Long story short, the band splitter adds a lot of complexity, which would take us a much longer time to reverse engineer than the waveshaping part. In addition to all that there are questions on where exactly the clippers are in the chain, how steep their cuts are, where the cuts are. It would also make the waveshaper respond differently once you go into the clipping regime. That would probably preserve more, but I kind of doubt that's what they're doing. We could omit the linear part (basically set curveA to 1) in the shaper and do that part on the input bus.
Since they wouldnt'be coherent with the non-bandsplit version, you'd get a lot of phase cancellation and you'd see notches in the response to white noise (which we do not see, right?). For curve > -50, it also changes the linear part of the slope meaning that you'd also get the fundamentals in the individual band signal. However, this story doesn't add up, because of the way the waveshaper is set up. This could explain why we see no latency or phase distortion in the low signal regime. Then, once they got a distortion contribution per band, they might sum it to the original unphased signal. Signal -> bandsplitter -> band signals (b0, b1, b2 etc) -> waveshaper on each of them -> subtract input to waveshaper for each band (so now you only have the harmonics/distortion products). So one thing I was suspecting then is that they might do something like this: It doesn't seem like they split the signal, then waveshape, then sum again, because afaik that would produce either latency (which we don't see) or phase distortion (which we also don't see). I don't really know what the plugin is doing. However, we could not detect any phase distortion either, which again makes minimum phase filter very unlikely. So a linear phase filter is probably out of the question. What exactly does Splitband do? The manual states that the signal is divided into 3 bands, but the plugin has 0 spls PDC. Out = sign(in) * s0 * wet + min(max(in * dry, -2 * dry), 2 * dry)
0 kommentar(er)
