Beads frozen delay questions

I’ve been having a wonderful time experimenting with looping audio slices in Beads, though I’d like to understand the relationship between Density, Time, and what in the buffer is actually played a little better.

Given delay mode, and all attenurandomizers at 12:00, with the Time knob all the CCW, it seems changing Density consistently changes duration while keeping the endpoint of the slice the same. When Time is elsewhere, the start and end time of the slices jumps around a bit. Notably, when Time is CW, the slice is only occasionally aligned with the beginning of the buffer.

As best as I can tell, Density is setting a duration. The buffer is then chopped into contiguous, mutually exclusive slices of that duration, aligned with the end of the buffer. Time then, as the manual says, selects which of these slices to play. (Now that I write this out, I realize that this is exactly what the manual says, save for the “contiguous, mutually exclusive” bit, which I suppose is implied). Consequently:

  • The distance from last the slice’s end time is the remainder of buffer length divided by slice duration.
  • As Density changes, the start and end times of slices change smoothly as long as that change doesn’t result in a change in the number of slices.
  • If changing density does change the number of slices, there may be a sudden change in the start and end times. This is easily audible when Time is all the way CW, but can happen elsewhere if Time’s position crosses an index boundary that result from the change the in the number of slices.
  • It’s impossible to change duration while keeping fixed start time. (Oh, I guess an exception to this is if you clock Beads, there are positions of Time where the start time will remain fixed across certain durations)

Is my mental model of how this is working correct?

A few other questions:

  • It often sounds like two copies of the audio are playing right on top of each other. For instance, if I play an unchanging saw wave, I hear phasing, and can see two saw waves interweaving on a scope. To ensure I wasn’t getting input audio leaking through (despite being fully wet), I confirmed it even happens when I remove the patch cable from my VCO. It happens whether or not the buffer is frozen. Is that intended?
  • I’ve noticed that, in addition to setting an envelope, turning the Shape knob will result in two copies of the slice being played, with the displacement corresponding to knob position. What’s going on there? Is that to get smoothly blending slices?
1 Like


Pseudo-code would be:

  • Use DENSITY (or the interval between clock ticks, possibly divided) to determine a slice_duration.
  • num_slices = buffer_duration / slice_duration
  • slice_index = value of TIME quantized into num_slices discrete values
  • loop between slice_index * slice_duration and (slice_index + 1) * slice_duration

Note that this is evaluated every time the loop end point has been reached (ie, changing the parameters won’t change the currently playing loop).

There are actual eight copies of the loop (or eight delay taps) being played and crossfading into each other. This is to account for:

  1. Changes in external clock rate: if a large deviation in the interval between clock pulses is detected, there will be a crossfade between the old interval and the new one.
  2. (a) The division ratio of the external clock controlled by DENSITY: there is a crossfade zone between ratios rather than a sudden jump from one to the other. (b) When the module is internally clocked and DENSITY is past 12 o’clock, there is a secondary, “polyrhythmic” clock superimposed to the main one.
  3. (a) When FREEZE is off, there is a crossfade between each of the multiplication factors controlled by TIME. (b) When FREEZE is on, the “bell” envelope shape controlled by SHAPE will actually play two loops with a 180° phase shift and a 180° phase-shifted envelope. This recreates the sound of rotary-head time stretchers, as found for example on old Akai samplers or the waveform scrubbing of some audio editors. The phase-shifted loop completely fades out when you are half-way to the next or previous envelope shape.

So that’s 2^3 = 8 simultaneous play heads, though in most situations, 7 of them will have a null gain.