance curve. The magnitude of the impedance thus determined is registered by a level recorder, and the phase angle by a second similar instrument. Another level recorder connected to other measuring instruments simultaneously registers the sound level at a distance of 2 m from the soundboard. In this manner, we obtain three curves for each measuring point representing the magnitude of the impedance (Z), and its phase angle (), and the sound level (L). These three curves summarise the acoustic properties of the soundboard.

Fig. 4. Block diagram of input impedance and sound level measurements.

Input impedance An example of measured input impedance and radiated sound level curves for an upright piano assembly consisting of the wooden frame, soundboard and metal plate, strung and tuned to normal pitch, is shown in Fig. 5. It can be seen that the impedance curve is characterized by prominent tall peaks and valleys in the lower part, superimposed on a continual decrease from 100 Hz to the highest frequencies. For those readers who are particularly interested in engineering units we may add that the impedance reaches values around 1000 kg/s at 100 Hz and 10 kg/s at 10 000 Hz. All measuring points are characterized by a uniform decrease in impedance above 1000 Hz at a rate of about 6 dB per octave, without any prominent resonance peaks. The phase angle is close to -90, which means that the input impedance is dominated in this region by the resiliency (springiness) of the soundboard. We find a similar situation in the extreme low frequency region (below 100 Hz); here too the phase angle is almost -90 and no resonance peaks show up (the peaks below 50 Hz are not generated by the soundboard but by the stands and crossbeam holding the electrodynamic shaker).

Fig. 5. Acoustic measurements on an assembled unit for an upright piano consisting of the wooden frame, soundboard, metal plate and strings, tuned to normal pitch (MP 7). Input impedance (top), phase angle (middle), and sound level (bottom). The 0 dB level for impedance corresponds to 1000 kg/s.

Sound radiation The corresponding sound level curve (Fig. 5, bottom) demonstrates that the soundboard is incapable of radiating sound below 100 Hz. Above this frequency, individual resonances, which may reinforce the sound radiation, become noticeable. At approximately 1000 Hz, the upper limit of the range of favored sound radiation is reached. Above this frequency the sound level decreases steadily. The connection between a soundboard resonance and the corresponding sound radiation can be demonstrated in a simple way. By knocking on the soundboard with a finger (preferably close to measuring point 9), we hear a thump sound with a definite pitch. The spectrum of this thump sound shows a strong peak at 102 Hz (see Fig. 6, top). This peak, which indicates the pitch of the thump, is due to the fundamental resonance of the soundboard. The same resonance can be seen as a minimum in the input impedance curve at the corresponding frequency (Fig. 6, middle). Also the following maxima (resonances) at 135, 150 and 165 Hz in the sound level curve correspond to minima in the impedance curve.

Fig. 6. Tap tone and influence of mass loading. The soundboard is tapped at the treble bridge (MP 9). Sound spectrum (top), input impedance (middle), and input impedance with a mass load (550 g) close to the measuring point (bottom). The vertical lines indicate the frequency of the fundamental resonance of the soundboard.

The input impedance curves can also be used to demonstrate changes in the properties of a soundboard. The resonances of the soundboard are controlled by the distribution of mass, stiffness, and damping. By changing the mass distribution of the soundboard in Fig. 6 with an additional mass at the treble bridge, we obtain a modified impedance curve (Fig. 6, bottom). The resonances below 100 Hz have not changed (properties of the shaker and beams), whereas the fundamental soundboard resonance has been shifted down from 102 Hz to 95 Hz. This is also easily heard by knocking on the soundboard.

