LISTENINGMINDLISTENING: …

                                            LISTENINGMINDLISTENING:
                                                           Untidy Mind

                                                         Tim Barrass

                                                  Melbourne University
                                                 School of Creative Arts
                                            t.barrass@pgrad.unimelb.edu.au

                         ABSTRACT                                  2.2.1.   moog+delay : ch27-VPVA, ch28-VNVB,             ch29-
                                                                            HPHL, ch30-HNHR (BarrassT27.wav)
The sonification of data in this piece is not consciously shaped   ch27-VPVA drives the frequency of a sawtooth wave fed to a
by any guiding metaphor, but has developed through tinkering       moog filter where the cutoff frequency is offset by ch28-
with and tweaking various synthesis methods while listening        VNVB and modulated by an LFO at a rate set by ch29-HPHL.
for coherences and differences in the sound field.                 The signal is sent to an echo with feedback level set by ch30-
                                                                   HNHR. The general positive slope of the data provides
                   1.    INTRODUCTION                              increasing echo intensity as the piece proceeds.

I divided the data channels into two main groups, with the         2.2.2.   bowed string model [1]: ch27-VPVA, ch28-VNVB,
1020 head signals in one and the remaining channels in the                  ch29-HPHL,        ch30-HNHR,         ch34-EDA
other.                                                                      (BarrassT28.wav)
This follows a pattern I noticed in a spectral analysis of the
data in which the 1020 head signals all showed similar strong
harmonic content, while the rest of the channels had no overall    ch27-VPVA : bow pressure
common characteristic. I processed the 1020 signals (ch. 1-26)     ch28-VNVB : bow velocity, modulated at a rate set by ch30-
in a uniform way, and took an ad-hoc approach with the rest        HNHR
(ch. 27-36), re-combining different channels of data in various    ch29-HPHL : vibrato frequency
syntheses. I hoped the redundancy would enhance a perception       ch30-HNHR : sets the rate of bowing, in multiples of the heart
of structure through several different relations in the overall    rate, determined by a tempo-finding algorithm
sound.                                                             ch34-EDA : bow position


                    2.   SONIFICATION                              2.2.3.   "blowtar" model [2]: ch27-VPVA, ch28-VNVB,
                                                                            ch29-HPHL, ch30-HNHR, ch32-OrbOcc, ch33-
                                                                            Mass, ch34-EDA (BarrassT29.wav)
2.1. Channels 1-26
                                                                   ch27-VPVA : jet/feedback coefficient, pluck frequency in
The energy in channels 1-26 was concentrated in bands              multiples of heartrate
between approximately 6 Hz and 32Hz. I pitch shifted the raw       ch28-VNVB : body size
data by a factor of 34 to make these frequency bands audible.      ch29-HPHL : jet/feedback frequency
I tried hetrodyne analysis and resynthesis, but preferred the      ch30-HNHR : flute/string frequency
results of a simple rotating buffer technique because it           ch32-OrbOcc : breath
produced more high-frequency detail than the resynthesis. The      ch33-Mass : pluck position
pitch-shifted signals had a strong constant drone beneath          ch34-EDA : vibrato frequency
changing harmonics. I used fft to subtract the average
spectrum in each channel, which de-emphasised the drone and
accentuated the fluctuating harmonics. The channels are            2.2.4.   Geiger-counter style "pingers"
spatially positioned according to their recording locations in
                                                                   ch32-OrbOcc rate of ping, with 466 Hz fundametal frequency
the 1020 system. (BarrassT01.wav ... BarrassT26.wav)
                                                                   and    a   sweepable    spectrum    set   by   ch33-Mass.
                                                                   (BarrassT30.wav)
2.2. Channels 27-36
                                                                   ch33-Mass : rate of unpitched click, processed with a short
I let the raw values of these signals provide a chance starting    delay (1-10 msec) set by ch32-OrbOcc with lots of feedback
point for tweaking usable sounds out a variety of syntheses. I     for resonance. (BarrassT31.wav)
tuned the sounds by ear and by watching the numerical values
produced while adjusting the multiplication and offset of the      ch34-EDA : rate of reverse-enveloped ping, with a tone tuned
raw data values on the fly. The output signals were positioned     by ch32-OrbOcc, and with noise band-pass filtered at a
and adjusted in volume with an ear for being able to hear each     frequency set ch30-HNHR. The filter oscillates an octave in
sound separately at home on a four channel system.                 range at a fixed frequency of 20Hz. (BarrassT32.wav)
The data channels were mapped to synthesis parameters as
follows:
2.2.5.   Miscellaneous mappings                                     from STK (by Perry Cook and Gary Scavone) by
                                                                    Dan Trueman, Computer Music Centre, Columbia
ch35-Resp sets the band-pass frequency of filtered noise,           University.
which is then processed with a short delay (1-5 msec) set by
ch30-HNHR       with   99%     feedback     for  resonance.
(BarrassT33wav)


ch36-ECG and ch31-Erbs each set the frequencies of sin
oscillators.   These signals are mixed and distorted by
waveshaping then sent to a flanger with the delay time
modulated by ch31-Erbs. This signal is then filtered with a
parametric equaliser to provide the deep throbbing bass that
drives the piece. ( BarrassT34.wav)

              3.   ADDITIONAL REMARKS

I think that in order to focus on details in the sonification, a
listener would really need to be able to interact with the piece,
at least by being able to adjust volume levels of different
sounds, but probably other parameters as well. An interface
would need to be designed along with the sonification,
probably with a way to meaningfully change many interacting
parameters simultaneously.

              4.   SOUND FILES/POSITIONS

BarrassT01.wav = (Fp1)
BarrassT02.wav = (Fp2)
BarrassT03.wav = (F7)
BarrassT04.wav = (F3)
BarrassT05.wav = (Fz)
BarrassT06.wav = (F4)
BarrassT07.wav = (F8)
BarrassT08.wav = (FC3)
BarrassT09.wav = (FCz)
BarrassT10.wav = (FC4)
BarrassT11.wav = (T3)
BarrassT12 .wav = (C3)
BarrassT13.wav = (Cz)
BarrassT14.wav = (C4)
BarrassT15.wav = (T4)
BarrassT16.wav = (CP3)
BarrassT17.wav = (CPz)
BarrassT18.wav = (CP4)
BarrassT19.wav = (T5)
BarrassT20.wav = (P3)
BarrassT21.wav = (Pz)
BarrassT22.wav = (P4)
BarrassT23.wav = (T6)
BarrassT24.wav = (O1)
BarrassT25.wav = (Oz)
BarrassT26.wav = (O2)
BarrassT27.wav = (1.0, 338, 24)
BarrassT28.wav = (1.0, 48, 60)
BarrassT29wav = (1.0, 312, 62)
BarrassT30wav = (1.0, 158, 25)
BarrassT31.wav = (1.0, 264, 31)
BarrassT32wav = (1.0, 0, 48)
BarrassT33.wav = (1.0, 96, 34)
BarrassT34.wav = (1.0, 177, 77)

                     5.   REFERENCES

     [1] [1, 2] These instruments are Pd (Miller Puckette,
         http://www.crca.ucsd.edu/~msp) externals ported