Dream Synth: wavetable *everything*, even the filter.

VST, AU, AAX, etc. plug-in Virtual Instruments discussion
User avatar
KVRist
175 posts since 12 Jan, 2004 from Boston, MA

Post Sun Jun 20, 2021 5:33 am

Behringer asked today for people to describe their dream product.

This was my idea. Feel free to laugh. :D

I would DEFINITELY take this as a virtual synth, though, if any developer who happens to read this is interested in the idea. Happy to give more feedback, even! :D

To avoid burying the lede: what makes this idea unique is that wavetables are used everywhere it's possible, and the especially cool trick here is applying them to make a kind of "convolution filter."

There's a strange balance between flexibility and character that needs to be struck. For this synth, though, I want that character to come through in the nature of the oscs and filter, NOT through the (lack of) modulation.

Overall, the structure is a 5-octave, 12-voice, 2-osc (plus sub and noise), 1 filter, four-envelope, four-lfo synth. There are six (!) dedicated FX slots: two before the filter, four at the end of the chain, plus a dedicated EQ and limiter at the very end. There are 59 (!!!) knobs on this synth, so it's a bit of a beast. (For comparison, that's about the same as the Waldorf Iridium).

7-knob oscillator section: you can control one at a time with a button to select which. The controls include a knob to select which table to use, a knob for position, a sensitive knob for tuning (+/- 8ve), a knob for panning, a dedicated knob for ENV depth applied to the table position (osc 1 from env 3, osc 2 from env 4) and another for LFO depth to the position (again, lfo3 → osc1, lfo4 → osc2). There's another knob for "skew," which shifts every wave in the table according to an algorithm selectable by another cycle button (pulse width/squish left/squish right/fold/highpass/lowpass). 1 button to select octave (cycles through 32' to 2').

There's no ringmod or sync on this synth, nor "proper" FM. I find these modes to be redudant with wavetables. Pick the right table!

9-knob mixer: gain knobs for osc1, osc2, transient, sub, and noise, plus a knob to select the wavetable position of the sub (you cannot select the table; it's a simple-shape table only), a knob to select the color of the noise (white at 12 o'clock, brighter clockwise and darker anti-), and a knob to select which transient sample to play.. Another knob controls the detune (of the two oscs, mixed, but not the sub or noise) with a button cycling between unison modes (1/2/4/6/8).

A 3-knob pre-filter FX section, with a DAC simulator (one knob with a button to select a mode, to control the bitrate, resolution, and jitter), and three buttons to select a drive type (none, saturation, drive, fuzz at least) and two knobs: depth and tone.

The 6-knob filter section would be--get this--CONVOLUTION-BASED. Kind of like wavetables, but instead of "waves", it's scanning through spectra. You would have control over the "offset" of the spectra (to affect the overall pitch, akin to a normal cutoff), the speed of the scan, and, of course, which "table" of spectra it's scanning. A "resolution" knob here would make the transitions smoother or more steppy, for some old-school digital goodness. The available spectrum tables would, of course, include "models" of classical filter sweeps of all sorts with specific resonances--these would be easy enough to produce--but also some interesting models of physical resonators. There are dedicated knobs for ENV2 depth and LFO2 depth.

Four slider, two-knob envelope section. I'm fine with a single set of sliders (or knobs, but I prefer sliders) and a switch between which I'm controlling. I'd rather have four envelopes than two with dedicated controls. ADSR is fine, but I would add two knobs, to control velocity and curvature. ENV1 is hard-wired to 100% depth of mixer amplitude (which is not a knob), but can be mapped to other destinations in the matrix.

Six-knob LFO section, controlling four total LFOS in two selectable groups. Three knobs each: fade in, shape, and rate. The "shape" would be a single dedicated scannable wavetable that includes the classic shapes, then some complex variations, then a few rhythmic patterns, then much "noisier" chaotic shapes.

Four big, different-colored "performance" knobs.


HIGH quality 14-knob (!!) FX section. I don't think manufacturers realize how much the FX make the unit! Since this is my dream product, I would want four FX: another versatile drive (sat/drive/fuzz/crush) with depth and tone knobs; a versatile modulator (chorus/phaser/flanger/trem) with three knobs: rate, depth, and tone; a REALLY nice delay with 4 knobs: time, mix, feedback, and mod; and a REALLY nice plate reverb (make Strymon blush, man!), with 5 knobs: mix, decay, size, mod, and tone. Yes, 14 knobs is a TON to dedicate to FX, but that's part of the character of this unit: it is as much about the effects as anything else. And, yes, you cannot control the reverb type: you get a plate, period. The other effects would have some controls that you can edit on-screen (like width or mod rate), but these cannot be mod destinations. Again: character.

A 6-knob post-processing section. Four knobs control EQ (lo/hi shelf and two parametric--you can control the frequency and Q on-screen, the knobs only control the levels of some canned defaults). One knob controls a limiter (you can affect the release time and pre-gain on-screen). One knob controls the overall output gain.

