Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: 89 db replaygain too much for classical music (Read 2022 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

89 db replaygain too much for classical music

 VA Fauré: Requiem 2002

Album gain +3.70  , quietest +14.69


I was listening the other day and heard a horrid distortion with crackling vocals.
I tried turning on clipping prevention in poweramp and it seemed to work. 
Originally, I encoded to mp3 via foobar with RG album mode.  Today I compared two encodings,
one with clipping prevent & one without.  The clipping prevent version has album gain of +3.70 while the other has +0.50.
The +0.50 is similar to most 'normal' albums . So 89db is too much here and 85db is the safer setting.
83db was the original proposal and EBU recommended.  So I set foobar to encode RG album mode + clip prevent. Or set everything to 83db ??





Re: 89 db replaygain too much for classical music

Reply #1
With "prevent clipping" it's always "safe".

But a higher target isn't as effective at volume matching.  The target volume is a compromise that allows most popular music to be volume-matched without making everything too-quiet.  (A lot of people complain that everything is too-quiet when using ReplayGain.) 

A lower target volume gives it more room to work with.   With a lower target more tracks can hit the target because tracks can always be turned-down, but some can't be turned-up (without clipping). 

A lot of tracks are already 0dB normalized (including ""quiet sounding" tracks, or tracks with quiet parts) so they can't be boosted at all to it a higher target, and others can't be boosted enough to hit the target loudness.

...You may not even need ReplayGain with classical.   Classical tends to be very dynamic (a large range between the loud and quiet parts) and ReplayGain doesn't change that, but it makes volume matching between track less useful.  You can still have a track that ends with a loud crescendo, followed by one that starts-out quiet.    And you probably aren't playing one short "song" after another so it's not as important that the volumes match.   If you do use it, you probably want to use Album Gain so you retain the relative volume between the movements.

If you are inter-mixing classical and other kinds of music, then yes...  You don't want a loud pop-song to come-on after a classical piece so you might want to set a lower target. You wouldn't want it the other way around either with a "quiet-dynamic" classical piece followed by a "loud compressed" modern song.

Re: 89 db replaygain too much for classical music

Reply #2
There are always some oddballs out there. Correcting gain for peak to get "clipping" absent fb2k's clipping prevention, I have some highlights sure.  One is John Cage, a recording of https://en.wikipedia.org/wiki/27_minutes_10.554_seconds , 8.9 dB clipping: 6.7 dB gain over a true peak of +2.2. Track peak is louder than album peak too.
But it isn't just classical music. This ambient live set also clips 8.9 dB. And bootleg sound quality ... uh well, no control over the peaks, so this black metal boot clips 7.5 dB.

And here is some nice electric guitar jazz for you: https://www.youtube.com/watch?v=PnjDoKAyqBE .  Album would clip 5 dB, this track slightly more.

Re: 89 db replaygain too much for classical music

Reply #3
VA Fauré: Requiem 2002

Album gain +3.70  , quietest +14.69


I was listening the other day and heard a horrid distortion with crackling vocals.
I tried turning on clipping prevention in poweramp and it seemed to work. 
Originally, I encoded to mp3 via foobar with RG album mode.  Today I compared two encodings,
one with clipping prevent & one without.  The clipping prevent version has album gain of +3.70 while the other has +0.50.
The +0.50 is similar to most 'normal' albums . So 89db is too much here and 85db is the safer setting.
83db was the original proposal and EBU recommended.  So I set foobar to encode RG album mode + clip prevent. Or set everything to 83db ??

I personally set a target of 79dB when using ReplayGain.

Re: 89 db replaygain too much for classical music

Reply #4
Of course, those of us who use foobar2000 can use the "Preamp" slider for up to +/-20 dB, that is enough for most of us. Separate for "With RG info" and "Without RG info", so one can go from 0 & -8 to  -10 & -18 if that is desired.

(Also I recall a long discussion over whether the EBU gain algorithm actually improves over the original RG ...)

Re: 89 db replaygain too much for classical music

Reply #5
I don't often encode classical music, but I'm aware of this problem and have encountered it from time to time in rock/pop music (hardly ever with more-recently-produced-or-remastered albums, which tend to be more compressed).

My old solution, when I was in early 20s and single, was to manually select each offending peak (from where that the amplitude line crossed 0 dB, to where it crossed back again, so peak about in the middle of selection) and scale that whole selected area so the peak was compatible with my target scaling on encode. As one might imagine, this was quite time-intensive, and the sort of thing a programmed loop or some other algorithm could have handled much more efficiently.

