In a mix - Gain vs. Track volume and mix LUFS for mastering

[Murray B]

Hi George,

Have you taken the time to watch Ian Shepherd's videos on Mastering?

I can't personally offer you any expert advice, but in the series he deals with a track that starts out quiet and gets louder. Might be helpful?

www.soundonsound.com/series/mastering-essentials-sos-techniques-guide

[RichardT]

Hi George,

You don’t have to use compression at all. Instead you can use automation to bring up the volume of the quiet parts. This is much better because it won’t change the quality of the sound.

[RichardT]

Hmm, I've read it and I still disagree! The mastering engineer doesn't do anything in relation to loudness normalization, he can only compress / limit. If you have a mix that sounds nice and dynamic to start with, you need to make sure that the mastering engineer doesn't overdo the limiting and take the life out of it (my mastering engineer says by default he will master to something higher than -14 LUFS because most clients want that 'loud' sound. I'm kind of the opposite). Perhaps we're at cross-purposes somehow?

Loudness normalization does change the sound if you have a mix which is quieter than the target for the streaming service, and has peaks close to 0dBFS. Compression or limiting will be applied by the streamer to stop the track clipping after it's turned up.

Indeed that's exactly what I wrote, so not sure what you disagree about!

But all good.

Good. I’m glad we actually agree!

[CS70]

Reading you makes me feel more and more unsure what could be the best approach in mastering phase for bringing volume levels up for streaming and CD purposes without squashing dynamics entirely with limiters.

Let’s take one of the pieces I am working on right now - it is 4 min long with bass solo and SATB choir. Most of it is in piano dynamics until close to the end where a big culmination part comes in forte (entire choir jumps an octave up and it gets really loud and then calms down again to pianissimo for finish).

So what should I do in mastering to preserve the dynamics but bring piano part up without clipping the culmination so that listeners are not forced to bring volume up for the quiet part and get surprised in the forte bringing quickly volume back down?

First of all, take a deep breath and forget mastering. Make a good mix. It's the mix that preserves the dynamics and gives the result you want. Mastering is a different job, and just the icing on the cake.

In your specific case, it really depends on the material. Remember that mixing is 95% about balance. The part that trips people is that often the "right" balance needs to change over the course of the track (it's "dynamic" - not in sense of level range, but in the sense that it changes over time).

Imagine people playing your material. If they're good players and singers, they'll continuously listen to each other and adjust their relative volume in different sections, so to give a good performance... in other words, they will mix themselves. That means that If you record well, making a good mix will in principle be a very simple job - just raise the faders and you're done. The exact same balance will work throughout the mix, because the band/ensemble is doing its own level dynamics, and as a mixing engineer you don't have to change much at all.

In your case, if the performers are good and the recordings are good, the volume relationship between the initial piano, bass and choir will be "right" out of the box: the initial (well rehearsed) piano player will set the initial volume, the bass player will have set (and change) his bass levels so that they're right both when he's setting the groove and when he's soloing, and the choirmaster will have helped the choir to find a suitable "fortissimo" (with respect to the piano level). And all that will have been capture by suitably positioned, well chosen microphones with the right gain settings for their sensitivity to faithfully capture the excellent performance. Bliss! And very little mixing to do.

Very often, however, for one reason or another things are not so. For example, music is made by overdubbing tracks taken at different times; the gain staging or positioning or selection of the various microphones and preamps was less than ideal; or the performers were a little to self-centred and didn't really respond to each other; stuff like that.

In that case, your primary job as a mixing engineer is to work the balance over time to compensate for that (and perhaps to compensate for timbral issues arising from bad choice of microphone or placement).

That means that you cannot set the faders at the start of the song and just leave them there - you need to change the balance as the track progresses.

Now moving many faders continuously and precisely over an entire performance is a daunting task. So it is in that context that you start thinking about compression, because compression is one tool that allows you to change the balance dynamically over time, in a relatively easy and efficient manner, without having to faff with the faders too much.

So in your case - say the bass player has gone overboard in the solo part and he's too loud with respect to the piano: you set up your initial balance (the fader level) for the bass - which is "right" everywhere but for the solo; then you add a compressor so that it's active during the "too loud" bits and it does nothing elsewhere. The result ist that you can happily leave your faders (initial balance) where they are and bass line will work throughout the mix.

