Description of feature (short and simple):
Vocoder Module
How will this feature improve the workflow or experience for all SSP users (keep it short and focused):
Just thinking of tools that SSP could do where other Eurorack modules fall short. Wondering if there is a VST that could be ported?
The SSP already has the required modules to patch together your own vocoder.
Iāve built this Patch in my head already, but have not tried how well it works in practice.
It could work as follows:
Feed an input signal to a number of SVFs in bandpass mode.
Each SVF-BandPass must be tuned to a specific frequency band. The number of frequency bands you want to use determine how many of the following chains you need:
1: INP ā SVF (In bandpass mode) ā EFL ā VCA (Control In)
2: OSC (tuned to same frequency as the Bandpass filter) ->VCA (Audio In).
Next you collect all the frequency band outputs from the VCAs and combine them in an OUT module.
The principle is to get a filter (SVF) to listen to a narrow frequency band. Use the Envelope Follower (EFL) to generate a control signal based on that frequencyās amplitude. Finally use that control signal to drive the amplitude (VCA) of an oscillator (OSC) that is tuned to the frequency band.
Give it a try and let us know hw you went. If I get around it, I will post a Patch in the forum.
1 Like
for superbooth, I created a 16 band filterbank with svf (*) , which is kind of the first half of thisā¦
(filter banks are a lot of fun!)
as you say, just add an EFL from the bands output, pass directly into the level of the OSC (they already have a ābuilt inā vca)
the fun, is tuning the bands (I based mine on the filterbank on fumana from frap tools) and also setting the parameters of the EFL.
this I guess is the āanalogā approach to vocodersā¦ (using filters)
the digital approach would be to use an FFT for analysis, this gives N frequency/amplitude outputs (as ābinsā) which you would drive N oscillators with. the āissueā is FFT are pretty cpu intensive (there is also an āanalysisā delay)
ā¦ of course, a āvocoderā would essentially be a combination of the fft analysis /vcas/ oscillatorsā¦ but I think in modular it be better to keep separatelyā¦ as you could then use with different oscillators and filters.
definitely worth playing with the analog version though, its not as āperfectā, but actually thatās likely to give it much more characterā¦ and allows for a little more āuserā tuning
( * ) as for using an existing VST, as Ive mentioned beforeā¦ its basically never worth itā¦
the issue is, you need to recreate a specific UI that works well with the SSPā¦ it also needs to compile for ARMā¦ and a host of other āadapationsāā¦ to do this you have to pretty much rip apart the VST and understand how it ticksā¦
by the time you have done that, its much easier to just look at the DSP code for inspiration etc (or find other similar open source DSP) and create the module from scratch.
(there are also other ways to create DSP code, for specific purposes like thisā¦ e.g. using something like FAUSTā¦ and then again, create the UI as required)
the issue is , VSTS are basically designed for desktops so have little in common with embedded hardware and also modular environmentsā¦
similar reasons why VCV rack module are generally not āportsā of vsts, rather fresh implementations.
(yes, I know the VST host module exists in vcv rackā¦ but that also demonstrates the limitations of that appoach e.g. automation mapping, but also only works because vcv rack is running on a full desktop, with mouse/keyboard)
i have given this module (vocoder) also a great deal of thoughts,
as i am looking for a way to superimpose/imply one sound source on another
-or play a carrier through a modulator so to say
all this above only is half of the story
since thereās no FFT (yet?) in the SSP this option is out of the question
rests the way of the filter, or to narrow certain frequencies so we can analyze and retrieve certain bands to extract the information we need/want from it
ideally this would be the frequency, amplitude and sequence/tempo/gate
i think amplitude & gate can be substracted through an EFL
but for the frequency of the signal to be analyzed into meaningfull information , the modulator signal needs to be (ideally) monophonic, which is not something iām interested in (compromising to)
you could tune the SVF into really small/narrow bands & end up with something like 128 bands, which puts us into FFT realms again
and all this to analyze the modulator signal, we havenāt spoken about the carrier signal yet, and how to superimpose it on the other signal
vocoders with a limited number of bands are very commonā¦
(actually, outside of vsts etc, Id suspect they are the most common)
first searchā¦found this
AnalogFX - VXC-2220 Vocoder, using the Synton Syntovox 222 vocoder, uses only 10 bands.
a GRP vocoder 22 (nearly 4500 euros!) uses 22 bands
the classic (again 5000+$) mood, had 16 bands
again, as analog these use the filter bank approach.
generally, Id say in modular, you are going to use a limited number of bandsā¦ since each band also becomes a kind of āvoiceāā¦ so 128 bands, would need 128 oscilllators/vcasā¦ so in physical modular, few are going to cope with even 16 bands 
of course, once we go digitalā¦ then its easier to create more bands, more oscillators etc.
(though of course, there, obviously, are limits with things like the SSP, compared to a desktop/laptop.)
also, I said, FFT is just ONE approachā¦ and it yields something very different, as the way it presents the bands (even if you use a limited number of bins) is very different, its not the same at all, as using filterbanks/bp filters.
basically there is no ārightā way to do vocoders (or frankly, any synth/fx) , all have different pros/cons, and also a different characterā¦
with vocoders, Ive tended to be driven by what they can do/sound like, rather than any idea of needing X/Y/Z, as I find that the more musical approach due to the fact they all vocoders have such an individual character.
honestly, though, not sure I have much interest in this areaā¦
as above, I do really enjoy filterbanksā¦ and I may well create one, as id like a spectral tilt function on itā¦ and whilst possible, by patchingā¦ it takes a few modules
but really, whilst I understand the various uses of vocoder and played around in this area before - Ive never really been excited much by the sound/effectā¦ perhaps Id need to play more in this area.
but as I said, filterbanks are much more āup my alleyā
I often use the vocoder on the Nord Modular I have and it is very useful for much more than your standard ārobot voiceā thing. Applying it to or using percussive loops can give very unusual results. The Nordās Vocoder is a 16 band one with the ability to assign any carrier band to any program band.
It looks like this -
3 Likes
Indeed Iāve used hardware vocoders for years and not once needed to make one ātalkā 
Instead of speaking or singing, Iāll use my voice to shape any sound with my mouth/lips/teeth/tongue (frankly the most powerful multi-mode āfilterā nature has provided us) in real timeā¦itās like having enough hands to adjust all the knobs on a fixed filter bank at the same time with great agility and precision.
Plus these human āfiltersā arenāt fixed at all, neither is the effective Q of each. The vocoder channels may (usually) be fixed, but your vocal Modulator input maps your complex noises across them, opening the available vocoder channels in subtle and organic ways which you could not replicate with two hands.
Instead of words, you make sounds like a child creating the explosions of an imaginary battle or rockets wooshing by. Anything. The Carrier can be simple white noise, a mixed piece of music, sound effects orā¦anything.
Another aspect of this non-verbal vocoding is that 8 or 10 bands can be plenty to achieve complex, evolving textures. You donāt necessarily need a 16+ channel vocoder to do musical things. It may not ātalkā convincingly, but you might find you donāt care.
1 Like