EBU R-128 meets K-System v2, a possible future for the loudness debate (Loudness War)

[HTBoogie]

Very interesting. Lots to read through and digest.

[AudioGuy720]

Honestly I think this EBU/K-16 standard is too little too late as far as music is concerned. This should have been something discussed and enforced over 25 years ago back when CD was becoming adopted by consumers.

Unless consumer playback electronics are forced to this EBU standard worldwide then you're still going to have people hypercompressing mixes. Personally, if the client is OK with it, I'll master at K-14 but 9 times out of 10 clients want it louder and as a business owner I have to give them louder even if sonics are compromised.

As long as mp3 is the predominate delivery format this idea is a bust. TV broadcasters will have to adhere to it for legal and quality reasons but the music industry will more than likely remain the Wild West.

Sorry if my messages across harsh...I would love to see this standard adopted. My speakers and headphones sound better when the volume knob is boosted past "2"...the typical spot for modern hypercompressed masters. Even at K-14 I still have a LOT more volume to go on even modest system.

I actually have a hard time listening to modern music because it sounds absolutely terrible...like a wash of sound that doesn't "kick" like music should. The more complaints consumers send to record labels I think the better off will be. Most importantly those complaints should include the words "I would have bought [musician X]'s new album but it sounds terrible." Money talks, BS walks.

[Compyfox]

I unfortunately have to agree that clients still want it loud. But I can slowly see a shift in that section as well.

More and more musicians and engineers say "it's enough". This is the very same reason why I mix/master at a low loudness and only raise it in the end. If it's desired of course, and then only up to a certain limit.

I even had it that a client came back to me and asked "do you still have that mixdown? I want it at lower volume please for a re-release". Luckily, I indeed had a backup of it.



Personally I still say that we should all work against overcompressed and too loud productions. And that starts with mixing already. The less workload there is for the mastering engineer, the better he can work, the more chance we have to see/hear actually great mixes and masters again over the course of the next decade.

And, if iTunes is really all about "-16LUFS" on average in the near future, we're somewhat future proof.



Agreed, we could have debated that topic over 20 years ago already. But engineers and producers all over the globe wanted to bend rules constantly which started this demise in the first place. However, it's not too late to pull back. And the shift is already happening as well.

The people are slowly regaining an open mind. All we have to do is offering proper education.

[@l3x!5]

This really is a fascinating topic. Many thanks to all involved.

I am confident that on the grand scheme of musical (and technological) history, the loudness war will be regarded as nothing but a brief anomaly, emerging alongside new means of production while causing a slight distortion of quality - before the community adjusts both on the technical and artistic ends. I like to believe that we are, if not near the end of it, at least/last at a turning point of sort.

I feel that more people start talking about it - or listening more closely when the point is raised. Surely we're gently moving to the point where it's becoming of public interest, hand in hand with the subject of audio compression? There's no doubt in my mind that the broadcast standards discussed in this thread will help BIG TIME. People have disliked over-limited tv spots for years, voices on some radios are barely endurable already etc. A little regulation is bond to lead to better practices overall, and I believe there's a public demand. I also think that it will help drive the demand for a better sounding music.

Compyfox, thanks to you I'm getting a better grasp at a lot of things!
I'm part of what I imagine is a silent majority that enjoys reading your posts but don't have anything of value to add to the conversation.
Same can be told of a number of quality topics here or on GS frankly, most probably a lot of folks like me, not having anything interesting to say, will only pop up to ask for clarification (with incidentally a 89% chance of it already being covered numerous times ) or blurp a thank you here and there.
Meanwhile we're learning! That's great

[DuX]

One of the things that would convince even the "non believers" is to put your -3dBFS average RMS "master" at the same RMS level as your "unmastered" mix at -14dBFS or so RMS, and listen to them side by side. Now tell me which one you prefer? That's how, I hope so, all the music will be listened to in the future, side by side, at the same RMS level. Then these "loud" masters will start to sound like an extremely unremarkable piece of smeared, blurred noise. Cause it *is* just a smeared, blurred noise without anything musical in it. Make a test. Hear the difference. It just sounds annoying, distorted, like an AM radio. I simply cannot believe people call that "mastered".

I think Fabfilter Pro-L has an auto-gain mode so you can listen to your song at the same RMS level while you're making up your mind how loud it should be? That can help to hear and evaluate what this practice is making to the sound better. The problem is that louder does always sound better to us humans. Even the management at many audio companies knows that and they use that fact amply in their advertisements and videos. But when all you hear is those transients going away, you begin to realise what it really sounds like.

Cheers!

[Compyfox]

Let's bring this one a bit futher up again.



Some specs for this particular topic:

So far this KVR thread was read almost 4000 times (around 3950 currently). Definitely some huge interest.


The white paper was downloaded:

Since the start of this thread (3rd January 2013):
78x (Jan), 31x (Feb) - 109x total

Since the original post on my techblog (4th July 2012):
68x (Jul), 31x (Aug), 38x (Sep), 32x (Oct), 51x (Nov), 36x (Dez) - 256x total

Since the white paper was first released, it was downloaded 365 times. Wow.



However...

So far only this KVR thread is listed on Google, and already a couple of pages back. I definitely got some visitors from ToneBoosters (since this was the first cross reference). But I did not find mentions about the K-System v2 on GearSluts, Google or Google listed Twitter/Facebook accounts or something. Yet it seems to have spawned enough interest to get that many hits due to forwarding or whatever.

Granted, could be more. But the interest seems to be there.

According to web references, it should have also been debated on the Cockos webboards? Yet, no trace to be found.



Anyway... I really thank you people for the interest. I even got note that Bob Katz gave some further input to ToneBoosters EBU Loudness (hence the PLR indicator with the most recent updates) and likes what Jeroen Breebaart ported in his tools. This is the biggest compliment someone could ever get (support from the original initiator).



Though let's discuss this topic further:

Did you adapt the K-System v2?
If you can answer this with a "yes", what did you discover?
What improved on your end?
If you have clients, what did they say?

If you're not using the K-System v2 - what issues do you have with it?
Do you understand it or are you lost in the ocean?
Or have you tried it and got somewhat scared off? If so, can you maybe elaborate on it?


Also regarding the prosed timeline for pulling down loudness again (much like the "ban of the lightbulb" timeline in Germany).

Do you think this schedule is too drastic?
Or do you say "ah funk it, let's drop right from the start"?



Again, please educate yourself, forward this thread (or my techblog), leave a comment in here, ask questions.

Thanks.

[Sounddigger]

Let's bring this one a bit futher up again.

...

Did you adapt the K-System v2?
If you can answer this with a "yes", what did you discover?
What improved on your end?
If you have clients, what did they say?

If you're not using the K-System v2 - what issues do you have with it?
Do you understand it or are you lost in the ocean?
Or have you tried it and got somewhat scared off? If so, can you maybe elaborate on it?

Right, let's pick up a few issues that were still laying around...

After having done a whole lot of testing of the K-System v2 proposal, I can now confidently say that in the real world this proposal is not quite as good an idea as it may look on paper.

But that also depends on what your ultimate goal is with this: is it your goal to really improve people's listening experience? Or is it merely your goal to (somehow) "punish" tracks that are uber-compressed and/or exceeding a true peak of 0 dBFS?

As I described in my earlier post, I checked this proposal on a number of tracks. In fact, what I did was to load 15 tracks into my DAW, each with particular very different dynamic profiles, and see/listen what would happen if I aligned their levels according to your proposal, as well as what would happen aligning them to the -23 LUFS EBU R128, both SLk and MLk, using Tonebooster's EBULoudness meter.

The tracks were very different, going from classical music by Ravel, Dvorak, to Rock & Roll from the Beatles, old skool House & Techno to a fairly recent electronic-music release by JM Jarre (very uber-compressed/limited, and very unpleasant to listen to at high levels, due to digital distortion - so much for recording with true analog synths, duh...).

