KVR Audio

Compyfox · Post by **Compyfox** » Sun Jan 12, 2014 2:39 pm

Again, here the EBU R-128/ITU-R BS.1770 standard really shines. Add to that my K-System v2 concept and that is an excellent weapong for mastering music.

With your last post, I think we can really say that you mean by "average loudness measurement", that it has to be:
a) within a given selected timeframe
b) post recording (read: full mix for example)
c) somewhat static and not measured in real time

Tools that do not do this in realtime are indeed wave editor built in analysis tools (Wavelab comes to mind), also the once available DR-Meter offline tool, that measured the max peak, and the dynamic range from random points in the song over the length of the track.

Tools like EBU R-128 can do this in realtime, who also exist as "offline measurement tools" (depending on the developer, faster than realime), which are just as precise.

So in the end, we can really say:
- Peak and Average signal strength measurement tools are Digital Meter, VU/RMS (needle/bargrapth)
- loudness measurement tools are then ideally: offline tools (wave editor analysis), tools with weighting filter (like EBU R-128 and the DR-Meter, c-weighted hardware SPL meter at 1m distance to the source signal)

Can we agree on that?

aciddose · Post by **aciddose** » Sun Jan 12, 2014 2:45 pm

If you want to call it "average loudness", sure.

I don't actually see where we've disagreed other than that I guess you misunderstood what I meant by "average", since you thought I was referring to standard RMS meters or VUs when I actually meant a "complete true average" of a full portion of the signal.

I wouldn't go with "average loudness" though because it implies that it reliably represents "loudness", which isn't really true. It will only give a measurement which combines several properties of the signal including density and intensity, although if you really want to I guess it's close enough. Whatever people call it, I'd like to see them talking about it more often because I find it incredibly useful.

Tricky-Loops · Post by **Tricky-Loops** » Sun Jan 12, 2014 3:01 pm

BTW, (I don't want to make a thread about it) what happened to Friedemann Tischmeyer and his Pleasurize Music Foundation (and the DR Meter)? On his homepage are only older informations from 2009 and 2011...

Did he gave up the project? Or isn't the DR Meter needed anymore because of EBU R 128?

I haven't seen any update of the DR Meter...

Compyfox · Post by **Compyfox** » Sun Jan 12, 2014 3:19 pm

aciddose wrote:I wouldn't go with "average loudness" though because it implies that it reliably represents "loudness", which isn't really true. It will only give a measurement which combines several properties of the signal including density and intensity, although if you really want to I guess it's close enough.

I just renamed it to Loudness Measurement Tools.

By your definition, the debated "true" average measurement would mean pure offline measurement, or measurement post program stream (the result), which the EBU R-128 meter is capable of (histogram) - again, if I understood you right.,

You should really write than down in an ultra simple form.

Tricky-Loops wrote:Did he gave up the project? Or isn't the DR Meter needed anymore because of EBU R 128?

I haven't seen any update of the DR Meter...

The DR-Meter was updated by Brainworx, with the once recommended weighting filter. The DR meter "idea" is now also ported into the EBU R-128/ITU-R BS.1770 concept. So it is indeed not a bit obsolete.

Certain ideas of the DR meter were hard to pull off to begin with (see my KVR marks), and didn't really make sense on the long run. Now it's basically just an indication how "big" the dynamic range of a program stream is.

kylen · Post by **kylen** » Sun Jan 12, 2014 3:44 pm

As long as 'The Machines' are beginning to adjust the volume of program material in the Post-Loudness Wars era we might as well just be given a plugin that Sound Check or Replay Gain or whatever governance uses. The ultimate AGC only now it's legal and will be controlled by machines, I'd hate to get fined for broadcasting some illegal underground loudness units...

EDIT: Speaking of Replay Gain...I guess I'll try out this Audacity plugin at some point:
http://forum.audacityteam.org/viewtopic ... 42&t=63067
This one is an implementation based on the 'standard' so I assume it could vary a bit.
Would be interesting to see how it stacks up against ears and other metering discussed.

kylen · Post by **kylen** » Sun Jan 12, 2014 6:44 pm

aciddose wrote:
Likewise with "8 feet away from a rock band sound system".