Influence of string tension In experimental work with upright pianos, a complete assembly of the instrument is normally used. Such an assembly consists of the wooden frame, soundboard, and metal plate, including the strings tuned to nominal pitch. This means that the soundboard is under the same load as in the finished piano. However, great care must be taken to ensure that the strings are muted so they cannot vibrate during the measurements. For this purpose, narrow strips of felt are woven through the strings as mentioned above. This operation wastes valuable time, and the question therefore arose as to whether the experiments could be conducted on a soundboard without strings, or at least with the tension brought down, without seriously changing the results. To answer this question, measurements were conducted on a completely strung assembly of an upright piano for which the pitches were successively lowered (A4 = 440, 415, 220, and 0 Hz). Finally, the strings were removed altogether. By these changes, the load on the soundboard was progressively reduced to zero, which would change the bending stiffness of the soundboard. When the strings were removed, the vibrating mass was also changed. Plate theory tells us that both the magnitude of the impedance, as well as the resonance frequencies will change when these parameters are changed. The experiments showed that detuning the strings by a semitone, or even an octave, produces only a small change in bending stiffness. This can be seen in the impedance curves for the first three conditions, which are practically identical (see Fig. 7). Not until the tension was completely let down could an influence on the impedance curve be observed (Fig. 7 d). In the lower frequency region, the soundboard resonances were shifted downward. The lowest resonance (leftmost valley on the Zcurve) was shifted from 115 to 90 Hz when the string tension was let down completely, while the resonances above 200 Hz remained essentially uninfluenced. When the strings were completely removed, the lowest resonance was shifted further downward (to 70 Hz), and now the upper resonances were also lowered (Fig. 7 e). However, the general configuration of the impedance curve remained practically unchanged.

Fig. 7. Influence of string tension on the input impedance (MP 7): (a) Normal tuning (A4 = 440 Hz),

(b) a semitone lower (A4 = 415 Hz), (c) an octave lower (A4 = 220 Hz), (d) slack strings, and (e) strings removed.

The sound radiation curves show a somewhat different picture. Above approximately 1000 Hz the sound radiation does not change to any appreciable degree - the changes are evident only in the middle and lower frequency regions. Again, the general configuration of the curve remains practically unchanged.

Summing up, stringing the back and tuning the strings to playing pitch are not absolutely necessary for obtaining representative measurements of the properties of a soundboard. The basic characteristics of the impedance and sound radiation curves remain essentially unaffected when the string load is removed. Only the lower resonances are influenced, resulting in an upward shift of about 50 Hz when the strings are pulled up to pitch. This can, however, easily be taken into account when analyzing the data.

Influence of ribbing The purpose of the soundboard is to radiate a large volume of sound over a wide frequency range. This would imply that as much as possible of the energy of the string vibrations should be transferred to the soundboard. However, this is not the whole story. If the transfer of energy is too efficient, the decay of the tone will be too abrupt. In order to achieve a reasonable sustain, the soundboard must reflect a large portion of the vibration energy back to the strings. Soundboard design is thus a compromise. A stiffening of the soundboard would improve the sound radiation efficiency, as a stiff soundboard is less inclined to subdivide into small vibrating areas. One purpose of the ribs is precisely to stiffen the soundboard, which indeed is thin in proportion to its size. Another purpose is to "homogenize" the soundboard by equalizing the difference in bending stiffness (elasticity modulus) parallel to and across the grain. The moduli of elasticity in the two directions are in a ratio of approximately 20:1 (anisotropism). If this anisotropism is not compensated for by the addition of ribs running across the grain, the effective vibrating area is reduced and the radiation efficiency is decreased over a broad frequency band.

Rib height The ribs should increase the bending stiffness, but not, on the other hand, load the soundboard with too much extra mass. For this reason, experiments were made to determine the general influence of ribs on input impedance and sound radiation. The experiments were conducted on a soundboard for which the height of the ribs was reduced in steps of 25% of the original height. The ribs were planed down using an electrical router, without removing them from the soundboard. Theoretically the moment of inertia of the ribs determines their stiffness. The moment of inertia (I) of a rib, the cross section of which is assumed to be rectangular, is:

where b is the width and h is the height of the cross section. If the height is reduced without changing the width, the result is a rapid decrease in the moment of inertia and hence in the stiffening effect of the rib, while the mass decreases much slower (see Fig. 8)

Fig. 8. Changes in mass and stiffness of a rib of rectangular cross section as rib height is reduced.

