Towards a virtual piano action

Alistair Riddell

This article first appeared in NMA 6 magazine. In it, the author reports on an Artists and New Technology project [1] to design and build a computer-controlled piano action.

Piano Music in Evolution

Since the turn of the century music for the piano has exploded with stylistic diversity and performance innovation. This has largely taken place where the inherent qualities of the instrument and their potential have been recognised and promoted. Not only in the West but in countries such as Japan, China and Korea, enthusiasm is manifest through performance, manufacturing and the fostering of study and social awareness through music schools and methodologies (Yamaha, Suzuki, etc). The instrument's historical repertoire (essentially from the Viennese classics to early Twentieth century) continues to feature prominently in most assessments of the instrument. However, it is clear that through many unique social and technological changes, its musical potential is being exploited by a far wider and larger cross section of world societies. This is evident, for example, in the scope of discussion in popular keyboard magazines. It may also be assumed that the process is continuing under the influences of today's technology.

In certain musical idioms, particularly those in the aural tradition such as Jazz, Blues, Rock, etc, piano technique has changed through the dissemination of recordings as well as through direct contact with the current influential practitioners. The dissemination of recorded music has supported many enthusiasts who, by emulating the definitive performers, have themselves developed further distinctive techniques. In a genre of music that appears to be less constrained by formalism and pedagogy, imitation becomes an underlying reactionary force that stems the tendency for particular styles to disintegrate as quickly as they emerge.

The situation in the Art music world - as distinct from other musical milieus: commercial, popular, etc - is somewhat different because most performers are bound to interpret some form of score. In this notated representation of the music, technical as well as stylistic innovation is dependent on the composer's practical and theoretical knowledge of the instrument. Thus, the success of music composed in this manner is largely attributable to the perspicacity of the composer in the treatment of both instrument and performer. This is evident in such works as the prepared piano works of John Cage; the Klavierstucke works of Karlheinz Stockhausen; Charles Ive's Concorde Sonata ; Elliott Carter's Piano Sonata of 1945/46; Pierre Boulez's Structures ; the Player Piano Studies of Conlon Nancarrow and many others from the twentieth century repertoire.

Recorded music is perhaps less influential as a conveyance for imitation in an art music context because direct imitation is not explicitly part of the evolution of that musical genre. However, recordings are evidence of the success of a musical idea and convey proof that certain compositional ideas work (or don't). Furthermore, the quality of contemporary music performance and interpretation has improved through its general availability in a recorded form. Those interested in contemporary music are able to further familiarise themselves with it between performances. Recordings also tend to establish personal criteria through which a comparison can be made between the memory of the recordings and live performances or other recordings. The ease with which it is possible to compare one interpretation of a work with another or one style with another, must contribute to a broadening of musical perspective and at least, perhaps introduces avenues for change.

Technical Evolution

Underlying the piano's social ethos throughout this century is the fact that the instrument had reached evolutionary stasis by the end of the nineteenth century. The following is a summary of the significant developments of last century (Grover 1973. p.210.):
  • 1821. Sebastien Erard produced the 'repetition action' with double escapement.
  • 1826. Jean-Henri Pape patented felt hammers.
  • 1843. Jonas Chickering patented the first one-piece cast-iron frame for a grand piano.
  • 1859. Steinway produced the first over-strung grand piano.
  • 1874. Steinway perfected the 'sostenuto' pedal.
Certainly, since last century the piano has undergone considerable refinement in component materials and production techniques yet it has essentially remained an instrument of 19th century technical achievement. It is perhaps, possible to attribute a considerable part of the diversity of piano music this century to developmental stability. Unlike the current state of electronic musical instruments, musical innovation is entirely up to the performer; the manufacturer has no part in encouraging new music other than through promoting artists who use their instruments.

This illusion of stability should not be interpreted in the first instance as an attainment of technical perfection. Given human nature and art, this is, of course, intrinsically impossible. The instrument, however, remains relatively constant in appearance and sound because of its social prominence and musical tradition. In order to remain a worthy interpreter of the historical repertoire as well as the piano music of tomorrow, the impact of even minor technical modifications and improvements are generally vetted against the standard repertoire, performance criteria and contemporary musical expectations. Transitions from wood or ivory to plastic, or from iron to alloys, do not significantly affect the perpetuity of the piano repertoire and are eventually accepted without regret. Yet the sum of the many minor changes must ultimately be manifest as an evolutionary step, even if obscured by the passing of many years.

