The XY DrumMeasuring the Timing DeltaThe HardwareConditioning the SignalThe Maple MicrocontrollerSoftwareDetecting Position With InterruptsGenerating Sound with PWM and WavetablesCoding a DrumPutting it all TogetherThe ResultThe FutureThe XY Drum December 10, 2009 Abstract Standard electronic drum pads are equipped with one piezoelec-tric sensor that converts stick strikes to electronic pulses which are measured and used to trigger pre-recorded samples or other sounds. Advanced models can measure the amplitude of this pulse and adjust the volume of the resultant sample or even select a different sample en-tirely. The state of the art electronic drums are equipped with several ’trigger zones’ on each head that allow for two to four different sam-ples to be mapped to different parts of the drum. The XY Drum aims to offer a new level flexibility to electronic drum designs by providing high resolution x,y strike position for each trigger event. Furthermore, the XY Drum uses this information to enrich the interface of electronic rhythm, affording an entirely new playing experience. This paper mo-tivates a fundamental approach to capturing x-y information using several piezoelectric sensors simultaneously and demonstrates its via-bility. Furthermore, an early prototype of the XY drum is described. However, further obstacles in the analog design of the system must be overcome before this prototype will be able to successfully generate the desired x-y strike positions. 1 The XY Drum The XY Drum is a nonprogrammable electronic drum head that captures stick strike velocity and x,y position on the pad. While information alone encapsulates a rich rhythmic interface, users are left to decide how best to use this information. Simple synthesis firmware for generating four sinu-soidal voices are provided, but the mapping between these oscillators and 1 Anonymous MIT studentthe trigger events are intentionally left blank so that users can experiment with how to use this new interface. This mapping is specified via a simple software interface that is uploaded to the XY Drum over USB. The classic use case, although this has not yet been demonstrated, is to map frequency on the vertical axis and decay on the horizontal. Thus pure tones at cer-tain frequencies and with certain envelopes can be produced by striking the pad at varying positions. Early data suggests that users should have no problem triggering several notes simultaneously. This is possible because no player is capable of producing truly simultaneously strick strikes. Incident impulses are measured with a phase resolution of 20MHz (wave peaks can be differentiated as long as they are 50ns apart). Furthermore, positioning information depends on 1ms timing differential between wave peaks, which is about the time it takes for sound to travel across the pad. Thus, unless a player can play simultaneous strokes accurate to the millisecond, the system should gracefully handle the gesture as two distinct strikes. 2 Measuring the Timing Delta The first step towards building the XY Drum was to verify that timing in-formation could be used to triangulate stick strike position. More complex methods of measuring this information are possible. For example, field inter-ference effects (technique used by most tablet displays) would be effective, but would require special sticks with inductive tips. Resistive arrays are also a possibility, and this approach is backed by several patents for x,y sensi-tive drum pads (none have come to commercial fruition). This mechanism is mechanically unsuitable, however, as the resistive top layer that does the sensing tends to ruin the desired strike dynamics that players prefer. Am-plitude triangulation is ruled out because it depends on the pad dampening wave pulses as they transmit across the surface. This dampening effect would make it impossible to recover velocity information about the strike, since am-plitude would necessarily depend on strike position. Measuring the timing delta between wave peaks is subject to non of the above issues. However, this approach requires high frequency sampling not available on most electronic drum interfaces. In fact, most drum triggers are samples at the audio rate of 44KHz, or sometimes slower, since that already exceeds the accuracy of the player. In high quality drum pads such as those produced by Roland, the piezoelectric sensor is coupled to the drum head with a thin metal interface. 2This decrease any dampening, but sound travels very quickly through this material. Depending on the metal, the speed of sound is usually near 3000 m/s. Thus wave pulses take about 3us to travel across a 10cm pad. While it is technically possible to capture 100 million samples per second (MsPs), the associated electronics cost would be too high (more than 200$). In a plas-tic ’rock-band’ style drum, sound travels substantially slower. Experimental results show that wave pulses take a full 1ms to travel across the pad. Un-fortunately, this comes at a cost of substantial dampening, making velocity retrieval difficult. However, this timing delta can be trivially measured to produce ¡ 1mm precision in x,y triangulation, which is far less than players can accurately differentiate. Rather than sample the piezoelectric sensors and search for wave peaks, the signal is amplified so that any tap will rail to the gain ceiling of 3.3V. This means that an ADC is not needed to measure strike position, as this 3.3V can be used to trigger a digital interrupt on most micro controllers. Velocity measurements are impossible in this configura-tion, but can be easily obtained by sampling any of the 3 triggers (or all three) and measuring the peak amplitude of impulses before they have been amplified. 3 The Hardware The piezoelectric sensors source the data pipeline, measuring stick strikes AC pulses. In order to generate x,y information, there are actually three iden-tical data pipelines sourced from three separate sensors. From the sensors, the signal is rectified and amplified and then passed into an STM32F103 Microcontroller. This ARM Cortex M-3 Processor was chosen because it has the requisite speed to measure the timing deltas while simultaneously synthesizing sound or replaying samples. Furthermore, the simple