Then I started listening switching from one track to another, leveling, metering, tweaking, comparing, and so on.

Well, I can tell you that aligning these tracks in any way that is being proposed - it be Kv2 or EBU R128 - does NOT improve my listening experience one bit!! And please note that I have tested things for many, many hours, several times, so this is not just an off-handed "easy-done" conclusion.

Compyfox, this makes me seriously wonder if you have actually tested your own proposal in the REAL world? And not just on the meters of your studio(s)?

Have you actually ever done a mix, lining up say 8 to 16 tracks at 0 LUFS K14v2 MLk, and listened to it, really taking the time to experience the sound and the music? Have you checked it in your living room, or in a room with a handful (or more) of people chatting? This could be only using mixes that have a nice dynamic profile, say comforming to the Katz system. THOSE are the ultimate tests in the end, there are none other...

Personally, I found that I still feel that some tracks are too loud, while others are too soft. And NOT the "right" ones! (as far as their dynamic profile goes) Even when leveled "perfectly" I'd still be found running for the gain knob to adjust the volume. But then again, maybe that's just me and who knows I simply have a volume-knob fetish

Granted, leveling according to the proposals can reduce level differences, but I find it actually augments annoyance due to the fact that it cheats you into thinking the level is right, when in fact it isn't (as felt), and would be better set manually at once.

Surely, if even excellently mastered tracks like Bruce Swedien's (using NO compression or limiting AT ALL) have to be leveled down as much as it would require, and sound too soft when compared to many other tracks once leveled at 0 LUFS K14v2 MLk, as is the case (generally) for the classical tracks, then it's not hard to see that something must be wrong with the K14v2 proposal.

And if you care to think it through a bit more, then really it's not too difficult to understand why the proposal simply will never be able to improve the listening experience, once and for good, for any and all cases. It might be able to improve things for a certain category/style of music, i.e. when taking them together (when they actually already fit), but even then only to a limited degree.

The reason for this is relatively simple: different sounds, with different timbres, result in a different "loudness" experience, even when heard at their exact same measurable SPL.

Is it difficult to understand that you will surely have a more pleasant experience listening to the voice of a sexy lady at 0 LUFS K14v2 MLk, than to the sound of a shrieking dentist drill, at precisely that same level? Would you bear listening to the drill as long as to the sexy voice?

See where I'm going with this?

Sounds that contain many high frequencies are experienced more aggressively than lower sounds, even when at the same level (as I know you're aware of), but even sounds with many high frequencies are experienced in different ways, depending on their character, timbre, or even sexual, psychological, emotional or political content! There's is simply no absolute technical way to measure this!!

Also, there might actually be a couple of knuckelheads who actually prefer to listen to the drill, for whatever reason. THAT's how much things can be different for different people...!

In short: with the K14v2 proposal you're trying to impose absolutes where there are no absolutes!

The only practically useable absolute we are dealing with is how much dB fit into a digital scale. We only want to make sure we are not going to deliver a master that will have a distorted sound. But you know what, some folks LIKE distorted sound, even if it's digital (yuck!).

That's the way of this world my friend... It's futile to want to change all and everything and impose one and one size only that should fit all.

If some people want to compress, limit and/or distort the audio of a music (or other kind of) track to death, then by all means, let them do it. At the other end, anyone has the freedom to hit that "OFF" button, unless being at Guantanamo bay or similarly hellish place (did you know they actually use music there to torture people?).

On the other hand, let people who know better and who have better taste deliver quality mixes, and let the quality do its work. Maybe that quality norm demands that no compression or limiting be used, maybe it does demand some of it, depending on the track. Both cases should not have the same leveling criteria be applied to at the user's end. Also, it should again be mentioned that tracks may be meant for very different listening conditions and different kind of monitors, which is why a club mix differs from a radio mix... Why on earth would you want to give them the same loudness/leveling treatment?

Let's not forget that the EBU R128 -23 LUFS norm has essentially resulted from the abusive compression/limiting practices used for the audio of TV and radio commercials, and its annoying effect on the audio of a broadcasting program, in the long run. Nothing to do with mastering music/audio of CD or (other) digital releases, really. It's a totally different issue: how to make TV & radio programming that is reasonably well leveled for the average consumer. Even people like Florian Camerer admit (in one of the videos) that the result is not perfect. It's just an objective way to avoid extreme disturbance from uber-compressed audio, and to obtain an objectively reasonable average, not an ultimate great listening experience.

Simply applying the old K-12/14/20 system norm will do just fine to deliver great music masters. And to answer your initial question: I will stick to working with it, together with the TT Dynamic Range Meter, mastering at a TP of (around) -1 dB FS.

On the other hand, the meters offered by Tonebooster's EBULoudness, it be the -23 LUFS EBU R128 or K14v2, both SLk and MLk meters, can definitely be useful tools in balancing tracks that will go together on the same release. Again: not perfect, but useful for sure. They can give you a fair idea of where things are situated.

It appears to me that the only viable (and democratic - yes, it's political as well) solution to remedy abusive uber-compression/limiting of CD or (other) digital masters lies at the consumer end.

Leave leveling ultimately to the channels who are broadcasting, it be on TV, radio, internet, or other.

At some channels/outlets music will be broadcasted without anything being changed, while other broadcasters will carefully level it to taste, or add (even more) heavy compression and limiting. Let it be THEIR business. They have clients to please, so ultimately will have to offer the best solution they can come up with, or are willing to.

That said, personally I find it perfectly normal to adjust the volume level when switching between CDs, channels or programs. Actually, I generally even extensively adapt EQ settings, and may add some limiting (or reduce it / take it off - generally have none on music audio though). But then again, I might be on the freakish side with that...

For people who are too lazy to bother with such a mundane matter, or who simply have no time or skill for it, I suggest that producers of audio gear should develop extra technology. This will actually be great for the audio gear industry. New functions/tech are always great for sales!

For this, it seems an audio processing solution could be applied that I occasionally used to work with, many summers ago. It's called a "compellor". It offers a mix of automatic gating, compression, limiting and (non compressed) leveling. This offers an astute mix of functions to handle big level variations.

Check out the Aphex Compellor 320D:
http://www.aphex.com/aphex-products/320d-compellor/

With video presentation at

(no audio examples though )

Incidentally, this Compellor was proposed as a solution for precisely the same reason EBU R128 was proposed, at the demands of a radio broadcaster who wished to adjust levels between normal programming and ads.

With advancing chips and technology, I'm sure it's already feasible to have an economically viable application of this concept in audio gear.

As a matter of fact, it's one of the few classic boxes I'm still waiting to see as a VST version of. I'm sure something will come up soon enough.

So there it is: maybe this is the ideal solution for dealing with unreasonably leveled music audio.

Always think about the music FIRST!

All the best

[williedscarface]

What's wrong with you people? If you don't like something, don't listen to it? I suppose your next move will be installing cameras in studios to make sure no one has a drink or gets high, thus preserving the integrity of the recordings? If Metallica wants to record an album of ear-piercing distortion, that's their business. Mind your own.

[Sounddigger]

What's wrong with you people? If you don't like something, don't listen to it? I suppose your next move will be installing cameras in studios to make sure no one has a drink or gets high, thus preserving the integrity of the recordings? If Metallica wants to record an album of ear-piercing distortion, that's their business. Mind your own.

In case this was not yet clear: I agree fully with what you're saying. And the technical solution I suggested (i.e. the compellor) is obviously meant to have an on/off switch, of course...

[Compyfox]

Phew, definitely a lengthy post, Sounddigger.
Let's see if I can get to all of this.


First and foremost:
Yes, I did use the K-System (both v1 and v2) in real world scenarios. Several times by now. And it's working like it should.


