I welcome the end of the “loudness war” as much as anybody, but when I try to put this into action I am met with various problems.
First example: I recently fixed bad audio on a video (in Sony Vegas) and it was rejected by the broadcaster on the basis that the integrated level (“I” in the WaveLab meter) was -24.7LUFS whereas the spec is -23LUFS+/-1LUFS with a maximum of -1dB. This was easily fixed by nudging the master output up 1dB and re-rendering, but how are you supposed to know that? Sony Vegas Pro 13 has a basic “loudness report” option but it’s just a text file with a line per time interval, with a summary at the end … I suppose you could find some way to load it into a spreadsheet and graph it, but it’s not really useful.
Second example: I need to deliver a DDP for CD, so I’m wondering should I do the usual, i.e. follow my ears, and then finally peak-normalise to -1dB, or should I follow the loudness meter? If I do that (and this is a fairly dynamic acoustic piece with little or no compression) then I end up with peaks less than -9dB and an audibly quieter track.
I can’t find any easy way to arrive at a final master that reads -23LUFS. Maybe it’s brain-freeze, but I just don’t get it.
To analyze it for adjustment, I normally render the entire program to a temporary wav file (render “whole montage” or “whole file”, “open resulting audio file”, uncheck “create named file”). After the temporary file opens in the Audio File workspace, from the Analysis menu run Global Analysis with “analyze loudness” checked on the Loudness tab (click the “Analyze” button to start the analysis). When it’s done, look at the Integrated Loudness reading, and go back and adjust your master for the difference.
I am currently mastering a CD and have picked -16 LUFS as reference. i expect it will be turned down for broadcast but stilll produces a reasonable Cd volume with full dynamic range.It is an acousyic folk CD- any comments?
A reference of -16LUFS is probably a good compromise for music production and conforms with a vague concensus for CD mastering and on-line mastering. iTunes Sound Check and Apple’s portable products use an integrated loudness of -16.7 LUFS. (Hypercompressed commercial CD releases are often mastered much higher than this. -10 to -8LUFS integrated). -1dBTP for peaks is a good idea to avoid the possibility of inter sample peak distortion. LRA might be expected to be between 4 and 12dB but varies enormously depending upon the dynamics and type of material.
You could also consider keeping a separate 24-bit audiophile version of your master for future use.
I see a problem with both using the “Integrated Loudness” as reference anchor and a low value like -23LUFS for music. If it’s material for TV or something, the both ILk and -23LUFS make absolutely sense.
If we talk about music, I actually recommend the “short term loudness” (SLk) and a value of up to -15LUFS for forte fortissimo passages. Average it should still be somewhere around -18LUFS.
Want a higher “loudness” while still retaining the transients/dynamic of the song, don’t go higher than -11LUFS for forte fortissimo passages (which is roughly -14LUFS on average). I might even recommend to Google for the K-System v2 concept (KVR Audio is a starting point).
Wavelab’s batch processor might definitely help you there (Multipass - Loudness Normalizer). Though sometimes the settings for the Short Term Loudness Normalization confuse me.
For music - either my own stuff or vinyl restoration or even a simple mix CD for the car - I keep it simple and mix into the K-System (meters using K14) and ensure the material is just tapping 0 on that meter.
Then by adding the IzoTope Insight metering into the WL master section - I am always end up in the -14 to -16 LUFS range - which results in a perfect blend of “just right” volume and overall loudness.
K-System (v1, the original concept) is based upon an unweighted meter (strong bass can interfere with the results) with integration time of 600ms and a custom reference level.
K-System v2 (the overhauled concept) is based upon the ITU-R BS.1770-x specs (read: EBU R-128), also with a custom level, focus on the SLk meter, but utilizing all of the other benefits of measurement as well.
A shift from several dBFS between the K-System and the EBU R-128 meter on the readout is apparent!
So K-14 on the K-System (v1) meter can result in -16LUFS or even lower LUFS valuies - depending on the production, how much bass intensive material was used, etc. Not great IMO for objective analysis or loudness normalization.
If you use a loudness normalizer like the built in Wavelab one, it makes more sense to use SLk with a custom level - especially for music. At least in my opinion. Your mileage may vary.
And well, I created v2 - which currently is not widespread enough yet.
To specify, I basically wrote down the concept, that builds upon the specs of ITU-R BS.1770-x, in an understandable manner. I expanded on Mr Katz ideas and comments I got through mails, which gave me the inspiration of a logical evolution for music uses - he is still the sole creator of the K-System.
Funny enough, my “ideas” (loudness limits) were also picked up in AES discussions (I got wind of that from befriended engineers). Much like my ideas for the old “Dynamic Range Meter” (especially the “score sheet”) secretly sneaked into the official specs after having heavy contact with the developers. My ideas/logical thinking did even coincide with ideas from Apple and co with their planned/taken over “Replay Gain/Loudness Normalisation Schemes”. I only got to know this after I contacted the original EBU R-128 creators.