Although it is difficult to clearly define the relationship between instrument evolution and musical style, music itself should not be assessed on the basis of technical progress. As Willi Apel pointed out:

"Nothing is more dangerous and misleading in the study of the arts than to regard achievements of the past from the standpoint of technical progress. A superficial observer sees only what has been gained in the fight and not what has been lost. The true historical mind, however, sees that in the history of humanity there is no possibility of perfection, and that there is only a faint hope of approaching it." (Apel 1953. p.86)

To some extent the continuing popularity of the piano repertoire of the 18th and 19th centuries is attributable to the fact that its performance has been widely accepted on contemporary instruments. Those with an understanding of the historical instruments would argue that certain original qualities of the music are consequently lost or misrepresented. This may very well be true but for the most part the musical intention and aesthetics are present where the interpretation is sincere and reverent.

This concern for historical accuracy has increased with the growth in popularity and scholarship of pre-baroque music this century. It has fuelled the underlying controversy and debate over, not only the use of authentic replications of historical instruments but musical practice and interpretation researched from primary sources. Stepping back from the issues surrounding early musicology, the net effect of this debate is perhaps to increase the musical public's awareness of instrumental diversity and evolution. If there is greater awareness of a musical continuity then contemporary instrumental practices should appear less disconnected from tradition.

The technical evolution of musical instruments is, of course, inevitable. What is perhaps different today is the nature of the association that is formed between the various instruments and the current musical styles. As the instruments themselves become 'virtual', that is able to adopt the characteristic sounds and behaviour of many other instruments, the relationship between instrument and musical idea that is not only difficult to define but transient. Combine this with the rate with which 'new' instruments are appearing on the market and the result is a instrument/music relationship that has never existed before.

The Action Project in abstract

In 1987 I began to consider the design and development of a high performance electro-mechanical action for a grand piano. The opportunity arose through the Artists and New Technology Program from the Australia Council. The concept had been maturing from the early 1980's when I first began working with pianos under microcomputer control. During those interim years, composition and research with less ambitious instruments and equipment (Riddell 1982, 1988) helped clarify my future intentions, aesthetics and the inevitable pragmatic considerations associated with the project.

When construction seemed imminent, expectations of the function of the action were largely influenced by the prospect of an unusual Performer/Machine Interactive system. One that could exploit a distinct relationship between the piano tradition and digital technology. At that time it was felt that mainstream music technology was not pursuing the same type of instrumental relationship in such a simple and direct manner. On a fundamental level, the difference is manifestly the acoustic production sound, while at a higher level, it is performance on a familiar instrument with an entirely different philosophy towards its control.

An integral part of the action's functionality allows it to be used in two performance contexts. In the first, it can operate through a live performer in real-time. In the second, by the computer alone, that is without any real-time human intervention. This range of functionality covered my interests in real-time performance and algorithmic composition.

From the perspective of a performer/machine interactive system, the action can be regarded as 'virtual' in that the performer's actions can be mapped to almost any pitch/rhythm combination on the instrument. Transposition, inversion and intervalic parallelism with all or partial input are some of the simple processing functions possible for keyboard actions. The more fascinating possibilities lie beyond these techniques where performer input is mapped to parameters other than pitch or rhythm. In fact they might not be parameters but functions.

The action has also a degree of operational autonomy that the traditional action could not possibly accommodate. The relationship between a hammer and its associated damper is flexible. They can operate synchronously or under separate control which permits two timbral possibilities: playing with the damper off or playing with the damper on the string.

Although the prospects for the research and performance of the action appear limited, it is an attempt to alleviate the difficulties of working with computer technology and acoustic instruments. To begin with the grand piano is a complex and yet ideal medium that is reasonably available at most performance venues. With the action in a modular state, the logistics of performing are considerably reduced. However, installing and adjusting the action will still take some time but does not require the piano to be altered or modified in any way other than the removal of the original action. This is a straightforward operation which takes only a few minutes.