For example:
In theaters (at the cabaret theater I worked a couple of years ago), the engineer didn't have to mess a lot with volumes anymore, but could rather focus on the show. With commercials (that I do the most recently), I usually don't go higher than K-12 - but on the long run it's easier material for broadcast compressors. With music, I worked in both K-14 and K-12, up until the Amber Zone +2 (or K-12 AZ+2 in short). It was still competitive, though a tad more quiet than say -5dB RMS productions like Metallica or even certain other speed metal material.



I understand your fears and issues with having such a production, and constantly reaching for volume knobs. But the positive side effect is that you keep the transients of the original production, rather than having a squashed sound. And, connected compressors/limiters (at home, at a club, whereever) don't have to work that much.

For example (yet again):
The AC3 format only has a certain loudness limit. Make sh*t as loud as you want and all you get in return is constant pumping sound on the disk. Not good. On the other hand, have the same song collection on DVD (like in DTS 2.0, or AC3 2.0) and then put in the 1:1 CD which was however loudness maximized... then you have to(!) reach for the volume knob.

Both has it's good and bad sides. Though I do not agree, that this is an artistic decition. Because you can be just as artistic at -18dB RMS, as you are with K-12. The difference is the loudness.



I also agree that the EBU R-128 -23LUFS proposal for music (as it was initially made) is far from perfect. With recent AES debates, this was adressed. Even -20LUFS. Hence my middle way with K-16v2, which came to me by chance btw. Then again, this is somewhat going back to recordings, or better said engineered releases, from end 90ies.

My proposal is not a completely new standard. It's built upon the EBU R-128 (or ITU-R BS.1770-2) ballistics, but with the same concept as the K-System v1. It's a natural evolution. Something that Mister Katz already hinted at in some of his posts shuffled over the internet, but in written form from my side. There is no drastic change. Other than a fourth K-System indicator: K-16v2, which correlates with the rumored iTunes limit.



So what is the right thing to do?

Personally I think we should stick to either K-System regardless, though I strongly recommend to use the K-System v2. Hence my time table proposal. Currently I stick to K-14v2, and it's working fine. I also see a drift towards that direction in the industry. Which is a good thing.

Now why do this?

You said in your post, that the broadcast stations should take care of the loudness in terms of streams. While I can agree with that, there is one bit of information missing. Broadcast studios asked (in the old days, sometimes even today) for non-mastered versions in order to have not as much load on their studio compressors. Or the need to mess with faders.

To some extend, this is automated nowadays with loudness normalisation tools, or at least an automated loudness normalisation process that pulls down the volume by at least -6dBFS prior to playback. No matter at which loudnes the original product resides.


You can try the effect yourself if you're using a multi band compressor that has maybe a so called "FM" preset. Unfortunately I don't have the old JB Broadcast Compressor anymore. By that, I mean v1. This one had a FM preset that was similar to compressors like the Orban OptiMod broadcast compressors.

But it clearly showed what happened if you ran a hot signal through it, compared to a more moderate loud one. Ideally, the loudness output was the same, only that the more moderate loud one was not as drastically compressed/limited.

And this is exactly what certain engineers that support the K-System (either version) try to aim at: provide healthy engineered productions. Also due to several reasons: retained transients, less distortion, usable on several kinds of mediums (DVD Audio, Vinyl, CD, Tape, MP3), broadcast ready without providing a second mix.


This is one thing.


Of course I also understand the problematic with loud listening environments. Like on the street, in a car, on a train/airplane. But for this, the Music Loudness Alliance is already working on a solution. And that one is called "Loudness Normalisation on Playback".

But even here, the limits are apparent. The hotter the original signal, the less transients, the softer it will sound compared to a full dynamic, loudness normalised production. This is however the thing with "broadcast" and "playback" - currently not our concern. To prevent this however, we should fix that at the source: music mastering.



OUR concern, as engineers, should be to drive back on loudness, stop bending the rules of technical limitations, focus on music again rather than making sh*t go "boom".

Since I adapted the K-System, I literally dropped multiband compression and strong limitation. With the K-System v2, I only use the limiter for catching rouge peaks and maybe raise the loudness by 2-4dB at max. Nothing more.

Though then again, I also use a proper gain staging process while mixing. But this lies on a whole different ballpark.




I don't force anyone to go this route. Or even believe me. But personally I think the listening experience is much better, more relaxing to your ears. Granted, not always ideal (loudness wise), but always better than distorted material which could have prevented in the first place.

Now... if you stick to the K-System v2 (which, again, is an evolution of the K-System v1), and still use an APHEX Compellor (or something like Melda Production MAutoVolume) for controlling the loudness of a certain stream, you're pretty much on the current "future track".

Of course you still need to use your ears and trust your instincs as to what is perfect for a CD release compared to an individual release. Then again, technology improves drastically - but this is what audio engineers are for after all: to port over the feeling of a production.

Else we'd all be jobless by now.

[Sounddigger]

Phew, definitely a lengthy post, Sounddigger.
Let's see if I can get to all of this.


First and foremost:
Yes, I did use the K-System (both v1 and v2) in real world scenarios. Several times by now. And it's working like it should.


For example:
In theaters (at the cabaret theater I worked a couple of years ago), the engineer didn't have to mess a lot with volumes anymore, but could rather focus on the show. With commercials (that I do the most recently), I usually don't go higher than K-12 - but on the long run it's easier material for broadcast compressors. With music, I worked in both K-14 and K-12, up until the Amber Zone +2 (or K-12 AZ+2 in short). It was still competitive, though a tad more quiet than say -5dB RMS productions like Metallica or even certain other speed metal material.

OK, fair enough.

However, let me ask you: what kind of music did you play at that/those occasions? Was it mostly in the same genre, or was it very varied?

The reason I ask is that your comment spurred me to give your system another good solid shot, to see if I could make it work. But to no avail...

This time, I tried to apply strictly the K14v2 system (using Tonebooster's EBULoudness meter), but in a somewhat more loose way, which would be in the advantage of the more dynamic tracks. In my earlier tests I did in fact a kind of peak alignment, where I'd align the absolute peak of each track to the K14v2 MLk 0 level. However, that doesn't really seem to be in the spirit of what your system proposes, as it should be acceptable for the level to go into the "amber" zone, as you call it. So this is what I did, looking for an average level of K14v2 MLk 0 LUFS on the majority of each track, and for the genuinely dynamic tracks I went as far as accepting a "true peak" for fff passages, at just below the (red) +3 level (on the MLk meter).

Well, this still doesn't work for me. I simply find the result not really musical.

Sure, it's a way to get levels kinda close, but I find the results hopelessly unsatisfying.

The simple reason for this is that tracks that are genuinely dynamic, and that have a genuine crescendo (God forbid!), are automatically punished, as the forte passage(s) will make it so that the tracks are leveled too low to stand a good chance at ending at a level that feels comparable to ubercompressed tracks, or even tracks that are not all that compressed, but that are simply rather flat throughout the track.

To illustrate the latter, I have a track by Kraftwerk "The Robots" (from an early 1990's release) which has an average dynamic range of about 14 (on the TT meter), which is pretty good. On the other hand, M. Jackson / Bruce Swedien's "Jam" has an average dynamic range of 15, and often much larger. But once you align the tracks in the most favorable way to the K14v2 system, "The Robots" track will always be louder, simply because it is mostly flat throughout the track, while "Jam" is not only ridiculously dynamic, but also has a significant crescendo going on from beginning to end! The same is the case when compared with other tracks that are more compressed/limited, for the same reasons.

The same problem occurs for aligning / "normalizing" classical music, the moment it has some kind of crescendo going on with a couple of fff passages. Either the level of the non-classical tracks has to go down substantially, or the classical tracks have to go a significant amount into the red zone.

So no, in short: I just can't make it work. By the way, I played around with this using it on at least 17 tracks.

