Typeset by REVTEX 4 for JASA(A,Y)

Touch and temporal behavior of grand piano actions

Werner Goebl
Austrian Research Institute for Artificial Intelligence (OFAI), Freyung 6/6, 1010 Vienna, Austria

Roberto Bresin
Department of Speech, Music, and Hearing (TMH); Royal Institute of Technology (KTH),
Lindstedtsvägen 24, 10044 Stockholm, Sweden

Alexander Galembo
Sechenov Institute of Evolutionary Physiology and Biochemistry, Russian Academy of Sciences,
M. Toreza av. 44, St. Petersburg 194223, Russia

(Dated: May 10, 2005)

This study investigated the temporal behavior of grand piano actions from different manufac-
turers under different touch conditions and dynamic levels. An experimental setup consisting
of accelerometers and a calibrated microphone was used to capture key and hammer move-
ments, as well as the sound signal. Five selected keys were played by pianists with two types
of touch (“pressed touch” versus “struck touch”) over the entire dynamic range. Discrete
measurements were extracted from the accelerometer data for each of the over 2300 recorded
tones (e.g., finger–key, hammer–string, and key bottom contact times, maximum hammer
velocity). Travel times of the hammer (from finger–key to hammer–string) as a function of
maximum hammer velocity varied clearly between the two types of touch, but only slightly
between pianos. A travel time approximation used in earlier work [W. Goebl, J. Acoust.
Soc. Am. 110(1), 563–572 (2001)] derived from a computer-controlled piano was verified.
Constant temporal behavior over type of touch and low compression properties of the parts
of the action (reflected in key bottom contact times) were hypothesized to be indicators for
instrumental quality.
PACS numbers: 43.75.Mn, 43.75.St

I. INTRODUCTION The roller falls back to the repetition lever, while the
hammer is caught by the back check. For a fast rep-
The universe of expressive music to be played on the etition, the jack slides back under the roller when the
modern grand piano1 is produced by sophisticated accel- key is only released half-way, and the action is ready for
eration of the (usually) 88 keys, none of which travels another stroke. More detailed descriptions of the func-
through a distance greater than one centimeter, com- tionality of grand piano actions can be found in the liter-
bined with the use of the pedals. The piano action pro- ature (Askenfelt and Jansson, 1990b; Fletcher and Ross-
vides the pianist the only point of contact to the strings; ing, 1998, pp. 354–358).
it is therefore both an extremely important as well as a
highly elaborate and complex mechanical interface. It
allows accurate control over the speed at which the ham- A. Temporal properties of the piano action
mer arrives at the strings over a vast dynamical range Temporal aspects of the piano action have been in-
from the very pianissimo to the ultimate fortissimo. Since vestigated recently by Askenfelt and Jansson (1990a,b,
not only the intensity of tone, but also the precise onset 1991).2 The time interval from the key’s initial position
timing of the outcoming sound is crucial to expressive to key bottom contact ranges from about 25 ms at a
performance, it can be assumed that trained pianists are forte keystroke (approximately 5 m/s FHV) to 160 ms
intuitively well acquainted with the temporal behavior of at a piano tone (or 1 m/s final hammer velocity FHV,
a piano action, and that they take it into account while Askenfelt and Jansson, 1991, p. 2385).3 In a grand pi-
performing expressively. ano, the moments of hammer contact (when the hammer
The piano action functions as follows: The movement excites the strings) are temporally shifted in comparison
of the key is transferred to the hammer via the whip- to key bottom contact. Hammer contact occurs 12 ms
pen, on which the jack is positioned so that it touches before key bottom contact at a piano tone (1 m/s FHV),
the roller (knuckle) of the hammer shank. During a but 3 ms after the key bottom contact at a forte at-
keystroke, the tail end of the jack is stopped by the es- tack (5 m/s FHV, Askenfelt and Jansson, 1990a, p. 43).
capement dolly (let-off button, jack regulator) causing However, Askenfelt and Jansson (1990a,b, 1991) provided
the jack to rotate away from the roller, and thus break- measurement data solely for a few example keystrokes.
ing the contact between key and hammer. From this The timing properties of the piano action can be mod-
moment, the hammer travels to the strings with a small ified by changing the regulation of the action. Modi-
deceleration due to gravitation and friction, strikes them, fications, e.g., in the hammer–string distance or in the
and rebounds from them (“free flight of the hammer”). let-off distance (the distance of free flight of the ham-

To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior 1
mer, after the jack is released by the escapement dolly), solves itself finally, into one basic question: the variations
alter the timing relation between hammer–string contact of force produced at the key-surface by the player.” And
and key bottom contact or the free flight time, respec- in order to produce the forces at the key surface that en-
tively (Askenfelt and Jansson, 1990b, p. 57). Greater tail a particular desired musical outcome, pianists have
hammer mass in the bass range (Conklin, 1996, p. 3287) to practice for decades. Over this time period, they de-
influences the hammer–string contact durations (Asken- velop a tacit tactile knowledge of how piano actions be-
felt and Jansson, 1990b), but not the timing properties have under the various physical forces they apply to it.
of the action. Thus, an integral part of what pianists perceive from a
Data on timing properties of a baby grand piano ac- piano is the haptic-tactile response of the keys (includ-
tion were provided by Repp (1996) who worked with a ing particularly key resistance and inertia) in relation to
Yamaha Disklavier (Mark II series, similar to the one the physical force they apply and to the acoustical result
used in the present study) on which the “prelay func- they hear (Galembo, 2001).
tion” was not working.4 This gave him the opportunity An article in Nature by Bryan (1913) was the start-
to measure roughly a grand piano’s travel time character- ing point of a lively discussion on piano touch. Bryan
istics. He measured onset asynchronies at different MIDI puts the “single-variable hypothesis” (the timbre of a pi-
velocities in comparison to a note with a fixed MIDI ve- ano tone determined solely by FHV) into question with
locity in the middle register of the keyboard. The time rudimentary experiments performed with a player-piano.
deviations extended over a range of about 110 ms for His contribution entailed a discussion of 6 Letters to the
MIDI velocities between 30 and 100 and were fit well by Editor and 3 further replies by Bryan (all to be found in
a quadratic function (Repp, 1996, p. 3920). Nature, Volumes 91–92, 1913).
A function similar to the Disklavier’s prelay func- A first profoundly scientific investigation to this con-
tion was obtained from a Bösendorfer SE290 computer- troversy contributed Otto Ortmann from the Peabody
controlled grand piano by Goebl (2001). He accessed Conservatory of Music in Baltimore (Ortmann, 1925).
data from an internal memory chip of the SE system that He approached the “mystery of touch and tone” at the
presumably stored information on the travel time inter- piano through physical investigation. With a piece of
vals for each of the 97 keys and seven selected FHVs. An smoked glass mounted on the side of a piano key and
average travel time function (against FHV) was derived a tuning fork, he was able to record and to study key
from these data (Goebl, 2001, Fig. 5, p. 568). It was ap- depression under different stroke conditions. He inves-
plied to predict the amounts of note onset asynchronies tigated various kinds of keystrokes (“percussive” versus
to be expected at given dynamic differences between the “non-percussive,” different muscular tensions, and posi-
voices of a chord. tions of the finger). He found different acceleration pat-
There have been several attempts to model piano terns for non-percussive (finger rests on the surface of the
actions (Gillespie, 1994; Gillespie and Cutkosky, 1992; key before pressing it) and percussive touch (an already
Hayashi et al., 1999), also for a possible application moving finger strikes the key). The latter starts with a
in electronic keyboard instruments (Cadoz et al., 1990; sudden jerk, thereafter the key velocity decreases for a
Van den Berghe et al., 1995). Gillespie and colleagues moment and increases again. During this period, the fin-
developed a virtual keyboard that simulates the haptic ger slightly rebounds from the key (or vice versa), then
feel of a real grand piano action using motorized keys re-engages the key and “follows it up” (p. 23). On the
(Gillespie, 1994; Gillespie and Cutkosky, 1992). Van den other side, the non-percussive touch caused the key to
Berghe et al. (1995) performed measurements on a grand accelerate gradually.
piano key with two optical sensors for hammer and key Ortmann (1925) found that these different types of
displacement and a strain gauge for key force. Unfortu- touch provide a fundamentally different kind of key con-
nately, they provided only a single exemplary keystroke trol. The percussive touch required precise control of the
of their data. Hayashi et al. (1999) tested one piano key very first impact, whereas with non-percussive touch, the
on a Yamaha grand piano. The key was hit with a spe- key depression needed to be controlled up to the very
cially developed key actuator able to produce different end. “This means that the psychological factors involved
acceleration patterns. The displacement of the hammer in percussive and non-percussive touches are different”
was measured with a laser displacement gauge. They (p. 23). “In non-percussive touches key resistance is a
developed a simple model and tested it in two touch con- sensation, in percussive touches it is essentially an image”
ditions (with constant key velocity and constant key ac- (p. 23, footnote 1). His conclusions were that different
celeration). Their model predicted the measured data for ways of touching the keys produced different intensities
both conditions accurately. of tones, but when the intensity was the same, also the
quality of the tone must be the same. “The quality of a
B. Different types of touch sound on the piano depends upon its intensity, any one
While physicists and technicians argue that the sole degree of intensity produces but one quality, and no two
factor that controls the sound and the timbre of the pi- degrees of intensity can produce exactly the same qual-
ano is the hammer velocity at which the hammer hits ity” (p. 171).
against the strings (Hart et al., 1934; Seashore, 1937; The discussion was enriched by introducing the aspect
White, 1930), it is of extraordinary importance for pi- of different noises that emerge with varying touch (Báron
anists, how they touch and accelerate the keys. As Ort- and Holló, 1935; Cochran, 1931). Báron and Holló (1935)
mann (1929, p. 3) puts it: “The complex problem of distinguished between “Fingergeräusch” (finger noise)
physiological mechanics as applied to piano technique re- that occurs when the finger touches the key (which is ab-

