Implementing Loudness Processing

How do file-based and real-time loudness processors fit into an overall facility audio system? Each has its place. If a facility is primarily file-based, then file-based processors can be integrated into the overall front-end workflow to conform content to house loudness and true peak targets. Real-time processors, on the other hand, are most appropriately used at the end of a signal chain just before transmission to catch and correct any errant loudness errors.

For ingest, file-based loudness processing is generally set up as part of an overall video and audio workflow that can include transcoding, quality control and more audio-specific processes such as channel verification and assignment, Dolby E decoding and encoding and metadata verification, besides loudness and true peak processing.

Workflow is generally set up via software interfaces instead of knobs and buttons on a front panel. The processors are integrated into the file-based environment via IP infrastructure.

Linear Acoustic AERO.2000 audio/loudness manager Software-only loudness meters are available as plug-ins for digital audio nonlinear editing systems for the craft editor to use during the edit session. However, there is a school of thought that advocates using a file-based loudness processor on a piece after it has been edited.

“Glancing every 17 seconds [at a loudness meter] is antithetical to getting work done and stultifies the creative juices,” said Oliver Masciarotte, director of customer experience at Minnetonka Audio Software Inc., maker of AudioTools. “Let audio mixers mix and stop worrying about loudness. Mix and make it sound good. Then give it to something to take care of loudness and not screw up the mix. Do this after the mix is locked.”

While post-edit loudness processing could be done in the edit suite, it does tie up that resource, which is why taking it out of the NLE environment will probably make more sense for most facilities. The edited piece can be copied to a “finished jobs” folder on the network, which would put it in a queue for automated “back-office” processing.

Simon Pegg, senior software pipeline architect for RadiantGrid, said he’s heard concerns from audio mixers about the introduction of loudness metering into the edit suites, fearing that it would take away from their creative process. Pegg’s response was that loudness processing actually gives audio mixers the freedom to do their job properly.

Jünger Audio’s T*AP TV audio processor

“We’re not taking away your skills,” Pegg said, referring to audio mixers. “Rather we’re liberating you to use your skills. Let the loudness process be automated. You use your ears to create, and we do the technical stuff, so you play well with everyone else.”

Compared to file-based processing, real-time loudness processing is best used at the end of the audio chain before final transmission. In this application, a real-time processor acts as a final gatekeeper to make sure no potential loudness violation sneaks by.

“Typically you want a real-time processor to do nothing for most of the time,” Pegg said. “But if something goes disastrously wrong, clearly going to a violation, then you want it to take effect.”

If every piece of content is checked and made to conform to the correct loudness level up front, then real-time processing at the end of the signal chain can be light and audio quality will be minimally affected.

“Scale everything first [with a file-based processor], then the real-time processor can be set to do less,” said Tim Carroll, Telos Alliance CTO and Linear Acoustic founder. “Otherwise the real-time processor has to be set for the worst offending audio and then apply it to all programs.”

Linear Acoustic has the AERO series of processors for television, as well as software versions for scaling and optional processing and upmixing for file-based content. Also, licensed versions of its processing can be found in products from Wohler/RadiantGrid, Axon, Cobalt Digital, Miranda, Ross, Snell and others, Carroll said.

One of the limitations of a real-time processor is adapting the processing to different types of audio content. As Pegg said, “loudness correction has to be applied sensibly, or it can squash the signal entirely. You can get a good loudness number, but the sound [quality] will be poor.”

While real-time loudness processors typically have less flexibility than file-based processors, they don’t necessarily have to be stuck with a worst-case setting. These types of processors come with factory-defined presets that cover many scenarios, and usually allow custom presets to be saved. Much of this can be done with front-panel controls. Just remember that, unless in bypass, a real-time processor is always doing something.

It’s probably best not to set and forget with a real-time processor, or take the setting that comes out of the box, but to consider the types of programs that may have to be dealt with. A live musical event will probably need a wider dynamic range than a prerecorded situation comedy, for example.

Automation can be a solution. A playlist can be set to trigger different presets for different types of content. Peter Pörs, managing director, Jünger Audio GmbH, suggested sending a trigger to a real-time processor when content is switched to interstitials and then back again. “This clears the measurement memory from the previous program and starts the measurement all over again,” he said.

Jünger Audio calls its loudness solution Level Magic II, and can be found in such products as T*AP Edition eight-channel TV audio processor and D*AP4 LM Edition four-channel digital audio processor.

A real-time processor can also be put in bypass if a piece of content is known to have the correct loudness, but take care. Carroll said that even if a program is scaled correctly with a file-based processor, it could still have a wider dynamic range than the viewer’s system can comfortably handle. Since this could then engender viewers’ complaints, some real-time processing may actually be needed.

Another caution regarding automation is that if it somehow gets out of sync or was programmed incorrectly from the start, the wrong preset could end up being triggered— processing when not needed and vice versa.

If a real-time processor must be set with only one configuration, then try picking a setting somewhere in the middle between hard limiting and really wide dynamic range.

“If it’s too far in one direction, then Hollywood will complain; but if you go the other way, the consumer will complain,” Carroll said.

While a typical form factor for a real-time processor is a 1RU chassis, there are products available in a modular format, especially for facilities that generate multiple channels. Examples are Cobalt Digital’s Compass and Fusion range of modular products that have an option for the inclusion of Linear Acoustic AEROMAX 5.1 and/ or 2.0 channel loudness processing for at least 17 Cobalt openGear cards. Jünger Audio’s C8000 modular line includes the C8086+ eight-channel Level Magic II processor card for 5.1 or 2.0 mixed mode or four 2.0 channels.

Any processing will leave its effect on the audio signal, real time generally more so than file-based. So be aware if any of these undesirable audio effects are heard—distortion, pumping, reduced dynamic range, collapsed soundstage, image shifts from front to back and left to right, spectral skewing and reduced intelligibility.

With a collapsed sound field, “you lose the sense of depth and panning,” Masciarotte said. “The center channel may be lost or get buried. The sound may shift from front to back for no apparent reason.”

Remember, intelligibility can be compromised any time that gain or power spectrum is changed.

Used wisely, loudness processors can keep a facility compliant with the latest loudness rules and regulations and still produce good quality audio.

Mary C. Gruszka is a systems design engineer, project manager, consultant and writer based in the New York metro area. She can be reached via TV Technology.