Loudness Links repository

[nucelar]

Hi all,
As part of my involvement in the Spanish workgroup for Loudness in Broadcast and Postproduction, I have collected some links to useful documents I'd like to share.
I hope this thread grows with your contributions and helps for the "evangelization" of this noble cause :)

Links marked with <======= are of special interest to "Loudness newcomers"

EBU's main Loudness page
Following can be found here:
EBU R128 Loudness Recommendation
EBU Tech 3341 Metering specification
EBU Tech 3342 Loudness Range descriptor
EBU Tech 3343 Practical Guidelines
EBU Tech 3344 Distribution Guidelines
Test material for compliance testing
"On the way to Loudness Nirvana" article <=======
2 Webminar recordings <======= the first webminar is simply excellent
EBU TECHNICAL - Loudness

10 things you need to know about Loudness
http://tech.ebu.ch/docs/testmaterial...s_loudness.pdf <=======

Article in Broadcast Engineering Magazine
Broadcast Engineering World - February 2011

The TV Technology Guide to Audio Loudness (free registration)
Registration

A/85 – Techniques for Establishing and Maintaining Audio Loudness for Digital Television
http://www.atsc.org/cms/standards/a_85-2009.pdf

ITU-R BS.1770-2 (03/2011) measurement standard:
Algorithms to measure audio programme loudness and true-peak audio level

Articles at Qualis Audio:
Qualis Audio Downloads


Some loudness metering products.
partial list Last update: March 2011

TC electronic TM7, TM9, LM2...- Hardware meters
DK Technologies MSD, PT0600...- Hardware meters
RTW Touch Monitor... - Hardware meters
Wohler Pandora - Hardware meters
Qualis Audio Sentinel - Hardware meters
Harris LLM-1770 - Hardware meters
Tektronix WFM6000/7000 - Hardware meters
Trinnov Smartmeter - Hardware meters
Nugen Audio VisLM - Plug-in for realtime metering and offline analysis
Grimm Audio LevelOne - file based analysis normalization
Pinguin - several software solutions
Dolby Media Meter 2, LM100, DP600 Plug-ins, standalone software and Hardware meters
Channel D Audioleak - File based analysis and normalization, standalone metering
Zplane PPMulator V3 - 2Q 2011
Flux Pure Analyzer System -2Q 2011
Audiocation Free PC VST Plugin
TB EBULoudness by toneboosters (VST)

Loudness feature request at Ideascale
http://protools.ideascale.com/a/dtd/...770/29859-3779

Open Source library:
libebur128
libebur128 - (yet another) EBU R 128 implementation - Hydrogenaudio Forums

Open source file scanner R128GAIN

(note: The BS1770 has been updated to BS1770-2 in March 2011. The measurement now includes relative gating at -10 dB. Check that your meter is up to date)

[Postman]

Great to see lots of links gathered together. Thanks!

Unfortunately, "noble cause" is not the first phrase that comes to my mind every time I start mixing with a loudness target.

"Pain in the buttocks"
"nuisance"
"pest"
"stinkin' POS"

[nucelar]

Hi Postman!
Thanks for your insight.
I'm really interested in why you think that way, especially from an experienced and respected pro like you. Could you briefly elaborate?
Let me guess: You have been mixing to specs like "-24 LKFS and -10 peaks" (facepalm). Just to be clear that is NOT mixing to loudness as it should be.
I genuinely think that when done well (dropping the limiter), loudness normalisation is the holy grail of audio quality for tv.
Thanks!

[Postman]

Hi nuclear! Okay, well, since you've asked...

Let me guess: You have been mixing to specs like "-24 LKFS and -10 peaks" (facepalm). Just to be clear that is NOT mixing to loudness as it should be.

Correct, -24LKFS or -23LKFS, +-1db, with peaks at -10. Finally this year I see some easement on the -10dbfs peak ceiling to a -2dbfs, but not consistently. PBS warns that going above -10 may clip the analog parts of their network. NatGEO requires the stereo mixes to be at -10dbfs, only the surround mix can go to -2 (as of Dec 2010) Spec sheets are what they are, though, so I'm not afraid to slap on the big honkin' limiters across my busses, and away we go! No, its not "mixing as it should be", it is "mixing by the meters".

