You may recall that last month I wrote about audio guru Neil Muncy's observations regarding television audio level problems he encountered in Canada. Like all of us, Neil is bothered by levels that vary significantly from channel to channel.
In particular, Neil noticed that things got worse around the beginning of October 2005. I also mentioned Neil has tried to measure these level problems and is having trouble reconciling his findings. I ended by saying that there is more to the story.
WHAT DOLBY SAYS
Jeff Riedmiller is the audio levels evangelist from Dolby. He has probably worked harder on the implementation of dialnorm and related metadata than just about anyone else.
Jeff wrote, referring to the Canadian Cable Telecommunications Associa-tion, "Since the implementation of correct dialnorm settings by the CCTA was carried out on a service by service basis, where some were correctly set and others were not. As Neil states, before the change, everyone was at -31 for dialnorm, which only allowed us to fine tune the level in one direction (down) for many of them. Now we have room to fine tune the levels in both directions for many of the services simply and easily (i.e. with no changes to production practices).
"Also during this effort, I brought up the fact that by setting dialnorm more appropriately 'on average' may expose certain program-to-program level differences that were there to begin with, on a single channel/service due to the fact the dynamic range control subsystem of AC-3 is not being constantly driven into compression and/or limiting because the dialnorm value was set at -31 (this behavior depends on the actual speech level and dialnorm setting). Hence, the DRC subsystem was (unbeknownst to many) offering a type of 'brute force' normalization that reduced any significant level differences from program to program (for a single channel/service) with the penalty of a reduced dynamic range."
Here we see a negative and unintended side effect of the metadata system. Dialnorm is not only an indication of the actual average dialog level of the program, but also a threshold setting for DRC (dynamic range control).
If dialnorm is set at -31 dBFS when the actual dialogue level is -21 dBFS, then most of the program level could be heavily compressed by the DRC subsystem, depending on how it's set. Happily, when dialnorm correctly indicates the actual level, the DRC subsystem functions correctly.
Jeff continued, "This does not address the problem of average channel-to-channel level differences among all the channels combined on a system. For some channels and services, -31 dBFS may be essentially correct over the long-term while in other cases it may be very wrong.
"If only some of the services set dialnorm correctly (meaning that their actual long-term dialogue level and transmitted dialnorm value match each other), their decoded dialogue level will unfortunately be quieter than an adjacent channel/service where their actual dialogue level and transmitted dialnorm value differ (sometimes greatly) from each other.
Programmers who are setting dialnorm correctly can, and often do, receive complaints that their programming is too quiet when compared to other services. The first step to resolve this problem is to get everyone on the same playing field over the long-term."
WHAT CCTA IS DOING
Meanwhile, CCTA is picking away at the problem. One initiative has been to log and publish the actual long-term dialogue levels of all the Canadian service providers at: http://www.ccta.com/CMFiles/Responses_from_programmers_09210564ORQ-9232005-1248.pdf
What we have here are a compendium of long-term dialogue LeqA averages for the various service providers. In theory, each provider could simply enter the indicated level as their dialnorm level, and all would be fine over the long-term.
Meanwhile, CCTA is also moving ahead to help members get dialnorm and DRC correctly implemented.
Michele Beck, vice president of technology at CCTA, has written, in several memos to the membership, "While we recognize that the introduction of digital audio metadata will require companies to invest time and resources, it will greatly improve the overall quality of the audio as it will reduce the variations in loudness levels between services. Audio loudness problems are the top complaint of television viewers. This project should greatly reduce the number of complaints received related to audio loudness levels..."
The project will work only if everyone cooperates. "As you will recall from previous memoranda, CCTA has been coordinating a project that aims to improve the quality of audio... Phase 1 occurred on Aug. 29 and involved the introduction of dialnorm on the digital services carried by BDUs. To date, the reaction has been positive... Phase 2 [commencing Oct. 5] involves the introduction of dialnorm on all the services that are typically distributed on the analog tiers on cable... This will ensure that the loudness levels of the programming services transmitted to all cable headends will be consistent at the output of the satellite integrated receiver/decoder (IRD)... cable companies should be prepared to make their adjustments beginning Thursday, Oct. 6.
"It must be noted that audio levels on cable companies' analog tiers will be significantly misaligned after the Oct. 5 introduction of dialnorm until cable companies verify the settings of the IRDs and adjust the BTSC modulators. Cable companies must be prepared to schedule site visits as soon as possible after Oct. 5. Once the adjustments have been made, cable companies should see improvements in the overall quality and consistency of audio levels.
"Off-air services may require adjustments to bring these in-line with the remaining services... Cable companies should use the digital services as a reference when adjusting audio modulators for the analog services. This will ensure that the majority of all analog and digital services are equalized. Cable companies should avoid using off-air services as a reference as these do not have dialnorm applied to them."
So here's what appears to have happened. The implementation of dialnorm in the digital tier went comparatively smoothly in late August, but when dialnorm was put into effect on the analog tier in October, there was a mad scramble to (A) determine what their analog tier dialog LeqA levels were and (B) to correctly set dialnorm for their digital simulcast services in real time, on-air. It wasn't a pretty sound for awhile there. Neil Muncy noticed.
The problem is not unique to Canada, of course. Next month, I'll take a much harder look at that part of it, and share with you some more of Jeff Riedmiller's remarkable thinking about how this should go. In the meantime, thanks for listening, even without dialnorm and with way too much compression!
