Loudness - Everything You Need To Know
Not sure what all this loudness stuff is about and how it affects you? Confused about the difference between LKFS and LUFS? In this article loudness expert Mike Thornton answers all these questions and many more, including what loudness is and how it relates to both music production and audio post-production workflows.
Why Do We Need Loudness?
Whether it’s music production or audio post-production, why can’t we carry on doing what we’ve always done? Surely everything is fine?
Well perhaps not.
In music production, we have seen the loudness wars, as producers and engineers tried to make their music track be louder than any other track, and the resulting annihilation of dynamic range.
When it came to broadcasting, loudness-related issues represented a significant number of complaints received by broadcasters. Viewers and listeners complain about jumps in loudness. For example, looking at one 40 day period here in the UK, BBC TV received 100 complaints relating to loudness issues.
61 related to the background sound was too high, which is an issue of intelligibility.
20 were about volume jumps between content, so program trailers or announcements being louder than the content around them. On commercial tv channels, the biggest issue is with adverts and trails being much louder than the programs they sit in.
19 were about the volume range within programs being too high and this often relates to dialog being quiet and indistinct and then suddenly everything gets much louder.
Also when viewers change the channel the volume can go up or down. So overall loudness has become a serious issue for consumers.
So before we go any further, how do we perceive loudness?
What Is Loudness?
Loudness is the perceived strength of a piece of audio, how loud we perceive it to be and this can depend on level, frequency, content and duration. It’s not just about level.
Level is an electrical measurement, of the audio signal, whether it’s analogue or digital.
Before we go any further let’s take a look at how our ear works.
Firstly, the ear-brain combination does not register all frequencies in the audible spectrum equally. For example, lower-pitched sounds require more energy into the eardrum if it is to be perceived as being equally loud as mid-range speech frequencies.
This is not news, back in the 1930s research was undertaken by Fletcher and Munson from listening tests and these curves were drawn up from the results…
Secondly, although our ears react to changes in level pretty quickly, our brain does not register a change in “loudness” unless the level change persists for at least 400ms.
The History Of Loudness
The earliest way we measured loudness was to measure the RMS value of an audio signal mainly using an analogue VU meter. The dynamics of the VU (Volume Unit) meter proved to be a remarkably close approximation to the perceived loudness of the audio being measured.
So in broadcasting, why did we ever choose to measure Peak Level rather than RMS?
In the late 1930s, as radio broadcasting was being established around the world, there was a real need for a reliable meter to display that signal was being sent to the transmitter. In the UK, the BBC developed their peak program meter (PPM). At around the same time, the German broadcasters were undertaking similar research, but because of the timing, could not collaborate with the BBC but they too developed a similar peak reading meter.
When we moved to a hybrid analogue-digital world, we set our reference so that we had around 10dB of headroom above our maximum peak level and that headroom was never used.
Because of all the history since the very early days of radio, the delivery specs from broadcasters have been built around peak level this has resulted in everyone matching (or normalising) programmes to a matching peak level, usually around -9 or -10dBFS.
To help make things sound louder but remain in spec, audio compression techniques, similar to what happened in music production, were developed to make content louder, without increasing the maximum peak level, with the end result that we have been experiencing loudness wars in broadcasting with producers and directors each wanting their content to be louder than everyone else’s, whilst not breaking the rules by going over the maximum peak level.
Now we have broadcast chains that are fully digital from end-to-end there is no need to leave all that headroom unused and we no longer need to concentrate our metering on peak level and so can go back to measuring loudness.
The Solution - Loudness To The Rescue
As you can see, what we needed was a system that can measure loudness in the same way our hearing perceives loudness, and provide a definite measurement for the loudness of a program and then normalise our content to a standard loudness measurement rather than normalising to peak level.
To achieve this, we needed a meter that could measure and display loudness, in the same way, our ears hear loudness and to achieve this, the standard loudness measurement BS1770 was developed and has become a single universal standard for measuring loudness and all broadcasters delivery specs being based on this BS1770 standard.
A number of governments and broadcasters across the world have addressed this problem with legislation. In the USA, the law is called the CALM act supported by the ATSC A/85 standard. France and Spain have also passed laws for the control of all their broadcast channels’ loudness, using the EBU R128 standard.
Whereas other countries like Germany, Switzerland, Austria, Norway and the UK have voluntarily implemented EBU R128 recommendations across all their TV broadcast channels. Austria has even made the bold claim that they have reduced their loudness complaints to…ZERO.
Check out this free tutorial, which includes audio examples to help you hear the differences…
One Loudness Standard For Music Production And Audio Post Production
Following many years of research into how our hearing works and how we perceive sound, together with many hours of listening tests by organisations including Communications Research Centre (CRC) and McGill University in Canada the standard BS1770 was drawn up.