My conclusion: the level of masters should merely be set so that the sound quality is optimal (i.e. no distortion, and such, and with good dynamics such as in the old K14 way). Comparative levels (between different productions) must be left to the programming for which these productions are used.

I understand your fears and issues with having such a production, and constantly reaching for volume knobs. But the positive side effect is that you keep the transients of the original production, rather than having a squashed sound. And, connected compressors/limiters (at home, at a club, whereever) don't have to work that much.

Well, if there's one master-engineer who is quite obsessed with transients, then it's surely Bruce Swedien. Let no one dare touch his transients!! Believe me... And I have not heard him complain that his beloved transients ever got lost on any CD that was released of his masters (unless I missed something). So I wouldn't say that mastering as hot as he does would result in any significant loss of transients, even when released on a CD at (slightly) over TP 0 dBFS. And I'm still using material from ancient sample CDs, that is uncompressed but often normalized or close, and full of great transients (if well recorded). So no, this doesn't fly with me...

For example (yet again):
The AC3 format only has a certain loudness limit. Make sh*t as loud as you want and all you get in return is constant pumping sound on the disk. Not good. On the other hand, have the same song collection on DVD (like in DTS 2.0, or AC3 2.0) and then put in the 1:1 CD which was however loudness maximized... then you have to(!) reach for the volume knob.

Those are interesting arguments, but I'm only talking about stereo mixes here. Also, these kind of arguments are moving away from your initially stated goal - as quoted from the opening post:

Last year, at the end of August, especially Europe and a lot of selected countries all over the world adapted the so called "EBU R-128" loudness standard. In short, it's a standard intended for a unified loudness while broadcasting. This saves you from heavy loudness jumps while watching TV, or even at cinemas. Even though it is still happening to a certain extend.

The thing is however, that the music realm was almost left out of the context. At first, it is totally logical since this standard is aimed at broadcast environments, not engineers. But the Loudness War doesn't find an end, if only the broadcasts are normalized.

And your test in the theater was about normalizing levels between tracks, not about what happens when you go from one format to another. With the latter come specific technical issues that should of course be dealt with, but only once such a transfer to another format occurs.

You said in your post, that the broadcast stations should take care of the loudness in terms of streams. While I can agree with that, there is one bit of information missing. Broadcast studios asked (in the old days, sometimes even today) for non-mastered versions in order to have not as much load on their studio compressors. Or the need to mess with faders.

To some extend, this is automated nowadays with loudness normalisation tools, or at least an automated loudness normalisation process that pulls down the volume by at least -6dBFS prior to playback. No matter at which loudnes the original product resides.

It's not only about compression, it's also about timbre, pitch, crescendos / decrescendos, emotional content, and so on, even when there's no compression / limiting at all.

Again, you're trying to put absolutes where there simply aren't. At least, not as much as technicians would like to think there are. Sure, you can align and "normalize" levels sorta, but there's not an absolute way to do it as your system would imply if it did what is claimed. However, even when applying your system there can only be a (somewhat) arbitrary result. For instance: how much "amber" zone is acceptable? This will never be sorted out. And even if you do it in a pretty standardised way, the effect of normalizing levels of music masters in this strictly meter driven way is just terribly unsatisfactoy, in my opinion, and far from the desired goal of being able to enjoy levels that are adjusted in a way that feels balanced. Actually, I'm quite surprised as to how unsatisfactory it is, after having seen & heard the example(s) given in the videos. I was really hoping this approach could offer a good way to balance volume between tracks. But clearly, music is simply not to be compared with dialogue, real-world noises and sound fx, especially when looking at how it behaves over time.

It's literally like the proverbial comparison between apples and oranges. Both are fruits that grow on trees, look very much alike, and taste great, but pressing them into a juice results into totally different stuff.

Still, the MLk and SLK meters can be a great help as a starting point, offering a visual aid for analyzing what's going on. And when leveling several track that are very much alike, same genre, and relatively flat, it can be enough to get the levels right.

You can try the effect yourself if you're using a multi band compressor that has maybe a so called "FM" preset. Unfortunately I don't have the old JB Broadcast Compressor anymore. By that, I mean v1. This one had a FM preset that was similar to compressors like the Orban OptiMod broadcast compressors.

But it clearly showed what happened if you ran a hot signal through it, compared to a more moderate loud one. Ideally, the loudness output was the same, only that the more moderate loud one was not as drastically compressed/limited.

It can still be found at
http://www.jeroenbreebaart.com/audio_vst_jb.htm

I guess you mean the preset called "AGC+Limiter". It significantly dulls the audio, in a typical "FM" channel way...

However, the problem you describe is purely a level issue. Turn the level down before putting the audio through (or raise the AGC threshold), and you avoid the activation level. A track that is already heavily compressed/limited will come out as the winner, loudness wise (even with dulled transients)...

And this is exactly what certain engineers that support the K-System (either version) try to aim at: provide healthy engineered productions. Also due to several reasons: retained transients, less distortion, usable on several kinds of mediums (DVD Audio, Vinyl, CD, Tape, MP3), broadcast ready without providing a second mix.

I'm all for it. But good sound can be made even with TP at 0 dBFS (though preferably a bit below, to avoid inter-sample peak calculation errors).

Now... if you stick to the K-System v2 (which, again, is an evolution of the K-System v1), and still use an APHEX Compellor (or something like Melda Production MAutoVolume) for controlling the loudness of a certain stream, you're pretty much on the current "future track".

That's a great tip. hadn't seen it yet. The Melda plugs are great. They have really nice meters as well. The EBU meter works quite well and is a bit nicer to work with for aligning short sounds than TB's (it seems a bit slower). I've planned to get their full bundle one of these days.

Of course you still need to use your ears and trust your instincs as to what is perfect for a CD release compared to an individual release. Then again, technology improves drastically - but this is what audio engineers are for after all: to port over the feeling of a production.

Else we'd all be jobless by now.

As far as loudness goes, there's clearly an "artistic" dimension (or should I say merely "sensory"?) that seems to be very hard to capture in technical terms. Rather interesting aspect of all this, for sure.

Probably the best thing is to limit technical decisions to where the technical aspect really counts, and not be afraid of this intangible and "artistic" (i.e. human sensory) side of things.

Also, I for one would not be able to work with a recording engineer who has no genuine artistic side to his work.

After all: life IS art, or it would be hardly more than a bore...

[Compyfox]

Lot of questions, a lot to answer. Let me try...

However, let me ask you: what kind of music did you play at that/those occasions? Was it mostly in the same genre, or was it very varied?

The reason I ask is that your comment spurred me to give your system another good solid shot, to see if I could make it work. But to no avail...

Funny enough, varying genres. And yes, I can confirm that there can be discrepancies in terms of meter outputs and going by plain numbers.

Even with the EBU R-128/ITU-R BS.1770-2 meter ballistics, what the K-System v2 is based upon as well, you can get varying readouts. Sometimes even drastically.

As you found out yourself, this happens depending on the density of the stream. Then again, I only encountered that with already processed material. Or if I had to align recordings from pre 70ies with mixes from post 90ies.

With unprocessed (however frequency balanced) material, or material that is not peak limited (read from the last 5-10 years), the variation is not as drastic and you can work by a "rule of thumb" with the K-System.

The same problem occurs for aligning / "normalizing" classical music, the moment it has some kind of crescendo going on with a couple of fff passages. Either the level of the non-classical tracks has to go down substantially, or the classical tracks have to go a significant amount into the red zone.

So no, in short: I just can't make it work. By the way, I played around with this using it on at least 17 tracks.

My conclusion: the level of masters should merely be set so that the sound quality is optimal (i.e. no distortion, and such, and with good dynamics such as in the old K14 way). Comparative levels (between different productions) must be left to the programming for which these productions are used.

Yes and no.

