How loud is too loud ? When DR values just aren’t enough

It’s almost time for Dynamic Range Day again, and as usual I’m getting some questions.

One of the most common is – “So you say you’re against the Loudness Wars, but you give an award to albums that measure only DR8 – why don’t you ask for more dynamics than that ?”

In other words, shouldn’t we be campaigning for far more dynamic range, for higher DR values – DR14, say, like the Pleasurize Music campaign ?

I wrote an answer to this a while ago in a post called It’s not about the numbers, but here I want to take a different approach, using… more numbers.

The limitations of the TT Meter

The image above shows the loudness values of Steven Wilson’s awesome new album “The Raven That Refused To Sing“. As you can see, it’s very dynamic – the loud sections are “loud”, with a crest factor of only 8dB or so, but there’s still plenty of variety and contrast throughout.

But it only measures as DR9. That’s only one dB off the DR8 minimum we recommend. So how can the album be as great dynamically as I say it is ?

The answer lies in the limitations of the “DR” value as calculated by the Offline TT Meter.

The truth is, it’s a VERY blunt instrument.

The DR value only tells us how loud the loudest moments of an audio file are, and averages these values over all the tracks of an album. This is a quick, convenient, familiar shorthand – which is why I still use it regularly – but it tells us nothing about the internal dynamics of a song or album, and in some cases can be highly misleading.

Below the waterline

A more sophisticated analysis is to use an ITU-based loudness meter. For example, here’s the loudness graph for one of my personal “benchmark” albums for great sound – “Wish You Were Here” by Pink Floyd.

You can see that the overall loudness is lower, but also that the “loudness range” of “Raven” is actually 16LU (loudness units) – 2LU more than “Wish You Were Here”, despite having an overall integrated loudness of -12LUFS, which is 4 dB louder than the Floyd album. (An LU is roughly the same as a dB)

And looking at the graphs you can see that beneath the tip of the DR iceberg, “Raven” is actually more varied in it’s dynamics.

Notice I’m not saying it sounds better – just that the “DR” value is almost entirely useless as a way of distinguishing between these albums. Even though “Raven” is louder on the CD, and the loudest sections have a much smaller crest factor than “WYWH” – it sounds just as dynamic, if not more so, than the older record.

When is DR8 not GR8 ?

Now take a look at this graph, of Blur’s 1997 album “Blur”.

This album measures DR8 – only 1 dB less than “Raven”. But it’s easy to see from the graph and the “loudness range” value that it’s much more restricted dynamically.

Again, I’m not saying it sounds worse, necessarily – just different. The style is worlds apart – deliberately dense, distorted indie rock textures as contrasted with Wilson’s lush, open prog soundscapes. It suits the material and I really like the sound of this album, too – but it goes to show how much information is hidden when we try to pigeon-hole an album by a single DR number.

So why am l always quoting DR values, still ?

So is it time to stop using DR values ? Maybe the goals of Dynamic Range Day and it’s Challenge should be updated to request a minimum “loudness range”, LUFS reading or some other metric ?

I don’t think so. Not all albums will have (or need) a wide “loudness range” – we’re not trying to stop people making dense, intense music, or creating heavily distorted textures. We just want people to realise they don’t have to make their music sound like that in order to “compete”. As ITU volume normalisation becomes common-place, we’ll be more free than ever to mix and master our music to sound exactly as we like it, and know that people will hear it the way we intended.

As I said above, DR values are quick, convenient and familiar. And in many cases they can still give a reasonable snapshot – if all the songs on an album read less than DR6, for example, the chances are it will sound pretty crappy – but even that isn’t guaranteed.

In general, lower DR albums tend to have more limited dynamics – especially beyond DR8, which is why I chose it as the threshold for Dynamic Range Day’s Award and Challenge – but it’s still only a crude estimate.

The way an album really sounds has far more to do with style, arrangement and production than it does with DR value, peak level or any other technical distinction.

To really understand loudness, to make better judgements and comparisons, we need to use more sophisticated tools – and of course, we need to listen.

facebook comments:


  1. says

    Almost ! A minimum of DR8, yes – but the LUFS guidelines are for playback only, not mastering.

    If you’re mixing for broadcast, -23 LUFS (integrated) is the guideline.

    Otherwise, an average LUFS of -11, peaking at -9, corresponds to my DR8 recommendation.

    Does that make sense ?

  2. says

    This is interesting but more difficult than I thought!

    If the LUFS broadcast standard (-23) will also become the standard in mp3 players, than lower values then -23 will mean that listening to the song on a mp3 player will compromise the dynamics?
    Or.. as long as my values have an average of -11 LUFS and the DR = 8 or higher the mp3 player will not effect the dynamics of the song?

    And is it not the goal of all mastered song to be broadcasted? (and eventually be listened to on MP3 players)

  3. gabe says

    Here’s a question:

    if two tracks (of the same music) have identical dynamic range (R128) LU units but different DR values, do they have the same dynamics?

    Would the differing DR values reflect how close the peaks to 0 are?

  4. says

    You mean if they measure the same R128 loudness ? That on it’s own doesn’t really tell you much about how squashed they are.

    If you calculate the peak-to-integrated loudness ratio (PLR) this will give you a number similar to the DR value, though.

  5. Julian H says

    Hey Ian! I don’t agree with your sentiments about the DR meter not being as useful as it should be…for me it doesn’t matter if a song has great macrodynamics if the loud parts are squashed! It’s squashed microdynamics that are annoying. If a song keeps the average level relatively constant, that isn’t so much a problem as the squashing of drums. For me personally, DR 9 is the minimum of sounding good for hard rock, whereas prog and art rock should have DR’s of 11 or more.

  6. says

    Hi William,

    DR and “loudness range” (LRA) are completely different measurements, there’s no way to connect the two.

    If you want an “LU equivalent” of DR, the closest you’ll get is PLR (peak to loudness ratio). To find this, measure the integrated loudness, and find the difference from the maximum peak level. So for example if the music peaks at -1 dBTP, and the integrated loudness is -10 LUFS, the PLR is 9, which is *roughly* equivalent to a measurement of DR9 with the TT offline meter – but only roughly…


Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>