PDA

View Full Version : Understanding Replay Gain



liffy99
2008-08-21, 01:29
Ok - now that I've recorded loads of music, I notice even more than before the huge volume differnces between albums. Looking at some tracks on Audacity I can see that some have been "maxed" to the nines during the production process - all high level with little dynamic range (e.g. Keb Mo's "Suitcase") whilst others are far less aggressively mastered.

When I tried using album gain with EAC, I noticed that levels always seem to be reduced by a few Db. How is this "gain" (shouldn't that be "negative gain". What is it using as a reference to reduce the gain ?

I'd ideally like to increase the volume level of some of my quieter albums but preserve the dynamic range - i.e. increase the min and max levels by the same amount.

Can I do this with files that are now stored in FLAC format ? What software would I need ? Any other tips ?

Cheers

Nonreality
2008-08-21, 01:55
Ok - now that I've recorded loads of music, I notice even more than before the huge volume differnces between albums. Looking at some tracks on Audacity I can see that some have been "maxed" to the nines during the production process - all high level with little dynamic range (e.g. Keb Mo's "Suitcase") whilst others are far less aggressively mastered.

When I tried using album gain with EAC, I noticed that levels always seem to be reduced by a few Db. How is this "gain" (shouldn't that be "negative gain". What is it using as a reference to reduce the gain ?

I'd ideally like to increase the volume level of some of my quieter albums but preserve the dynamic range - i.e. increase the min and max levels by the same amount.

Can I do this with files that are now stored in FLAC format ? What software would I need ? Any other tips ?

CheersI would say that most modern albums (guessing 95% and up) need a negative replay gain. That means that the volume needs to be lowered to match up with other albums and not get into overiding peak values and causing clipping. Clipping is bad and can cause problem the least of which is damage to your speakers. Most replay gain programs are set to around .89db so you normally have negative values in the replay gain. Using dbpoweramp or foobbar you can have it tag your files with album gain and track gain. Album gain will keep albums that you are playing to the same level while allowing songs to vary as they do on the original album. Track gain is useful for playlists to keep individual songs in the same volume range. It's the most important for mixes and playlists. You need to set volume control in the settings on the SB to smart gain and the SB will use track gain when needed and album gain when needed. It works rather well. Make sure what ever program you use does both album and track gain. Track gain probably being the most important if you do mixes. If you happen to have Itunes soundcheck tags then don't use replay gain as the Squeezecenter software doesn't work right with both tags. It adds them both together and causes too much volume reduction or gain if you have a postitive replaygain number. Most new music runs between -5db and -12db because of the volume wars. Everyone wants to be noticed and dynamic range be damned. Louder is better. I recommend dbpoweramp and let it do it's thing.

EDIT: Once you have replay gain working, just use your volume dial to adjust. The big round one usually. :)

liffy99
2008-08-21, 03:34
Thanks Nonreality - very clear explanation.

I've used Foobar 2000 to add replay gain (as single album option) to a Keb Mo recordimg - overall level reduction of 9.04 db (yikes). I'll listen later but assume that the level of this album will now be 9db below many others so I just need to wind the volume up when needed.

Of course this can't sort out the woeful mastering - how did this mess ever arise ?

I remember the good old days when the source was the source and we all spent many happy hours tweaking and changing equipment to get the best out of it. Now not only do the manufacturers keep changing the platforms but seems they also mess up the source itself.

aagaaghh

Rick B.
2008-08-21, 03:52
I remember the good old days when the source was the source and we all spent many happy hours tweaking and changing equipment to get the best out of it. Now not only do the manufacturers keep changing the platforms but seems they also mess up the source itself.

aagaaghh


Of couse, even in the good old days of vinyl there could be large variances in volume btween albums depending on how "hot" they were cut and on how long each side was. Put on Todd Rundgren's "A Wizard, A True Star" album that clocks in near 30 minutes a side and compare it to any album with 20 minutes a side and the difference in loudness is obvious. Longer albums had to have their level reduced so that the squiggles on the vinyl were narrower.

At least today we have tools like Replaygain to deal with these things in the digital world!

Skittler
2008-08-21, 04:28
I'm not necessarily a fan of Wikipedia, but this is worth a read:

http://en.wikipedia.org/wiki/Loudness_war

Mark Lanctot
2008-08-21, 06:23
I've used Foobar 2000 to add replay gain (as single album option) to a Keb Mo recordimg - overall level reduction of 9.04 db (yikes).

Try a Red Hot Chili Peppers album...-12 dB across the board.


I'll listen later but assume that the level of this album will now be 9db below many others so I just need to wind the volume up when needed.