The dramatically easier way I've found to do this is to encode to +1.5 or +3 or +4.5 (etc.) db above the target volume, and then use fb2k or mp3gain to lower the resulting file by that same amount. I'm doing this with FB2k, and this may engage some anti-clipping algorithm that I'm not aware of and for which I can't find a switch. I haven't tried it with commandline or LameDropXPd to see if baseline LAME encoder does the same thing.
God kills a kitten every time you encode with CBR 320

Re: 89 db replaygain too much for classical music

Reply #6
The dramatically easier way I've found to do this is to encode to +1.5 or +3 or +4.5 (etc.) db above the target volume, and then use fb2k or mp3gain to lower the resulting file by that same amount. I'm doing this with FB2k, and this may engage some anti-clipping algorithm that I'm not aware of and for which I can't find a switch. I haven't tried it with commandline or LameDropXPd to see if baseline LAME encoder does the same thing.

Here's an example of a song that I'd encoded years ago (original was with LAME 3.92) that I just left much quieter than I'd prefer it to be. Recently I scaled it up to volume that sounded right (it's one of these weirdly-mastered indie songs where RG really misdiagnoses the perceptual loudness) and used the method I described above to get the same RG volume, without any clipping. This peak-lowering only decreased the RG value by 0.12 dB.

Note: I'm unable to ABX-differentiate the no-clipping version from the WAV, even though I did a fair comparison with scaling them down to the volume where the WAV isn't clipping, and just turning up my speakers.

Attached are screenshots showing the RG and peak values, as well as wav exports of those two mp3s (the 100% one has clipped peaks that go up to 189%)
God kills a kitten every time you encode with CBR 320

Re: 89 db replaygain too much for classical music

Reply #7
I thought about several ways proposed here.  Using album gain + clip prevent doesn't appear to alter the recording so that appears to be the obvious way. Most other genres will be normalized as they should.  The other is adjust pre amp as needed and doesn't change the files.  Or not apply gain to classical at all - not sure about that yet.

Re: 89 db replaygain too much for classical music

Reply #8
I also tried ebu 128 normalizer . Its like track mode RG but no clipping since dynamics are smoothed out.
I think its not bad and even better than RG in some ways but obviously not 'true' like album mode rg.
Would this way be better ?

Re: 89 db replaygain too much for classical music

Reply #9
Or not apply gain to classical at all
^ This!

So far as I am concerned, RG is intended to even out the perceived overall volume of individual tracks, presuming each track is itself reasonably consistent in volume, to save the user repeatedly reaching for the volume control.  How is it supposed to assess an overall volume for a track (or an album) if it has a significant dynamic range (typical of classical pieces)?

If you are in a listening environment where quiet passages are inaudible when loud passages are not too loud (and vice versa), then what you need is dynamic range compression not RG.  I've done that for in-car listening (using Audacity).

It's very difficult getting the full dynamic effect of classical when there is background noise or if there are neighbours not to annoy.  Presuming that's not the case: the original recording isn't (or shouldn't be) digitally clipped, and applying digital gain won't magic any extra detail out of the least significant bits, so just set the volume on your amp!
It's your privilege to disagree, but that doesn't make you right and me wrong.


Re: 89 db replaygain too much for classical music

Reply #11
Never heard about the album gain and album peak scanning!?
Yes.  That changes nothing.  If there are significant (deliberate) differences between tracks in an album, the album is not dissimilar to a single track with significant dynamic range.
It's your privilege to disagree, but that doesn't make you right and me wrong.

Re: 89 db replaygain too much for classical music

Reply #12
... and then the peak value can prevent it from clipping.

Also, since most non-classical music requires attenuation, then I will set the player to attenuate files which do not have RG info.
Hence, deleting RG info from classical music will only make matters worse.

Re: 89 db replaygain too much for classical music

Reply #13
I have another problem with android and replaygain.  With 89gb my volume is around 80%  All good until
I switch sources to say another non-RG aware app - they often use close to full scale loudness so a few times already my
ears nearly exploded .  In that case the volume needs to be %20 but then poweramp is way too soft.  So I need a system wide
normalizer for android like ms-windows 'loudness EQ' or like many TV's and cars have .  Anyone know of one ?

Re: 89 db replaygain too much for classical music

Reply #14
I have another problem with android and replaygain.  With 89gb my volume is around 80%  All good until
I switch sources to say another non-RG aware app - they often use close to full scale loudness so a few times already my
ears nearly exploded .  In that case the volume needs to be %20 but then poweramp is way too soft.  So I need a system wide
normalizer for android like ms-windows 'loudness EQ' or like many TV's and cars have .  Anyone know of one ?

if you're using RG on lossy files (mp3 or aac for example), then in foobar > ReplayGain > Apply gain to file content... thus actually changing the files. Then you could remove RG tags, so they'd be consistent over different players.