The choir starts too low, catches up and gets to the right spot? Same thing: set the balance that works for most of the part, and rein in the rest with a compressor.

In other words, in this context you use compression to try to go back to what the performance should have been (as opposite to what you have on the recorded tracks).

And so on. There's literally books written on how to mix and deal with these things, so we could go on for years. And the whole thing of timbrical changes - compressors can play a big role there as well. Since life's not simple, timbre also affects the preceived loudness, but that's another subject so won't go there.

The gist is that dynamics are, in general, good. They're what makes music worth listening. Reducing 'em is only needed for fixing thing, or to overcome technical limitations of certain playback equipment or environments (for example, radio station routinely compress everything, to ensure max compatibility with even dreadful playback gear and environments).

So you really want your mix to have the low parts low, and the final choir high - and the mix mirror (or claw back to) the great performance.

You can, of course, set a target dynamic range (you had to do that, for example, with vinyl records because of their limitations) and that's fair in mastering or even in playback processing, but you should do it only because of concrete reasons, like radio, or because you think your target market won't be able to cope with the real performance dynamics.

That is where either mastering or master bus processing comes into play, but first thing: make a good mix.

You may notice I haven't mentioned LUFS once. Loudness normalization is a volume change applied uniformly on the entire track, so it won't change the balance (i.e. the relationship between levels of each part at every moment) even if, of course, will reduce the loudness (i.e. the absolute level of each part at every moment). If at a point T in time the level of the piano is 1/3rd of the level of the bass, for example, it will stay that way no matter the loudness (and reducing loudness will simply round the amount of "decimals" of the division, if you like the arithmetic).

As mentioned both in my blog post and Richard's entry, you just need to ensure you have enough loudness to start with, because some services only attenuate but do not increase volume. I routinely get out mixes which play louder than the recommended LUFS for streaming... but of course I make sure I listen to them - in calibrated monitoring conditions - so that I know how they sound at reduced volume.

That's why you dont need to worry about LUFS, exactly like you don't worry about the listener volume knob. It's out of your hands, and all you can do is to do the best mix you can.

I use Adobe Audition CS6 as DAW and plugins from FabFilter and Ozone 5. Let’s say the mixdown is with 6db headroom from the loudest part. In such case the piano part (most of the song actually) becomes very quiet for listening in a car for example.

If you think that your music will be mostly listened at in a car, it's a good reason to reduce the dynamics, yes. Another reason is of course aesthetics: you like a little riff or accent or section and want it to stand out for a little while.

There are 3 options that I see:
1. Make EQ, compressors, gain plugin for bringing volume up, maybe a maximizer and reverb
2. The same as above but without gain. All effects are applied in a way that there is no change in volume. Then use something in Audition called Match where you set target LUFS and it brings the whole waveform up but it’s using a limiter to prevent clipping
3. Apply effects without gain plugin while keeping the same volume level. Then apply simple normalization up to -1 dbfs and skip LUFS matching part to avoid additional limiting. I guess this would leave you with whatever LUFS it gets.

All possible if they get you where you want to be. It's like a toolbox - there's many tools and many ways of combining them to get you to a specific result you want.

2. will not change the dynamic relationship between the parts (the balance) so if your piano was too quiet with respect to the choir, it will still be too quiet (or now the choir will be too loud). Remember: loudness is about turning the volume knob uniformly for the entire track, not changing the balance between its components.

Just to be precise, it's true that increasing LUFS will be usually followed by a bit of limiting, but that'll be so small to be insignificant. Again, LUFS is "turning up the volume knob" so 99% of the action will be an uniform amplitude increase - to reach the required LUFS level from a lower level. How much the volume knob is "turned up" is calculated as "just enough to reach the target level" and if you do the math, due to the way the algorithm calculates loduness, even with the craziest dynamic music (almost silence all the time, then a gigantic peak) there's no chance you'd need more than 1 or 2dB of limiting. Very transparent.

What do you think? Which option would give best results? Or you would approach such use case very differently?

