Lawrence E Murr & Everaldo Ferreyra Tello. American Scientist. Volume 88, Issue 1. Jan/Feb 2000.
In 1946, Ellie Mannette wanted to build the world’s biggest and best steel drum for an upcoming music contest on the island of Trinidad. So he hid a 55-gallon drum in a sugar shack and pounded its lid into a concave surface. He then tuned 14 notes into the surface of this steel drum, which was gigantic compared with the standard 35-gallon drums. When Mannette went on stage at the contest, pulled out his big steel drum and played Ludwig von Beethoven’s Fur Elise and Johannes Brahms’ Lullaby, a screaming crowd acknowledged the birth of a new era in Caribbean music.
The Caribbean steel drum-or steel pan-is the only musical instrument developed in the 20th century. It evolved in street bands in Trinidad and Tobago that used various percussive rhythm instruments, including skin drums of African origin and a bamboo instrument called tamboo bamboo, a hollowed-out piece of bamboo that can be pounded on another surface to make sounds. Members of these bands also built various metal drums, beginning in the mid-1930s, when they used paint or bakery tins to make some of their first examples. They pushed out the bottom of the tins to make a convex surface, and then crafted three to eight outward bulges of different sizes to produce sounds of various pitches. Mannette changed the pan design by pushing the drum head inward to make a concave surface. Then he made “notes” by creating convex bulges of various sizes on the drum’s surface. Shortly after the development of concave drum designs, Trinidad native Bertie Marshall and Barbadian-bom Joe Griffith discovered a harmonic tuning process (explained below) that created the chromatic sounds that we associate with modem Caribbean steel drums.
Today’s steel-drum heads contain from three to 32 notes in patterns that, like the instrument itself, evolved by trial-and-error and intuition. Similar to a marching band using more than one kind of drum-say, a snare and a bass drum-steel-drum bands also use various drums that have different arrangements of notes. For example, a lead drum consists of a dozen U-shaped notes lying on the drum’s periphery, with an inner circle of another dozen notes. Finally, four notes lie on the bottom of the drum. A bass steel drum, on the other hand, has just three notes that are arranged as if the drum’s head is divided in three large slices, like a pie. A so-called cello pattern includes half a dozen notes on the periphery and two inside of that. So different numbers and arrangements of notes create different kinds of steel drums.
Despite the alluring sounds of steel drums, we know rather little about the materials-science side of their acoustical properties. Some of the most intriguing studies available came when David S. Hampton, then a Master’s student at Northern Illinois University, and his colleagues used time-averaged holographic interferometry to reveal the acoustic mode patterns along with their movement and coupling effects of the note zones. Nevertheless, many other qualities of these drums remain unexamined. For example, no one knows how deformation and crystal imperfections, which are created during drum fabrication, affect the vibrational spectrum or timbre of a drum. In addition, no systematic studies have examined the effects of heat treatment on the development and tuning of notes.
We have examined the acoustical and materials-science properties of steel drums at their various processing stages. These studies reveal that the art of Caribbean steel-drum making captures a collection of fine musical qualities in the materials being used. Nevertheless, advanced techniques and materials might make even finer drums.
Pounding Out the Patterns
Making a steel drum involves a complex process that can vary depending on the orchestral voice of the drum to be built and on the builder’s fabrication procedure. For any modern steel drum, though, the process begins with a standard 55-gallon drum, which has an initial diameter of about 57 centimeters and a barrel length of 86 centimeters. The following fabrication procedure applies to steel drums, such as lead drums, that produce higher acoustic frequencies.
To get started a drum maker draws concentric circles to mark out approximate note zones and to aid in shaping. Then the maker takes a sledge hammer-usually weighing at least 2.7 kilograms-and pounds in a spiral pattern to create a nearly hemispherical concavity 15 to 20 centimeters deep, with the deeper areas being used to create higher-frequency notes. Next, the maker lays out the notes, scribes their shapes and pounds them into convex ellipsoids on the concave surface. Once all of the notes have been pounded out, a steel-drum maker forms a groove around each one by using a flattened punch. A groove partially isolates a note’s acoustics from surrounding notes, and it also levels the convex dents of all the notes so that they are pushed out to about the same distance.
