Whether you like it or not we have ended up with a Loudness war in cinema sound, even though there is a fully calibrated system to prevent this, the reality of the situation is that the loudness levels in movies have gone up and consequently, due to complaints from cinema goers, playback volumes in cinemas have consequently come down. As a result, many cinemas keep the playback level low for all movies, which puts pressure on post-production facilities to adjust their playback level.
But it's not just about overall level, there is also a growing trend for cinema mixes to be mixed with less dynamic range than TV mixes, to the extent, that there have been occasions where the remix for TV has ended up increasing the dynamic range rather than the expected need to reduce the dynamic range of a theatre mix for TV broadcast.
That is just plain wrong.
One possible solution is to adopt the BS1770 loudness standard we already use in broadcast workflow and music streaming and apply it to the cinema workflow, with a suggestion to set a maximum for the Program Loudness of -27 LUFS (-21 LUFS for ‘loud cinemas’) and a maximum Short Term Loudness of -6 LUFS. The expected result of this solution is that cinema playback volume can then return to a normal reference level. Eelco Grimm, one the team behind the Loudness Petition Group that we have covered recently regarding standards and workflows for music streaming has also been studying cinema workflows and with his permission, here is a paper he presented to the AES in May 2013. Over to you Eelco...
In the days of 35mm film, someone from Dolby had to be present during the mastering of a cinema release. Masters were always printed with the volume set to 7 on the Dolby box. The trend at the end of the 35mm era was already towards very loud mixes. People present during the mix of movies like Tron were advised to wear hearing protection. The audience does not appreciate this level anymore, there are lots of complaints.
In Belgium, there has even been a case of hearing injury which led to legal action by the Belgian government. This will put an official limit to the volume and cinemas routinely now put the volume setting much lower than 7.
With the DCP workflow, the Dolby official is no longer present during mastering and playback volume control during post-production is also often lower than 7. The result is that headroom suffers. The current dynamics in the cinema are now smaller than that with a R128 television broadcast. Is there anything we can do to turn back to normal?
This is a graph that shows all weighting curves that matter to us. ‘A’ weighting is originally based upon the 40 phon curve, but nowadays is mainly used in Sound Level Meters to predict potential hearing loss from environmental noise. The ‘C’ curve that was originally based upon the 100 phon curve, finds some use in electro-acoustic system calibrations. The ‘M’ curve stems from Dolby. It is based upon the CCIR curve, which was developed for measuring the annoyance of noise. Ironically it is now used for measurement of commercials in cinema. ‘K’ weighting is the now well known ITU1770 curve, developed for subjective loudness estimates of electrical signals. In this article, I will freely juggle with mixing A, C, K and M weighting aware that this is a compromise which is dependent on the spectrum of the sound.
Loudness measurements are routinely used with gating: either a foreground gate (PL) that takes all but the low levels into account or a dialogue gate (VL) that only measures dialogue. Which one to choose depends on the goal of the measurement.
Since the director wants to have control of the loudness that his audience experiences, cinemas are level calibrated. The SMPTE standard says that when the Dolby box volume set at 7, the special ‘Dolby noise’ signal should give 85dBC per speaker. The bandwidth limited signal is designed in such a way that it also gives 85dBC. It is 0 VU on the meter and its loudness is - 21 LUFS.
This level is intended for a forte passage, it is quite loud. The average level of a movie should normally be approximately 6 dB below that, in film work, using digital audio,
it was generally agreed that dialogue levels were consistently running 30 dB below full scale, giving film audio 30 dB of “emotional” headroom. Dialog is therefore 7 to 9 dB below the reference level. This would be equivalent to app. 76 dBA.
|Dolby Box Volume Control|
|Dolby Scale||SPL dBC|
This table shows the scale of the Dolby box volume control. The reference noise is 85dBC at "7", 80dBC at "5.5" and 75dBC at "4". Below 4 the scale becomes 10 dB per 0.5 step. At 3.5 the ref noise is at 65dBC. From the Dolby CP650 manual I quote:
The main front panel fader knob on the CP650 controls the volume level in the theatre. It works in both normal and bypass operation. When the CP650 has been correctly installed, setting the fader to 7.0 plays the film at the level at which it was mixed. This is the proper level for any Dolby encoded film. Although a minor adjustment in playback level might be required under unusual circumstances, avoid significant deviations from the correct level 7.0 established by the installer. If the playback level is set too low, dialogue may be hard to understand; too high a level may cause complaints from the audience, and under extreme circumstances, can damage the theatres sound system.
