Andrew Margules, May 14, 2004
Virtual Lunga Project
Our project for this class was to create a Virtual Lunga. The purpose of this was to be able to analyze the language that is used with this "Talking Drum," when played by Dolsi-naa Abubakari Lunna, a Ghanaian prodigy who helps to teach the African Drumming class here at Tufts. The hope was that this Virtual Lunga would not look like a traditional Lunga, but sound like the real instrument when played. When formulating construction procedures and order, we analyzed the drum and broke it into three basic parts. The first part is the stick used to play the Lunga. The second integral part of the Lunga is the drumhead. The strings that run up and down the length of the Lunga are the third part of the drum that we needed to emulate in our construction. We wanted to realistically emulate each of these components using sensors and other electronics.
The first thing that was done to emulate the Lunga was to try and mimic the sounds that are produced when the drumhead in its natural state. We acquired a practice pad and took that apart. I cut out six pieces of Piezo film from the large sheet. I them made three different pressure sensors each composed of two pieces of the Piezo film. Each sensor was constructed as follows:
Diagram of Piezo Film Sensors in Drumhead
The point of making each sensor in this manner was to have the pieces of Piezo film separate and when the drumhead was stuck, the pressure created from the impact would complete the circuit and would send controller data to the MAX patch via the Doepfer Box.
The Piezo sensors were originally put on the face of the drumhead, to be able to provide position of the strike, and to generate controller data when stuck with the stick. After attempting this, and finding that we were met with little success, I decided that a new approach needed to be taken with regard to the sensors on the drumhead. I drilled a hole in the back of the pad to be able to feed the wires from the sensors through. I then took off the pads top and put the sensors between the foam in the center of the pad and the heads membrane.
Again, with this new set up, a great amount of noise was still present and sent to MAX from the sensors. At this point, we tabled the idea of having sensors in or on the practice pad and decided to concentrate on wiring up the Lunga stick. The fist thing that was attempted was to shave off part of the outer edge of the stick head, making it more flat for us to attach sensors.
The sensors that we had planned to attach to the now smoother parts near the base were FSR sensors, or Force Sensing Resistors. These were purchased at the advice of Roberto Aimi, who is a PhD candidate at the MIT Media Lab. He told us that he has had great success with FSRs and that we should give them a try. Fortunate for us, when the FSR sensors arrived, the actual sensor was small enough to still fit in the area that we had previously shaved off. We tried different types of glue and tape to try and affix the sensors to the stick head and produce the kind of data that we could us in the MAX patch. Unfortunately, there was a discrepancy in the data sent to both MIDI SCOPE and to our MAX patch with the FSRs. When the sensor was hooked up to the Doepfer Box, we were receiving data in a range that had a great enough difference to be able to scale it to our needs. When the sensor was attached to the stick, the data range was not consistent and was too small for us to manipulate. The big issue we encountered at this point was that the values were going up and down, too rapidly, in a staircase manner. This caused the MAX patch to not be able to determine the maximum value at a specific time and to not play the appropriate note through Reason.
To fix this problem, we enlisted the help of Warren from the Electrical Engineering Department. He made for us a circuit that was to act, in a way, as a sort of op-amp. As he demonstrated on his oscilloscope, his circuit smoothed out the waveform, and was able to effectively change the amplitude of the wave with a variable resistor. When we tested this circuit with the FSR sensors, we achieved great success. The data coming in from the FSRs were in an appropriate range that we could then scale, and the staircase effect was all but eliminated. Also, when tested with the Piezo film, the data read on MIDI SCOPE was at a stable value when at rest and would go to a maximum value, and hold it for a moment, when struck.
The next problem we encountered was that when the group next met, none of the sensors worked as they previously had, and it seemed that the circuit built by Warren had stopped functioning as it had been. Unsure of what had happened, we took the circuit back to Warren and told him what had happened. He added a Germanium Diode to the circuit and we tried the sensors again. This time, they worked perfectly.
The next step, once we had all of the sensors working, was to implement them on the actual Lunga. We acquired the head of a Lunga from Dolsi-naa, and we took and attached a Piezo sensor to the back of the drumhead and ran the wires through a hole drilled in the side of the Lunga. The drum was taken back to Dolsi-naa where he strung the drum back up and returned it to us. Upon its return, we hooked the Piezo to the Doepfer Box and tested it out. At this time, we found that the manner in which we attached the Piezo sensor greatly muffled the sound of the drum and we were uncertain that Dolsi-naa would even play the drum, as I was told that Dolsi-naas playing the Lunga is a full mind and body experience and the muffled sound could make him feel uncomfortable. Upon further testing of the drum at this time, the Piezo fell off of the drumhead and into the cavity of the drum itself. This was extremely unfortuneate as we would need to bring the drum to Dolsi-naa, have him unstring the drum, bring the drum back to the lab and reattach it, bring it back to Dolsi-naa and have him string the drum up and get it back to the lab all before he had to leave to go back to Africa.
At this point, we tried to mimic the Lunga with a set to toy Bongos, only to find that the Piezo was extremely unresponsive. Also, we began to develop some trouble with interpreting the data from the FSR on the head of the stick. This became a problem as when force was applied to the sensor with our hands, we got good data values. On the other hand, when hitting the drum or the table, the values were much less than ideal. As the device stands now, we have a practice drum pad with a Piezo sensor under the head. This part does not deviate far from the original design only that instead of three sensors, only one is full operational at this time. As far as the stick is concerned, the geometry of the stick has changed. It is now not as deep, as approximately 3mm were cut off of the face, and both of the sides have been cut to be flat. A single FSR sensor is attached to the face of the stick and a push button is attached to the Doepfer Box in its place.
The MAX patch that was made for this project has two main functions. The first is to take in the controller data provided from the sensors attached to the Doepher Box. Once the MIDI data comes in, the MAX patch scales the data range to fit in the MIDI range of 0 to 127. This in turn, depending on the parsed data, sent information to the NN19 module in Reason and told it to play a specific Lunga sample. This function was originally implemented for the sensor that was to be in place on the face of the stick and on both sides to facilitate the side hits that occur in the Lunga play. There was also a sensor attached to the strings of the Lunga. This proximity sensor was supposed to mimic the movement of the strings when the Lunga is played. To do this, the controller that the Proximity sensor was assigned was mapped to control Pitch Bend in Reason, as to help create that same sound of the tension increase when the strings are pulled on.
If I were to start this project again from the start, I would do a couple of things differently. First, I would get range of different sensors that I would think would work for a specific application. I would then make sure that I read all of the literature that I could find on the specific sensor and knew its pluses and minuses. I would then try each sensor, for a specific task, and compare the data from each. I would then make a decision about which sensor to use in the situation based upon which sensor gave the best data. That means data that is consistent, easy to manipulate, and repeatable. I would perform this testing process depending on the number of sensors that I planned to affix to the drum. Second, I would attach all of the sensors but fully assemble the drum. This would help in that I would be able to test out and make sure that all of the sensors were still fully operational and were all operating in the manner for which I had chosen them. Third, and finally, I would assemble the drum and make sure that the sensors were still operating correctly. Another construction parameter that I would try to change or be more wary of is the type of wire being used. In instances where space is limited or a large change in space would be detrimental to the testing, I would use a very high gauge wire. I would then connect it to a smaller gauge wire further from the sensor. Another parameter that I would try to make more consistent would be the type of and quality of the soldering connections between wires and sensors and wires and wires. One thing that did set us back a bit was the need to be continually re-soldering specific joints and the inconsistency of the wire used. A more uniform wire scheme would help to not only alleviate stress on the wires, but also on the joints. •