There is no sequencer, no arpeggiator on this synth. It's part of its character, and this is meant for studio use. Also, no inputs and just a single pair of outputs plus a headphone jack, and, yes, they share the same overall gain knob.


5-octave keyboard, mod/pitch, poly aftertouch. Nothing fancy, but it really needs to feel quality. I like the IDEA of touch strips and X/Y pads and the like, but I never USE them. 😕

A nice, half-a-cellphone-quality screen, because this synth will involve looking at lots of wavetables and spectra. Four buttons each on the top and bottom of it, plus a clickable jog wheel or rocker to the right (I don't have a preference, really), to navigate within it.

For modulation: anything that is a knob can be a destination and nothing without a knob can be; no exceptions. Sources are "limited" to the env, lfos, wheels, knobs, keyrange, velocity, aftertouch, and a dedicated "random" function.

A big "mod matrix" button with a big "depth" knob to the right of it. When you hold it and wiggle a knob, the screen shows you what sources are available (and at what depth, if active) for modulating that destination. Push one of the buttons around the screen (env 1-4 at top, lfo1-4 at bottom) or wiggle one of the performance knobs or wheels to select a source, then use the depth knob to select the amount (pos/neg). The screen's jog wheel also lets you select keyboard range, velocity, aftertouch, and random as additional sources.

Sorry, software developers, but: I would suggest you do your best to allow a vast number of mappings. 16 at least, 32 would be more acceptable, but "unlimited" would be best. :-\

An SD card reader, where all of the wavetables, transients, and convolution tables would live, as simple files (I'm not sure what the simplest format is for each of these, but stick to the most common!) in three simple directories.

I would imagine this costing somewhere in the US$2000 range, plus or minus 20%.
Last edited by Introspective on Sun Jun 20, 2021 5:56 am, edited 2 times in total.

KVRist
86 posts since 31 Aug, 2020

Post Sun Jun 20, 2021 5:43 am

Are you dreaming of Dune 3 ? it does have that phat matrix, effects, oscs and mixers you describe :)

User avatar
KVRist
377 posts since 26 Jan, 2020

Post Sun Jun 20, 2021 5:45 am

I thought Genesis Pro was the one.
If the Earth was flat, cats would have pushed everything off the edge a long time ago.

KVRAF
1729 posts since 2 Jul, 2010

Post Sun Jun 20, 2021 7:03 am

How does cutoff modulation and feedback work with a convolution filter? Interpolating between kernels could be quite glitchy.

User avatar
KVRist

Topic Starter

175 posts since 12 Jan, 2004 from Boston, MA

Post Sun Jun 20, 2021 7:22 am

imrae wrote:
Sun Jun 20, 2021 7:03 am
How does cutoff modulation and feedback work with a convolution filter? Interpolating between kernels could be quite glitchy.
The convolution / spectrum "table" would contain a filter sweep, so you would "control" it by an envelope which controls travel through the spectra. Letting the table "run normally" (for whatever that means, given the table) should sound quite natural, but speeding it up or slowing it down or changing the "offest" (effectively the "pitch" of the convolution) would produce artifacts.

Indeed, I imagine this synth will be VERY glitchy by nature. That was kind of the intention: thus the DAC emulation in front of the filter, too, so your wavetable scanning can sound very old-school and PPG-like. This isn't meant to be an "all in one" by any stretch of the imagination, it's meant to have a very specific "sound," and I think that would largely come from the filter.

Interesting, neh? I was trying to suggest something a bit off the beaten path. I think it has potential.

KVRist
52 posts since 19 Jun, 2021

Post Sun Jun 20, 2021 10:40 am

Aside from practicality issues, considering that convolution based filters would be completely linear I don't think there's a real point to it as there wouldn't be much variation possible compared to a real time filter. And those got really good anyway. It mostly sounds interesting on paper.

I guess you should find Acoustica Audio products interesting though since this is their type of approach and they do actually go further than bare convolution.

User avatar
addled muppet weed
80673 posts since 26 Jan, 2003 from through the looking glass

Post Sun Jun 20, 2021 11:03 am

id want something purple.
with a load of knobs and sliders, all unlabelled, and all randomly connect with destinations, each time you tweak one.
id call it "the angry unicorn mk2" :)

KVRAF
6399 posts since 12 Feb, 2006 from Helsinki, Finland

Post Sun Jun 20, 2021 11:55 am

Introspective wrote:
Sun Jun 20, 2021 7:22 am
imrae wrote:
Sun Jun 20, 2021 7:03 am
How does cutoff modulation and feedback work with a convolution filter? Interpolating between kernels could be quite glitchy.
The convolution / spectrum "table" would contain a filter sweep, so you would "control" it by an envelope which controls travel through the spectra. Letting the table "run normally" (for whatever that means, given the table) should sound quite natural, but speeding it up or slowing it down or changing the "offest" (effectively the "pitch" of the convolution) would produce artifacts.
You can do a pretty clean implementation if you compute your oscillators and filters completely in the spectral domain, interpolate from one modulation frame to the next still in spectral domain and then construct the final waveform by using additive synthesis (or optimized DFT, but it's roughly the same thing). Time-domain manipulation of oscillator waveforms (eg. stretching, distortions, FM, whatever) is not entirely trivial using such an approach and I'm not sure if any synth specifically has "wavetable" filters, but otherwise it's a proven technique.
Preferred pronouns would be "it/it" because according to this country, I'm a piece of human trash.