At this point, the maker cuts the drum sides to make a drum skirt of a specified length. It is left full length for bass drums and is correspondingly shortened for higher octave ranges or acoustic frequencies. For example, a C lead drum, which has 29 notes that cover more than two octaves, has a skirt length of only 15 centimeters. It should also be noted that the bottom lid is cut out of the drum so that the final instrument has a skirt, or resonator, that is open at the bottom.
Next, the drum gets heated. Heat treatment of the patterned drum head began initially for burning off the paint and excess oil by turning the drum over an open fire. Someone discovered that this firing produced better sounding and more tunable drums, and the process continued as a crucial part of steel-drum fabrication.
Finally, a maker again takes up a hammer to tune a drum. During tuning, glancing hammer blows near a note’s edges manipulate the elastic properties of the metal membrane. This can be envisioned as being similar to stretching a guitar string, but the string is a thin and irregular metal membrane-a two-dimensional “string.” During this process, a maker tunes the center of a note to its fundamental frequency. Then two sides of the note get tuned to one octave higher than the fundamental frequency, or twice the fundamental. The remaining two sides get tuned to a harmonic, or three times the fundamental frequency.
Musicians often hang drums on stands and then play them with a mallet that consists of neoprene rubber wrapped around an aluminum tube about 25 to 40 centimeters long.
Sliding Sheets of Atoms
Beyond shaping a steel drum’s surface, the pounding during fabrication also changes other characteristics of the metal. For one thing, this processing reduces the thickness of the drum head. We cut in half a fully patterned, heat-treated and partially tuned drum and measured its thickness. From the drumhead perimeter to the drum bottom, the steel gets thinner but not at a constant rate. Instead, the steel thins most quickly near the perimeter. The thickness reduction slows considerably along the drum’s bottom, where it reaches a minimum thickness of about 50 percent compared with the original metal, which is usually 18 gauge, or 1.15 millimeters thick. In other words, the thickness varies regularly but nonlinearly from the perimeter to the bottom. In addition, the grooving around individual notes can reduce this crosssection by another 10 to 15 percent. Consequently the so-called engineering strain-which is roughly equivalent to the thickness reduction-in the grooved cross-section around the highest-pitch notes in the drumhead bottom would be expected to be around 65 percent.
The fabrication also changes the microscopic structure of the metal. Prior to any processing, we swabbed a barrel’s polished metal surface with three percent nitric acid in ethanol, which selectively etches the boundaries between grains or crystals of the metal. Light microscopy then revealed slightly different-size grains in the drum surface and skirt, but all grains generally formed approximately regular polygons. Then we took 3-millimeter-thick disks from hammered sections of a surface and a skirt, polished them and examined their surfaces with transmission electron microscopy, bright-field imaging. The samples looked very similar, and both included dislocation substructures, or crystal defects.
To understand the concept of a dislocation, imagine that planes of atoms arranged in layers make up a crystal. A dislocation consists of an extra plane of atoms that stops partway through a crystal. The edge of this extra plane bends the neighboring planes and creates a disturbance running through the crystal. When a very thin section of metal containing a dislocation is viewed in an electron microscope, the bent neighboring atomic planes scatter the transmitted beam. The image appears as a black line-or many complex, interacting lines when there are many such defects. Dislocations accumulate during rolling or other forms of deformation, including hammer blows, by allowing the atoms in the metal to slip. Other-wise, the metal would simply break. In fact, if the deformation is too severe, this slip process cannot accommodate it and the metal will break.
The dislocation density is measured in terms of lines intersecting a unit area of surface, say one square centimeter. Metals that are not heavily deformed contain dislocation densities of around one million lines intersecting one square centimeter of surface. After hammering a steel drum, it contains from 100 million to 1 billion dislocation lines per square centimeter.
