Replay Gain question?

[Eleven13]

Hey everyone,

Sorry if this question is in the wrong place, I just wasn't sure where else it would fit in.

I've had dbpoweramp for a while now, and I use it to rip all of my CDs to FLAC format which I put on my iPod with rockbox.

When initially setting up dbpoweramp ripper, I just followed the setup guide exactly as the pictures showed, so when I got to the DSP "Replay Gain" section, I left the setting "EBU R128 Calculated Gain -23 LUFS Target Volume" just as it was in the picture at -23.

I sort of forgot all about that setting for a while until I had an Mp3 file that was really loud in comparison to my other files, and I used the batch converter to add the Replay Gain DSP tag to it and it fixed the volume.

However, it was also then that I realized I never looked into what the optimal setting for the Replay Gain should be. After reading around, a lot of people are saying it should be instead set to -18.

What is the difference between -23 and -18? Is it just that -18 will be able to be played at a louder volume? I did notice that even maxed out on my iPod the music doesn't get overtly loud, so it got me thinking that maybe I screwed up and should have changed that setting a long time ago.

So overall, my questions are:

1.) Would a change from -23 to -18 make a very significant different in sound volume/quality with my music files?

2.) Since I've already ripped probably around 150 CDs, would the process of changing all of the tags be a nightmare if I wanted to change them all to -18?

Thanks for any help and sorry if this is a dumb question!

[dbfan]

Re: Replay Gain question?

The difference is just louder, however to iPod will not use replaygain tags, only itunenorm

[AndersHu]

Re: Replay Gain question?

An iPod running Rockbox can use replaygain tags.

EBU R128 is a broadcast standard for loudness normalisation, and it recommends to normalize audio at -23 LUFS.
The ReplayGain uses a different algorithm for measuring loudness, and defines a set of tags that the player can use.
Now we are writing ReplayGain tags, but use the R128 algorithm to measure the loudness.

You should chose one target for all your files.

User tests have shown that if you set R128 to -18 LUFS the result is close in volume to the old ReplayGain(89 dB) results.

[garym]

Re: Replay Gain question?

An iPod running Rockbox can use replaygain tags.

and one can also use dbpoweramp DSP or an action in mp3tag, to create SOUNDCHECK tags (iTunNorm) based on either track or album RG values. These SC values are then used by an ipod/iphone etc. This works with standard apple software (no requirement for rockbox).

[Eleven13]

Re: Replay Gain question?

Thanks for the replies everybody! Currently I have all of my music stored on my HD in separate folders according to bands and then each album in a separate subfolder. Would there be a way to change all of the tags at once without needing to really do any rearranging, or is it just not worth the hassle and I should just keep everything at -23?

[garym]

Re: Replay Gain question?

Thanks for the replies everybody! Currently I have all of my music stored on my HD in separate folders according to bands and then each album in a separate subfolder. Would there be a way to change all of the tags at once without needing to really do any rearranging, or is it just not worth the hassle and I should just keep everything at -23?

if everything is already -23 I'd just leave it. The only reason to change is if 1/2 is at the old RG level (approx. -18) and the later 1/2 is -23. You might want them consistent. But in any case, you can easily use the batch RG DSP to reapply RG values at -18 to your entire collection in a batch manner given that your tagging and organization is good.

[Eleven13]

Re: Replay Gain question?

if everything is already -23 I'd just leave it. The only reason to change is if 1/2 is at the old RG level (approx. -18) and the later 1/2 is -23. You might want them consistent. But in any case, you can easily use the batch RG DSP to reapply RG values at -18 to your entire collection in a batch manner given that your tagging and organization is good.

Thanks for the information, all of my songs are -23 so they're uniform, but I just wasn't sure if I'd be better off having them at -18 and rip any new CDs at -18 from here on out.

I took a look at the batch converter and it does seem like it'd be fairly easy to convert it all at once. I did notice though that when I click on the box next to an artist folder, two yellow dots appear in the box, but when I click the box again it changes to one yellow dot, then once again clears back to blank. What do the dots mean? And if I were to convert them all using the batch converter will it keep my folder structure exactly as it is?

I'm still debating whether I'll switch them over, but if I do I just have two more quick questions:

1. Will simply applying the new tag using the batch converter overwrite the previous tag, or will I have to somehow delete it first?

2. I have all of my songs FLAC compressed level 5 currently. After doing some thinking I feel like if I were to do all the tag changes I might be better off while I'm at it to just convert them all to FLAC compressed level 8 to save more disk space as well.

If I were to do that, would going from level 5 files to level 8 files do anything weird or would it be a seamless conversion? You never lose any sound quality with FLAC regardless of its compression level right? So in theory it is best to have it as compressed as possible to save space?

[garym]

Re: Replay Gain question?

