iTunes Radio Normalizes Playback Volume with Sound Check

I’ve seen some reports online suggesting that iTunes Radio uses Sound Check, Apple’s volume-normalizing feature, to keep music streamed on iTunes Radio at a consistent relative volume. Since these claims present no data or examples to back them up, I decided to have a listen and see if I could determine whether iTunes Radio is, indeed, using Sound Check.

For those unfamiliar with Sound Check, it’s a feature in iTunes that normalizes the volume when you play music through the app. You can activate it be checking a setting in iTunes’ Playback preferences, and it’s also available on iOS, in Settings > Music > Sound Check, and on the Apple TV, in Settings > Audio & Video > Sound Check.

002.png

If you turn Sound Check on, iTunes examines all your music to determine how much it deviates from a norm:

examine.png

When it’s finished, you can see, for each track, whether the music’s volume needs to be increased or decreased when using Sound Check; this song needs to be reduced by 5.0 dB if Sound Check is active:

volume.png

When you play music or podcasts from your iTunes library, iTunes takes into account those volume changes, playing music softer or louder as required. Volume differences also show up for audiobooks,

You can notice Sound Check if you play, say, an opera, where tracks segue into each other. You’ll notice obvious changes in volume from track to track, if their average volume is very different.

So what is iTunes Radio doing? Quite simply, it’s lowering the volume of loud songs, sometimes drastically. In order to show this, I recorded some iTunes Radio streams, and compared them with previews from the iTunes Store, where Sound Check is certainly not turned on. I then compared some CD rips and purchased music I have with iTunes Radio streams. In all of these cases, iTunes’ volume was the same.

I’ll just show you one example; Miley Cyrus’ Wrecking Ball, an ode to overly loud production techniques. The first part of the waveform below is the chorus of the song, from an iTunes Radio stream; the second part is the middle of the chorus, then the quieter part, from an iTunes Store preview:

wrecking.png

As you can see, the volume is drastically higher in the iTunes Store preview. It’s also a bit more flexible; look at the arrows, which show the same points in the song. That little notch is where Miley’s voice pauses: “Yea you … Wre-e-eck me.” You can see in the second sample that the waveform is less flat than in the iTunes Radio sample.

None of this is definitive. You could argue that it’s really not clear if I’m comparing the same things. That there’s no way to know exactly what the iTunes Radio stream contains; whether it’s the same format as iTunes Store previews, or the same format as the music in my iTunes library that I compared (AAC 256 kbps). But listening to iTunes Radio, it is very obvious when certain songs come on – again, the current hits by Lady Gaga and Miley Cyrus – that they are muted.

But it’s also obvious in tracks such as The Allmann Brothers’ Ramblin’ Man, Bob Weir’s Black Throated Wind, and several others that I compared. The volume change isn’t only for the latest over-produced hits; it’s for anything that has a volume that is a bit higher than normal.

iTunes Radio also raises volume of softer music. Here’s a piece by Steve Reich, the second movement, Slow, of his Double Sextet. The first part is from an iTunes Radio stream, the second is from my iTunes library, a file that I bought from the iTunes Store. With Sound Check on, iTunes says that this track needs a 6.0 dB boost, and you can see it well.

reich.png

The arrow highlights the louder first chord of the movement. The first one is the iTunes Radio stream, the second the beginning of me playing my own file. There is a marked difference both in the volume of that chord, and in the entire playback.

Given both anecdotal evidence (loud songs by certain artists sound softer on iTunes Radio), and the above tests, it’s pretty clear that iTunes Radio is using Sound Check. Is this a good or a bad thing? If you like the sound of Miley Cyrus and Lady Gaga, you’ll probably be turning up the volume when their songs come on. You’ll also find that quieter music is played a bit louder.

I don’t use Sound Check because I’ve found that some music gets distorted if Sound Check alters its volume too much. But this is radio; I’m less likely to care exactly how something should sound.

On the other hand, Apple is clearly saying that overly loud music doesn’t have a place on iTunes Radio. Will music suddenly become less loud? I doubt it; iTunes Radio doesn’t have the power to change the way music is produced. However, it’s possible that some music producers will take note of this, and some may just reconsider how much volume they’re putting into their music.