What the thread is about is adjusting the final product (the master) so that the same relative intensity / perceived loudness is going to be apparent to the listener on the same system, say a pair of ipod headphones.

@aciddose - What I meant by that is that I prefer not to set perceived loudness, or listen to tracks with perceived loudness that have been set to sound as if a band was overplaying the room. For me that's a distortion that can't be cured with a volume knob on playback.
I wasn't clear about that.

There's a lot to this topic but I think some of yours and compfox's fine posting on the topic help along with other folks in these type of threads. Especially the details on ITU and metering (I have Toneboosters so that's what I'm trying to use more and more: http://www.toneboosters.com/tb-ebuloudness/).

A while ago compyfox posted some demos (I think in his KVR remarks) that set levels based on the ITU/Kv2 that confused me a bit (because they don't go digital full scale, only TP -1LU or something), the perceived loudness sounded good but the rms (if pushed to full scale) would seem to be a louder reading than I shoot for, that's because I'm used to setting tracks to touch -0.3dbfs a few times over the 'average' of the song. So it's taking me a bit to get into this ITU metering (granted it's only for broadcast now, not cd/dvd) but I think it will eventually win out over my [differently weighted and filtered] vu and rms metering schemes. But I'm also interested in that Replay plugin for Audacity to see how it compares to the new ways. I've been trying to 'normalize' tracks using metering and tools ever since CoolEdit96 first gave me the option. The ears have always found the correct levels to level match tracks, now ITU based levels gives some type of 'standards', even legalities. But, if plugins could set the gains I'd delegate that task for sure!

Tricky-Loops · Post by **Tricky-Loops** » Sun Jan 12, 2014 7:08 pm

Isn't Replay Gain an older technique to approximate equal loudness?

I used MP3Gain which is pretty good, and it's non-destructive, too. Since then, I can listen to MP3 playlists without adjusting the volume for every song.

But I never tried it with WAV files.

kylen · Post by **kylen** » Sun Jan 12, 2014 7:19 pm

Tricky-Loops wrote:Isn't Replay Gain an older technique to approximate equal loudness?

I used MP3Gain which is pretty good, and it's non-destructive, too. Since then, I can listen to MP3 playlists without adjusting the volume for every song.

But I never tried it with WAV files.

Could be - I heard that it was used for playback gain on Spotify. I know its been out for sometime but the implementation on Audacity is somewhat new, 2012 or thereabouts.

I'm just looking for ideas on automated ways to set playback gain, will look at mp3gain (seems like you like it) but I didn't think about it since it has the word mp3 in it, I guess that's not a problem though since the stuff I upload to the various services gets some level of codec compression anyway.

Compyfox · Post by **Compyfox** » Sun Jan 12, 2014 9:02 pm

Compyfox wrote:By your definition, the debated "true" average measurement would mean pure offline measurement, or measurement post program stream (the result), which the EBU R-128 meter is capable of (histogram) - again, if I understood you right.,

You should really write than down in an ultra simple form.

Actually - it is pretty simple:

Peak Meter, VU Meter (needle)/RMS meter (bargraph) equals: "Peak and Average Signal Strength" in real time (tracking/recording, mixing)
DR-Meter, ITU-R BS.1770/EBU R-128 meter, SPL Meter (at 1m distance to signal source) equals loudness measurement in real time (peak max, average loudness on histogram and over the course of the program, Dynamic Range, etc)
Offline Measurement Tools (and realtime tools with histogram and averaging calculation, i.e. EBU R-128 post stream analysis) equals "definite/true average" and "maximum" signal strength measured over a specific timeframe

Example:
The old DR-Meter offline tool could be considered "True Average", since it analysed the whole file, looked at specific parts of a track and created a arithmetic average. Similar tools can be seen in analysis tools for example in Wavelab, Grimm Audio's "LevelShow" or Nugen Audio's VisLM offline tools.

Everything that is running in real time and do not give out any arithmetic average values can then be considered "plain average". At least to my understanding, and I definitely know where aciddose is aiming at with that.

Interesting concept, excellent discussion. Never thought of that actually. Hope I made it as simple as possible.

kylen wrote:EDIT: Speaking of Replay Gain...I guess I'll try out this Audacity plugin at some point:
http://forum.audacityteam.org/viewtopic ... 42&t=63067
This one is an implementation based on the 'standard' so I assume it could vary a bit.
Would be interesting to see how it stacks up against ears and other metering discussed.

