generating random reflections with a DSP

[elektromeister]

Hello to all

I would have a question about reverberation generation:

First of all about the motivation: I have extensive knowledge in calculations with FEM for mechanics, had recently dealt with torsion and sound reflection calculations in materials and there very in-depth knowledge. This gave me the idea to program a small music application myself. I would like to implement some ideas of mechanical reflections for music generation and stumbled upon reverberation algorithms. I have access to a DSP system that can be used for real-time audio and also intend to buy something like MiniDSP or Hypex ...

My first idea to create something new would be artificial random reflections. I looked around and understood that random reflections (in time and angle, as far as I understand) could be very interesting and effective to improve the sound. According to some articles on the net this is a partially known thing, on the other hand it seems to be partially patented.

In any case it is tricky, because despite my knowledge I have no idea how to do it in detail:

If I assume that I have to create reflections by continuously writing and reading to a RAM and the data stream at the input never stops, the question arises how to create random directions. For example, if one uses panning via delay or volume, then the sound will inevitably break off...
as soon as you change the parameter. Parts of the sound will be moved, so to speak.

Any idea about this? I tried smoothing and several channels in overlay but had no success.

[elektromeister]

total silence? no programmers in here?

[Blue Hell]

Well .. i don't know about finite elements (I assume you meant with FEM) in relation to reverb and I don't know what methods could be used to make directional reverb.

What I do know is that when you set up delay lines of acertain lengh you canfreely modulate the delay time from zero to that length - this is not artefact free as there either ill be a click from a jump in time or the audio frequency/ies will change (as readout will go in a tempo differing from write-in).

I've made two reverbs for my synth software, one modelled after Freeverb and the other after GVerb .

Both implement a sort of 3D effect by making little differences between the left and righ audio channel. Also both use comb filters instead of plain delay lines, so there will be coloring.

And both are approximations from the era where real-time physical modelling would have been too CPU intensive ... which changed with the years but 've nver tried to do anything with that.