Have I misunderstood volume levelling?

It works for me apart from the psychoacoustic effect described above. I rarely reach for the volume except if I really do want to turn the volume up or down because my mood changes or something. Otherwise, when it just plays along, the leveling seems to do what it needs to do for me

@Suedkiez - You are a lucky guy :slight_smile:

It would be nice if you could share your settings. THX


Nothing special, I use “Auto” and that’s it. I think that the only way to get to the bottom of this is for those who say it’s not working to set up a microphone and measure the loudest part of two tracks that come from different albums. Once with “track” leveling and once with “auto” leveling. If the leveling works as expected technically, then the loudest part of each track should measure the same on the mic with both of these settings.

@Suedkiez - May I ask you test this:

against any other album.


That’s a quite a difficult example because a) I have it only on Qobuz so I can’t see the waveform and b) there are huge differences already in the first track. The beginning has very low volume but at the end of the first track there is a much louder part. Comparing the first part of the first track to another album has very low volume in comparison but much less so at the end of the first track (although the other album is still clearly louder). Can you tell me where to look for the loudest part of the album?

I’m wondering if some of you don’t want volume leveling but compression :stuck_out_tongue_winking_eye:

I don’t know if the information is on the site


Can’t find loudness info on your links but I skipped through the album. I would say that its loudness is overall quite low and even when I enable track leveling and try to find the loudest parts, they are still somewhat quieter than other albums, subjectively.

But it’s also noteworthy that there are many very quiet parts within the album and some short much louder parts. If I enable track leveling within the album, the 3rd track gets a +7.2 db boost, but when I enable album leveling, the whole album overall only gets a +1.2 boost. (I use -14 LUFS). So that’s a very big difference, and one shouldn’t, e.g., compare the very quiet beginning of track 3 to another album that has a more even loudness throughout, because necessarily the difference will be big.

Also in track 3, it starts very quiet but there are some rather loud bass notes starting at 1:02. Short parts like this, and if it’s just one much louder note, will set the maximum level for the track or even the album, and if everything else is much quieter music, then volume leveling won’t undo that. And it shouldn’t, unless what you really want is compression :slight_smile:

Therefore, I still maintain that a discussion of whether Roon’s leveling works only makes sense if you find the one loudest moment of a track or an album (depending on the setting of volume leveling) and compare that objectively to the loudest moment of another track/album.

In addition, it would make sense to do this for local tracks that Roon analyzed itself. If it relies on the information from a streaming service, Roon’s leveling might work just fine but the information from the streaming service might be inaccurate. I compared the track 3 of your Alma album between Qobuz and Tidal, using Roon “track” leveling, and the Qobuz version gets boosted by +7.2 dB as mentioned above, while the Tidal version gets reduced by -0.4 dB. And I’d say that the track-leveled Qobuz version with the boost plays clearly louder than the track-leveled Tidal version. So I have to conclude that either the streaming info might not be entirely accurate or Roon might be doing something wrong, but which one it is is indistinguishable when using streaming tracks.

1 Like

On the contrary, according to EBU tech paper 3342 loudness range calculation explicitly excepts loudest and lowest portions to prevent skewing the result, see short excerpt …

LRA is defined as the difference between the estimates of the 10th and the 95th percentiles of the
distribution. The lower percentile of 10%, can, for example, prevent the fade-out of a music track
from dominating Loudness Range. The upper percentile of 95% ensures that a single unusually loud
sound, such as a gunshot in a movie, cannot by itself be responsible for a large Loudness Range.

But volume leveling values are derived from the integrated average loudness.
That’s, why a strongly compressed track and a very dynamic track, both replayed at levels so that their integrated average loudness is the same, will still be perceived as being differently loud.
There’s nothing one could do about it, other than introducing dynamic range compression/expansion techniques.

1 Like

Ah OK thanks for the correction, got that wrong. So then one might either record the whole track and maybe Audacity has some ways to do calculations or integrals (I dunno) or find some not-quite-the-loudest parts and compare those for an approximation.

And that’s certainly how it is

… build something like this.

The outcome will still be as I described in my last post:
You’ll likely measure the same integrated average loudness, but subjectively perceive one being louder as the other, because it’s dynamic range had been compressed too much due to the loudness wars…

1 Like

I turn off volume leveling after some experimentation for a while, I’ve personally found it detrimental to the music, making it sound dull and flat in comparison to Linn Kazoo/App. Maybe it is specific to my Linn network streamer.

Had a blip with volume levelling loudness data so perhaps worth checking this is pulling through correctly I.e. not defaulting to correction level when not known.
I use streaming services primarily and find I have no need to amend volume which I did before using.
Using Decibel X app to monitor from listening position extremes are there but averages are pretty consistent.
There are always going to be volume differences never mind between tracks but also within - that’s just music.
Perhaps you have very sensitive hearing which clearly will have pros and cons……

1 Like

One caveat (which I don’t think applies in @CrystalGipsy’s case) is that Roon’s analysis also keeps track of peak level, and will limit any increase in level to avoid clipping; ie will not exceed -0dBFS.

If you want to do your own file analysis, try out the Orban loudness meter…

1 Like

why exactly -23?

-23LUFS is the (proposed) EBU broadcast standard. There are caveats. The industry is not there yet, but it’s where it’s heading. An average perceived loudness 23dB below digital full scale should allow enough headroom for the most dynamic programme material, although I’m sure someone could contrive an exception. Actual playout is normally hard limited to -1dBFS.

1 Like

Wish they all stuck to it for video content as they don’t each service has their own constantly have ot adjust volume between Apple TV+, Netflix and Amazon.

1 Like

They likely don’t seem to see the need since they’re not in the E(uropean)B(roadcast)U(nion), claiming 68 members representing 112 organizations in 56 countries …

Absolutely, I’ve been there too. Volume leveling is meant to balance out volume differences between albums, but it might not completely eliminate disparities. LUFS adjustments affect overall loudness, not album balance. Roon does its best, but recording variations can be challenging. You’re not doing anything wrong. Fine-tune settings or manually adjust for specific albums if needed. It’s all about finding the best compromise for your collection.

I was sceptical at first to use volume leveling, but I gave it a try. Contrary to the OP, I am finding it very good, in the sense that I need to reach for adjusting the volume wayyyy less now. I never expected to be perfectly level matched, but it is great job done by Roon in my opinion. I am using default -14.

It’s an AES recommendation too. Netflix et al are likely to comply in time, but there are many reasons why it remains a work in progress.

How loud any individual track/work/piece should be to feel ‘right’ is a different question, and is maybe subjective. Do you want to feel that the orchestra is playing in your living room (I don’t…) or that you are in the concert hall? Acoustic guitar in your living room? Sat in the control room in front of the mixing desk? Standing in front of the bass bins at a rock concert? These are all ‘correct’ at different sound pressure levels, but what is encoded on your chosen medium is at the mercy of the mixing and mastering engineer. It is both science and art.