IPB

Welcome Guest ( Log In | Register )

 
Reply to this topicStart new topic
foobar2000's ReplayGain misunderstood?
cowls192
post Aug 8 2012, 02:23
Post #1





Group: Members
Posts: 7
Joined: 8-August 12
Member No.: 102097



I've learned about a decade ago about the Normalize function, and i always thought there is no easy solution to "level" an album.

Pretty recently, i heard about ReplayGain, and it sounded like a good solution.

So i finally figured out EXACTLY how to use this ReplayGain function on foobar2000, and i'm starting to fuss about where this is going.

When i scan per-file, the results show Track Gain as
-6.46 dB for the quietest song i can find,
-11.91 dB for the loudest song i can find, and
-7.98 dB for the actual song i would like it to be louder.
In my uneducated perception, the loudest song had the biggest number, 11.91, but obviously, it is negative and i don't understand why.

After applying the RG value to the tag, i didn't think it got any louder, so i disregarded it.

Long story short, i duplicated that -7.98 dB song and deliberately lowered the volume by 50%, about -6.02 dB, using GoldWave. I applied album gain on these duplicate songs, and the result is that even though they sound equally loud, they were not as loud as the loudest song i found. And when i say "not as loud as", i mean the difference of mosquito flying noise and a shouting noise....

Maybe i didn't "specify" an option hidden somewhere in the option menu, and the default value happens to be matching the songs to the quietest song in an album. I don't know any better because i didn't design ReplayGain.

If anybody can give me a solid explanation about this RG issue, i will really appreciate it.

Bottom line is, how can i apply album RG so that all of the songs in an album are as loud as the loudest song in that album, or all the songs in an album are as loudest as possible in general, which was my initial intention?

And yes, when i say "as loudest as possible", i do not expect clipping, distortion, or any other audio nuisance. I just have a loud song that doesn't give me an earsore, and i just want to make few other songs to be loud like that. I hope nobody misunderstands my point.

This post has been edited by cowls192: Aug 8 2012, 02:24
Go to the top of the page
+Quote Post
pdq
post Aug 8 2012, 02:31
Post #2





Group: Members
Posts: 3407
Joined: 1-September 05
From: SE Pennsylvania
Member No.: 24233



For starters, album replaygain adjusts ALL songs on the album up or down equally, it does not make all tracks equal loudness. For that you want track gain.
Go to the top of the page
+Quote Post
saratoga
post Aug 8 2012, 02:32
Post #3





Group: Members
Posts: 4971
Joined: 2-September 02
Member No.: 3264



QUOTE (cowls192 @ Aug 7 2012, 21:23) *
When i scan per-file, the results show Track Gain as
-6.46 dB for the quietest song i can find,
-11.91 dB for the loudest song i can find, and
-7.98 dB for the actual song i would like it to be louder.
In my uneducated perception, the loudest song had the biggest number, 11.91, but obviously, it is negative and i don't understand why.


Its usually not safe to increase the volume of tracks (since most modern music is peak normalized already), so replaygain generally lowers volume to a common level.

QUOTE (cowls192 @ Aug 7 2012, 21:23) *
After applying the RG value to the tag, i didn't think it got any louder, so i disregarded it.


Well no, all of those tags are negative values, hence they make the file quieter. Unless you have a track that is normalized to a very low peak level, this will almost always be the case.

QUOTE (cowls192 @ Aug 7 2012, 21:23) *
Long story short, i duplicated that -7.98 dB song and deliberately lowered the volume by 50%, about -6.02 dB, using GoldWave. I applied album gain on these duplicate songs, and the result is that even though they sound equally loud, they were not as loud as the loudest song i found. And when i say "not as loud as", i mean the difference of mosquito flying noise and a shouting noise....


I don't understand what you did and what you expected to happen. Could you explain in more detail? Lowering the volume by 6 dB should result in a replaygain value of -2dB and no net change in the loudness.

QUOTE (cowls192 @ Aug 7 2012, 21:23) *
Bottom line is, how can i apply album RG so that all of the songs in an album are as loud as the loudest song in that album, or all the songs in an album are as loudest as possible in general, which was my initial intention?

