gpud
Mind Map by , created more than 1 year ago

The second grouping of criteria from D jaffe's 1995 paper evaluating synthesis from the perspective of purpose, situation and aims for the user, rather than listing synthesis methods individually. Examples have been given to demonstrate each criteria. Part II focuses on the implementation and efficiencies that need to be considered for the processing environment, platform and system architecture, the limitations, constraints, power and memory.

21
0
0
gpud
Created by gpud over 10 years ago
Rate this resource by clicking on the stars below:
1 2 3 4 5 (0)
Ratings (0)
0
0
0
0
0

0 comments

There are no comments, be the first and leave one below:

Close
Jaffe Synthesis Evaluation part II criteria 6 & 7how EFFICIENT is the ALGORITHMhow SPARSE /DENSE is theCONTROL STREAMIaMEMORY REQUIREMENTSIb PROCESSING DETAILSc CONTROL STREAM ATTRIBUTESwhat CLASS of SOUND can be representedwhat is SMALLESTpossibel LATENCYdo ANALYSIS TOOLS existEXTREMELY IMP. - FOCUS OF MY DISSERTATIONDETERMINES:1. REALTIME2. IN NON-REAL-TIME CONTEXTS -TIME OF WAITING FOR RESULTS OF COMPUTATIONSIN A FIXED AMOUNT OF TIME TO COMPLETE A PIECE AN EXTREMLY LONGFEWER ITERATIONS == LESS-REFINED RESULTSTURN AROUND TIME =QUOTE: "determining theefficiency of an algorithmis more complicated than itmight at first appear. It isnot merely a matter ofcomparing processingbenchmarks. Numerousaspects of a technique andits implementation comeinto play,"The aspects of the GS techniques and their implementation that come into play canbe divided into 3 categories:- a memory, b processing requirements c control streamheavinessTRADE OFFwell known axiomSINE E.G.Amem -WAVETABLE SYNTHA single period of a waveformstored in memoryuse a non-interpolating'drop-sample' oscillatorNOS. VOICES AVAILABLEON GIVEN HARDWAREstored in memory ina HUGE TABLEEGA1memEGA2memstored in memory in aSMALLER TABLEa more expensive OSC thatinterpolates between samplesin the tableGreaterMemorylessprocessingOptimisedMem - butMEM < EG1PROC>EG1EGA3memPrecomputedresultant. WAV: ƒWAVE * A ENVMAX MEMMIN RTPROCSINE.G.Bproc -ANALYTICALCOMPUTATIONEGB1procEGB2procEGB3procmarginally stable 2-POLE FILTERSAlt. techniques for sinewave generationEVAL. of COMPLEX PHASORWAVEGUIDESGordon,Smith'85Smith,Cook 92Mathews '69VARIABLE MEM REQ. Changeswith the parameter valuesVm1 Karplus-Strong(plucked str)(waveguidebasedMODELLING): MORE MEMLO Pitch LESS MEM HIPitchII (b) PROCESSING POWER1. quiteCOMPLEX"2. depends on thePROCESSOR ARCHITECTUREpoorEXPENSIVE:good EFFICIENTFinite ElementModelling(numericalINTEGRATION ofDIFFERENCEEQUATIONS thatDescrb. Masses &SpringsEGIIBBUT wellsuited to: ARRAY -PARALLEL PROCESSORS1. active code fits withinthe CACHE of a RISCCHIP RUNS FASTER thanan active code set thatOVERFLOWS THE CACHEAFreed, Rodetand Depalle 932. With MINIMUM CHANGESin PROGRAM FLOWwellsupportedby:a. DSPprocessors& other b. HEAVILYPIPELINED ARCHITECTUREPURPOSE BUILTHARDWAREINCREASESENORMOULSYTHE EFFICIENCYOF A TECHNIQUEB Some techniqueshave a PROCESSINGREQUIRMENT thatCHANGES with thePARAMETER VALUESTime domainimplementation ofCHANTFormant WaveFunction FOFVm2 N harmonic partials in Trainlet synthesis -brightest timbre (chroma) = bandlimited pulse,softer timbre (low pass filtered, ie higherpartials attenuatedGOODdensity/heaviness of control streamProblem esp. for RT impl.may need 2 processors - then expensiveprocessing to get data from control processor tothe sound procesoreven wthjust 1processor -is theirenough DISKSPACE/BWfor the cntrlstream dataie is controlstreamdensity >RAMDispersion pattern:sys. FAIL wth sporadic,CLUMPED, LARGE DENSEBURSTS of parameterupdate mess,more easily manageable:REL STEADY STREAM OFWELL-SPACED FAIRLYDENSE CNTRL MESS.synth. byaddingoverlappingvocal tractIResp.more EXPENSIVE as FREQ RISESMORE PITCH PERIOD/SEC.MORE ADDONS and more TABLELOOKUPS PER OUTPUT SAMPLEwith MORE MEM/MORE PROC= NARROWER PEAK with 32Harmonics than with 8 -changes behaviour of thetechniquee.g. ENTIRE AMP ENV chngs withnote data, at be of each noteMore EFF. feed ENV DATA BY BREAKPOINT THRU THE NOTE 1 POINT AT A TIME1 to 5 relate to parameter behaviour, physical, intuitive, wellbehaved, perceptiblechanges, sound ID robust e.g. sounds good on more than 1 pitch/resonance of the instr.Double click this nodeto edit the textClick and drag this buttonto create a new node