Not if all your music has RG tags. They say how to adjust playback to achieve 89 dB average volume. So your Keb Mo recording will be played back at -9 dB, resulting in 89 dB average volume, and another album with say -6 dB will be played back at -6 dB, also resulting in 89 dB average volume, and say you have Red Hot Chili Peppers Californication (one of the worst examples of loudness wars) at -12 dB, well, that'll be played back at -12 dB, resulting in 89 dB average volume.

Of course, this is for whole albums - that the average volume of the whole album will be 89 dB. Individual tracks may be louder or quieter as the artist intended. It's much easier to understand this concept if you only consider track gain, say in a random playlist composed of songs from various artists out of many albums. Then each track is played back using an adjustment to bring it to roughly 89 dB, so there's no volume jump between tracks.


Of course this can't sort out the woeful mastering - how did this mess ever arise ?

Bad things tend to happen when record labels get involved. ;-) The Wikipedia article is a very good read.

What's most disturbing about loudness wars is the fact that the damage done to the recording is NOT corrected by ReplayGain. The loss of dynamic range is permanent as is digital clipping. You can turn it down, but if the samples are cut off, they won't come back.

SuperQ
2008-08-21, 07:48
Can I do this with files that are now stored in FLAC format ? What software would I need ? Any other tips ?

All you need is the metaflac tool that comes with the normal FLAC software. I wrote a simple shell script to go through all my album dirs and add the tags if none were found. If you're on windows, you will need a separate tool to do the processing and tagging because the windows CLI is crap for scripting.

liffy99
2008-08-21, 07:53
Hmmmm - very interesting.

One thing that still has me a little baffled is this concept of clipping. If I look at the waveform of a recording (say for that Red Hot Chilli peppers stuff) I assume that the peak levels will still remain at or below 0db. So it can't "clip" can it ?
The only clipping I could imagine is if you were simply turning the wick up too far and letting the amp run past its comfort zone.

So still not sure why record levels are being reduced, though I see the lure of trying to get albums to sound equally loud at their loudest points - why not raise the quieter ones to the same level as RHCP (any compression in the mastering will still be unalterable).

bobkoure
2008-08-21, 08:05
Anyone know for sure how to turn RG-adjustment off for sure when listening album-by-album?
I listen to MusicIP mixes a lot - and RG's a godsend for mixes. But, when I'm listening to an entire album, I'm oft-times listening critically and want any volume adjustment off.
It seems like this is what smart gain does, but maybe not.
What happens if there are no RG-album tags? Is RG-adjustment then based on the track tags? Not what I want. What if I just write 0 to the RG-album tags? The album adjustment isn't used at all in mixes, is it?
Thanks!

Mark Lanctot
2008-08-21, 08:07
One thing that still has me a little baffled is this concept of clipping. If I look at the waveform of a recording (say for that Red Hot Chilli peppers stuff) I assume that the peak levels will still remain at or below 0db. So it can't "clip" can it ?

Well it can go right up to 0 dBFS and if you take a look at the waveform, it flatlines there over a series of samples. That means there *were* samples higher than the flatline, and they're gone for good.

It's digital clipping. It's not the same as analog clipping but has that same pop/crackle sound. It's not as damaging to speakers but it's still damaging to the sound.

bobkoure
2008-08-21, 08:20
Maybe it's time for the RIAA to get back to its original charter and do for this what they did for vinyl.
Not that I'm holding my breath.

Meanwhile, it'd be great for places like allmusic to start adding the RG album peak/gain numbers for each album - along with a link to something that explains what it's all about.

Nonreality
2008-08-21, 12:51
Anyone know for sure how to turn RG-adjustment off for sure when listening album-by-album?
I listen to MusicIP mixes a lot - and RG's a godsend for mixes. But, when I'm listening to an entire album, I'm oft-times listening critically and want any volume adjustment off.
It seems like this is what smart gain does, but maybe not.
What happens if there are no RG-album tags? Is RG-adjustment then based on the track tags? Not what I want. What if I just write 0 to the RG-album tags? The album adjustment isn't used at all in mixes, is it?
Thanks!My understanding is that Smartgain will use album gain if x amount of tracks are from the same album. I don't know how many it takes to kick it in but if you play an album then it will use album gain. If they are from multiple albums then track gain is used. If there are no album gain tags I do believe that it will use track gain which isn't ideal. If you don't want any replay gain just shut it off. If you have both album and track replay gain tags you really don't need to worry about it. It works great.