Initial Research and Observations of the Traditional Grand Piano Action and the Proposed Action

The preliminary research towards the project began in mid 1987 in the Department of Physics at La Trobe University [2] . This initial research was not directed towards the immediate construction of the new action but an examination of the physical behaviour of the traditional grand piano action. The research was conducted jointly with Dr Michael Podlesak [3] and centred around the measurement and investigation of the velocity, acceleration and power in a hammer when propelled into motion. Experiments were carried out using an ONO SOKKI FFT (Fast Fourier Transform) analyser. This instrument recorded, calibrated and plotted values that were transmitted to it from a tiny accelerometer mounted on the hammer shank.

The information gathered from the experiments was used to convert the hammer's mechanical energy into electrical energy. The power in the hammer motion was quantifiable in terms of Watts and these values could be used comparatively with the electronic solenoids also under examination.

The results finally revealed that, within certain limitations, the physical size and power rating of the proposed hammer solenoids would be acceptable. This was particularly important because it meant that the existing Pianocorder solenoids and control components could be used, thus reducing the cost and complication of the overall system. The major restrictions, however, arose in relation to the adequacy of the power supply from a standard domestic source (2400 - 3000 watts) and whether the solenoids would be capable of an effective dynamic range in considerably less than optimal conditions.

The experiments on the instrument focused on two dynamic ranges: moderately loud and extremely loud and the differences between human and electro/mechanical performance for this experimental range was considered from a number of perspectives.

The more demanding end of the performance spectrum was of more interest since the lower end is demonstrably less problematic for either system, attention could therefore be directed towards the extreme case to at least establish a possible upper limit.

The experiments at this dynamic level revealed that the power in terms of electrical energy was approximately equivalent to 25 watts. However, the occasions for such extreme performance (if any) are only likely to be found in later twentieth century music and even then, in order to achieve such a dynamic, the performer would generally have to forgo speed, elegance and repetition for brute strength. It might be expected that even the most extreme contemporary works would not require the performer to sustain or distribute this sort of power to large groups of notes over a considerable time. Neither is it possible to apply such power to complex passages with any degree of subtlety or accuracy (although the Jazz musician Cecil Taylor might come close to being an exception).

It has been speculated that the experimental solenoids could take an equivalent power (25 watts) and possibly deliver the same effect. As desirable as it may be to emulate this level of performance, it is doubtful that the existing solenoids and support electronics could sustain repeated use under such conditions and also whether many of them could be supplied from a domestic source. This result was eventually considered too extreme to pursue as necessary for the electro-mechanical system yet it helped to define a practical upper limit.

The moderately loud attack resulted in power ratings between approximately 7 and 10 watts. This considerably reduced range was closer to estimated realistic performance figures for the solenoids. The solenoid experiments were carried out with relatively conservative power supplies and it was speculated that their performance may be improved through more efficient and large power supply strategies.

The issue of power supply and usage was further complicated by the differences between the mechanical operation of two hammer systems - conventional and electro-mechanical. The hammers for the electro-mechanical arrangement (discussed in the next section) are heavier than the heaviest piano hammer by about 7 grams. The heaviest bass hammers on the conventional action are approximately 13 grams while the iron solenoid cores alone (without the actual hammer tip) weigh 20 grams each. However, a larger hammer mass with less velocity could result in a similar dynamic to a conventional lighter hammer travelling at a greater velocity. This was not experimented with directly but was recognised as an important factor in the performance of the new action.

In addition to the weight differences the operation of the two mechanisms necessitates an alternate assessment of the mechanics of the hammer movements.

FFT plot

An FFT plot of the hammer striking the lowest A on the grand piano.

The FFT plot shows that the time taken by the hammer from initial impulse to string impact was approximately 18 milliseconds (ms) for the moderately loud case. From that time onwards the hammer is returning to a rest position. The further oscillations recorded on the plot are the movement of the hammer on its wooden shank. For the initial thrust of the hammer the plot does not show the point where it is free from the impulse of the keystroke force and travelling without further propulsion. Although this period of free flight may well be negligibly small, by contrast the solenoid remains under power and acceleration until it impacts on the string and potentially interferes with its vibration.