Take say the piano - set its fader in the mix at the level that's "right" for most of the track, probably the quiet bits, and compress the rest; or alternatively, gain up the quiet bits.. which is best depends on the quality of the recording. Increasing level will bring up noise; compression on a loud signal without makeup gain, will perform better from a noise point of view. Both work with the right material.

Then move to the bass and look at the relationship with piano and choir and again, find the fader position which works for most of the track, and use compressors or level changes to bring the rest in line.

Another common way to deal with these things is to automate the fader (which is the same that a compressor does, but you have much more control of the exact way that the changes happen).. but it's very very time consuming so it's a type of processing usually left only for the really important stuff, like vocals in a song.

Also, as above, timbre changes (whatever mean you use to achieve them) will also change the perception of loudness, and so does the music arrangement itself... so there's really many more ways of achieving what you want.

If you want the final result to have a narrower specific dynamic range, tell the mastering engineer... or place a compressor on the master bus, bring the DR where you want it and use makeup gain to get the peaks back where they were.

In this latter case, if you are really really sure that your music will be loudness-normalized when played back, you can use just enough makeup gain to reach the LUFS you want (as opposite to bring the peaks back to whatever natural level they had). But it doesn't really change much at all.

[george_vel]

Hi George,

Have you taken the time to watch Ian Shepherd's videos on Mastering?

I can't personally offer you any expert advice, but in the series he deals with a track that starts out quiet and gets louder. Might be helpful?

www.soundonsound.com/series/mastering-essentials-sos-techniques-guide

Thanks, definitely I'll watch the videos. Actually I've already watched the first one, and once again I see that there are almost no such videos or knowledge share articles about how to make mixing and mastering with classical / choral music. Most of the examples and techniques are applicable for pop / rock / r&b, etc. genres, but almost nothing about classical / choral music.

Hi George,

You don’t have to use compression at all. Instead you can use automation to bring up the volume of the quiet parts. This is much better because it won’t change the quality of the sound.

But maybe just a little, to glue things together?

...

Wow, I appreciate your effort in writing so detailed explanation. Thanks for that, I'll take your advices into count.

[Drew Stephenson]

CS70, not that it changes your advice, but I think George is talking about piano as a musical term rather than musical instrument. Likewise the 'bass solo' is a solo by the bass singer, not the bass player!

But I could be wrong...

[george_vel]

CS70, not that it changes your advice, but I think George is talking about piano as a musical term rather than musical instrument. Likewise the 'bass solo' is a solo by the bass singer, not the bass player!

But I could be wrong...

You're not wrong at all!
By piano I've meant piano dynamics in performance, not the instrument.
And yes, bass singer as a soloist.
So it's pure a capella material, except 3 pieces where also a church organ comes on "the sound stage".
Sorry, I've missed somehow this detail from CS70 explanation, but as you said - it's not changing the nature of what was said.

[Hugh Robjohns]

Let’s say the mixdown is with 6dB headroom from the loudest part. In such case the piano part (most of the song actually) becomes very quiet for listening in a car for example.

Headroom is something that you need when tracking, so that there is no risk of overload when recording the raw tracks.

It's also something that is useful to have when mixing because some processing can result in unexpected signal peaks, and while the DAW can't be overloaded internally, the converters in the monitoring chain definitely can.

But once you have a finished (mastered) mix, headroom becomes irrelevant and necessary and is almost always reduced as the signal levels are now completely known. I wouldn't recommend removing the redundant headroom completely -- as was the norm in the days of CD -- because that puts you at risk of inter-sample overloads in the end-user's converters, especially if the material is likely to end up being data reduced (as an mp3 file etc). For that reason, I would say never peak higher than -1dBFS, ever... and often -3dBFS is a safer bet.

Okay, so headroom is not necessary in the finished (mastered) mix. So the next question issue is dynamic range.

In the days of CD, everyone wanted their record to sound louder than everyone else's and, since the format had a hard ceiling, the only way to achieve that was to compress and limit the whotsits out of the track to drive the average level as high as possible. end result, horrible lifeless, boring, unpleasant music...

The good news is that with loudness normalisation now the norm, that has all gone away. Everything is now replayed such that the average long-term loudness sits around -14LUFS... meaning you can now have dynamic peaks 12dB or so higher than the average level again. Hurrah...