Without getting into all the technical details, a K-weighted filter curve, based on the research results from all the listening tests builds a bridge between our subjective impressions of loudness and the need for an objective measurement we can use to make comparisons.
The K-weighting method is an essential part of the global, open standard BS1770 defined by The International Telecommunication Union. That said, BS1770 has gone through a number of tweaks along the way.
Meters measuring loudness using K-weighting produce a measurement with a unit called LUFS or LKFS. Note that there is no difference between LUFS and LKFS, they are two labels for the same thing and are completely interchangeable.
It also helps to use a relative scale so our target loudness becomes 0LU making it easier to read the meter, a bit like 0dBU as an absolute measurement and -10dB being a relative one. Also, it is useful to know that a change of 1 LU is equivalent to a 1dB change.
When the BS1770 standard was set they determined how a compliant meter should measure the various parameters but they rightly chose not to specify how a compliant meter should look. As a result, there are a range of meters to choose from.
All BS1770 meters have a minimum of 3 loudness measurements, all of which are averages.
The first two, Momentary and Short Term, are tools to help us mix content that will be loudness compliant but they don’t normally form part of the delivery specs.
Momentary (M) averages the loudness over the last 400mS
Short Term (S) averages over 3 seconds.
Integrated (I) is an average over the complete program so far.
At the end of the program, it is the Integrated measurement that is one of two measurements that will determine whether a program will pass or fail. The other measurement that will pass or fail a program is the true peak level.
What Is True Peak?
Interestingly, with the introduction of loudness metering, peak metering is not dead. In fact, because we are working so much closer to digital headroom we need much more accurate peak metering, we just are not using it as a way of matching programs anymore.
True Peak differs from Sample Peak in that true peak looks at what might have happened between the audio samples. To be absolutely sure exactly what is going on, to measure what we now call the true peak value requires a meter to read 4 times faster, so making 192 thousand measurements per second enables us to establish what is happening between the samples.
This is really important as we are now working within 1 or 2 dBs of digital headroom and true peak readings can be up to 6dB higher than sample peak measurements. Because a program can pass or fail on True Peak it is essential we measure the peaks of the audio using a true peak meter and where limiters need to be used to make sure the audio never goes above the maximum permitted and that all limiters be true peak limiters too.
The green blocks represent the Loudness Range for different types of content
What Is Loudness Range?
Coming back to our loudness meter, there is another measurement BS1770 compliant meters take and that is Loudness Range.
So what is Loudness Range? It’s another measurement that we can use in conjunction with Momentary, Short Term, Integrated, and True Peak and gives us a measurement of the range of the loudness in a program, so how much light and shade there is. Normally a program will not be failed for too much or not enough loudness range. Rather it is a tool to help us get a sense of the amount of variation in loudness in a program. However, more recently some broadcasters and publishers have started to recommend Loudness Range guidelines.
The range is described in LU, and to avoid occasional extreme events from affecting the overall result, the top 5% and the lowest 10% of the total loudness range are excluded from the LRA measurement. For example, a single gunshot or one long passage of silence in a movie would result in a higher Loudness Range and skew the result.
Loudness Range is also a useful indicator of potential dynamics reduction processes in a signal chain, perhaps if the program is also going to be delivered on a mobile platform it might be necessary to take steps to reduce the loudness range to suit that delivery platform.
One Standard - Several Delivery Specifications In Audio Post-Production
Although we have the benefit of one universal worldwide loudness measurement standard, we have a range of delivery specs around the world. The two major groupings of delivery specifications are ATSC A/85 for the US and Canada and the EBU R128 for Europe. There are other delivery specs in other territories like Japan and Australia but ATSC A/85 and R128 are the key profiles that most delivery specs are based on.
As you can see from this table, they are all very close. When it comes to broadcast delivery specs, we will look at online and streaming later in this article.
The key difference now is 1dB or 1LU! ATSC A/85 uses a loudness reference of -24LKFS, whereas EBU R128 uses a reference of -23LUFS. R128 specifies a maximum of -1dBTP, whereas ATSC A/85 specifies -2dBTP.
There were bigger differences in early versions of A85 and R128 but now they have converged significantly. For example, early versions of ATSC A/85 didn't specify the gate, but the gate is now in the current version of A/85.
Misconceptions About Working With Loudness
Just before we finish, I would like to touch on a few misconceptions about working to this loudness standard.
Firstly, some people have assumed wrongly that content normalised to loudness has no dynamic range. I believe this misconception has grown out of a misunderstanding about normalization. We have got so used to normalizing up to a maximum, with peak normalization where as Loudness is about a centre of gravity, the average over the whole program. To have an average, there must be lower and higher values, which produce an average somewhere inside the range between the highest and lowest. Just as every object has its own centre of gravity, which won’t always be in the middle, so every program will have an integrated loudness figure (the average over the whole program), but the overall position of that average will vary, depending on the shape of the program, and that is OK, it’s more than OK, it is the way it is meant to be.