I even forwarded this concept to Steinberg at the time the R-128 metering plugin was still around. But I never got an answer, and there is also no presets existing in WL. Absolutely no problem on that end.
The only VST Plugin that actually took over the settings/concept in preset form is ToneBoosters EBU Loudness (Jeroen Breebaart). Eelco Grimm (Grimm Audio), Thomas Lund (tc.electronic), Tim Flohrer (zplane), Jon Schorah (Nugen Audio) and Dave Gamble (DMG Audio) know and read the white paper I wrote, even positively commented on the concept. Yet didn’t port “presets” since pretty much all of their tools can be setup independent. Ralph Kessler (Pinguin Ing.) and Bob Katz also know about this. Mister Katz even gave me a “silent agreement”, but both said that they don’t have time to read the papers. Though from a summary I gave them, they were like “interesting and logical thinking”.
BTW: I pretty much mentioned almost all of the “Music Loudness Alliance” staff in this paragraph, and also almost all of the original creators of the EBU R-128 spec!
The possibilities were already there with the ITU-R BS.1770.x specs… and it is a logical evolution of the LEQ, DR and K-System meters - pretty much also eliminates it’s major flaws. Just a “manual”/guide for musicians was missing.
Though granted, I still have to update this paper at some point since the original document still says “focus on MLk for music”, but in KVR Audio discussions, we debated and agreed on the issue to rather focus on SLk for music. Which did result in a better usability for loudness normalization.
v2 can be easily setup in Wavelab’s R-128 meter window (custom settings), though sometimes (really sporadic) I have an up to 1-2dB offset compared to other EBU R-128 meters I use. I filed this as bug report weeks ago already.
Summed up, v2 of the K-System is the following:
ITU-R BS.1770-x specs are used (corrently we’re on 1770/3)
custom reference levels (-20LUFS, -16LUFS, -14LUFS and -12LUFS - the last one for a “transition period”, K-20 is a compromise for those that don’t want to go to -16LUFS, which is somewhat stabbing at -18dBFS/0VU unweighted)
custom color coding (-inf to -3LU = dark green, -3LU to 0LU = light green/hotspot/mezzoforte passages, 0LU to +3LU = amber zone/forte fortissimo passages only, +3LU and up = red zone, don’t go there)
For very dynamic musical content (orchestra), I even concepted a K-16v2 “Dynamic” preset, which uses:
Reference level - 16LUFS
color codes: -inf to -7LU/-23LUFS = dark green, -7LU to 0LU = light green/hotspot, 0LU to +3LU = amber zone/forte fortissimo passages, +3LU and up = red zone
This preset is basically for a “transition” from movie to music or (as mentioned), or simply orchestra material. The “light green area” is for the average signal ideally hovering around the -2LU to 0LU value.
Should be simple to understand once you see it in action/have used it a couple of times
I couldn’t find any plug-ins that actually use a ‘version2’. Is the proposition being specifically considered by the EBU loudness group?
Just to point out also that the loudness meter custom setttings in Wavelab do not cater for four colour zones. There are only three, so we cannot set it specifically as you suggest.
I enjoyed reading your paper and tried out some of the settings. Do I assume that we are looking at the R128 momentary meter when using your settings? If so, are you suggesting that when set to a reference of -16LUFS we should never see the momentary meter go into the red? As far as I can tell, seems like an impossibility with most music I have tested, since peaks can easily make the momentary meter go way above the reference, and to be honest, the short term meter also tends to go well into the red on occasion.
TB EBULoudness … see the third paragraph in the description. It’s just not referred to as “v2” literally.
[EDIT] Actually, from the manual: Meter mode sets the display and metering modes to one of:
• LU EBU R128,
• LU EBU +9,
• LU EBU +18,
• LU EBU +27,
• LKFS ATSC A/85,
• LKFS ITU-R BS.1770-0
• LKFS ITU-R BS.1770-2
• LU K20 v2 (-20 LUFS)
• LU K16 v2 (-16 LUFS)
• LU K14 v2 (-14 LUFS)
• LU K12 v2 (-12 LUFS)
• LU K16 v2 (-16 LUFS)
ToneBoosters EBU Loudness and EBU Compact use K-System v2 presets. The first and so far only plugins on the market - most other tools can be “customized” according to needs. Back in the days, I was contacting Jeroen Breebaart about the possibility for customizations, which weren’t ported yet (maybe around v3.2 or something, currently it’s v3.0.x). But he liked the concept and created presets instead. Along with the affordable price, we tried to go this route to bring awareness to the users.
