The No.1 Website for Pro Audio
More a musing than a moan... stream normalizing
Old 26th November 2019
  #1
More a musing than a moan... stream normalizing

So, for much of the last two decades I've been hoping for the widespread adoption of a pre-indexed average-loudness-leveling system. (Something like the Replay Gain system.)

You'd want to be able to turn it off for playing a whole album straight through or turn it on to even up the levels in a mix of tracks from different albums, sources, and/or eras).

Many years ago now, I checked out the 'normalizing' in Spotify. I'm not sure how closely I really tested it out, but I got the impression it was a relatively simple compressor/limiter option. Meanwhile, the stream subscription I did use had no normalization/leveling options in the desktop version.

But since the streamer (Google Play Music) I've been using is going to be scuttled soon (supposedly in order to move subscribers over to YouTube Music -- def not my thing) -- and since those nice hippies at Amazon (whose Prime delivery service is a bit of a supply line for me) rolled out a new lossless tier to their own stream service (and while it's got a few rough spots I've got into a groove with it), I decided to jump ship.

Meanwhile, one of my old bandmates decided he needed to comp me in to his own Tidal lossless family plan (I've always thought of him like a father, though he's a decade younger than me ) so we could exchange playlists -- so it's a real embarrassment of lossless riches.

ANYHOW, back to the point... both Tidal and Amazon have a 'normalization' option. I tried it out and it sounded OK.

I couldn't hear any pumping/breathing/etc and I started wondering if maybe the stream companies had implemented the kind of indexed level-setting (not compression/limiting) I'd been on about for so long.

While I only tested one track off one album (via bitwise stream rips), I did test the streams from both players, Amazon's dedicated desktop executable on the one hand and Tidal's browser based player on the other. Both lossless. Recorded with and without player normalization set on.

The results were virtually identical, well within any sensible MOE for the testing method, an RMS reduction of 4.9 dB in one and 4.88 dB in the other. Here's what I wrote in response to someone on reddit...
Quote:
[An online acquaintance who knows his consumer sound] suggested that the streamers currently use straight level reduction rather than leveling/compression.

Since I'm currently on both Amazon HD and Tidal Lossless, I decided to run a test, doing stream rips of the same track from both, with and without their respective normalization features turned on.

When I analyzed the results, there was a 4.88 dB reduction overall in the track on Tidal and a 4.9 dB reduction in Amazon (probably within the measurement and calculation MOE). The RMS average for both was unchanged (save for overall level).

When I then increased the gain by 4.9 db on the normalized Amazon track and compared it with the non-normalized track in ABX comparison, I ended up giving up during the sighted/training phase because I simply couldn't hear a diff, either cutting back and forth or zeroing in on likely 'tell' sounds.

So, it would appear that the normalization in both is a) very similar or identical, perhaps based on the same volume index) and b) seemingly transparent.

But, of course, that's just one test; perhaps someone else will replicate it.
That might sound like a bit of effort, but I figured if I didn't go to the mats, my imagination would be working on me and I'd be sitting there worrying I was messing up the sound.

By the way, while the Amazon desktop app now allows you to access your local files -- it doesn't apply the normalization to them (or let you create playlists with them or a mix of streamed and local; hoping they fix that).


SO... anyhow... since these days I mostly seem to listen to an ever-changing eclectic mix of tracks, I've been listening with the 'normalization' on... and it seems to mostly work quite well. Every now and then there's a streamed track that's a little too loud or too quiet for my notion of how it should work, but mostly, it's slick.

BUT... one thing I have sort of noticed is that, listening to an evened up mix, songs don't 'stick out' in my consciousness or grab my attention as readily.

I mean, I'm GLAD to not have to race across the room to hit the volume (though I'm usually in front of the computer reading, anyhow) but it also means I'm not continually adjusting the volume control -- and, annoying as that was, it was also an act of physical engagement that took me back to the 3DW for a few moments and made me think about what I'd been listening to...


tl;dr: I think I need to get out more.

Last edited by theblue1; 27th November 2019 at 03:46 AM..
Old 26th November 2019
  #2
Lives for gear
 

I've wondered about this too. In the Spotify preferences it states the following:

For: Premium
Set a baseline volume level for all songs for a more consistent listening experience, or disable it to hear the songs at the level they were originally recorded.
Click the arrow in the top-right corner and select Settings.
Under Music Quality, switch Normalize volume on , or off .
You can also adjust the normalized volume level for your environment. Next to Volume level, choose from these options:
Loud - Handy for when you’re in a noisy environment, or if you just like your music loud! You may notice less dynamics in the levels though.
Normal – This is the default setting.
Quiet - This is best for when you’re in a quieter environment. You’ll notice more dynamic levels
.


[my italics and bold]

Which seems to suggest that they do apply some sort of compression at 'normal' and more compression at 'loud' setting? They don't give any more detail Would be nice to know.

Last edited by tedmanzie; 27th November 2019 at 08:40 AM.. Reason: Made it clear this was Spotify
Old 27th November 2019
  #3
Quote:
Originally Posted by tedmanzie View Post
I've wondered about this too. In the preferences it states the following:

For: Premium
Set a baseline volume level for all songs for a more consistent listening experience, or disable it to hear the songs at the level they were originally recorded.
Click the arrow in the top-right corner and select Settings.
Under Music Quality, switch Normalize volume on , or off .
You can also adjust the normalized volume level for your environment. Next to Volume level, choose from these options:
Loud - Handy for when you’re in a noisy environment, or if you just like your music loud! You may notice less dynamics in the levels though.
Normal – This is the default setting.
Quiet - This is best for when you’re in a quieter environment. You’ll notice more dynamic levels
.