To be honest, I haven't tried this particular tool. But I have my issues with Replay Gain tools in general. For example R128gain or MP3gain. The former being a project by HydrogenAudio btw (yes, the community that insists on hearing the grass grow in MP3s!).

I took a dive at it compared to analysis tools like they are built into Wavelab 8 (Full version) these days, which also writes track and album gain in the ID3 tags. I found out that the ballistics are totally wrong, and the conversion is not really great to pull off "manually". Unless you let the tool (like Wavelab) handle the offset values.

And this is all there is to it:
Let's say the playback system is setup to -17dBFS, but the track's loudness (example: SLk meter, reference level is -16LUFS = 0 LU) has a readout of +6LU on forte fortissimo parts, while the average signal strength is about +4LU or something. The MP3 ID tags then needs a value to compensate. Example with the given values, and in order to have the same loudness over the whole stream: Replay Gain -4dB.

BUT, and this is a big but, the concept ultimately falls apart as soon as there is another system with a different reference level. Let's say system B uses -18dBFS as refrence level. What sounds consistend on system A would be 2dB off.

Not to mention that Replay Gain is not hardcoded, but can be changed.

Here is where the so called "Music Loudness Alliance" steps in. The creators are the same people that worked on the EBU R-128 specification, with a handful more engineers that think a bit different.

Their concept is on the "playback engine" side of things. Instead of relying on Replay Gain values in metatags, the system analyses the track either on the fly, or offline prior to playback, then adjusts the loudness automatically according to set rules. The user does not need to do anything, neither does the mastering engineer. Carefree usage.

But (there is it again, the evil word!) - this is unfortunately still a bit far from reality. And it's still not known how this will be a "built in" feature of future players (hardware are darn hard to upgrade for this, software wise it's simpler).

kylen wrote:A while ago compyfox posted some demos (I think in his KVR remarks) that set levels based on the ITU/Kv2 that confused me a bit (because they don't go digital full scale, only TP -1LU or something), the perceived loudness sounded good but the rms (if pushed to full scale) would seem to be a louder reading than I shoot for, that's because I'm used to setting tracks to touch -0.3dbfs a few times over the 'average' of the song. So it's taking me a bit to get into this ITU metering (granted it's only for broadcast now, not cd/dvd) but I think it will eventually win out over my [differently weighted and filtered] vu and rms metering schemes.

I think you mean the "Chameleon Jazz Connection" promo CD I posted in Everything Else (Music). The data that I presented in the "K-System v2 proposal" thread were randomly selected files.

I did indeed limit the output of these files (the jazz album) to -1dBTP, but the signal peaks clearly exceeded 0dBFS prior to limiting. Some tracks by merely 0.5dB, others by up to 3-4dB. The tracks clocked in at about K-12v2 (SLk bargraph of the EBU R-128 meter). By request of the client.

Now... if I'd turn down the signal by 4dB post mastering to go down to -16LUFS (K-16v2), the signal peaks would have still been clipped away. The percieved "punch" would be gone due to changed/missing transients. Granted, not as strong as if the loudness would have been -8LUFS on average (mezzoforte passages, -4LUFS at forte fortissimo passages - which I do not recommend). But definitely noticable. Especially if you compare a similar track with it's full dynamics at this signal strength.

The other way around(!) would be more suitable. Let's assume that the average signal strength of the album (mezzo forte passages) are at -16LUFS, and the loudest parts (forte fortissimo) are at -13LUFS. The maximum digital peak would barely touch -1dB True Peak, sometimes go over up to +0,5dB. So effectively only the really strong peaks would be clipped (1,5dB compared to the upper mentioned up to 5dB! at K-12v2) - barely noticable.

We now have a dynamic range of 15dB(!) with the K-16v2 version compared to the K-12v2 version of 11dB. Side by side, at the same loudness (i.e. -16LUFS), the K-12v2 version would sound more quiet due to the missing transients (+3dB more gain = doubled output, therefore it is felt louder, but in order to stay within technical limits, the signal will be compressed!), while the K-16v2 version sounds louder and healthier. The transients are what really give us the impression "now wait a minute, this thing is loud, yet still pleasant sounding, not pumping, etc".