Added: I think you can put a 0 in the album gain if you don't want Smartgain to use the track gain on albums. I've heard people say they do this in the forums. Just remember that if you play multiple albums you will have differences in volume.

schiegl
2008-08-21, 13:42
Anyone know for sure how to turn RG-adjustment off for sure when listening album-by-album?
I listen to MusicIP mixes a lot - and RG's a godsend for mixes. But, when I'm listening to an entire album, I'm oft-times listening critically and want any volume adjustment off.
It seems like this is what smart gain does, but maybe not.
What happens if there are no RG-album tags? Is RG-adjustment then based on the track tags? Not what I want. What if I just write 0 to the RG-album tags? The album adjustment isn't used at all in mixes, is it?
Thanks!

This piece of code from ReplayGain.pm is responsible for smartgain (and similar applies to smart-crossfade)



if (defined $album->replay_gain() && ($class->trackAlbumMatch($client, -1) || $class->trackAlbumMatch($client, 1))) {
return $album->replay_gain();
}
return $track->replay_gain();


1. Check if album-rg is set, if not use track-gain. Therefore always use album-rg if you intend to use smartgain. You should set it to +0.0db otherwise trackgain will be used, which is probably wrong.

2. Check if the previous OR the following title is from the same album. If it is, use album-rg otherwise use track-rg. It has nothing to do with how many tracks from the same album are in your playlist, just what comes next or what was before.

kind regards,
Markus

mark95841
2008-08-21, 14:42
I would say that most modern albums (guessing 95% and up) need a negative replay gain. That means that the volume needs to be lowered to match up with other albums and not get into overiding peak values and causing clipping. Clipping is bad and can cause problem the least of which is damage to your speakers. Most replay gain programs are set to around .89db so you normally have negative values in the replay gain. Using dbpoweramp or foobbar you can have it tag your files with album gain and track gain. Album gain will keep albums that you are playing to the same level while allowing songs to vary as they do on the original album. Track gain is useful for playlists to keep individual songs in the same volume range. It's the most important for mixes and playlists. You need to set volume control in the settings on the SB to smart gain and the SB will use track gain when needed and album gain when needed. It works rather well. Make sure what ever program you use does both album and track gain. Track gain probably being the most important if you do mixes. If you happen to have Itunes soundcheck tags then don't use replay gain as the Squeezecenter software doesn't work right with both tags. It adds them both together and causes too much volume reduction or gain if you have a postitive replaygain number. Most new music runs between -5db and -12db because of the volume wars. Everyone wants to be noticed and dynamic range be damned. Louder is better. I recommend dbpoweramp and let it do it's thing.

EDIT: Once you have replay gain working, just use your volume dial to adjust. The big round one usually. :)

If dbpoweramp was free I would not cringe every time you have recommended dbpoweramp, which you have done in at least 10 different responses I have seen so far but you really come off like you are getting commission from these guys for your constant recommendation. Recommend some free programs once in while since dbpoweramp is not the only choice out there or necessarily the best for everything...please. : )

GeeJay
2008-08-21, 20:16
If dbpoweramp was free I would not cringe every time you have recommended dbpoweramp, which you have done in at least 10 different responses I have seen so far but you really come off like you are getting commission from these guys for your constant recommendation. Recommend some free programs once in while since dbpoweramp is not the only choice out there or necessarily the best for everything...please. : )

His mention of "foobbar" in the post you quoted was a reference to foobar2000, which is a free program.

cliveb
2008-08-22, 00:43
So still not sure why record levels are being reduced, though I see the lure of trying to get albums to sound equally loud at their loudest points - why not raise the quieter ones to the same level as RHCP (any compression in the mastering will still be unalterable).
I suspect that you may be confused over what factors are involved in the perceived loudness of a track. It is important to appreciate that the loudness of a track is not due to its peak level, but to its average level.

Suppose you have a track that sounds quiet (because it has a low average level) but which peaks at the maximum possible. You can't increase its playback level without either: (a) clipping off the peaks; or (b) compressing its dynamic range. Both operations will introduce distortion. Therefore it is necessary instead to reduce the loud tracks. This is a simple linear volume adjustment, and causes no distortion.

liffy99
2008-08-22, 00:52
Think I follow that CliveB, but take the following;
1 track is recorded with peaks at full volume. 0db (there may be clipping that was introduced in the mastering process which we are powerless to correct).
2nd track peaks at -4db.
So whu can't track 2 be amplified (normalised ?) to reach 0db peaks ?

Talking of normalising - is this a good thing to do when ripping ?

