Thanks, this would be a great solution. Meanwhile I collected the references for the relevant parts (below, if someone is interested ). The unfinished reverb will be omitted for simplicity and clarity, pickup replaced. I will take some time for the changes, if someone finds a part in the code the should be either more concise or more verbose, or a msitake, let me know. I also realized that I rewrote some things that seem to have equivalents in the library, I will most likely replace them.
Gabriel // Modulation synthesis with sparse convolution filter and distortions. // ==================================================================== // // A "3D" oscillator oscillating on x,y,z axis, with the radius of x,y going // into a hyperbolic distortion. // // The y axis oscillation is set by MIDI pitch, x and z are detuned by simple just tuned ratios. // Feedback acts on the individual sine oscillations. // // // Three weighted copies with time varying shifts are summed in a lossy integrator // ( sparse convolution ), followed by a peak filter and shaped by an asymmetric polynomial. // // The convolution tabs would give a (variyng) triangle impulse response if integrated twice, // with a -12 dB/octave rolloff and varying regular notches. // Here only one integrator is used. // // // The envelope is hard wired to the oscillation amplitudes and the rise time of the filter. // // An LFO is wired to pitch. // // // Inspired by the history of sound synthesis, namely Trautonium, Mini Moog, Phase Modulation Synthesis, // Variophon Wind Instrument Synthesizer, Physical Modeling, and the work of Thams D. Rossing. // // References: // Kot, Vítězslav. (2006). DIGITAL SOUND EFFECTS ECHO AND REVERB BASED ON NON EXPONENTIALLY DECAYING COMB FILTER. // https://en.wikipedia.org/wiki/Variophon // Parker, Julian & Zavalishin, Vadim & Le Bivic, Efflam. (2016). Reducing The Aliasing Of Nonlinear Waveshaping Using Continuous-Time Convolution. // Nicholas G. Horton, Thomas R. Moore. (2008). Modelling The Magnetic Pickup Of An Electric Guitar. // https://www.musicdsp.org/en/latest/Effects/86-waveshaper-gloubi-boulga.html, see comment from 2005-09-22 01:07:58 // Frei, Beat. Digital Sound Generation I & II, ICST Zurich University of the Arts // Smith, J.O. Physical Audio Signal Processing, http://ccrma.stanford.edu/~jos/pasp/, online book, 2010 edition On Thu, Sep 11, 2025 at 10:57 AM Stéphane Letz <[email protected]> wrote: > Hi Gabriel, > > I suggest we do it simple for now. If you can cleanup and document the DSP > code, then I can put in the examples/misc section: > https://faustdoc.grame.fr/examples/#misc > > Thanks. > > Stéphane > > > > Le 10 sept. 2025 à 18:43, ga <[email protected]> a écrit : > > > > Thanks > > > > I will look into installing Faust locally, I am bit deterred by the vast > amount of dependencies > > and my little experience with installing such projects. > > > > I also don't have much experience with make and compiling and C, > > but I think faust2rpialsaconsole might be onther option I have to look > into > > as running it on a Pi seems a reasonable solution for hardware. > > I do have a Pi 400, on which unfortunately the Patch OS which might be a > > good choice for OS does not run (or I didnt get it to run ). > > > > to 4) > > The code and concept is public and libre from my side, but maybe > licenses of third parties have to be considered. > > > > So I reused and altered code from the Faust library ( by Julius Smith I > think ) for the allpass delay, > > and the idea for the triangular filter was originally inspired by the > historic Variophon triangular oscillator, etc. > > so at least a proper note with history and references would be > desireable. > > Since the concept has a really long history with many sources and > variants, and is floating on my desk since years, > > it's