As dislocations accumulate in a metal, it hardens. Consequently, the thin metal near the bottom of the drum head is harder because it is more deformed. So drum makers place the smallest and highest frequency notes in the harder, thinner drum bottom. The larger, lower frequency notes get placed toward the top of the drum. We measured the dynamic and static hardness of individual notes on a steel drum, and that revealed an apparent softening in the larger notes and the more rigid smaller notes.
Extracting the Ideal Note
We conducted a simple series of experiments to examine the effect of deformation on the acoustic response of free circular plates. In this case, we deformed the metal by squeezing it. We cold-rolled-by multiple-pass rolling (back and forth)—a 6.35 millimeter thick plate of annealed 316 stainless steel to reductions of 10, 20, 30 and 40 percent of its original thickness. These rolled samples and an unrolled plate were all milled to a uniform thickness of 3.4 millimeters, and we then cut 7.3 centimeter-diameter disks from each sample. Finally, we drilled a small hole in each plate, hung it by a loop of wire, struck it with a tungsten-alloy mallet and recorded its acoustic spectrum. We selected 316 stainless steel as a model material because it resembles the material in steel drums. Its carbon content is similar to that of drum steel, and its initial grain structure roughly matches the low-carbon drum steel, although the dislocation microstructures are significantly reduced compared to the low-carbon drum steel. The starting plate hardness is also similar to drum-steel hardness.
To test the disks, we milled the 3.4 millimeter-thick stainless steel disks to a thickness of 1.6 millimeters and acoustically tested them. Then we milled them in half again, to 0.76 millimeters, and tested those disks. Although the thicker disks showed some acoustic-signal variations when comparing the O-percent deformation to the 40-percent deformation, the acoustic spectra of the very thin disks exhibited much greater changes with the degree of initial reduction or deformation.
The acoustic signal of the thin disks changed dramatically when their thickness was reduced. Initially, the disk created a fundamental frequency and one harmonic, or higher frequency. When thickness was reduced by 20 percent, the harmonic peak split in two, separated by 60 hertz. The disks reduced in thickness by 40 percent showed harmonic peaks separated by 160 hertz. In addition, the O-percent reduction’s harmonic peak was approximately 24 hertz higher than the centroid of the split peaks at 20 and 40-percent reductions. That shift arose because the unrolled disk was somewhat thicker than the others initially (it was difficult to mill such thin disks uniformly), and the frequency is proportional to the ratio of the disk’s thickness to the square of its radius. Consequently, given equal radius, a thicker disk produces a higher frequency.
Our tests on these disks also showed that the harmonics dominated the acoustic signals. That is, the amplitude of the fundamental frequency, or ideal pitch, of the note was relatively weak and inaudible in contrast to the first and second harmonics in the 3.4-millimeter thick disks, and to the first harmonic in the 0.76-millimeter disks. Furthermore, the amplitude of the split harmonic is greatest for the thin disk at the 40-percent reduction.
These observations for relatively ideal, free circular disks suggested that the variations in deformation from the top to the bottom of a steel-drum head might influence its overall sound spectra. In fact, the 0.76-millimeter disk corresponds to the drum-head thickness where it has been reduced by 34 percent, or in the intermediate hardness zone of the drum head. Consequently, the results from these tests illustrate the very dramatic, if not complex, effects that deformation has on the overall timbre of a steel-drum note.
Turning Up the Heat
The firing or heat treatment of a drum continues to be guided almost entirely by observing color changes or color evolution of the heated drum head. These features arise from the development of oxide, which is similar to color changes on a water puddle when different thicknesses of oil films are present. The time to achieve the requisite color depends on the heat source, or temperature, but normally takes I to 10 minutes. Following the heating to a requisite drum color, the drum is removed from the heat and either air cooled or quenched in-water.