But actually, how dynamic the music recording actually is comes down to a combination of the performance, the genre and the end-user expectations.

In a live situation, a performance with huge dynamics can be fantastic, and it's easy these days to capture those dynamics intact in a recording. But as a recorded track, huge dynamics are a pain in the back-end. We live in a noisy world and most people, listening at home in the kitchen, or in the car, or whatever, don't really want the loud bits to scare the neighbours, and the quiet bits to get lost under the type noise or the cooker's extractor fan.

So you're almost always going to have to reduce the recorded dynamics to make a track listenable domestically. How much reduction comes down to taste and expectations again -- it requires an artistic decision.

In the pop/rock field, compression and limiting are part of the sound, but in classical and choral works, it can often be a bit too mechanical, and I find better results can be achieved with some skilful fader-riding, either manually or through automation.

So you set the appropriate level for the loudest bits, and then raise the quieter bits manually in such a way that the listener is unaware that you're manipulating the level.

Of course, you can still use a limiter to help reign in any transient peaks, and I often use either some parallel compression or, more usually, very gentle compression (1.5:1 or less) starting from a very long way down (-50dBFS, say) to help squeeze things in a transparent way.

If you establish your monitoring level to sound 'right' when playing similar genre tracks from your preferred streaming service, you'll find it fairly easy to mix your own tracks with a similar integrated loudness value.

Once you are happy with the mix, you can run a quick check to determine the actual Integrated LUFS figure and True Peak value, and raise or lower the overall track level accordingly, if necessary. But if the LUFS comes out at -14LUFS and the True Peak at -6dBTP, that's fine -- it doesn't need to hit the -1dBTP limit, and actually being a little low is often a good thing. All that matters is that the mix sounds great, with nothing getting lost becuase it's too quiet for the end user, and that the integrated LUFS is in the right ball park.

[CS70]

CS70, not that it changes your advice, but I think George is talking about piano as a musical term rather than musical instrument. Likewise the 'bass solo' is a solo by the bass singer, not the bass player!

But I could be wrong...

You're not wrong at all!
By piano I've meant piano dynamics in performance, not the instrument.
And yes, bass singer as a soloist.
So it's pure a capella material, except 3 pieces where also a church organ comes on "the sound stage".
Sorry, I've missed somehow this detail from CS70 explanation, but as you said - it's not changing the nature of what was said.

Ahah thanks Drew! Yeah, they were just examples anyway, the ideas are irrelevant of the specific musical content.

[Zukan]

Let’s say the mixdown is with 6dB headroom from the loudest part. In such case the piano part (most of the song actually) becomes very quiet for listening in a car for example.

Headroom is something that you need when tracking, so that there is no risk of overload when recording the raw tracks.

It's also something that is useful to have when mixing because some processing can result in unexpected signal peaks, and while the DAW can't be overloaded internally, the converters in the monitoring chain definitely can.

But once you have a finished (mastered) mix, headroom becomes irrelevant and necessary and is almost always reduced as the signal levels are now completely known. I wouldn't recommend removing the redundant headroom completely -- as was the norm in the days of CD -- because that puts you at risk of inter-sample overloads in the end-user's converters, especially if the material is likely to end up being data reduced (as an mp3 file etc). For that reason, I would say never peak higher than -1dBFS, ever... and often -3dBFS is a safer bet.

Okay, so headroom is not necessary in the finished (mastered) mix. So the next question issue is dynamic range.

In the days of CD, everyone wanted their record to sound louder than everyone else's and, since the format had a hard ceiling, the only way to achieve that was to compress and limit the whotsits out of the track to drive the average level as high as possible. end result, horrible lifeless, boring, unpleasant music...

The good news is that with loudness normalisation now the norm, that has all gone away. Everything is now replayed such that the average long-term loudness sits around -14LUFS... meaning you can now have dynamic peaks 12dB or so higher than the average level again. Hurrah...

But actually, how dynamic the music recording actually is comes down to a combination of the performance, the genre and the end-user expectations.