The experiments showed that at the upper end of the treble bridge, no noticeable change in the input impedance occurred when the rib height was reduced. Here, the ribs do not play a decisive role, either in stiffening or in adding mass. Instead, the boundary conditions at the joint between the soundboard and the frame exert a greater influence. In the central region of the soundboard, on the other hand, we may observe the development of pronounced resonances (valleys) in the impedance curve below 200 Hz when the rib height is reduced (see Fig. 9). The resonance frequencies are gradually shifted downwards with decreasing rib height. At the same time, the resonances become more sharply defined. The difference in level between peaks and valleys increases, while the average value for the impedance decreases below 200 Hz. Above 500 Hz, the impedance curve is only slightly influenced, and no shift in the upper resonant frequencies can be observed. Fig. 9. Influence of ribbing on the input impedance (MP 7): (a) Original rib height (100 %) (b) 75 % of original (c) 50 % of original

(d) 25 % of original and (e) no ribbing.

No uniform relation was found between the successive stages of reduction of the rib height and the changes in the input impedance. The influence of the first reduction (from 100 to 75% rib height) was considerably less than that of the second (from 75 to 50%). The influence seems to be greater once the basic stiffening has been

removed and there is no longer any compensation for the anisotropism in the moduli of elasticity. Turning to the sound radiation, the same effects were observed as for the impedance; the lower resonant frequencies are shifted downwards and sound radiation is reduced. However, unlike the effect on the impedance, the reduction in sound radiation extends over the entire frequency range, even above 500 Hz. This is due to a short-circuiting phenomenon, as mentioned in the introduction. Above a certain frequency, the soundboard no longer vibrates as a unit but divides into a number of vibrating areas, which results in a poorer radiation efficiency. In addition, the lack of compensation for the anisotropism causes a reduction in the total vibrating area and hence a poorer radiation. From the measurements it can be concluded that the two most important functions of the ribs are to stiffen the soundboard and to compensate for the differences in bending stiffness parallel to and across the grain (anisotropism). The bending stiffness exerts a greater influence than the mass. For this reason it is more advantageous to use narrow, high ribs than ribs with a low and wide section.

Number of ribs We also investigated the changes in the acoustic properties of a soundboard when the number of ribs, rather than the rib height, is reduced. An upright assembly of the same type as in the experiment with rib height was used. Every second rib was removed by routing it down to the level of the soundboard, and input impedance and sound radiation were measured. A comparison of these data with those of the previous experiment showed that a reduction of 50% in the number of ribs is equivalent to a 25% reduction in the rib height. This applied both to the input impedance and the sound radiation. The measurements clearly show that a change in the number of ribs exerts less influence on the acoustic properties of a soundboard than a change in rib height, all other parameters being equal.

Decay of the piano tone

As explained by Gabriel Weinreich in the preceding lecture, the decay of a piano tone is complex. The early decay, the so-called prompt sound, is of major importance in judging the tone quality. Therefore it seems reasonable to measure the decay time at the beginning of the note, from the maximum level to -20 dB (T20). This decay should depend mainly on the soundboard, more specifically on the impedance matching between the string and the soundboard motion in the vertical direction.

This is so, since the initial blow of the hammer sets the string in vibration mainly in this direction. To test this hypothesis, a complete assembly with frame, soundboard and plate was fitted with a single string at a few measuring points. The impedance matching was varied by tuning the string to frequencies at and close to a soundboard resonance (see Fig. 10). The change in the impedance matching was registered by observing the change in the decay time of the fundamental when the string was plucked. Starting with a string on the treble bridge, the string was first tuned to D4 = 294 Hz (point 1 in Fig. 10, top) and the string tension was then slowly let down to 210 Hz (point 3). By means of these detunings, we were able to compare the decay times at peaks, valleys and at intermediate points with the same impedance value. Of the two intermediate points, one is located on an upward slope of the curve corresponding to a positive phase angle (point 5), while the other is on a downward slope with a negative phase angle (point 4).

Fig. 10. Changing the impedance matching between string and soundboard by detuning; treble bridge MP 8 (top), and bass bridge MP 12 (bottom). Numbered points indicate the successive frequencies to which a single string was tuned.

