replay gain - sounds compressed
Oct 12, 2020 at 5:43 AM Thread Starter Post #1 of 12

magicalmouse

500+ Head-Fier
Joined
Oct 21, 2013
Posts
721
Likes
118
Location
uk
i know it should make no difference but comparing using replaygain and then turning it off, when off the music seems less compressed and more dynamic.

is this real or just my imagination

i am using fiio x7/foobar/mojo (coax) and hd25 aluminiuim (modded)

d
 
Oct 12, 2020 at 8:00 AM Post #2 of 12

VNandor

500+ Head-Fier
Joined
Oct 17, 2014
Posts
567
Likes
226
You could try a null-test. Capture the sound with replaygain turned off and then turned on. After that level match them, align them in time, invert one of the tracks add them together and see what's left. If you don't know what any of that means or how to do that, I can explain more or just do it myself and tell you what I've found.
 
Oct 12, 2020 at 9:45 AM Post #3 of 12

castleofargh

Sound Science Forum Moderator
Joined
Jul 2, 2011
Posts
9,448
Likes
4,836
i know it should make no difference but comparing using replaygain and then turning it off, when off the music seems less compressed and more dynamic.

is this real or just my imagination

i am using fiio x7/foobar/mojo (coax) and hd25 aluminiuim (modded)

d
On average the sound will be a good deal louder without replaygain, and loudness differences can cause soooooo many subjective impressions beside feelings of loudness. It's the very reason why we tend to dismiss or look very skeptically at any feedback not involving matched level outputs.

Replay gain in itself does not compress anything. The basic stuff is just an extra metadata(so the track is still the same) telling the player to digitally reduce(or increase) the amplitude by XdB on playback.

If you apply it to very dynamic tracks like most classical music, then you need to pay extra attention or the result will often be a boost and the track could clip. In foobar you just have to pick the replaygain setting with prevent clipping in it and so long as the replay gain scan was done with some amount of oversampling(to catch intersample peaks), you should be fine. If not, maybe you're getting clipping or compression because some other DSP is handling the clipping(some form of compressor could do that, but you have to put it there?).
In any case, it's easy to check if the song you find "compressed" has a positive or negative replaygain value after being scanned. Right click on the track pick properties, and go to the second tab called "details" in the pop up window. If replaygain lowers the volume, then I don't see how it's going to compress anything. It just sound quieter and your brain makes up a story about how that feels but it's probably just quieter.

I don't remember if you can show gain and peaks just with a right click on the columns line by default, but it's foobar so you can always add stuff:
I have this pattern showing track gain %replaygain_track_gain%
And this one showing what the replay gain scan saw as track peak %replaygain_track_peak_db%
those can be added in "preference"->"display"->"playlist View". And just "add new" in the custom column if you're slightly paranoid like I am ^_^.
Just note that the peak value is that being scanned by the replay gain tool, it isn't necessarily the true peak value. You can select the amount of oversampling and even the type of resampler in the gain settings and the resulting peak value displayed will change based on those settings. The higher the amount of resampling, the more accurate the value. But at the same time you radically increase the CPU usage while scanning for replaygain, and increase the duration of the scan. so, many people will just chose a lighter scan option and don't worry about potential small clips(most actually occure at low freqs and aren't perceived as clipping). Or they just remove 2dB or 3dB on the preamp ("preferences"->"playback") to deal with the possible error in the true peak reading and just stop thinking about that.

I'm guessing I went well beyond your specific situation, but we never know what you are actually experiencing. And that might still interest someone using foobar.

edit , I don't know how the X7 handles replaygain, but I'm guessing it's still based on foobar's metadata.
 
Last edited:
Oct 12, 2020 at 3:14 PM Post #4 of 12

bigshot

Headphoneus Supremus
Joined
Nov 16, 2004
Posts
22,926
Likes
4,656
Location
Hollywood USA
Louder sounds better, even if it is exactly the same.
 
Oct 13, 2020 at 2:59 AM Post #5 of 12

magicalmouse

500+ Head-Fier
Joined
Oct 21, 2013
Posts
721
Likes
118
Location
uk
Thank you for your interesting observations, i have another question based upon these, is every program that codes the replaygain tags the same or are some better at it, i use media monkey to set the tags as it is easier to do both track and album at the same time, maybe i should be doing it with foobar or musicbee if these do the oversampling mentioned above?

d
 
Oct 13, 2020 at 5:17 AM Post #6 of 12

dazzerfong

1000+ Head-Fier
Joined
Jul 5, 2014
Posts
1,396
Likes
645
Location
Sydney
Thank you for your interesting observations, i have another question based upon these, is every program that codes the replaygain tags the same or are some better at it, i use media monkey to set the tags as it is easier to do both track and album at the same time, maybe i should be doing it with foobar or musicbee if these do the oversampling mentioned above?

d

It's complicated.