2 To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior
sent when the finger velocity is zero as touching the key key touches the keybed. At a very pianissimo tone, the
— in Ortmann’s terminology “non-percussive touch”), force hardly exceeds 0.5 N.
“Bodengeräusch” (keybed noise) that emerges when the In the piano acoustics literature, there are other ways
key hits the keybed, and “Obere Geräusche” (upper of categorizing touch, as e.g., Suzuki (2003) who intro-
noises) that develop when the key is released again (e.g., duced a “hard–soft” antagonism, relying on a profes-
the damper hitting the strings). As another (and in- sional pianist’s intuition how this distinction is realized
deed very prominent) source of noise they mentioned the on the piano. However, he did not control for this vari-
pianist’s foot hitting the stage floor (or the pedals) in or- able (human factor) in his experiments.
der to emphasize a fortissimo passage. In a later study, In the present study, two prototypical types of depress-
Báron (1958) advocated a broader concept of tone qual- ing the keys are used based on the criterion of the finger’s
ity, including all kinds of noise (finger–key, action, and speed when beginning the keystroke. These two types
hammer–string interaction), which he argued to be in- (“struck touch” and “pressed touch”) are identical to the
cluded into concepts of tone characterization of different categories introduced by Askenfelt and Jansson (1991).
instruments (Báron, 1958). However, the terminology was deliberately changed from
More recent studies investigated these different kinds “legato–staccato” (that was also used in earlier studies
of noise that emerge when the key is struck in different by the authors, i.e., Goebl and Bresin, 2003a,b; Goebl
ways (Askenfelt, 1994; Koornhof and van der Walt, 1994; et al., 2003) to “pressed–struck” (Goebl et al., 2004) in
Podlesak and Lee, 1988). The hammer–impact noise order to draw a clear distinction between terms referring
(“string precursor”) arrives at the bridge immediately to touch and those referring to articulation (that is the
after hammer–string contact (Askenfelt, 1994) and char- length of each tone relative to its nominal value in the
acterizes the “attack thump” of the piano sound without score, thus referring to the connection of tones). Espe-
which it would not be recognized as such (Chaigne and cially in conversations with performing musicians, they
Askenfelt, 1994a,b). This noise is independent of touch get very quickly confused by “legato–staccato” used in
type. The hammer impact noises of the grand piano do a double sense. Nevertheless, there might be parallels
not radiate equally strongly in all directions (Bork et al., between these two meanings of “legato–staccato.” E.g.,
1995). As three dimensional measurements with a two- articulated and short tones may be more likely played
meter Bösendorfer grand piano revealed, increased noise with a struck touch and legato tones smoothly overlap-
levels were found horizontally towards the pianist and in ping with each other might be more likely played from
the opposite direction, to the left (viewed from the sitting the key surface (pressed touch). However, these parallels
pianist), and vertically towards the ceiling. occur only in very typical situations; pianists will have
Before the string precursor, another noise component no difficulty in producing opposite examples, e.g., a short
could occur: the “touch precursor,” only present when staccato tone played with a pressed touch and vice versa.
the key was hit from a certain distance above (“staccato
touch,” Askenfelt, 1994). It precedes the actual tone by
20 to 30 ms and was much weaker than the string pre- II. AIMS
cursor (Askenfelt, 1994). Similar results were reported by The present study aimed to collect a large amount of
Koornhof and van der Walt (1994). They called the noise measurement data from different grand pianos, different
prior to the sounding tone “early noise” or “acceleration types of touch, and different keys, in order to determine
noise;” it occurs closely in time with finger–key contact. and provide benchmark functions that may be useful in
They performed an informal listening test with four par- performance research as well as in piano pedagogy. The
ticipants. The two types of touch (“staccato touch” with measurement setup with accelerometers was similar to
the early noise and “legato touch”) could be easily iden- that as used by Askenfelt and Jansson (1991). However,
tified by the listeners, but not anymore with the early in order to obtain a large and reliable data set, the data
noise removed. Unfortunately, no further systematic re- processing procedure and the reading of discrete values
sults were reported (Koornhof and van der Walt, 1994). was automated with purpose-made computer software.
In a recent perception study (Goebl et al., 2004), mu- Each of the measured notes was equipped with two ac-
sicians could hardly identify what type of touch piano celerometers monitoring key and hammer velocity. Ad-
tone samples were played when the finger–key noises ditionally, a microphone recorded the sound of the pi-
were included (only half of them rated significantly bet- ano tone. With this setup, various temporal properties
ter than chance), but not at all, when finger–key noises were determined and discussed (travel time, key bottom
were removed from the stimuli. This evidence suggests time, time of free flight). Moreover, the speed histories of
that finger–key noise, which occurs only with a percus- both key and hammer revealed essential insights into the
sive (“struck”) touch, is responsible for pure aural touch fundamentally different nature of the two types of touch
recognition. examined in this study.
The different kinds of touch also produced different In a study on tone onset asynchronies in expressive
finger–key touch forces (Askenfelt and Jansson, 1992b, piano performance (“melody lead,” Goebl, 2001), finger–
p. 345). A mezzo forte attack played with “staccato key onset times were inferred from the hammer–string
touch” typically has 15 N, very loud such attacks show onset times through an approximation of the travel times
peaks up to 50 N (fortissimo), very soft touches go as low of the hammer (from finger–key to hammer–string con-
as 8 N (piano). Playing with “legato touch,” finger–key tact) as a function of FHV. This travel time function was
forces of about one third of those obtained with “stac- obtained from data of an internal chip of a Bösendorfer
cato touch” are found, usually having a peak when the SE290 reproducing system. The present study addition-