From these experiments, a number of observations were made. As expected, the decay time for the string vibrations is considerably longer at a peak in the impedance curve (off resonance), than in a valley (at resonance). Further, the decay is longer when the phase angle is positive than when the phase angle is negative. Hence, a good match between the string and the soundboard is found when the input impedance is high and the phase angle is positive. The same results were obtained for a string on the bass bridge (see Fig. 10, bottom).

Sharp valleys (resonances) in the impedance curve exert the strongest negative influence on the decay time. Consequently, one characteristic of a high quality soundboard is that the impedance curve exhibit as few dips as possible, avoiding sharp zigzagging. In addition, the overall level of the impedance curve should be high enough to ensure sufficient reflexion of the string energy at the bridge, resulting in a long decay time. The decay times for the notes on a small upright piano are plotted in Fig. 11. As can be seen, the decay times vary drastically among the notes. In order to judge the perceptual consequences of the irregular curve we may compare two neighbouring notes, F#4 and G4. From the graph, we obtain the corresponding T20-values as 3.5 and 0.7 s, respectively. This large difference in decay time (5:1) suggests an aggravating difference in sustain between these two adjacent notes.

Fig. 11. Decay times (T20) for the notes of a small upright piano, corresponding to a 20 dB drop from maximum sound level

Another example of the consequences of the irregular decay times of this particular piano is the differences between the two chords C3 - C4 - G4 (3.8, 1.7, 0.7 s) and D3 D4 - A4 (3.4, 3.3, 2.3 s). These relations between the decay times mean that the first chord will consist mainly of the lowest note (C3) shortly after the attack, since the two higher notes will fade away very quickly. In contrast, the spectrum of the second chord will not change as fast, sounding brighter and fuller compared with the first chord.

Modal analysis

The results presented so far have illustrated that the measurements of input impedance and sound radiation are influenced by the position of the point of excitation. In order to understand this influence we need to know how the soundboard vibrates. Modal analysis is a convenient method for this purpose. With this method, it is possible to map the vibration modes (resonances), and to measure their frequencies and damping. The motion of the soundboard when vibrating at the individual modes can also be observed in slow motion on a computer display, which gives a good understanding of how the soundboard vibrates.

Fig. 12. Modal analysis, sketch of measurement method.

The preliminary measurement data from the analyzer (the complex transfer functions) are later transmitted to a small computer, which calculates the motion of the soundboard at the resonances and shows an animated picture of the vibrating soundboard in slow motion. By a proper choice of perspective and gain, the vibration distribution of the entire soundboard can be seen clearly.* * During the lecture, video recordings of the vibrating soundboard were shown. In the figures that follow, the soundboard resonances are illustrated by "snapshots" of the deflections at the positive and negative maxima, respectively.

Method The principle of modal analysis is explained in the block diagram in Fig. 12. First, a net of measuring points are marked on the structure under investigation - in this case the soundboard of a grand piano. The measuring points are then excited in succession by blows of a small hammer, featuring a built-in force gauge. A small contact microphone (accelerometer), which is fixed to the soundboard at a certain

point, registers the resulting vibrations. The electrical signal from the accelerometer is sent to a computerized analyzer together with the force signal from the hammer.

Soundboard resonances Using the technique described above, the modes (resonances) of the soundboard of a concert grand (length 2.90 m), with the iron plate and strings included, were investigated (see Fig. 13 - 16).

Fig. 13. First and second mode of the soundboard of a concert grand piano (length 2.9 m).

The first mode has a resonance frequency of 62 Hz, and the deflection distribution shows only one vibration maximum with equal phase all over the soundboard (see Fig. 13, left). The vibration maximum is located in the front left third of the soundboard. The treble area at the front end of the treble bridge hardly vibrates at all. In the second mode at 90 Hz, the front half vibrates in opposite phase compared to the far half, with a zone of small deflection in the middle between these areas (Fig. 13, right). In this zone there is a nodal line running parallel to the keyboard along

the middle of the soundboard. (Modes with this orientation of the nodal lines are sometimes referred to as longitudinal modes.)

