top of page

Reverb

Reverb is a mixing technique that alters clean studio mixes to sound as if they were performed in a particular setting, such as a concert hall, theater, or outdoor setting. While reverb can be applied algorithmically through tuning many knobs to create a theoretical room, the approach we have taken uses the impulse response of the actual space to convolutionally apply reverb.

Screenshot 2022-04-27 213303.png

Convolution

The key property of signals, systems, and convolution that allows us to apply reverb is that if we have the impulse response of a system, we can obtain the output for any input of the system through convolution of the former with the impulse response. The physical room acts as a "system", and the input is any sound put into it, with the response being what is heard afterward. By inputting a short and loud sound, like a clap, into a room, we can then convolve that response with any signal to simulate the room, even when we no longer have access to it.

Approach

The approach to reverb was rather simple - an impulse response library was found [2], for each audio clip the impulse was filtered out, leaving only the response, and the clip was then convolved with the desired track to emulate how it would sound in that space. With some samples like the concert hall (left) the reverb was applied cleanly, however others, such as the cathedral (right), introduced much distortion into the track. This seems to depend on how easy the impulse is to filter out, as a harsher and longer impulse seems to cause more distortion.  

Original Track
00:00 / 05:50

[3]

Concert Hall Impulse Response
00:00 / 00:01
Cathedral Impulse Response
00:00 / 00:02
Concert Hall Reverb
00:00 / 05:51
Cathedral Reverb
00:00 / 05:51

[1] Image courtesy of University of Michigan EECS 351 Fall 2020 Lecture Notes

CONCLUSIONS

The reverb application works in general, but seems to be dependent on the quality of the presented impulse response. It seems possible that through more modifications to that response, such as more precise attenuation, filtering, or perhaps some type of learning algorithm that over time could tell the difference between the impulse and its response more precisely, this algorithm will work more consistently.

Citations

1. Image Courtesy of https://umich.instructure.com/courses/488640/files/folder/F20%20Lecture%20Notes?preview=24760492

​

2. EchoThief Impulse Response Library http://www.echothief.com/

​

3. Mussorgsky, Modest. "Pictures At An Exhibition: X. The Great Gate of Kiev." https://www.youtube.com/watch?v=Itdm46ZVLfA

​

Images not cited are from the Wix template images

bottom of page