You are here

Loudness contribution of various frequency bands

For everything after the recording stage: hardware/software and how you use it.

Moderator: Moderators

Loudness contribution of various frequency bands

Postby ultraGentle » Sun Feb 21, 2021 10:01 pm

Hello hello,

I'm trying to understand loudness better, and in particular, how different frequency bands seem to contribute non-linearly to loudness as it relates to (1) LUFS and (2) clipping.

I've been studying this by listening while watching graphic EQ, and so far have made the simple observation that low sounds seem to have an outsize impact on levels.

I'd like to understand the relationship between frequency, loudness, and clipping more systematically though, whether through maths or metering tools.

I'm using Youlean Loudness meter, which gives a fine holistic picture, but I'm wondering if there's a multiband equivalent, so I could analyze the contribution of specific bands. A workaround would be to band pass the whole mix prior to Youlean so that it only measure the contribution of a single band at a time -- I may try this -- but it's a bit clunky.

As an example application of this query, subjectively, the low notes in a piano part I have are at a reasonable level compared to the melody, yet I need to turn them down considerably to avoid the whole track getting too loud. I imagine that what I really need to do is to keep the *impression* of them being loud, while actually scooping out portions via EQ -- however to do this effectively, I need to understand some of the above.

Really, though, I want to understand the relationship as stated initially. Thoughts?
ultraGentle
Poster
Posts: 33
Joined: Tue Feb 16, 2021 1:09 am

Re: Loudness contribution of various frequency bands

Postby blinddrew » Sun Feb 21, 2021 10:06 pm

The most direct way is probably to read the specification https://tech.ebu.ch/publications/r128/
User avatar
blinddrew
Jedi Poster
Posts: 13525
Joined: Sat Jul 04, 2015 11:00 pm
Location: York
Ignore the post count, I have no idea what I'm doing...

Re: Loudness contribution of various frequency bands

Postby ultraGentle » Sun Feb 21, 2021 10:10 pm

Thanks for the spec link! I shall read it.

To clarify, independently of current loudness conventions, I'm interested in the math of frequency/power, or interested in any suggestions for ways to visualize this via metering.
ultraGentle
Poster
Posts: 33
Joined: Tue Feb 16, 2021 1:09 am

Re: Loudness contribution of various frequency bands

Postby n o i s e f l e ur » Sun Feb 21, 2021 10:20 pm

Martin was kind enough to post a link to one of Eddie Bazille's articles in the mag that explained this quite succinctly:

https://www.soundonsound.com/techniques ... -reference
n o i s e f l e ur
Regular
Posts: 271
Joined: Sat Aug 04, 2018 5:20 am

Re: Loudness contribution of various frequency bands

Postby Hugh Robjohns » Sun Feb 21, 2021 10:24 pm

ultraGentle wrote:'m trying to understand loudness better, and in particular, how different frequency bands seem to contribute non-linearly to loudness as it relates to (1) LUFS and (2) clipping.

Perceived loudness is calculated as a rolling average over time and is predominantly affected by the upper midrange bands and least determined by the LF content. So the BS.1770 Loudness Normalisatian standard employs a weighting filter (k-weighting) which essentially removes signal content below 100Hz and boosts it above 2kHz, as shown here:

Image

Clipping occurs instananeously when a signal waveform exceeds the maximum amplitude capability of the system, and since most sources generate far more energy in their low frequencies than high frequencies, clipping often occurs when the signal has a lot of LF going on... Although that being said, most mixes tend to have prominent midrange signals -- vocals, snare drums, guitars, etc -- because thats what we like, so these sources are typically more prone to clipping...

I've been studying this by listening while watching graphic EQ, and so far have made the simple observation that low sounds seem to have an outsize impact on levels.

Yes. Typically music spectra exhibits a falling frequency response approximating 3dB/octave, much like pink-noise.

I'm using Youlean Loudness meter, which gives a fine holistic picture, but I'm wondering if there's a multiband equivalent, so I could analyze the contribution of specific bands.

No, because loudness is calculated across the entire (weighted) spectrum, not in separate bands, and involves averaging over time. Monitoring individual frequency bands would be pointless and misleading.

As an example application of this query, subjectively, the low notes in a piano part I have are at a reasonable level compared to the melody, yet I need to turn them down considerably to avoid the whole track getting too loud.