Advantages Of Mixing To A Loudness Standard
To conclude, what are the advantages of mixing to a loudness standard?
As audio professionals we get to determine what a program sounds like and have the confidence to know that the program we mix will be the same as the program the consumer hears and that it won't be messed about with downstream by any processing.
Because of this, it’s very important there are no black boxes in the TX chain. Some broadcasters have been sold a black box that, if inserted in the TX chain, will make them loudness compliant; in my opinion, they cause more problems than they solve and, most importantly, fly in the face of the spirit of the concept of loudness normalization and bringing the control back to the audio professional.
We can now make programs with more dynamic range, not less, because we no longer need to hit the ‘zero’ as hard as possible to make our program stand out.
We are no longer constrained by quasi-PPMs. Calibrate your monitoring, and then trust your ears. They are still the best loudness tools we have.
Why Calibrating Your Monitors Is A Key Part Of Delivering Loudness-Compliant Content.
In my article Are TV Mixes Getting Too Big For The Domestic Living Room? I took an initial look at this issue. The National Association Of Broadcasters Engineering Handbook states that larger rooms with a higher reference SPL will yield wide dynamic mixes, whereas smaller rooms with a lower SPL will yield more constrained mixes.
We know from the various standards, including the ATSC recommendations, that the recommended sound pressure levels for larger rooms are higher.
Categories of Audio Control Rooms Used In Television Production
Sound Pressure Level For Different Size Audio Control Rooms
Please use these tables as a guide to start form, but don’t be afraid to deviate from them if you find your mixes routinely too loud or too quiet. The key is to find a calibration point that is right for you so that you deliver loudness-compliant mixes without looking at the meters.
It doesn’t matter if you are working in surround or stereo; the principles are the same, but for the purposes of this demo, I am going to use a stereo system.
Your speakers should be all the same, whether for stereo, surround, or immersive. However, on a surround system, if space or budget is a constraint, the front three speakers should be the same, but the surrounds could be smaller speakers from the same brand.
The Role Of The Subwoofer In Professional Studio Monitoring
Bass management and LFE are two different issues that are often handled by the same speaker, especially in smaller systems, and this can cause confusion.
As audio professionals, it is generally recommended that we try and avoid using bass management. Bass management is necessary where the main speakers are unable to handle the very low frequencies in normal audio content, usually because they are too small. In this case, the system is designed to redirect the very low frequencies to the subwoofer, a single speaker designed to handle very low frequencies, and because low-frequency sound is not directional, it doesn’t really matter that it is all coming from one speaker.
In surround monitoring systems, the LFE channel requires the use of at least one subwoofer in the monitor system, but this should only reproduce what is in the Low Frequency Effects or LFE channel.
The Importance Of Monitor Level Calibration
Monitor Calibration is really important. Because our ears are excellent at perceiving loudness, if we set our monitoring so that we play sounds at the same volume level, we will naturally mix loudness-compliant content without having to be glued to the metering. We will be able to work to that consistent loudness centre of gravity.
We use a sound pressure level meter to calibrate our monitoring, but although there are recommended levels for different sizes of rooms, my advice is not to feel bound by them. Essentially, the smaller your room is, the lower your listening level should be. Unless you’re in a mixing space that is the size of a movie theatre (and has similar acoustics), 85dB-SPL will simply be too loud. Again we are picking up a standard that has come from the film workflow, but most broadcast sound is not mixed in film dubbing theatres, we work in much smaller spaces hence the smaller the room the quieter your monitors will need to be.
You can see from the table above some recommended levels for different sizes of mixing rooms. Use them as a guide to start off with, but don’t feel tied to them. Start off with the most appropriate sound pressure level for the size of your room.
Then start mixing and see how you get on. The key aim is to find the monitor volume that works for you where you naturally mix loudness-compliant programs without looking at the metering. If you find that, having calibrated your monitors based on this table, your mixes are coming out too high or too low, then adjust your monitor volume by the corresponding amount until you mix bang on the mark. If you get this right, you should be able to mix content within 1LU of your target loudness.
How To Level Calibrate Your Monitors
You’ll need two things to calibrate your monitors: a pink noise file and an SPL meter. A low-cost meter like this one is absolutely fine. These meters come with several different ‘weighting’ systems built into them, which are similar to, but different from, the K Weighting used in the BS1770 standard. For this application, you should use a “C-weighted” curve and the “slow” response setting.
You can use an app on your smartphone, but be aware that they may not be calibrated, so they won’t give you an accurate reading. The ones on Apple devices will tend to be more accurate because they can be calibrated to compensate for the mic in the device where as there are too many different devices in the Android market.