Guerrilla tactics, so to speak.
Like the old TT-DR Meter did years before.
The presets in EBU Loudness/Compact are tagged:
LU K-20v2 (-20LUFS), LU K-16v2 (-16LUFS), LU K-14v2 (-14LUFS), LU K-12v2 (-12LUFS) and LU K-16v2d (-16LUFS, dynamic) respectively.
(sadly) No, this concept was based upon my ideas, logical analysis and further thinking of “what could be done” to bridge the gap towards mastering music with the ITU-R BS.1770-x specs.
But as I earlier wrote, the people that worked on the EBU R-128 specs in the first place are aware of my concept. And some of this sneaked into discussions at various AES events.
I am aware of that.
The “light green” color zone is actually for easier understanding and simpler usage. It’s not mandatory, merely a recommendation. You can even use own color schemes if you like. As long as you stay withing -3 LU to 0LU for mezzoforte to forte passages, and don’t exceed +3LU on forte fortissimo passages, you’re fine.
Like I earlier wrote, this is something I have to rewrite at some point. Maybe I can finally get that done this month. Currently, the white paper recommends to focus on the Momentary Loudness meter (MLk), but in practical terms it doesn’t really make sense for suitable loudness analysis, and especially not for loudness normalization of music content. The posts on KVR Audio made that clear.
So the focus is indeed on the SLk meter, while ignoring the MLk meter.
Furthermore most SLk meters are also connected to histograms as well (Radar, Horizontal scroll, etc). So is the Loudness Range measurement (including custom gate settings). A logical step is to also use it for measuring music instead of a program stream only.
According to the official papers for the EBU R-128 specs, the MLk meter can overshoot of up to about +9 LU (sometimes more, depending on the material). Though transients (both digital peak and “loudness” peak) are easier to catch at a lower average signal strength (what we call “Loudness” these days) than having a constant level at (say) -8dBFS (RMS).
Just focus on an average signal of your feed being around the -3LU to 0LU SLk area, while the forte fortissimo (read: extremely loud parts, like breakdowns, bridges, etc) should ideally not exceed +3LU SLk. The MLk bargraph can still clearly exceed that value (as expected with a shorter measurement time) - but the important section is the SLk meter.
Want to test it yourself:
Try “Phil Colling - In the Air Tonight” and “The Police - English man in NY”. Adjust them both to K-16v2 so that the loudest parts (the breakdowns) don’t exceed +3LU SLk - you’ll notice that the program stream (both tracks played in row) will result in a fairly pleasant listening experience without reaching for a volume control. But you’ll also notice that momentary loudness peaks clearly exceed +3LU (see breakdown in the “The Police” track).
Again this is still according to specs of the EBU R-128 / ITU-R BS.1770 papers - and these “peaks” can be treated easily if desired.
So… if you want to go the “old” K-System route with the ITU-R BS.1770 specs, you can still use the MLk meter and not exceed +4LU for forte fortissimo passages.
But Loudness Normalization focused on ILk and SLk respectively in the last 1,5 years. Better let us focus on these meters.
And in the end, and this is what I recommend up and down especially to not so trained users, you only need to learn and understand three metering tools. For Recording and Mixing, use a PPM (digital) and VU (with reference level) combo. For mastering an ITU-R BS.1770 type meter.
Each to their own method, of course, but I would tend to use a combination of digital (k-system or standard digital) and loudness meters when mastering. A loudness meter can also be useful when mixing, especially the LRA figure.
Yes, an ITU-R BS.1770 meter can be useful during mixing if you for example have several vocals takes that need a certain alignment. Definitely no doubt about that.
But I see the ITU-R BS.1770 specs as definite future in terms of mastering as it combines several meters into one (and this is the reason why i created the concept in the first place), while I still use the “old ways” (digital PPM, VU) during mixing.
Getting back to this after some time, some more experiments, and my “brain freeze” has thawed out; first off, my second example in the original post can be disregarded, thanks stingray:
EBU R-128 is a broadcast standard so it doesn’t apply to CD masters … seems obvious when you say it, but hey, as I said, brain freeze. I shall continue to trust my ears therefore. I have to thank all contributors, especially StudioCompyfox for giving me much food for thought, and indeed, education!
Regarding my original actual problem, thank you bob99 for pointing me in the right direction:
In fact it’s the “Loudness Normalizer”, which is now fully EBU R-128 aware (when did that happen?) and there’s even a factory preset called “Broadcast recommendations (EBU R-128)”. I feel some embarrassment for not having seen this earlier. My task now is solved very quickly by rendering the audio from the video clip in Vegas, processing it in WaveLab 8.5 using the Loudness Normalizer, and re-importing it into Vegas. Sweet!