Re: 89 db replaygain too much for classical music

Reply #15
I have another problem with android and replaygain.  With 89gb my volume is around 80%  All good until
I switch sources to say another non-RG aware app - they often use close to full scale loudness so a few times already my
ears nearly exploded .  In that case the volume needs to be %20 but then poweramp is way too soft.  So I need a system wide
normalizer for android like ms-windows 'loudness EQ' or like many TV's and cars have .  Anyone know of one ?


For Android I use just one app for all my recorded music (foobar2000 Mobile).  Android is kind of at where desktop Windows was 20+ years ago with it's media apps, iPhone is even worse in that regard as it's more lockdown than Android is.

Re: 89 db replaygain too much for classical music

Reply #16
I'm trying to understand this thread and replaygain better, but I haven't been able to find an answer after searching. I'm used to seeing volume in relation to full scale (0dBFS). When you say in the title 89dB, what is that down from/reduced by? Is it 7dB down from 96dB (16 bit)? 11dB down from 100dB?
Think millionaire, but with cannons.

Re: 89 db replaygain too much for classical music

Reply #17
I'm trying to understand this thread and replaygain better, but I haven't been able to find an answer after searching. I'm used to seeing volume in relation to full scale (0dBFS). When you say in the title 89dB, what is that down from/reduced by? Is it 7dB down from 96dB (16 bit)? 11dB down from 100dB?

The best way to understand ReplayGain is understanding what loudness itself actually is and how to more accurately measure it.  It's basically a way of analyzing a recorded audio signal based on how loud it sounds to human ears when played, not the relation of any peak or volume control setting and saving that result into a tag based on what the relation is to desired loudness level or the average is over time without altering the actual recorded audio signal stored in the file itself.

The ReplayGain target level of 89 dB correlates roughly to -18 LUFS which cannot be changed because it's part of the specification itself and that's what is stored in the file tags.  Changing the target level of the player however is possible by adjusting the preamp settings of any player that implements it, which foobar2000 in fact implements (Preferences > Playback > Replaygain Section of Playback Tab).  Changing the preamp, changes the target level in the player itself, giving the user some control based on personal preferences or needs.  For example: in another post in this thread i said I like to use a target level of 79 dB (or about -28 LUFS) which is basically the preamp set to -10 dB.

In ReplayGain terminology: The target level is basically where you want all audio to average at in terms of human perceived loudness measurements not the loudness of the signal itself.

Re: 89 db replaygain too much for classical music

Reply #18
Thanks for the explanation. Somehow I'm even more confused though.
I understand what ReplayGain does and I've been using it for almost two decades now. I'm just confused what the numbers are in relation to. What is -18 LUFS? What dB would 0 LUFS be? If I had preamp set to +/-0 with no replaygain, and a tracked peaked at 0dBFS, what dB would that be in comparison to the 89dB OP mentions as being too much?
Sorry, I don't want to derail the thread, but I'm having trouble understanding. Perhaps there is a resource where I can learn? I tried reading all of the info I could find on various wikis about ReplayGain, but I can't find an answer to this. Should I instead make a new thread?

*EDIT*
I will start a new thread instead. My apologies for the interruption.
Think millionaire, but with cannons.

Re: 89 db replaygain too much for classical music

Reply #19
I have another problem with android and replaygain.  With 89gb my volume is around 80%  All good until
I switch sources to say another non-RG aware app - they often use close to full scale loudness so a few times already my
ears nearly exploded .  In that case the volume needs to be %20 but then poweramp is way too soft.  So I need a system wide
normalizer for android like ms-windows 'loudness EQ' or like many TV's and cars have .  Anyone know of one ?


For Android I use just one app for all my recorded music (foobar2000 Mobile).  Android is kind of at where desktop Windows was 20+ years ago with it's media apps, iPhone is even worse in that regard as it's more lockdown than Android is.

My problem is internet radio streams from other apps or browsers.

Re: 89 db replaygain too much for classical music

Reply #20
My problem is internet radio streams from other apps or browsers.
You can simply try turning the volume of the phone all the way down after you're done listening to any local files and before listening to any internet radio streams  You can't change what volume level an internet radio station is broadcasting at but you can try using the preamp sliders for stuff without ReplayGain if your player supports it (i.e. foobar2000 mobile).  For mobile browsers like Chrome, you're basically out of luck in that regard and would be better off turning the volume on the phone down before opening an internet radio stream, which I what I recommend regardless of any app you're using to listen to internet radio streams.