[MLA] Normalizing on Tidal

[Grimm | Eelco]

Hi Kevin,

> Unfortunately all the ReplayGain implementations I've seen don't actually adjust the system volume; They just attenuate or amplify the player's output in accordance with the loudness measurement.

What I meant is that the controller app of the media player often has a volume control slider. This data is merged with the replaygain values and used for digital volume control right before the DAC of the media player. If there's a simple analog volume control behind the media player, this is of course the best you can get. In fact digital volume control often is of higher quality than analog volume control, so it is already quite optimal. In our LS1 system, we do not affect the audio in the media player but forward the 'volume + replaygain' data to the DSP in the speakers, so the DSP can work on unattenuated audio and the volume is set after the crossover filters, just before the DAC's. I agree that is rarely seen, unfortunately. But it still is a digital volume control. 

> Sound Check on portable devices works the same. There is no happiness for anyone here. With this approach, modern loud recordings are attenuated and play at a lower maximum volume when normalization is enabled. Amplification on older dynamic recordings introduces clipping or limiting at any playback level.

I now see what you mean: with norm-l completely implemented, the volume + replaygain level would never cause clipping, but just 'loose normalization'. With media players that simply merge replaygain with volume settings, this is only guaranteed if the analog gain after the media player dac is big enough to play at your preferred maximum level with the most dynamic track in your collection. All the other tracks would be attenuated always. But that's no problem. It is in fact what we advocate for mobile use as well: "please make sure there is enough gain to playback -23LUFS tracks at the max allowed Cenelec level". From then on norm-l would never go 'out of normalization', except for tracks that are lower in loudness than -23 LUFS. In case devices do not offer this headroom, of course the full norm-l approach including 'going out of normalization at high playback levels' is of course preferable (although a build in limiter would also perform nicely).

Cheers,
Eelco