If you'd send a never-compressed or lossless-compressed version (WAV or FLAC, probably, before Zoom sees it), then it'd be easier to verify the periodic-silence hypothesis.Īudacity is free, and it records as well as editing, analyzing, and playing back. It *would* produce a human-perceptible frequency, to have short-periodic silence like that, and so the MP3 compression algorithm would have kept that frequency while throwing away something else, and then the reconstruction on my end creates an entirely new waveform just from the frequency information that *is* still there. But since I'm looking at a lossy-compressed file that only saves the human-perceptible frequency content and makes no effort to preserve the original waveform, I can't tell if it originally had a few samples of silence every 5ms or not. Zooming in on the waveform, there does appear to be a "bump" every 5ms or so, which would be kind of a small buffer, even for internal transport, but entirely possible.
(USB devices may or may not count as different machines, depending on how each of them works - the USB spec provides a way to synchronize the clock, but it's more complicated than just shoving a bunch of samples into a buffer and sending it up the wire, which is also a USB-acceptable way to do it) More accurately, match the clocks, as different gear with the same setting can still drift out of sync with each other, but if you're doing it all on the same machine, you're probably okay there. Of course, the correct way to fix it regardless, is to match the sample rates all the way through the system. Resampling done right is completely transparent. If that's the case, then I'm surprised that the "glue" cable doesn't resample. But for live audio, it speeds up each buffer individually and puts silence in between. If you were playing a recording, that would appear as ~8% fast, with a corresponding pitch shift. Like the destination expects 48kHz but the source is 44.1kHz, and it just blindly plays out each small buffer as it becomes available. My second thought is a mismatched sample rate, without resampling. My first thought on that is a noise reducer that isn't set right.
0 kommentar(er)
