I was looking around on the internet and came across automation software for the Mac (looked at it out of curiosity, as I don’t have a Mac). In the feature list was something that I would welcome in mAirlist as well.
It had AutoGain. It’s description said:
Automatic calculation of a track's true volume
Now that would be absolutely great! Not just a normalize – which doesn’t calculate the track’s true volume – but a way to actually play every file at a similar volume level.
Would it be possible to build something like that into mAirlist?
mAirList supports Replay Gain, which is a similar technique. However, the calculation must be performed by a third-party software, e.g. foobar2000, and the Replay Gain tag must already be present when the file is imported into mAirListDB initially. (The Replay Gain information is transformed into mAirList’s Amplification setting and stored along with the other metadata in mAirListDB.)
Oh, and by the way - did it mention any details about how that algorithm is supposed to work? Because usually, most software that claims to have a feature like that is simply using normalization internally, or at most some sort of compression.
But I assume that with Replay Gain, as the tag needs to be present, it’s still a manual setting per file, rather than an automatic action?
That the calculation has to be done through a third-party program, like Foobar, doesn’t seem a big problem to me. However it does raise the question whether that will slow down the process and also WHEN does that calculation process take place?
Alas, the people who made the Mac software only listed it as a feature on their site. They didn’t reveal the algorhythm behind it. The site is: http://www.dj1800.com/benefits.php . However, normalize will never calculate the true volume of a file. It will just limit the whole file on the basis of the highest peak, to a pre-set setting. If that is what they mean by ‘autogain’ it’s pretty useless really. But if they actually mean calulating the file’s true volume so that all files play at exactly the same volume… that’s a different matter…
That’s an important subject you’re talking about. It would be great if mAirlist would have also an automatic RMS option for the database, especially music.
If you have a music rotation which includes dynamic music for example from the 70s and really hard limited tracks from today, it’s a hard game for an amateur to normalize them in a good way. Most of the people doesn’t know about RMS Level oder Replay Gain.
I’m realizing my database with Sound Forge Pro. It has a great RMS Level Normalizer. You can save presets and make job lists with a batch converter.
RMS level for music is at -16dB RMS. It’s important to have an eye on this. Then the sound processing isn’t that much difficult in the end.
Generally, the Replay Gain standard only defines the way the gain level for each track (or album) is stored in the various file tag formats (ID3, APE, Vorbis Comments etc.). But not how the gain is actually calculated.
For proper gain adjustment, RMS is probably the way to go. The RMS algorithm is pretty simple, I could just add a step to the file import routines that scans the file, calculates the RMS level, and adjusts the amplification setting so that the RMS level matches a particular value, say -16dB(fs?), as Michel suggests.
This would slow down the file import process significally though. Just like Auto Cue, but even worse, because for Auto Cue, one doesn’t need to scan the whole file, only the beginning and the end.
Another thing to keep in mind is peaks and possible clipping. The amplification calculated by the RMS algorithm can make single peaks louder than 0dB, which will result in clipping. This is ok if you’re using a (software) limiter before the audio signal reaches the sound card, but is a major problem if you don’t (which is the case for most mAirList users, unless you’re using a VST limiter and play directly into the encoder). Replay Gain addresses this problem by storing the peak value in the file tag as well, and a playout software that doesn’t have a limiter can reduce the gain accordingly. This is what the “Prevent clipping when using ReplayGain amplification” setting in mAirList is about.
Personally, I run ALL music tracks through mp3Gain, doing Track analysis and setting, to the default 89.0dB.*
However, I do normalise jingles, trails, promos, etc. This naturally makes them sound ‘louder’ than the music. For that rason, I personally would NOT want mAirList to start messing around with gain internally.
In my personal case, I run mp3Gain with ‘no undo,’ so changes I make are permanent and don’t rely on a ReplayGain tag in the file, so the changes will work in any software or hardware mp3 player.
(* Apparently, this is some weird measurement used principally in the movie industry. I suspect it’s the RMS A-weighted SPL at 5m, but I’m not sure about that or exactly what it means.)
I run all music through MP3Gain aswell, with a default of 93, but this does not measure or change the files true volume. A song from 1967 for example would still sound louder than a song from 2010 (Not all songs; it’s an example), regardsless of the MP3Gaining. In other words, MP3Gain gives you a guide, but not more than that.
If that ‘AutoGain’ can actually measure the TRUE volume of the file and adapt that upwards or downwards, that would be great, because then every file will have the same volume, whether it’s old or new, regardless.
Peakmeter doesn’t reflect the loudness a human beeing hears and feels. It’s just a technical info.
Maybe we could do some research on this, Torben. At the moment I’m grabbing all my music in FLAC format. I did not reached the CDs with big dynamics. If they came along I will test a really dynamic and a hard limited track to suggest some RMS Level which should be fine in most cases.
Nevertheless Sound Forge RMS Normalizer works great, you can set up a dynamic processor which stops clipping at the end of the process.
Correct. In the ‘old days,’ peak levels were more important than today, becuase it was easy to literally damage transmission equipment (and/or transmitter valves/tubes) with a signal that was ‘over peak level.’
That said, the BBC standard is to mix so that (on a PPM) music averages 4 and peaks at 5, and speech averages 5 and peaks at 6.