And yes, when i say "as loudest as possible", i do not expect clipping, distortion, or any other audio nuisance. I just have a loud song that doesn't give me an earsore, and i just want to make few other songs to be loud like that. I hope nobody misunderstands my point.


Generally speaking, just play the songs. Most modern music is mastered to be a loud as possible without distortion. No further processing is necessary or possible.

If thats still not loud enough for you, you can look into something like dynamic range compression, which will reduce dynamics in order to make a track sound louder. That said, a lot of people tend to hate this since it will change how the track sounds.

Go to the top of the page
+Quote Post
cowls192
post Aug 8 2012, 03:11
Post #4





Group: Members
Posts: 7
Joined: 8-August 12
Member No.: 102097



QUOTE
Well no, all of those tags are negative values, hence they make the file quieter. Unless you have a track that is normalized to a very low peak level, this will almost always be the case.


Well, this was one of the experiment i did. Since i lowered the vol to 50%, the peak level of altered song should have positive gain using Track RG, in my imagination, but it didn't.

QUOTE
Long story short, i duplicated that -7.98 dB song and deliberately lowered the volume by 50%, about -6.02 dB, using GoldWave. I applied album gain on these duplicate songs, and the result is that even though they sound equally loud, they were not as loud as the loudest song i found. And when i say "not as loud as", i mean the difference of mosquito flying noise and a shouting noise....


QUOTE
I don't understand what you did and what you expected to happen. Could you explain in more detail? Lowering the volume by 6 dB should result in a replaygain value of -2dB and no net change in the loudness.


Oh yeah, this one, i duplicated the -7.98 dB song by lowering 50% vol, selected those two songs (which are really same songs, just dif vol) and applied album gain. As you explained, when i played them, i guess they were "proportionately" adjusted, in terms of volume. Then, in the Preference -> Playback -> ReplayGain, in the Source mode, i chose Track. That's when these duplicate songs were equally loud.

QUOTE
Lowering the volume by 6 dB should result in a replaygain value of -2dB and no net change in the loudness.


Sorry, it's not easy uploading pictures in the posts.

Album gain / Album peak for these duplicate tracks were -5.99 dB / 1.03 respectively,
while Track gain / Track peak were independent,
-1.96 dB / 0.52 for the duplicate,
-7.98 dB / 1.03 for the original,
which is kinda accurate, based on -6.02 dB vol change.

Hmm, i think when you said -2 dB, i assume it's the dif between Album gain's value and original's -7.98 dB.

QUOTE
For starters, album replaygain adjusts ALL songs on the album up or down equally, it does not make all tracks equal loudness. For that you want track gain.


I noticed that when i Track RG on all of the songs involved in this post, they were very similarly loud. This brings me to a new issue on how to tweak this Track RG values higher so that RG affected songs become louder...any ideas?

This post has been edited by cowls192: Aug 8 2012, 03:16
Go to the top of the page
+Quote Post
saratoga
post Aug 8 2012, 03:19
Post #5





Group: Members
Posts: 4971
Joined: 2-September 02
Member No.: 3264



QUOTE (cowls192 @ Aug 7 2012, 22:11) *
Hmm, i think when you said -2 dB, i assume it's the dif between Album gain's value and original's -7.98 dB.


Ok sounds like you got that figured out.

QUOTE (cowls192 @ Aug 7 2012, 22:11) *
I noticed that when i Track RG on all of the songs involved in this post, they were very similarly loud. This brings me to a new issue on how to tweak this Track RG values higher so that RG affected songs become louder...any ideas?


You can change the preamp value in foobar to target a higher level, however, if you do not want distortion/clipping, your tracks will virtually always be quieter.
Go to the top of the page
+Quote Post
cowls192
post Aug 8 2012, 03:29
Post #6





Group: Members
Posts: 7
Joined: 8-August 12
Member No.: 102097



Thanks for all the help and tips. smile.gif
Go to the top of the page
+Quote Post
Kohlrabi
post Aug 8 2012, 07:26
Post #7





Group: Super Moderator
Posts: 1050
Joined: 12-March 05
From: Kiel, Germany
Member No.: 20561



QUOTE (saratoga @ Aug 8 2012, 03:32) *
Most modern music is mastered to be a loud as possible without distortion.
There, fixed that for you.

