... or the "big stairs," as my Aussie friends called it, is literally the face of a cliff with stairs carved into it, i believe of aboriginal origin. I was fortunate enough to get a recording of this soundscape with a well calibrated pair of condenser mics through some kind of pre-amp. Mitch hooked me up with the gear and persuaded my sister and me to come out to the site and be tourists. since then i've used the material as fodder for a few projects, including my recent studio retrospective, "Codecs, 2007-08 ."
this time, i used the material as input for my spectral tracing patch in supercollider. the patch consists of a filterbank that sends osc messages back to sclang regarding the spectrum of the incoming signal. on sclang i declared a function that would respond to the osc messages by writing csound score instructions into a file. the result was a csound score that was 102.5 megabytes! it is so big, i was unable to open it in JEdit! on the commandline, i could cat the file contents to the console, but the resulting data vomit lasted about a minute. what a huge file! i used a pre-filter to get slices of the spectrum and dynamic range, and smeared the spectral components by statistically re-tuning the grains by various octaves.
this tracing function is a bit like a charcoal rubbing. it often can result in a pretty close resynthesis if one takes care to get the appropriate slice (which is monitored and performed in realtime) and providing one sets up the appropriate resynthesis instructions on the other side (which can be performed in realtime but in this case was not). of course, i have better ways of making a clean resynthesis that reside completely in scsynth and thus are sample accurate and totally realtime, but the point here was to generate a csound score for grain instructions.
it is interesting to note here that csound is used for the mpeg codec. i have been looking at csound patches for wavelet analysis and they seem pretty simple and easy (for csound anyway) to implement.