In a live situation, a performance with huge dynamics can be fantastic, and it's easy these days to capture those dynamics intact in a recording. But as a recorded track, huge dynamics are a pain in the back-end. We live in a noisy world and most people, listening at home in the kitchen, or in the car, or whatever, don't really want the loud bits to scare the neighbours, and the quiet bits to get lost under the type noise or the cooker's extractor fan.

So you're almost always going to have to reduce the recorded dynamics to make a track listenable domestically. How much reduction comes down to taste and expectations again -- it requires an artistic decision.

In the pop/rock field, compression and limiting are part of the sound, but in classical and choral works, it can often be a bit too mechanical, and I find better results can be achieved with some skilful fader-riding, either manually or through automation.

So you set the appropriate level for the loudest bits, and then raise the quieter bits manually in such a way that the listener is unaware that you're manipulating the level.

Of course, you can still use a limiter to help reign in any transient peaks, and I often use either some parallel compression or, more usually, very gentle compression (1.5:1 or less) starting from a very long way down (-50dBFS, say) to help squeeze things in a transparent way.

If you establish your monitoring level to sound 'right' when playing similar genre tracks from your preferred streaming service, you'll find it fairly easy to mix your own tracks with a similar integrated loudness value.

Once you are happy with the mix, you can run a quick check to determine the actual Integrated LUFS figure and True Peak value, and raise or lower the overall track level accordingly, if necessary. But if the LUFS comes out at -14LUFS and the True Peak at -6dBTP, that's fine -- it doesn't need to hit the -1dBTP limit, and actually being a little low is often a good thing. All that matters is that the mix sounds great, with nothing getting lost becuase it's too quiet for the end user, and that the integrated LUFS is in the right ball park.

Jedi!!!!

[CS70]

CS70, not that it changes your advice, but I think George is talking about piano as a musical term rather than musical instrument. Likewise the 'bass solo' is a solo by the bass singer, not the bass player!

But I could be wrong...

You're not wrong at all!
By piano I've meant piano dynamics in performance, not the instrument.
And yes, bass singer as a soloist.
So it's pure a capella material, except 3 pieces where also a church organ comes on "the sound stage".
Sorry, I've missed somehow this detail from CS70 explanation, but as you said - it's not changing the nature of what was said.

Something else that can help you perhaps, is that in order to evaluate your mix from a dynamic perspective ("is that part too quiet? This other too loud?") you really want to a set a listening level and stick to that (mostly) when mixing.

That listening level should be around the ballpark of what the listeners will, in average listen to... in old, "proper" studios, I think it was found that using 83dB ass a reference listening level you got good results for the final product, but that was with the sound coming off big monitors positioned quite far from a big console... so one tends to monitor at a much quieter level for small studio nearfields, 6-10dB less.. surely Hugh can tell all about it )

But regardless, find a playback level where you think what you mix is similar to what eventually the listener will hear, and mostly stick to it. Just do occasional "very quiet" and "very loud" runs to see what's what. For example, both for Fletcher-Munson and the actual balance, certain parts will disappear when the volume knob is set very low, so you want to make sure that what survives is what you think it should survive; and the same with very loud playback, certain timbral issues may become apparent only when you push the level a bit.

[george_vel]

...

Jedi!!!!

I'd say - a Master Jedi!

@Hugh, @CS70, thanks for your invaluable inputs.

I have one last concern to ask about, and then I guess I have to go and try all your advices.

My issue comes from the fact that I'm trying to achieve a reasonable wide stereo effect. So, in the mix the choir is in a formation where sopranos are far left, basses far right and the rest in between. The soloist is in the middle. This is achieved with panning sopranos 45 degrees left, basses 45 degrees right and the rest in the middle between these and the soloist.

But what I experience is that, because of this panning, my right ear is almost all the time exposed more to the lows and respectively my left ear more to the highs. It sounds good, especially on headphones, but after listening for some time (and again especially on headphones) my ears start feeling uncomfortable.
Any suggestion for better practices when panning such type of performance and keeping the stereo effect wide?

[Hugh Robjohns]

But what I experience is that, because of this panning, my right ear is almost all the time exposed more to the lows and respectively my left ear more to the highs. It sounds good, especially on headphones, but after listening for some time (and again especially on headphones) my ears start feeling uncomfortable.
Any suggestion for better practices when panning such type of performance and keeping the stereo effect wide?

