Monday, August 26th, 2019

In search of loudness

By editorMarch 15, 2015


vumeterYou’re sitting in front of your DAW system, whether that be laptop based with headphones or a dedicated room set aside for recording and mixing, and you’ve got to the point where you are really happy with your final mix of a track.  You want to hear how it sounds real world, so you down sample to 44.1kHz 16 bit and maybe burn a CD to have a listen in your front room audio system.

And it sounds really quiet!  What’s happened to the wonderful mix you have just crafted?  Maybe something’s gone wrong in the down sampling?  Maybe it’s your hi-fi system?  So you pop in a well known commercial release of a similar genre and it sounds … just fine.

The temptation is to go straight back into the 2-buss and reset the final buss compressor from your carefully tempered 2:1 soft knee ratio and compress the life out of the mix in search of that commercial sound.  Sadly some of the commercial releases out there do appear to have been given that treatment, but there is another way …


Peak Limiting

Peak limiters are now available in most DAW systems, either as part of the standard toolset or as separate third party plugins.  The concept is simple.  In many genres, most well crafted mixes have a small percentage of content which peaks well above the average level of the mix.  If those peaks can be gain reduced without unacceptable audio degradation then the whole mix can be pushed up in level by a valuable few dBs.  How much will depend on the ‘natural’ dynamics of the material and how it has been mixed.

My advice is to have a visual look at the track waveform before bringing a final stage peak limiter into play.  Let’s take a look at a track which I have recently done as a client-mix for the artist.  Relatively simple four track song with stereo acoustic guitar, main and ambiance vocal tracks.


The track has been normalised to peak at -1dBFS which gives a sensible safety headroom margin.  I’ve set level markers at -4dBFS so you can get a visual idea of how much of the audio sits in that 3dB window.  Having reviewed the waveform display, I was pretty confident I could peak limit by 3dB without having any negative impact on the quality of the mix.

Nowadays, in-the-box peak limiters use what’s called a look ahead technique.  By inserting a small audio delay into the processing path, the limiter sidechain is able to see, in advance, what’s coming along in terms of peak levels and plan how to deal with them before the peaks arrive at the limiter’s gain reduction stage.  Usually, the look ahead delay is fixed, typical values are around 1.5ms, but there are some ‘off-line’ limiters which offer variable look ahead times so you can choose how much time you want the peak limiter to have to bring the peak within the maximum limit you set.

Let’s put the track in question through an off-line peak limiter which has some useful additional tools to help visualise the operation of the peak limiter.  In this case from the classic Syntrillium Software ‘Cool Edit Pro’ Audio Editor (purchased by Adobe in 2003 and now marketed as Adobe Audition).

loudness-2The advantage of an off-line peak limiter is that you can send the whole track through a pre analysis stage as shown opposite.  Knowing that my track currently has a peak level of -1dBFS, I’ve dialled up a potential level boost of 4dB in order to have the software calculate how many samples would clip (usually this means three consecutive samples of 0dBFS) if the peak limiter was not in circuit.  (4db to clipping is the equivalent sample count for a 3dB level boost, limiting at -1dBFS).

You can see from the figures in the screen shot that the software has calculated figures of;

Left Channel = 0.002%

Right Channel = 0.009%

This is the percentage of samples which would clip were you to gain boost without putting the peak limiter into circuit.

As the software can tell me that the entire track has 25,009,879 samples for each channel, and I know I’ve recorded at a sample rate of 88.2kHz, some calculator crunching gets us to;

Left Channel, 500 samples, means 5.7ms of audio would have clipped

Right Channel, 2,250 samples, means 25ms of audio would have clipped

This is a good indication that I’m not planning to dial in too much gain boost, pre-limiter.

In reality the number of samples recalculated by the peak limiter will be quite a bit higher due to the attack and release time of the limiter, necessary to avoids clicks caused by too sudden changes in level but it’s still a reasonable visual guide to plan your provisional limiter settings.

Just to complete the visual picture, here are ‘before’ and ‘after’ waveform images of the peak limiter in action on a single peak with 3dB of gain applied before the limiter.  On the left hand side of the waveform displays, you can see  that, after the gain boost and limiting have been applied, the level of the waveform is indeed 3dB higher (have a look at the first major waveform peak).  As you move along the timeline, you can see the look-ahead peak limiter in action,  smoothly gain reducing so that the peak  (where I’ve placed a yellow marker) which would have clipped,  is nicely controlled to sit just within the -1dBFS set in the ‘limit max amplitude to’ box.