To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior 3
ally aims to reconsider that approximation. distance above, thus with the finger touching the key al-
ready with a certain speed (struck touch). Parallel to the
III. METHOD accelerometer setting, the two computer-controlled grand
pianos recorded these test tones with their internal de-
A. Material vice on computer hard disk (Bösendorfer) or floppy disk
Three grand pianos by different manufacturers were (Disklavier).
investigated in this study. Two of them were computer- For each of the five keys, both players played in both
controlled pianos, the same as in an earlier study (Goebl types of touch from 30 to 110 individual tones, so that a
and Bresin, 2003b). sufficient amount of data was recorded. In case of the two
computer-controlled devices (Bösendorfer and Yamaha),
1. Steinway grand piano, (model C, 225 cm, se- the internally recorded file was reproduced by the grand
rial number: 516000, built in Hamburg, Germany, piano immediately after each recording of a particular
in 1989),5 situated at the Dept. of Speech, Music, key, and the accelerometer data was recorded again onto
and Hearing (TMH) at the Royal Institute of Tech- the multi-channel DAT recorder. However, for the sake of
nology (KTH) in Stockholm, Sweden. clarity and due to limited space, we restricted this paper
2. Yamaha Disklavier grand piano (DC2IIXG, to the human data. For the Steinway, 595 individual
173 cm, serial number: 5516392, built in Japan, attacks were recorded, for the Yamaha Disklavier 996,
approximately 1999), situated at the Dept. of Psy- and for the Bösendorfer 756 (not counting the keystrokes
chology at the University of Uppsala, Sweden. repeated by the reproducing devices).11
3. Bösendorfer computer-controlled grand pi- D. Data analysis
ano (SE290, 290 cm, internal number: 290–3, In order to analyze the three-channel data files, dis-
built in Vienna, Austria, 2000), situated at the crete measurement values had to be extracted from them.
Bösendorfer Company in Vienna. Several instants in time were defined as listed below and
automatically read off with the help of Matlab scripts
Immediately before the experiments, the instruments prepared by the first author for this purpose. This
were tuned, and the piano action and—in the case of the method allowed to obtain timing data without specially
computer-controlled pianos—the reproduction unit ser- having to install additional sensors or contacts into the
viced and regulated. The Steinway grand has been regu- piano action (as e.g. done by Askenfelt and Jansson,
larly maintained by a piano technician of the Swedish Na- 1990b), only by processing the key and hammer trajec-
tional Radio. At the Disklavier, this procedure was car- tory and the sound information.
ried out by a specially trained Disklavier piano technician The hammer–string contact was defined as the mo-
from the Stockholm “Konserthus.” At the Bösendorfer ment of maximum deceleration (minimum acceleration)
company, the company’s SE technician took care of this of the hammer shank (hammer accelerometer) which cor-
work. responded well to the physical onset of the sound, and
conceptually with the “note on” command in the MIDI
B. Equipment and calibration file.12
The tested keys were equipped with an accelerometer The finger–key contact was defined to be the mo-
on the key6 and another one on the bottom side at the ment when the key started to move. It was obtained by
end of the hammer shank.7 The sound was picked up a simple threshold procedure applied on the key veloc-
by a sound-level meter microphone8 placed about 10 cm ity track. In mathematical terms, it was the moment
above the strings. The velocities of key and hammer and when the (slightly smoothed) key acceleration exceeded
the sound signal were recorded on a multichannel DAT a certain threshold. Finding the correct finger–key point
recorder (TEAC RD-200 PCM) with a sampling rate of was not difficult for struck tones; they showed typically a
10 kHz and 16-bit word length. The data was transferred very abrupt initial acceleration. However, automatically
to a computer harddisk and analyzed with computer soft- determining the right moment for soft pressed tones was
ware written for this purpose. The recorded voltages more difficult and sometimes ambiguous. The thresh-
were transformed to obtain required measures (m/s and old was optimized iteratively by hand. It was found
dB SPL). The measuring equipment and the calibration that softer tones required a smaller threshold than louder
procedure was identical as in Goebl and Bresin (2003b, ones; therefore it was coupled to the hammer velocity by
p. 2274–75), so we do not repeat further details here.9 a linear function. When the automatic procedure failed,
it failed by several tens of milliseconds—an error easy to
C. Procedure discover in explorative data plots.
Five keys distributed over the whole range of the key- The key bottom contact was the instant when the
board were tested: C1 (MIDI note number 24, 32.7 Hz), downwards travel of the key was stopped by the keybed.
G2 (43, 98.0 Hz), C4 (60, 261.6 Hz), C5 (72, 523.3 Hz), This point was defined as the maximum deceleration of
and G6 (91, 1568.0 Hz).10 The first two authors served the key (MDK). In some keystrokes, the MDK was not
as pianists to perform the recorded test tones. Each the actual keybed contact, but a rebound of the key after
key was hit at as many different dynamic levels (ham- the first key bottom contact. For this reason, the time
mer velocities) as possible, with two different kinds of window of searching MDK was restricted to 7 ms before
touch: one with the finger resting on the key surface and 50 ms after hammer–string contact. The time win-
(pressed touch), the other hitting the key from a certain dow was iteratively modified depending on the maximum

4 To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior
 fk kb hs fk kb hs
Amplitude (−1/+1) Hammer velocity (m/s) Key velocity (m/s)

Amplitude (−1/+1) Hammer velocity (m/s) Key velocity (m/s)

0.5 0.5

0 0

−0.5 hs−fk: 43.8 ms −0.5 hs−fk: 24.7 ms

kb−hs: −2.7 ms kb−hs: −0.2 ms

2 2

0 0

−2 −2
maxHv: 2.976 m/s maxHv: 2.975 m/s

0.2 0.2

0.1 0.1

0 0

−0.1 −0.1

−0.2 SPL: 99.32 dB −0.2 SPL: 98.87 dB

−50 −40 −30 −20 −10 0 10 20 −50 −40 −30 −20 −10 0 10 20
Time (ms) Time (ms)
a) b)

FIG. 1. Two forte keystrokes (C4, 60) played on the Yamaha grand piano with a pressed touch (left) and struck touch (right).
The top panels show key velocity, the middle panel hammer velocity, and the bottom panel the amplitude of the sound signal.
Both keystrokes exhibit similar MHVs and peak sound levels. Finger–key contact time (“fk”), hammer–string contact (“hs”),
and key bottom contact (“kb”) are indicated by vertical lines.

hammer velocity until the correct instant was found. The pend strongly on their regulation so that generalization
indicator MDK was especially clear and non-ambiguous to other instruments of the same brands may be prob-
when the key was depressed in a range of medium inten- lematic.
sity (see Fig. 1).
The maximum hammer velocity (MHV, in meters A. Two types of touch
per second) was the maximum value in the hammer ve-
locity track before hammer–string contact. The recorded three-channel data of two example
The escapement point was defined as being the in- keystrokes performed on the Yamaha grand piano are
stant after which the hammer travels freely (with no fur- plotted in Fig. 1. They both exhibit an almost identical
ther acceleration) towards the strings. It was approxi- MHV (2.976 and 2.975 m/s, respectively) and a similar
mated by fitting a line onto the hammer velocity track peak sound level (99.32 and 98.87 dB, resp.).13 In mu-
between the point of MHV and hammer–string contact. sical terms this corresponds roughly to a forte dynamic.
The slope of this line was set to the theoretical deceler- The first keystroke (Fig. 1a) was played from the key with
ation caused by gravity (−9.81 m/s2 ), disregarding any a “pressed touch.” From the beginning of the keystroke,
influence of friction. This instant in time was measurable the key velocity increases gradually; the hammer veloc-
only at soft and very soft touches. At MHVs exceeding ity grows in parallel. The hammer reaches its maxi-
approximately 1.5 m/s, it virtually coincided with the mum velocity immediately before it arrives at the strings.
moment of MHV. Hammer–string contact is characterized by a very sud-
To inspect the recorded key and hammer velocity den deceleration (Fig. 1, indicated by vertical solid lines).
tracks and the sound signal, an interactive tool was cre- Key bottom contact shows a slightly less abrupt deceler-
ated in order to display one keystroke at a time in three ation and occurs immediately before hammer–string con-
panels, one above the other. The user could click to the tact. On the other hand, the keystroke produced with a
next and the previous keystroke, zoom in and out, and “struck touch” (Fig. 1b) shows a very sudden jerk at the
change the display from velocity to acceleration or dis- beginning of the key movement that has no correspon-
placement. Screen shots of this tool are shown below (see dence in the hammer movement, but can be seen in the
Fig. 1). The data was checked and inspected for errors audio data (“touch precursor”).14 The hammer starts its
with the help of this tool. travel to the strings with a delay of several milliseconds.
It receives a first, larger acceleration by this initial blow
applied to the key; later the key “catches up” (Ortmann,
IV. RESULTS AND DISCUSSION 1925, p. 23) and brings the hammer to its final speed.
In this section, measurement results of the three in- The whole striking procedure needs roughly 20 ms less
vestigated pianos are presented and compared. Recall time with a “struck touch” compared to the “pressed
that these data apply to specific instruments and de- touch,” both with almost identical intensities.

