I'm trying to master a track that is hard rock. Looking at the waveform it undulates somewhat. I keep getting two overs that I haven't been able to quell with a limiter while maintaining -14 LUFS, but the strange and complicating part is that these overs occur on the Yulean Loudness meter at a point in time that this between two of these undulations of the waveform, which are perhaps one second from crest to crest.
How is this possible? One would expect a peak dB reading on the loudness meter to correspond with a peak on the waveform being measured. Any thoughts?
Strange behaviour of Yulean Loudness Meter
Re: Strange behaviour of Yulean Loudness Meter
Even the momentary LUFS loudness meter is averaging over a 400ms period, so there is a delay before the loudest value appears after a big transient. And the sample windows overlap by 75%, so each window starts 100ms after the previous one.
You already have your instantaneous sample peak meter and true peak meter to tell you how big the transients are, and the instantaneous LUFS meter tells you how loud it sounds. Think of it as measuring the area under the waveform line. A very sharp transient with sample peak at -6dBFS is going to sound quieter than a very broad waveform peaking at -6dBFS.
E.g. a waveform with sample values either side of the peak (in dBFS) of -35, -34, -33, -12, -6, -13, -14, -15 and -17 is going to have more total energy and sound louder than one with values of -35, -34, -33, -12, -6, -20, -35, -40 and -58. But you don’t know that until after the peak has passed. So there has to be a delay in the ‘momentary’ LUFS value display.
So both sides of the waveform either side of the peak value are important in determining how much energy/how loud that wave sounds, so by the nature of the way it works, the LUFS result will always come after the peak sample value has passed.
You already have your instantaneous sample peak meter and true peak meter to tell you how big the transients are, and the instantaneous LUFS meter tells you how loud it sounds. Think of it as measuring the area under the waveform line. A very sharp transient with sample peak at -6dBFS is going to sound quieter than a very broad waveform peaking at -6dBFS.
E.g. a waveform with sample values either side of the peak (in dBFS) of -35, -34, -33, -12, -6, -13, -14, -15 and -17 is going to have more total energy and sound louder than one with values of -35, -34, -33, -12, -6, -20, -35, -40 and -58. But you don’t know that until after the peak has passed. So there has to be a delay in the ‘momentary’ LUFS value display.
So both sides of the waveform either side of the peak value are important in determining how much energy/how loud that wave sounds, so by the nature of the way it works, the LUFS result will always come after the peak sample value has passed.
Reliably fallible.
Re: Strange behaviour of Yulean Loudness Meter
Wow, that explains a lot! Excellent description! I've seen a lot of weird behavior like that using delays - but never managed to properly figure out what was going on because they weren't consistent. So you need to look behind the clipping point to see they causal waveform?
Adrian Manise
Faith in Absurdity
https://adrianmanise.bandcamp.com/
https://soundcloud.com/adrian-manise
A Hazelnut in every bite
Faith in Absurdity
https://adrianmanise.bandcamp.com/
https://soundcloud.com/adrian-manise
A Hazelnut in every bite
Re: Strange behaviour of Yulean Loudness Meter
For LUFS, yes. To fully interpret what’s going on you need to combine the LUFS reading info with the sample peak values, and the shape of the waveform around that peak reading.
A limiter may ‘cut’ the top off a sharp transient, but if you then use make-up gain to bring the signal level up again to a value that’s below the previous peak value, because the overall shape of the wave peak is now broader, it may well have a momentary LUFS value that’s higher than before.
A limiter may ‘cut’ the top off a sharp transient, but if you then use make-up gain to bring the signal level up again to a value that’s below the previous peak value, because the overall shape of the wave peak is now broader, it may well have a momentary LUFS value that’s higher than before.
Reliably fallible.
Re: Strange behaviour of Yulean Loudness Meter
Use a look-ahead limiter, or an oversampling limiter, and with a lower threshold. ...or do a manual edit to reduce the level of the offending transients.
The Integrated loudness indication is a long term averaging arrangement, so the peak Integrated Loudness value will always be after transients, not synchronous with them. Same applies to Momentary and Short term loudness readings, but to a lesser extent because they have shorter averaging windows.
-
- Hugh Robjohns
Moderator -
Posts: 43685 Joined: Fri Jul 25, 2003 12:00 am
Location: Worcestershire, UK
Contact:
Technical Editor, Sound On Sound...
(But generally posting my own personal views and not necessarily those of SOS, the company or the magazine!)
In my world, things get less strange when I read the manual...
(But generally posting my own personal views and not necessarily those of SOS, the company or the magazine!)
In my world, things get less strange when I read the manual...
Re: Strange behaviour of Yulean Loudness Meter
What peak level are you mastering to? If you’re getting overs it suggests you’re limiting to 0 dBFS which (some YouTube pundits notwithstanding) is not a good idea.
Aim for no more than -1 dBTP, ie limit using true peak levels.
Aim for no more than -1 dBTP, ie limit using true peak levels.