This post has been edited by Kohlrabi: Aug 8 2012, 08:01


--------------------
Ceterum censeo Masterdiskem esse delendam.
Go to the top of the page
+Quote Post
mjb2006
post Aug 8 2012, 08:05
Post #8





Group: Members
Posts: 811
Joined: 12-May 06
From: Colorado, USA
Member No.: 30694



I'm not sure the original misunderstandings were addressed.

QUOTE (cowls192 @ Aug 7 2012, 19:23) *
When i scan per-file, the results show Track Gain as
-6.46 dB for the quietest song i can find,
-11.91 dB for the loudest song i can find, and
-7.98 dB for the actual song i would like it to be louder.
In my uneducated perception, the loudest song had the biggest number, 11.91, but obviously, it is negative and i don't understand why.

The ReplayGain (RG) value you get when scanning is a measure of what change is needed to bring the track or album to an average loudness of 89 dB. This relatively quiet target level is the loudness of a reference pink noise signal, and is about what you will find many pop CDs were mastered at in the 1980s and early 1990s. Those CDs will have RG values close to zero, i.e. their volume won't be reduced when playing with default RG processing enabled. Newer CDs tend to be mastered louder, to the point where any increase in volume whatsoever will add clipping and distortion. Tracks from these CDs will have negative RG values like you are seeing; they'll be played with the volume reduced by default.

You can subtract the RG value from 89 to get the song's "actual" loudness.
89 − -6.46 = 95.46 dB
89 − -11.91 = 100.91 dB
As you can see, your quietest song is still about twice as loud as the target level, and your loudest song is nearly four times as loud as the target. So when you have foobar2000 play these files with RG processing enabled, it's going to cut the volume by 50%-75% in order to make them all be 89 dB.

QUOTE (cowls192 @ Aug 7 2012, 19:23) *
how can i apply album RG so that all of the songs in an album are as loud as the loudest song in that album, or all the songs in an album are as loudest as possible in general, which was my initial intention?

That shouldn't be your intention. ReplayGain is making all songs or albums be the same loudness. There's no such thing as "as loud as the loudest song" then. If you're using track gain, every song is played at 89 dB; no song is any louder than any other. If you're using album gain, tracks will be played at 89 dB plus or minus a few dB, depending on how much louder or quieter each track is relative to the other tracks on the album. By definition, album gain is not going to make all tracks as loud as possible; the quieter tracks are going to remain that much quieter than the louder tracks, and they'll average out to 89.

In other words, you really don't want to bring the sub-100.9 dB tracks up in volume; you want to bring the louder tracks down. Otherwise you are going to get distortion and clipping, because there's likely no room for boosting the level of any of your tracks.

If you are also playing tracks which aren't RG-tagged, they'll be playing quite loud in comparison to the RG-tagged ones. To mitigate that, you can set the "Without RG info" Preamp level to, say, -11.9. This will make the player pretend they have -11.9 dB album gain. Thus if you played a non-RG-tagged copy of that loudest track, it would be played at 89 dB instead of its natural 100.91 dB, and would thus match the level that all the RG-tagged tracks are played at. However, then the quieter non-RG-tagged tracks would still be that much quieter. So you may find -8 or so to be a better "without RG info" preamp level, on average.

Personally, I think you should learn to love album gain. See how you feel about these settings:

Source mode: album
Processing: apply gain and prevent clipping according to peak
Preamp with RG info: 0 dB
Preamp without RG info: -8 dB

Don't do anything to your files in audio editors, just scan and tag albums with RG info, and let the player handle the volume adjustment. If you want it louder, turn up your amp.
Go to the top of the page
+Quote Post
cowls192
post Aug 8 2012, 16:09
Post #9





Group: Members
Posts: 7
Joined: 8-August 12
Member No.: 102097



QUOTE (mjb2006 @ Aug 8 2012, 03:05) *
The ReplayGain (RG) value you get when scanning is a measure of what change is needed to bring the track or album to an average loudness of 89 dB.

