We Won The Loudness Wars?
I was asked to write a guest blog post for local Seattle Live Certified trainer Isaac Cotec, thought I would share since it might pertain to some people here. Or at least explain what the fuss is about for those that care:
http://subaqueousmusic.com/production-articles/sound-design-tech/223-we-won-the-loudness-wars
I’ll bite.
It’s an interesting subject, and I wasn’t aware of some details but there’s a few discrepancies.
I get the whole normalization happening with streaming services (pandora, spotify, etc.) but doesn’t the fact that 90% of all the music they are streaming still adhere to current volume trends sort of ignore the effort (if there is one)?
You’re a mastering engineer, and I’m not going to act like I know more than I do; but don’t you think this is a bit of a sidestep? I mean, our work as engineers has always been to make lasting work that sounds intentional at whatever volume, on whatever listening system. To then go: “No wait, this is going to be streamed, it needs a special version for that.”, doesn’t that seem counteractive?
“The new 1770 standard only allows for raising and lowering of the overall song volume, the audio content itself is not being altered at all.”
So basically, peak normalization. No dynamic changes, just volume.
(And let’s be real, everything these days peaks at 0dB, or close, so we’re turning down volume most likely)
“In short, more and more ways we consume audio are being set up to automatically adjust the playback volume in order to achieve consistent loudness.”
This sounds like compression, or RMS normalization. Which begs the question, what are we talking about here?
If the 1770 standard does peak normalization, as you first stated, then I don’t believe there’s a need for concern. If 1770 is RMS, then OK, time to panic. If the RMS normalization you’re speaking of is a separate entity like the kind that we already have a la: iTunes, Youtube, Streaming, etc. then how is this news? Any audiophile worth their salt knows to turn normalization off on their playback devices, so where’s the concern?
I personally find it quite scary that your clients have requested 2 separate masters because of this (to me still not altogether clear) concern. Again, in my opinion, a master is a master. Just like how I don’t expect a master to compress the hell out of dynamics because it needs to sound loud, I don’t expect it to sound thin because it needs more dynamics at low volume. It just doesn’t make much sense to me; not to mention, they’re complete opposites!
Now for the whole “loudness wars”, I don’t want to write a whole thesis, so I’ll just ask, how is this related? It sounds like this 1770 standard, is just that, a volume standard. How will this derail us from current dynamic/volume practices? It’s already factual based on psycho-acoustic analysis and the fletcher munson curve that we as human beings perceive sound non-linearly; which as an effect makes us believe that louder sounds better.
Whether we take this practice too far by over compression, is debatable, but you can’t argue against peak volume consistency, with the ceiling being 0dB.
To wrap this up, though I realize I probably started a few discussions, I’m confused by the concern. I’d still like to believe that the 1 final master coming out of the studio should be the reference point, not to over complicate things.
It’s an interesting subject, and I wasn’t aware of some details but there’s a few discrepancies.
I get the whole normalization happening with streaming services (pandora, spotify, etc.) but doesn’t the fact that 90% of all the music they are streaming still adhere to current volume trends sort of ignore the effort (if there is one)?
>>> At the moment, sure. But that’s going to change quickly once people realize that mastering to the current volume trend ends up sounding worse than leaving a bit more dynamics in. <<<
You’re a mastering engineer, and I’m not going to act like I know more than I do; but don’t you think this is a bit of a sidestep? I mean, our work as engineers has always been to make lasting work that sounds intentional at whatever volume, on whatever listening system. To then go: “No wait, this is going to be streamed, it needs a special version for that.”, doesn’t that seem counteractive?
>>> It’s one more step in the process FOR SOME PEOPLE, but I don’t think it’s a sidestep. It’s more like an intermediary step for those people who are concerned with how their music will sound in both systems. Most artists will probably just ignore it until they can’t any longer, but some of my clients clearly hear the difference and want a version that sounds the best for both delivery methods. <<<
“The new 1770 standard only allows for raising and lowering of the overall song volume, the audio content itself is not being altered at all.”
So basically, peak normalization. No dynamic changes, just volume.
(And let’s be real, everything these days peaks at 0dB, or close, so we’re turning down volume most likely)
>>> Correct, it’s peak normalization. <<<
“In short, more and more ways we consume audio are being set up to automatically adjust the playback volume in order to achieve consistent loudness.”
This sounds like compression, or RMS normalization. Which begs the question, what are we talking about here?
If the 1770 standard does peak normalization, as you first stated, then I don’t believe there’s a need for concern. If 1770 is RMS, then OK, time to panic. If the RMS normalization you’re speaking of is a separate entity like the kind that we already have a la: iTunes, Youtube, Streaming, etc. then how is this news? Any audiophile worth their salt knows to turn normalization off on their playback devices, so where’s the concern?
>>> The measuring of the loudness of the files is the only part of this that involves any sort of RMS aspect. The track is analyzed using the guidelines in the spec to determine it’s loudness based on methods that approximate the way the human ear works. It’s assigned an RMS value if you will, and then all tracks are scaled with peak normalization so that they all playback at the same perceived loudness. <<<
I personally find it quite scary that your clients have requested 2 separate masters because of this (to me still not altogether clear) concern. Again, in my opinion, a master is a master. Just like how I don’t expect a master to compress the hell out of dynamics because it needs to sound loud, I don’t expect it to sound thin because it needs more dynamics at low volume. It just doesn’t make much sense to me; not to mention, they’re complete opposites!
