Does using this method of measuring LUFS make the SWS method unnecessary?

sjs94704 · 07-14-2018, 10:29 AM

I was watching the following video by Kenny Gioia regarding setting LUFS:

https://www.youtube.com/watch?v=0wPnlND92QU

I'm just wanting to confirm that if I do it as he instructs that using the SWS extension for loudness is not necessary. Is that right?

ashcat_lt · 07-16-2018, 10:07 AM

I know you don't actually want to hear from me, but I waited a day and nobody else tried to help, so...

Kenny's method is not exactly ideal for a couple reasons:

1) RMS is not the same as LUFS. There are important differences in the way they are calculated. It's often "close enough" with normalish music as a source, and will probably work fine if you're just "eyeballing for YouTube", but if you really have to meet a LUFS standard, you should really measure LUFS;

B) I'm pretty sure all of those streaming services are looking at the "Integrated Loudness", but the method in the vid can only measure "Short Term". And yes, Kenny seems to be cross-comparing reaper's short term against the plugin's integrated LUFS. This matters most in very dynamic material. On rock song that just gets loud and stays loud, the two numbers will probably close, but if you've got a gentle intro and mellow verse and then a big climactic chorus, what you see on reaper's meter will depend on what part of the song you're playing. Worse, the integrated value can't be expected to match any of your short-term readings except by accident. Like, if you measure the loudest point, you might turn it down more than it needs to be, but if you measure the quietest part, YouTube will end up turning it down for you.

But the ultimate answer is (as almost always) that it depends on what you're trying to do. If you're really trying to play the Price Is Right (as close as possible without going over) in order to deliver a product to tightly specified standards - like a TV show for broadcast, use SWS or some other method to get real integrated LUFS.

If you're really just trying to get it "right" for streaming media, Kenny's method will definitely get you close enough. I use a very similar method (with a slightly larger RMS time) for most of the things I do. Its the way I've always done since my first analog recordings 30 years ago.

If OTOH, you're actually trying to extrapolate from this to "gain staging" on individual items or tracks, the answer is pretty much "yeah, that's exactly what those meters are for".

I really do hope that some of this helps.

karbomusic · 07-16-2018, 10:20 AM

Pretty much everything aschat said. If you are wanting to say "nail it" for something like YT using LUFS integrated will work. I've been able to get YT to react (or rather not react) within .2 dB or so by meeting the number at the time - let's pretend that was -15 LUFS.

Keep in mind some of the services only turn audio down, not up but again, what you are looking to accomplish matters here. Integrated is what all the services use to my knowledge FWIW.

Also on an academic side note, the LUFS algo ignores the bottom 10% and top 5% of the audio levels (add a big IIRC to that but I'm pretty sure I saw that when reading the spec last year).

ashcat_lt · 07-16-2018, 10:31 AM

@karbo - yeah, that's some of the "important differences" that I glossed over. I've never bothered to read the spec, but from things I've read I know that you can't "fool" it by having a long period of silence before/after your super loud mess. I hadn't ever heard that it ignores the top, though. That almost doesn't make sense to me, but if it's true, it kind of changes the game.

nofish · 07-16-2018, 11:25 AM

For the nerds among us, here are the measuremnt specs

https://www.itu.int/rec/R-REC-BS.1770-4-201510-I/en

karbomusic · 07-16-2018, 11:45 AM

Quote:

Originally Posted by ashcat_lt

I hadn't ever heard that it ignores the top, though. That almost doesn't make sense to me, but if it's true, it kind of changes the game.

Yea, and I can't say with certainty I read that in the spec itself, I took a Saturday last year and it was "LUFS Day" for me so... However, here is a reference from Steinberg that mentions it but still need to see if the spec agrees *and* if I even understood correctly:

Quote:

The top 10 % of the quiet blocks and the top 5 % of the loud blocks are excluded from the final analysis. The calculated loudness range is the ratio between the loudest and quietest remaining audio blocks. This measurement helps you to decide if and how much compression or expansion you can or should apply to the audio.

https://steinberg.help/cubase_pro_ar...urement_c.html