4 ms·

Since we're at the PSAs, here is one about Opus: The codec is fixed at a 48kHz sample rate. For your audiobooks that were already low quality it probably doesn'

Since we're at the PSAs, here is one about Opus: The codec is fixed at a 48kHz sample rate. For your audiobooks that were already low quality it probably doesn't make a difference that there's now one more processing step applied to the audio, but I would never choose Opus as an archival format for music, as there is a lot of content using different sample rates (CDs in particular are 44.1kHz, so there's always a resampling step involved when converting them to Opus). Storage is cheap these days, so for serious archiving FLAC is the only logical choice (but obviously not when the original material was a 64kbps MP3 of course).

iggldiggl 5y ago
In theory, the resampling shouldn't matter, though, should it? As per Nyquist, as long as the sampling rate is higher than double the highest frequency you care about, the original signal can always be perfectly reconstructed. Therefore going to a higher sample rate shouldn't affect the originally recorded content. While the "perfect" reconstruction per Nyquist does indeed assume things that aren't actually possible in reality, like infinitely steep filters, AFAIK the best approximations we have are still good enough to not cause any audible issues (all the more so if you're subsequently applying lossy compression anyway), though of course it might be possible that a particular encoder uses a faulty implementation that does indeed mess things up.
- TonyTrapp 5y ago
  Yes, perfect filters don't exist, that's one issue. The other issue is that using Opus now forces you to use that one filter the encoder chose to use. As weird as it may seem, there are various reasons for preferring a lower-quality filters if your source input signal has a low sample rate to begin with (the result may sound fresher due to non-perfect filtering, which can be preferable in some situations). With Opus, the decoder would first have to downsample the 48kHz to the original sample rate, only to upsample them again with a lower-quality filter to achieve the same thing. I think avoiding all those extra steps has its advantage.

iggldiggl 5y ago

In theory, the resampling shouldn't matter, though, should it? As per Nyquist, as long as the sampling rate is higher than double the highest frequency you care about, the original signal can always be perfectly reconstructed. Therefore going to a higher sample rate shouldn't affect the originally recorded content. While the "perfect" reconstruction per Nyquist does indeed assume things that aren't actually possible in reality, like infinitely steep filters, AFAIK the best approximations we have are still good enough to not cause any audible issues (all the more so if you're subsequently applying lossy compression anyway), though of course it might be possible that a particular encoder uses a faulty implementation that does indeed mess things up.

TonyTrapp 5y ago
Yes, perfect filters don't exist, that's one issue. The other issue is that using Opus now forces you to use that one filter the encoder chose to use. As weird as it may seem, there are various reasons for preferring a lower-quality filters if your source input signal has a low sample rate to begin with (the result may sound fresher due to non-perfect filtering, which can be preferable in some situations). With Opus, the decoder would first have to downsample the 48kHz to the original sample rate, only to upsample them again with a lower-quality filter to achieve the same thing. I think avoiding all those extra steps has its advantage.