Pink noise is a signal that has equal power across each octave. In particular, you want a -20dBFS pink noise file. You can get these in a number of ways. One way may be to create it using a plug-in in your DAW or NLE. For example, in Pro Tools, you can use the Signal Generator plug-in, but you can also download a useful set of test/calibration files from Blue Sky, and on their site, they cover all the nitty gritty of monitor calibration in much more detail.
The advantage of using the bandwidth-limited signals in the Blue Sky folder is that they limit many of the room interaction effects often associated with measuring SPL and broadband pink noise, so my advice is that they are well worth using, especially if the acoustics of your space are not as good as you would like.
When you download the test files from Blue Sky, you get four files…
1kHz file recorded at -20dBFS for electrical calibration
40Hz to 80Hz bandwidth limited pink-noise file recorded at -20dBFS
500Hz to 2500Hz bandwidth limited pink-noise file recorded at -20dBFS
Full-bandwidth pink-noise file recorded at – 20dBFS
These test files are all mono files. You need to hard assign them to the left and then the right. Don’t use a pan control, and don’t feed more than one channel at a time.
The reality of most setups these days is our reference is around -20dBFS, so there is little need to use the 1K reference tone. You can go straight to the 500 to 2.5kHz band-limited pink noise.
As my studio fits in the 40 to 140 cubic meter category, I am going to demonstrate how to calibrate your monitors to 78dBC SPL. If you are calibrating for other spaces, use the appropriate SPL target.
Set your main monitor control to an appropriate point. The exact position will depend on your system and whether you have active or passive speakers.
Set your SPL meter on a stand at eye height at the position in the room where you would be sitting. Effectively, it should occupy the space where your head should be. Rotate the spl meter so it is pointing at the left speaker.
Assign the 500-2.5kHz pink noise signal to just the left channel. Ensure nothing is coming from the right channel or any other channels.
Now, adjust the gain control on the monitor speaker so that the SPL meter reads 78dB. Repeat for each channel, but if you have a 5.1 monitor system, then repeat for the centre, right, left surround and right surround until all the channels have been calibrated. If you don’t have bass management, then you are done.
If you have bass management, then feed a 40-80Hz pink noise signal to the left channel only. Adjust the subwoofer level control until the subwoofer reads 78dBC (slow) at the mix position. The meter will bounce around a little, so you will need to do a mental average (I tend to filter out the peaks in my mind, so I don’t set the sub too hot). Check that when you feed the right channel with the 40-80Hz pink noise you get a similar result, all being well you shouldn’t need to make an adjustment.
There is also a speech sample from the ATSC, which you can download and playback. You should find that this speech sample should be at a comfortable volume when played back over a calibrated system.
Now, start mixing content on the principle that if it sounds too loud, it will be too loud, and if it sounds too quiet, it will be too quiet. However, if using your calibrated monitoring, you find that your mixes come out too loud, say 2LU high, then readjust your monitor calibration and bring the level up 2dB.
Continue mixing and see if that fixes the problem. The key point is to calibrate your monitors to suit you, not to a specific number.
Creating Loudness Compliant Mixes For Broadcast And Netflix
Before the move to loudness-compliant deliverables, I had to mix programs so they would survive the transmission processing rather than create the mix I wanted.
Once again, we can trust our ears because when we set the monitoring level to the appropriate level and start mixing, the amazing thing is that our mixes come out right and, with a little practice, will hit the appropriate target loudness. Because our hearing is very good at perceiving loudness by setting the monitor levels appropriately, it just works. But hitting the target by ear is much harder if your monitoring isn’t calibrated. For example, I was mixing a program quite early in the morning, and so as not to disturb the family, I had my monitoring lower than normal. I missed the loudness target and had to adjust it in post-production.
People often ask how they will know how to get mixes to hit the target. There is an interesting fact that not a lot of people know, and that is if you set your speech level as before so that it peaks around -10dBFS or PPM 6 here in the UK, then measure it with a BS1770 loudness meter, it will come out at around 0 LU - the target loudness. Obviously, compressed speech will be perceived as louder, so it will also read slightly higher, but uncompressed speech at a normal level will come out at target loudness. In fact, one of the ATSC A/85 test files is a sample clip of speech that hits the ATSC target loudness of -24LKFS.
Peak Reading Meters - Stay Or Go?
