Originally presented at the International Computer Music Conference and Festival
(Delphi July 1992)

NeXT, MusicKit, DSP 56001 and the Grand Piano:
A Study in Interfacing Computing Technology to an Acoustic Instrument

Alistair M. Riddell
Department of Music
Princeton University
Princeton New Jersey 08544
amr@silvertone.princeton.edu

Preface

Any attempt to integrate computer technology with traditional acoustic instruments in a way that aims to extend the instrument will yield some interesting results. Insight illuminates areas of the project, certainly when least expected. On an initial reflection, one might begin to sense the difficulty of the task, and from there, to relations and correspondences, in an artistic sense, between the mechanical and informational ages from which the technologies originate. That is, what the exercise promotes is an awareness of the differences between not only the instruments but the thought processes behind them. For instance, one may care to further speculate on potential disturbances to a powerful extant aesthetic, the piano tradition. What one is meeting head on is history and the inevitability that accepted ideals will attack innovation where it dramatically crosses the boundary from 'necessity' (technical or logistical) into that of a new 'language' and 'concept' domain.

However, for most projects requiring considerable practical or intellectual effort in order to become manifest, it is unlikely that they begin from such obscure contemplation alone. Only later, as the model grows in complexity and contextual richness, does one, perhaps, begin to consider such inherent questions in more detail.

The `Meta-Action' (MA) thus began as the result of research and performance with acoustic pianos under computer control carried out during the 1980s (Riddell 1982, 1989). Once again, existing paradigms, certain musical achievements and ambitions but not exclusively those from the piano tradition drive technological development.

Fundamental Objectives

The following functional and design considerations were worked out over a two year period from 1987 to 1989. In March of 1992 the action was installed for the first time in an arbitrary instrument, a Mason and Hamlin 6ft grand. Given these circumstances, installation was, on reflection, much easier than one could have imagined. Future installations will be less problematic.

As a portable replacement action the MA radically departs from the traditional grand piano action model with an emphasis on mechanical autonomy, speed and subtly. Although it might accommodate conventional performance paradigms, it is in the area of extended functionality that it departs from its 19th Century precursor. Here its reason for existence is more acutely felt. The above performance expectations are based on different performance expectations, not necessarily those of the piano tradition. Overall design and mechanical simplicity, contribute to a unique perspective on performance. This is most dramatically seen in there being only 3 moving parts per hammer/damper assembly: the hammer solenoid, the damper solenoid and the damper lever. The MA further permits operational autonomy between hammers and dampers. Not only does the operation of `pedalling' disappear (or at least become more sophisticated) but interesting combinations of hammer/damper operation are accessible including activating the hammers but not the dampers and vice versa, and raising dampers on strings that are not being struck thereby inducing sympathetic resonance in ways not possible before. The net effect of this is a re-evaluation of the piano from both a compositional and performance perspective.

Unlike many so called computer controlled piano systems the MA is portable and, due to design considerations, should fit a large number of existing instruments. The advantages of this are immediately obvious. Less obvious is that there are tens of thousands of grand pianos around the world with considerable timbral variety and an action that can access, at least, a small number of these is preferable to one that is permanently installed. However, the disadvantages are that it requires installation for every piano and that takes time. The original keyboard and hammer mechanism are temporally put aside and not used.

While the MA's design allows considerable performance functionality, the counter side to this is that precise control becomes a critical issue, in that, what is being controlled, the solenoids, tend also to be subject to the dynamics of their inherent physical condition. The control system is not completely causally accountable for all events. As a result of this inherent complexity, specifically with regard to hammer solenoid management, it is not possible to control the action with MIDI directly. In fact, the timing demands are so critical that it needs a dedicated processor, in this case, the DSP 56001. This comes about not only because it is crucial to turn the hammer solenoids off at the right time but that the dynamics are determined by how long they remain under power. Timing is in fractions milliseconds and must be rigorously observed for each hammer solenoid in order to obtain uniformity of articulation. Due to its performance potential and adjustability, quick installation of the MA is not possible. Even accounting for the awkwardness of most spaces in which it is placed, considerable attention has to be given to stable positioning and adjustment of each section and solenoid. Inevitably, this would be the case for any portable action that offered such installation flexibility.

The current development platform is the NeXT computer which provides the DSP 56001 processor as the controller and the main processor (68040) as the Host. The NeXT machine provides good development facilities for both the `040 using a MIDI kernel and the DSP 56001.

From the point of view of computer processing power the situation is only going to improve. This translates into greater throughput of information and the potential for even more subtle performance scenarios.

Some Physical Particulars

The MA consists of approximately 160 solenoids-one for each hammer and damper-mounted on 2 unique parallel rails. Both rails assemble from three sections into seamless units upon which the solenoids can be freely positioned to suit the instrument. The rails are mounted on 4 or 5 support brackets which also function as adjustable feet. From these 5 brackets the two rails can be tilted or raised in relation to the strings.

The forward and taller rail has the hammer solenoids mounted on only one side in two vertical rows. This facilitates the necessarily close grouping. No other arrangement with the existing hardware is possible. The damper rail has the solenoids mounted on either side and a lever from each passes underneath to lift the dampers. Again, this is a critical arrangement determined by the spacing of the strings. Gravity returns the hammers and dampers to their quiescent state.

