Krista Ehinger and Justin Kao
We carved marimba bars and simulated them in Matlab.
The marimba bars were made from 1''x2''x12'' and 1''x2''x18'' pieces of East Indian Rosewood. The curves of the tuned bars were determined experimentally by making shallow cuts and then measuring the pitch. The 12'' turned bar has a fundamental of A5 and a second partial of about E7. The 18'' tuned bar has a fundamental of E4 and a second partial of E6.
Since a marimba bar is much longer than its width and much wider than its height, we model the marimba bar in one dimension only: the vertical displacement from equilibrium as a function of distance along the length. In this model of the bar, the mallet exerts a force on the center of the bar causing it to bend, and elasticity provides the force causing the bar to vibrate thereafter. The vibration decays in time due to damping forces.
Equations for this were adapted from Chaigne and Doutaut [1], and predict the vibration of the marimba bar given only physical constants [2] and measurements of the marimba bar and mallet. 1
The sound radiation of the vibrating bar is modeled as radiation of a vibrating line of spheres, as described in Doutaut, Matignon and Chaigne [3].
We deviate from [1] and [3] in our bar's supporting force. We model this as two localized forces, representing the strings holding a real marimba bar in place.
The simulation treats the bar as a string of rectangular slices taken along the length. This is a discrete version of our model's differential equations. The discretized equations can then be solved by computer at each time step, yielding the motion of the mallet and marimba bar.
The code for this was written in Matlab.
We created several virtual bars: an ideal bar given in [4], two curved bars based on our real bars, and two flat bars based on our real bars. Since wood is variable in its physical properties, we estimated density from dimensional measurements and weight of the flat bars. Using a formula2 for the fundamental frequency of a flat bar cited in [5], we calculated the Young's modulus of our wood from its fundamental frequency.
Sound radiation is approximated by considering each slice of the bar as a vibrating sphere. The simulation calculates the actual sound pressure at the observer's location.
Our simulation approximates the fundamentals of the bars very well, but error increases for the higher peaks. Part of the problem may be that the simulation generally yeilds fewer peaks than the real bar, so it can be difficult to identify and match up peaks.
|
The model that we used does not take into account non-vertical vibrations, so this may account for some of the error. Non-vertical vibrations can be ignored in a long, flat bar, but they are significant in a short, thick bar such as the 12'' bar. In addition, there are probably variations in height, density, and elasticity along the length of the real bars, which would also contribute to error.
An obvious extension would be to consider the marimba resonator as modeled in [3]. The accuracy of the simulation could be improved by better measurements of the rosewood's elasticity, density, and other properties.
This document was generated using the LaTeX2HTML translator Version 2002-1 (1.68)
Copyright © 1993, 1994, 1995, 1996,
Nikos Drakos,
Computer Based Learning Unit, University of Leeds.
Copyright © 1997, 1998, 1999,
Ross Moore,
Mathematics Department, Macquarie University, Sydney.
The command line arguments were:
latex2html -split 0 marimba
The translation was initiated by Justin Kao on 2003-06-06