Thanks for the information, all of my songs are -23 so they're uniform, but I just wasn't sure if I'd be better off having them at -18 and rip any new CDs at -18 from here on out.

I took a look at the batch converter and it does seem like it'd be fairly easy to convert it all at once. I did notice though that when I click on the box next to an artist folder, two yellow dots appear in the box, but when I click the box again it changes to one yellow dot, then once again clears back to blank. What do the dots mean? And if I were to convert them all using the batch converter will it keep my folder structure exactly as it is?

I'm still debating whether I'll switch them over, but if I do I just have two more quick questions:

1. Will simply applying the new tag using the batch converter overwrite the previous tag, or will I have to somehow delete it first?

2. I have all of my songs FLAC compressed level 5 currently. After doing some thinking I feel like if I were to do all the tag changes I might be better off while I'm at it to just convert them all to FLAC compressed level 8 to save more disk space as well.

If I were to do that, would going from level 5 files to level 8 files do anything weird or would it be a seamless conversion? You never lose any sound quality with FLAC regardless of its compression level right? So in theory it is best to have it as compressed as possible to save space?

First understand that -18 vs -23 is equivalent to a minor adjustment of your volume knob on your stereo. Nothing more nothing less. There is nothing magical about this. The only issue is that if some things are -18 and some are -23, then you'll lose the benefit of RG playback (which after all is to try to "equalize" the playback volume of different CDs so you don't have to change the volume knob on every song/album).

Also, you can point at your parent music directory and reapply RG to ALL your folders in one single batch. I have > 6,000 albums. But they are all under one parent directory
music\
music\artist
music\artist\album\tracks....

So If I point at the top directory (music) and run the RG DSP on it, it will handle all 6,000 albums at once. no problem.

1.. adding new tag will auto overwrite the old tag, no need to delete.
2. why change from 5 to 8. I can't recall the calculations, but the space saved would be minor. I looked at this once because I have so many CDs, but I calculated that the difference between 5 and 8 across even 6,000 CDs was a trivial amount of space.
3. going from any FLAC to any FLAC (compression change or anything) changes nothing about sound quality, etc. That's why these are called LOSSLESS files. Lossless=lossless=lossless, etc.

If I were you, I would continue ripping files using -23 RG (since all is at that level already), not redo my exisiting files, and maybe start using -8 compression going forward. But no need to convert all your old rips to -8.

[schmidj]

Re: Replay Gain question?

As someone who has worked in professional audio his entire life, I want to tip you off about an issue concerning R128 levels set to -23 dB LUFS for home and car playback systems. R128 was developed for use in professional broadcast plants which have, since the dawn of the digital broadcast era, used different audio level standards than consumer equipment. Consumer analog audio, since the dawn of its existence, has always "normalized" audio levels as loud as could be obtained without "objectionable" distortion. This was initially done to reduce noise in media with poor signal-to-noise ratios, and for radio and later television, to maximize coverage areas. The FCC required that peaks of the audio resulted in modulation between 85 and 100 percent, a couple of dB. (6 dB down is 50 % modulation.) Psycho-acoustic studies, easily reproduced, soon also discovered that music played louder, even slightly louder, sounded "better" to most human ears. Advertisers and record labels took advantage of this by compressing the dynamic range of commercials and recordings and keeping the peaks at the same high levels, raising the average level, the perceived loudness. Advertisers carried it to extremes, resulting in "loud" commercials and listener complaints.

In the early days of digital consumer audio, with the birth of the CD in the early 1980's CD's began being mastered the same as records, as loud as possible. The audio was compressed or limited and then the audio level was adjusted or "normalized" so the peak audio level encoded on the CD exactly hit 0 dBFS, the peak level that could be encoded in the audio file. In practice, this placed the "average" level, as measured by a VU meter about 8 to 12 dB below the peak level as measured by a true peak meter for most well mastered material that was not overly compressed. The CD was now considered to be the same level as the phonograph record or radio signal. CD players set the gain of their playback electronics accordingly so that CD's played on music systems played at about the same perceived level as the radio tuner, cassette player, or phonograph input on home stereo systems and car radios.

In the analog audio days, professional broadcasters and recording studios also kept levels, particularly on "controlled" parts of their systems (anything after a meter where someone presumably would adjust the levels to keep them optimized) about 10 to 18 dB below the point where distortion became objectionable. This was the level which the setup tone, typically at +4 dBu in recording plants, often at +8 dBu or dBm in broadcast plants was adjusted to. However with the dawn of the digital audio equipment area in the broadcast world, the Society of Motion Picture and Television Engineers (SMPTE) realized that noise was not as big an issue then, but digital clipping was, and set the set-up or "reference" tone to be set to -20 dBFS in early digital video tape recorders. This eventually became the USA standard for professional digital audio levels. The Europeans came up with a reference level 2 dB hotter, -18 dBFS. (a minor annoyance to international broadcast organizations to this day.)

