Serial Reverb Techniques

Locked New Topic
RELATED
PRODUCTS

Post

I'm really curious if anyone here feeds their reverbs into other reverbs on a regular basis?

Sean, I really enjoyed your blogs about pitch shifting and I know you are the reverb guru so I thought you might have some insight about interesting or unusual uses for this technique.

Post

xmodz wrote:I'm really curious if anyone here feeds their reverbs into other reverbs on a regular basis?

Sean, I really enjoyed your blogs about pitch shifting and I know you are the reverb guru so I thought you might have some insight about interesting or unusual uses for this technique.
First of all, sorry for the late reply! I just don't scroll down to my own forum very often. I should have named the company "Aardvark DSP" or something like that...

As far as cascading reverbs, I designed ValhallaShimmer to work well with this technique. In the stereo modes, each channel of Shimmer is essentially "colorless," ignoring any filtering and feedback. From the strict DSP perspective, each channel is allpass, in that the frequency response is 0dB for all frequencies. This allows for multiple instances of Shimmer to be cascaded in series, without creating metallic artifacts.

For most reverbs, the input/output path isn't allpass. ValhallaRoom, for example, takes its outputs from a variety of delay taps, depending on the reverb mode selected. The frequency response of the input/output path of ValhallaRoom is very complex, but it isn't allpass. Real rooms aren't allpass either - they have a ton of peaks and valleys in the frequency response. Lexicon reverbs are built around allpass delays, but they generally don't have an allpass response, due to using anywhere between 3 and 15 output taps per channel.

When you cascade reverbs that don't have an allpass response, the frequency responses will cancel out in weird ways, and usually the sound is VERY colored. If a peak in one response coincides with a valley in the other response, they will cancel each other, while matching peaks will result in very loud peaks. The result is that a lower number of peaks end up dominating the sound, creating a metallic effect.

The time-domain effects of cascading reverbs are determined by the laws of convolution, and tend to have somewhat predictable responses:

- The total reverb time will be the sum of the reverb time of the cascaded reverbs.
- The attack of the reverb tends to be smoothed out. By cascading 4 or more stages, the attack/decay tends to get closer to a Gaussian curve.

My recommendation is to use ValhallaShimmer as a building block for cascading reverbs. With the feedback turned down to 0, you can cascade as many stages as you like, without causing cancellations in the frequency response. Turning up the feedback won't cause undue coloration until you cascade a lot of stages. You can also run ValhallaShimmer before or after other reverbs, in order to get the cascaded sound with these reverbs without too much coloration.

Sean Costello

Post

Wow. Thanks for the response. That's alot to take in!

Hmmm... Do you know if any of the MidiVerb II algorithms lend themselves to cascading like that? How about EOS? I'm guessing no on EOS.

Locked

Return to “Valhalla DSP”