User avatar
KVRAF
3254 posts since 22 Aug, 2019

Post Sun Jun 20, 2021 12:03 pm

This is the SOFTsynth forum...

User avatar
addled muppet weed
80673 posts since 26 Jan, 2003 from through the looking glass

Post Sun Jun 20, 2021 12:08 pm

e-crooner wrote:
Sun Jun 20, 2021 12:03 pm
This is the SOFTsynth forum...
first post says
I would DEFINITELY take this as a virtual synth,

User avatar
KVRAF
3254 posts since 22 Aug, 2019

Post Sun Jun 20, 2021 12:28 pm

vurt wrote:
Sun Jun 20, 2021 12:08 pm
e-crooner wrote:
Sun Jun 20, 2021 12:03 pm
This is the SOFTsynth forum...
first post says
I would DEFINITELY take this as a virtual synth,
Oh, sorry, I had not read that. In view of the length of the text I only ran over it, and I got the impression he or she was talking about a hardware synth...

KVRAF
28813 posts since 11 Apr, 2010 from Germany

Post Sun Jun 20, 2021 12:32 pm

You know... I think there's some truth to it. I thought more times about a hardware synth now that I can count, and, it always comes down to the possibilities in the virtual world vs. the limitations of possitibilities in the physical world.

I'm not sure if Mr. Uli really wants to do a virtual synth though. I'm not sure if it makes sense. There's so much competition, and, I doubt that Behringer has some top DSP people, that's why I'd stick to analog hardware, if I was him/them. Wouldn't make much sense for them to come up with a soft synth now.
Plugins and a DAW. On an operating system. Which runs on a computer.

rj0
KVRist
450 posts since 18 Sep, 2010

Post Sun Jun 20, 2021 1:35 pm

Introspective wrote:
Sun Jun 20, 2021 5:33 am
...

9-knob mixer: gain knobs for osc1, osc2, transient, sub, and noise, plus a knob to select the wavetable position of the sub (you cannot select the table; it's a simple-shape table only), a knob to select the color of the noise (white at 12 o'clock, brighter clockwise and darker anti-), and a knob to select which transient sample to play.. Another knob controls the detune (of the two oscs, mixed, but not the sub or noise) with a button cycling between unison modes (1/2/4/6/8).

...
I don't want a "mixer". Too mundane. I want a "combiner", something like LION has, but on steroids, maybe a morpher or two (with various algorithms), some cross-modulations (FM, etc.), then maybe filters, effects, etc.. Or just a configurable, modular routing, including all of this.

User avatar
KVRAF
5838 posts since 6 Jan, 2017 from Outer Space

Post Sun Jun 20, 2021 9:08 pm

chk071 wrote:
Sun Jun 20, 2021 12:32 pm
You know... I think there's some truth to it. I thought more times about a hardware synth now that I can count, and, it always comes down to the possibilities in the virtual world vs. the limitations of possitibilities in the physical world.

I'm not sure if Mr. Uli really wants to do a virtual synth though. I'm not sure if it makes sense. There's so much competition, and, I doubt that Behringer has some top DSP people, that's why I'd stick to analog hardware, if I was him/them. Wouldn't make much sense for them to come up with a soft synth now.
He wants to run his DAW in a keyboard workstation. And that DAW needs a super synth which can do everything… The Roland/Yamaha killer… Can’t say I am interested…

KVRian
979 posts since 21 Nov, 2005

Post Mon Jun 21, 2021 1:16 am

I am quite interested in the idea of convolution synthesis but the current concept of convolution, AFAIK, only really applies to static filter types (amp sims, etc) or reverb. The wavetable idea was one I had envisioned a while ago, as it would allow the spectral characteristics of one instrument to be used as a filter over another so that you could, for example, be able to play a voice through a piano's spectral and dynamic envelope. This sort of effect has been mostly acheived through spectral vocoding but so far I haven't heard a great example of this beyond some of what I've heard acheived with Kyma.

I think potentially that this might be more easily acheived through some sort of subtractive synthesis using one wavetable modulator (e.g. the piano) as an evolving filter removing unshared frequencies from the carrier (e.g. voice), rather than working out real-time spectral morphing (although admittedly I'm not a DSP designer so I may well be wrong). I'm sure there was an option to do this in the Emulator X (and the earlier hardware versions) but this was a destructive audio edit rather than a real-time operation.

Return to “Instruments”