Master bus gain vs limiting

[Dave B]

So my poor brain is currently struggling with wrapping itself around what should happen on the master bus when mixing. Like a fair few of us (I expect), I work on a variety of stuff for my own personal amusement and it will inevitably find it's way to a variety of playouts - soundcloud, direct to people's ipods/itunes/etc, and youTube is the next platform. I think I'm relatively happy with mixing for loudness rather than peak levels. It's just the last leg that I struggle with.

I tend to mix so that my mix bus is working around the K12 / K14 standard (usually the former and I'm a little bit lazy, but not too bad). So the bus will peak with lots of headroom. If I apply any 'glue' or 'shine' (compresssion or exciter/eq), then that will be the first thing I insert. I then stick in my k-meter so that I'm keeping an eye on the values post bus processing.

It's what comes next that I'm not sure about. My standard tactic is to place a limiter, followed by a gain, followed by the LUFS meter. When talking to one of the clever bods about this, I was advised to limit first, then simply add the required, fix-level gain to the result to get the whole thing to peak at 0 (ok .. a little under but you know what I mean). I'm sure there was a reason for that, but I can't remember it. I can then push the limiter to get the loudness of the audio into the right LUFS level.

My question is : does this make a difference? Could I not simply apply a whole heap of input gain into the limiter until it's nicely peaking and reading the right LUFs level? Am I worrying too much over nothing? Or is there really a good reason to limit then gain?

I think I'm struggling because I'm using Fabfilter's ProL limiter and, after a little experimentation, I'm not sure that simply setting desired output level to -12dB is the same as saying 'limit with -12dB as your peak value'. It's quite possible I've missed something so I'll do a more thorough test, but I thought I'd check if I was missing something more fundamental first...

[CS70]

My question is : does this make a difference? Could I not simply apply a whole heap of input gain into the limiter until it's nicely peaking and reading the right LUFs level?

It sure does! So long you don't clip, gain will not change the sound and feel of the music at all. It's basically a volume control, and can be defeated by the final listener simply by reducing the volume If you increase the DAW gain but decrease the playback volume correspondingly, your mix will sound exactly the same.

Placing a limiter (and lowering the threshold), on the other side, can change the sound, and possibly quite a bit: a limiter is a high ratio compressor, typically with very fast attack, lookahead and automatic makeup gain to keep the limited peaks at the desired level (usually a couple cents below 0dBFS). Like any compressor set up to compress your material, where you raise the makeup gain to bring the peaks at the same level pre- and post-, it will both lower the dynamic range and raise the overall loudness, essentially make the lower sounds more audible and squashing the peaks, which are most often drum hits. So it'll change the sound of your drums and will make the lower-level details of your mix more audible at a normal listening volume. This can be good or bad depending on the quality of your recordings and mixing.

However. so long the limiter doesn't limit (i.e. there are no sounds that would clip, and the compression circuit doesn't do anything) most limiters act simply as gain devices, so until then they're pretty much equivalent to a gain control. For analog gear or simulation, of course, there can be different coloration due to the fact that the audio path can be a bit more complicated than just a gain circuit, there may be i/o transformers involved and lots of stuff that people with actual electronic competence here can describe way better than myself. In that case, it's not different than using a compressor with 1.0 ratio, 0 threshold and just some makeup gain as a preamplifier. So really no need to pre-gain the input, even if of course it doesn't do any harm and sometimes it may make it easier to accurately control the limiting.

[Dave B]

Cheers CS. I've been doing some more thinking and a bit of audio testing and I'm now starting to doubt what I'm doing altogether.

My theory with the above signal chain was to end up with a final result where the peak is actually limited at -12dBfs (assuming a K12 mix). I've been playing with the limiter and it seems that it always limits peaks at 0dBfs - which is why I've had to increase the input gain to at least the reverse of the output level. Reducing the output level does not - as I thought it did - reset the point at which the limiting occurred, but simply turned the results down. So in order for my (broadly) -12dBfs output to hit the limiter, I had to drive it 12dB extra to force it up to 0dBFS and then pad the results back down to -12. And then add gain to get it back up to 0 again.