>>> I have some clients now who ask for specific Soundcloud versions too, knowing their steaming encoders do weird things to bass heavy tracks sometimes. Or people who want alternated Mastered For iTunes versions. At the moment I don’t see wanting a seperate 1770 version as any different. Some people will care enough to bother with the versions, some won’t until they have no choice.
The concern here is that a track mastered to current loudness levels likely has a LOT of limiting and compression done to it, and often not for artistic reasons, but in order to “compete” with other tracks while browsing (say Beatport). When you play back these tracks in a peak normalized system, instead of being louder than other tracks, they end up quieter and sounding flat by comparison. Because they are flat in comparison, they’ve had a lot more dynamic range removed already. <<<
Now for the whole “loudness wars”, I don’t want to write a whole thesis, so I’ll just ask, how is this related? It sounds like this 1770 standard, is just that, a volume standard. How will this derail us from current dynamic/volume practices? It’s already factual based on psycho-acoustic analysis and the fletcher munson curve that we as human beings perceive sound non-linearly; which as an effect makes us believe that louder sounds better.
Whether we take this practice too far by over compression, is debatable, but you can’t argue against peak volume consistency, with the ceiling being 0dB.
>>> I think I answered that above. <<<
To wrap this up, though I realize I probably started a few discussions, I’m confused by the concern. I’d still like to believe that the 1 final master coming out of the studio should be the reference point, not to over complicate things.
>>> In an ideal world, that’s definitely the case. And I think the shift to this new way of working will lead us to that faster than anything else too. Right now there is no consistancy in how albums and songs from different artists are played among various media outlets. This will fix that issue if enough get people onboard, one master should translate the same in terms of loudness compared to other songs by other artists. It’s an attempt to solve an issue now, not create a new one in the future.
In the short term, yes it makes a little more work with alternate versions (for those that actually care), but in the long term this should give the artist a LOT more flexibility in how they mix and master their tracks. I think it’s important to realize there’s very little differences in the versions too. It mainly eliminates the need to over limit for volume’s sake alone, allowing the artist to focus on more creative uses for buss processing. These aren’t drastically different versions, or a requires new way of writing a song. Most people will just limit their mixdowns less and everything else will be the same. <<<
I made some examples to put this in perspective. Don’t analyse or anything first, just play the first one and turn up your monitors until it sounds nice and loud like you’d normally listen to your music:
Now without touching the volume, play the second one:
Both of these were peak normalized to read -20 LUFS (Loudness Units Full Scale), just like they would be in the ITU-R 1770 spec we’re talking about. The actual number doesn’t matter, just the fact they they are both set to the same LUFS value.
At first listen, not too much different right? So big sigh of relief people, we’re not talking a huge change here. When you listen a little closer, hopefully you can hear that the second file is a little less punchy, just as loud but not quite the same impact from the sounds. A tiny bit distorted too, a side-effect often times of the volume levels we master to.
So there’s definitely an audible difference, but it’s not huge. At least in this example, I’ve heard worse with examples like these. Like I said, some people won’t care given that it’s not a huge difference. For those that do, likely it means you’re just going to be mixing and mastering the same as before, you just won’t use a peak limiter at the end. That’s it.
Same EQ and colorful compression if you want, but no need to slam it to make it loud, as you can hear it’ll be just as loud as one you do limit. So why bother? Aesthetic reasons perhaps, but it won’t be a knee jerk reaction that you apply to every song as a matter of course like it is now.
Does that help?
Just to make sure I understand things right: LUFS is just a concept that is kind of like RMS?
Because this is what I was thinking at first:
Let’s say we have two tracks, both peaking at 0 dB FS.
Now let’s one has a RMS level of -10 dB FS, and the other one has an RMS level of -5 dB FS.
In order to bring both tunes to the same perceived loudness (-> RMS level) you’d have to bring the second track down by 5 dB, so they both would now have an RMS level of -10 dB.
Now both would be playing at the same perceived volume, which means you could listen to both track in a row and not touch the volume control of your stereo without any noticeable change in playback volume.
However… the peaks of the track that had the original RMS of -10 dB FS would be much higher than the peaks of the other track. Which means that, even though the track seems to be at a reasonable volume, you could be be clipping your stereo with the peaks, since you’ll be adjusting levels by perceived volumes…
But I guess that problem only exists in my over-simplified understanding of this matter?
Hey Tarekith, once again thank you for sharing your insides and taking the time to respond to the people of our community!
It’s very much appreciated!
Correct, LUFS is more or less the same concept as RMS, both measure the average loudness. RMS literally stands for Root Means Squared, which is a statistical way of determining an average of a set of numbers. Unfortunately, RMS doesn’t really correspond to how our ears determine loudness. Which is why all this LUFS stuff was invented in the first place. 
Your example is more or less correct in how all this will work, except that there’s headroom built in to the spec so that peaking is never really an issue. The examples I posted above are probably a little closer to what the real volume of all this will be, which is why most people probably had to turn up their monitors a little bit to get them at a comfortable volume.
If I would have realized it was going to be this long, I would have suggested it for the blog Happy to help though, good karma and all that