Taming Convolution Reverb #98
rerdavies
started this conversation in
Tips and Tricks
Replies: 0 comments
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Convolution reverb is expensive. It's easy to push a Raspberry Pi 4 to the limits with convolution reverb, or consume far too much of the total compute budget to be reasonably useful. For very long reverbs, you many need to increase the size of audio buffers to avoid under-runs.
There is, however, a way to substantially reduce the CPU use of convolution reverb. The Max T control affects how much of the reverb impulse file is used. If the impulse file is shorter than Max T, then the entire impulse response is used. If it is longer, the tail of the impulse file is discarded, and TooB Convolution reverb uses feedback to approximate the rest of the reverb tail.
Most of the character of a reverb impulse response is carried in the early part of the impulse file. Setting Max T to 1 second usually has little effect on the sound of the reverb, while substantially reducing CPU use. For most reverb impulse response files, you can set Max T to one second or even less -- perhaps 0.5 seconds -- without affecting the character of the reverb at all. Even more aggressive trimming of impulse response files usually produces nice reverb effects, even if it doesn't faithfully reflect the sound of the original.
As a point of reference, reverb impulse responses can range from about 0.5 seconds long for typical room reverbs, up to 6 seconds or longer for cavernous cathedral reverbs. The Jack Lyons Hall impulse response file (a cavernous reverb) is 4 seconds long. The St. Margaret's Church impulse file is -- a fairly long church reverb -- is 2 seconds long, with most of the impulse response in the first 0.6 seconds. The Arthur Sykes Rymer Auditorium impulse file (a tight room reverb) is 1.5 seconds long, with almost all of the impulse in the first 0.4 seconds.
Beta Was this translation helpful? Give feedback.
All reactions