Many thanks

Nonreality
2008-08-22, 04:02
If dbpoweramp was free I would not cringe every time you have recommended dbpoweramp, which you have done in at least 10 different responses I have seen so far but you really come off like you are getting commission from these guys for your constant recommendation. Recommend some free programs once in while since dbpoweramp is not the only choice out there or necessarily the best for everything...please. : )Sorry I just really like it and if you read all my posts you will realize that I recommend mp3tag all the time also. If it bugs you so much just skip my posts. And you can use quite a bit of dbpoweramp without buying it I think. I use it, and I like it and recommend it. I don't get a kickback or anything but usually I tell people that have a reason to know about it. I also tell them that EAC is very good if they want a free one. If you want to repost and chew me out for telling everyone about mp3tag then go ahead. But I forget, that's free. So as long as it's free you won't cringe. I'll try to follow your posting guidelines. Thanks.

Hey Spoon,do I get a discount on the next version?

Nonreality
2008-08-22, 04:05
Think I follow that CliveB, but take the following;
1 track is recorded with peaks at full volume. 0db (there may be clipping that was introduced in the mastering process which we are powerless to correct).
2nd track peaks at -4db.
So whu can't track 2 be amplified (normalised ?) to reach 0db peaks ?

Talking of normalising - is this a good thing to do when ripping ?

Many thanks
If I understand you the reason is you want the differences between tracks to remain as intended. Some tracks are soft and some are loud, that it why you want album gain when listening to an album as it maintains the original differences in the tracks.

Nonreality
2008-08-22, 04:12
If dbpoweramp was free I would not cringe every time you have recommended dbpoweramp, which you have done in at least 10 different responses I have seen so far but you really come off like you are getting commission from these guys for your constant recommendation. Recommend some free programs once in while since dbpoweramp is not the only choice out there or necessarily the best for everything...please. : )I'm also glad that this is the only thing you got out my post. I was trying to help someone not sell them something, as the foobar reference shows. Critics always know best. I also recommend Mediamonkey. Now that's a tough one. It can be free and works well but you can also pay for it and get added stuff you may or may not need. Can I still do that one? Please...

Nonreality
2008-08-22, 04:16
This piece of code from ReplayGain.pm is responsible for smartgain (and similar applies to smart-crossfade)



if (defined $album->replay_gain() && ($class->trackAlbumMatch($client, -1) || $class->trackAlbumMatch($client, 1))) {
return $album->replay_gain();
}
return $track->replay_gain();


1. Check if album-rg is set, if not use track-gain. Therefore always use album-rg if you intend to use smartgain. You should set it to +0.0db otherwise trackgain will be used, which is probably wrong.

2. Check if the previous OR the following title is from the same album. If it is, use album-rg otherwise use track-rg. It has nothing to do with how many tracks from the same album are in your playlist, just what comes next or what was before.

kind regards,
Markus

So if the next song is from the same album then it uses album gain? Like I said I wasn't quite sure how it determined to use album gain. This is good to know. Thanks

cliveb
2008-08-22, 05:02
Think I follow that CliveB, but take the following;
1 track is recorded with peaks at full volume. 0db (there may be clipping that was introduced in the mastering process which we are powerless to correct).
2nd track peaks at -4db.
So whu can't track 2 be amplified (normalised ?) to reach 0db peaks ?
In principle it can. If you're using ReplayGain, and if the calculated perceived loudness of track 2 (according to its average loudness) was less than the reference level, then you'd get a positive ReplayGain setting, which would indeed cause a volume boost.

Let's take this step by step:
1. ReplayGain analyses the tracks and determines what their perceived loudness will be.
2. It has a "reference" loudness. In principle you can set the reference level to whatever you like. There is a default level[1] that was chosen because it's a reasonable compromise.
3. If the analysed track is louder than the reference level, a negative ReplayGain tag is added. If the track is quieter than the reference level, a positive[2] ReplayGain tag is added.

[1] The default is a specific level relative to pink noise at -20dBFS, and is somewhat confusingly referred to as "89dB". (I believe this is because in professional cinema installations this particular level will play at 89dB SPL - at what distance from the speakers I know not).

[2] Note that there is currently an outstanding bug concerning the Squeezebox's handling of positive ReplayGain settings. See here for the gory details: http://bugs.slimdevices.com/show_bug.cgi?id=5119


Talking of normalising - is this a good thing to do when ripping ?
No. Normalisation applies a linear amplitude change, as does ReplayGain. If you use the latter, there's no point in doing the former, and if you *do* apply normalisation when ripping then you're permanently changing the audio data in an irreversible way. ReplayGain operates on playback so it's far more benign.

Teus de Jong
2008-08-22, 07:33
If dbpoweramp was free I would not cringe every time you have recommended dbpoweramp, which you have done in at least 10 different responses I have seen so far but you really come off like you are getting commission from these guys for your constant recommendation. Recommend some free programs once in while since dbpoweramp is not the only choice out there or necessarily the best for everything...please. : )

Why would you cringe if someone recommends a program that's not free but works very well (and let's be realistic, the ultimate version is still only $36).