To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior 5
 Pressed Pressed
4 f
1 mp
3
0.5 2

0 1
mHvB = 1.151
mHvY = 1.160 0
−0.5 mHvS = 1.145 mHvB = 4.130
−1 mHvY = 4.151
mHvS = 4.119
Hammer velocity (m/s)

Hammer velocity (m/s)

−1 −2
−3
0 0.1 0.2 0.3 0 0.25 0.5 0.75
Struck Struck
4 f
1 mp
3
0.5 2

0 1
mHvB = 1.180
mHvY = 1.149 0
mHvS = 1.176 mHvB = 3.506
−0.5
−1 mHvY = 3.455
mHvS = 3.512
−1 −2
−3
0 0.1 0.2 0.3 0 0.25 0.5 0.75
Key velocity (m/s) Key velocity (m/s)

FIG. 2. Touch trajectories of mezzo-piano keystrokes at the FIG. 3. Touch trajectories of forte keystrokes at the C5 (72)
C5 (72) on three pianos played with pressed touch (upper on three pianos played with pressed touch (upper panel) and
panel) and struck touch (lower). Upward-pointing triangles struck touch (lower). Symbols and axes proportions as in
denote the escapement point, diamonds hammer–string con- Fig. 2.
tact, and downward-pointing triangles key bottom contact
times. Small filled circles are plotted on the trajectories in
2-ms intervals to indicate time.
gle), hammer–string contact (diamond), and key bottom
contact (downward triangle), as well as elapsing time
B. Relation between key and hammer movement (filled circles every 2 ms). Fig. 2 contains mezzo-piano
In order to demonstrate the behavior of the hammer keystrokes, Fig. 3 forte, and Fig. 4 fortissimo, which can
in relation to the key movement under two touch condi- only be achieved with a struck touch.
tions, so-called “touch trajectories” of pressed and struck The pressed keystrokes develop fairly linearly until the
keystrokes are plotted in Figs. 2–4. These plots depict the escapement point (top panels in Figs. 2 and 3). The av-
progression of hammer velocity against key velocity from erage slope of this part of the trajectory for all recorded
finger–key contact to key bottom or hammer–string con- pressed tones is 5.6 for Steinway and Yamaha, and 5.3
tact (depending on which of these two points was later). for the Bösendorfer. In the softer pressed example (Fig. 2
Each panel compares keystrokes with almost identical top), the key–bottom occurs after hammer–string contact
MHV values played at the C5 on all three pianos. Marked (the trajectories drop at the right before going leftwards),
on the trajectories are escapement point (upward trian- while at the forte example (Fig. 3 top) the key–bottom is

6 To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior
 Struck Pressed Struck
1 1

Mean correlation coefficients

mHvB = 6.868 G6 (91)
ff mHvY = 6.781 0.8 0.8
C5 (72)
C4 (60)
8 mHvS = 6.777 G2 (43)
Hammer velocity (m/s)

0.6 0.6 C1 (24)

6 0.4 0.4

0.2 0.2
4
0 0
S Y B S Y B

2 FIG. 5. Mean correlation coefficients of the touch trajectories

(key and hammer velocity histories from finger–key through
the escapement point) by touch (panels), piano (x axes: Stein-
0 way, Yamaha, Bösendorfer), and pitch (marker shape). Error
bars denote 95% confidence intervals.
0 0.4 0.8 1.2
Key velocity (m/s)
Bösendorfer requires more playing effort at struck touch
forms to achieve the same dynamic level than the two
other pianos.
FIG. 4. Touch trajectories of fortissimo keystrokes at the C5
(72) on three pianos played with struck touch. Symbols and
In order to quantify the transformation effectivity of
axes proportions as in Fig. 2. a keystroke, the correlation coefficient between the key
and hammer velocity track (starting from finger–key con-
tact until the escapement point) was introduced. The
mean correlation coefficients for touch, pianos, and key
before hammer–string (trajectory moves left before drop- are each plotted separately in Fig. 5. As this measure de-
ping downwards). The exception here is the keystroke at termines linearity between key and hammer movement,
the Steinway, at which the key–bottom contact is still it may serve as a “touch index,” distinguishing clearly
after hammer–string contact (though the time difference between the two types of touch. It turned out that all
is negligible) and therefore the trajectory still drops first. pressed touches display coefficients beyond approx. 0.6
Struck tones show very different trajectories (bottom and struck ones below that value. In this sense, pressed
panels in Figs. 2–4). They deviate clearly from the di- touch is a more effective way of transforming finger force
agonal; the initial acceleration of the key pushes the tra- into hammer velocity than playing with a struck touch.
jectories rightwards, before the hammer starts to move. Moreover, this index may also hold for a measure of
After this first blow, the key stops for a moment and re- tone control for the pianist. With a struck touch, the
accelerates again, while the hammer still gets faster. This action decompresses after compression (relaxing of com-
pattern is quite consistent across pianos and intensities pressed cushions, bent key, and the hammer shank). At
(see Figs. 2 and 3). However, at very loud keystrokes the very loud keystrokes, this must be the reason of the high
second acceleration of the key does not occur anymore acceleration of the hammer (the key decelerates clearly
(Fig. 4) so that the whole keystroke consists of one strong while the hammer accelerates up to 7–8 m/s, see Fig. 3).
impulse and the acceleration of the hammer during re- Therefore by striking a key, the tone intensity is con-
tardation of the key (diagonal trajectory to the upper trolled through the initial key or finger velocity; by press-
right). The exception is the Steinway which still exhibits ing a key, the tone intensity is controlled through the key
a second key acceleration phase. or finger velocity until the escapement of the jack (“early
The struck touches compared here display almost iden- versus late impulse,” cf. Askenfelt and Jansson, 1991).
tical MHVs (bottom panels Figs. 2–4). However, the
effort spent for the keystrokes does not appear to be sim- C. Travel time
ilarly identical: the initial blow (maximum key veloc- The time interval between finger–key contact and
ity) at the Bösendorfer is larger than at the other pianos hammer–string contact is defined here as the travel
for all three intensities; relatively the most in Fig. 2. A time.15 The travel times of all recorded tones are plot-
larger initial amplitude to the right denotes a larger en- ted in Fig. 6 against MHV separately for the three grand
ergy loss in a keystroke due to compression of the parts pianos (different panels), different types of touch (filling
in the action (e.g., cushions, dunnage) and bending of of symbols), and different keys (denoted by symbol).
the key and hammer shank. Therefore, the Bösendorfer Some very basic observations can be drawn from this
action exhibits the largest degree of compression (espe- figure. The two pianists were able to produce much
cially with soft tones) and the Steinway the least. To higher MHVs on all three pianos with a struck attack
draw more profound conclusions from this finding, mea- (almost 8 m/s), whereas with a pressed tone, the MHVs
surements on the touch form would have to be performed hardly exceeded 5 m/s. There was a small trend to-
that include monitoring of finger speed and force applied wards higher hammer velocities at higher pitches (due to
to the key. However, the present data suggests that the smaller hammer mass, see Conklin, 1996). The highest

To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior 7
 STEINWAY C YAMAHA DISKLAVIER BÖSENDORFER SE290
−0.7147 −0.5959 pr st
250 ttpr = 98.57 × HV 250 ttpr = 89.41 × HV 250 G6 (91)
ttst = 65.19 × HV −0.7268 ttst = 57.43 × HV −0.7748
C5 (72)
C4 (60)
200 200 200 G2 (43)
Travel time (ms)

C1 (24)

ttpr = 89.96 × HV −0.5595

150 150 150 ttst = 58.39 × HV −0.7377

100 100 100

50 50 50

0 0 0
0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Maximum hammer velocity (m/s)

FIG. 6. Travel time (from finger–key to hammer–string contact) against MHV for the three grand pianos (three panels),
different types of touch (pressed and struck ), and different keys (from C1 to G6, see legend). The two types of touch (“pr” and
“st”) were approximated by power functions (see legends).

