Hector Martin

Why is it that half the time I read a HN article about some cool research in a field I'm familiar with, I go "huh, that sounds impossible, I wonder how they're doing that"... then it turns out they aren't and the entire thing will never work in practical applications.

The latest one is a remote noise cancelation system where, after pages and pages about the approach, it ends with "oh yeah we're using the original noise signal, as played via speakers, as an input into the noise canceling algorithm, we know real ANC systems don't do that"

Why yes, when you collapse an entire sound field to 2-4 sources you have perfectly characterized, noise cancellation becomes very easy doesn't it?