Wavetable scanning: is this possible?

[elassi]

Wondering if scanning of a wavetable has to be always from bottom-to-top or vice versa when we look at a 3D-representation of a table like shown in numerous WT-synths.

Why not - alternatively - from left-to-right or vice versa? *scratches head*

See the well known scanning we're used to:


Now ... would it be impossible to scan from left to right or right to left here - meaning implement such a readout alternative?

[EfreetiSultan]

Yes, pwm

Edit. Maybe that's not what you wanted now that I think about it. Pwm would only affect the one single oscillation of whatever position is currently being scanned. What you want is to basically flip the whole wavetable by 90 deg? In that left/right would now be up and down..

But to do it simultaneously would be a cool effect

[elassi]

Yep, exactly, flip by 90 degrees.

I guess then the sample length (say 2048) must be the same in both directions, so the wavetable has to be a perfect square.

[Funky40]

This VCO does it in an arrangement of 3 axis.

http://www.industrialmusicelectronics.com/products/21

it´s at first a thing of how you arrange the WTs. Then add the appropriate controls.
They could do this at any time.
Nothing new for the eurocrackers. This exists since loong.

musically wise ?
I prefer the one-axis approach. it´s way way more predictable.
For example to set up modulations. Or doing manual jamms.
There, you can develop "muscle memory".
This will happen much much harder with a two or three axis approach. If anything at all.
Yet, the three axis approach has other advantages.......it´s minor, for my taste.

i rate "to have a predictable control" higher, rather than to have more tonal options.
Thats the two sides of the spectrum. More tonal options vs. a more precise control ( and memorisable). You chose ....if there was any, haha

[GruvSyco]

I think it would be hard without converting the entire table to something else. Maybe re-synthesis? Current wave-tables are literally that. It's a linear group of single cycle waves. You scan across the waves just as you would rub your finger across something. You are talking about reading a bit from each wave in the table at a time. The 3D view that many wave-table synths give you is kind of misleading in that way. It's not really just a 3 dimensional grid of points.

[gentleclockdivider]

Wondering if scanning of a wavetable has to be always from bottom-to-top or vice versa when we look at a 3D-representation of a table like shown in numerous WT-synths.

Why not - alternatively - from left-to-right or vice versa? *scratches head*

See the well known scanning we're used to:


Now ... would it be impossible to scan from left to right or right to left here - meaning implement such a readout alternative?

Because this way the shape of the individual waveforms are clearly shown .
If you would rotate the wavetable the waves would either be subject to perspective or orthographic perspective ( if you want to keep their shape intact ) but this also means waste of green estate

[elassi]

I'm not talking about the visualization but indeed about the resulting audio readout of the waveform. It's defo another result when you scan the table from left to right instead of top to bottom (and each vice versa).

Therefore I suspected that the table has to be a square in that case to provide the same specs (i.e. 2048 samples).

[elassi]

This VCO does it in an arrangement of 3 axis.

http://www.industrialmusicelectronics.com/products/21

Interesting.

[whyterabbyt]

I think it would be hard without converting the entire table to something else.

Honestly, no.

Current wave-tables are literally that. It's a linear group of single cycle waves.

Yes, but its basically one index into a linear array of single-cycle waves, (which are themselves a linear array of single sample values). Basically its already a 2-dimensional array. The actual code could even be treating it as such...

newsample = array[wavetable, sampleposition]

And multidemensional arrays, ie with multiple indices, are just as easy.

newsample = array[wavetableX, wavetableY, wavetableZ, sampleposition]

edit : (sorry, just noticed this bit)

You are talking about reading a bit from each wave in the table at a time.

Not necessarily, though I may be misinterpreting you. Its entirely possible that the full single cycle is read in its entirety, and the wavetable index is only updated once per read through that whole single cycle.

[whyterabbyt]

I'm not talking about the visualization but indeed about the resulting audio readout of the waveform. It's defo another result when you scan the table from left to right instead of top to bottom (and each vice versa).

You cant scan a linear table 'up and down' and 'left and right' as well. Its either/or. For the table to be scannable in two dimensions it needs to be a two-dimension table in the first place.

Or treatable as one. If you treat a 'common' linear table of single cycles as a 2D array of N wavetables of X samples per wavetable, and increment the wavetable/sample indices in the 'wrong order' you'll get something but TBH Im not convinced you'll get anything like a useable result.

Therefore I suspected that the table has to be a square in that case to provide the same specs (i.e. 2048 samples).

same specs as what?

[Gamma-UT]

I'm not talking about the visualization but indeed about the resulting audio readout of the waveform. It's defo another result when you scan the table from left to right instead of top to bottom (and each vice versa).

Therefore I suspected that the table has to be a square in that case to provide the same specs (i.e. 2048 samples).

Without some fancy interpolation, I think you'd probably wind up just getting noise or DC by scanning in the orthogonal direction, unless you do something akin to what Ensoniq did with transwaves to make sure their wavetables were always nicely behaved. You might get somewhere with a sinc(x,y) waveform for example but once it gets a bit more arbitrary, none of your waves line up nicely.

[Olnium]

Toybox Audio has a wavetable module for Reaktor in its Synth Pack that lets you scan a wavetable in 2 dimensions.

[Ah_Dziz]

Look into phase distortion synthesis combined with wave table scanning. The phase distortion would be changing the shape along the x axis while the wave table scan would be along the x axis as usual. Not a super complicated process in any modular type system.

[gentleclockdivider]

I'm not talking about the visualization but indeed about the resulting audio readout of the waveform. It's defo another result when you scan the table from left to right instead of top to bottom (and each vice versa).

Therefore I suspected that the table has to be a square in that case to provide the same specs (i.e. 2048 samples).

I must be missing something but almost every wavetable can be read in the opposite direction
Just set you pointer to the last index and invert the amount (mod index ) of the envelope

[pekbro]

This is from the Anyma V manual, wave table type: Artin WMAP:

Artin WMAP (WMAP)
2D wavetable synthesizer.
Pitch source PITCH
The pitch source to follow.
Pitch (semitones) SEMI
The pitch correction on the oscillator.
‐64 to +63 semitones
Pitch (fine) FINE
The pitch correction on the oscillator.
‐100 to +100 cents
X scanning X
The horizontal position in the grid.
0 to 100%
Y scanning Y
The vertical position in the grid

I like Anyma V a lot, it's got much cool stuff in it, and it's ridiculously efficient since it's designed
to run on an ARM chip I guess. Seemingly, my core i5 is quite a bit more powerful than the chip
inside the Anyma V HW synth.