The loudness metering systems already largely ignore lower frequencies as explained above. If the whole track is exceeding your target loudness, while the mix sounds tonally balanced, then you just need to turn the whole mix down. Simples! ;-)
User avatar
Hugh Robjohns
Moderator
Posts: 30067
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK
Technical Editor, Sound On Sound
In my world, things get less strange when I read the manual... 

Re: Loudness contribution of various frequency bands

Postby RichardT » Sun Feb 21, 2021 10:58 pm

ultraGentle wrote:Hello hello,

I'm trying to understand loudness better, and in particular, how different frequency bands seem to contribute non-linearly to loudness as it relates to (1) LUFS and (2) clipping.

I've been studying this by listening while watching graphic EQ, and so far have made the simple observation that low sounds seem to have an outsize impact on levels.

I'd like to understand the relationship between frequency, loudness, and clipping more systematically though, whether through maths or metering tools.

I'm using Youlean Loudness meter, which gives a fine holistic picture, but I'm wondering if there's a multiband equivalent, so I could analyze the contribution of specific bands. A workaround would be to band pass the whole mix prior to Youlean so that it only measure the contribution of a single band at a time -- I may try this -- but it's a bit clunky.

As an example application of this query, subjectively, the low notes in a piano part I have are at a reasonable level compared to the melody, yet I need to turn them down considerably to avoid the whole track getting too loud. I imagine that what I really need to do is to keep the *impression* of them being loud, while actually scooping out portions via EQ -- however to do this effectively, I need to understand some of the above.

Really, though, I want to understand the relationship as stated initially. Thoughts?

My first thought is - is doing this going to be worthwhile? I think the best thing to do is simply to create a good mix and then try out different loudness settings on it until you're happy with it.
RichardT
Frequent Poster
Posts: 644
Joined: Thu Aug 12, 2004 11:00 pm
Location: London UK

Re: Loudness contribution of various frequency bands

Postby ultraGentle » Mon Feb 22, 2021 6:14 am

Ah, this is helping me tease apart issues I conflated!

May I get feedback on the thought process below?

(1) Track sounds balanced, but
(2) Track clips because of excess bass energy
(3) Turn track down
(4) Track no longer clips, but now below target loudness;
To raise track back up to target loudness,
(5) Treat bass to reduce energy while maintaining subjective balance,
(a) either via EQ to remove energy via bands,
(b) compression to remove dynamic range,
(c) or peak limiting to remove momentary clipping.

In particular, do step (5) and (a, b, c) make sense? Am I missing a better (5)?

Forgive the basic-ness -- I'm trying to "talk through" this so I can sort it out for myself. I welcome refinement or outright upending!
ultraGentle
Poster
Posts: 33
Joined: Tue Feb 16, 2021 1:09 am

Re: Loudness contribution of various frequency bands

Postby Hugh Robjohns » Mon Feb 22, 2021 11:19 am

ultraGentle wrote:May I get feedback on the thought process below?

(1) Track sounds balanced, but
(2) Track clips because of excess bass energy

...no, it clips because you are working with insufficient headroom!

(3) Turn track down
(4) Track no longer clips, but now below target loudness;

Then your track has an excessive Crest Factor for the specific medium you are intending to optimise the track for -- essentially the ratio of peak level to average level is inappropriate for that medium.

To raise track back up to target loudness,
(5) Treat bass to reduce energy while maintaining subjective balance,
(a) either via EQ to remove energy via bands,
(b) compression to remove dynamic range,
(c) or peak limiting to remove momentary clipping.

Option (c) is the usually the least noticeable and most appropriate solution. From your description, it seems most likely that you have wayward transients in the track that really need to be reduced, which is usually achieved by applying some peak limiting. Option (b) could also be used -- some gentle compression from a lower threshold will reduce the overall dynamic range without changing the sound character in an obvious way. Option (a) will inherently change the sound character...
User avatar
Hugh Robjohns
Moderator
Posts: 30067
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK
Technical Editor, Sound On Sound
In my world, things get less strange when I read the manual... 

Re: Loudness contribution of various frequency bands

Postby RichardT » Mon Feb 22, 2021 11:40 am

Hi Ultragentle,

There's another option you didn’t mention - reducing the volume of the channels in question, either by simply setting the channel gain, or using automation if you need to.