I'm all for free programs and have a lot of respect for people who make them. I make scripts and small programs for a totally different area and they're free.

But it is perfectly reasonable to ask money for so excellent a program as dbPoweramp. When I started ripping I tried EAC. Although the results were good, I found the program a real PITA to use. I the tried dbPoweramp: it just did what it was supposed to do. Much faster and more intuitive than EAC and with more functionality. So I never looked back, even when the price increased slightly with the release of version 13.

Teus

liffy99
2008-08-24, 00:20
[2] Note that there is currently an outstanding bug concerning the Squeezebox's handling of positive ReplayGain settings. See here for the gory details: http://bugs.slimdevices.com/show_bug.cgi?id=5119

Is this fixed yet ? Now that I've found a number of albums (mostly classical) that do have positive RG added, I presume I'm in danger of introducing clipping ???

Thanks for all the help - really feel I understand RG now.

Nonreality
2008-08-24, 04:39
[2] Note that there is currently an outstanding bug concerning the Squeezebox's handling of positive ReplayGain settings. See here for the gory details: http://bugs.slimdevices.com/show_bug.cgi?id=5119

Is this fixed yet ? Now that I've found a number of albums (mostly classical) that do have positive RG added, I presume I'm in danger of introducing clipping ???

Thanks for all the help - really feel I understand RG now.What is scarier than this bug is how for some reason SC handles having replaygain tags and Soundcheck tags from Itunes. They are basically the same thing, just itunes version of replay gain but for some reason someone at Slimdevices decided if you have both they should be added together. Now for most files it just doubles the negative values as most music has negative values. You just sit and wonder why it's so quiet. But for the odd positive values, well you can figure it out. I've done a couple of bug reports but can't find them. One they didn't understand and put it with another bug about soundcheck. It wasn't the same and had nothing to do with what I wrote. I tried again but for some reason it didn't go through. I got tired of trying and just eliminated all my soundcheck tags. Not happy about that as now I can't use volume adjustment on my ipods but oh well. As far as I know it hasn't been fixed as far as 7.1 So make sure you don't have both types of tags or you do risk maybe speaker damage. It would be rare to have a positive 10 replay gain but if you did it would be a bad surprise to have your volume raised by 20db if you had it very loud to begin with. :(

snoogly
2008-08-24, 14:39
Hmmm. I used to use replay gain, but read somewhere (here or in the documentation somewhere) that it's best not to use it. I may have imagined it, but I am pretty sure I must have had a good reason for disabling it in SC settings.

Am I mad, or is there a reference somewhere that it's best not to use it?

Nonreality
2008-08-24, 17:27
Hmmm. I used to use replay gain, but read somewhere (here or in the documentation somewhere) that it's best not to use it. I may have imagined it, but I am pretty sure I must have had a good reason for disabling it in SC settings.

Am I mad, or is there a reference somewhere that it's best not to use it?
Really no reason not to. Maybe there is some audiophile reason but a lot of audiophiles use it too. It's really needed for mixes. If you listen to 1 album at a time then it's not.

liffy99
2008-08-25, 02:09
Well, gradually going through the whole collection with Foobar 2000 and adding replay gain (question : why doesn't Foobar show the RG details in the RG column ? They appear if I look at Track Properties and if I try to add RG again it tells me they are already there).

Most albums end up with a suggested level reduction, far fewer with an increase. Typically many rock albums seem to get a 5-10db decrease and jazz and classical a 0-5db decrease. A few (mostly classical) albums get a small increase.

So, this will mean that when I listen to them with RG on, levels will be much the same for the average perceived music loudness but I'll have to crank the volume control round more to get back to the original volume levels of the tracks that have had gain reductions.

One thing I'm still a little hazy about is the selection of this 89db default. 89db from what ? If I were to alter the default level by the amount that many albums have had their gain reduced by (say 9db) - to 98db, what would be the result ? All albums would still be at the same average loudness, but louder overall (meaning less cranking of the volume control on the amp). Although I assume I'd have to re-analyse and tag all the files to conform to the new 98db default.

nearly there . . . .

Nonreality
2008-08-25, 02:31
Well, gradually going through the whole collection with Foobar 2000 and adding replay gain (question : why doesn't Foobar show the RG details in the RG column ? They appear if I look at Track Properties and if I try to add RG again it tells me they are already there).

Most albums end up with a suggested level reduction, far fewer with an increase. Typically many rock albums seem to get a 5-10db decrease and jazz and classical a 0-5db decrease. A few (mostly classical) albums get a small increase.