a bit difficult to be accurate in this regards, and to do this > justice. > > > > The code also still needs some minor tweaks and cosmetic changes before > it is released in a 'final' version. > > For instance it uses two SVFs in series at the moment with very similar > corner frequencies, > > which could probably be replaced by a single SVF with a 'morphing' > output. > > > > A previous version had roughly antialiased synched noise (windowed with > a quarter sine wave) in a addition to the osciallator, > > to mimick a corpus impulse response, and to enhance piano and string > reminiscent sounds. > > > > I now tried to replace this with short allpass delays but it sounds less > convincing and "boxed", and setting > > the length of the allpass chain is also too arbitrary att the moment. > > > > Also noise has the interesting property that it has fluctuations, so a > seed could be matched > > to produce a sequence that resembles the derivative (or 2nd derivative) > of a real corpus impulse response. > > > > I would like to keep the paramter set to 4 though, as the idea for a > hardware interface is to have > > two rows with 4 push and turn encoders each, one row for synth and one > for EQ and other effects, > > with each encoder serving also as a button to select a set of 4 > parameters that belong together, like ADSR. > > ( sketch : > https://assets.steadyhq.com/production/post/c0d7b8ae-4d1f-4afa-afe8-8bcce17883ac/uploads/images/5prochhxdb/UI.jpg?auto=compress&w=800&fit=max&dpr=2&fm=webp > ) > > > > (Pressing two encoders in the corners simultanously could be used for > saveing and laoding presets.) > > > > This is one reason why the noise was omitted in this version. > > > > Such a controller should be seperate from the computing hardware and be > useful for many things, > > and could be easy to build from two I²C breakout boards from Adafruit, > > but I do not have the tools and funds for this at the moment, and it > requires > > additional code for interfacing, which I do not have experience with. > > > > The idea defintively is to make it an all open source and somewhat > flexible synth concept. > > > > An interesting aspect for me is that it touches and fuses many aspects > of the history of synthesis, > > and synthesis approaches, starting with the Trautonium, modulation > synthesis, subtractive, > > aspects and findings of physical modeling, etc, in a very compact but > meaningful parameter set and combination. > > ( less paramter than a Mini Moog I think, from which it also borroughs > of course). > > > > By this it is also a good simplified model to learn and teach I think, > for instance you could > > examine what makes a sound "pianoide" and then expand on this with real > pianos and real accurate modeling > > of real phyiscal forces etc., and then again examine their perceptual > significance and compare to this "cartoon" > > version, and many similar things. > > > > I dont know whats the best way to publish this so others can contribute > and expand on this. > > Maintaining and ovreseeing a project on Sourceforge or similar requires > a lot of work and energy and experience > > which I do not have. > > So I am also looking for interested people I can hand this idea over, > > Including it with Faust examples would be interesting in this regards, > but I am not sure it is fundamental and also simple enough for this, etc. > > > > Gabriel > > > > > > > > > > On Wed, Sep 10, 2025 at 3:14 PM Stéphane Letz <[email protected]> wrote: > > Hi, > > > > Thanks for this interesting code. For exporting the code, you have > several options: > > > > 1) exporting the DSP for a standard plugin format. > > > > - you can possibly use the JUCE export for that, as an > intermediate step: > https://github.com/grame-cncm/faust/tree/master-dev/architecture/juce. > For maximal