Hello,

I am working on a Freeverb using vanilla objects. In contrast with [freeverb~] it should match with samplerates other than 44.1 KHz.

Freeverb is explained here by Julius O. Smith:

http://www.dsprelated.com/dspbooks/pasp/Freeverb.html

Apart from small details [vfreeverb~] is operational. It needs (performance-)testing. I'd be eager to know if it works on tablet platforms.

Katja

http://www.pdpatchrepo.info/hurleur/vfreeverb~.pd

I think the damping is supposed to be a 1 pole lowpass filter that controls the relative decay time at Nyquist. At low values it should be quite dark sounding. If it's not doing much, it might not be implemented correctly.

edit: maybe this won't work so well if you're using a high sample rate...

It's helpful to have additional tone controls, but I wouldn't use a [vcf~] for that. I prefer using a [lop~], some higher order filter with a very steep cutoff (for a "vintage" sound) and some form of parametric EQ.

Anyway freeverb is a basic Schroeder reverberator. It's not that difficult to make better algorithms.

Hello,

Patch vfreeverb~ is a literal translation of freeverb~, and damping is just as bad as in the original. Most important is incorrect placement of the filters: after the delay line outputs. So the first reflection is not filtered. Further, the frequency range doesn't go low.

Attached vfreeverb~alt.pd is a quick fix in this sense. By the way, freeverb (the routine) is put in the public domain by the original author, Jezar.

Acreil, do you happen to have good fdn reverb algorithms? I'm currently trying to understand gverb/gigaverb but having a hard time with the code. Freeverb~'s diffusor is good, and I'd like to try this in combination with fdn. The aim is to write an efficient yet nice sounding reverb in C, for small devices like Raspberry Pi.

Katja

http://www.pdpatchrepo.info/hurleur/vfreeverb~alt.pd

Maybe you could (if you haven't already) check the website of Jeroen Breebaart (een landgenoot van jou als ik me niet vergis) who makes very nice plugins, among which the cool and very cpu efficient omniverb?

D.

Thanks for the hint Domien. Jeroen seems to be an interesting developer indeed. Unfortunately for us, he has many patents on his name, and no open source code.

Katja

@katjav said:

Acreil, do you happen to have good fdn reverb algorithms? I'm currently trying to understand gverb/gigaverb but having a hard time with the code. Freeverb~'s diffusor is good, and I'd like to try this in combination with fdn. The aim is to write an efficient yet nice sounding reverb in C, for small devices like Raspberry Pi.

Not really. I've experimented with a 16x16 FDN. I wanted to make a variable matrix using nested rotation matrices, so that the diffusion is adjustable. I think it never really got finished. I also made a 4x4 allpass FDN (from a Mark Poletti patent). It would be good as a diffusor for an FDN, but I'm not convinced that it's really very useful.

Allpass loop topologies (like the Dattorro ones) are also efficient and nice sounding.

The FDN matrix thing was basically nested 2x2 rotation matrices at 45 degree increments. A 16x16 matrix is 4 levels of that, so you'd have 4 diffusion controls. If it's restricted to 45 degree increments, you can do the whole thing with addition and subtraction, and rescale later (lumped in with the decay time scaling). And I think if you can arbitrarily reorder the columns and rows, and switch their polarity, you can make about any matrix you want, not that it would necessarily be intuitive.

It ended up having too much signal routing stuff to really be sensible, but I think a C implementation would work better.

>Freeverb~'s diffusor is good
Indeed, it's a great effect in itself! With small amounts it makes artificial sounds more interesting. But what you really got to try is higher amounts to feed a distortion...

@EvilBird said:

Hi katjav,

This patch sounds great. I was looking for a nice verb to use on an iOS app, so first I want to ask what the license is (BSD/GNU)? Second, I wasn't getting much contrast in the verb sound across the range 0-1 for values of the damping parameter. I added a vcf~ object after the output with Q=1 and center frequency in the range of 200-20000 Hz, then synced this with the damping control. This produced a nice range of verb colorations, but I was wondering if there's anything I might be missing about the patch that would cause damping not to work? Thanks!

I am also interested in whether or not I can use this in an app?

KatJa thank you for making this! I really like the lingering color that it's producing in the low level. I had been looking for a breakdown of the Freeverb abstraction that I once had to use it in Vanilla but so far had not found. Thank you for making this variation.

Jim