D U N E user waveforms

[Kriminal]

the CZ uses no samples.

[nix]

No, but the 8 waveforms r digital.
My clear cut understanding was that analog can be sampled, digital can not.
I think this may not be so.
Can u sample a digital additive synth for example?
I suggest u may be.
hmm, I dunno really.
I'm gonna hunt one of these things maybe. I like Casios,
have the MT-40, which has a cool little bass keyboard along w/- the main keys.
You definitely can't sample wavetable synths,
if the synth manufacturer says no.
I wanted to rip the Virus, so I asked.
I even have about a hundred tables from a NI user's instrument,
it used to be up on the web, may still be.
They said no, hmm, I asked elsewhere,
and recieved permission from 3 devs to cut their stuff.
It might usually be wise to ask with digital-
but I imagine most devs send a C&D before taking u to court.

[jammie1]

i can generate any wavetable in any size waveform samples 2048 is fine but does dune use an additive engine for the wave osc as 2048 samples in the sampling domain when up sampled through a transpose engine will give aliasing after about 2-3 octaves

being f0-24 cents in tune for a sampler instrument it would only give you a transpose engine of f3 or f4 before it started to fold back on its self with band limited waveforms

i can make very smooth transistions

i do a lot of sync of wavefroms as asr10 does not have sync so to get that sound i have to create wavetables to do that

same for resonant filter emulations as the ensoniqs dont have resonant filters

i can do robotic voice versions aswell or speaking voice but 70 waves limit would be dificult for that but the wavefusion would be fine

https://rapidshare.com/files/2300887779 ... vetable.7z

this is a wavetable that i did for the prophet vs emulation in the sharc dsp card of the scope system

i sent them this wav file so they could here it in full then i sliced it up into 256 samples for them to use as single cycle spectra

the only thing that is missing from scope is a user wave scanning osc like in wavefusion but with open scope comming it will be easy as you can write your own c++ dsp code and impliment its use on the dsp cards

but i can write smoth single cycles so that the morph over time for use in the dune

will dune read full audio path wavetables like the wavefusion or will it have to be single cycle 70 single waves of 2048 each

also what bit depth are they are they 32bits floating point as i do mine in 16bits due to being used in hardware samplers

[Mark]

Well it's all hypothetical at the moment what the format of the wavetable for DUNE will be if it is ever opened up. All I know is what is in the current DUNE so 70*2048 waves.

I'm guessing that Rich gets around the aliasing problem by dropping higher harmonics as you go higher in frequency.

Just downloaded the Prophet VS to have a look at. Thanks for uploading

Note that (AFAIU) WaveFusion prefilters all wavetables to reduce aliasing, which is even more of a problem with 256 sample waveforms than with 2048 sample waveforms.

[jammie1]

basically the same as hardware manufacturers did adding lowpass reconstruction filters to reduce the steping of the dac and to remove any content of aliasing from the frequency playback engine

i do the same with my modular patches in scope i add fixed lowpass filters that cut any frequencies above 16khz

as i dont see the point in having any higher frequency content as it becomes noise when transposed

i like doing 256 sample as thats the real size you can sample from digital and analog systems as any higher they get bandlimited by the reconstruction filter and start to resemble sin and saw waves

and in the ensoniqs with there oversampling dacs you can transpose up by 3 octaves with no aliasing which meeans you can get a full keyboard of waveforms

i also use them in the t1 but have to down sample them to 16khz so that i can transpose them up by 2 octaves as the antialiasing engine wont allow any samples above 64khz

[Kriminal]

No, but the 8 waveforms r digital.
My clear cut understanding was that analog can be sampled, digital can not.
.

a digital osc that produces its own waveforms is fine to sample, its only sample based osc that cannot be sampled and redistributed

[nix]

Thanks for clearing that up Krim.
It's kinda odd how that works,
because for me it's harder to generate an algorythmic waveform,
than to render a wavetable.
It's the circumstance,
I guess, but still, seems unfair for those who wish to not have there calculated forms emulated.
I like the CZ sound, have just had a listen.

[Christophe]

thanks for the ultran and wavefusion waveform !
and welcome to the forum jammie1

[jammie1]

i sent about 10 wavefusion wavetables and 1 test ultran file

im busy sampling at the moment to provide a 1gb+ soundware file for ultran

[Christophe]

@jammie1, if you have some time to provide some sound design hints to build wavetable, i will listen religiously because i have done a lot but in a chaotic way and i love Ultran

[jammie1]

im going to do an in depth explination from sampling sounds to creating a multilayer sample file using the in built sampler and then taking that file and creating a ultran text file

i do lots of single cycle sounds but they are static

my experiments at the moment are doing pwm of the static syngle cycle waves to make multicycle looped spectra that changes over time and is looped perfectly

this makes the single cycle waves sound very analog but with any type of digital waveform

these i am replacing the d50 pcm sounds with my multicycle spectra and using the actual la synth sounds

the choir test ultran file is a choirs created this way from my childrens singing voices i take the vocal files and put them in an fft partial program that gives me the additive spectra of 128 partials per slot size i choose of any sample

i then choose a slot and generate a waveform from the additive data for that given slot of the sample

from this i get variating waveforms over time

i then put them in a python script which which does a random read of the waveform over time so that each cycle it increments or decrements the start position of the single cycle wave when it gets to the end of its cycle it then adds 0 crossing data for the single cycle size until it reaches the destination sample size where it starts its next read cycle of the single cycle wave

by adding the number of samples increments per read cycle you can simulate many different pwm type modulations

it interpolates from the zero crossing to the next level of the read cycle

this is not done in real time as the math is high cpu usage to render

when rendering these files it can take minutes to calculate and use 100% cpu on a i7 comp

but the samples are great and sound fat on there own as they have animated chorus

and by layering them you can make really fat sounds

me for the ultan what i would like to see is a change for each of the 4oc to have there own text file that can blend into other sample files so that as it is you could set up a 4 layer sf2 file and then a txt file that can morph from 1 layer of the sf2 file to other layers

and then the other osc 2-4 could have different txt files that also blend into other files and have a seperate filter for each os

and a seperate vca for each osc that is controled by lfo or envelope then they each go to a seperate vca for transient vector mixing

that way you could have 4 osc each mix blending there own txt file and then being filter morphed any way you like with different filter topologies and then being transiented by the envelope and vca and then being able to be vector mixed

and also a vector mixing section for pitch like the sy22-35 range

that would be a special synth 2 vector pads with sequenced control so that you could generate patterns to create modulation of the vector mixers

[Christophe]

thanks you for this very long and detailed response
i hope this give Rich some ideas

[Richard]

i can generate any wavetable in any size waveform samples 2048 is fine but does dune use an additive engine for the wave osc as 2048 samples in the sampling domain when up sampled through a transpose engine will give aliasing after about 2-3 octaves

The standard way of dealing with this is to multisample the waveforms. Octave spacing is common, which is what Dune 1 uses. Interestingly this seems to already give a higher quality than hardware. For Dune 2 we'll use a yet higher density though, to further increase the quality of wavetable synthesis, so that it's as good as say a standard sawtooth.