- Upsampling DOES introduce latency (we also have zero-latency mode, but that's not linearphase).
- Upsampling the whole thing is most likely better idea - the problem is with filtering. Upsampling is basically a very very steep accurate filter, which requires access to many samples in a row, so processing just a few samples wouldn't really give you any improvement in CPU performance, and would lead to unstable results in between of critical areas.
- My test was with 4x MDynamicsLimiter. Anyway as I said, Maxwell ended up with lots of intersample peaks, I don't blame it, it's impossible to remove it completely without analytic approach (which would suffer from other consequences).
- Q1 - How can be a difference between R128Gain and MLoudnessAnalyzer? Because none of these plugins actually measure true true peaks (2x true is intentional
). Please check my original long post .- Q2 - Again check my post. Anyway you upsample audio, perform limiting, downsample it back - the downsampling could raise new peaks (not only true ones), so another clipping section is necessary "just in case". And there you create another set of potential true peaks... If you do that again, same situation
...- Q3 - CPU consumption would be for a very long conversation. But the latency is basically cause by the filter. Usually the better the filter, the more latency (and CPU) it requires...
)
)