flexibility the best would be to compile and install a local > Faust version. > > > > - another option is to use the Fadeli project > https://github.com/DISTRHO/Fadeli > > > > 2) you may find more info on this page > https://faust.grame.fr/community/powered-by-faust > > > > 3) you can connect to the Faust developer/user community on Discord > channel, see https://faust.grame.fr/community/help/ > > > > 4) You wrote « I am proposing the attached synthesis engine. » : Is the > code public ? Are you interested to contribute it in the Faust examples: > https://faustdoc.grame.fr/examples/ > > > > Thanks. > > > > Stéphane > > > > > > > Le 4 sept. 2025 à 11:53, ga <[email protected]> a écrit : > > > > > > Hello > > > I am proposing the attached synthesis engine. > > > It uses a "3D" oscillator that oscillates in x,y, z ( similar to FM / > AM) > > > The radius of x,y is fed into a pickup distortion, which goes into a > triangular filter ( 3 phase offset copies going into an integrator) an > asymmetric distortion. > > > It has only 4× 4 parameter, including classic ADSR and LFO, envelope > hardwired to oscillation amplitudes. > > > Its capable of a variety of semi- realistic sounds. > > > Sound demo is here: > > > https://youtu.be/7CBhMcYDWac?feature=shared > > > > > > > > > I would need some help to streamline the code mor Faustian, > > > to export including GUI, and to export including the effect, > > > and maybe ideas how to port this to some small hardware, Pi or Daisy > Pod ( though I doubt it will run there ). as well as opinion on the method > and ideas. > > > > > > Code: > > > > > > declare options "[midi:on][nvoices:8]"; > > > declare options "[-vec]"; > > > declare name "Paradigma_9 v007"; > > > declare version "0.0.7"; > > > declare author "gabriel"; > > > declare copyright "https://steady.page/en/voxangelica/";; > > > declare license "DWTW"; > > > // a synthesizer with "philonic" 3D spin oscillator and triangular > filter > > > import("stdfaust.lib"); > > > import("maths.lib"); > > > > > > // frequency ratios table > > > frtonum = waveform{1,16,9,6,5,4,7,3,8,5,7,15}; > > > frtodiv = waveform{1,15,8,5,4,3,5,2,5,3,4, 8}; > > > > > > // MIDI > > > midigrp(x) = hgroup("[1]MIDI",x); > > > f = nentry("freq",200,40,2000,0.1) ; > > > kmidi = nentry("key",69,0,127,1) ; > > > bend = ba.semi2ratio(hslider("bend[midi:pitchwheel][style: > knob]",0,-2,2,0.01)) ; > > > gain = nentry("gain",0.6,0,1,0.01)<:* ; > > > master = hslider("volume[midi:ctrl 7]",1,0,2,0.01) ; > > > gate = button("gate") ; > > > > > > // spin oscill params > > > rtogrp(x) = hgroup("[2]philonic",x); > > > rto1sel = rtogrp(hslider("[1]x[style:knob]",-12,-24,24,1)); > > > rto2sel = rtogrp(hslider("[2]z[style:knob]",19,-24,24,1)); > > > fbka = > rtogrp(hslider("[3]excentric[style:knob]",0.4,0,1,0.01)<:*:*(1/ma.PI)); > > > detune = > rtogrp(hslider("[4]warble[style:knob]",0.125,0,0.5,0.005)/ma.SR); > > > pickd = > rtogrp(hslider("[5]distance[style:knob]",0.7,0.25,1,0.0625))<:*:si.smoo; > > > > > > // LFO and Envelope Parameter > > > lfogrp(x) = hgroup("[3]envelope & lfo",x); > > > enva = (lfogrp(ba.db2linear(hslider("[1]A[style:knob]",20,15,66,1) > )/1000)); > > > envd = (lfogrp(ba.db2linear(hslider("[2]D[style:knob]",74,26,100,1) > )/1000)*envpscal); > > > envs = (lfogrp(hslider("[3]S[style:knob]",0,0,1,0.01) )); > > > envr = (lfogrp(ba.db2linear(hslider("[4]R[style:knob]",50,26,100,1) > )/1000)*envpscal); > > > lfof = lfogrp(hslider("[5]LFO Hz[style:knob]",3,0.1,12,0.1)); > > > lfvibra = lfogrp(hslider("[6]Vibrato[style:knob]",0.125,0,2,0.01))<:*; > > > > > > env = en.adsre(enva,envd*envpscal,envs,envr*envpscal,gate); > > > envg = env:_* gain; > > > > > > lfosn = qsin(mphasor(lfof/ma.SR)); > > > > > > // Triangular Filter Parameter > > > fltgrp(x) = hgroup("[4]triangulation",x); > > > wid = fltgrp(hslider("[1]rise[style:knob]",4.89,1,9,0.001)):2^_:1/_; > > > edge = fltgrp(hslider("[2]fall[style:knob]",6,1,9,0.001)):2^_:1/_; > > > fiq = fltgrp(hslider("[3]q[style:knob]",1.18,0.5,3.87,0.01))<:*; > > > hpon = fltgrp(checkbox("[4]highpass")); > > > drive = > fltgrp(hslider("[5]drive[style:knob]",0,-6,36,0.1)):_/20.0:10^_; > > > > > > rto1oct = rto1sel / 12 : floor; > > > rto1semi = rto1sel + 24 : _% 12; > > > rto1a = frtonum, rto1semi : rdtable; > > > rto1b = frtodiv, rto1semi : rdtable; > > > rto1 = (rto1a/rto1b)*(2^rto1oct); > > > rto1r = min((1/ rto1),1); > > > > > > rto2oct = rto2sel / 12 : floor; > > > rto2semi = rto2sel + 24 : _% 12; > > > rto2a = frtonum, rto2semi : rdtable; > > > rto2b = frtodiv, rto2semi : rdtable; > > > rto2 = rto1*(rto2a/rto2b)*(2^rto2oct); > > > rto2r = min((1 / rto2),1); > > > > > > // fve > > > lg2f = ma.log2(f/440); > > > stretch = 0.0333*lg2f; > > > envpscal = ( - 3 * lg2f ):ba.db2linear; > > > fplus = f*bend + lfosn* lfvibra*f * 0.5/12*envg + stretch; > > > > > > w = f/ma.SR; > > > w2 = rto1 * w; > > > w3 = rto2 * w; > > > wplus = fplus/ma.SR; > > > > > > fbk1 = fbka*(0.5 -w)^4; > > > fbk2 = fbka*(0.5 - w2)^4*rto1r; > > > fbk3 = fbka*(0.5 - w3)^4*rto2r; > > > > > > // modulation reduction per frequency > > > redux1 = ((3.3 -((rto1+1)*w) )/3.3),0: max:_^3; > > > redux2 = ((3.3 -((rto2+1)*w) )/3.3),0: max:_^3; > > > modep = envg; > > > modep1 = envg * redux1 *rto1r * gain ; > > > modep2 = envg * redux2 *rto2r * gain ; > > > > > > // sine oscillator > > > wrap(n) = n-( floor( n +0.5)) ; > > > qsincurve(x) = 1 - ( (x*x)<: *(1.2253517*16),(_<:*:* > (-3.60562732*16)):>_ ); > > > qsin(x) = x+(0.5): wrap <: (abs:-(0.25):qsincurve),_:ma.copysign; > > > // feedback depth reduction curve > > > fbcurve (x)= x:abs:-(1) <:^(3):_,(x):ma.copysign; > > > > > > // oscillator > > > mphasor(fw) = (+(fw) ~ (wrap)); > > > oscsn(fw, off) = mphasor(fw) + off:qsin:+~*(0.5); > > > osc1(fw, off) = ((fw),+(off):(oscsn)) ~ > (*(fbk2):fi.pole(0.5):_*fbcurve(fw)); > > > dcrem(x) = x <:_,_': -: +~*(0.999773243); > > > > > > // 3D > > > oscy(fw, off) = (osc1(fw, off )*osc1(fw*rto2+2*detune,0.75 + > off)*modep2)*modep; > > > oscx(fw, off) = (osc1(fw*rto1+detune,0.25 + > off)*osc1(fw*rto2+2*detune,0.25 + off)*modep2)*modep1; > > > oscxy(fw, off) = (oscy(fw, off)<:*),(oscx(fw, off)<:*):+:sqrt: > fi.zero(1.0);//dcrem; // > > > //oscxyb(fw, off) = (oscy(fw, off):fi.zero(1)) <:_,(_^2),((oscx(fw, > off):fi.zero(1):_^2)): _, (_+_):_,(_+0.1:_^(3/2)):_/_; > > > // with pickup > > > oscxyc(fw, off) = oscxy(fw, off) <:_,(_^2:_+pickd:_^(3/2)):/; > > > // > > > //synthvox(fw, ph2, ph3, g1, g2, g3) = (oscxy(fw, 0):_*g1), (oscxy(fw, > ph2):_*g2),(oscxy(fw, ph3):_*g3):>_ ; > > > synthvox(fw, ph2, ph3, g1, g2, g3) = (oscxyc(fw, 0):_*g1), (oscxyc(fw, > ph2):_*g2),(oscxyc(fw, ph3):_*g3):>_ ; > > > // triangulation > > > widredux = w <:+:_^3:1.0-_; > > > // diff to max f in octaves, reduced for higher octaves > > > dwo = ( 0.25 / wid ):max(_, 1): ma.log2: ma.inv: _*widredux: ma.inv; > > > //edge = 1/7; // falling triangle edge, > > > egderto = edge / wid; > > > wid2 = wid * (2^(dwo * (1-envg ))): > > > _* (2^(dwo * (1- gain ) )): > > > min( _, 0.25): max( _, 4 / (ma.SR/fplus)); > > > wid2e = edge: min( _, 0.25): max( _, 4 /(ma.SR/fplus)); > > > > > > fiw = wplus/wid2; > > > fiwtail = wplus/wid2e; > > > // triangle coefficients > > > apg0 = fiw; > > > apg1 = - apg0 - fiwtail; > > > apg2 = fiwtail; > > > // integration freq > > > igpole = 1.0-5.0/ma.SR; > > > resf = (fplus /( wid2 +wid2e) ): min( _, (0.249 * ma.SR)); > > > > > > // shaper > > > // x - 0.15x²-0.15x³ > > > tubicclip = _:min(_, (1.19419)):max(_,(-1.86086)); > > > //tubic(x) = x - 0.15*(x^2)-0.15*(x^3); > > > tubicilo(x) = x, > > > // normal for in < 1.2e-4 > > > ( x - 0.15*(x^2)-0.15*(x^3) ), > > > // ILO: > > > (( 0.5*(x^2) - 0.05*(x^3) - 