LUFS, dBFS, RMS… WTF ?!? How to read the new loudness meters


 
 
LUFS, dBFS, R128, dBTP, ITU BS 1770, ATSC A/85, TT DR

WTF ?!?

There are now so many ways to measure your music these days, it’s not surprising that one of the most common questions I get asked these days is

What do all these acronyms mean ?

In particular, people want to know about the new “ITU-based” loudness meters, which use mysterious new “Loudness Units” – LU for short, or LUFS to be more technical. Especially as mainstream software like Cubase 7 and Ozone Insight begin adding support for them.

You know they’re important, you know that meters using the new system are appearing in more and more places – but how do you know what they mean ?

Well the good news is, it’s not as complicated as it seems. There’s a pretty simple relationship between the new LUFS measurements and the loudness measurements we’re used to, like the TT Dynamic Range Meter’s DR value, or good old RMS average levels.

The video above demonstrates all of this, as clearly and simply as I could make it – as well as an important complication with RMS levels that everyone needs to know about.

You can download the “pink noise” test file I use in the video to try this for yourself here.

The loudness meter I use in the video is the excellent Meterplugs LCast, and you can find out more about the Music Loudness Alliance here.

I hope you find the video helpful, and let me know if you have any questions !
 

Appendix: Get your RMS right !

I mention in the video that some DAWs and meters use the wrong method to measure RMS for music, and so many people have asked me about this I thought I’d add some more detail.

RMS can be calibrated using either a full-scale square wave, or a sine wave. The correct way to do this for a music signal, as defined by AES Standard AES17-1998, is to use a sine wave peaking at 0dBFS – for more details, click here.

If a meter is calibrated using the wrong standard, the RMS level will under-read by 3dB, and you’ll be tempted to mix or master everything too loud.

Here is an (incomplete) list of DAWs and meters this has been tested in:

Incorrect

Slate FG-X
Waves PAZ - 4dB off (?!)
Waves WLM
Logic 9 Multimeter
T-RackS (Older versions)

Correct

Wavelab
TT Meter
Ozone 5
Waves Dorrough
Voxengo SPAN

(Actually by default it’s wrong but this can be corrected using the “RMS +3dB” preference)
Reaper is about right, but the time constant needs to be correctly set to 300ms
T-RackS CS Meter (Latest version)

 

Test for yourself

Simply download the pink noise file here and try it in your own DAW or meter – the RMS level should read very close to -11.5 dBFS. If it doesn’t, something is wrong.

And, please feel free to report the results of your own tests in the comments.
 
 

facebook comments:

Comments

  1. says

    Thanks for the overview. I understand most of this a lot better now, but I have two questions:
    1) You lost me when you compared the meters with the TT Meter. You were comparing LU, Dynamic Range, and RMS, but all of your values were at or around 9, so I couldn’t correlate anything. Is it possible to give an example where the values are different?
    2) -23LU seems like it’s really really quiet. Doesn’t that mean that we’re going to be mixing and mastering at much lower levels? I get that I can mix/master as loud as I want and it will just be turned down, but if I aim for -23LU when mixing, that’s about like mixing at -23dB RMS, isn’t it? That’s roughly 13dB quieter than we mastered at in the HMM.

  2. says

    Hi Tim,

    1) If the signal is peaking at close to zero, the values will always be similar. So if the RMS is -14, the LUFS will be roughly -14, and the DR will be roughly 14.

    I say roughly, because both LUFS and DR employ “gating”, meaning they assign less priority to low-level signals than pure RMS, and LUs also take account of the ear’s frequency-sensitivity, to a certain extent – but as a rule of thumb, it’s not too bad.

    The exception is the DR – if you have a DR10 signal, and turn it down by 6dB, both the LUFS and RMS readings will also fall – but the DR will stay the same, assuming it wasn’t heavily limited or clipped to start with.

    2) -23 LUFS seems quiet in comparison to the absolute levels on CDs because it is – so there is enough headroom for classical and especially feature film material at a similar playback volume. I don’t recommend -23 LUFS as a guideline for mixing or mastering – I think somewhere between -14 and -18 is fine, assuming there’s no clipping.

