Limiting, perceived loudness and normalisation in streaming
I always find it interesting discussing limiting audio. This article will explore peak reduction and why it is a distraction to the goal of increasing perceived loudness in mastering; but it is one of the tools required to work in the digital realm.
Digital brickwall limiters are a key aspect of this, but a clear comprehension of the side-effects of peak reduction, its imposed change on tonality and the possible degradation of mastered material is required to effect positive change.
There is often a lack of familiarity around these aspects; some of this stems from an absence of understanding in the practice of equal loudness but also from the way manufacturers present digital limiting plug-ins. Prevalently, these work on the basis of turning up the program material, just making it louder pushing into the limit process. They are not focused on equal loudness; delivering effective peak reduction relative to the original source with minimum impact on the audio. The Waves ‘L’ series is one of the exceptions to this having the ability to link the reduction and output controls. However, this is not its standard mode. This again makes the first engagements with the use of a limiter for new learners hard to not over limit as louder does sound better. Until they start to comprehend listening at equal loudness, but once bad habits are learnt, they’re harder to undo.
Another factor is the misunderstanding of the term ‘limiter’. It was originally associated with analogue limiting. This is not brickwall as with many digital limiters. Analogue limiters, with the exception of some modern additions, cannot ‘look ahead’ and do not have an ultra fast attack time, 1/15ms at best. The term limit means high ratio gain reduction and an aggressive slope/knee in practical terms. This is a much softer outcome than that of a digital brickwall limiter but the latter is required to avoid clipping within the digital domain. Not an issue within the analogue domain.
If we ask ourselves what we are trying to achieve in the mastering process we should be thinking about making a track ‘fit for purpose’ in relation to the output format. This could be CD, Vinyl, Digital Download / Streaming, lossy or HD. If we’re truly concerned with dynamic range, we should be focused on making this optimum in terms of power for it’s final consumer listening. If going to an analogue format such as vinyl, we’re not restricted by the digital ceiling. There is no ‘headroom’, clip maximum as with digital, the focus is towards the rms. So why should we be concerned with digital brickwall limiting when the principle of dynamic range is the same in both realms. This being the case, the digital limiter is there to stop clipping in the digital mastering process - not to make the audio perceptively louder.
Our focus and concern should be with the average level and not the peak in the mastering process. The limiter should be there in a digital system to catch the peaks because we have a digital maximum. The perceived loudness of a track doesn’t come from the direct reduction to the peak transient with a digital limiter (which will only dull the sound) it comes from good frequency balance and use of effective compression to even the rms (power) of the audio. This effective use of manual gain riding, downwards compression and upwards (parallel) compression processing in addition to classic limiting (high ratio gain reduction with a fast attack) can achieve the required amount of dynamic range reduction to make most audio ‘fit for purpose’.
There is a rationale to using digital limiting when we are trying to even out the peaks to achieve a smooth sounding mid/high end, especially with drum peaks. This is often the case in Metal, Rock and some Dance music genres. With an even peak to drum transients it is adventitious especially when consumed at louder volumes. Making it more impactful. Often I find this type of limit is best placed earlier in the transfer path chain, if not a couple of times to reconfirm the smoothing. But again, we should ask ourselves what the limiter is actually doing to the sound. We spend hours and hours using compression in a mix to put the attack (definition) into our mix to achieve the clarity we require. If we just apply a limiter, all we are doing is to remove this good work by removing that very definition! If you’re getting more and two decibels gain reduction on the limiter I’d say you’d be applying too much at that point in the transfer path.
With the advent of loudness normalisation on some streaming platforms it has abated somewhat the quest to making music louder than other programmed material. Which was definitely an outcome from the era of compact disc. Though as this normalisation is not universal or even default in terms of the user interface the requirement for relative loudness to other material still exists. This negates the directive to the mastering engineer to embrace an era of wider dynamic ranges especially as these services are delivered with post ingest processing. One day normalisation standard might be -14dbfs to be changed the next to -10dbfs. Or a different selectable set of levels as with some. Hence, we also want to avoid the post software applying a limiting process to our mastering audio, which it will if our dynamic range is wide when it levels it up to the narrower dynamic range selected by the user.
What makes a track sound ‘Mastered’ is not a brickwall limiter squashing the living daylights of a track but good attention to frequency balance and the average weighting of the rms. In smoothing out the average rms to enhance the listening experience the quieter sections of a song will cut though in noisy playback environments and, equally, the louder sections won’t override this average. Both things, if left uncorrected, would cause the listener to reach for the volume control. If we just ‘limit’ a track to make it perceptively louder we still have the same issues with the rms, hence the track will still sound unmastered but possibly perceptively as loud a ‘commercially’ mastered track in some sections. Though obviously if we make an equal loudness comparison back to the original it would show the track has been dynamically hyper-compressed and dulled. It doesn’t sound as good as the source. Hence, we haven’t met our basic mastering goal of making the track sound better!
If we take any limiter and reduce the peak detail of a mix more than a couple of decibels and make an equal loudness comparison we will hear the dulling effect of this limiting. Different limiters will sound more or less intrusive than others but they will all dull the tone, the more the reduction the more the dulling to the point where the depth of field and general definition starts to be reduced. This is a basic error to make in trying to master tracks. If we want to make our tracks ‘loud’ we need to look at frequency balance (tone) and average track rms.
We need to employ focus around the tone of the track as well as seeking to even the dynamic. Often this needs a combination of dynamic and tonal processing, including parallel and downwards compression; soft clipping can be a helpful tool in controlling stronger percussive elements in conjunction with limiting.
I use these approaches and classic analogue style limiting everyday in commercial mastering. I’m wholly aware of how these tools affect the sound on the source material and how to compensate. I use the digital brickwall limiter as an application at the end of the chain after the main mastering process relative to the output format to keep within the restrictions of the digital realm and not for the sake of loudness.
John Paul Braddock
Mastering Engineer