Here is an example of what's happening
http://www.youtube.com/watch?v=3Gmex_4hreQ

So... Replay Gain, even if it's years old, still has it's disadvantages (see the "system reference level" thing). Having a too high reference level on playback with loudness normalisation systems, or a too low one, also results into problems. Especially considering the loudness variety of available current releases.

Now, if we'd at least agree on K-16v2 for music, or at least K-14v2 for the time being until people got used to it, it would be a great compromise for everybody. Only the strong rouge peaks are cut off, while still having a healty signal. It would be like mid 90ies CD releases again, we won't overload CODECs either.

And certain tools would be obsolete, or used like initially concepted - not contantly being abused. Multiband Compressors and Limiter come to mind.

Everything has it's positive and negative effects. But if there is at least one person that picks up some bits of knowledge from all this, maybe even starts to work different - then I'm more than happy. Proper education is something that got lost in the last couple of decades (due to elitism, or like after wars - "the winner writes history"). And my toenails constantly roll up while reading magazines, modern tutorials on the web and what not.

kylen · Post by **kylen** » Sun Jan 12, 2014 10:54 pm

As usual, your posts on loudness exceed all expectations so it will take a bit of time for me to work that in to my thinking but it's getting there.

Yes, this thread has made me want to consider the 'playback engine' more as you noted and how to prepare tracks in the most optimal manner. Playback is not just a mix-cd you hand off to folks anymore. Now we have SoundCloud, iTunes and YouTube to name a few, each with their own playback engine that one may have lesser or greater control over as we go on.

So the seeming simple issue of getting each song to the optimal loudness and then combining multiple tracks into a collection of similar loudness is easy to detect when it's wrong. Takes a 'little' more effort to actually perform. Every one with ears, sound engineer and not, can say when the loudness of a song or collection is wrong. They just have different ways of communicating it...

Compyfox · Post by **Compyfox** » Sun Jan 12, 2014 11:14 pm

Another thing that wasn't considered with these "playback engines"...
If these engines don't stream "your" created file, chances are that they are recoded into something more usable for them.

Example:
Radio streams using AAC, but relying on WAV and MP3 prior to the webstream conversion. In case of YouTube, recoding can happen. Sometimes even audio streams being cut off or worse sounding than before. I don't know how it is with iTunes these days, but I still remember early days where you were forced to "recode" to AAC.

So even if we did all we could to "prevent" issues in an upcoming release, there is no guarantee that it won't be messed up at a later state. Especially with streaming services where we don't have any further control about.

All this is adding up - especially if your content is already at it's limits, or at the limit of the particular used audio codec in question.

A whole topic all in itself. But it's one of the side effects we have to consider.

matthew_wehttam · Post by **matthew_wehttam** » Sun Mar 16, 2014 1:56 pm

Nerd alert. Nerd alert. Master track - hard gate - problem solved. Then the question is -40 or -60. Surely?

Compyfox · Post by **Compyfox** » Sun Mar 16, 2014 2:12 pm

Hard gate?
In terms of "cutting" noise?

Sorry, but I don't follow what you mean.

matthew_wehttam · Post by **matthew_wehttam** » Sun Mar 16, 2014 2:22 pm

A gate without filters or anything will cut out all sounds below a certain threshold. Not just unwanted noise. So if this is on your master track without any decay then your sounds will always be within a set dynamic range. Then you actually can rely more on your ears as apposed to being concerned about various forms of over generalised terms and "suck my d@$&" audio terminology and student programmer orientated plugin experiments. Such as kantancorous is the same as grumpy old. Grumpy old has far fewer syllables and means the same.

Compyfox · Post by **Compyfox** » Sun Mar 16, 2014 2:48 pm

If you aim at specific measuring, then this is what the EBU R-128 spec is already doing.

These type of meters gate the signal below a certain threhsold (-60dB), amongst other things (like summed signal, rather than individual channels).

And I still say that EBU R-128/ITU-R BS.1770 meters are definitely more suitable for long-term stream measurement (or analyzing a track in general) rather than "momentary" measurement like with an unweighted RMS meter (and a timeframe of 300ms) and focusing on the loud parts only.

Unless I'm still understanding you wrong here?

Good ways to make your songs approximately the same volume?