    But really, the absolute level when mixing shouldn’t be critical in general – it’s only when mastering that it becomes an issue.

    At that point, my recommendation is to aim for an *integrated* value of -12 LUFS (over an album with a fair bit of contrast) averaging a momentary -10 LUFS reading during the loud sections and peaking at -8 LUFS – which corresponds closely with what I suggest on the Home Mastering Masterclass course.

    Whereas plenty of recent releases are up at -6 LUFS or higher almost all of the time… BUT as the ITU guidelines are implemented more and more widely, no-one will ever hear that !

    Ian

  3. says

    Thanks, Ian. that helps a good bit. I wish the meter manufacturers were making recommendations similar to yours. The meter you used in the video had adjustable thresholds, but most of the cheaper/free ones don’t, and working in the -12 to -10 range is squarely in the red on the meters. That gets confusing for folks. I know how to read my TT meter, but all of the LU meters confused me and seemed to be saying “YOU’RE WAY TOO LOUD!” Maybe I can start using them more now that I have a better understanding.
    Oh… but I do love that TT Meter, 32-bit bridge on Logic and all….

  4. says

    Yeah, the thing is those recommendations are using the broadcast standard – since so far there IS no standard for CD. It’s confusing, I agree.

    And yes – the original TT meter is still one of the best !

  5. says

    Great video Ian.

    This is all excellent news and could rescue recorded music as a commercial art form. I don’t think there is much wrong with the CD format. I believe people started thinking CD sound sucked simply because of the audio fatigue syndrome associated with hyper loud masters.

    Sure, 24 bit/44.1 (etc) is notably more enjoyable and revealing on well recorded and mastered music but I’d settle for properly dithered 16/44.1 if future CD releases were mastered to the new standard.

    With the current state of consumer technology we are now at a point in time to deliver, commercially, fantastically dynamic and well recorded music to the masses and given these new standards we may see Joe Public start to appreciate music from an audiophile perspective.

    As an aside, it’s interesting how films have been forced down the route of music where people are seemingly content to watch films on Youtube etc, in dreadful quality – just like they do with music and mp3s.

    One final point: Part of me wishes they’d pick -20LUFS over -23, just because it falls in line with K-20 nicely which is a good standard to work to, allows for good dynamic range across many platforms of music and matches up with 0VU/-20dbFS if your hardware is calibrated that way.

    I look forward to punters using their volume knobs again when a song excites them so much that they want to hear it louder – and with the dynamics intact!

  6. Stone Walters says

    Hey Ian. I thoroughly enjoyed this video and it shed much light on what can be a perplexing subject, so thank you for explaining it in such simple terms.

    My one question was going to be about whether to aim for -23LUFS when mixing for radio but Tim beat me to it! If I understand your response correctly somewhere between -18 to -14 LUFS is a good level for mixes to come in at providing there is no clipping. Based on your ‘Headroom’ article mixes should ideally also have headroom of -6dBFS (when measuring peaks). Shooting for these targets whilst mixing should allow the mastering engineer enough scope to do what they do best, creating a scenario in which mixes are compliant with current/new broadcast standards. Do I have this correct?

    Thanks!

    Stone

  7. says

    Hi Stone,

    Almost ! If a mix is reading -18 or even -14 LUFS, I think it’s safe to peak right up near zero.

    If you want to play it safe and have 6dB of headroom, I’d make sure the integrated loudness is adjusted down accordingly – between -24 and -20, say.

    Ian

  8. Stone Walters says

    Hey Ian,

    So if the mix comes in at -18 to -14 LUFS that means it’s OK to peak close to zero and still be one of your mastering engineer’s favorite clients! I think I’ve finally got it!!!

    Thanks again!

    Stone

    PS You also mentioned that you cover this info in more depth in one of your courses. Which video/ebook is this?

  9. Mark Hawkins says

    With the release of Logic Pro X, I was playing with the Waves WLM today with the pink noise file since the TT DR Meter no longer works with X’s 64-bit mandate. Watching this video, it seems to suffer from the -3 db lower issue you described. I’ve tried the various options in the plugin but don’t really see a way to adjust the reading. Why does WLM use the lower reading? Is it a problem? Is there a way to fix it other than use another meter. :-)

    Thanks in advance!

