How to Randomize Step Sequence LFO

[CableChannel]

Another problem

I would like to have a step sequence LFO that produces random value steps (of fixed length) throughout the song, never repeating a pattern. In the LFO editor I can create step sequences of random values with up to 32 steps. This sequence is repeated. Is there a way I can retrigger the random sequence generation during the song? Or any other way to create an endless sequence of random steps?

I tried using the "Random" tab and then discretizing the LFO into steps via the transfer shape. That works but the Random thing does not really produce random discrete values, it produces a random continuous movement. So I tried to modulate the "Random" LFO onto a regular Noise LFO but that sounded kinda odd, somehow not only the step values but also their length seems to be random.

So... maybe forget all that stuff above. Question is: how to generate an endless sequence of random step values (with fixed step length)?

[MeldaProduction]

Well, this isn't really easily possible. But here's a little hi-tech scientific trick :

1) Use one modulator as step sequencer, attach it to whatever you want to modulate.

2) You can now modulate/automate 32 steps of each step sequencer, so attach all of the 32 steps (or how many you need) to a multiparameter (currently you have to do that manually, I'll check if it can learn it, it's too "deep" though)

3) Set the multiparameter to "banks" mode, use many banks and for example randomize each of the banks.
Now if you move the multiparameter, you get a random step sequence.

4) Use another modulator to control the multiparameter, again in step sequencer mode, but with much lower sync rate than the main step sequencer. In fact, if the main step sequencer was in say 1/2 sync and you have 2 steps here, you should sync it to 1/1, if 4 use 2/1 etc. So you may for example use 128/1 .

See what I mean? You basically use one step sequencer to switch sequences in another step sequencer via a multiparameter. This won't be an infinite random sequence, but it can be very very long . If you like math, you may enjoy complicating things by using "non-straight" sync to switch "in the middle", where you can get extremely long pseudorandom sequences .

[Gone soft]

- dropping by and taking notes -

2) You can now modulate/automate 32 steps of each step sequencer, so attach all of the 32 steps (or how many you need) to a multiparameter (currently you have to do that manually, I'll check if it can learn it, it's too "deep" though)

If those parameters can't be learned I'm thinking that process could be a little more efficient if we could add more than one parameter at a time, for instance by pressing shift or some other appripriate key while adding more parameters. That would save us the scrolling through the list of parameters every time etc. It should work fine as long as we want the same range for all parameters as we do here, right?

[MeldaProduction]

"Multiadd" would be quite problematic to do, but the good news is, it will be available for learning from next version .

[CableChannel]

Wasn't there a rule "Don't modulate the modulators, because it's going to create a black hole?" or something?

Thanks Vojtech for the explanation, I will try it!

[Gone soft]

"Multiadd" would be quite problematic to do, but the good news is, it will be available for learning from next version .

Is it possible to add learning for the shape of the LFOs too, the parameter for the LFO's journey from below sine to noise and mess? Would be swell if it could.

Wasn't there a rule "Don't modulate the modulators, because it's going to create a black hole?" or something?

Does that mean I can get rid of my annoying neighbour and all the shouting from next door? I'll keep modulating those modulators while directing the sound over there and see what happens...

[MeldaProduction]

V-GER: Yep, all of the stuff from modulators should be "learnable" (is that a word?? ) with next version .

And about modulating modulators , well, you must not create a cycle - is you use modulator 1 to modulate modulator 2 and modulator 2 to modulate modulator 1, then the results may be, ehm, a little unpredictable .

[CableChannel]

1) Use one modulator as step sequencer, attach it to whatever you want to modulate.
2) You can now modulate/automate 32 steps of each step sequencer, so attach all of the 32 steps (or how many you need) to a multiparameter (currently you have to do that manually, I'll check if it can learn it, it's too "deep" though)
3) Set the multiparameter to "banks" mode, use many banks and for example randomize each of the banks.
Now if you move the multiparameter, you get a random step sequence.
4) Use another modulator to control the multiparameter, again in step sequencer mode, but with much lower sync rate than the main step sequencer. In fact, if the main step sequencer was in say 1/2 sync and you have 2 steps here, you should sync it to 1/1, if 4 use 2/1 etc. So you may for example use 128/1 .

Yeah that works out! I was fooled for a while when playing one bar in a loop in Cubase, that repeats itself exactly, but if you play many bars in a row the sequence always changes.

One thing: the random generator in the Multiparameter banks setup behaves very oddly. It produces a lot of -100% and +100% values for my setup, I have to hit the random button quite several times to get a nice random looking set of values. Also, if I hit the random button a dozen times the generated values seem to repeat themselves.

[MeldaProduction]

Yep , it's great, because it always behaves the same way .

About the randomization - it uses the "smart randomization", which generates sort of "reasonable values". In most random buttons you can use ctrl/shift/alt combinations to force some behaviour (e.g. true randomization, which generates totally random values), but it's not there, I'll add it to to-do list.

[CableChannel]

About the randomization - it uses the "smart randomization", which generates sort of "reasonable values". In most random buttons you can use ctrl/shift/alt combinations to force some behaviour (e.g. true randomization, which generates totally random values), but it's not there, I'll add it to to-do list.

That'd be awesome!