Then somewhere something went terribly wrong. The original SMPTE standard for commercials and trailers was 82 LeqM, which is approximately 6 dB above the dialogue level. Next, the standard nudged to 82 LeqM for commercials and 85 LeqM for trailers, or 9 dB above dialogue.
Today commercials and trailers are all normalised to 85 LeqM. This is because from experience the trailers and commercials are played back at a much lower level than 7 anyway. In fact in the Netherlands 3.5 is used for commercials and trailers, or 20dB lower than '7'...
It's All Too Loud!
It's not just commercials and trailers that are played back much lower than '7'. Because the audience started to complain more and more about the main movie being too loud, cinemas have responded and reduced the level of the main movie too and there are good reasons for the complaints.
The Flemish Association for Tinnitus and Hyperacusis measured sound levels in several cinemas in Belgium. They measured peak levels up to 118 dBA! How is that possible? Well: one channel is calibrated for 85 dBC at -21 LUFS and therefore peaks at approximately 105 dBSPL. But there are 4 channels (surrounds are measured as one), which means 6 dB more. With full-scale clipping distortion, you gain another 3 dB per channel, and of course close to a speaker the level can be higher than in the middle of the room. 118 dBA seems insane, but in theory, it is possible. And the future is interesting. With Dolby Atmos, every single channel is calibrated to 85 dBC. The Atmos peak demand is 115 dBC per loudspeaker. A system can have 64 speakers, or in other words, the theoretical maximum SPL can be 133 dBC...
The Damage Is Done
In Belgium, in September 2010 a 17-year-old girl suffered permanent tinnitus when visiting the movie Inception. This received a lot of attention from the press and the government. In Kinepolis' cinemas, the absolute maximum is 5.5, with position 5 as the norm. This summer (2013) the Belgian government will make a law for a maximum allowed loudness levels in cinemas. They might even put a legal maximum level to the volume control.
But in general, it is not the risk of hearing damage that makes cinemas turn the volume down. It's the audience complains that the film is too loud, even long before it becomes a potential threat for their hearing. Another cause for complaints is that due to poor sound isolation loud scenes of the neighbour's movie disturb their own. Because of this, many cinemas turn the volume down and also modify the original calibration of the cinema. LFE's are often turned down to limit crosstalk to the neighbouring room and many times the balance between C, L, R and surrounds is off. Nevertheless, it seems like many cinemas are at least calibrated once in a while for a festival and I think the average cinema will be quite close to SMPTE calibration. Deviations from this are certainly not the main cause for a playback lower than 7.
We should emphasise that SMPTE calibration is important and should remain the reference, even with DCP workflows.
3. Indications Of The Reality
I called a couple of operators of Art Houses and Multiplexes. My informal survey already shows a solid trend: no movie is ever played at 7 anymore, 5 is standard. Commercials and trailers are routinely played back at 3.5, maximum 4. A check of the playlist notes of arthouse Studio K in Amsterdam reveals playback levels for the main movie of between 4.2 and 5.
The operator of arthouse Louis Hartlooper in Utrecht tells me that every Thursday various parts of all movies are viewed in their cinema (this is much easier with DCP than 35mm). Even when movies move to a new cinema room, they are viewed again. The optimum level is set by ear and programmed into the DCP server. Almost always that level is between 4 and 5. Another arthouse in Utrecht, 't Hoogt, also previews every movie on Thursday's. On average they end up at 5.5, with 4.5 being the softest. Once in a while, they come across a movie that can be played at 6.3. Their goal is that soft sounds should be audible, loud scenes not too loud. They want to avoid complaints from the audience.
The experience is that in larger rooms the level can be a little higher than in small rooms. In the new film museum Eye in Amsterdam new movies are also previewed in the cinema. Because of its main task, most movies are digitised old movies. The level varies, but many movies are projected at 4, the maximum after one year of operation is 5.4.
In Multiplexes, the problem is that many of the bigger theatres do not employ true operators anymore. Everything is completely automated, with a handful of people - mainly bar personnel - running the complete facility. The operational manager assembles the projection program on his office computer and most of the time he sets the playback volume of the main movie to a default position. This can be 5 or 4.5. Trailers and commercials are often played back at the same setting as the main movie (which can be really loud), or at 3.5 to 3.8. In special cases such as the premiere in a major cinema like Pathe Tuchinski in Amsterdam, the movie is played back according to the wishes of the director. But for the normal viewings, the level is always adjusted. For instance, the recent James Bond (in 2013) premiered in Tuchinski Cinema 1 at 7, and was played at 5.5 for the normal viewings and still led to complaints.