On a whole, "replay gain" is split into three camps: MP3 gain, replay gain and music player-based gain normalisation.

MP3 gain works universally: that's because in each frame of MP3, an offset is applied. This offset is read as part of the MP3 decoding, and good thing is that this offset is non-destructive (i.e. undoes it).

Replay gain in other formats applies a tag in the footer for programs to read: sometimes, the program choose to ignore it.

Then we have the music player normalisation, where the player 'guesses' how to normalise. This is usually the one that compresses and destroys the dynamic range, as it doesn't necessarily apply the gain universally throughout the whole song, but rather through snippets in an attempt to normalise. In Musicbee, all three of the above can be used: see picture below.

1602580577812.png

MP3 gain is automatic if you used it on your files.

First red line is the 3rd type: auto normalising. I really don't recommend that one.

Second red line is the replay gain tag, or the 2nd type.
 
Oct 13, 2020 at 7:16 PM Post #7 of 12

old tech

500+ Head-Fier
Joined
Jul 8, 2015
Posts
809
Likes
397
Location
Sydney, Australia
It's complicated.

On a whole, "replay gain" is split into three camps: MP3 gain, replay gain and music player-based gain normalisation.

MP3 gain works universally: that's because in each frame of MP3, an offset is applied. This offset is read as part of the MP3 decoding, and good thing is that this offset is non-destructive (i.e. undoes it).

Replay gain in other formats applies a tag in the footer for programs to read: sometimes, the program choose to ignore it.

Then we have the music player normalisation, where the player 'guesses' how to normalise. This is usually the one that compresses and destroys the dynamic range, as it doesn't necessarily apply the gain universally throughout the whole song, but rather through snippets in an attempt to normalise. In Musicbee, all three of the above can be used: see picture below.

1602580577812.png
MP3 gain is automatic if you used it on your files.

First red line is the 3rd type: auto normalising. I really don't recommend that one.

Second red line is the replay gain tag, or the 2nd type.
I'm using Bluesound which has a choice of track gain, album gain or smart gain. Which of these should I use? Would smart gain be similar to the music player issue described above?
 
Oct 14, 2020 at 1:53 AM Post #8 of 12

dazzerfong

1000+ Head-Fier
Joined
Jul 5, 2014
Posts
1,396
Likes
645
Location
Sydney
I'm using Bluesound which has a choice of track gain, album gain or smart gain. Which of these should I use? Would smart gain be similar to the music player issue described above?

I would assume so - the software tries to dynamically compress the dynamic range, which describes what you're facing.

I personally use album gain for gapless albums, and track gain for everything else.
 
Oct 14, 2020 at 1:29 PM Post #9 of 12

castleofargh

Sound Science Forum Moderator
Joined
Jul 2, 2011
Posts
9,448
Likes
4,836
I'm using Bluesound which has a choice of track gain, album gain or smart gain. Which of these should I use? Would smart gain be similar to the music player issue described above?
The pretty vague explanation on their website suggest an either/or use of track and album gain. My fertile imagination thought that maybe it depends on the playback mode picked? Like when you shuffle it always goes for track gain, and when you just use default playback, it will notice an album and then switch to album gain?
Zero clue if that's correct. ^_^
 
Oct 15, 2020 at 3:46 PM Post #10 of 12

bigshot

Headphoneus Supremus
Joined
Nov 16, 2004
Posts
22,926
Likes
4,656
Location
Hollywood USA
I wish they were smart like that. I shuffle a lot. But most of the time, they just go through and apply a volume level in the data fork of the file. It plays that level however you choose to play it. So Dazzle is right about how to apply it.
 
Oct 15, 2020 at 4:02 PM Post #11 of 12

castleofargh

Sound Science Forum Moderator
Joined
Jul 2, 2011
Posts
9,448
Likes
4,836
I wish they were smart like that. I shuffle a lot. But most of the time, they just go through and apply a volume level in the data fork of the file. It plays that level however you choose to play it. So Dazzle is right about how to apply it.
Sure. That's the logical approach. In Foobar (can't remember if it's a component or baked in), I use something called "replaygain overide mode", where I set it to apply album gain when the playback mode is "default", and track gain when the playback is set to "shuffle". So I guessed it could be possible that the Bluesound "smart gain" might do something similar based on how they describe it on the website. But again, that was just a wild wild guess.
 
Oct 15, 2020 at 4:07 PM Post #12 of 12

bigshot

Headphoneus Supremus
Joined
Nov 16, 2004
Posts
22,926
Likes
4,656
Location
Hollywood USA
I think the gain correction is saved in the file something like "+3dB" in the file. When you play it back in override, it just ignores that and plays it back at the 0 level. For it to know to play it back without gain correction in album play, there would have to be communication between the player and the gain corrector. The one I use appears to be completely separate from the player. It goes through and analyzes the file and assigns a level all on its own.
 

Users who are viewing this thread

Top