Controlled Randomization

[PaulSC]

Hi everyone,

I've just begun using Numerology and am very impressed by how powerful (and fun!) it is. I'm going to stick my neck out and make a feature request. If I've overlooked an existing way to achieve these results in N2, please guide me. Meanwhile, here's my idea…

Many of the existing modules offer the option of playing steps in random order. What I'm requesting here is a couple of new control modules that could be used to drive the existing modules, in order to provide varying degrees of controlled randomness.

For the sake of generality and efficiency, I've conceived these as separate modules that could be applied to any of the Note or CV Sequencers. One issue with this approach is handling the case where the random-control module has a larger number of steps than the sequencer to which it is applied. An easy solution is for the excess steps to "wrap around" in this case. Related mapping issues have equally straightforward solutions.

RandWeight module

A fixed random weight is assigned to each step 1 to N.

Example (the weight values can be increased/decreased by dragging; for better visual feedback and real-time control they might have associated sliders).

http://pnauert.googlepages.com/randweight.png

If a RandWeight module is configured as shown and applied to a MonoNote, then the MonoNote will play its first step 20% of the time, its second step 10% of the time, its third step 5% of the time, and so on.

StateTransitionMatrix module

The random weight of each step depends on the index of the preceding step. Thus this module implements a first-order Markov process.

Example

http://pnauert.googlepages.com/stm.png

If a StateTransitionMatrix module is configured as shown and applied to a MonoNote, then if the MonoNote plays step 1, there is a 50% chance that it will repeat this step, a 20% chance that it will proceed to step 2, a 15% chance that it will jump to step 3, and so on. Similarly, if the MonoNote plays step 3, there is a 50% chance that it will repeat this step, a 20% chance that it will jump to step 1, a 15% chance that it will jump to step 2, and so on.

At the beginning of the process, there is not yet a "preceding" step on which the random choice can depend; a provision is needed for the user to set this initial condition.

[PaulSC]

Minor point -- I would appreciate if a moderator could shorten the thread title to "Controlled Randomization", since my original title is too long and the interesting part is lost in truncation. I tried to make the change myself; I was able to alter the title at the top of the first post, but not what appears in the list of forum threads...

Cheers!

[jim]

Hmm, some interesting options, I will have to think about these for a while, but there are definitely some interesting ideas here. I'm definitely curious to hear more ideas in this area, as it does offer some interesting new directions.

In the meantime, you can 'fake' the weighted step option by using the random step mode with a sequence. In that case, you can add more weight to a certain value (or note for a note sequencer), by repeating that value (or note). For instance, lets say you have a four-step CV sequence, with the values:

1.0, 2.0, 3.0, 4.0.

If you set the sequence to play randomly, then each step has equal weight. If you repeat one step, and increase the size of the sequence:

1.0, 2.0, 3.0, 3.0, 4.0.

Then one value (3.0), now has twice the weight of the others. This definitely isn't as precise as assigning specific weights to each step, but is still pretty useful.

Also, I renamed the thread, as requested...

Cheers,
Jim

[blurk]

I quite like the idea of the random weight module because unlike N2's current random step approach, what I would like to do is keep the order of steps intact (i.e. preserve the essentials of the melody / rhythm) but occasionally play or not play a step.

I'm not sure I've expressed that clearly enough, but I hope you get the drift. With Jim's "repeated notes to fake the weightings" approach, you will get the random weightings, but the current random step will keep shuffling the melody.

[jue]

Hello,

please have a look to this:

http://five12.net/showthread.php?t=5

possible this is some kind for what you are looking.
The controlled randomness is possible with Numerology by setting up big CV sequencers to simulate the % by putting in the stepvalues as often as you need them. Value C3 twenty times means 20% .... if you take a CV seq with 100 steps.
A little bit of work but it is working for me.
Even some kind of weight.
Most of my projects contain this:
A VelocitySeq with 32 steps, 24 steps are equally distributed over the velocity range and the other 8 dublicate some values were I like to have the velocity most of the time. This Seq runs in random mode and for each note there is a kind of controlled random velocity with a weight to some values.

best
jue

[PaulSC]

Jim and all,

Thanks for the replies and the additional ideas.

I'm going to enjoy exploring the technique of weighting events by duplicating them in the sequence. The main disadvantage of this approach is that it's not practical to adjust the weights in real-time (except by modulating the length of the sequence to exclude some of the duplicates). Still an interesting approach.

[CF3]

+1 for this idea. CGS does a module like this, would love to see it applied to Numerology. also take look at the FutureRetro Revolution and Orb sequencers. the way you can have 256 variations of a pattern is in itself a kind of random sequencer and would be awesome for remix work