velocities on the Yamaha and the Steinway were obtained D. Key bottom time
at the G6, but at the middle C on the Bösendorfer. The
lowest investigated key (C1) showed slightly lower MHVs Fig. 7 displays the key bottom contact times relative to
by comparison to the fastest attacks (loudest attacks on hammer–string contact (tkbrel = tkb − ths ). Negative val-
the Steinway: C1: 6.8 m/s versus G6: 7.5 m/s, on the ues indicate key bottom contacts before hammer–string
Yamaha: C1: 6.4 m/s versus G6: 7.8 m/s, and on the contact, positive values key bottom contacts after the
Bösendorfer: C1: 6.0 m/s versus G6: 6.6 m/s and C4: hammer hits the strings (see overview display in Fig. 9).
7.6 m/s). The variability between the intensity distribu- The keybed was reached by the key up to 35 ms after
tions of the keys could be due to the fact that the tones hammer–string contact in very soft tones (up to 39 ms
were played by human performers. at the Bösendorfer) and as early as 4 ms before in very
strong keystrokes. This finding coincides with Askenfelt
The travel times ranged from 20 ms to around 200 ms and Jansson (1990a,b)’s results, but since much softer
(up to 230 ms on the Steinway) and depicted clearly dif- tones were measured in the present study (as low as
ferent patterns for the two types of touch. The travel 0.1 m/s), the key bottom times extended more after
time curves were independent of pitch although lower hammer–string contact.
keys have much greater hammer mass than in the high
Key bottom contact times varied with the type of
register (Conklin, 1996).
touch. Keys played with a pressed touch tended to reach
The data plotted in Fig. 6 were approximated by power the keybed earlier than keys hit in a struck manner. This
curves of the form tt = a × HV b separately for each type was especially evident for the Bösendorfer and for the
of touch (“pr,” “st”) and each of the three pianos. The Yamaha, but not for the Steinway. Askenfelt and Jansson
results of these curve interpolations are shown in the leg- (1992b, p. 345) stated that the interval between key bot-
ends of Fig. 6. Struck touch needed less time to transport tom and hammer–string contact varies only marginally
the hammer to the strings than a pressed touch which between “legato” and “staccato touch.” They obviously
smoothly accelerated the key (and thus the hammer). refer with this statement to one of their earlier studies
The travel times were more spread out when the tones (Askenfelt and Jansson, 1990b), where the investigated
were pressed, indicating that there was a more flexible grand piano was also a Steinway grand piano.16
control of touch in this way of actuating the keys (also Power functions were fitted to the data as depicted in
reflected by the lower R2 values of the curve fits). On Fig. 7, separately for the two types of touch and the dif-
the Steinway, the struck data showed higher variability, ferent pianos (see legends). Since the data to fit contains
almost similar to the pressed data. also negative values on the y axis, power functions of the
The present data were generally congruent with find- form kbt = a × HV b + c were used. The data spread out
ings by Askenfelt and Jansson (1991) and Hayashi et al. more than in the travel time curves (reflected in smaller
(1999). The travel time approximations used in Goebl R2 values) and showed considerable differences between
(2001, tt = 89.16 × HV −0.570 ) were very similar to the types of touch, except for this Steinway, where touch did
curve fit of the Bösendorfer’s pressed data. The impact not divide the data visibly. This finding suggests that
of this updated travel time function on the melody lead struck keystrokes tend to compress the parts of the ac-
predictions was rather negligible; it is discussed elsewhere tion more than pressed ones that this behavior was least
(Goebl, 2003, p. 74 onwards). at the Steinway piano.

8 To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior
 STEINWAY C YAMAHA DISKLAVIER BÖSENDORFER SE290
40 40 40
−0.3936 −0.4158 pr st
kbtpr = 19.09 × HV −12.30 kbtpr = 14.63 × HV −11.05
35 35 35 G6 (91)
kbtst = 59.57 × HV −0.1131 −51.19 kbtst = 10.15 × HV −0.6825 −3.74
C5 (72)
Key bottom time (ms)

30 30 30 C4 (60)
G2 (43)
25 25 25 C1 (24)

20 20 20 kbtpr = 11.59 × HV −0.4497 −9.98

kbtst = 13.96 × HV −0.3559 −10.15
15 15 15

10 10 10

5 5 5

0 0 0

−5 −5 −5

−10 −10 −10

0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8 0 1 2 3 4 5 6 7 8
Maximum hammer velocity (m/s)

FIG. 7. Key bottom time relative to hammer–string contact against MHV. Negative key bottom values denote instants
preceding hammer–string contact. Legends list power curve fits of the data separately for pressed touch (“kbtpr ”) and struck
touch (“kbtst ”).

Askenfelt and Jansson (1990b) considered key bottom fk hs kb

times as being sensed with the fingertips by pianists and Amplitude (−1/+1) Hammer velocity (m/s) Key velocity (m/s)
0.05
thus as being important for the vibrotactile feedback in
piano playing. Temporal asynchronies of the order of
30 ms are in principle beyond the temporal order thresh- 0

old (Hirsh, 1959), so at very soft keystrokes key bottom

hs−fk: 126.9 ms
contact and hammer–string contact could be perceived as −0.05 kb−hs: 15.3 ms
two separate events by the pianists. But for the majority 0.5
of keystrokes these time differences are not perceptually Vmax ep
distinguishable; however, they may be perceived subcon-
sciously and perhaps as part of the response behavior of a 0
particular piano. Especially, the different key bottom be-
havior for the different kinds of touch might be judged by
maxHv: 0.338 m/s
the pianists as part of the response behavior of the action −0.5
(Askenfelt and Jansson, 1992b). Hammer–string contact 0.01
occurs earlier relative to key bottom contact when the
key was struck compared to when it was pressed. For
0
a pianist, a struck touch produces a tone earlier than a
pressed touch with comparable intensity, both relative to
key bottom contact and relative to finger–key contact, −0.01
SPL: 72.32 dB
and thus may be perceived as being louder and more di- −150 −100 −50 0 50
rect. Time (ms)

E. Escapement point
FIG. 8. A struck pianissimo keystroke at C2 (24) played
Shortly before the hammer crown arrives at the strings, on the Steinway grand piano. Additionally indicated are the
the tail end of the jack gets pushed away from under point of MHV (“Vmax”) and the estimated escapement point
the roller by the escapement dolly and the pianist loses (“ep”). The skewed line in the middle panel denotes expected
physical contact with and thus control over the hammer, deceleration according to gravity.
which is then moving freely along a circular path to the
strings. This measurement point was comparatively dif-
ficult to extract automatically from the data, since at
many keystrokes this point was not obvious at all. The of MHV (“Vmax”) and the escapement point (“ep”) are
higher the hammer velocities, the more it tends to coin- sketched, as well as the line of gravity fitted in the ham-
cide with the instant of MHV. Only at soft and very soft mer track between escapement point and hammer–string
dynamics, the hammer might reach its maximum speed contact. At this pianissimo tone, the hammer reaches
considerably earlier than the jack is escaped. its maximum speed quite soon after the begin of the
An example of a very soft keystroke is displayed in keystroke (after 31.9 ms), travels another 78.7 ms with
Fig. 8. In addition to the display in Fig. 1, the point connection to the jack, and decelerates approximately by

To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior 9
 the three tested piano actions (line style) against time
TABLE I. Power curve approximations of the form fft = a ×
(in seconds) relative to the hammer–string contact. The
HV b for the free-flight time data, separately for pressed and
struck touches and the three pianos. temporal differences between extremes in intensity were
largest for the finger–key times and smallest for key bot-
fft = Pressed Struck tom times (both relative to hammer–string contact). The
Steinway 1.63 × HV −1.403 3.04 × HV −1.581 differences of the curves between the pianos by different
Yamaha 2.78 × HV −1.266 5.27 × HV −1.384 manufacturers were small compared to the differences in-
Bösendorfer 3.21 × HV −1.353 4.32 × HV −1.404 troduced through the type of touch. The finger–key curve
of this Steinway action was the left-most except for loud
8 pressed tones. Also our Steinway’s key bottom curve was
Steinway C
7 Yamaha Disklavier
the right-most of the three actions. Thus, the Steinway
Bösendorfer SE290 Pressed touch action needed more time for the attack process than the
6

5 Key bottom
other two pianos, except for very loud pressed tones. At
the free flight time approximation, the Steinway showed
Maximum hammer velocity (m/s)

