> There is one more refinement you could use to limit the amount of > CPU use. Of course the IRs for different source directions will be > different. But the *signficant* differences are only in the first 100 > ms or so. In many cases the 'tail' of the reverb will sound just the > same even if the IR is 100% decorrelated with one for another direction. > That means you need to do the direction-dependent convolution only for > the initial 100 ms or of the IR. For the rest you can share a single IR > for all sources.
Thanks Fons that makes a lot of sense. When preparing the tail-IR, is this matter of taking the file into an editor and silencing the first 100msec of the channels, perhaps using a ramp function to slope up to the 100msec point? Also, for the early-reflection-IRs, deleting everything after 100mcsec? > You could also experiment with changing > the relative delay of the direct sound and the reverb sends (no standard > mixer will allow you to do that easily, but if you use Pd or a similar > tool it is entirely possible. > Do you mean some sort of variable delay? Not sure what you mean. > Since you mention jconvolver, there are two preset files in the source > distribution that are designed to do this: sala-concerti-cdm and > santa-elisabetta. Both use ambisonic IRs which you can download from > my website. For the concert hall there are IRs corresponding to various > on-stage source position, and for santa-elisabetta (a small church) there > are 8 early reflection IRs spaced 45 degrees apart. Thanks! Iain _______________________________________________ Sursound mailing list Sursound@music.vt.edu https://mail.music.vt.edu/mailman/listinfo/sursound