A rosey ideal of a world where every mix sounds equally loud and yet there is plenty of headroom for wonderful natural dynamics, well that doesn't really exist when you put real people in the world.

Networks are just big switchboxes, shuttling content about from one media server and then the next. Most don't want to take responsible for the moment to moment loudness levels that they broadcast if they don't aboslutely have to. That takes intelligence, or a big honkin' limiter, so instead they have this great new weapon of spec'd loudness targets. They can shovel that responsibility onto their vendors. This is all about money.

Once the networks figure out that we can measure loudness "consistency" or "range", which is a feature of several loudness meters now, I predict they will adopt a very narrow loudness range into their specs, something like 4 to 3 db! THAT is what most of them obviously want, not a glorious high fidelity aural treat. HBO is a noteable exception. Of course, new loudness specs coincide with a general reduction of production/post production budgets. (I met with an executive of a large network production unit last year, who was only slightly appologetic when he told me the purpose of the meeting. He said that although their budgets were going to drop about 30% this year, their requirements for quality sound work could not change, and so I should just know that in advance...thanks for the meeting!) We sound mixers are faced with less money and the additional time required to meet loudness specs. Depending on how a spec is written that can be many hours or even more in my experience.

I am not some primadona who wants to have "freedom for my art." My first network show came with a producer who wanted to be sure that our show was just as loud as the commercials (this was over 20 years ago!). I've worked hard to keep my tv work loud and consistent, and I didn't need the constant worry of a +-1db window to do it. Loudness meters lie, and that is why keeping within the ever-moving window robs me of valuable time.

[nucelar]

Hi postman, thanks for your detailed explanation. It seems you have had a quite some experiences to make you skeptic, but have faith!! have dreams!! At least here in Europe, tv stations are slowly "getting it". This debate is necessary because many engineers may feel like you, but it's certainly is getting better.
We are still in the beginnings of the loudness metering system, and many networks still haven't embraced it fully because they have to deal with mixed digital-analog distribution and/or legacy equipment and are not sure what to to themselves. But when analog disappears forever from broadcasting, i feel that tv audio quality will improve considerably.

Mixing to a loudness target is way easier than you describe, though. You said

...and I didn't need the constant worry of a +-1db window to do it. Loudness meters lie, and that is why keeping within the ever-moving window robs me of valuable time.

But that's not the idea. That would leave you with a 2dB dynamic range :) The loudness is measured on average for the whole programme, be it a 20s spot or a 2 hour movie. I just mix away like i think it should sound (hardly looking at meters) and only when finished, I analyze the whole mix offline and normalize to exact -23 LUFS (takes 1 min). On a calibrated room, you will land around -23 anyway, so I find it the most natural way to mix.
I don't think networks will impose a narrow loudness range, what would they gain?
Anyway, from my experience it's been a positive experience to be able to mix "freely". If at the end I have to normalize to a common loudness, I'm very happy with that.
Cheers!

[Postman]

I analyze the whole mix offline and normalize to exact -23 LUFS (takes 1 min)

Huh? How does that affect your peak readings? Do you normalize all stems as well? Really? What happens if there is a change later?

Much or my anger about the misuse of loudness comes from personal experiences. If everyone from here on adopts the same reasonable specs then I will have no future issues.

I have had to work to a spec that stated dialog throughout the program had to be maintained within a certain relatively narrow limit. I don't recall now what the limit was, but if I get a few minutes to look through old deliverables spec sheets perhaps I'll find that one and update this post.

The specs I mixed for last week state that each section of program must measure -24LKFS +-1db. That is each act, for this particular show. Some acts were mostly talk, some were mostly action. The dialog is NOT at the same level in each act, in order to make the results correct, and that takes time to figure out.

Any spec sheet that specifies -23 or -24 LKFS with peaks stopped at -10, is really saying they want a loud show with no dynamics, and if the show is a talkie, the dialog WILL be highly compressed. If they wanted dynamics they could have easily specified -27LKFS or even -29. So, once the artificial peak limit is raised from -10 to -2dbfs, it is logical they will require a narrow loudness range to assure the same results. Will they? I don't know, but that is my prediction.