A special case is Kinepolis in Belgium. For a Multiplex organisation, they do care a lot about sound quality, installing high-quality loudspeaker systems and taking care to have proper acoustics. But they care even more for their visitors. The board of directors decided to set the standard playback level at 5, maximum 5.5, but even wth that, they still check every movie in its cinema. Often playback is then adjusted to even lower levels. Trailers and commercials are projected at 3.5. Kinepolis' main theatre, Metropolis 1 in Antwerp, has a continuous measurement system that monitors dBA levels and sends reports to the main office. This week (May 2013) an official report about SPL's in cinemas has been given to the Belgian government upon which a law will be based on later in the summer. Unless the industry offers an alternative soon, the maximum playback volume in Belgium will be limited to probably 5 by law.
4. What This Means For Your Mix
All this has a major influence on the mixes that are made today. The Belgian Art House hit "The Broken Circle Breakdown" was mixed in two versions. One for television with a Program Loudness of -23 LUFS, and one for cinema with a Program Loudness of -20 LUFS. The cinema version had a lot more compression to get dialogue at the intended level with the maximum allowed playback level of 5.2 for this movie in the Belgian cinemas.
At the largest post-production facility in The Netherlands, Cinemeta, playback level is now set to 5 by default. Last year a compromise setting of 6.3 was used, but this year (2013) complaints from directors that their films were too soft in the cinema was getting too strong.
I called the director of Shooting Star Productions - Dave Schram. He told me...
I always visit at least ten cinemas when my movie is out. In the past I always mastered at 7. But these days projections have a maximum at 6, more usual 5.5, sometimes even 4.7. This is mainly caused by audience complaints about loud movies. I have surrendered and now work at 5.5 myself.
I have even found cases where cinemas have considered buying broadcast dynamics processors to control loudness automatically. Let's forget about this potential nightmare as soon as possible and look at what solution we can think of.
5. Measured Levels In Movies
First, we need to have some objective figures. A first data point is the trailers and commercials. These are all normalised to 85 LeqM and then played back at 3.5, or 65 dBA. The current trend is that commercials and trailers are screened with only dimmed cinema lighting, so people can conveniently chat a bit as well as find their seats. Apparently, 65 dBA matches this well. Some cinemas use a '4' setting for trailers, which is 10 dB louder or 75 dBA. These are viewed with lights off. Mark that this is approximately the originally intended dialogue level and since trailers are very compressed, their dialogue will indeed be at this level.
Another data point would be the long term dose that's allowed for hearing protection. Usually, the figure used for that is 80 dBA for 8 hours, which is equivalent to 86 dBA for 2 hours. This means the limit for the Program Level should be approximately at reference level.
But as we now know, cinemas do not adjust their levels for hearing protection but mainly for audience comfort. A dose of 85 dBA over 2 hours is very loud for a movie and will not be accepted by the audience. A more appropriate standard level would be around 80 dBA. The Program Level of the SMPTE ref noise is just below -21 LUFS. 5 to 6 dB below that would be about -27 LUFS for Program Level.
Finally, in Europe limits do apply for maximum sound pressure levels in clubs. In Belgium, 100 dBA average level is used, in The Netherlands 105 dBA. However, a movie is not a pop concert.
In my opinion, it would make sense to limit the short-term (3 secs) exposure to 100 dBA. That would be a max S (short-term) of approximately -6 LUFS. I repeat that I am aware of the differences between the various weighting curves. Measuring electrical A-weighted levels, however, does not make sense either since that's not a true representation of the acoustical levels in the cinema. I think that A weighting should only be used with an acoustical Sound Level Meter, and the K-weighted LUFS measure should be used in the electrical domain. A maximum of -6 LUFS for max S (short-term) is a rough estimate, research is still needed to find the figure that is right most of the time.
With this in mind, let's check a few movies. Together with Michel Schöpping, a well-known film mix engineer in The Netherlands, I analysed 24 (mainly Dutch) movies and they varied wildly. We found that the program Level goes from -38 LUFS to -20 LUFS, with the maximum Short-term level varying from -29 LUFS to -8 LUFS and the maximum True-Peak level varying from -7 to +3.5 dBTP. Dialogue levels varied from -41 to -25 LUFS. The huge spread of about 20 LU with these levels gets much smaller when we take the playback level during mastering into account.
