Our music reader was not functioning as expected. It would work one day but stop working the next day, which made us recalibrate our photocells each time we worked on it. When we learned to create the chart (recording each trial, location of the trial, outcome(s), correction(s)), we kept having troubles because our reader was not consistent. It would make weird beeping noises, play notes that were not on the music scroll, etc. We first had to fix the calibration problem.
We brainstormed for possible solutions: better (more constant) LED/source of light inside the sensor box, using black tape (circles) for notes versus drawing the notes with black markers (consistent black without any ink fading?), using multiple LEDs versus using one (PICO Cricket LED), creating an automatic calibration program that would allow the reader to calibrate itself before playing the notes by taking multiple readings on white surface (no black notes) and storing the maximum/greatest readings for each photocell.
We talked to our professors and discovered that the PICO Cricket LED was not very constant. We were not able to notice with our eyes but the light was constantly flickering, which led to great ranges in readings (that also constantly changed in values in significant magnitudes). Thus, putting two or more of these lights was not a good idea.
We also abandoned the black tape circles idea because that would only make the scroll thicker and could possibly jam and stop the music reader from working smoothly.
Therefore, we experimented with different LEDs. We made a series circuit with a red and white LED and attached them in our sensor box (replacing the PICO Cricket LED). The readings (numbers) were a lot stabler, but when we were trying to calibrate melody 3 (note 3), the differences between the readings for white and black were not so great (we graphed the readings as the music reader read the black notes). The reader was still not as consistent.
Punzi (Julia) and Professor Banzaert came up with the idea of having LED lights UNDER the board (that supported the paper as the photocells detected the notes). They created a cardboard model of 8 LED lights (series circuit), each one for each photocell. This was a great idea and was VERY effective because I observed significant differences between the black and the white surfaces. After measuring, I drew the LED board on Solidworks and printed it (Thanks, Essie!). I had about 2-3 trials (not very many but will do more), and our reader was a lot stabler and significantly more consistent.
One of the things Punzi tried was using taping 2 LED lights next to the bottom board (versus inside the sensor box). This method did not work well, but this idea led to the LED board idea.
The cardboard LED board model.
The cardboard LED board model.
The cardboard LED board model.
Graph from using 2 LED (one white and one red) lights inside the sensor box. The peaks are where the notes are.
Graph using LED under sensor2. Compare with previous graph. There is a significant difference between the white and black surfaces.
Tools to make the holes for the cardboard LED board model.
Our LED board (wooden model)
(back view) We'll make it neater.
Music Scrolls (Experimented with spacing, intensity of black ink, etc.)