As we moved from peak metering to loudness metering, people asked about whether to remove the old peak reading meters. For me, it is a matter of personal preference; although some people are adamant that all peak reading meters should be removed, some folk prefer to retain them as a known reference point. Ultimately, it’s up to you. I had a BBC PPM in my studio, but I found myself looking at it less and less, so in the last upgrade, I chose to take it out,
Mix to -3dBTP To Be Ready For Lossy Codecs
Then, when it comes to peak level, consider only going to -3dBTP because lossy codecs used to deliver content digitally, like mp2 or mp4, will distort if the input audio goes much above -3dBTP. With peak normalization and having a ceiling at around -10dBfs it wasn’t an issue because we weren’t close to digital headroom, but now because we can go up to -1dBTP with the EBU R128 spec or -2dBTP for the ATSC A/85 spec if we don’t do something about it when it gets to the coding stage in the transmission chain, the peak levels will need to be restricted to -3dBTP to meet the transmission specs to protect the lossy codecs, so why not create content that only goes up to -3dBTP so that the audio doesn’t get modified before it goes through the lossy codecs.
Hardware Loudness Meters For Live Mixing
For areas handling live audio, like studio control rooms and transmission areas, the TouchMonitor loudness meter is becoming very popular. You also might like to consider including the TC Electronic loudness radar option for your hardware loudness meter. The radar histogram is very helpful in giving a sense of where the loudness trends in your mix are going, especially as the nature of the radar display effectively means the histogram folds back on itself.
Use the meter in relative mode so that the target loudness is 0LU. It is easier to aim for, and if you are high or low, it is clearer as it will be displayed in negative numbers below and positive numbers above the target loudness.
Colour Choice Tips For Hardware And Software Loudness Meters
Do experiment with the display preferences, especially the colours for different loudness to help you get a sense of loudness with just a quick glance at the loudness meter.
A number of sound mixers have settled on a colour scheme, as shown in the diagram above on the left-hand side. Using green bands for each of the three loudness meters makes it easier to hit the target loudness. A popular configuration is having different green zones for each meter.
On the momentary meter, which is averaging the loudness over 400ms, have a green zone 3 LU above and below your target. With the short-term meter, which averages over the last 3 seconds, set the green zone to plus & minus 2LU, and the Integrated loudness, the average for the whole program, set your green zone to just 1LU above and below your target loudness.
Then use red above your green zones and blue below, all of which will make it very easy to see how close you are to the target. Remember that the momentary loudness will vary the most so significant swings in momentary loudness are absolutely fine. The short term will vary but somewhat less that the momentary meter and the Integrated will stabilize very quickly and only move quite slowly.
Where it isn’t possible to change the colour zones for each meter, I have configured my VisLM plugin to have a green zone for +2LU to -3LU.
Always Use A True Peak Limiter
In live situations, you should always have a true peak limiter in your studio output. Most digital limiters are sample peak limiters and so will allow intersample peaks through unlimited. Because true peaks can be anything up to 6dB higher than sample peaks, and we are now working within 3dB of digital headroom, it is essential that your output limiter is a true peak limiter and not a sample peak limiter.
There aren’t that many hardware true peak limiters out there, and if yours doesn’t shout it from the rooftops, then it almost certainly won’t be a true peak limiter. For a long time, there was one main contender, and that was the TC Electronic DB6, but more recently, the TC Electronic LM2 has been finding favour as a more cost-effective solution.
So that we can be confident that the mix we create is the same mix the consumer enjoys the true peak limiters should be the only processors in the delivery chain, all other dynamics processors should be taken out otherwise we end up fighting the transmission processing all our again.
Remember to set your true peak limiter to -3dBTP so as not to overload the lossy codecs.
Loudness Planning
Whether you are mixing live, or working in post-production, loudness planning should become an integral part of your workflow just as planning for channel and track counts, desk layouts etc, are all essential parts of good preparation, so loudness planning should become a critical part of the process whether you are working in a live or post-production context.
As you plan your mix, consider the comfort zone. The comfort zone has come out of all the listening tests that have gone into the development of the loudness standards. What they found is that there is a zone within which consumers don’t feel they need to reach the remote control to adjust the volume. This comfort zone is between +3LU and -5LU around the target loudness and is a useful guide to consider for the short-term meter. If the loudness goes outside this comfort zone for a significant period of time, then consumers will reach for the remote control, which defeats the object of the loudness workflow.
We recommend when mixing for post-production to do their dialog mix pass first the the aim of getting the dialog to stay fairly close to the target loudness. Then, on subsequent passes, add and mix the music and effects, setting them around the foreground sounds (usually the dialog) as feels appropriate. If you have calibrated monitoring then if an element feels loud then it will be loud, if it comes across quite it will be quiet, you can trust your ears.
Because it is harder to bring the average down, especially on the louder shows, consider running the opening credits and any other packages that will be played throughout a show, slightly lower. Because they are pre-recorded, consider running them through a loudness meter before the show and then playing them in a little quieter. Because these packages will now be a known loudness, this will also also help you keep track of the loudness and you will find it easier to mix live shows very close to target loudness. Established mixers can mix live shows to within 0.5LU using these techniques together with calibrated monitoring.