[my italics and bold]

Which seems to suggest that they do apply some sort of compression at 'normal' and more compression at 'loud' setting? They don't give any more detail Would be nice to know.
That's from Spotify, yeah? It sounds kind of familiar.

And, you bet, it sure sounds like 'AVC' style compression/limiting, to me, too; just like in your old VCR. (Although you gotta give them credit for giving the user different compression curves.) Handy to HAVE, for sure, on a mobile or just if you like to, say, segue from Amadeus Quartet to Skrillex, but not the way most of us want to listen to hi fi music.


That's why I was pretty delighted to determine that both Tidal and Amazon's lossless tiers appear to use an indexed system (that presumably preprocesses and stores an average loudness index number for each track in their library and, then, if the 'normalization' is set on, raises or lowers the overall level accordingly).

It's the most transparent way to deliver evened-up loudness in a mixed source media scenario, basically equivalent to an output volume control in linearity/transparency. (Mind you, this is all informed assumption based on performing diagnostics on stream rips from the two services with normalization on and off.)

The fact that both Tidal and Amazon lossless reduced the 'normalized' volume of my test track by a virtually identical 4.88/4.9 dB while the RMS value for each track was identical before and after [adjusted for the reduced level] seemed to suggest that either they were buying their code libraries from the same place or that maybe there had been some sort of industry-wide adoption of an RG-like indexing system for use in streaming. (Silly me, THAT last would require music industry entities to work together for the good of all. And THAT ain't gonna happen. )

Whatever, it IS really nice to be able to listen without continually shooting my hand out to adjust the volume (or, worse, lunging across the room) -- and I'm pleasantly shocked that they actually gave us an optimal-fi solution.


EDIT: And, of course, even though the normalization seems as transparent as such a system can get, I always turn it off when listening to full albums (assuming they're properly mastered, which is not always the case, I've heard some amazing level jumps between [non-adjacent] songs on the same album -- happened more in vinyl days because of adjustments to level trying to accomodate circumference 'loss' issues as the needle nears the center; but some contemporary repackaging bottom-feeders ('Your dad's favorite hits all on one album!') are incredibly sloppy and or heavy-handed about source material quality, mastering, and, ugh, 'restoration.' I firmly believe some of these 'mastering engineers' working for fly by night repackagers may effectively be deaf. Or simply have NO idea what an album of music is supposed to sound like.

EDIT/UPDATE 2: I found this info -- from 2017 -- that appears to describe a compression scheme in Tidal as it was then -- but a compression scheme with variable settings. The Android app no longer has these variable settings, normalization is just on or off. That could mean that they changed to an indexing system in the two years since that was written -- or it could mean I'm misinterpreting the results of my experiment. I suspect (and hope) the former. I've been wrong before; still, I poked and prodded pretty good to come to my conclusions.

Last edited by theblue1; 28th November 2019 at 12:08 AM..
Old 8th December 2019
  #4
This Tape Op article from last year addresses changes in the stream world with regard to 'normalization' issues in relation to our practice as creators of recorded media:

https://tapeop.com/interviews/126/modern-loudness/

From the section on streaming:
Quote:
Apple and various streaming services also want to provide their listeners with a consistent volume, but it's impractical for them to demand that every mix turned in to them averages -23 LUFS. Instead, when tracks are uploaded, the loudness of each track is measured, then a playback gain offset is calculated for each track so that they all play back at similar volumes (usually -14 LUFS, because your headphone amp on your device isn't very powerful). If you upload a track that has been brickwall-limited to sound "louder," it's just going to be turned down. You might have crushed all the life out of it to make your track average -4 LUFS; but it will still play back at -14, and will now likely sound like garbage.
https://tapeop.com/interviews/126/modern-loudness/
Old 13th December 2019
  #5
Lives for gear
 
doom64's Avatar
Quote:
Originally Posted by theblue1 View Post
This Tape Op article from last year addresses changes in the stream world with regard to 'normalization' issues in relation to our practice as creators of recorded media:

https://tapeop.com/interviews/126/modern-loudness/

From the section on streaming:
https://tapeop.com/interviews/126/modern-loudness/
Excellent! It's amusing to hear my mixes compared to the top level competition. Gain staging is something not respected anymore. Loudness normalization is the best thing to happen to recorded music in the 2010s
Old 13th December 2019
  #6
Quote:
Originally Posted by doom64 View Post
Excellent! It's amusing to hear my mixes compared to the top level competition. Gain staging is something not respected anymore. Loudness normalization is the best thing to happen to recorded music in the 2010s
I'm kind of amazed a more or less standardized approach has emerged -- and that it's not just shoving everything through a program limiter/compressor.

(I do wish the Amazon player, which can access local files, and mix them with streams in the queue [but not include them in playlists, at least not yet] could apply the normalization to local files. I think I'm going to throw a Replay Gain (RG) encoded FLAC or MP3 in my Amazon player feeder folder and see if it makes a difference.)
Topic:
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Forum Jump
Forum Jump