PDA

View Full Version : ReplayGain - How effective?



JJZolx
2011-07-23, 19:12
I've been using ReplayGain consistently for about the past six months. My library is in Flac and I've added RG tags, both track and album, to all tracks in the library.

Typically, for background listening I have a Transporter (living room), an SB2 (kitchen), and a Radio (bathroom) kept in sync. All of them are set to 'Smart Gain'.

My experience has been that ReplayGain is much less effective than I'd anticipated, and only offers some nominal degree of volume normalization. Particularly when listening to Random Mix, I'm still constantly fiddling with volume on the player in the room that I'm in. If someone were to sneak in and disable ReplayGain on all players, I'm not sure I'd notice.

Your experience?

GeeJay
2011-07-23, 21:25
Every once in awhile I will hear a track that doesn't sound like it has the right adjustment, but by and large it works well for me. I still remember how often I had to change the volume before I had ReplayGain tags; I don't have anything approaching that problem anymore. I used to go back and manually change the value if I thought the track was too loud or too soft, but I haven't done that in awhile, either.

BTW most of my listening is also random, and most of my collection is FLAC. I do seem to recall that when I first applied RG tags to my collection, somehow not all my tracks got tagged. It became very noticeable which tracks didn't have a RG tag, and the volume variation drove me nuts. I went back and retagged the entire collection (and every time I rip a new CD I have dbPowerAmp tag the tracks), and things got much better. On the rare occasion where I buy a track online, if I forget to go back and apply an RG tag it becomes very obvious the first time the track comes up in the mix.

JJZolx
2011-07-23, 21:36
The other thing I find about using ReplayGain, even worse than my disappointment with its effectiveness, is that it makes listening to internet radio very inconvenient. Virtually every RG tag in my library is a negative value, with most somewhere between -5 and -7dB. When internet radio is played it BLASTS on all three systems, forcing me to run around the house turning everything down. If I want to switch back to local library listening, it's the reverse. I listen to internet radio much less since setting RG on my Squeezeboxes.

GeeJay
2011-07-23, 21:41
Could it be the type of music you listen to? Most of my listening is pop, rock, country...the 4 minute song where there isn't a lot of variation in volume within a track. I imagine volume equalization would be much more difficult with classical or jazz.

MeSue
2011-07-23, 22:20
I definitely can tell a difference with replay gain, and I listen to a lot of mixes. Switching to a music service usually requires a volume adjustment or when the playlist switches from an album to a mix.

Sounds like you need iPeng for adjusting relative volume on all players at once. ;-)

amcluesent
2011-07-24, 01:59
I use album RG on all pop/rock, since that's been afflicted by the 'loudness war' and needs 6/7/8 dB knocked off. Never needed for classical.

simbo
2011-07-25, 06:19
The other thing I find about using ReplayGain, even worse than my disappointment with its effectiveness, is that it makes listening to internet radio very inconvenient. Virtually every RG tag in my library is a negative value, with most somewhere between -5 and -7dB. When internet radio is played it BLASTS on all three systems, forcing me to run around the house turning everything down. If I want to switch back to local library listening, it's the reverse. I listen to internet radio much less since setting RG on my Squeezeboxes.
Completely agree. Even services like Napster and Last.FM stream their tracks un-normalised (or using a value different to the standard 89db), meaning you still need to adjust your volume control regularly.

Rick B.
2011-07-26, 06:18
I also think RG works great. From time to time, when I think a song or album is too loud, it is usually because I forgot to apply RG to that album when I obtained it.

Phil Leigh
2011-07-27, 06:01
I've been using ReplayGain consistently for about the past six months. My library is in Flac and I've added RG tags, both track and album, to all tracks in the library.

Typically, for background listening I have a Transporter (living room), an SB2 (kitchen), and a Radio (bathroom) kept in sync. All of them are set to 'Smart Gain'.

My experience has been that ReplayGain is much less effective than I'd anticipated, and only offers some nominal degree of volume normalization. Particularly when listening to Random Mix, I'm still constantly fiddling with volume on the player in the room that I'm in. If someone were to sneak in and disable ReplayGain on all players, I'm not sure I'd notice.

Your experience?

Never had that experience; never felt the need to reach for the volume after initially setting it for a session (except to gradually turn it up of course! - never to turn it down).

Are you sure RG is active on your setup/player? :-)

I don't keep switching between radio and library so the difference in relative levels doesn't bug me.

servies
2011-07-27, 08:07
What happens if you have your volume set at 100% and the track that's playing has a replaygain value of (par example) +8 dB

verypsb
2011-07-27, 09:10
What happens if you have your volume set at 100% and the track that's playing has a replaygain value of (par example) +8 dB

Replay Gain tries to play everything at +89dB. A track would get a +8dB RP value, if the track's normal value is +81dB.
So yes, it's best to set your volume to 100%.

