--opcode-lib=/path/to/csound-nam.so -+rtaudio=null -o dac -d -m0 ; invoke: csound --omacro:AUTOV=1 --omacro:OUTF=live.txt nam_bench.csd (AUTOV=0 -> seed.txt) ; must run realtime-paced (-+rtaudio=null): the async model loader is wall-clock sr = 48000 ksmps = 128 nchnls = 2 0dbfs = 1 #ifndef AUTOV #define AUTOV #1# #endif giSeg init 22 ; seconds per model segment giMeas init 6 ; measure window at segment end (settled) giN init 25 ; number of models instr 1 ; source: repeating 110 Hz saw plucks, identical for every model ktrig metro 2 aenv init 0 if ktrig == 1 then reinit PLUCK endif PLUCK: aenv expseg 1, 0.4, 0.05, 0.09, 0.0001, 10, 0.0001 rireturn asrc vco2 0.28, 110 a1 = asrc * aenv a2 = a1 ; step to the next model every giSeg seconds kt timeinsts kseg = int(kt / giSeg) kidx = (kseg < giN ? kseg : giN - 1) ; model directory path is machine-specific: edit before running aoL, aoR NAMProcess a1, a2, kidx, \ "/path/to/models", $AUTOV, 0 ; peak 300 ms RMS in the settled window (last giMeas s of each segment) krW rms aoL, 0.53 ; ~300 ms time constant krD rms a1, 0.53 kinwin = (kt - kseg*giSeg) > (giSeg - giMeas) ? 1 : 0 kpkW init 0 kpkD init 0 if kinwin == 1 then kpkW = (krW > kpkW ? krW : kpkW) kpkD = (krD > kpkD ? krD : kpkD) endif ; at each segment boundary, write the result for the finished segment klast init 0 if kseg != klast then SName NAMActiveName kerr = dbamp(kpkW + 1e-9) - dbamp(kpkD + 1e-9) fprintks "$OUTF", "%d\t%s\t%.2f\n", klast, SName, kerr kpkW = 0 kpkD = 0 klast = kseg endif outs aoL, aoR endin i1 0 573 ; 26 * 22 + 1 : one extra segment so model 24 gets written e