    -Mark

  10. Kahlbert says

    Hm, I see a difference between “true peak” and “normal” meters but I’m not sure about that math there.

    The level meter (plugin) in Studio One normally shows the correct level, but is also 3 dB off when switched to TP. So that might be the reason why WLM is off as well …?

    On the other hand, K-meters show the correct level, too, but aren’t they meant to be true peak?

    Dang, I really thought I had understood the whole level thing years ago …

  11. Jeroen Wolff says

    Hi Ian, what do you mean by “Reaper is about right, but the time constant needs to be correctly set to 300ms”
    Because in my Reaper it shows almost 0 dB FS RMS, far from the supposed -11.5 ?

  12. says

    Hi Ian,

    Came across this article some time ago, and a few days I got back to it for a bit of reference.

    I noticed there’s no readings for Sony Sound Forge (I’m using SF Pro 11), so here are some readings on that Pink Noise file…

    Minimum sample value (dB): -2.513
    Maximum sample value (dB): -2.081
    RMS Level (dB): -14.680
    Average value (dB): -51.128
    Maximum true peak sample value (dB): -1.499
    Maximum filtered true peak sample value (dB): -1.499

    LOUDNESS
    Integrated: -14.45 LUFS
    Loudness Range: 4.90 LU
    Maximum True Peak: -1.50 dBTP
    Maximum Short Term: -14.29 LUFS
    Maximum Momentary: -14.06 LUFS

    I guess it’s off by 3dB. Now I’m wondering how can I use this information to benefit from it.
    Cheers!

  13. Adyrhan says

    Hi Ian.
    I’ve been interested into measure dynamic range with the new metrics(LUFS)and I found an opensource VST plugin called klangfreund LUFS Meter. Have you tried it? What’s your opinion on the accuracy of it?

  14. says

    Hi Ian

    Just to say I have just tested the pink noise file though Waves WLM Plus and it gave me a reading of -11.5 LUFS Long Term and -11.6 LUFS short term so it looks like they have fixed the -3db problem!!
    I was also testing Waves WLM V TC Electronics LM2n and they have me almost identical results +/- .1db

    Best

    R

  15. Jim says

    Hi Ian,
    I have been fortunate enough to be able to work with a mate who has the Waves WLM and I brought him the TT Meter and showed him your video. He was shocked to find the discrepancy of around 2-2.5db RMS.
    His question was “How can a £300 plugin be wrong compared to a free plugin?” When we looked at the prices of those other meters you referred to as inaccurate – it’s amazing!
    How do the companies explain this? Would they say something along the lines of “well obviously you need to calibrate your system to our product”?

  16. Nicolas Godefroy says

    Hi Ian,
    I used your pink noise in logic 9.1.8 with both the multimeter and the plugin Nugen VLM-H, the multimeter shows a RMS peak at -15dB as opposed to VLM which shows a value of -11.4dB (INT average RMS)

    Now following your videos presentation regarding R128 new standard for TV broadcast, when mastering for TV I should make sure that my true peak value is maximum -1dB and my INT average LUFS value is -11.5dB is it correct?
    Also when mastering using Waves L316, if the celling is at -0.2dB how come the multimeter shows the correct FS value of -0.2dB but not the plugin Nugen VLM-H?
    Does that mean the waves plugin are not calibrated too?
    Thank you so much for your time and thanks again for your amazing advice.

  17. Phil says

    I also downloaded the pink noise file and compared with LCAST, Waves WLM and the Cubase meter LUFS and they all give a reading of 14.4 LUFS. In desperation I downloaded a demo for Waves Dorrough and that gives 11.5dB. Does anyone have an idea to help me? I would have thought that LCAST should have given me 11.5 LUFS? Help???

  18. says

    This is a mono test file – to get the correct value, it needs to be played by a mono track in most DAWs. If you play it on a stereo track and pan it to the centre, chances are the DAW will reduce the level, usually by 3dB (otherwise sounds would get louder as we pan them to the centre of the stereo image).

    Hope that helps ?