Phil Leigh
2011-07-27, 09:33
What happens if you have your volume set at 100% and the track that's playing has a replaygain value of (par example) +8 dB

Nothing bad or odd will happen. The playback volume and the RG tag value calculations are unrelated. If RG was +8, the track would be rather quiet, peaking at circa 81dB... so even after RG boost of +8 it will now be peaking at 89dB...

RG cannot induce clipping if calculated properly.

servies
2011-07-27, 10:46
Nothing bad or odd will happen. The playback volume and the RG tag value calculations are unrelated. If RG was +8, the track would be rather quiet, peaking at circa 81dB... so even after RG boost of +8 it will now be peaking at 89dB...

RG cannot induce clipping if calculated properly.
Ok, I was under the impression that with the playback volume set to 100% there was no room for replaygain to make it louder...

Phil Leigh
2011-07-27, 11:08
Ok, I was under the impression that with the playback volume set to 100% there was no room for replaygain to make it louder...

Just to reiterate, RG tries to make everything the same level (usually 89dB)... and that happens BEFORE the digital volume control - which can only make things quieter - is applied, not after.

MrSinatra
2011-07-28, 01:00
i've been using it for several years now, and i find it works pretty well.

one thing to look for, is peak values, the forgotten RG tags. using those is how it avoids clipping, and so a track might be played back quieter than it otherwise would be, to avoid the clip. usually not an issue, but if something sounds too quiet it might have a momentary loud spike in it or something like that.

the internet radio thing is an old complaint, vote for this bug:

http://bugs.slimdevices.com/show_bug.cgi?id=2431

and

http://bugs.slimdevices.com/show_bug.cgi?id=3152

and

http://bugs.slimdevices.com/show_bug.cgi?id=4335

and

http://bugs.slimdevices.com/show_bug.cgi?id=4362

having said all that, i do find that sometimes some things seem louder or quieter than all the rest, BUT for the most part, i think it does its job and does it well.

ps. i would also like to see a RG "pre-amp" so a user can adjust the 89db reference level. this is available in winamp.

Phil Leigh
2011-07-28, 01:28
...
ps. i would also like to see a RG "pre-amp" so a user can adjust the 89db reference level. this is available in winamp.

That would be a great enhancement request...

MrSinatra
2011-07-29, 01:57
here's a pic of the winamp RG implementation.

interestingly, along with apply gain, the dropdown also has "normalize" but i don't use that or really know how its different in practice.

what winamp does not have, is a smartgain feature, where it uses track or album gain as appropriate as per the playlist.

Ryssen
2011-07-30, 12:35
Doesn´t the use of RG comparable with turning down (or up) a digital volume control,you lose resolution?

Bytec
2011-07-30, 12:44
I use Foobar2000 to add ReplayGain info and since v1.1.6 ReplayGain scanner uses libebur128 for improved accuracy.

I'm quite satisfied with the results.

MrSinatra
2011-07-30, 16:40
Doesn´t the use of RG comparable with turning down (or up) a digital volume control,you lose resolution?

imo, RG is exactly like turning up or down the gain on an analog amp volume knob, except the gain is applied in the digital realm, and so less distortive.

firedog
2011-09-02, 23:25
I added track and album gain tags to all my files b/c I often listen in random mode and got tired of constantly adjusting the volume.

In general, using the "smart gain" setting seems to work fairly well. However, I find that for some tracks (source is very loud?) I still feel the need to turn the volume up. In fact, on these tracks I have to turn the volume up massively to get the level I want. Not something I'm sure I want to do to my speakers.

What I don't understand, is that if I play the same tracks in a whole album setting (not during random play) I still notice the same problem. I'd think that when the playlist is one entire album that the relative volumes of the tracks would be preserved and I wouldn't hear this same phenomenon.

bobkoure
2011-09-07, 08:06
Doesn´t the use of RG comparable with turning down (or up) a digital volume control,you lose resolution?
A digital volume control is down only.
Loss of resolution depends on how far down the control is set. There's been a lot of discussion about this on this forum over the years, but to (potentially over) simplify, a digital volume control basically shifts bits downward and the bottommost bits are discarded. If those bottommost bits are all zeros, then you didn't lose any resolution. With 16 bit samples, being handled as 24 bit, there are always some discardable bits, so no worries, at least not in the ranges that RG will get you to with player volume set to 100%.
If that's over simplified, search for those forum conversations, there were pages of it. As I remember (and IMO, of course), the best explanation came from Sean Adams (user seanadams), which I can't find right now (anyone got a link?)