Delaytrix - feedback delay network with 4 stereo delays

Hello I’ve finished a new patch that runs on Genius and Lich (and any other OWL2 device, I imagine). It’s a feedback delay network with four stereo delays that gives direct control over the feedback matrix values (positive or negative) in addition to input level and low pass filter for each delay. It’s clockable, has a freeze mode similar to Mimeophon’s, and an internal delay time modulator for vibrato/tape warble effects. Parameter assignment will make the most sense on a MIDI controller that has knobs in a grid, like the Midi Fighter Twister.

Here’s a long, but thorough video walkthrough:

Parameter A: the length of the first delay from 2 ms to 0.25 sec, 128th to 16th note when clocked

Parameter B: the overall feedback level (i.e. attenuator on the feedback matrix)

Parameter C: the amount that the four delay times are spread out, snaps to musical durations when clocked

Parameter D: dry/wet

Parameter E: a “skew” value that offsets the left and right delay times in each stereo delay by up to 48 samples in opposite directions, which can be used to stereoize mono signals

Parameter H: an internal delay time modulator with a frequency based on the first delay duration. above 50% is a sine wave and below 50% is smooth random. the modulator waves are output to parameters F and G.

Parameters AA-AD: gain for external audio going into each delay

Parameters AE-AH: low pass filter cutoffs for each delay output

Parameters BA-BD: delay matrix sends for Delay 1: 1 → 1, 1 → 2, 1 → 3, 1 → 4
Parameters BE-BH: delay matrix sends for Delay 2: 2 → 1, 2 → 2, 2 → 3, 2 → 4
Parameters CA-CD: delay matrix sends for Delay 3: 3 → 1, 3 → 2, 3 → 3, 3 → 4
Parameters CE-CH: delay matrix sends for Delay 4: 4 → 1, 4 → 2, 4 → 3, 4 → 4

Gate In 1: clock input, will drop into unclocked mode if the clock is slower than 60 BPM
Gate In 2: toggle freeze mode

Gate Out 1: a gate pattern derived from the four delay lengths, similar to how Mimeophon gate out works. In clocked mode these will be kept in sync with the clock.
Gate Out 2: high when freeze mode is enabled

1 Like


Our quality control department spotted slight artifacts when quickly changing delay length (i.e. a few times after 15 minutes mark in the video). Particularly noticeable in clocked mode when you change them with parameter C. So a bit extra smoothing may be required to avoid jumping too far in the buffer.

Btw, I’ll also be experimenting in FDN direction a bit, planning to make a quadraphonic reverb.

Yeah, I spent a little bit of time working on delay time smoothing. It’s tricky because I found that using the same smoothing value at all times doesn’t work well. It’s either too slow to follow parameter changes or too fast to prevent a lot of jitter when parameters aren’t changing. I have an idea about how to improve it and might take a pass.

Quadraphonic reverb sounds really cool! I mean quadraphonic in general sounds fun to play with, but I don’t have the setup for it.

I had the misfortune to listen to a talk by Susanne Ciani a few years ago which gave me the idea to try it. Now that we have a 4 channel OWL hardware (AC/DC, plus my port to Daisy Patch) it’s not unreasonable to give it a go. Btw, she doesn’t use a true 4 channel reverb, but 2 sets of stereo from Eventide H9’s, so there’s no spread between front and back planes inside the wet sound. But with FDN topology it’s not a problem to add more channels and I think that some older designs (i.e. Chowning) are also available in quad.

I’ve implemented what I think is a decent sounding solution for the flanging/glitching when sweeping delay times pretty quickly. Rather than updating delay times every block, I determine how many blocks to wait before actually setting a delay time based on the size of the change in delay time. This provides much cleaner transitions from very short to very long times on the fourth delay, particularly when sweeping time with spread all the way. The internal delay time modulation is still applied every block so you can get smooth pitch-shifting effects even with long delay times.

I’ve also tweak the “skew” parameter slightly so that when turned past 50% it starts to apply some cross-feedback in each stereo delay. In doing this, I discovered I had a bug in the feedback code that was using the left feedback path for both left and right feedback!

Thanks @antisvin for your feedback on this patch and others, they are better patches as a result!

I’d love to run this on a Befaco Lich or AC/DC. I imagine the AC/DC would work fine with some mapped MIDI controllers (since it’s based on the newer Xibeca board). I’m not sure if this would work on the Lich (there’s a comment to that effect on the demo video).

Has anyone had success trying this patch on either of the Befaco devices?

Yes, I’ve run it on my Lich and it should work. I didn’t notice that comment on the YouTube video, not sure why they are having trouble.

1 Like