Fig. 14. Third and fourth modes of the same soundboard.

The third mode at 105 Hz (Fig. 14, left) is not a following third longitudinal mode, but a transversal second mode. In the third longitudinal mode, three sections with antiphase motions, separated by two transversal nodal lines, should develop. This third mode at 105 Hz has the same deflection distribution as the second mode at 90 Hz, except that the nodal line now runs longitudinally (perpendicular to the keyboard). Because of the higher stiffness of the treble area, an asymmetry exists in favor of the bass section. The fourth mode at 127 Hz is again a longitudinal mode - four vibrating areas moving in opposite phases divided by three transversal nodal lines (Fig. 14, right).

For the following two modes (187 and 222 Hz), the low stiffness of the left part of the soundboard causes a strong motion of this area compared to the right part (see Fig. 15). In addition, the left part is divided into three and four zones of vibrations, respectively. In the following two modes (245 and 325 Hz), vibrating areas can still be found and identified (see Fig. 16), but for higher modes the vibration zones become harder to recognize.

Fig. 15. Fifth and sixth modes.

Fig. 16. Seventh and eighth modes.

Important information can be extracted from the modal analysis with regard to the input impedance for different strings, and the decay of the corresponding notes. For example, a mode pattern with a large vibration amplitude at a certain point on the bridge implies a low impedance and hence an efficient energy transfer from a string crossing the bridge at this particular point. This means that the decay time for partials close to this mode frequency will be relatively short.

Simulation of changes in the design

I will end my presentation with a short discussion of how modal analysis can be used to predict the decay times. The modal analysis program models the properties of the real soundboard by a set of hypothesized mathematical resonating systems (with one degree of freedom). Any changes in the properties of the soundboard can

be simulated in the computer by changing the values of the mass, the stiffness, and the damping in the measuring points. The technique is illustrated in Fig. 17, which shows the deflection of a soundboard at its second mode of 90 Hz (upper panel). The largest vibration amplitude is found in the front left section and the smallest around the nodal line (broken line). The nodal line crosses the bridge near a point (0) where one of the bass strings delivers its vibration energy to the soundboard. The small vibration amplitude at that point indicates a high input impedance to the string and consequently a very long decay time of the prompt sound.

Fig. 17. Second mode at 90 Hz of a grand piano soundboard (top), and as modified in the computer model, at 101 Hz (bottom). The change in the modal pattern and frequency is due to a simulated increase in stiffness at the front left corner (shaded area). Nodal lines are marked with broken lines. The circles indicate a point on the bridge where a certain bass string crosses, see text.

If the stiffness of the front, left corner of the soundboard is increased in the computer model (lower panel, shaded area), the shape of the second mode will change. The frequency will be shifted up to 101 Hz, and the nodal line turned toward the longitudinal direction of the soundboard. The motion at the front left section will decrease as would be expected. In contrast, the vibration amplitude at the point

where the bass string crosses the bridge will increase. This means that the early decay time will drop as intended - how much is, however, harder to predict. This example illustrates how modal analysis followed by computer modifications can assist in the development of pianos, by predicting the general effects of proposed changes in the design. By means of this technique, the need for expensive test prototypes can be minimized, although as yet, not completely excluded.

References
Wogram, K. (1984): "Akustische Untersuchungen an Klavieren," in H. Junghanns: Der Piano- und Flgelbau (Verlag Das Musikinstrument, Frankfurt); English version "Acoustical Research on Pianos: Vibrational Characteristics of the Soundboard," Das Musikinstrument, Vol. 24, pp. 694-702, 776-782, 872-880 (1980).

About the author

Klaus Wogram received his doctorate (Dr.-Ing.) at the Technical Faculty at the University of Braunschweig in 1972. From the beginning of his professional career he has been with the Acoustics Department of the Federal Institute of Physics and Technology (Physikalisch-Technische Bundesanstalt) in Braunschweig, West Germany, where he became the head of the Musical Acoustics Laboratory in 1985. His research has dealt particularly with the brass instruments, an interest which he gained in his early years and has been continuously maintained by the playing of the jazz trombone.