  19. Daniel says

    Hi Ian, I’m seeing the pink noise file’s RMS read -14.5 in DP 8.06 MW Limiter, TC’s LM2n, MetricHalo Channel Strip, and Spectre… Kind of disturbing. I appreciate the link to the standard, and I’d love to see a video on the RMS discrepancy topic alone, and/or references to other explanations if you have them. Thanks!

  20. says

    Hi Ian,

    I’m more into podcasting than mastering, but I’m trying to get all of my podcasts to hit -16LUFS, which seems to be a reasonable target for mobile devices according to some research by Thomas Lund from TC Electronic and indeed the apparent level that Apple’s soundcheck feature loudness normalises to.

    I tried your test file in Audacity and it gives -14.5dBFS, no real surprise there.

    The real problem is in Cubase (LE4) which gives -11.5dBFS on the TT meter for a mono track and a stereo track either hard panned OR centered when the stereo pan law is set to 0dB.

    Doing the same test with stereo pan law set to-3dB across the board EVERYTHING reads -14.5dBFS.

    I thought I understood how the LU system works, RMS and the theory of stereo pan law etc, but I seem to be getting bizarre discrepancies with mono vs stereo particularly as I’m trying to podcast in mono not stereo. Cubase just seems to be doing a 3dB level shift across the board, or am I missing something?

    Cheers, Richard

  21. Brian says

    Hi Ian just found your site, last week.

    RE ^ Reaper. The RMS scale need’s to be selected by right clicking on the MasterFader, and like Span, the 3db offset must to be selected also. I’m getting -11.2 RMS and – 2.1 Peak.

    I find metering to be one of the most confusing aspects of recording in general, but thanks to vids like this it’s becoming clearer!

  22. Phil Hershkowitz says

    Hi Ian, I downloaded the Meter Plugs LCast Loudness Meter Demo and installed it into Pro Tools 11. With its default settings it shows the pink noise with an RMS of -14.4 LUFS, not the -11.5 you show in the video. Waves Audio WLM + also shows -14.5. The Ozone 5 Demo shows it at –11.4

    Now I’m really confused. I need some simple way to anaulyze my Voice Over files to see if I’m complying with the ACX audiobook production standard:
    “Each uploaded file must measure between -23dB and -18dB RMS” described on this page http://www.acx.com/help/acx-audio-submission-requirements/201456300

    I am using this pink noise file from your webiste: https://dl.dropbox.com/u/8441718/audio/Pink_Noise.wav

    Any thoughts on what might be wrong?

  23. Andy says

    Hi Ian, I just subscribed to your newsletter.

    I’m using Logic 9, and Voxengo Span as a meter. The RMS values for your test wav are showing -14.5 stereo, and as -17.5 as a mono. Any comments PS Span is set to DBFS on the metering options.

    Thanks

  24. says

    Thanks for a great post. I teach Sound & Music production at a college and I’ll recommend my students to read and watch this as supplementary material to my loudness measurement lectures.

    I also tested the RMS of your pink noise file in Logic X and both the “Multimeter” and the “Level Meter” failed to produce a correct reading. Thanks for explaining why.

    /Henrik

  25. David Oliver says

    Interesting read! I tried your pink noise sample in DSP-Quattro with its built-in “Find Peak and RMS” function, and it registers -14.68 dB RMS.

  26. says

    Sorry I can’ answer all the comments individually. Please make sure you download the most recent (stereo) version of the pink noise file if you’ve noticed anything unexpected in your tests.

    At least one person has reported that the Waves WLM has fixed the issue

    Jim asked “How can a £300 plugin be wrong compared to a free plugin?”

    Good question ! To be clear, the 3dB lower reading is only wrong for music, not technically incorrect for RMS (there are two standards). But there are music apps and the developers used the wrong standard for music…

    Nicholas asked “when mastering for TV I should make sure that my true peak value is maximum -1dB and my INT average LUFS value is -11.5dB is it correct?”

    No – for TV the INT LUFS should be -23 LUFS

    “when mastering using Waves L316, if the celling is at -0.2dB how come the multimeter shows the correct FS value of -0.2dB but not the plugin Nugen VLM-H?”

    The Nugen shows True Peak – including inter sample peaks, whereas I suspect the L316 doesn’t take these into account.

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>