It should be noted that the average audio levels as measured on a VU meter for digital audio played on a -20 dBFS professional system will be seen to be about 6 to 12 dB below the audio levels of the same material played from the digital output of a CD player or the digital output of a computer or other consumer device playing ripped CD's or commercially released MP3's or most other audio files., because of the lower reference levels used in professional broadcast (and recording) installations. This does not effect the levels coming from the broadcast transmitter, because the processing equipment at the input to the transmitter modulator makes up the additional gain.

Now we get to LKFS levels. Over the years, people became mightily annoyed at the "loud" commercials (actually heavily dynamically compressed, so they played loud in a peak level controlled system). They complained to government regulators, who complained to licensed broadcasters. The European broadcasters, through their principal standards-setting organization, the EBU, began an investigation into perceived loudness and discovered that, lo and behold, the average audio level was much closer to the perceived level than the instantaneous peaks (which often contained little energy). Further study tweaked things and they came up with a new loudness measuring algorithm, and a new loudness measurement for digital audio, the LKFS. Further study found that in a broadcast plant that was using -18 or -20 dBFS set-up tones and peak or VU meters calibrated accordingly, if you used a meter (and a person with a fader adjusting accordingly) or a loudness controller set for an output level of -23 LKFS you would get typical levels for most program material similar to what you had before, but with overly compressed material (loud commercials) now softer, and wide dynamic material somewhat louder.

The European broadcast regulators and the then the FCC adopted this as a new TV audio loudness standard, and those of us in the TV broadcast business ran around for over a year installing new (expensive) meters and controllers, and trying to teach people how to read and use them.

But none of this applied to consumer equipment which never used the professional -18 or -20 dBFS set-up level. If you normalize your digital audio collection to -23 LKFS according to the new standard, you will find that, on consumer equipment, it will play back much softer than other material recorded to typical consumer "standards". If you are listening on your receiver or car radio to your audio files, and then switch to a radio station without adjusting the volume control, the radio station will be too loud, annoyingly too loud. That is because the design of the consumer device you are using was not based on audio being played at -23 LKFS, but rather much "hotter" levels. Furthermore, if you are using a personal listening device, such as an Ipod or smart phone, or even a tablet or laptop with headphones, you may find that the material plays back too soft, even with the volume control turned all the way up. Many of these devices purposely limited the audio playback gain so you couldn't blow your ears out, either accidentally or on purpose, bu turning the level way up.

So -23LKFS is not suited to consumer storage systems. People here have reported that subjective tests have shown that -18 LKFS seems to work well. That is 6 dB hotter, twice the voltage, somewhat less than twice as loud (10 dB). Based on my experience, this is probably somewhat low for most compressed pop material. but it does give you more headroom for wide dynamic range material (18 dB if the analog part of your system can handle full level digital signals), and is probably a good compromise between matching levels with other sources and maintaining improved audio quality.

I've rambled on quite a bit here, and I've left out the effects of pre-emphasis (which impacts FM and analog TV, records, and analog tape formats, but not so much digital) and analog tape saturation, which essentially provides an automatic and a typically pleasing peak level limiter.

If you read through all of this, I hope you get a better understanding of why -23 LKFS doesn't work in the consumer world, and the history of audio levels.

[Dat Ei]

Re: Replay Gain question?

Hey Eleven13,

I did notice though that when I click on the box next to an artist folder, two yellow dots appear in the box, but when I click the box again it changes to one yellow dot, then once again clears back to blank. What do the dots mean? And if I were to convert them all using the batch converter will it keep my folder structure exactly as it is?

the meaning of the yellow dots and yellow dot is explained in the toolbar of the program - there is a legend.
If you keep your folder structure or not, depends on how you configure the conversion. You can keep your structure, but you don't have too.


Dat Ei

[garym]

Re: Replay Gain question?

As someone who has worked in professional audio his entire life, I want to tip you off about an issue concerning R128 levels set to -23 dB LUFS for home and car playback systems.

[...]

If you read through all of this, I hope you get a better understanding of why -23 LKFS doesn't work in the consumer world, and the history of audio levels.

Thanks for the excellent explanation. I see why -18 makes the most sense.

[Eleven13]

Re: Replay Gain question?

First understand that -18 vs -23 is equivalent to a minor adjustment of your volume knob on your stereo. Nothing more nothing less. There is nothing magical about this. The only issue is that if some things are -18 and some are -23, then you'll lose the benefit of RG playback (which after all is to try to "equalize" the playback volume of different CDs so you don't have to change the volume knob on every song/album).