So, this will mean that when I listen to them with RG on, levels will be much the same for the average perceived music loudness but I'll have to crank the volume control round more to get back to the original volume levels of the tracks that have had gain reductions.

One thing I'm still a little hazy about is the selection of this 89db default. 89db from what ? If I were to alter the default level by the amount that many albums have had their gain reduced by (say 9db) - to 98db, what would be the result ? All albums would still be at the same average loudness, but louder overall (meaning less cranking of the volume control on the amp). Although I assume I'd have to re-analyse and tag all the files to conform to the new 98db default.

nearly there . . . .

You'll risk clipping. No check that, you'll have clipping. Just use the volume dial, it's not that big of deal. It's a good level of correction. It allowed for generous peaks without too much risk of clipping. It's the same thing as recording too hot if you increase it. Replay gain is not just peak volume it's more complicated than that. You hit 0 db and you get clipping, square peaks. A lot of new music has that already. You need to do a search on the loudness wars. If you set yours to 98db you are going to have all kinds of problems. Any album that has a replaygain of -2 or more will have clipping and not just the peaks. You really need to trust the people that created this and stay with their value, if you must raise it go to 91db at the most.

liffy99
2008-08-25, 03:00
Wow - 2db ? So virtually all of my rock recordings will have been clipped during mastering ?

Ho hum

Ok, I'll stick with defaults.

The thing I find confusing is the apparent use of different scales. i.e. we talk about 0db as being the "peak" and then 89db as a level below this that allows some headroom. But how much headroom ? What is the gap between 89 and 0db ?

Phil Leigh
2008-08-25, 03:30
Wow - 2db ? So virtually all of my rock recordings will have been clipped during mastering ?

Ho hum

Ok, I'll stick with defaults.

The thing I find confusing is the apparent use of different scales. i.e. we talk about 0db as being the "peak" and then 89db as a level below this that allows some headroom. But how much headroom ? What is the gap between 89 and 0db ?

erm... 11dB (which is a lot: +3dB = twice the "volume" - or more accurately twice the power or energy required)

liffy99
2008-08-25, 04:08
11db ? So peak volume =100db (which is the same as Odb ?) ?

How does that then relate to the dynamic range of the CD medium at 96db ?

And 11db isn't much when looking at peaks at all (about 12 times louder). Think of the contrast in a drum kit - one moment the snare is being tickled with brushes, the next a tom tom given a thwack with a stick - far more than 11db.

pfarrell
2008-08-25, 09:06
liffy99 wrote:
> Wow - 2db ? So virtually all of my rock recordings will have been
> clipped during mastering ?

Clipped during mastering is such an ugly concept.
But virtually all rock recordings are compressed past death as part of
the Loudness Wars.

Nonreality
2008-08-25, 13:41
11db ? So peak volume =100db (which is the same as Odb ?) ?

How does that then relate to the dynamic range of the CD medium at 96db ?

And 11db isn't much when looking at peaks at all (about 12 times louder). Think of the contrast in a drum kit - one moment the snare is being tickled with brushes, the next a tom tom given a thwack with a stick - far more than 11db.
Yes but the snare being tickled by the brushes is not starting at 89. I would hope not at least. Like I suggested before, read up on the loudness wars, it will really help you understand dynamic range and what's going on now in pop and rock recordings. As you use replay gain you'll see that older recordings have smaller numbers and the newer ones much larger. The are making all the sound louder with no regard to dynamic range. In the older albums you could have a difference in what you were talking about of maybe 25dbs and you would really notice the snap of the drum, and now that might be 6 or 7. It doesn't have the impact that it use to have. Everything in the mix is loud. It makes for more fatiguing listening too. Sorry if I'm not explaining this so it makes sense. pfarrell could probably do a much better job of explaining all this. :)

cliveb
2008-08-25, 23:45
[2] Note that there is currently an outstanding bug concerning the Squeezebox's handling of positive ReplayGain settings. See here for the gory details: http://bugs.slimdevices.com/show_bug.cgi?id=5119

Is this fixed yet ? Now that I've found a number of albums (mostly classical) that do have positive RG added, I presume I'm in danger of introducing clipping ???
AFAIK it's not fixed yet. I reported it a long time ago, until recently nobody in Slim Devices seemed to think it was that important. Then a few weeks ago Sean stumbled across it and added his comment. I think if you read between the lines he's basically saying "come on guys, this should not happen and it must be fixed". When it does get fixed is anyone's guess, though.

