Eclipse: the dsize parameter?

Diffusion is the amount of smear that is introduced as time progresses.  It’s essentially a dense field of delays created by allpass filters connected in series.  Because they are allpass filters the higher frequencies tend to get delayed, while lower frequencies don’t (this is due to the phase shift at the centre frequency).

This is what the manual says about diffusion.

0 to 1.0. The amount of diffusion. A value of 1.0
yields maximum diffusion or recirculation in the
allpass filter. A value of zero yields no diffusion.

This is what the manual says about the diffusion size

0 to 1.0. The size control scales the delays of all
of the allpass filters. A value of 1.0 gives the largest
size and a value of zero gives the smallest.

I hope this helps.

That’s pretty much right.

Diffusion, kind of emulates what happens to the higher frequencies in  reverberent spaces, where they lose energy and coherence as they bounce off surfaces, and get absorbed.

Diffusion is normally a part of the late stages of a reverb.

Generally reverbs get modelled like this.

Lets try to accurately model sound behaviour and reflections just after the sound is made. Phew, that was tricky, but I got pretty good results.

OK – lets try to accurately model the sound and reflections of reflections for a bit longer.  Ooooo, this is getting hard and it hurts my head, and I’ve suddenly run out of processing power.

OK – lets instead try to blur everything together and get it to kinda do the same thing that real spaces do by using some fancy maths like allpass filters and random delay stuff – Diffusion falls into this category.  It’s a part of the late reverb model.  Generally the more diffusion and smaller size will make the late reverb denser and more smeary.  Larger size and less diffusion will allow better definition of the sound in late reverb.

My last paragraph was referring to the diff size parameter.  Normally room size affects early reflections, and sometimes it is also wired into the diffusion network, so there is one control that controls both. Sometimes there are seperate controls for them.

Diffusion is more to do with the surface, and how frequencies roll off and reflections propagate. Depending on how this is acheived (and the diffusion block is only one of many mays to acheive this), similar parameters may be called adsorption, and sometimes there is a filter in the feedback path, so there may be a rolloff control.

Size tends to control the space between reflections.  Diffusion tends to control what happens to the “colour” of the reflections.  A large room with reflective surfaces may have a high time, but low diffusion. A small room covered in a highly irregular surface may have a low time but high diffusion.

It’s hard to say how diffusion is used in the particular algorithm. My impression from the documentation (I haven’t used this particular block in any of my algorithms yet) is that diffusion is kind of like a wet/dry for the diffusion network.  There is also a delay amount parameter and a g parameter, which control the delay for each allpass section, and g is the amount of feedback for each of the allpasses.

Your settings above are mostly right, but for a hard surfaced room, your decay will probably be long, as the soundwaves will bounce around for longer.

And as always, use your ears. Sometimes parameters can be counterintuitive when you hear them.  For example for a hard room reverb, you may just use and extend the early reflections set with lots of feedback, and turn off the late diffusion, which will keep things clean and create a better sense of space, but you can also create a sense of space with very little early reflections and a long spacious diffusion.

Yeah, it’s similar to density. Density is sometimes a bit of a combination of diff size and diff amounts but it’s a similar concept. How many reflections in the late reverb network

Sorry, missed your earlier reply.

The manual I’m pulling stuff from is the VSIG documentation for the individual programming blocks.

Bear in mind that the algorithm designer may have wired a knob to more than one destination, or may have maths between them and they can call the knobs whatever they want, so it may not exactly match the VSIG block.