I didn't need the constant worry of a +-1db window to do it. Loudness meters lie, and that is why keeping within the ever-moving window robs me of valuable time.

But that's not the idea. That would leave you with a 2dB dynamic range :)

Yes I know that the latest recommendations from EBU and ITU suggest measuring the entire show. I did not make my point clear. When I said loudness meters lie, I mean that what sounds equally loud to my ear may not be the same to the meter. For instance, a section of narration without music might measure something like -25 or -26 if there are lots of spaces. A section of archived sound, like the calls of a sports game commentator with crowd roaring, will easily read -22 or even higher, for what sounds like the same volume to my ear. So, IF there is a loudness range expectation that will be part of a Quality Control measurement, then I am forced to mix badly and make the meters read correctly, or I need to be prepared to deal with the consequences later. By the way, I have had this exact problem be the reason a mix was bounced by QC. I had kept the volume down, to make the meter read within the allowed range (this was a spec that said all sections of their show had to be within the same narrow range, I did not have the luxury of measuring only the full show duration). The client complained about it, and rightly so. We made the mix more correct and got tapped by the QC report. So, as I said, the meter lies. If one has the added burden of narrow loudness range, as I have had in the past (and that I fear will continue in the future), one finds himself in a quandary.

Here's what happened just last week. I mixed a show for a company that I've done a lot of work for in the past, but this was a new producer. I kept the wall to wall music generally low under voices, as I am certain the other producer at this company would demand. To my surprise, this producer wants stronger music and I gladly raised the music up to competitive levels. But, of course, my carefully crafted -24 LKFS (as per spec sheet) now reads -22.9. So I make global adjustments. That's fine, it's no big deal, but I'm not sure what purpose was served other than to make a meter happy, which takes time and attention away from other potential problems.

[nucelar]

Huh? How does that affect your peak readings? Do you normalize all stems as well? Really? What happens if there is a change later?

At the moment I only normalize the final master, not the stems. But granted, we don't create stems for our clients. (We just use them
As for peak readings, normalizing to -23 *and* limiting at -10dB IS certainly a PITA. You cannot brickwall an existing -23 LKFS mix without raising it's loudness 1-2 or even 3 dB. But thankfully more clients are dropping the -10 ceiling. Of course, if you need to raise a full mix, lets say 3 dB, you still have to make sure your peaks don't hit the Full Scale.

For instance, a section of narration without music might measure something like -25 or -26 if there are lots of spaces. A section of archived sound, like the calls of a sports game commentator with crowd roaring, will easily read -22 or even higher, for what sounds like the same volume to my ear

That is very true. The next revision of the ITU-R. 1770 meter will include a gating algorithm which is designed to solve precisely that problem. VisLM has this already built in. Of course, there's no perfect Loudness meter, but at least now we have one standard agreed internationally. Compared to the level and meter chaos we have now, good times ahead.
We are now in the middle of a transition period, and we are probably suffering from badly written specs from "over conservative" networks who don't want to risk a thing or are tied to legacy workflows. I didn't know about the exception being HBO, thanks for pointing that out. I recently watched Boardwalk Empire and enjoyed the sound very much.
The ideal universal spec should be "-23 LUFS, -1dBTP". that's it. If a network wants this segment for segment, I don't blame them, but I don't see a reason why they would want to restrict dynamic range further.
Best regards!
PD: Others are also allowed in this thread !!

[Postman]

The next revision of the ITU-R. 1770 meter will include a gating algorithm which is designed to solve precisely that problem.

Gating will only partially solve the problem. The real problem, if I may be so bold, is that our perception of volume is apparently not linear with power measurements. We need a better method. No, I don't know what that is. Yes, someone else should chime in on this thread.

[zzstation]

Just one silly question:
For movie theather, is it ok if I have some peaks at -5, -6 dB?

I mixed using DMM an a -25 as infinite level. Most peaks at -10, but some (only few) at -5, -6 and -7
I didn't use limiter. Should I use always to prevent that?

Thanks in advance,
zzstation

[Postman]

For movie theather, is it ok if I have some peaks at -5, -6 dB?

Off topic, but Yes.