Grids Pattern-Sorting Etc

This is one for @pichenettes.

I’ve been thinking again about the Grids firmware, and wondered if you cared to share any insights on the method you used to determine the arrangement of the 25 base patterns used in Grids.

Also, did you have an equally clever way to determine the trigger probabilities for each step/channel, or was that done by ear?


To sort the patterns:

  • Train a self organizing map using patterns of average complexity.
  • Arbitrarily divide your map in cells, and pick the pattern that falls the closest to the center of each cell.

To get the trigger probabilities:

  • Start with the pattern of average complexity at the center of your cell.
  • Search for the most similar pattern with one more BD.
  • Set the pattern you found as the new starting point…


How do you define ‘complexity’, in this context?


Did you try just arranging them arbitrarily, too?


Could the same principles be applied to patterns of notes/octaves/accents/slides?

I guess accents and slides, being simple on/off events could be treated in the same way, but you’d need some other method of arranging patterns based on note number.

Maybe you could redefine complexity in terms of variability in pitch, and apply the same principles.


May I ask, where the patterns originally came from? Were they programmed by hand? Extracted from audio (pichenettes, you seem to have quite some knowledge in this field)? Taken from some sort of rhythmic database?

They were manually transcribed from audio (extracted from commercial sample CDs), and used in some research I did around 2003 - 2004 ( ;

1 Like

I thought the patterns were organised using a Kohonen SOM? (Which is not the question that was just asked and answered, but also interesting and novel.)

Hopefully someone can help me here:

If Grids channel 1, 2 or 3 have external modulation to their respective fills is this:

  • added to the current choas ‘cv’ setting?


  • does the insertion of external modulation break the normalling of the choas modulation?

Eg: dummy cable into channel 1 fill, LFO into channel 2 fill and channel 3 modulated by internal chaos cv.

I am asking because I am interested in buying a Grids.

Also, do the fill inputs respond to negative cv or only positive cv?


  1. The “chaos” modulation is not disabled by patching a cable into the CV input of an individual channel. There is no normalling.

  2. The fill inputs respond to both positive and negative CV. If the FILL knob is fully CCW, you’ll need a +5V CV to “virtually” turn it fully CW. Reversely, if the FILL knob is fully turned CW, you’ll need a -5V CV to “virtually” turn it to fully CCW.