With playback level taken into consideration, we find something interesting...
- The average playback level for movies mastered at '7' is -28 LUFS (-29 to -25).
- The average playback level for movies mastered at '6.3' is -23 LUFS (-25 to -21). They are projected 3 dB softer, so if we corrected the average to a '7' level, it would be -26 LUFS.
- The average playback level for movies mastered at '5' is -20 LUFS (all were -20). They are projected 7 dB softer, so the corrected average would be -27 LUFS.
This means that the intended average playback level of all these movies, regardless the volume setting during mastering, is about -27 LUFS. This is approximately 79 dBA in the theatre. There were a few exceptions, like the movie with a playback level of -34 LUFS. This movie, however, was soft on purpose, it was a "silent movie".
Is there anything we can do to turn back to normal?
We could use a similar strategy to EBU R128. It's not by coincidence that two of the people behind that standard are sitting behind this desk here at the AES Conference in Rome, although I must stress that we are by no means representing EBU here.
In my opinion, there are four requirements to start with:
- A solution must be compatible with the current cinema hardware, which are often Dolby boxes. No new investments should be needed, only a software update in the DCP server.
- All SMPTE calibration demands should be kept intact.
- The directors must stay in control of the level, but at the same time, an objective measurement is needed.
- A comfortable overall loudness level should be guaranteed towards the cinemas, potentially damaging levels should be made impossible.
And this is my proposal of how it could be done:
- Measure the Program Level and the maximum Short-term level of the movie in the DCP server.
- If the Program Level is at or below -27 LUFS: just play it as it is. With low levels, the movie is probably intended as a "soft movie".
- If the Program Level is above -27 LUFS, check if the cinema has set the "loud movie playback" option. If not, attenuate the soundtrack during playback to -27 LUFS. If yes, check if the Program Level is above the absolute limit of -21 LUFS. In that case, attenuate to -21 LUFS.
- Check if maximum S is above -6 LUFS and also if the maximum True Peak is above -1 dBFS. If so, attenuate accordingly.
- The Dolby box is now set to 7 for a normal playback level. In small cinemas, a lower level can be selected. Such a calibration in principle only needs to be performed once.
Commercials and trailers undergo the same procedure but may get a standard offset by the cinema since they are often projected in dimmed light and conversation should then be possible.
- To get the current '4' level, this would be an offset of '1' Dolby position lower (6 instead of 7 or 5.5 instead of 6.5 etc).
- For '3.5' more attenuation is needed. I should mention that the foreground gate was designed to match wide loudness range material like movies with compressed material like commercials. So maybe no extra attenuation is needed.
This whole procedure can be easily done by a software update in the DCP server, which of course is just a computer, Windows if I am not mistaken.
The advantage of this proposal is that directors who like to use dynamics will have their artistic freedom back again. Cinemas and their audience will not encounter unpleasant loud surprises anymore. Loud playback is still possible in selected cinemas, with only a slight limitation to protect the hearing of the audience. No investment would be needed for the cinemas. The ITU loudness standard is open source and many broadcast tools can easily be adapted to indicate this cinema limits.
Obviously, more research is needed and the proposal would need to be tested in practice.
In my opinion, this topic can only be researched in co-operation with the cinemas since they will only use a recommendation if it really solves their problem. I am pleased to inform you that the Belgium Multiplex company Kinepolis has offered to participate in a pilot project testing a measure like the one I am proposing here. If this pilot is successful, they will help us to convince the Belgian government to use a solution like this in their upcoming cinema loudness laws.
To those who are opposing any type of regulation, I would like to say that you should be aware that paradoxically refraining from loudness measures to not limit creative freedom will definitely lead to a limited creative freedom, as is being proved right now.
Thank you, Elco for allowing us to reproduce your paper here. It is shocking to me that this was presented in May 2013 and here we are 4 years later without a proper solution like the one that Eelco proposed in this paper.
In Part 2...
In part 2 of this series on Loudness and Dynamics in cinema sound, we are going to come bang up to date and look at the results of a survey undertaken in May 2017 by Steven Ghouti AFSI AES. Steven is also part of a workshop with the French sound associations AFSI and ADM and he has been working in film sound for around 23 years.