Also, you can point at your parent music directory and reapply RG to ALL your folders in one single batch. I have > 6,000 albums. But they are all under one parent directory
music\
music\artist
music\artist\album\tracks....

So If I point at the top directory (music) and run the RG DSP on it, it will handle all 6,000 albums at once. no problem.

1.. adding new tag will auto overwrite the old tag, no need to delete.
2. why change from 5 to 8. I can't recall the calculations, but the space saved would be minor. I looked at this once because I have so many CDs, but I calculated that the difference between 5 and 8 across even 6,000 CDs was a trivial amount of space.
3. going from any FLAC to any FLAC (compression change or anything) changes nothing about sound quality, etc. That's why these are called LOSSLESS files. Lossless=lossless=lossless, etc.

If I were you, I would continue ripping files using -23 RG (since all is at that level already), not redo my exisiting files, and maybe start using -8 compression going forward. But no need to convert all your old rips to -8.

Thanks for the help, you're right, 5 to 8 is useless. I figured it would make a significant difference, but after looking at some crunched numbers online it seems like after hundreds of albums ripped it only makes enough room for ~1 extra CD.

I will however change the tags, I know that it is only loudness, but for instance with my iPod I need to crank the volume up to beyond 3/4th before I get to a good listening volume, and cranking it fully maxed isn't even that uncomfortable to listen to. Not that I want to blow my ears out, but I figure the extra ability to go louder may be a good thing especially since changing the tags is such an easy process.

[...]
So -23LKFS is not suited to consumer storage systems. People here have reported that subjective tests have shown that -18 LKFS seems to work well. That is 6 dB hotter, twice the voltage, somewhat less than twice as loud (10 dB). Based on my experience, this is probably somewhat low for most compressed pop material. but it does give you more headroom for wide dynamic range material (18 dB if the analog part of your system can handle full level digital signals), and is probably a good compromise between matching levels with other sources and maintaining improved audio quality.

I've rambled on quite a bit here, and I've left out the effects of pre-emphasis (which impacts FM and analog TV, records, and analog tape formats, but not so much digital) and analog tape saturation, which essentially provides an automatic and a typically pleasing peak level limiter.

If you read through all of this, I hope you get a better understanding of why -23 LKFS doesn't work in the consumer world, and the history of audio levels.

Wow, thanks a ton for the excellent explanation, I understand it much better now. I was on the fence about changing the tags before but I will definitely change to -18 now and rip with -18 from here on out, thanks again!

Hey Eleven13,



the meaning of the yellow dots and yellow dot is explained in the toolbar of the program - there is a legend.
If you keep your folder structure or not, depends on how you configure the conversion. You can keep your structure, but you don't have too.


Dat Ei

Thanks for the information, I didn't realize there was a legend so I'll check that out. I'm happy to hear that I can keep my folder structure because that'll make it much easier to simply cut/paste the files back onto my iPod when I make the switch.

[NinthWave]

Re: Replay Gain question?

I've rambled on quite a bit here, and I've left out the effects of pre-emphasis (which impacts FM and analog TV, records, and analog tape formats, but not so much digital) and analog tape saturation, which essentially provides an automatic and a typically pleasing peak level limiter.

If you read through all of this, I hope you get a better understanding of why -23 LKFS doesn't work in the consumer world, and the history of audio levels.

Long after, I have read all of it and want to thank you for explaining all the backgroung.

[NinthWave]

Re: Replay Gain question?

4-5 years ago, I ripped all my CD's with replay gain enabled. We are talking about 20K tunes.

For a few months now, I have been buying FLAC on the internet and it's just recently that I figured that bought music doesn't have replaygain tags.

Questions
1) Since everything is already FLAC, If I "batchrip" my root music directory and apply no compression and just add "replay gain", what are the odds of corrupting the files if just adding IDTags?

2) Can someone re-explain why I would choose either "album", "track" or both?

Thanks

Nicolas

[garym]

Re: Replay Gain question?

4-5 years ago, I ripped all my CD's with replay gain enabled. We are talking about 20K tunes.

For a few months now, I have been buying FLAC on the internet and it's just recently that I figured that bought music doesn't have replaygain tags.

Questions
1) Since everything is already FLAC, If I "batchrip" my root music directory and apply no compression and just add "replay gain", what are the odds of corrupting the files if just adding IDTags?

2) Can someone re-explain why I would choose either "album", "track" or both?

Thanks

Nicolas

1. Odds of corrupting files by adding tags is essentially zero. I do it all the time. If paranoid, run a conversion to TEST CONVERSION afterwards and it will tell you if files are corrupted.
2. Add both. They are just tags. My server/player uses album gain when Im playing an entire album (preserving the intra-album track relative volume) but track gain when playing a mixture of tracks from different albums.