Hacker News new | ask | show | jobs
by scott01 437 days ago
There's a problem with f32 due to devices using two ADCs to implement it. Essentially, there's certain volume threshold when device switches DACs, and it's audible. If you record sources with varying loudness, you may see noise floor raising around loud bursts, and typical noise suppression tools won't work on files like this (there was a video or a blog post, I can't recall where I saw this). At least this is how Zoom implemented f32, don't know about Sound Devices.
1 comments

I was wondering how they avoid being swamped by analog noise in the lower ranges with such a high dynamic range.

I suppose they use different gains per DAC and map that to the respective dynamic ranges of the f32 signal?

To my understanding, yes, but I'm just a hobbyist.