6 thoughts on “iTunes Radio Normalizes Playback Volume with Sound Check

  1. Step by step, Apple is getting it together when it comes to improving sound quality. Increasing data rates, improving AAC encoding, Mastered for iTunes, Sound Check . . .

    iTunes Radio may not yet (and may never) “have the power to change the way music is produced” but iTunes itself might. It’s the dominant platform. Artists often check their product in iTunes during production, worry that their music is not loud enough in comparison and send it back to the mastering engineer to squash it even further. This is making a mess of the sound of music.

    What if Sound Check were turned on by default? Suddenly the music is loudness normalized and it’s much easier to hear the extraordinary amount of distortion in those over compressed/limited tunes. This COULD change the way music is produced. And it would be a good thing.

    The other word out very recently (I’ve not checked this myself as I use a Squeezebox to play my iTunes library) is that Apple have recently more or less properly implemented Sound Check for Album playback. If you are in shuffle mode or playlist mode “songs” are normalized individually. If you are in Album view it will play with the Album normalized to others, i.e. it will respect the track to track levels of the original mastering of the album. This is a great news. You should be able to hit play at the beginning of the opera, set your playback volume and just enjoy the music with no jarring level jumps.

    The implementation of Sound Check is improving. I no longer see any reason to turn it off – other than to demonstrate its usefulness.

    • Sound check works fairly well at reducing volume; it works less well at increasing volume. If I play a soft piano recording with sound check on, iTunes’ volume increase almost always makes certain passages audibly distorted.

  2. For me, soundcheck is awful; 16bits is already a tight space for dynamics. Throw in compression used for artistic reasons, then throw in some more compression at the mastering phase that’s needed to bring things to a volume necessary to play nicely on a wide range of consumer electronics and the radio, and you have almost NO dynamics left. Take a song that’s been reduced by 5db, then turn the volume up on your amp by 5db. It sounds like crap. It’d be better to use loudness relative to the last couple of songs played, and make a decision on the output from there. The problem isn’t ‘loudness’ or how loud music is being produced. It’s the tight space that engineers are working with to begin with. iTunes policing the issue isn’t going to help, it’s just skirting the bigger issue. The only solution I can see would be to first standardize on 24 bits, then start using higher quality data compression formats when converting to whatever format you store your music in. 16 bits and CDs should be hucked out the window, never to return. There’s just not enough dynamic range.

    • I am all for 24 bits. No one working in professional audio – at least not in music – has been working with less than 24 bits of resolution (in terms of the capability of storing data or processing it) in the past 10 years or more. (We’ll set aside the fact that it’s physically impossible to get 24 bits of resolution out of an ADC for the moment). So there are almost no practical restrictions on dynamic range in the professional domain.

      MFiT requires that a 24 bit master be used for delivery.

      But I don’t agree that 16 bits is “just not enough dynamic range” when it comes to consumer delivery. We’re talking about 96 dB here and that’s ignoring the fact that proper noise shaping can give an effective dynamic range well beyond that.

      “It’d be better to use loudness relative to the last couple of songs played, and make a decision on the output from there.” Sorry. I don’t understand how this would work nor how it would help the situation.

      Loudness has very definitely been a problem but the tide is turning thanks to technologies such as Sound Check (not the best of the technologies but widely deployed!) there is less of a perceived “need” to take a 96 dB capable carrier and reduce it to 3 dB of dynamic range.

      http://www.tcelectronic.com/loudness/loudness-explained/

      • 96db minus your noise floor. It’s not that wide a range in my opinion. Aside from that,I think you missed my point: take a 16 bit file and reduce the level by 20db or so (to exaggerate what Apple is doing) now turn your amp up 20db. What you’re listening to will sound like mud. There just isn’t enough information left in the wave form to give you an accurate representation of the music. If you do the same with a 24 bit file, there will be. And if that resolution is available, compression and loudness will become less of an issue, because in some cases audio will not need to be compressed in order to give a similar effect of articulation. Reducing things by a set number of decibels does not address the issue without reducing the quality of the music. 24 bits would allow engineers to address the issue without ruining their sound.

        • I retract my statement; 16 bits is beyond the range of human hearing. I feel pretty silly for arguing now. After a lot of listening and reading, I think you’re absolutely right. There was an article in particular at xiph.org someone sent to me titled “24/192 Music Downloads are Very Silly Indeed” that, I admit, I was fully ready not to hear because of the title, but it made a lot of good points.

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.