This helps me understand that ReplayGain is adjusting volume to a Target Level. The issue was that

QUOTE
i duplicated that -7.98 dB song and deliberately lowered the volume by 50%, about -6.02 dB,

And after i Track gain the 50% vol song, which would have been leveled to 89 dB as you said, it was not necessarily as loud as the original song. For the original song,

QUOTE
Most modern music is mastered to be a loud as possible without distortion.

so this proves that the Target level of 89 dB is not necessarily the loudest level without clipping, distortion, etc.

In the end, i was looking for a combination of ReplayGain and Normalize. This would be helpful when i process audio files that were not normalized to the maximum level, adjust them to the Target level so they are equally loud, then finally raise the volume as much as possible without clipping, distortion, etc. I was thinking that there may be an RG option where it will automatically find the highest Target level without clipping, distortion, etc.
Go to the top of the page
+Quote Post
sld
post Aug 8 2012, 16:53
Post #10





Group: Members
Posts: 1017
Joined: 4-March 03
From: Singapore
Member No.: 5312



QUOTE (cowls192 @ Aug 8 2012, 23:09) *
In the end, i was looking for a combination of ReplayGain and Normalize. This would be helpful when i process audio files that were not normalized to the maximum level, adjust them to the Target level so they are equally loud, then finally raise the volume as much as possible without clipping, distortion, etc. I was thinking that there may be an RG option where it will automatically find the highest Target level without clipping, distortion, etc.

Solution:

Source Mode: Track
Processing: Apply gain and prevent clipping according to peak

Preamp - With RG Info: anywhere between +5 to +10 dB (89+5=94, 89+10=99)


This will probably cause your classical-instrumental albums to have a reduced dynamic range during playback, but

1) I presume most of your albums are pop/rock/techno.
2) You may not be able to hear any on-the-go dynamic range compression if it happens.


I do this because I see the point of Replaygain as bringing all or most songs to the same perceived volume via complicated but sound (pun intended) psychoacoustic-mathematical principles. But I control Replaygain... it controls the volume of the music, it doesn't control my music habits.

This post has been edited by sld: Aug 8 2012, 16:57
Go to the top of the page
+Quote Post
greynol
post Aug 8 2012, 18:21
Post #11





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



QUOTE (cowls192 @ Aug 8 2012, 08:09) *
In the end, i was looking for a combination of ReplayGain and Normalize. This would be helpful when i process audio files that were not normalized to the maximum level, adjust them to the Target level so they are equally loud, then finally raise the volume as much as possible without clipping, distortion, etc. I was thinking that there may be an RG option where it will automatically find the highest Target level without clipping, distortion, etc.

The peak value of the track with the smallest correction that is dynamic is going to tell you how loud you can go without clipping.

You didn't make it clear why you want everything as loud as possible, but between this and adjusting your amp to your desired listening level or applying dynamic range compression to the quieter tracks (both have already been stated), you will achieve your goal.

The apply gain but prevent clipping option pretty well undermines the whole point of replaygain. Applying some arbitrarily chosen level of preamp gain is also not an optimal solution when the peak and correction information are at your disposal.

This post has been edited by greynol: Aug 8 2012, 18:30


--------------------
I should publish a list of forum idiots.
Go to the top of the page
+Quote Post
cowls192
post Aug 8 2012, 22:18
Post #12





Group: Members
Posts: 7
Joined: 8-August 12
Member No.: 102097



QUOTE (greynol @ Aug 8 2012, 12:21) *
You didn't make it clear why you want everything as loud as possible,

Oh no, i don't want to just make everything as loudest as possible. Even if i wanted to, i would just batch-normalize audio files.

My key point was to make audio files similarly loud first, and then raise vol for those audio files as much as possible w/o clipping, distortion, etc. That's because, like i mentioned before, sometimes songs or albums are processed poorly (often by consumers, not music producers) and it's not easy to guess from the head on how much vol change i have to make for each of those music files to sound equally loud. Applying RG makes things easy by setting a Target level that's same for all the files.

I don't know much about dynamic range compression, FYI.