4
the shortest of all tested actions.
3 Finger−key
These data apply only to the tested instruments and
2 Esc. Point
temporal behavior changes considerably with regulation
1
(especially key–bottom contact and the time of free flight,
−200 −175 −150 −125 −100 −75 −50 −25 0 25 see Askenfelt and Jansson, 1990b; Dietz, 1968). We do
not know how different the temporal behavior of other in-
8
struments of these three manufacturers will be. Changes
7
Struck touch in regulation (hammer–string distance, let-off distance)
6 resulted in changes of the key–bottom timing and the
5 time interval of the hammer’s free flight, respectively, of
4 up to 5 ms (for a medium intensity, see Askenfelt and
3 Jansson, 1990b, pp. 56–57). The differences between pi-
2 ano actions in the present data are approximately of the
1
same order.18
It can be concluded that the temporal behavior of the
−200 −175 −150 −125 −100 −75 −50 −25 0 25 tested piano actions by different manufacturers were sim-
Time relative to hammer−string contact (ms) ilar. However, no definitive conclusions can be drawn
whether or not these (comparably small) differences in
FIG. 9. Temporal properties of three grand piano actions. temporal behavior were crucial for the pianist’s estima-
Power curve approximations for the three pianos (line style), tion of the piano’s quality and whether they apply also
the two types of touch (panels), and for finger–key (left), es- to other instruments of these manufacturers.
capement point (middle), and key bottom times (right) rela-
tive to hammer–string contact.
V. GENERAL DISCUSSION
This study provides benchmark data on the temporal
gravity for a period of 16.3 ms. properties of three different grand pianos under two touch
conditions (pressed and struck touch). Prototypical func-
F. Free flight of the hammer tions were obtained for travel time, key bottom time,
The time interval from the escapement point until and the time of the hammer’s free flight by fitting power
hammer–string contact is called here “the free flight of curves to measured data. The temporal properties varied
the hammer.”17 The individual data points are not plot- considerably between type of touch, only marginally be-
ted here due to space limitations, but they were fitted tween pianos, and not at all between the different tested
by power curves separately for the type of touch. The keys. The latter was not surprising, since piano tech-
formulas are provided in Table I. The free flight of the nicians generally aim to adjust a grand piano action so
hammer ranges from almost zero at louder tones up to that all keys show similar and consistent behavior over
20 ms at very soft keystrokes, with some outliers up to the whole range of the keyboard.
around 40 ms at the Yamaha piano (struck touch). Gen- Different kinds of actuating the keys produced differ-
erally, pressed touches exhibit shorter free flight times ent ranges of hammer velocity. Very soft tones could only
than struck touches. This finding coincides with the be achieved with a pressed touch (minimum 0.18 m/s or
above stated proposition that a pressed touch provides 50.0 dB–pSPL) and the extremely loud attacks only with
generally a better control over the tones than a struck a struck touch (maximum 6.8 m/s or 110.4 dB–pSPL).
touch. Of all three actions, the Steinway action showed Playing from the keys (pressed) did not allow MHVs be-
the shortest free flight times (see Fig. 9). yond around 4 m/s, thus for some very loud intensities
hitting the keys from above was the only possible means.
G. Comparison among tested pianos The free flight times were shorter for pressed touch than
In Fig. 9, all power curve approximations reported for struck touch which suggests a better tone control for
above (Figs. 6, 7, and Table I) are plotted in a single the pianist when playing with pressed touch. Moreover,
display, separately for the type of touch (panels) and depressing a key caused less touch noise than striking a

10 To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior
key which is commonly regarded as a desired aesthetic is of the order of or just beyond just noticeable differ-
target in piano playing and teaching (cf. e.g., Gát, 1965). ences for perceiving two separate events (Askenfelt and
The two types of touch (in the present terminology Jansson, 1992b, p. 345). Also as Figure 9 made visu-
pressed and struck touch) do represent two poles of a va- ally evident, the travel times were far larger than the
riety of possible ways to actuate a piano key (i.e., late time differences of the other readings (escapement point,
acceleration versus early, hesitating in between, or ac- hammer–string contact, key bottom contact), so it can
celerating directly at the escapement point). It must be assumed that the pianist (especially in the dynamic
be assumed that a professional pianist will (even un- middle range) only senses two points in time: the start
consciously) be able to produce many different shades of the keystroke (finger–key) and its end which coincides
of touch between pressed and struck. with the beginning of the sound.
The travel times and the key bottom times changed Conceptually, the key bottom contact has to be after
considerably with intensity of key depression. A soft hammer–string contact. If it were the other way round,
tone may take over 200 ms longer from the first actu- no soft tones could be played at all. The fact that key
ation by the pianist’s finger to sound production com- bottom contact moves towards and beyond (that is be-
pared to a very sudden fortissimo attack. Moreover, fore) hammer–string contact with increasing hammer ve-
travel times and key bottom times changed considerably locity (cf. Fig. 7) was due to the bending of the hammer
with touch. A struck tone needed around 30–40 ms less shank and the compression of various parts in the ac-
from finger–key to hammer–string than a pressed tone tion (i.e., cushions, dunnage) and a later unbending and
with a similar MHV. These findings were not surprising decompression of those. The three actions showed dif-
(since they follow a very basic physical law), but the per- ferent behavior as to when the key bottom line crossed
forming artist has to anticipate these changes in tempo- the hammer–string line with changing hammer veloc-
ral behavior while playing in order to achieve the desired ity (Fig. 9). According to the power curve approxima-
expressive timing of the played tones. The pianist not tions of the data (Fig. 7), the Bösendorfer’s crossed at
only has to estimate before playing a tone, how long the 1.4 m/s (pr) and 2.5 m/s (st), the Yamaha’s at 2 m/s
keystroke will take for what desired dynamic level, but (pr) and 4.3 m/s (st), and the Steinway’s at 3 m/s (pr)
also for what intended way of actuating the key. These and 3.8 m/s (st). If we considered these values to be
complex temporal interactions between touch, intensity a measure of compressivity of the action, the Steinway
and the tone onset are dealt with and applied by the would be the least compressive action (of the three), and
pianist unconsciously; they are established over years of the Bösendorfer the most for pressed touches. At struck
intensive practising and extensive self-listening. Imme- touches, the Yamaha showed the least compressive be-
diately, musical situations come to mind in which loud havior. A smaller compression behavior might be con-
chords tend to come early with pianists at beginning or sidered a criterion for the subjective quality of a piano
intermediate level; or that crescendo passages tend to action (see discussion further below). However, further
accelerate in tempo as well, because each keystroke is investigation would be necessary to verify this hypoth-
performed with a harder blow and thus quicker in order esis (e.g., measuring static compression behavior of the
to achieve the crescendo, but the time intervals between investigated keys).
finger activity were not correspondingly increased. Furthermore, senso-motoric feedback is considered an
A keystroke starts for the pianist kinesthetically with utmost important factor for pianists not only for judging
finger–key contact (the acceleration impulse by the the action’s response, but also to judge the piano’s tone
finger)19 and ends at key bottom, but it starts aurally for (Galembo, 1982, 2001). In an extended perception exper-
pianist and audience at (or immediately after) hammer– iment, Galembo (1982) asked a dozen professors from the
string contact. Typical intensities (at an intermediate Leningrad Conservatory of Music to rate the instrumen-
level) in expressive piano performances (i.e., as measured tal quality of three grand pianos under different condi-
in Goebl, 2001) fall between 40 and 60 MIDI velocity tions. The participants agreed that the Hamburg Stein-
units (0.7 to 1.25 m/s) and thus typical travel times are way grand piano was superior, followed by the Bechstein
between 80 and 108 ms, thus varying as much as about grand piano, while the lowest quality judgment received a
30 ms. At such keystrokes, the key bottom times are grand piano from the Leningrad piano factory. In differ-
between 3.5 and 0.5 ms before hammer–string contact, ent discrimination tasks, the participants were not able
thus a range of the order of 3 ms. It can be assumed here to distinguish between the instruments (although all in-
that with such moderate intensity levels (and a default dicated to be able to) only by listening to them when
touch which is likely to be pressed rather than struck), played by some other person behind a curtain. But they
the changes in travel times due to varying intensity might could very well discriminate between instruments when
not be directly relevant for the player since they are small they played on them blindly or deaf-blindly (Galembo,
and at the threshold of perceivability. Nevertheless, they 1982, 2001). This study implied that the hapto-sensorial
are sufficiently large to produce the typical melody lead feedback of the piano action to the playing pianist is cru-
(Goebl, 2001). cial for the estimation of the instrumental quality.
At that typical dynamic range, key bottom times are Another important factor altering the hapto-sensorial
even more unlikely to be perceived by the pianist sepa- feedback sensed by the pianist is the room acoustics
rately from the sound (hammer–string), since those tem- (Bolzinger, 1995; Galembo, 1987). A piano action might
poral differences are there of the order of a few millisec- feel easily to handle in a room with reverberant acous-
onds. However, the differences between key bottom and tic, while the same action feels intractable and tiring in
hammer–string can be up to 40 ms in extreme cases which a room without any reverberation. Similarly, the tim-