It is possible to go by the plain K-System (either of them btw), but you have the problem with the crescendo and drastic volume jumps. I clearly showed that with the track comparison from a couple of pages back (Sting - English Man in NY, this one is jumping heavily).

I created this by accident, but the light green zone is ideally for that. If the music is within that section, and the ff to fff parts go up until red, then it's ideally mixed and only needs fine tuning.

But... you still need to use your ears and trust your gut. This is a setback that I maybe need to write down in the PDF.

So I wouldn't say that mastering as hot as he does would result in any significant loss of transients, even when released on a CD at (slightly) over TP 0 dBFS. And I'm still using material from ancient sample CDs, that is uncompressed but often normalized or close, and full of great transients (if well recorded). So no, this doesn't fly with me...

I have the HIStory Collection by Michael Jackson, and this one is fairly hot to begin with. Granted, not as drastic as modern productions. But it is peak limited already.

Still it sounds fairly good. One of the prime example where it is a good balance of a loud, yet still dynamic CD. Though I find "Scream" (the duet) really a bit dull sounding from all the tracks on the CDs.

Those are interesting arguments, but I'm only talking about stereo mixes here. Also, these kind of arguments are moving away from your initially stated goal - as quoted from the opening post:

...

out normalizing levels between tracks, not about what happens when you go from one format to another. With the latter come specific technical issues that should of course be dealt with, but only once such a transfer to another format occurs.

Not quite.

First and foremost: DTS and AC3 are lossy encoders, unless we talk DTS HD Audio (which is lossless) or HD-AAC (the same). The 2.0 means stereo without additional sub frequency channel. It's just a different encoded audio stream, but channel wise it's the same. And this is what I was aiming at, and I wanted to show certain limitations.

Then regarding the work at the theater - I had to juggle with both CDs (redbook) and MiniDisk's (ATRAC codec). Or better said prepare for both formats. Prior to my edits, the playback of the shows were a constant reach for the faders rather than "set once and then forget".

But in order for this to work, I did of course need to take into consideration how loud each track is compared to each other. Or if one track lacks bass (recordings from pre 70ies) while others were overpowered (mid 90ies and forward). Read: I had to manually adjust the stream, fix really old recordings or pull them into more modern times. Some tracks were a tad hotter on readout, some not. But the overall loudness was pretty much constant in the end.

And this is where the Integrated measurement comes into play (which works fine with streams longer than 15 minutes and different tracks) additional to proper set ML values.


At least to my experience. Depending on the limitations of the streams of course. But if we go by plain "modern" technology, it should work.

But my focus is still ML for individual tracks. Integrated only for CD montages or podcast's as additional measurement tool.

... However, even when applying your system there can only be a (somewhat) arbitrary result. For instance: how much "amber" zone is acceptable? This will never be sorted out. And even if you do it in a pretty standardised way, the effect of normalizing levels of music masters in this strictly meter driven way is just terribly unsatisfactoy, in my opinion, and far from the desired goal of being able to enjoy levels that are adjusted in a way that feels balanced.

This was and still will be a constant dilemma. Music is subjective, how long a part is mf, one is ff, one mp, one fff - this is subjective.

The TT-Dynamic Range Meter can't measure that, just like the EBU R-128 meter can't do that either. Each user has to find a middle way.

As earlier mentioned, I usually use the light green area (K-14v2) for mf passages, ff passages don't go higher than AZ+1 for me, and fff usually don't exceed +3. Though I had one production where it went up to +5, but average it was still within... er... specs.

Actually, I'm quite surprised as to how unsatisfactory it is, after having seen & heard the example(s) given in the videos. I was really hoping this approach could offer a good way to balance volume between tracks. But clearly, music is simply not to be compared with dialogue, real-world noises and sound fx, especially when looking at how it behaves over time.

I can agree to that by plain EBU R-128 metering (where the Integrated value is most important).

But if you're so dissapointed by the K-System v2 (ML meter, color codes), why aren't you of the K-System v1? Only the ballistics differ.

Still, the MLk and SLK meters can be a great help as a starting point, offering a visual aid for analyzing what's going on. And when leveling several track that are very much alike, same genre, and relatively flat, it can be enough to get the levels right.

This is why I have a focus on the MLk meters. If each metering tool would have histograms for ML rather than SL, then loudness "aligning" would be a tad simpler. Though I already contacted the one or another company to implement that. Or port an affordable offline tool.

It can still be found at
http://www.jeroenbreebaart.com/audio_vst_jb.htm

I guess you mean the preset called "AGC+Limiter". It significantly dulls the audio, in a typical "FM" channel way...

However, the problem you describe is purely a level issue. Turn the level down before putting the audio through (or raise the AGC threshold), and you avoid the activation level. A track that is already heavily compressed/limited will come out as the winner, loudness wise (even with dulled transients)...

A matter of preference, really. Then again, I never liked multiband broadcast systems. Oh and... I meant v1 of this plugin. The original version had 4 bands.

I'm all for it. But good sound can be made even with TP at 0 dBFS (though preferably a bit below, to avoid inter-sample peak calculation errors).

But if you go by plain peak values, and maybe PLR (peak to loudness ratio), or in case of the TT-DR Meter, the DR value of >12, then you still have differing loudness.

Well, at least you don't go by plain peak normalisation.

That's a great tip. hadn't seen it yet. The Melda plugs are great. They have really nice meters as well. The EBU meter works quite well and is a bit nicer to work with for aligning short sounds than TB's (it seems a bit slower). I've planned to get their full bundle one of these days.

IIRC, the MLoudness plugin has a fixed reference level and doesn't stick to certain specifications (especially with TP). Unless that was fixed by now. Else, nice plugin - definitely on the right track.

If the TB plugin seems slow to you, and you're using Wavelab for measurements - this seems to be a bug that Jeroen Breebaart couldn't track yet. We both assume it's due to the Wavelab engine. It works fine in most other hosts.

If you want a cross-platform one, freeware (though stand alone, but usable with every recording card I know) and fully configurable measurement tool, then maybe take a closer look at the ORBAN Loudness Meter.

As far as loudness goes, there's clearly an "artistic" dimension (or should I say merely "sensory"?) that seems to be very hard to capture in technical terms. Rather interesting aspect of all this, for sure.

Probably the best thing is to limit technical decisions to where the technical aspect really counts, and not be afraid of this intangible and "artistic" (i.e. human sensory) side of things.

Also, I for one would not be able to work with a recording engineer who has no genuine artistic side to his work.

After all: life IS art, or it would be hardly more than a bore...

I can definitely agree with this.

Though, we need to find a middle way. And fusing the K-System with the ballistics of the EBU R-128 meters is, IMO, a good start.

[tanabarbier]

I follow this thread since post one, and with a lot of interest, I wanted to thank Compyfox and Sounddigger to create and feed this realy interesting debate.

The note about crescendo made me think about a situation I recently lived, it is slitghly (or totaly?) off topic but has to do with the subject too:

I played a show in a really nice sounding place, with really good PA and everything, and the result was awfull, at least to me, because of a sad intent of normalizing loudness from the sound ingeneer.

When doing the soundcheck I told him:

"My set has a big dynamic, it starts at about -30 dB rms and finishes at about -3. I have a limiter on my master track, wich works only for a few peaks and the final 3 mn (crescendo until explode). What I need is steady level, no EQ, nothing."
He told me that it is really unusal for electronic music, I said, well yeah, it is more like electroacustic.

And he putted a compresor that I heard pumping through allmost the whole show.... Wich made me incapable of playing with the volumes like I had to, and I played horrible! Everytime a sound came a bit up, all the others went down, etc etc. It is HORRIBLE!!!

Ok so, sorry for the off topic (but a bit in topic, or not?). My question is: How can I avoid that kind of situation to occur once again? What can I say, or do, to avoid that?

Thanks again!

