advertisement


ReplayGain, is anyone else using it?

tuga

tuga

Legal Alien
I find the huge disparity in level between '80s to early '90s and subsequently released rock CDs very anoying.
In extreme cases, as the example below (Neil Young's "Harvest" vs. Radiohead's "Nude"), that difference reaches 12dB (perceived as over 2x as loud).

0DdQw3a.jpeg


ipB4MS0.jpeg


I have my files stored locally, and since HQPlayer will losslessly (using 64/80/128-bit and arbitrary precision floating point processing) upconvert to DSD, adjust gain and EQ in one step I decided to try 'ReplayGain'.

As described in the Hydrogen Audio Knowledgebase, ReplayGain is the name of a technique invented to achieve the same perceived playback loudness of audio files. It defines an algorithm to measure the perceived loudness of audio data. (more info here)
The metadata is added to the file and the playback software will decrease or increase the track or album gain.

And I'm sold. No more massive volume adjustments between albums required (or any kind of volume adjustments).
The downside, in my case, is that 'ReplayGain' works with FLAC but not AIFF, so I had to convert all my rock library to FLAC (an easy task with XLD) before I could perform the analysis and tagging and I can no longer use iTunes/Music for browsing.
Files can be tagged for 'Track Replay Gain' and 'Album Replay Gain' but HQPlayer only has the option for the latter, which is useless with Classical music.

Don't know about other streaming services but Spotify provides a similar option ('Normalise volume'), although I suspect it will be lossy. Apple's iTunes/Music also has its own version of 'ReplayGain', called 'Sound Check', and again it's likely to be lossy.
 
Last edited:
Yeah, been using it (and similar) since soon after I first started ripping my CDs and creating a music server of flac files over 12 years ago. I used ReplayGain whilst using Squeezebox server and have been using Roon's Volume Levelling since getting that in 2019. I've found it invaluable and would never go back to the days of having to keep jockying the volume control each time I start a new album.
 
Roon will do the same. I dont use it as adjusting the volume is so trivially simple I prefer to leave the data unmolested as far as possible. Yes I know that normal signal processing involves significant data manipulation but why add to it.
 
Colin L said:
Roon will do the same. I dont use it as adjusting the volume is so trivially simple I prefer to leave the data unmolested as far as possible. Yes I know that normal signal processing involves significant data manipulation but why add to it.
Click to expand...

It's no different than using digital volume.
If you use HQPlayer you include it in the DSP calcs (I already adjust the output to -6dB to avoid intersample overs).
With other equipment it may indeed add to other DSP.
 
AIUI - The Replaygain method uses meta data, adding ID Tags, either per Track or per Album. It doesnā€™t manipulate the actual data file in any way, it tells the playback software to bump/drop/leave the volume as is.

It uses algorithms/calculations and aims to get the output standardised to 89dB. Google for more info on EBU R128 and target volume -18LUFS, they are the usual choices.

If the playback software doesnā€™t understand the Replaygain ID Tags you can manipulate the actual data file using a Replaygain DSP or use Volume Normalisation.
 
Amber Audio said:
AIUI - The Replaygain method uses meta data, adding ID Tags, either per Track or per Album. It doesnā€™t manipulate the actual data file in any way, it tells the playback software to bump/drop/leave the volume as is.
Click to expand...

That is it, although to be absolutely correct the playback software adjusts the gain, not the volume, and that can be a lossy operation depending on the implementation (HQPlayer and Roon are able to perform lossless adjustments).

It requires software that is able to analyse and add the Track ReplayGain and Album ReplayGain metadata and the use of a file format and a playback software that are compatible.

The link I posted earlier is very informative:

ReplayGain - Hydrogenaudio Knowledgebase

wiki.hydrogenaud.io wiki.hydrogenaud.io
 
Not knowing the details, but simply looking at the plots the second example seems grossly clipped. If more than two samples in succession are clipped then reconstruction becomes quite dubious. FWIW If I want to adjust level I use sox. This also lets you upsample if you wish, scaling down the overall level. It can also let you DIY the applied filter for the upsampling. But if the clipping lost real info, you can only guess what to replace it with.
 
Jim Audiomisc said:
Not knowing the details, but simply looking at the plots the second example seems grossly clipped. If more than two samples in succession are clipped then reconstruction becomes quite dubious. FWIW If I want to adjust level I use sox. This also lets you upsample if you wish, scaling down the overall level. It can also let you DIY the applied filter for the upsampling. But if the clipping lost real info, you can only guess what to replace it with.
Click to expand...

The plots above are made from the original files (extracted/ripped from the CD). Unfortunately a lot of modern rock music and remasters are pushed to the limit.
Clipping is another matter, not being addressed by ReplayGain.

Might be worth reading about ReplayGain in the link above.
It doesn't touch the file/signal, it merely calculates the amount of gain or attenuation a file needs to meet the user-defined LUFS (Loudness Units relative to Full Scale) and adds that information to the metadata, which subsequently read by the music player which will adjust the gain. No Socks šŸ˜ required.
This is what you see when you open the FLAC file with a text editor:

SfokUXA.png
 
I misguidedly thought replay gain would mess up my bits. Wish I'd known.

The difference in levels without replay gain is a PITA.
 
I have all of my tracks processed with ReplayGain (both track and album numbers). I can use "Smart Gain" with LMS (Logitech Media Server) to utilize this. However, I find that it negatively affects the sound quality, so I have the feature turned off for the two players where I do my own listening.

For all the other systems, though, I leave the feature enabled on the player. That way if someone queues up a bunch of music from multiple albums, the volume is consistent for their background listening.
 
Mike Hanson said:
I have all of my tracks processed with ReplayGain (both track and album numbers). I can use "Smart Gain" with LMS (Logitech Media Server) to utilize this. However, I find that it negatively affects the sound quality, so I have the feature turned off for the two players where I do my own listening.
Click to expand...

I don't notice any difference in sound quality between having HQPlayer adjust the gain and using either it's digital volume control or the amplifier's.
It might be a limitation of LMS' 'Smart Gain'.
 
tuga said:
I don't notice any difference in sound quality between having HQPlayer adjust the gain and using either it's digital volume control or the amplifier's.
It might be a limitation of LMS' 'Smart Gain'.
Click to expand...
I can't recall all the particulars, but I believe that LMS reduces the volume of the louder tracks to bring them in line with the volume of the quietest track in the playlist. In doing so, if there was a particular track that was very quiet, some of the other tracks sounded rather squashed. This was likely being done in the digital realm, and probably not as well as HQPlayer.
 
If youā€™re going to use replaygain tags you should probably use an EBU R128 compliant tool as it determines gain by reference to perceived loudness rather than a fixed reference like that used in the old replaygain 2.0 spec.

When listening to albums I have replaygain disabled, when letting LMS DJ for me (typically when Iā€™m toiling and need something to occupy my mind) I enable replaygain to avoid massive jumps in volume
 
You must log in or register to reply here.


advertisement


Back
Top