0.0375*(x^4) ),(x <:_,_':- > :_<:(abs:max(_,1.2e-4)),(ma.signum):ma.copysign):/): > > > // select > > > ba.if( (_:abs:_<= 1.2e-4), _, _ ):dcrem; > > > // > > > superfbp = 1 - sin( 2 * ma.PI * w ); > > > > > > // make sound > > > process = synthvox(wplus, wid2, wid2e, apg0, apg1, apg2): > fi.dcblockerat(10.0): fi.pole(igpole) : fi.svf.peak( resf, fiq) <: > > > ba.if( hpon, fi.svf.hp( fplus/(wid+wid):min(_, > ma.SR*0.249), 0.707 ),_): > > > _*drive: tubicclip: tubicilo:_*(1/drive); > > > effect = _ * master:rev; > > > > > > // > ############################################################################################### > > > // CIELverb > ###################################################################################### > > > // minimalist reverb > > > // > > > // UI > > > revgrp(x) = hgroup("[5]reverb",x); > > > sizem = > revgrp(hslider("[1]size[style:knob]",0,-1.5,1.5,0.02)):(2.0)^_:_*16.7:si.smoo; > > > revt = revgrp(hslider("[2]revTime[style:knob]",60,40,80,0.1)): > ba.db2linear:_*0.001; > > > bright = revgrp(hslider("[4]brightness[style:knob]",90,52,112,0.1)): > ba.midikey2hz; > > > earlyl = revgrp(hslider("[5]early/late[style:knob]",0,0,1,0.01)); > > > drywet = revgrp(hslider("[6]dry/wet[style:knob]",0.5,0,1,0.01)) <:*; > > > > > > // reverb settings > > > // change revt with size > > > revtadapt = revt * ( 0.161*(sizem^3)/(6*sizem^2 )); > > > // diffusion delay times > > > revd0 = ma.SR * (sizem / 334); > > > revd1 = revd0 * 1 / ( 2 - log(2)); > > > revd2 = revd0 * 1 / ( 3 - log(2)); > > > revd3 = revd0 * 1 / ( 4 - log(2)); > > > revgn = 10^(-3*(( sizem/ 334 )/revtadapt)); > > > // diffusion allpass coeficients > > > revc = 0.707;//0.61803; // > > > revc1 = -revc * 10^(-3*(( revd1/ ma.SR )/revtadapt)); > > > revc2 = -revc * 10^(-3*(( revd2/ ma.SR )/revtadapt)); > > > revc3 = -revc * 10^(-3*(( revd3/ ma.SR )/revtadapt)); > > > // post (early) > > > revdp = revd3 * 1 / ( 4 - log(2)); > > > postd1 = revdp * 1 / ( 2 - log(2)); > > > postd2 = postd1 * 1 / ( 3 - log(2)); > > > postd3 = postd2 * 1 / ( 4 - log(2)); > > > postc = 0.382;//1/3;//0.61803; // > > > postc1 = -postc * 10^(-3*(( postd1/ ma.SR )/(revtadapt * 1 / ( 4 - > log(2))))); > > > postc2 = -postc * 10^(-3*(( postd2/ ma.SR )/(revtadapt * 1 / ( 4 - > log(2))))); > > > postc3 = -postc * 10^(-3*(( postd2/ ma.SR )/(revtadapt * 1 / ( 4 - > log(2))))); > > > > > > // left right delay time offsets > > > postdlroff = ma.SR * 0.15 /334 ; > > > > > > lfo1 = os.oscsin(0.13)*8.0; > > > apcomblp(maxdel,N,g) = (+ <: > (de.fdelay1a(maxdel,N-1.5)<:_,_':+:_*0.5),*(g)) ~ *(-g) : mem,_ : +; > > > // post diffusion (early reflections, placed after reverb loop) > > > postdiff( in ) = in <: > > > (apcomblp( 4096, postd1, postc1): apcomblp( 4096, > postd2 + postdlroff, postc2): apcomblp( 4096, postd3 - postdlroff*0.382, > postc3)), > > > (apcomblp( 4096, postd1 + postdlroff, postc1) > :apcomblp( 4096, postd2 - postdlroff*0.382, postc2): apcomblp( 4096, > postd3, postc3)); > > > > > > // feedback filter > > > dampp = sin( 2 * ma.PI * bright/ma.SR); > > > fidamp = _*(dampp) : +~*(1-dampp) <:_,_':+:_*0.5: *(revgn); > > > > > > > > > // reverb > > > rev = _<: +~ (apcomblp( 4096, revd1 - lfo1, revc1):apcomblp( 4096, > revd2 + lfo1, revc2): apcomblp( 4096, revd3,revc3): de.fdelay1a( 4096, > revd0 ):fidamp ),_:(_*drywet:postdiff),((_*(1-drywet))<:_,_):route(4, 4, > (1,1),(2,3),(3,2),(4,4)):>(_+_),(_+_); > > > > > > // END REVERB > ############################################################################ > > > > > > _______________________________________________ > > > Faudiostream-users mailing list > > > [email protected] > > > https://lists.sourceforge.net/lists/listinfo/faudiostream-users > > > >
_______________________________________________ Faudiostream-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/faudiostream-users