I also discourage myself from trying pre-amp, because it sounds much like another hit-or-miss on clipping, distortion, etc.
Go to the top of the page
+Quote Post
saratoga
post Aug 8 2012, 22:26
Post #13





Group: Members
Posts: 4971
Joined: 2-September 02
Member No.: 3264



QUOTE (cowls192 @ Aug 8 2012, 17:18) *
QUOTE (greynol @ Aug 8 2012, 12:21) *
You didn't make it clear why you want everything as loud as possible,

Oh no, i don't want to just make everything as loudest as possible. Even if i wanted to, i would just batch-normalize audio files.

My key point was to make audio files similarly loud first, and then raise vol for those audio files as much as possible w/o clipping, distortion, etc.


That is exactly what batch normalization does. If you just want to be as loud as possible without clipping, that is all you need.
Go to the top of the page
+Quote Post
mjb2006
post Aug 8 2012, 22:35
Post #14





Group: Members
Posts: 811
Joined: 12-May 06
From: Colorado, USA
Member No.: 30694



QUOTE
Most modern music is mastered to be a loud as possible without distortion. so this proves that the Target level of 89 dB is not necessarily the loudest level without clipping, distortion, etc.

Err, well, not exactly. What they do in modern mastering is apply dynamic range compression/limiting at the same time as increasing the volume, so that what would be clipped is instead rolled off more gently, which doesn't mean the damage is necessarily less audible than clipping, just less "harsh".

It's not clear when you say "clipping, distortion, etc." if you are actually OK with the sound being modified / made less dynamic. If you can't abide any clipping or rolling-off of peaks, then you are constrained by the existing peak, which I'm confident is going to be right near maximum in all your modern files, and not much lower in your older ones. But if you are OK with some clipping, or with applying dynamic range compression when you make the quieter tracks louder, then yes, you have more room to boost the volume.

QUOTE
I was thinking that there may be an RG option where it will automatically find the highest Target level without clipping, distortion, etc.

"...and prevent clipping according to peak" means that if the volume is being boosted, it won't be boosted any more than the peak allows, just like "100%" normalizing, but it's not something you want to have kicking in all the time. Most all your music has peaks quite close to maximum already, with the modern recordings having those peaks much more frequently than the older ones, so the modern recordings are at greater risk of suffering audible damage if you don't do some kind of clipping prevention. If you follow sld's advice, I would turn off the 'prevent clipping' and instead use the Advanced Limiter DSP to apply dynamic range compression..
Go to the top of the page
+Quote Post
greynol
post Aug 9 2012, 00:58
Post #15





Group: Super Moderator
Posts: 10000
Joined: 1-April 04
From: San Francisco
Member No.: 13167



Modern recordings are far less likely to suffer additional clipping at the hands of the RG/Preamp/don't clip options than older dynamic tracks because the are still going to be attenuated under sane settings, whereas the dynamic tracks that still have peaks at or near full-scale are surely going to be amplified.

We already know that many jazz and classical recordings clip when the 89dB reference is used. Guess what happens when one of these tracks is played back with clipping prevention. Answer: it is played at a lower level and will be quieter than other material (so much for equal loudness!). Guess what happens when one of these tracks is played back with preamplification and clipping prevention. Answer: with a dynamic track preamplification does nothing unless it is used to attenuate that track below what would have been applied with just clipping prevention.

Again, if you want equal loudness and clipping prevention you really have just two options at your disposal if you have both heavily compressed and dynamic material. You can use either or both.
  • Play the tracks with RG and choose a level of preamplification so that no clipping occurs based on the needs of the most dynamic material
  • Employ dynamic range compression
One really should consider whether the second of these is really necessary. IMO, hardware and listening environment are good reasons; whereas fear that compressed tracks no longer use all the headroom afforded is not a good reason. To me personal taste is an adequate reason as well but with this is the expectation that the individual applying DRC understands what it is doing.

This post has been edited by greynol: Aug 9 2012, 02:15


--------------------
I should publish a list of forum idiots.
Go to the top of the page
+Quote Post

Reply to this topicStart new topic
1 User(s) are reading this topic (1 Guests and 0 Anonymous Users)
0 Members:

 



RSS Lo-Fi Version Time is now: 22nd September 2014 - 08:07