To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior 11
bre of that instrument might be judged differently with ACKNOWLEDGMENTS
changing room acoustics. A pianist is usually not able to This research was supported by a START Research
separate the influences of room acoustics from properties Prize by the Austrian Science Fund (FWF project No.
of the instrument and directly attributes room acoustics Y99-INF), by the city of Vienna (project “Interfaces to
to instrumental properties (Galembo, 1987, 2001). Music”), and by the European Union [Marie Curie Fel-
The reported temporal properties of the piano actions lowship, HPMT-GH-00-00119-02, the Sounding Object
were derived from isolated piano tones (without pedal) project (SOb), IST-2000-25287, and the MOSART IHP
such as they rarely occur in piano performances. For a network, HPRN-CT-2000-00115]. The Wenner-Gren
new keystroke, the key does not necessarily have to come Foundation provided a visiting professorship grant to the
back to its resting position, but, due to the double repeat- third author during 2001/02. The OFAI acknowledges
ing feature of modern grand piano actions, the hammer basic financial support by the Austrian Federal Ministries
is captured by the check and the repetition lever stopped for Education, Science, and Culture and for Transport,
by the drop screw (Askenfelt and Jansson, 1990b). When Innovation, and Technology. Special thanks to Anders
the key is released approximately half way (of the ap- Askenfelt, Erik Jansson, Simon Dixon, Friedrich Lach-
proximately 10 mm touch depth), the jack is able to re- nit and the Bösendorfer company, Tore Persson, Alf
sile back underneath the roller and another keystroke can Gabrielsson, and two anonymous reviewers for essential
be performed. This occurs usually some 2–4 mm below help during the experiments and valuable comments on
the key surface. For such keystrokes, the key can travel earlier versions of this manuscript.
only 6–8 mm, so the travel times can be expected to be 1 The term “modern grand piano” refers to what is nowadays com-
shorter than with a pressed touch from the key’s resting monly used by pianists in concert halls. However, it is not modern
position. Also for such repeated keystrokes, it would be anymore, because, e.g., the Steinway model D grand was developed
impossible to calculate or to determine a finger–key con- and introduced already in the second half of the 19th century and
tact point in time. The study of such repeated keystrokes has remained essentially unchanged since then.
2 The first three paragraphs are taken from Goebl (2001) and re-
has to remain for future investigation. peated here for the sake of completeness.
3 Askenfelt and Jansson (1990b) measured the C4 on a Hamburg
An interesting issue with respect to the reported data
Steinway & Sons grand piano, model B (211 cm).
is whether there is a relationship between the actions’ 4 The “prelay function” compensates for the different travel times
temporal properties and the instrumental quality of the of the action at different hammer velocities. In order to prevent
tested grand pianos. The authors’ personal opinion as pi- timing distortions in reproduction, the MIDI input is delayed by
anists was that from the three investigated grand pianos 500 ms. The solenoids (the linear motors moving the keys) are then
activated earlier for softer notes than for louder notes, according
in this study the Steinway grand piano was qualitatively to a pre-programmed function.
superior to the other two (in terms of the actions’ respon- 5 This particular piano was used in Askenfelt and Jansson (1992a).
siveness), although the Bösendorfer was a high-standard 6 Brüel & Kjær Accelerometer type 4393 (2.4 g).
7 Brüel & Kjær ENDEVCO Accelerometer Model 22 (0.14 g).
concert grand piano as well. The small Yamaha baby
8 Brüel & Kjær Sound Level Calibrator Type 4230.
grand was the least interesting instrument also due to its 9 In order to account for the differences in radius between the ac-
size. However, all pianos were on a mechanically high celerometer placement on the hammer shank and the actual strik-
standard and they were well maintained and tuned. It is ing point at the hammer crown, the hammer velocity data was
assumed here that one of the most important features of corrected for that (resulting in values increased by 14%). This
a “good” piano is a precise and responsive action. correction was not applied in Goebl and Bresin (2003b).
10 Only three keys were tested at the Steinway piano (C1, C5, G6).
In the data reported above, some differences between 11 The recordings were done between May 2001 and January 2002.
12 This measurement was also used to find the individual attacks in
the pianos could be observed that might influence the
subjective judgment of instrumental quality and that a recorded file. All accelerations below a certain value were taken
as onsets. The very rare silent attacks were not captured with this
support the authors’ subjective preference for the Stein- procedure, as well as some very soft attacks.
way. The Steinway showed (1) much less difference in key 13 The peak sound level was calculated by taking the maximum of
bottom times due to touch than the other two pianos; (2) the sound energy (max. RMS with a 10-ms sliding window).
14 The intensity of the touch precursor depends strongly on the way
late crossing of the key bottom approximations and the
hammer–string contact line, indicating a low compressiv- a particular keystroke was played. It is possible to produce loud
struck tones without clearly visible touch precursors (see also Goebl
ity of the parts of the action; (3) shorter time intervals et al., 2004).
of free flight (already almost zero at keystrokes with a 15 This terminology might be misleading, because “time” refers to a

MHV of more than 1.5 m/s, while for the Bösendorfer point in time, although in this case a time duration is meant. Terms
it was around 2.5 m/s, for the Yamaha above 3 m/s). like “travel time” or “time of free flight” were used according to
the term “rise time” that is commonly used in acoustic literature
Moreover, the Yamaha showed many very early hammer (see, e.g., Truax, 1978).
velocity maxima at velocities between about 1 and 2 m/s, 16 Askenfelt and Jansson (1990b) used a Steinway Model B, serial
the Bösendorfer some, the Steinway almost none. number 443001, built in Hamburg 1975.
17 In Goebl et al. (2003) the time interval between the points of MHV
Although further evaluative investigations would be re- and hammer–string contacts are plotted under the label “free flight
quired to be able to state more conclusively any hypothe- of the hammer.” We consider the present data display (escapement
ses on the relation of temporal behavior of grand piano through hammer–string contact) to be more appropriate.
18 Note that all three pianos were maintained and regulated by pro-
actions and instrumental quality, it seems likely that a
fessional technicians before the measurement so that all pianos were
constant behavior over type of touch and late hammer in concert condition before the tests.
velocity maxima are crucial for precise touch control and 19 Certainly the performing pianist may in some cases hear the
a subjective positive appreciation of instrumental quality. finger–key noise as well.