This doesn't seem like a sensible approach to me. It's like playing table tennis with gain levels.

I'm going to have to think about this some more..

Out of interest, how do other people set up their mix bus so that their resulting files end up at the recommended -14LUFS-i level?

[Hugh Robjohns]

My standard tactic is to place a limiter, followed by a gain, followed by the LUFS meter. When talking to one of the clever bods about this, I was advised to limit first, then simply add the required, fix-level gain to the result to get the whole thing to peak at 0 (ok .. a little under but you know what I mean).

This seems an odd way of going about things to me! I wonder if you are trying to integrate an out-dated peak-normalisation approach with a modern loudness-normalised one, and actually losing all the benefits along the way?

So, you need to decide if you're working to a peak-normalised paradigm, or a loudness-normalised one -- but I would suggest the latter is the more sensible approach these days. As we're chewing our way through a format cross-over period some compromise will be necessary, but only to a modest degree.

In the loudness-normalised world, peaks are theoretically allowed up to -1dBTP maximum, but if the release format is a lossy-codec like MP3 then the advice is to limit to -3dBTP -- and that's measured with a TRUE-PEAK meter (a standard element of a LUFS meter) to take account of potential inter-sample peaks. So you'll need a decent over-sampling peak-limiter near the end of your mastering plugin chain.

With your transient peaks under control, your next decision is your target Integrated loudness. Most streaming services are currently operating in the -14 to -16 LUFS range, but if you want to retain some sense of comparable 'volume' with 'legacy CDs' (ie. peak normalised) releases you could compromise a bit and aim for an integrated loudness target of -12 LUFS.

So, I would configure a master bus signal flow with a buss compressor for glue, an over-sampling peak limiter set to -3dBTP, and an LUFS meter across the final output to monitor your Integrated loudness and true peak levels.

Personally, I also like to use the Dynameter to keep an eye on the mix dynamics. I think the K12 (or K14) meter is obsolete when it comes to mixing in the loudness-normalised world, but I still like to use the K20 meter when tracking.

So, with that plug-in chain in place, you simply mix and process as necessary to achieve the desired target loudness while retaining and optimising your mix dynamics.

H

[Hugh Robjohns]

My theory with the above signal chain was to end up with a final result where the peak is actually limited at -12dBfs (assuming a K12 mix).

The whole point of mixing on a K12 meter is that you'll end up with your average level somewhere close to -12dBFS (ie. around -12LUFS), so why would you also limit to -12dBFS?
If you do that your peak to average ratio will be tiny, and while that might work for a peak-normalised CD (once you've stripped off 12dB of redundant headroom), that's really not a good idea for material played over a loudness-normalised streamer!

I had to drive it 12dB extra to force it up to 0dBFS and then pad the results back down to -12. And then add gain to get it back up to 0 again.

Er... bonkers! If you're aiming for peak-normalisation and you're already limiting at zero, why would you pad it down and then amplify it again? You've gained nothing from that double-process.

H

[Wonks]

We've always said he needs help.

[Dave B]

I think the K12 (or K14) meter is obsolete when it comes to mixing in the loudness-normalised world, but I still like to use the K20 meter when tracking

Ah, I think that this is what is confusing me. I was assuming that the loudness normalisation was the logical next step after a k-metered mix, and was confused about what to do with the headroom left on the mix bus.

In my defence, I had worked out that what I was doing was fairly daft! See, I do learn ... slowly .. but I get there ...

We've always said he needs help.

Yep! And still much more needed ...

[Hugh Robjohns]

I was assuming that the loudness normalisation was the logical next step after a k-metered mix, and was confused about what to do with the headroom left on the mix bus.