If you recorded the choir in real life with a stereo mic array you'd have the same effect... and many orchestral recordings have the first violins on the left and cellos and basses on the right!

But in these cases the violins or cellos aren't a 'point source' panned to a specific position, but a spread of sound covering a certain angle of 20 degrees or so, overlapping the more central second violins. And the cellos and basses are similarly spread out and overlap the central violas.

So I suspect much comes down to how you recorded this choir and how much stereo width you have for each of the SATB sections. If you only have single mics for each section it might sound a bit isolated and spartan rather than a spacious wide blend.

But in general, if it's sounding unbalanced it's either because your mix levels are wrong, or the musical arrangement is working against you...

Careful use of artificial reverb can help, though, introducing some bass weight to the opposite side without detracting from the imaging... and it could also enhance the sense of a cohesive spread of sound.

[george_vel]

But what I experience is that, because of this panning, my right ear is almost all the time exposed more to the lows and respectively my left ear more to the highs. It sounds good, especially on headphones, but after listening for some time (and again especially on headphones) my ears start feeling uncomfortable.
Any suggestion for better practices when panning such type of performance and keeping the stereo effect wide?

If you recorded the choir in real life with a stereo mic array you'd have the same effect... and many orchestral recordings have the first violins on the left and cellos and basses on the right!

But in these cases the violins or cellos aren't a 'point source' panned to a specific position, but a spread of sound covering a certain angle of 20 degrees or so, overlapping the more central second violins. And the cellos and basses are similarly spread out and overlap the central violas.

So I suspect much comes down to how you recorded this choir and how much stereo width you have for each of the SATB sections. If you only have single mics for each section it might sound a bit isolated and spartan rather than a spacious wide blend.

But in general, if it's sounding unbalanced it's either because your mix levels are wrong, or the musical arrangement is working against you...

Careful use of artificial reverb can help, though, introducing some bass weight to the opposite side without detracting from the imaging... and it could also enhance the sense of a cohesive spread of sound.

This time the recording was not done by me And yes, single mic per section.

When you say “careful use of artificial reverb” I guess you mean applying it on the mix bus or...?

[Hugh Robjohns]

This time the recording was not done by me And yes, single mic per section.

I suspected as much... Is there much spill between sections/mics?

When you say “careful use of artificial reverb” I guess you mean applying it on the mix bus or...?

Several possible ways of tackling this. You could send the individual mics to four mono reverbs, and pan their outputs to the opposite side from the panned mic. Or you could use a stereo reverb on the stereo bus and possibly L-R swap the outputs (depending on how defined the reverb is spatially).

But basically it comes down to the choice of reverb algorithm and the setting of its parameters to come up with something that helps to balanced and blend the section mics together, without losing definition.

[george_vel]

This time the recording was not done by me And yes, single mic per section.

I suspected as much... Is there much spill between sections/mics?

Yes, unfortunately too much spill. This makes EQ at this point very difficult because strange sonic build ups appear.

Several possible ways of tackling this. You could send the individual mics to four mono reverbs, and pan their outputs to the opposite side from the panned mic. Or you could use a stereo reverb on the stereo bus and possibly L-R swap the outputs (depending on how defined the reverb is spatially).

I’ve tried sending single tracks to buses and place them opposite but without reverb. But when bringing up volume in mastering the result was an appearance of some (I don’t know how to describe it) tension or too much density in the choir sound that I quit this idea. I guess, because of doubling the channels, some frequency build ups happened that made the sound so tense, although duplicate tracks were at least 15db lower than originals.

But basically it comes down to the choice of reverb algorithm and the setting of its parameters to come up with something that helps to balanced and blend the section mics together, without losing definition.

Well, yeah, and the hardest part is “choice of reverb algorithm and the settings of its parameters”...

[CS70]

Just splat a Lexi RandomHall on there and be done!

(ducking the shoe that Hugh's throwing at me )

Just kidding of course, but yeah it's not unusual to go thru many reverbs to find the one that clicks, and maybe the 4 mono reverbs can be easier in your case.