This is especially important on the louder types of shows like light entertainment with a live audience because it is very hard to bring the average down especially once you have passed the 10 minute mark. Check out the video for an example, which is the exception that proves the rule.
Consider Reducing The Attack Of Your Compressors
Because we can have a wider dynamic range and we are no longer constrained to a maximum peak level, we can now configure our compressors with a slower attack, which will allow the transients through but with manage and control content if it consistently stays loud. We can still use compressors and limiters to manage surprises, especially on live or as live shows but having more transients come through the mix will help to open up our mixes.
Mixing Tips For Adverts Trailers and Promos
When it comes to advert and promos often called short form content, all the rules still apply as well as all the opportunities for increased dynamic range and impact moments. However because of the intense loudness wars of I want my ad to louder than anyone else’s all signs of dynamic range have been squashed out of short form content. I am aware that clients have rejected mixes with just 4dB of dynamic range as being too much! They were only satisfied when the ad was remixed with no more than 2dB of dynamic range.
The loudness standards mean that there is no point in having content that is so heavily compressed to make it louder. If you do it will simply be turned down until it meets the target loudness.
However, people started looking for other ways of creating adverts that had an impact but met the spec. One trick was to create an ad where the first 25 seconds or so was very quiet and then drop in a really loud payoff line that, when averaged over the whole ad, would hit the target loudness.
A Change To The Delivery Specs
Consequently, in a number of territories, they have added a requirement for a maximum short-term or momentary loudness for any short-form content. For example, the PLOUD group responsible for the EBU R128 standard published a revision to R128 for short-form content, that is, anything less than 1 minute in duration, which specifies that the highest the short-term loudness can go is +5LU. Remember that normally, there are only two pass/ fail criteria: the integrated loudness and the maximum true peak. The momentary and short-term measurements were only there as a guide to help us mix loudness-compliant content. Now, a maximum short-term or momentary figure has been added to the pass/fail criteria for short-form content.
However, just as for long-form content, we can add more dynamic range so we can for short-form content, too. In the video, I play you two mixes of the same spoof ad, one mixed to a maximum peak level of -10dBFS and then the other to a target loudness and with more dynamic range; take a listen and watch the quasi-peak and loudness meters as well.
You can see from the histograms that the first version has very little dynamic range and is much louder than the second loudness-compliant mix, which also has more dynamic range. In the second mix, the second half of the mix is quieter and more restful, which sits better with the story. In addition, the explosions cut through the mix and have much more impact than they do in the first mix because they simply bash into the peak limiter.
As we get more experience and these concepts are introduced into the creative brief, we are seeing more opportunities for more dynamics in ads and promos. Take a look at the histogram for two different mixes of a Pet Rescue ad. The ad opens in a museum before the music and effects kick in halfway through. However, in the peak normalised mix, the museum sounds are as loud as the music and effects, whereas, in the loudness-compliant mix, the museum section is much quieter, as you would expect, and then when the music and effects kick in, they have much more impact.
In Conclusion
To recap, the opportunity to go through the roof beyond the old -10dBFS ceiling is very liberating and enables us to have those moments of impact where we can make things louder if we want to as long as the overall loudness for the whole program averages out.
We can add dynamics to our mixes and get away from the uniformity that we used to end up with when normalising and working to a maximum peak level, often referred to as ‘the sausage mix’.
Working to these new loudness standards is liberating, we an set our monitor level and mix trusting our ears, knowing it will come out right and we will hit our target loudness.
What About Loudness Workflows For Netflix?
What we have talked about so far largely relates to audio post-production for broadcast content, whether that is delivered to the consumer by cable, terrestrial transmission or satellite transmission. But with more and more content being delivered by streaming services like Amazon or Netflix. Do we have to amend the workflows outlined above to meet the requirements of Over-The-Top (OTT) and by Online Video Distributors (OVD)?
The Audio Engineering Society (AES) published Preliminary Loudness Guidelines in September 2016 that aimed to address the fundamental concern of audio Loudness in the developing segment of on-line television and video content delivery, from creation through distribution and to the consumer experience.
Then in October 2017, the AES followed that up with the full Loudness Guidelines for OTT and OVD Content.
In 2018, Netflix published their Netflix Sound Mix Specifications and Best Practices, which brought back the use of Dolby’s Dialog Intelligence software that was used by some broadcasters in the 90s as part of the Dolby Digital AC3 workflow using Dolby’s Dialogue Intelligence, their speech-gating technology, which measures loudness only on the segments of a program that contain dialogue.
This workflow was built around using the dialog loudness as the anchor or reference point and normalising to that, to enable content creators to set dialogue to a fixed loudness and mix the other content around it, based on the idea that viewers typically adjust their television volume controls according to the audibility and intelligibility of dialogue.