[Sounddigger]

So can the EBU R128 standard successfully be used to balance music, or not?

Being intrigued by the whole EBU R128 proposition, I dug in still further today/night to clarify the issue for myself, and decided to do a bit of "reverse engineering": I took a number of tracks, balanced them in a way I felt was "right", then carefully looked on the (Tonebooster's EBULoudness) meters to see what could be the method behind it.

Yes, there actually seems to be a way to use the EBU R128 standard in a way that gives results that are musical and quite satisfying, but no, it's (mostly) not as simple as merely aligning peaks on a given level of the MLk K14v2 scale (using Tonebooster's EBULoudness meter), and yes, it needs a bit of artistry to get it right.

I found that combining the MLk and SLk readings of the 16kv2 scale of the Tonebooster's EBULoudness meter can give quite excellent results when aligning tracks to a level that will be experienced as homogenous, for music of ALL types. It could be done with the basic R128 scales, but the 16kv2 scales proved to be the most useful and universally applicable.

Say you want to do a big mix of tracks of all sorts, or even taken from the same genre, then this offers a way to align the levels in a pretty fool-proof way, without adding the use of compressor, limiter or "compellor" type devices.

This is not about introducing a modified standard to master CDs or (other) digital releases, but for using such material into a mix, it be for an album, a compilation or for use in an AV production.

The trick is to make a difference between tracks, and between the parts within these tracks, then align levels using a consistent method. Sounds simple? It is somewhat, once you work it out and put some practice in it.

For EDM, EM, Pop and Rock, the best results are obtained by setting maximum acceptable average and peak levels for (what would correspond to) the verse of a song/tack, and another set of average and peak levels for (what would correspond to) the chorus/bridge/hook of a song/tack. This is simple enough on itself, and it works!

Take any number of tracks that fall in this category, and adjust the level of each track in the way that I will describe next.

First, if you haven't already, then download Tonebooster's EBULoudness meter at http://www.toneboosters.com/tb-ebuloudness/
You can buy it, but even the demo can be used with practically no limitations, only settings can't be saved with it.

Find the menu top-left of the window of the plugin called "Meter mode", and select "K16v2".

Note that you'll have to frequently click the "Reset" button, to refresh the readings of the MLk and SLk meter, to set levels accurately throughout the track.

For adjusting the level of a track, the verse part is the most important part. Get this right, and the rest follows. During the verse you should see an average level of 0, and absolute maximum of +1 on the Kv16 SLk scale (so a level of 1.1 is NOT acceptable here!). As a guide for when there are some dynamic elements, such as a voice: set the groove and "pad" of the verse peaking at 0 SLk, and let a dynamic voice or other instrument peak at maximum +1 SLk.

The MLk level should be on average at +1, but the more dynamic a track is, the more this can jump around (no real limit here, the correct average SLk level should imply that MLk doesn't peak excessively).

Make sure you've got indeed the verse (or the part with what amounts to the feel of a verse), and that you're not already looking at the levels of what is a bridge/chorus/climax. With uber-compressed EDM tracks it's particularly easy to get confused, especially when there is very little change of theme.

And here is the trick to handle uber-compressed/limited tracks: while for a dynamic track it is acceptable to see (during the "verse") the level jumping up and down between 0 and +1 on the Kv16 SLk scale, if a track is uber-compressed, you should keep it at an average of around 0 on the Kv16 SLk scale, which will be almost the same as the peak level, due to the uber-compression, unless there is a temporary peak-sound, such as a loud voice or hit that has not been smashed by the compression/limiting. The way to verify that the setting is correct is to look at the MLk scale. The uber-compressed track will/should have an average level around +1, or somewhere between 0 and +1 on the MLk scale, without ever jumping over (approximately) 1.5.

Typically, for an uber-compressed track you will see very little difference between the verse and (what would correspond to) the chorus/bridge/hook of a song/track, especially in today's EDM tracks.

If the level of the verse is set right, then leave things where they are for the rest of the song/track, it will correspond to other material that is aligned this way. This way of aligning the verse works universally, both for normal very dynamic tracks (such as a B. Swedien mix) as well as for uber-compressed tracks!

It even works very well for finding the right level for Metallica's "The Day that Never Comes"! What's even more intriguing, I found that this track is not quite as "anti-dynamic" it is generally accused of. For sure, the track is compressed/limited to death, and the wave looks (mostly) totally uniform, but there's actually an excellent dynamic difference between the verses and the chorus/bridge parts. Check it out, you'll see what I mean! You should see the level on the SLk K16v2 scale move from average 0 to a peak at +3 on the SLk scale, when going from the verse section into the chorus/bridge section!

Due to this huge dynamic difference, I would actually classify it as a "dynamic" track, much more so than typical tracks by for example deadmau5 and A. v Buuren, of which I also checked and aligned some tracks, and which show hardly any crescendos from verse to bridge/hook (aside from the break-downs and intros/build-ups), but which still sound good in their own typical uber-compressed way though. Of course, once the Metallica track goes into double-tempo, it never comes down again, so there's something to say for treating the intial verse as a "cheat" verse, and bring the whole track down even more. But setting the levels as described here seems to feel "right" when aligned with other tracks.

Interestingly, I found that what seems the optimum peak level for chorus/bridge for very dynamic tracks to be precisely +3 on the SLk scale, with on the MLk scale a maximum peak of +6 to +7 (in rare cases higher is acceptable for VERY dynamic mixes). An uber-compressed track will obviously not get such a high peak, and Metallica's track accordingly peaks at around 3.5 MLk. If there's a peak that goes higher than +3 SLk, than overall level must be lowered (but for orchestral classical tracks I found maximum peaks of up to +4 SLk can be the "right" level).

There is a bit more to all this, and there are a few tricks that can be used for careful extra tweaking, and I'm still discovering new things, but this is the basic approach that can give totally smooth results.

If you guys want me to, I can elaborate further on this and give a more detailed explanation of how to work with this. It can be quite useful for leveling tracks in a mix, in a universal way, or even for using various parts in a mix.

Aligning classical music is both the easiest and the most difficult. The trick here is to make a difference between leveling an orchestra and small ensembles, and identifying what parts can be said to be mp up to fff. Jazz can also be very tricky, but again, much depends on correctly identifying the parts that work as verse vs chorus/bridge (soloing parts can often be interpreted as verse parts, up to climax sections). In rare cases, tracks can be crazy dynamic, which would either necessitate a squirt of limiting, or bringing all other tracks down by the necessary amount of dBs. Some exotic ambient tracks might best be treated like classical music.

Anyway, I'll leave further details for another time. Let me know if you guys find these kinds of ideas useful!

Also, if anyone finds useful methods that can be applied in a universal way, using the Tonebooster's EBULoudness meter or any other one, then let us know!

Now back to the replies:

As you found out yourself, this happens depending on the density of the stream. Then again, I only encountered that with already processed material. Or if I had to align recordings from pre 70ies with mixes from post 90ies.

With unprocessed (however frequency balanced) material, or material that is not peak limited (read from the last 5-10 years), the variation is not as drastic and you can work by a "rule of thumb" with the K-System.

What I described above should work across all genres, from any time-period. Classical music and Jazz (accoustic or electric) need some special care and extra criteria, and then can generally be made to work.

What doesn't work - for me - is to simply align peaks at a set level on K14v2 (or other) MLk. You have to identify which peaks belong to mf (i.e. "verse" feeling), and which belong to fff (i.e. chorus/bridge/climax). And sometimes, entire pieces are merely mp up to mf. This must also be identified to put it in the right context and make it work with the other tracks.

Those are some of the tricky things... oh yeah

It is possible to go by the plain K-System (either of them btw), but you have the problem with the crescendo and drastic volume jumps. I clearly showed that with the track comparison from a couple of pages back (Sting - English Man in NY, this one is jumping heavily).

