MVerb 7 - yet another reverb?
Posted: Sun May 26, 2019 7:10 amHi gang,
my newest creation is this reverb device. Don't judge this book by its cover - I am not exactly known as a GUI artist.
Under the hood is a rather new reverb design: a single delay with feedback followed by a convolver. I highly recommend this device for churchy organ sounds. 
I have spent many hours experimenting with reverbs and often got frustrated by the result. The challenge with reverbs is a high echo density per second combined with a high mode density per Hertz. These two requirements seem to be in conflict. The second requirement has very little attention in the literature, although it is more important for the final outcome. Ever thought about why many reverbs sound metallic?
While it is easy to create many echos per second, the frequency response is usually much harder to tweak. The FDN architecture is just one example. Here, the recommendation is to modulate the delays to get rid of sharp resonances. Hm.
It is striking how embarassingly simple and low CPU a white noise generator is in comparison. So why not use convolution of the dry signal with white noise?
Thre are some nitty gritty details like how do you convolve two infinite streams (the signal and the continuous white noise)? You can't. The closest is to convolve the (infinite) signal with a block of white noise and renew the block when done. To suppress the block transitions one can do some staggered crossfading.
So we have a "gated" reverb of a fixed length - the block size. For an exponentially decaying reverb tail we repeat this block by a delay with feedback. It is important to place the delay before the convolver. (Why? )
Enough said, now check it out.
my newest creation is this reverb device. Don't judge this book by its cover - I am not exactly known as a GUI artist.
Under the hood is a rather new reverb design: a single delay with feedback followed by a convolver. I highly recommend this device for churchy organ sounds. I have spent many hours experimenting with reverbs and often got frustrated by the result. The challenge with reverbs is a high echo density per second combined with a high mode density per Hertz. These two requirements seem to be in conflict. The second requirement has very little attention in the literature, although it is more important for the final outcome. Ever thought about why many reverbs sound metallic?
While it is easy to create many echos per second, the frequency response is usually much harder to tweak. The FDN architecture is just one example. Here, the recommendation is to modulate the delays to get rid of sharp resonances. Hm.
It is striking how embarassingly simple and low CPU a white noise generator is in comparison. So why not use convolution of the dry signal with white noise?
Thre are some nitty gritty details like how do you convolve two infinite streams (the signal and the continuous white noise)? You can't. The closest is to convolve the (infinite) signal with a block of white noise and renew the block when done. To suppress the block transitions one can do some staggered crossfading.
So we have a "gated" reverb of a fixed length - the block size. For an exponentially decaying reverb tail we repeat this block by a delay with feedback. It is important to place the delay before the convolver. (Why?
)Enough said, now check it out.
Thank you