In the meantime, my notes in the bug report explain how to modify the MySQL database post-scan to avoid clipping due to positive RG values.

liffy99
2008-08-26, 00:43
Thanks Nonreality

I think I do understand the Loudness Wars and the problem with dynamic range (and that brush strokes probably would not start out at 89db - a rather extreme example I'm sorry). What I struggle with is the different scales being used.

One moment I am told that 0db equates to full output and anything that tries to break this "barrier" is clipped. OK - that's clear.

Then this arbitrary level of 89db is touted around, but db is a relative scale, so 89db compared to what ? It surely can't be 89db below full output ! Or is it 89db above the -96db that might equate to the lower threshold of Red Book resolution (which, presunably, could also be represented as -7db).

Then the gap between 89db and "something" is stated as being 11db - so presumambly full output has now been expressed as 100db (and with a fixed maximum dynamic range of 96db, how does this fit in with CD ?).
So this would suggest a max difference of 11db between "average" and "peak" loudness.

Why can't we all use the same scale, for example;

0db = full output
-15db = arbitrary average loudness that Replay Gain would seek to match
-96db = lowest level that CD can reach

So, on a "good" recording, the real dynamic range might span, say, -50db to -2db (giving a 48db range)with an average loudness level of, say, -20db. With the headroom available above -2db it is unlikley that a recording like this has been "clipped".

But what we are often getting might be a range of -20db to 0db (where peaks may have been clipped during the mastering process to ensure no output beyond 0db) with an average loudness level of, say, -9db. Lousy dynamic range, loud and clipped !

Or am I completely missing the boat somewhere ?

Nonreality
2008-08-26, 01:46
Thanks Nonreality

I think I do understand the Loudness Wars and the problem with dynamic range (and that brush strokes probably would not start out at 89db - a rather extreme example I'm sorry). What I struggle with is the different scales being used.

One moment I am told that 0db equates to full output and anything that tries to break this "barrier" is clipped. OK - that's clear.

Then this arbitrary level of 89db is touted around, but db is a relative scale, so 89db compared to what ? It surely can't be 89db below full output ! Or is it 89db above the -96db that might equate to the lower threshold of Red Book resolution (which, presunably, could also be represented as -7db).

Then the gap between 89db and "something" is stated as being 11db - so presumambly full output has now been expressed as 100db (and with a fixed maximum dynamic range of 96db, how does this fit in with CD ?).
So this would suggest a max difference of 11db between "average" and "peak" loudness.

Why can't we all use the same scale, for example;

0db = full output
-15db = arbitrary average loudness that Replay Gain would seek to match
-96db = lowest level that CD can reach

So, on a "good" recording, the real dynamic range might span, say, -50db to -2db (giving a 48db range)with an average loudness level of, say, -20db. With the headroom available above -2db it is unlikley that a recording like this has been "clipped".

But what we are often getting might be a range of -20db to 0db (where peaks may have been clipped during the mastering process to ensure no output beyond 0db) with an average loudness level of, say, -9db. Lousy dynamic range, loud and clipped !

Or am I completely missing the boat somewhere ?Like I said before, I'm probably not the best at this. I think it was that 89db was the max to allow headroom for vinyl recordings. It is just an 11db allowance for peaks. It was to mantain good dynamic range. If you go much more you are not going to like the results with some albums. Actually many albums. If you get a chance, go to hydrogenaudio.com and search a lot and read what you find. It will be much better than me trying to explain. I run my replaygain with the defaults and never have problems so that is part of why I say to use the defaults. I do know that some press it to 91 with ok results and I think that I shared that. I think you need to read up on it from better sources than myself. I've kinda wrapped my head around it but that doesn't mean I can explain it well. :) Oh and as other have said I think, replay gain will not fix bad recordings, you will still have clipping on a lot of modern albums.

cliveb
2008-08-26, 04:17
Then this arbitrary level of 89db is touted around, but db is a relative scale, so 89db compared to what ? It surely can't be 89db below full output ! Or is it 89db above the -96db that might equate to the lower threshold of Red Book resolution (which, presunably, could also be represented as -7db).
It's none of those. We'll get to what it actually is in a moment, but first a bit of background. Take a step back and consider what ReplayGain is trying to achieve. It's aiming to find a way to make the perceived loudness of various tracks and/or albums sound the same. In order to do that, it needs to choose a reference level. So what level should be chosen? There is no standard level in the audio industry, but there is in the movie industry. That standard is called SMPTE RP 200, and it states that a comfortable listening level is one whereby pink noise at -20dBFS RMS is played at 83dB SPL. ReplayGain originally used that level as its reference. In other words, after analysing the music track, it defined the volume change that would be required to make it sound the same loudness as pink noise at -20dB (and which, were it to then be played in a properly calibrated movie theatre, would be at 83dB SPL). Hence the use of the term "83dB" as the reference level.

