I mapped data to sound in three ways. First, I used the montly temperature timeseries (which is more or less periodic) as an LFO to control a low pass filter and provide a wobble sound. Next, I developed a bass sound by setting my target frequency to MIDI-40 the shifting by +/- 20 in accordance with the data of the stock market returns, then I added in some delay. Finally, I added a lead section which took the same data and implemented them as a phasor and sine-oscillator combintion with an added 5th and some reverb.
After recording the three sonificatiosn into Ableton Live, I different mixed parts of them that I found interesting to produce the final result.
(1) The values for the gain were squared to account for negative amplitudes in data. Perceptually, these negative amplitudes
are just as loud as their positive couterparts. The negative amplitudes therefore need to influence the gain applied to the
oscillator as well.
(2) MIDI keynums were used for frequency values to easily divide the notes into recgonizable pitches based on the musical octave system,
thus making intervals consistent perceptually with observed data. Had the data been mapped lienarly, we would have heard sweeping frequencies
with no clear recgonizable pitches or perceptual audio correspondance with the data.
(3) If the 100ms update rate was increased or decreased significantely we would have had a one-to-one increase or decrease of note duration. This update controls the time it takes
for the envelope tracking the data to reach it's target value and the speed the data is input to the sonification.
I used two tiem series in this project, all came from http://datamarket.com/.
Mean monthly temperature, 1907 through 1972
Sources: Time Series Data Library (citing: Hipel and McLeod (1994))
Monthly returns of value-weighted S and P (standard and poor) 500 stock from 1926-1991
Time Series Data Library (citing: Pena, Tiao and Tsay (2001))