This differs from the BS 1770 based workflows, which use the loudness of the full mix, rather than just the dialogue, to normalise the loudness for each piece of content.
Here in Europe, Dolby’s speech gating technology never really got a foothold in broadcasting, we went from peak level normalisation straight to BS1770 loudness normalisation working to EBU R128.
Whereas in the US and Canada, Dialnorm and Dolby’s Dialog Intelligence and the use of the Dolby Media Meter were in common use but slowly phased out as broadcasters moved over to full mix normalisation.
Rolling forward to 2018, the industry was shocked when Netflix published their delivery specs to find that…
“Average loudness must be -27 LKFS +/- 2 LU dialog-gated. Peaks must not exceed -2db True Peak. Audio should be measured over full program according to ITU-R BS.1770-1 guidelines.”
Because we hadn’t used dialog gating for so long and Dolby had discontinued their Dolby Media Meter in August 2018, the phrase dialog-gated caused a lot of confusion. The other point was the normalisation number was -27LKFS, rather than -23LUFS or -24LKFS. Again an echo back to the old Dialnorm number.
If you want to get the full story and find out what we thought of this move by Netflix then check out our very comprehensive article Has Netflix Turned The Clock Back 10 Years Or Is Their New Loudness Delivery Spec A Stroke Of Genius?
Following this article, Scott Kramer, Manager, Sound Technology | Creative Technologies & Infrastructure at Netflix reached out to us to respond to our article, and here are some excerpts from that response…
“In writing our specifications, we seek to preserve the best possible experience for Netflix members, while protecting the creative freedom of the creators. In the case of the new loudness specification, both groups were aligned. Many Re-Recording Mixers asked for dialog-based measurement to simplify their workflow and to ensure loudness consistency for the audience….
… With the new spec, we sought to change how mixers measure, but not how teams mix. We measured our content which was compliant for the -24 LKFS +/- 2 LU full program spec, and found that dialog levels clustered around -27 LKFS. This approach, by using the dialog loudness, allows for as little workflow change as possible, while providing more dynamic range for creators that feel they need it. The choice of stating a 1770-1 measurement comes from the Dolby Media Meter terminology for a dialog based loudness measurement. Dialog based loudness measurement does not require the use of the relative-level gate specified in BS.1770-4, as the DI algorithm already applies a gate (a dialog-based gate) to the audio.”
You can read the response in full in our article.
Netflix Loudness And Dynamic Range Developments Update
You can read the response in full in our article. All this talk of Netflix brings us nicely to other online content…
Netflix continues its work to improve the experience for its subscribers across a variety of devices. In this article, drawn, with permission, from a post on the Netflix Technology Blog, we explore how they measure program dynamic range and then use dialog normalisation and dynamic range reduction to improve the experience for subscribers, especially those using mobile devices.
In this article, we cover the following…
Dialogue Levels and Dynamic Range
Loudness Management
Peak Audio Sample Metadata
Putting it Together
Seamless Switching and Adaptive Bit Rate
Deployment, Testing and Observations
Volume Level
Volume Changes
Audio Sink Changes
Conclusions
Acknowledgements
You can learn more in our article Netflix Loudness And Dynamic Range Developments Update.
How To Optimise An Audio Mix For Delivery To Netflix
To help the growing number of audio professionals creating content for Netflix, the streaming company has produced a tutorial video designed to provide useful advice on how to optimise a mix for their streaming audiences, insights on how Netflix members consume Netflix content, as well as sound mixing guidelines and deliverable requirements for original version sound mixes, M&E mixes and alternate language sound mixes.
In this video, Netflix talks about sound mixing for Netflix, aiming to answer these questions…
How should I prepare sound mixes for Netflix?
How should I prepare M & E mixes for Netflix?
How should I prepare dub mixes for Netflix?
What is the purpose of Netflix sound specifications?
How may I ensure a sound mix plays well on the service?
How does Netflix process and stream sound mixes?
How does Netflix optimize sound experiences for members at home?
The goal is to enable a clean sound pipeline from the microphone onset to the speaker in a Netflix member's home by protecting the sound pipeline, we're able to provide members with the sound mix that the creator's intent.
If you would like to also read a full transcript of this video with stills from the video then check out our article How To Optimise An Audio Mix For Delivery To Netflix.
What About Loudness For Online Content?
Ever since we started using -23LUFS or -24LKFS for broadcast content, there has been a consensus that -23/-24 would not be suitable for portable devices because there wasn't enough gain in the headphone amps to deliver an acceptable volume, with content made to -23/-24.
A number of services, including Apple, settled on -16LUFS and we deliver our weekly podcast at -16LUFS too.