Separate driver logic controls power switching for each solenoid. This hardware is in turn activated by yet more dedicated circuitry which receives data from the control processor.
The hammers themselves were made initially with a hard surface with the expectation that when used in dense configurations, some sense of identity for each pitch will be apparent. However, the issue of tone, in general, is one that remains open to exploration. The Meta-Action was constructed through a grant from the Australia Council's `Special Projects Unit' and the work carried out in the engineering workshop of the Physics Department at La Trobe University in Melbourne Australia between June 1987 and August 1989. It was designed and constructed by Marshall Maclean and Alistair Riddell. Current research in control and performance is taking place in the Department of Music at Princeton University.

Performance

The action can only be accessed via the computer. Consequently, there are two possibilities for performance: either by the computer alone or by initiation through some external device (a MIDI instrument) which communicates through the computer. However, for the purposes of this research the MA is considered a performer/interactive system. Like other real-time systems this puts considerable strain on computing resources and the sound producing mechanisms. Therefore, the assumption from the outset is that delegation of control, at a low level, is to be handled by a dedicated processor. This processor receives data from the host but has a priority to manage the activated solenoids. This means that there is no facility to ªrecordº from the action just as there is no means to record keyboard activity from a conventional piano. The MA itself largely prohibits this facility. To some degree, performance information could be derived from another part of the interactive setup, probably the host processors (68040). The implication here is, of course, that the host and control processors are very closely connected so as to minimise data transfer time. Now, since there is a special communications link between the control processor and the action interface hardware, a proprietary protocol can be employed that again optimizes performance by being tailored to the specific requirements of the action.

Efficiency Considerations

I am currently using the NeXT Music Kit to translate and expand incoming MIDI into DSP data and commands. Preliminary investigations suggest that this process does not introduce any significant latency and is able to maintain the necessary throughput. The DSP program currently requires waits on the output side. This is possibly because of the relatively slow opto-isolators (1 Mbit/s) used between the action and the NeXT. The interface logic should be able to handle the direct output from the DSP, since TTL logic switching rates are in the order of Nanoseconds. However, there is a slight kludge with the data connection in that there is no busy signal available on the DSP side. The signal lines are eight data and one strobe which take all the nine lines available from the DSP Port C. This makes it difficult to determine when the interface hardware has successfully received data and when to send more. This may become a problem, however, under the current test conditions it has not appeared to be so. Note also that there is no provision for dialog between the interface hardware and the DSP. The interface hardware does not return status information. The following DSP code shows the data out routine:

outData

movep y:dataout,x:m_pcd ; move data out

do #400,d_out ; pause momentarily
do #2,d_out2
move X1,X1 ; do nothing much
move X1,X1
d_out2
move X1,X1
d_out

bchg #8,y:dataout ; change 8 to 1
movep y:dataout,x:m_pcd ; toggle the strobe line

do #200,strb_out ; once again, wait for things to stabilize
do #2,strb_out2
move X1,X1 ; do nothing much
move X1,X1
strb_out2
move X1,X1
strb_out
rts

In recent tests, the potential problem of accumulated system delay (MIDI keyboard, Unix, DSP, action) turned out to be less than anticipated. Even hand clusters showed considerable simultaneity. However, these experiments were not conducted with the entire system in operation. The data throughput has only been examined for damper control. When the hammer interrupt routine is fully operational some delay may become apparent.

Some Final Thoughts

This particular research has provided a unique opportunity to consider a range of problems in interfacing which include the physical, the historical and the aesthetic. The physical world once again demonstrates a reluctance to function as anticipated, and yet to reveal on occasion results that have interesting ramifications. The next challenge is to compose for this system in a way that explicates these conditions.

Glossary of Terms

Solenoid - An electro-magnetic mechanical device. When supplied with electricity, an iron rod will move within an outer coil of copper wire in which a magnetic field has been induced.

Protocol - In computers, an agreed upon procedure and syntax for communication between different electronic equipment.

Interface - In this case, electronic hardware that interprets and converts incoming signals to another series of signals.

Opto-isolators - Integrated circuits that provide separation of electrically incompatible systems. These IC's are used to allow MIDI to connection with most standard serial ports.

References

Jaffe, D. A. 1989. "An Overview of the NeXT Music Kit". Proceedings of the 1989 International Computer Music Conference, Columbus, Ohio. International Computer Music Assoc., pgs. 135-138.

Jaffe, D. A. 1991. "An Overview of the Sound and Music Kits for the NeXT Computer". With L. Boynton. 1989. Computer Music Journal, MIT Press, 14(2):48-55. Reprinted in book form in The Well-Tempered Object, ed. Stephen Pope. MIT Press.

Riddell, A. M. 1982. "The Computer Controlled Piano - New Instrument, New Performer". NMA 1. NMA Publications. Melbourne, Australia.

Riddell, A. M. 1989. "Towards a Virtual Piano Action". NMA 6. NMA Publications. Melbourne, Australia.

Riddell, A. M. 1989. A Perspective on the Acoustic Piano as a Performance Medium Under Machine Control. MA Thesis. La Trobe University. Melbourne Australia.

Riddell, A. M. 1990. "A META-ACTION for the Grand Piano". Proceedings of the 1990 International Computer Music Conference, Glasgow, Scotland. International Computer Music Association. California.