In theory, a mix that 'looks right' on a K12 meter should have an integrated level around -12LUFS (or -14LUFS for a K14 meter, or -20LUFS for a K20 meter) -- Bob Katz' idea was that the meter should encourage the creation of a mix with the appropriate average level and headroom... but in practice I find it easier to keep an eye on the integrated level display of a BS1770 (R128) meter, as that gives more direct feedback and confidence of hitting the right target level.

In my defence, I had worked out that what I was doing was fairly daft!)

[Zukan]

Dave, did I not send you my book?

There are chapters devoted to this plus a ton of video guides.

Not as hard as it looks and quite intuitive compared to ppm metering or k standard referencing.

However, the deity known as Hughness has nailed it... of course.... as usual....followeth him for him knoweth.

[trunkdog]

Dave, did I not send you my book?

snipped...

Would that be "MixBus Strategies"

[Zukan]

Yes Trunk.

[Dave B]

And it is (as all Zukan's books and tutorials are) thoroughly recommended. Zukan knows his beans and is excellent at presenting them clearly and simply.

[hooty2]

Is the book available to buy?

[Matt Houghton]

Is the book available to buy?

Yep. It's a PDF eBook with loads of video and audio to accompany it. Available here

[Ariosto]

In theory, a mix that 'looks right' on a K12 meter should have an integrated level around -12LUFS (or -14LUFS for a K14 meter, or -20LUFS for a K20 meter) -- Bob Katz' idea was that the meter should encourage the creation of a mix with the appropriate average level and headroom... but in practice I find it easier to keep an eye on the integrated level display of a BS1770 (R128) meter, as that gives more direct feedback and confidence of hitting the right target level.

I assume from this Hugh, you would recommend the use of a R128 meter for tracking and for final adjustments in a DAW, as opposed to using a peak level meter. I've experimented with -20 LUFS and - 23 LUFS and it seems to work well on classical chamber music. (None of this was tracked in this way).

I take it that the "Loudness Range" on the meter is the dynamic range in LUFS - equivalent to the same amount in dB's?

[Hugh Robjohns]

I assume from this Hugh, you would recommend the use of a R128 meter for tracking and for final adjustments in a DAW, as opposed to using a peak level meter.

Never for tracking -- utterly inappropriate tool for that. I use the stock sample-peak level meters on the recorder for tracking, working with 20dB nominal headroom and colour coded with green to -20dBFS, yellow from there to -10dBFS or so, and red from there on up, and when tracking I aim to keep things mostly in the yellow region.

When mixing I use a combination of things depending what I'm mixing for, but always with one eye on the R128 Integrated Loudness meter.

I've experimented with -20 LUFS and - 23 LUFS and it seems to work well on classical chamber music. (None of this was tracked in this way).

Yep... somewhere around there, depending on exactly what the content is. The guys at BBC Radio 3 suggested that something even lower might be more appropriate (-27LUFS?)

I take it that the "Loudness Range" on the meter is the dynamic range in LUFS - equivalent to the same amount in dB's?

Yes, broadly, but the actual LRA calculation is a little more complicated than it might appear. The LRA value is derived by analysing a continuously sliding three-second loudness assessment window, with any values more than 20LU below the current programme loudness level gated out (to ensure the LRA figure reflects the range of foreground audio signals not background noise or the system noise floor). The distribution of these short-term loudness levels is further quantified by setting 10 and 95 percentile limits so that the LRA figure is actually the difference between the loudness levels at the 10 and 95 percent points in the total measured range. The lower 10 percent limit prevents things like a fade-out from dominating the LRA value, while the upper 95 percent limit ensures that an isolated loud sound -- such as a gunshot in a film soundtrack -- won't cause an excessively large LRA value.

A noisy crashy-bangy feature film might have an LRA of 25, typical TV programmes would be closer to 10, and squashed pop music could have an LRA value of 5 or less. There's more detail in the box-outs in this feature if you're interested:

https://www.soundonsound.com/techniques ... udness-war

H

[Zukan]

My 'go to' article . Thanks Hugh.

[Ariosto]

Many thanks Hugh for your reply and link to that fantastically brilliant article.