If I apply some limiting to a track I sometimes find a particular channel is triggering limiting consistently, for example a snare drum. Reducing the volume means limiting is not triggered so much and the loudness will go down a little bit. Plus the sound is not altered by the limiter.

But to be honest, I do this to maximise sound quality. I don’t think the effect on loudness is very significant.
RichardT
Frequent Poster
Posts: 644
Joined: Thu Aug 12, 2004 11:00 pm
Location: London UK

Re: Loudness contribution of various frequency bands

Postby ultraGentle » Mon Feb 22, 2021 2:29 pm

> not working with enough headroom
Ah, so I *should* be mixing with the master fader down, yes? I've been doing that as a stop gap, but now it seems that that was the right move all along!

> automation
Silly me, of course, thanks for pointing that out. Indeed an option (5d).
ultraGentle
Poster
Posts: 33
Joined: Tue Feb 16, 2021 1:09 am

Re: Loudness contribution of various frequency bands

Postby Hugh Robjohns » Mon Feb 22, 2021 3:00 pm

ultraGentle wrote:> not working with enough headroom
Ah, so I *should* be mixing with the master fader down, yes?

It's workable, but not ideal.

You *should* be mixing with the master fader at its zero mark, and all the channel faders close to their zero marks too. If the level is too high when set up that way, introduce attenuator plugins as the top insert of each source channel or reduce the level of each source track clip in some other way (depending on what your DAW allows).

Basically, your source tracks are recorded with too high an initial level -- inadequate headroom margin during tracking.
User avatar
Hugh Robjohns
Moderator
Posts: 30067
Joined: Thu Jul 24, 2003 11:00 pm
Location: Worcestershire, UK
Technical Editor, Sound On Sound
In my world, things get less strange when I read the manual... 

Re: Loudness contribution of various frequency bands

Postby ultraGentle » Tue Feb 23, 2021 5:34 pm

> You *should* be mixing with the master fader at its zero mark, and all the channel faders close to their zero marks too.

> Basically, your source tracks are recorded with too high an initial level -- inadequate headroom margin during tracking.

I did some reading on the above, and want to make sure my understanding is correct:

The first reason to follow this advice is so as not to push any analog recording gear too hard (e.g. preamps). Record at perhaps -18, with -12ish peaks.

The second reason is to streamline the mixing workflow.

Now, that said, as long as the recorded signal is clean (noise free, no clipping), there shouldn't be any problem with using tracks recorded too hot, since, as mentioned, these can be turned down with an attenuator.

Is that about right?
ultraGentle
Poster
Posts: 33
Joined: Tue Feb 16, 2021 1:09 am

Re: Loudness contribution of various frequency bands

Postby RichardT » Tue Feb 23, 2021 6:27 pm

The main reason is to prevent your DAW from clipping when recording an input channel. While DAWs can process audio at values above 0dBFS internally, they can’t record above this without clipping. You do need to ensure that your AD converter is not clipping and that your analogue equipment is not driven into distortion, but those are quite different issues!

A second reason is that some plugins expect their inputs not to be too hot.

A third reason is to allow yourself space to add processing to the track without risking clipping.

If you have a hot recording then yes you can attenuate it.

It’s best to have the channel faders close to unity gain because they are most accurate in this range.
RichardT
Frequent Poster
Posts: 644
Joined: Thu Aug 12, 2004 11:00 pm
Location: London UK

Re: Loudness contribution of various frequency bands

Postby Sam Spoons » Tue Feb 23, 2021 7:39 pm

In simple terms, analogue mixers/recorders have meters that show a max level of around 22dB above the reference 0 dB level. Digital systems don't because they use 0 dBFS (the max level the system can reproduce as their reference level). When we recorded into analogue systems we were happy to allow the signal to stray into the yellow much of the time and even into the red occasionally. But the yellow started at at 0dB and the red at +16 dB os somewhere in that region. On your DAW the yellow starts at maybe -18dBFS and the red at -6dBFS. The numbers are different but the warning implied by the colours is not so different.

Hugh explains it much more eloquently than I can here (with diagrams) https://www.soundonsound.com/sound-advice/q-how-much-headroom-should-leave-24-bit-recording
User avatar
Sam Spoons
Jedi Poster
Posts: 15118
Joined: Thu Jan 23, 2003 12:00 am
Location: Manchester UK
Still taking this recording lark seriously (and trying to record my Gypsy Jazz CD)........