It is expected that the solenoid can be powered up for an approximately equivalent period of time as the normal hammer - about 18ms. But that also depends on the intended dynamic. The dynamic range for the electro-mechanical action is achieved by turning the solenoid off at various stages after activation. The loudest attack will result from the solenoid remaining powered up fractionally beyond the time of impact. This may be 20ms while the softest attack might only be a 6-7ms period. An influential factor in the timbral quality of a struck string is the reflected waves from the shortest end of the instrument which return in approximately 9ms. The longer the solenoid remains in contact with the string, the more the hammer interferes with the evolving spectra.

Construction of a New Action

The question of performance was momentarily put aside when time came to consider construction of the action. After all, abstract performance questions were of little consequence if it turned out to be impossible or impractical to construct a mechanism - using the resources at hand - that could fit into the complex cavity that normally houses the conventional action.

Through the theoretical principles of operation and those existing components that were inextricably part of the 'grand design', some idea of its form was known in advance. However, from February of 1988 to late June, many ideas and approaches were examined and discarded as the search for a workable design was given priority.

scale drawing

A scale drawing showing space restrictions inside the piano.

The action began to take shape in the hands of Marshall Maclean [4] in the Physics workshop at La Trobe University from June of 1988 onwards. Many of the technical and design problems that sprang up during construction were resolved by him (see the working sketches ). Without recourse to a piano at every critical point, he produced an action that constantly met not only with theoretical expectations but proved successful during intermediary tests.

The action benefited from many important decisions made along the way. Some of those were:

  • No specialised solenoids were to be constructed either for the hammers or dampers. Those used in the Pianocorder mechanism were considered to be suitable provided they could perform to an acceptable level and be located in the already cramped space. Development went ahead on those premises.

  • The action could be made to fit a number of instruments by being modular and adjustable. The major obstacle, however, that this attempted to overcome in the first instance was the intense asymmetry and irregularity of the piano's construction. It could not be assumed that internal structure, for example, would remain constant or that it would change in a linear fashion. Consequently, consideration was given to the possibility that part of the action mount might cause interference with the action frame, i.e. frame struts or sostenuto bar.

  • A preliminary test installation revealed that there was more space available than initially thought. This meant that the action could now reside well within the limits and permitted greater variation between instruments.

An examination of the photographs of the first (bass end) of the three modular sections identifies the support structure for the hammer and the damper solenoids. These are arranged in two parallel rails mounted on adjustable feet. The hammer/damper configuration met the functional specification of minimum moving parts and autonomous control. The forward vertical rail is the mounting for the hammer solenoids which are staggered in two vertical rows. The rear shorter rail holds the damper solenoids in a similar staggered configuration but this time back-to-back on either side of the rail. The front rail has circular holes drilled at intervals to allow the wiring that runs between the solenoids and the driver boards to pass around any of the moving parts and also reduces the weight of the system.

Each of the three sections has at least one locating foot. For the full action length there will be five: one at each end and three located a various points in between. Since each module locks together, two feet per module are unnecessary and only add to the complexity of installation. Feet can be added or subtracted during an installation depending on the particular instrument. They are also locatable anywhere along the mounting rail where the damper levers are absent. This will usually occur when the frame struts separate the strings into sections.

The solenoid cores (the iron part which moves under the influence of the magnetic field) will move 15 millimeters for hammer operation and 5 - 7 millimeters to lift the dampers. The hammer and damper movements are actually in contrary directions, that is, the hammer solenoid cores move up while the damper solenoid cores push down.

The dampers are operated in the same way as the original action. The solenoid pushes down on a lever which raises the damper. It is expected that these solenoids, while on for considerable periods of time, will require less power. Consequently they will not heat up so quickly nor be subject to aberrant electro/mechanical behaviour such as core 'chatter'.

