Using Logic Pro X, I've recently completed a mix and bounced it to a single audio file to be mastered in a separate project. I checked the short-term loudness level, which peaked at -15LUFS, with a True Peak reading of +0.1dBTP. My understanding of the mastering process for streaming suggests I should apply a 'brick wall' limiter set to ensure true peaks no higher than 0dBTP (although I've been using -0.2dBTP). The song is intended principally to be streamed and I don't want to compress the mix unnecessarily and lose any quality or dynamic range. Based on this goal, it would seem that I need to employ a limiter only to ensure the mix does not have a true peak higher than 0dBTP and to ensure the mix does not exceed standard LUFS level guidelines (eg. -14LUFS is commonly used). Also, if the loudness sits below -14LUFS, (let's say a hypothetical -21LUFS), can I simply rely on the streaming services to apply their normalisation process to bring up the level when the song is streamed?
Please note, I'm asking this question only from the perspective of loudness and not other mastering benefits (eg. phase issues, compatibility with other mixes on an album). All guidance and recommendations would be greatly appreciated.
SOS Forum post
SOS Technical Editor Hugh Robjohns replies: Different streaming services use different loudness targets, and unfortunately almost none currently employ the ITU-R 1770 standard. However, in general the integrated loudness value, measured with an ITU-standard meter, should be around -14LUFS, and the short-term level shouldn't peak higher than -9LUFS. So you're probably in the right ball–park for loudness... but you aren't for the True Peak level! The True Peak value should never, ever, be higher than -1dBTP — and there is a valid argument to keep it below -3dBTP. Note that this means either 'minus one' or 'minus three' decibels, not 'minus point-one'!
In general the integrated loudness value, measured with an ITU-standard meter, should be around -14LUFS, and the short-term level shouldn't peak higher than -9LUFS.
The reason I state those figures is that although the True Peak metering format is a lot more accurate than conventional sample-peak meters, total accuracy requires a much higher oversampling rate than is practical. Most True Peak meters use 4x oversampling, which has a potential error margin of 0.6dB, so the ITU has ruled that the maximum acceptable True Peak measurement should be no higher than -1dBTP. That way, even if the meter under-reads the audio by 0.6dB, the actual reconstructed waveform will still be 0.4dB below converter clipping. When using lossy data codec systems, though, like MP3 or AAC, the True Peak level should be kept below -3dBTP to ensure the codec has sufficient headroom to process the signal without internal transient clipping.
As for the issue of raising quiet tracks, some streaming services do and some don't — so check with the specific services you want to use. But all will turn down tracks which are louder than their own specific target level.