I created this by accident, but the light green zone is ideally for that. If the music is within that section, and the ff to fff parts go up until red, then it's ideally mixed and only needs fine tuning.

But... you still need to use your ears and trust your gut. This is a setback that I maybe need to write down in the PDF.

You rightly identified the problem, and worked with the green, amber and red zones. But an ideal approach has to be more methodic. Identify "verses" and "chorus/bridge/hook/climax", and it all gets much more accurate.

First and foremost: DTS and AC3 are lossy encoders, unless we talk DTS HD Audio (which is lossless) or HD-AAC (the same). The 2.0 means stereo without additional sub frequency channel. It's just a different encoded audio stream, but channel wise it's the same. And this is what I was aiming at, and I wanted to show certain limitations.

Then regarding the work at the theater - I had to juggle with both CDs (redbook) and MiniDisk's (ATRAC codec). Or better said prepare for both formats. Prior to my edits, the playback of the shows were a constant reach for the faders rather than "set once and then forget".

But in order for this to work, I did of course need to take into consideration how loud each track is compared to each other. Or if one track lacks bass (recordings from pre 70ies) while others were overpowered (mid 90ies and forward). Read: I had to manually adjust the stream, fix really old recordings or pull them into more modern times. Some tracks were a tad hotter on readout, some not. But the overall loudness was pretty much constant in the end.

And this is where the Integrated measurement comes into play (which works fine with streams longer than 15 minutes and different tracks) additional to proper set ML values.


At least to my experience. Depending on the limitations of the streams of course. But if we go by plain "modern" technology, it should work.

But my focus is still ML for individual tracks. Integrated only for CD montages or podcast's as additional measurement tool.

The integrated meters can definitely be most useful tools for this kind of work. For music it's just not as simple as for general (other) audio of an AV production.

... However, even when applying your system there can only be a (somewhat) arbitrary result. For instance: how much "amber" zone is acceptable? This will never be sorted out. And even if you do it in a pretty standardised way, the effect of normalizing levels of music masters in this strictly meter driven way is just terribly unsatisfactoy, in my opinion, and far from the desired goal of being able to enjoy levels that are adjusted in a way that feels balanced.

This was and still will be a constant dilemma. Music is subjective, how long a part is mf, one is ff, one mp, one fff - this is subjective.

Try to think of it in terms of "verse" and chorus/bridge/climax. That should help.

pp is maybe the most difficult to identify, but needs the least of adjustment as its level is necessarily set by the peaks at higher levels. However, in some cases they might have to be raised in level, although that depends on the liberties that can be taken when using the track.

As earlier mentioned, I usually use the light green area (K-14v2) for mf passages, ff passages don't go higher than AZ+1 for me, and fff usually don't exceed +3. Though I had one production where it went up to +5, but average it was still within... er... specs.

That goes very much in the direction of what I've found to work, but it can be even more methodic.

Actually, I'm quite surprised as to how unsatisfactory it is, after having seen & heard the example(s) given in the videos. I was really hoping this approach could offer a good way to balance volume between tracks. But clearly, music is simply not to be compared with dialogue, real-world noises and sound fx, especially when looking at how it behaves over time.

I can agree to that by plain EBU R-128 metering (where the Integrated value is most important).

But if you're so dissapointed by the K-System v2 (ML meter, color codes), why aren't you of the K-System v1? Only the ballistics differ.

That's ultimately what I will stick to for mixing down, the "old" K-14 system.

However, I have to say that when I'm creating tracks, I don't worry so much about mixing per se and finetuning the output level, as I'm more concerned with correct gain staging and rough individual levels and occasional EQ and compression/limiting. Since I do everything in the box, while what is done OTB is recorded and then kept ITB, any mix can be fully recalled, I can set a different time for mix-down, grouping the work that way.

Still, I think I've figured out now how to make the K16v2 meters work for me, and will use that in any number of ways. It will be VERY useful for leveling tracks that will go together on an album, which is the reason why I dug into this

Still, the MLk and SLK meters can be a great help as a starting point, offering a visual aid for analyzing what's going on. And when leveling several track that are very much alike, same genre, and relatively flat, it can be enough to get the levels right.

This is why I have a focus on the MLk meters. If each metering tool would have histograms for ML rather than SL, then loudness "aligning" would be a tad simpler. Though I already contacted the one or another company to implement that. Or port an affordable offline tool.

Hmmm... I have no problem with the MLk not having a histogram. I'd rather see a small improvement for the SLk, so that it's possible to disengage the peak "hold" function of the cipher that indicates its level, so that it refreshes even when the level goes down.

As it is now I'm constantly using the "Reset" button... (but the "hold" function is useful the other half of the time, of course).

If the TB plugin seems slow to you, and you're using Wavelab for measurements - this seems to be a bug that Jeroen Breebaart couldn't track yet. We both assume it's due to the Wavelab engine. It works fine in most other hosts.

No no, I meant the Melda meter seemed a bit slower. Could be 600 ms ballistic or so. It's helpful for what I was referring to.

If you want a cross-platform one, freeware (though stand alone, but usable with every recording card I know) and fully configurable measurement tool, then maybe take a closer look at the ORBAN Loudness Meter.

Looks interesting, but since it's not VST it's a bit awkward to use on my system, due to the fact that my PC with Win7 is wired so that once I have audio running in my DAW, it doesn't accept to run it through the ORBAN, or to run any other audio device at the same time for that matter. I made it work on Winamp though (and that could well come in handy!), and for other players it probably works as well.

As far as loudness goes, there's clearly an "artistic" dimension (or should I say merely "sensory"?) that seems to be very hard to capture in technical terms. Rather interesting aspect of all this, for sure.

Probably the best thing is to limit technical decisions to where the technical aspect really counts, and not be afraid of this intangible and "artistic" (i.e. human sensory) side of things.

Also, I for one would not be able to work with a recording engineer who has no genuine artistic side to his work.

After all: life IS art, or it would be hardly more than a bore...

I can definitely agree with this.

Though, we need to find a middle way. And fusing the K-System with the ballistics of the EBU R-128 meters is, IMO, a good start.

Definitely appears to be a very good step forward.

I follow this thread since post one, and with a lot of interest, I wanted to thank Compyfox and Sounddigger to create and feed this realy interesting debate.

The note about crescendo made me think about a situation I recently lived, it is slitghly (or totaly?) off topic but has to do with the subject too:

I played a show in a really nice sounding place, with really good PA and everything, and the result was awfull, at least to me, because of a sad intent of normalizing loudness from the sound ingeneer.

When doing the soundcheck I told him:

"My set has a big dynamic, it starts at about -30 dB rms and finishes at about -3. I have a limiter on my master track, wich works only for a few peaks and the final 3 mn (crescendo until explode). What I need is steady level, no EQ, nothing."
He told me that it is really unusal for electronic music, I said, well yeah, it is more like electroacustic.

And he putted a compresor that I heard pumping through allmost the whole show.... Wich made me incapable of playing with the volumes like I had to, and I played horrible! Everytime a sound came a bit up, all the others went down, etc etc. It is HORRIBLE!!!

Ok so, sorry for the off topic (but a bit in topic, or not?). My question is: How can I avoid that kind of situation to occur once again? What can I say, or do, to avoid that?

Thanks again!

Not at all off topic, but right on, it appears to me.

Thanks for the nice words. And yeah, that's really lousy of that engineer to do that to you. He was actually taking the risk of you turning around and doing something really ugly in return, like stopping the sound and pointing to him, with moves telling "this is not right", or so...

Maybe it was more a communication issue though. Did you let him check and hear (i.e. experience) what your final level was going to be at the end of your particular track? If it's just the explosion/climax that is at top-level, that might be a bit short, and you may want to have a logical number of bars previous to that at full level, so that everybody knows that this IS the climax.

