Skip to main content
Topic: Tracks with a lot of high-frequency content are noticeably quieter at 96kbps (Read 983 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Tracks with a lot of high-frequency content are noticeably quieter at 96kbps

Using libopus 1.2.1.

Average bitrates:
  • Sample1: 87kbps
  • Sample2: 86kbps

 

Re: Tracks with a lot of high-frequency content are noticeably quieter at 96kbps

Reply #1
They sound about the same in Firefox.

Re: Tracks with a lot of high-frequency content are noticeably quieter at 96kbps

Reply #2
I'm not hearing any issue. Try decoding the Opus file to a 44.1 kHz wav like the original and play the two side-by-side. It could be your sound setup that's playing 48 kHz audio differently from 44.1 kHz (e.g. one of the two rates is using a bad resampler).

Re: Tracks with a lot of high-frequency content are noticeably quieter at 96kbps

Reply #3
After some testing, I found that only <112kbps Opus files in foobar2000 and MPC-HC have this problem. VLC and Google Chrome work fine.
If I decode the Opus file to a 44.1kHz or a 48kHz WAV, I don't have this problem in any program. It doesn't seem to affect the foobar2000 ABX comparator.

My motherboard is an ASRock G31M-VS2 (VIA VT1705 Audio Codec). I didn't notice this issue when I was doing ABX tests with my previous motherboard (ASRock G31M-S, Realtek ALC662) a few months ago, but it died so I can't check if my problem can be solved by replacing my motherboard.

Re: Tracks with a lot of high-frequency content are noticeably quieter at 96kbps

Reply #4
If I decode the Opus file to a 44.1kHz or a 48kHz WAV, I don't have this problem in any program. It doesn't seem to affect the foobar2000 ABX comparator.

Probably have a resampler or other DSP effect enabled for higher sampling rates that applies a gain reduction to avoid clipping during processing.

Re: Tracks with a lot of high-frequency content are noticeably quieter at 96kbps

Reply #5
The reason for the loudness differences is the Windows mixer's clipping prevention. The mixer automatically limits signals when they are near clipping and generally the lower the bitrate the more the waveform shape changes and peaks can increase from the original.

For example the tracks you provided are over 2 dB above digital fullscale. If you play the files at maximum volume (in a player that uses floating point decoding and output, like foobar2000), don't use ReplayGain and don't clip the peaks, the mixer will utilize its clipping protection and the tracks will be quieter.

This issue comes up every now and then. There are occasional topics about people claiming for example that foobar2000 sounds worse than WMP and I'm quite sure they run into this issue. Unfortunately they aren't allowed to be educated and the topics get removed.


Re: Tracks with a lot of high-frequency content are noticeably quieter at 96kbps

Reply #6
The reason for the loudness differences is the Windows mixer's clipping prevention.
Good reason to use WASAPI exclusive, BTW. Regardless many people here, on HA, keep saying "There is no difference". (and, yes, i prefer simple clipping over windows' limiter, because windows' limiter can result in "floating" loudness)

Re: Tracks with a lot of high-frequency content are noticeably quieter at 96kbps

Reply #7
I prefer to add Advanced limiter at the end of DSP chain.

(...actually, I also always use replaygain, and it's enough to prevent Windows limiter to interfere)

Re: Tracks with a lot of high-frequency content are noticeably quieter at 96kbps

Reply #8
So another possibility is just that the player decided to "normalize" the audio to avoid clipping, so it made the Opus version slightly quieter (which is sometimes perceived as less HF even though there's less of everything).

 
SimplePortal 1.0.0 RC1 © 2008-2018