In late 2015, the Audio Engineering Society (AES) published Technical Document - AES TD1004.1.15-10 - Recommendation for Loudness of Audio Streaming and Network File Playback in which they recommend a window of -16 to -20LUFS. Looking at the graphic above, many of the key players in streaming have gone outside the AES recommended window, with Amazon, Spotify, TIDAL and YouTube have all gone for -14LUFS.
In 2017, the Loudness Petition Group was set up by Eelco Grimm, Bob Katz, Matt Mayfield and Ian Shepherd with the express aim of bringing peace to the loudness wars. If you want to know more then we explored this in detail in these two articles…
Bring Peace To The Loudness Wars - Recommendations For Loudness Normalisation - Part 1
Bring Peace To The Loudness Wars - Recommendations For Loudness Normalisation - Part 2
In the first article, we looked at the Loudness FAQ from the Loudness Petition Group, which expands on the issues and covers some of the misconceptions and urban myths that have already grown up around this whole subject.
In the second article, we looked at the research that Eelco Grimm undertook with HKU Muziek en Technologie, University Of The Arts in Utrecht in the Netherlands, in cooperation with the music streaming service TIDAL and presented at the Berlin 142nd AES convention.
More recently, in April 2020 the European Broadcasting Union (EBU) updated their R128 Loudness Standard to include specs for Streaming Services. R128 s2 (Loudness in Streaming) recognises that streamed content - both live programmes and files - has become “an increasingly important distribution method for EBU members”.
It points out that conventional broadcasters are now competing with the likes of Facebook, Apple, Amazon, Netflix, Google and Disney, as well as music or other audio-only services such as Spotify, TIDAL and podcast platforms, which typically use a higher Target Loudness than is used in broadcasting.
You can read more about this in our article EBU Loudness Standard Updated Including New Spec For Streaming Services.
When it comes to Loudness Range the higher the target loudness is the less headroom there is to support content with a good dynamic range. That said, most of this content is going to be consumed in noisy, or at best, less than ideal environments and so I would recommend that rather than drop the target loudness to accommodate the additional dynamic range that content destined for consumption on mobile devices is mixed with a relatively low loudness range, it which case the reduced headroom is no longer an issue.
In reality, broadcast-related content is only one part of the content that is streamed to consumers. Music streaming services are playing a greater and greater role in this sector and so this leads us to how loudness-based workflows are impacting music production.
What About Loudness In Music Production For Streaming Services?
Over the last 20 years, the phrase “loudness wars” has become enmeshed in conversations about music and audio, and for good reason.
We’ve been hearing the phrase for decades, long before we had digital limiters. We now have limiters that allow us to push level in a way that we never could before. With or without these tools, the concept of making more impactful and compelling records that contain good amounts of dynamic contrast and loudness isn’t going away.
Given that this is the case, we should dig a little deeper into the craft of music production to think harder about what contributes to the experience of loudness.
There are two key considerations in loudness, average level, and the distribution of energy across the spectrum. If you make a record focused in the upper mid-range, it’s going to be louder by definition (if the average levels are the same) than a record containing more low-frequency information, proportionally speaking.
When we talk about dynamics, whether that is dynamic range or musical dynamics, very often what we are really talking about is contrast. We mean that something’s louder than something else—a nearby truck horn is louder than one at a distance, a verse is generally not as loud as a chorus, and the chorus is louder than the verse.
When the time comes to consider pushing the perceived loudness in a mix or master, we need to keep this dynamic contrast between sections in mind.
Is there enough of it? When we compress, will it diminish the sense of dynamic contrast?
If we are concerned about the perceived loudness overall, where is the energy focused? Is it focused in such a way that the track will appear loud enough to the listener?
So how loud do I make my track for streaming?
The question of how loud should your master be if you’re distributing to Spotify, Apple Music, Amazon Music, vinyl, and so on, might be the wrong question to ask. Perhaps the right question should be…
“Can I make this louder and have it sound at least as good—if not even better—than it sounds now?”
Adjusting the way we conceptualize good and loud enables us to start to truly develop the craft and skills of mastering (and mixing), managing the tradeoffs, recognizing when we've got to the place where the music is as loud as it should be, sounding as good as it can be so when it's adjusted for playback, either turned down or if we have to turn it up for certain kinds of applications, it will still sound as good as we can possibly make it sound in that moment.
The music distribution world has changed. Revenue from audio streaming services is higher than all other music delivery formats combined, it’s even more important to understand the impact that streaming has on audio production and how best to optimize our mixes to make the most of the way streaming services deliver our content to the consumer.
In this series of 3 articles and free video tutorials in conjunction with Nugen Audio, our own loudness guru Mike Thornton takes you through the history, why loudness is important for music, and how to prepare your mixes ready for music streaming services.