On the other hand, if you keep having problem such as you describe, then you could actually start your set with a couple of sounds at full-blast level, so that everybody knows where it's at. It cuts the build up, by giving away some of what you have in store though.

I have to say that the EDM tracks I've checked are very, very uber-compressed, so this is the present trend. But notwithstanding that, intros and build-ups are generally still dynamic to a point.

If you want to blend in, you'll have to work with the reduced dynamic range. So that means that when you're at the most silent, there still must be "enough" for engineer and crowd to feel something is going on that is still part of the show (and not just the sound of some TV set at the neighbours, lol).

If you set the level for your verses (or what amounts to that) the way I described above, a good pp level to work with might be K16v2 scale -3.5 SLk and -3 MLk. If you go lower, that might disappear in the mist...

The RMS scale proved to be very useful as an addition to finding the right levels. If you're starting your set at -30 RMS, it seems to me this is way too low, although you might pull it off if you raise the level fairly quickly to what would be equivalent of what I indicated. On my meters, going from -30 RMS to around -15 RMS in 8 bars could be a good starting point, but that might be much too fast for what you have in mind. If so, then simply start at a higher level. It might help to make sure you get the feel of a "verse", with beats at average level of 0 SLk.

Make sure you deliver level increases in ways that feel logic and that are telling a good story, keep those filters moving and introduce new ingredients every 4 to 8 bars.

You might also experiment with headroom, depending on what is possible. If there's an agreed top-level for beats, you might have that at what would be equivalent to K16v2 scale 0 MLk on average for "verse", up to +3 SLk for chorus/bridge/hook/climax, with explosion/climax peaking up to +6 MLk. That will probably be too much though, and the beat will usually peak during climax around +1 SLk if everything is uber-compressed, but with explosion peaking up to +3 SLk you should have a big bang for your buck

Of course, the EBU scale levels I give here will have to be translated at a higher overall output level for the kind of stage you're at. Between DAWs and stage mixer levels there are a whole lot of variables, which I can't all line up here. You'll have to work it out by experimenting with the meters and adapt the final output level. Know exactly what signal you should feed the stage mixer, and what the absolute peaks can be. You may need reference levels. Maybe have a couple of sine tones available for this as well.

Hope this helps

(pheww... this IS now a lonnngg post for sure... may it be entertaining )

[Compyfox]

Ok so, sorry for the off topic (but a bit in topic, or not?). My question is: How can I avoid that kind of situation to occur once again? What can I say, or do, to avoid that?

In live environments, it's definitely a bit tricky since engineers usually don't have devices like at broadcast environments. Thought that might change.

The pumping however could result due to a wrong setup limiter/compressor. Too strong threshold, too fast release.


There is currently (at least to my knowledge, which is limited for live) no real solution to work with this. Shows still need to be loud, or loud enough. There were some schemes once planned for live as well to limit the maximum output - but it somewhat faded over time.

So I think your only solution is to take care of the levels yourself before you send your mix over to the FOH. Sorry if this is not of much use to you.

So can the EBU R128 standard successfully be used to balance music, or not?

[an excellent explanation for a possible usage following this statement]

This is indeed an interesting concept you show here, and is actually how it was planned to be used. Somewhat at least. Though the focus should be on the ML meters, not the SL meters.

The SL meters (3s measurement time) can give you a more closer readout for short term loudness for individual tracks compared to Integrated, which is a middle way of the whole stream (broadcast). But as you found out yourself, if the SL reaches up to +3, chances are that the ML reaches up to +7 or higher. So the focus should be the ML meter.


And this is where the color codes come into play. Also yes, I do agree that we should focus on the felt loudness of a chorus/verse/bridge/etc rather than plain mf to fff definition. Like in the old days, just without the definition of "loud" from score sheets.



I usually go by this:

I try to find the second to loudest part of a song, and then aim to pull it into the scale (with a gain plugin pre the metering plugin) so that this particular part (which I declare forte-to-fortissimo on loudest peak) hovers around 0LUFS.

The hotspot of the song, or the average loudness, should then be between -3LUFS (MLk) and 0LUFS (MLk) - the light green area.

Should this song have uber-dynamic jumps, I still have a headroom of +3LU on top of the +3LU from the light green area (mf to f), though I don't punish myself if it's going to +4 or even +5 once in a while. If we talk about K-16v2 or even K-14v2, it's not as drastic as with K-12v2 or even lower RMS values.

Funny enough, if you render a mix that was edited with a reference level of -18dBFS (VU), you're almost in the K-16v2 specs already. Chances are that you only need to pull it up by 2-3dB.



I currently use the SL meter only to check the overall loudness of a track.

If there would be a histogram for ML, or a "focus meter" (like with Nugen Audio), better even an offline meter prior to aligning things, then the workflow would speed up drastically. The thing is with either K-System version, unless you don't trust your ears - which can be fooled, it means constant "eyes on the meters".

The EBU R-128 meters were made in combination with broadcast compressor/limiter systems to have a "check once in a while, forget it otherwise" type system. But this doesn't work for music.


IMPORTANT:
I do not(!) use the K-System while mixing. For this I trust my Digital/VU meter combo with a reference level of -18dBFS (VU) and a maximum peak of -6dBFS (digital meter, per channel) or better said -3dBFS (digital meter on the summing bus).

I only use the K-System v2 while mastering or doing audio montages.



But to sum up your question if the EBU R-128 meter ballistics can be used for measuring music. The answer is a definite yes.

You rightly identified the problem, and worked with the green, amber and red zones. But an ideal approach has to be more methodic. Identify "verses" and "chorus/bridge/hook/climax", and it all gets much more accurate.

The thing is, and I think this is what all engineers and bedroom-engineers kind of expected... there is no holy grail. There is no slamming on a measurement plugin and shut off your brain.

You need to think, make decisions - even more then before. You have to think in a musical way, not plain technical. But you need to understand the technical aspects in order to grasp what you're trying to do.

The meters are only a guide, there are some rules to follow. The rest should be in good sense to not bend or even break the rules again.


Porting that into the heads of the users of this measurement system was and still will be very hard.

Hmmm... I have no problem with the MLk not having a histogram. I'd rather see a small improvement for the SLk, so that it's possible to disengage the peak "hold" function of the cipher that indicates its level, so that it refreshes even when the level goes down.

As it is now I'm constantly using the "Reset" button... (but the "hold" function is useful the other half of the time, of course).

May I ask why?
You could use the histogram for that. Or do you hit the reset button since the meters kind of distract you?

Implementing a arithmetic mean might be an idea, but it still doesn't tell us where the ideal mf resides. Though granted, these so called "loudness variance" displays in VisLM (Nugen) and the DP MeterPro (BlueCat Audio) can give us a good estimation of where the most energy of a song resides.

Looks interesting, but since it's not VST it's a bit awkward to use on my system, due to the fact that my PC with Win7 is wired so that once I have audio running in my DAW, it doesn't accept to run it through the ORBAN, or to run any other audio device at the same time for that matter. I made it work on Winamp though (and that could well come in handy!), and for other players it probably works as well.

I'm using a RME Digiface. The Orban can pickup the main playback stream of the recording card without changing anything. I just threw it in there as additional available tool.





Else... Yes, again a lengthy post. But it seems to me that you found a solution for you which can be further optimized.

Definitely some good input for me, which might result in an update of the PDF. Maybe even a slight tutorial in either form how to use the meter "in my opinion".



Which leaves the question with the time table I made earlier.

Currently I recommended to drop K-14v2 in favor for K-16v2 by the end of this year. Since I don't see an imminent implementation or release of Loudness Normalisation devices in modern playback devices yet (at least not by end of 2013), is my idea to go down to K-16v2 ASAP (by 2014) to far fetched?

Or should I/we extend the "grace period" until end of 2014 so that even the squallers can adapt?