To investigate the effect of heat treatment on steel-drum heads, we cut samples from undeformed drum steel and from representative notes on pounded but untreated drums. The samples from notes had thickness reductions of 10 percent-at the top of a drum-and 50 percent-at a drum’s bottom. We cut numerous coupons from these sections and used them for spectroscopic analysis for carbon and other elements and to test for hardness. Then we heated other representative coupons at constant temperatures for times ranging from 1 to 30 minutes, or for constant times at different temperatures, ranging from 177 to 1,000 degrees Celsius. After heating, we quenched the samples in water and measured their hardness.
These experiments revealed that high temperatures produced dramatic hardening in samples from a note where the steel had been reduced in thickness by 50 percent and its carbon content was 0.09 percent. In fact, applying a temperature of about 370 degrees for roughly 10 minutes increased the metal’s hardness by about 20 percent in comparison with an untreated drum bottom. Moreover, a secondary hardening peak arose at about 800 degrees, where the base hardness increased by more than 50 percent. Optical metallography analysis showed no perceptible microstructure variation after the 370-degree treatment, but the 800-degree treatment caused the original grains to recrystallize-into larger ones.
This lower-temperature hardening, which corresponds to the normal way steel-drum heads are heat treated, represents so-called strain aging. In classical strain aging, carbon atoms move to regions of dislocations and thereby prevent slip or movement in response to additional straining. This increases the hardness. At the much higher temperature, the grains divide into many smaller ones, which produces considerable hardness because it is ideally proportional to the reciprocal square root of the mean grain size.
Hardening also depends on the carbon content in the steel. Even after half an hour of 1,000-degree heat, no tempering appeared in a sample with about 0.02 percent carbon that was reduced in thickness by 50 percent. A sample with 0.07 percent carbon that had not been reduced in thickness at all, on the other hand, showed no notable hardening at 370 degrees, but at 800 degrees and above there was considerable hardening. In fact, between 10 and 20 minutes of 1,000-degree heat increased this material’s hardness by 2.4 times. We also noted that the prevalence of very tiny crystals, or grains, increased significantly at or above 800 degrees, and this accounted for the dramatic hardening.
Consequently, heating steel drums at 800 degrees would probably have a much more dramatic effect on timbre than the conventional heating or firing at around 370 degrees or less. Even heating at 800 degrees for one minute dramatically affects the hardness. Nevertheless, deformation could cause changes in the orientation of the grains in the steel that create the harmonic splitting mentioned above. So heat-related changes in the grains, such as recrystallizing to form smaller grains, might or might not enhance the quality of the sound. We are currently testing the effect of heat treatment on the harmonic splitting.
The hardening depends on many factors, including the carbon content, the temperature and duration of heating, and the level of deformation. For example, less deformed metal hardens slower at a given temperature. Consequently, the different notes, which have different levels of deformation, would not harden uniformly if exposed to consistent heat. However, in measuring the temperature distribution over the drum surface, we found that the temperature near the upper drum surface-where the larger notes are patterned-is nearly twice the temperature in the drum bottom, which is situated farthest from the heat source.
As a result, the temperature differential over the drum surface is self-compensating and in effect allows for nearly uniform hardening in all of the notes. This is another example of how circumstantial and serendipitous the hand fabrication of steel drums has been. Not only does the “burning” of drum heads optimize the hardness through aging, but because of the dome geometry and deformation variations, the aging is nearly uniform when drums have the requisite carbon content. On the other hand, too much heat will over-age a drum and too little carbon prevents it from aging at all. So the conventional making of fine steel drums with superior musical quality is analogous to the making, or aging, of a fine wine.
The fact that the variations in drum thickness produce corresponding variations in deformation that are self-compensating and produce a uniform hardening over the drum surface might argue against spin-forming or other automated forming operations that might produce a uniform drum-head thickness. Although this might permit a more uniform note thickness, it could seriously alter note timbre by yielding a perfectly uniform hardness. In effect, this might provide some scientific rationale for maintaining the handmade folk art of steel-drum fabrication.
Tapping in the Tuning
The fact that all of the notes coexist on the same drum head or acoustic platform creates complex, nonlinear coupling. For example, when a drummer strikes a note that has a specific fundamental frequency, other notes with equivalent octaves or harmonic frequencies get excited sympathetically. These excitations increase with amplitude, and at typical percussion amplitudes nearly the full drum surface vibrates and radiates the unusual music that is characteristic of the Caribbean. Appropriate tuning to take advantage of the overtones must consider this complex interaction.