12 To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior
Askenfelt, A. (1994), “Observations on the transient components tional Computer Music Conference, San José, USA (Interna-
of the piano tone,” in Proceedings of the Stockholm Music Acous- tional Computer Music Association, San Francisco), pp. 77–80.
tics Conference (SMAC’93), July 28–August 1, 1993, edited by Goebl, W. (2001), “Melody lead in piano performance: Expressive
A. Friberg, J. Iwarsson, E. V. Jansson, and J. Sundberg (Publi- device or artifact?” J. Acoust. Soc. Am. 110(1), 563–572.
cations issued by the Royal Swedish Academy of Music, Stock- Goebl, W. (2003), “The Role of Timing and Intensity in the Pro-
holm), vol. 79, pp. 297–301. duction and Perception of Melody in Expressive Piano Perfor-
Askenfelt, A. and Jansson, E. V. (1990a), “From touch to string mance,” Doctoral thesis, Institut für Musikwissenschaft, Karl-
vibrations,” in Five Lectures on the Acoustics of the Piano, Franzens-Universität Graz, Graz, Austria, available online at
edited by A. Askenfelt (Publications issued by the Royal Swedish http://www.oefai.at/music.
Academy of Music, Stockholm), vol. 64, pp. 39–57. Goebl, W. and Bresin, R. (2003a), “Measurement and reproduc-
Askenfelt, A. and Jansson, E. V. (1990b), “From touch to string tion accuracy of computer-controlled grand pianos,” in Proceed-
vibrations. I. Timing in the grand piano action,” J. Acoust. Soc. ings of the Stockholm Music Acoustics Conference (SMAC’03),
Am. 88(1), 52–63. August 6–9, 2003, edited by R. Bresin (Department of Speech,
Askenfelt, A. and Jansson, E. V. (1991), “From touch to string Music, and Hearing, Royal Institute of Technology, Stockholm,
vibrations. II. The motion of the key and hammer,” J. Acoust. Sweden), vol. 1, pp. 155–158.
Soc. Am. 90(5), 2383–2393. Goebl, W. and Bresin, R. (2003b), “Measurement and reproduc-
Askenfelt, A. and Jansson, E. V. (1992a), “From touch to string tion accuracy of computer-controlled grand pianos,” J. Acoust.
vibrations. III. String motion and spectra,” J. Acoust. Soc. Am. Soc. Am. 114(4), 2273–2283.
93(4), 2181–2196. Goebl, W., Bresin, R., and Galembo, A. (2003), “The piano ac-
Askenfelt, A. and Jansson, E. V. (1992b), “On vibration and finger tion as the performer’s interface: Timing properties, dynamic
touch in stringed instrument playing,” Music Percept. 9(3), 311– behaviour, and the performer’s possibilities,” in Proceedings of
350. the Stockholm Music Acoustics Conference (SMAC’03), August
Bolzinger, S. (1995), “Contribution a l’étude de la rétroaction dans 6–9, 2003, edited by R. Bresin (Department of Speech, Music,
la pratique musicale par l’analyse de l’influence des variations and Hearing, Royal Institute of Technology, Stockholm, Swe-
d’acoustique de la salle sur le jeu du pianiste,” Unpublished den), vol. 1, pp. 159–162.
doctoral thesis, Institut de Mécanique de Marseille, Université Goebl, W., Bresin, R., and Galembo, A. (2004), “Once again: The
Aix-Marseille II, Marseille. perception of piano touch and tone. Can touch audibly change
Bork, I., Marshall, H., and Meyer, J. (1995), “Zur Abstrahlung piano sound independently of intensity?” in Proceedings of the
des Anschlaggeräusches beim Flügel,” Acustica 81, 300–308. International Symposium on Musical Acoustics (ISMA’04) (The
Báron, J. G. (1958), “Physical basis of piano touch,” J. Acoust. Acoustical Society of Japan, Nara, Japan), pp. 332–335, CD-
Soc. Am. 30(2), 151–152. ROM.
Báron, J. G. and Holló, J. (1935), “Kann die Klangfarbe des Gát, J. (1965), The Technique of Piano Playing (Corvina, Bu-
Klaviers durch die Art des Anschlages beeinflußt werden?” Z. dapest), 3rd ed.
Sinnesphysiologie 66(1/2), 23–32. Hart, H. C., Fuller, M. W., and Lusby, W. S. (1934), “A precision
Bryan, G. H. (1913), “Pianoforte touch,” Nature 91(2271), 246– study of piano touch and tone,” J. Acoust. Soc. Am. 6, 80–94.
248. Hayashi, E., Yamane, M., and Mori, H. (1999), “Behavior of piano-
Cadoz, C., Lisowski, L., and Florens, J.-L. (1990), “A modular action in a grand piano. I. Analysis of the motion of the hammer
feedback keyboard design,” Comp. Mus. J. 14(2), 47–51. prior to string contact,” J. Acoust. Soc. Am. 105(6), 3534–3544.
Chaigne, A. and Askenfelt, A. (1994a), “Numerical simulations of Hirsh, I. J. (1959), “Auditory perception of temporal order,” J.
piano strings. I: A physical model for a struck string using finite Acoust. Soc. Am. 31, 759–767.
differences methods,” J. Acoust. Soc. Am. 95(2), 1112–1118. Koornhof, G. W. and van der Walt, A. J. (1994), “The influence
Chaigne, A. and Askenfelt, A. (1994b), “Numerical simulations of of touch on piano sound,” in Proceedings of the Stockholm Mu-
piano strings. II: comparisons with measurements and system- sic Acoustics Conference (SMAC’93), July 28–August 1, 1993,
atic exploration of some hammer-string parameters,” J. Acoust. edited by A. Friberg, J. Iwarsson, E. V. Jansson, and J. Sund-
Soc. Am. 95(3), 1631–1640. berg (Publications issued by the Royal Swedish Academy of Mu-
Cochran, M. (1931), “Insensitiveness to tone quality,” Austral. J. sic, Stockholm), vol. 79, pp. 302–308.
Psychol. 9, 131–134. Ortmann, O. (1925), The Physical Basis of Piano Touch and Tone
Conklin, H. A. (1996), “Design and tone in the mechanoacoustic (Kegan Paul, Trench, Trubner; J. Curwen; E. P. Dutton, Lon-
piano. Part I. Piano hammers and tonal effects,” J. Acoust. Soc. don, New York).
Am. 99(6), 3286–3296. Ortmann, O. (1929), The Physiological Mechanics of Piano Tech-
Dietz, F. R. (1968), Steinway Regulation. Das Regulieren von nique (Kegan Paul, Trench, Trubner, E. P. Dutton, London,
Flügeln bei Steinway (Verlag Das Musikinstrument, Frankfurt New York), Paperback reprint: New York: E. P. Dutton 1962.
am Main). Podlesak, M. and Lee, A. R. (1988), “Dispersion of waves in piano
Fletcher, N. H. and Rossing, T. D. (1998), The Physics of Musical strings,” J. Acoust. Soc. Am. 83(1), 305–317.
Instruments (Springer, New York, Berlin), 2nd ed. Repp, B. H. (1996), “Patterns of note onset asynchronies in ex-
Galembo, A. (1982), “Quality evaluation of musical instruments pressive piano performance,” J. Acoust. Soc. Am. 100(6), 3917–
(in Russian),” Tech. Aesthetics 5, 16–17. 3932.
Galembo, A. (1987), The Quality of Piano Tones (in Russian) Seashore, C. E. (1937), “Piano touch,” Scientific Monthly, New
(Legkaya Industria Press, Moscow). York 45, 360–365.
Galembo, A. (2001), “Perception of musical instrument by per- Suzuki, H. (2003), “Analysis of piano tones with soft and hard
former and listener (with application to the piano),” in Pro- touches,” in Proceedings of the Stockholm Music Acoustics Con-
ceedings of the International Workshop on Human Supervi- ference (SMAC’03), August 6–9, 2003, edited by R. Bresin
sion and Control in Engineering and Music, September 21– (Department of Speech, Music, and Hearing, Royal Institute of
24, 2001 (University of Kassel, Kassel, Germany), pp. 257–266, Technology, Stockholm, Sweden), vol. 1, pp. 179–182.
http://www.engineeringandmusic.de. Truax, B. (1978), Handbook for Acoustic Ecology, vol. 5 of World
Gillespie, B. (1994), “The virtual piano action: Design and imple- Soundscape Project (A.R.C. Publications, Vancouver, BC).
mentation,” in Proceedings of the 1994 International Computer Van den Berghe, G., De Moor, B., and Minten, W. (1995), “Mod-
Music Conference, Århus, Denmark (International Computer eling a grand piano key action,” Comp. Mus. J. 19(2), 15–22.
Music Association, San Francisco), pp. 167–170. White, W. B. (1930), “The human element in piano tone produc-
Gillespie, B. and Cutkosky, M. (1992), “Dynamical modeling a tion,” J. Acoust. Soc. Am. 1, 357–367.
the grand piano action,” in Proceedings of the 1992 Interna-

To appear in J. Acoust. Soc. Am. Goebl et al.: Piano Action Temporal Behavior 13