The action was not conceived as being a permanent part of any particular piano. The full implications of this desire for portability emerged as the research progressed and it was soon realised that it was going to take a considerable amount of time to assemble in an instrument. More over, it was going to be tedious and frustrating due to the massive amount of adjustment required to align damper levers and hammer solenoids. However, to make the task easier, much of the adjustment and assembly was envisaged as taking place before the action was installed. Every attempt therefore will be made to alleviate complications with the wiring and final adjustments before it enters the instrument. A clearer understanding of solutions to these types of problems are likely to result from installation experience rather than speculation.

action frame

A cross section of the action frame.

The action frame was machined from 16mm and 25mm aluminium plate. This provides considerable rigidity and yet is relatively light compared to the all up weight with the solenoids.

Testing and Performance Expectations

Once the mechanical component is completed the system will be tested with existing electronic hardware to verify that it is operationally successful. This testing stage will not initially take place in an instrument since that adds unnecessary complications. When the system has satisfactorily completed the initial tests and been examined for any potential problems, it will be installed within a piano and the testing will begin again under the intended operational conditions.

Underlying the success of the project will be its ability to perform to fundamental expectations. Since the movement of each part of the action is very small, it should in theory at least, function with considerable rapidity. This has been proven to some extent on an earlier system (Riddell 1988). The particular layout of the solenoids (vertically) introduces no new forces other than the existing functional ones of electro-magnetism and gravity. However, rapid operation may be inhibited due to the sophistication of some of the co-ordinated operations of the mechanism, where the dampers and hammers are able to operate synchronously or autonomously during a performance. In a normal operating mode, for example, a damper should be free of the string by the time the hammer strikes. The system is therefore required to have the damper solenoid activated in time to avoid the hammer striking the string while the damper is still relatively close.

Once the mechanical component has been tested, attention will turn to the most critical aspect the system - real-time control. An initial consideration of real-time operation results in the view that a distributed approach may be necessary. That is, where the management of complex scheduling and event structures is made easier by partitioning and distributing computation and I/O to subordinate processors. This simplifies and solves problems in one area but unfortunately introduces problems in others. Nevertheless, while control of one piano may not cause major problems, a multiple instrument configuration would at least increase the complexity of control by an order of magnitude for a single microcomputer system and thus justify distributed processing.

Irrespective of what approach is adopted, the system will remain subject to change if circumstances warrant it. Consequently, every effort is made to maintain flexibility throughout, hopefully allowing change to one area without necessitating extensive change in others.

Software is particularly suited to modularity and revision. If, for example, the lower level processes manage the note by note activity they need not change when software is changed at a higher level. This is a common practice these days in software development but difficult to carry out if the lower level functions become to embedded or too specific.

Conclusion

From the outset of this project many obstacles and problems arose that made it appear impossible. It was not simply the scope of the project but the perceived inherent difficulties within each major section. Months later when those obstacles can be considered with some hindsight, it is clear that some apparently insurmountable dilemmas were resolved with less effort than initially imagined. Therefore to have even arrived at this point in the project is very encouraging and although much work remains to done, the next stages can be viewed less as a series of problems to be faced and more as milestones to be passed.

Footnotes

1. The author gratefully acknowledges the financial assistance of the Australia Council in this project. return

2. I would like to thank Professor Keith Cole, Dr Ron Miller and Horst Dressel of the Department of Physics at La Trobe University for their co-operation and encouragement in getting this project underway. return

3. The action owes its existence to Dr Micheal Podlesak who contributed his expertise and interest during the experimental stage and Marshall Maclean of the Physics workshop, through whose skill and insight the action takes its present form. return

4. (see 3. above). return

Bibliography

Apel, Willi. The Notation of Polyphonic Music. 900-1600. Cambridge, Massachusetts: The Mediaeval Academy of America. Publication No. 38. 1953.

Grover, David S. The Piano. Its Story from Zither to Grand. London: Robert Hale. 1976.

Riddell, Alistair. The Computer Controlled Piano: New Performer, New instrument. NMA1 . Melbourne : NMA Publications 1982. pp.6-9.

Riddell, Alistair. M.A. Thesis (in progress). Department of Music. La Trobe University. Bundoora. Victoria. 1988


© 2001 NMA Publications and Alistair Riddell.

Composer Alistair Riddell has worked on a range technology-related projects and created many works for his own computer controlled piano action.
His more recent work includes numerous computer-generated pieces.

back to NMA magazine index.