After a while, actual implementations of ReplayGain gravitated towards aiming for 6dB higher than that, so we now have an unofficial "89dB" reference level. Since you're at liberty to set your playback volume control wherever you like, that 89dB figure is pretty meaningless. For the consumer of digitally-encoded music, it would make more sense to define the reference level as being relative to 0dBFS. On that scale, for a typical music signal, this reference level is about -24dBFS RMS.

Mark Lanctot
2008-08-26, 06:35
So, this will mean that when I listen to them with RG on, levels will be much the same for the average perceived music loudness but I'll have to crank the volume control round more to get back to the original volume levels of the tracks that have had gain reductions.

Sure, but why would you want to? The average CD is so loud that if you cranked up the volume to its "original volume level" it would be way too loud.

RG is attempting to adjust the volume for you - it's like it's moving the knob. With RG, you start playing and settle on a comfortable volume level. You can then leave it alone and as playback moves from track to track, RG will attempt to make sure that the playback volume is even between tracks*, which means you don't have to fiddle with the volume as much.

Without RG, if you had a nice CD mastered in the mid-80s (in my opinion, the golden age of the CD) and then moved to one produced recently, you would almost certainly have to make a huge volume adjustment.

* assuming track gain.

liffy99
2008-08-26, 09:42
Ahahhh - "I can see clearer now . . . . "

Thanks CliveB. So 89db "SPL" is an absolute measure, not a relative one. I quite agree that if we all referred levels to the 0db full output mark it would make a lot more sense.

And the 11db thing is just a complete red herring.

And as for "cranking the volume round more" it's just a psychological thing. After many years of analogue volume controls I've been happy runnning with the knob at the 11am-ish position (which always felt there was a lot more in the pot if needed).

Now I have an all digital set-up with a digital vol. control calibrated in db. Up until replay gain, serious listening was done with the control at about 60-65 (OK the number should be reducing as the volume increases if we are relating everything to the 0db full output mark but hey ho). With RG switched in I'm now listening at abou 70 for the same loudness as before. Just feels kind of close to the max output on the dial of 89 (even though, logarithmically, it is using only a little more than one hundredth of the max power available (on average).

Thanks everyone - I'll butt out of this one now (and write to complain to record companies about their mastering . . . . )

PS Downloaded Stravinsky's Firebird Suite from www.hdtracks.com - now that DOES have some dynamics !

cliveb
2008-08-27, 00:56
So 89db "SPL" is an absolute measure, not a relative one.
I failed to explain that "SPL" stands for "Sound Pressure Level". 0dB SPL is defined as a pressure change of 20uPa (microPascals), and is generally accepted as the threshold of human hearing. (For reference, normal atmospheric pressure is about 100 kiloPascals, so the quietest noise a human can detect is one that varies ambient pressure by about one part in 5 billion).


With RG switched in I'm now listening at abou 70 for the same loudness as before. Just feels kind of close to the max output on the dial of 89 (even though, logarithmically, it is using only a little more than one hundredth of the max power available (on average).
If you're referring to the volume scale on a Squeezebox, then it goes up to 100, not 89. Each number represents 0.5dB, so when you're listening at 70, that's 15dB below the Squeezebox's maximum output.

Mark Lanctot
2008-08-27, 06:01
0dB SPL is defined as a pressure change of 20uPa (microPascals), and is generally accepted as the threshold of human hearing.

Err, shouldn't that be 1 dB, not 0?

cliveb
2008-08-27, 08:48
Err, shouldn't that be 1 dB, not 0?
Well, I can't say with 100% certainty. You could be right. But I was fairly sure the threshold of hearing is defined to be 0dB SPL.

Remember that 0dB doesn't mean "nothing" in an absolute sense. It's just a reference point on a logarithmic scale.

TurnipFarmer
2008-08-31, 13:38
Im just about to replaygain all my files using Foobar2000 however should I select all the files and select them all together and add replaygain tag data as albums.

The reason why I am asking is that I thought about doing a few albums at a time and add replaygain tags to each album which I suppose is the same way but seems to be a bit quicker. By doing this way does it affect the replaygain?

ModelCitizen
2008-08-31, 14:22
Im just about to replaygain all my files using Foobar2000 however should I select all the files and select them all together and add replaygain tag data as albums.
The reason why I am asking is that I thought about doing a few albums at a time and add replaygain tags to each album which I suppose is the same way but seems to be a bit quicker. By doing this way does it affect the replaygain?

Doesn't this help then?
http://forums.slimdevices.com/showpost.php?p=334276&postcount=8

MC