To see how such coupling affects the notes on a steel drum, we performed a complete acoustical analysis for each note in a partially tuned drum. Specifically, we used the left drum from a pair that makes up a so-called double tenor drum. This drum head included 17 notes. We examined them for evidence of harmonic features noted in the ideal, circular notes, especially the apparent splitting of harmonic frequency peaks.
For instance, we compared the acoustic spectrum for a 7.3-centimeter disk of 316 stainless steel with one of this tenor drum’s partially tuned A-flat notes, which had a maximum dimension or note length of about 7.3 centimeters. The note thickness varied from about 0.7 to 0.8 millimeters in contrast to the 0.76-millimeter thick 316 stainless steel disk, which had been reduced to 40 percent of its original thickness by cold rolling. Despite the physical similarities in these materials, their fundamental frequencies differed by about 280 hertz. A closer look at their spectra showed other striking differences. For one thing, they both showed a split in the first harmonic, but the harmonic peaks were separated by 160 hertz in the stainless-steel disk and by 170 hertz for the drum note. Moreover, most of the energy in the circular disk’s spectrum came from the split harmonic, and the fundamental frequency had almost no acoustic energy, or was barely audible. The steel-drum note, though, produced most of its energy in the fundamental frequency, not in the split harmonic, which had low but audible energy. After tuning, the split harmonic for this note would produce more energy, or louder sounds.
The tuning process actually enhances the multi-harmonic sound of steel-drum notes. These multi-harmonic features-a strong fundamental and weaker, split harmonic-have been demonstrated for many of the steeldrum notes. In fact, steel-drum notes usually contain audible tonics, or fundamental frequencies, and harmonics.
As mentioned briefly above, drum makers tune the steel by more hammering to shape and tune the note center to the fundamental frequency Then they use glancing blows-aimed at strategic parts and usually directed away from the note’s center-to “pull” the steel membrane to tune the octaves at the right and left ends of the major axis of the elliptical note and one or more harmonics at the top and bottom of the ellipse. Sometimes, a steel-drum maker uses a torch to relieve key zones in order to enhance stretching during the hammering.
Although the skirt length affects the overall tuning of a steel drum, it is not usually deformed during fabrication or tuning. Moreover, the skirt does not get nearly as hot as the drum head during heat treatment, so it is not significantly hardened either. It is likely that, just as deformation influences note timbre, variations in skirt deformation or even heating effects might also have some influence on the overall drum timbre, or voice quality. We are experimenting with decorative hammering-created with a ball-peen hammer–of the skirt at various levels of hammer impact to investigate this prospect.
What we have learned about the acoustic spectra of steel-drum notes could be used during tuning. For instance, tuners might make a careful comparison of tuning sequence changes from these spectra and use that informarion to simplify the tuning process. In addition, more acoustic-signal analysis of steel-drum notes along with variations in heat treatment might eventually lead to an understanding of very subtle harmonic qualities that contribute to superior or even special sounds for a variety of steel-drum voices.
Beyond advanced tuning techniques, steel drums could also be made from different materials. Fabricating a Caribbean steel drum requires a strong material that is very hard to stretch. The 316 stainless steel used for the free circular notes surpasses conventional drum steel in these features, and it might make a superior steel drum. Since the skirt material has not been shown to be of much importance, a suitable stainless-steel plate could be fabricated into one end of a standard 55-gallon barrel and tested. Moreover, this stainless steel would not harden or temper the way low-carbon drum steel does, so it might not be necessary to heat treat the patterned drum head.
No one knows if different techniques or different materials would make a superior steel drum. In fact, trial-and-error clearly guided Caribbean drum makers to an approach that creates marvelous music. Nevertheless, there is certainly an opportunity to explore a number of possibilities to expand the fabrication potentials and musical prospects for other metal alloys.