When I built the Meng Qi Rollz-5 PC board, among the many modifications I made was to simplify the rolls oscillators by making them all just two-transistor circuits, i.e. ‘2-rolls’.  The two nodes were exposed on brown banana jacks, like the Plumbutter.  I was disappointed with the outcome for a number of reasons.

1. I made the cycles too fast in relation to the delays built into the AV Dogs and Gongs.  Those both have fairly long internal delays (adjustable on the Gong), but longer than the period of the longest Rolls.
2. The intent of the Rollz-5 was to get audio frequency glitches out of the odd-numbered Rolls, due to the paradoxical nature of the feedback loop.  I prototyped 3- and 5-Rollz circuits, but all they would do is oscillate around 100 KHz.  That’s why I didn’t initially build them.

I decided to experiment and turn the fastest 2-roll into a 3-roll.  It was easy enough to just remove one jumper wire I had put in to shorten from 3 to 2 and add the two resistors, transistor and capacitor.  When I tried it out I was surprised to see that it oscillated in a well-behaved way.  So I made a second.  The fireworks started when I patched the nodes of those together.  Now I started getting audio bursts.  I continued to modify two more to get four 3-rolls.  Now half are 2 and half are 3-rolls.  The outcome is a success!  I also modified the slowest 2-roll to be much slower, adding capacitors to get about 27uf.

In the following experimental recordings, the audio path goes from the Rollz oscillators into the Gong input (which I had added), so the Gongs act as low-pass filters.  Other paths go into the Ultrasounds and another takes the Gong outputs through AV Dogs.  I manipulated the mix, the Gongs, and also messed with the node patching.

The second recording has a different node patch, using all four 3-rolls.  I created a cycle of going from one node to the next and then repeating, the last one going back to the first, so all 8 brown jacks are patched.