And here is the entire waveform having undergone a 3dB gain boost and peak limiting.  It’s important to note that the vast majority of the audio track has been untouched by the limiter but I have gained 3dB in level.



A word on stereo imaging

Traditionally, when compressing stereo material, you would normally link left and right channels together so that gain reduction is equally applied to left and right channels to avoid stereo image shifts.  As peak limiters operate for very short durations, there is a case for not linking channels so that only the channel requiring limiting (may be just one, may be both) gets processed.


Audio Quality

I tried out two software peak limiters on the track I was mixing.

The Cool Edit Pro limiter, I set with an attack time of 7ms and a release time of 100ms, and this seemed to cope well with the peak dynamics.  I tested these settings with gain boosts of 3dB and 4dB and it was difficult to tell which was which during ‘blind’ listening tests when the 1dB of level difference was accounted for.  Beyond 4dB of gain boost audio quality quickly deteriorated as the peak limiter started to work on significant proportions of the track’s audio, so I settled on 4dB for this track.


As I was currently undertaking a review of the new Sonar Artist DAW, I also tried out the Boost11 Peak Limiter which is included in the software.  Boost11 has a fixed look ahead time of 1.5ms and a nice ‘programme dependant’ release time which means you don’t have to worry about setting the correct release time.  Again I found that 4dB of gain boost was as far as I could push into this particular track.  Note that I have dialled up an excessive amount of boost only so I could snapshot a view of the little waveform displays which show audio being limited in red.

These are just the two peak limiters I had to hand.  Pretty well every DAW and plug-in company have look-ahead peak limiters available within their product range so its worth having a good look around should you need to add one to your audio toolset.


What if I want to mix out of the box?

The huge advantage of using a software based peak limiter is that the look ahead facility means that the limiter has enough time to begin to gain reduce ahead of reaching the peak which is something very difficult to achieve in an analogue peak limiter.  I guess you could put a pair of UA 1176s across the two buss.  These have an incredibly fast attack time of 20uS (that’s less than 2 samples when working at 88.2kHz in the digital domain) but they impact a definite audio signature to the audio all of the time.  Although wonderful for some applications, this may not be what you want to add after spending a lot of time getting your mix to where you want it to sound.   And even the 1176 cannot guarantee to capture every potential ‘over’.

I tried processing the 2-buss into a relatively fast transparent limiter with an attack time of around 100us but it simply could not grab every peak reliably, even when reducing the limiter threshold beyond what was acceptable to preserve audio quality.

A few analogue designers have attempted the impossible, to create a 100% analogue limiter with predictive look ahead facilities.  One technique I designed into a limiter was based on dynamically altering the limiter threshold depending on the acceleration of the audio level increase as it hurtled towards the nominal limiter threshold.  It was incredibly complex to design but worked pretty well.  However, considering that most material is going to be released digitally, software based peak limiting is the sensible approach for most situations.


Where to sit a peak limiter within the 2-buss?

The simple answer is right at the end.  Get the mix so you are happy with it within your studio listening environment, then see what you can usefully achieve in terms of gain boost.

Occasionally, after peak limiting, you might decide to revisit your 2-buss compressor settings, feeling that you can compress a little less now that you have bought some level boost through peak limiting, but I suspect that road leads to manic uncertainty … or worse!


How much is too much?

I’m of the opinion that final stage peak limiters should only be operating for a very short durations and occasionally throughout the audio track.  I don’t see the logic in using a peak limiter on every beat of a bass drum or bass guitar unless you want to use it as a creative effect.  There are compressors which can do a much better job, either inserted in the track or across a stereo sub mix, as appropriate to the track and your mix down workflow.

So do give this technique a go.  It works really well and will get you closer to the typical levels achieved on commercial releases.  If you are going for commercial release but are doing your own mixing, then leave this job to the mastering house who will have the tools and experience to optimise your levels and dynamic range.

Robert Campbell
March 2015

Share on Facebook
Share on LinkedIn
Tweet about this on Twitter
Email this to someone
share with your friends

© audio times 2018. All rights reserved.

Website by Small Business Marketing