Cutting out the noise

In the studio, noise contamination can be controlled through increasingly elaborate arrangements of baffles, absorbers and diffusers. Onstage, however, the options for limiting spill are few and far between, with the most common approach today being to remove as much sound from the stage as possible, replacing traditional monitors with in-ear systems.

Even then, background noise entering a vocal microphone remains one of live sound’s most persistent problems. In houses of worship, spill from drums, backline and loudspeakers can muddy vocals, reduce intelligibility and force engineers to balance performers’ freedom of movement against the risk of feedback. Although many tools are designed to tackle feedback, they do little to isolate the voice from unwanted sound already reaching the microphone.

“I’ve been involved in touring and houses of worship for many years, and one of the problems that’s never been solved is removing background noise from vocals,” explains Ryan John, director of product management, software, at L-Acoustics. “Other products often focus on feedback removal because feedback is a very audible problem that everyone in the room can hear. What audiences rarely put their finger on – but nonetheless feel – is just how much other noise is coming through that vocal microphone, and the compromises engineers have to make to keep it in check.”

L-Acoustics developed Source Intelligence to address that gap. Available as an exclusive licensed application for the L-ISA Processor II and integrated into the L-Acoustics workflow, it uses machine learning to continuously identify the voice within a microphone signal and remove everything else. The result, according to the manufacturer, is a cleaner, more intelligible vocal – even on noisy stages or when a performer moves in front of the PA.

Source Intelligence began life in the R&D department as an extension of the L-Acoustics DJ project. At which point, John recognised the possibility of repurposing it into a solution for engineers to remove background noise. “Feedback can take a while to build up which gives the engineer more time to process it, but you don’t have that time if there’s a snare drum behind a vocalist which comes straight out of the PA,” he says. “Source Intelligence has been designed to deal with immediate problems rather than ones that can ramp up.”

Source Intelligence can be used to strip away PA bleed, crowd noise, stage instruments and room reverb, leaving a cleaner, more natural-sounding vocal without the restrictions that have long been a fact of life on live productions. The result is described by the manufacturer as a vocal that preserves the natural tone, character and intelligibility of the source.

John explains that while there are plenty of other products available on the market, they are mostly threshold-based. “If the signal is above a certain threshold, it allows the signal to pass through and when it’s below that threshold, it reduces,” he continues. “This tends to work if the background noise isn’t too bad. But the challenge is that when the signal passes that threshold, everything comes through and a snare drum that’s behind the vocalist can sound louder than the voice. Source Intelligence is built to recognise what is a voice and what is not a voice.”

Once installed, Source Intelligence requires no threshold setting, no ongoing adjustment and no changes to the existing console workflow or signal routing. End-to-end latency North Point is the world’s first house of worship to deploy L-Acoustics Source Intelligence is less than 8.5ms over MADI, fixed whether processing is active or bypassed, so that automation and snapshot changes never affect system timing. Full PSU, network audio and DSP redundancy are standard through the L-ISA Processor II core.

“It’s entirely self-serviceable,” explains John. “The customer opens up their L-ISA controller software, types in the new licence they’ve been given and it’s ready. It’s the exact same routing that you would have otherwise used for your L-ISA processor, but now you’re enabled on the software side to use a completely different process. If you’re already familiar with L-ISA, you’ll understand every bit of the interface but, even if you’re not familiar with the L-ISA workflow, it’s incredibly intuitive.”

Source Intelligence is the first product L-Acoustics has released that turns the L-ISA Processor II into a platform rather than just an L-ISA device. “One of the key tenets for us is reliability – you need to be able to trust that the product is going to work all of the time,” says John. “Historically, you’d just use the L-ISA Processor to run L-ISA, or you could run L-Acoustics DJ or Ambiance on it in conjunction with L-ISA. But when you switch on Source Intelligence mode, it only runs Source Intelligence. It’s about making sure that there’s enough memory and headroom available because reliability is so important.”

The technology has been designed for any application where a voice is being amplified. It works with handheld, headset, lapel and distant-pickup microphones, and processes up to four channels simultaneously. Each instance can handle one vocal at a time and can even be placed in the path of a group of channels so that a single instance can serve all handheld microphones in any workflow. “By cleaning up the sound, the intelligibility goes up and the word is heard clearly, whether that’s in a loud worship environment or an incredibly cavernous liturgical church,” explains Joshua Maichele, applications project director, house of worship at L-Acoustics.

From a broadcast and multisite perspective, one of the main challenges that engineers face is taking the main pastor’s voice from the broadcast location and ensuring it’s reproduced clearly at all of the campuses. Often, the pastor is being captured in a much larger environment than the size of the remote locations. Because Source Intelligence can handle four discrete processing channels, the engineer can process what’s happening in the main room separately to what’s being sent out to the other locations. “For the broadcast room, we want a very high level of refinement, but maybe at the remote locations, we might want that feed to have a little more of the ambient energy in it,” describes Maichele. “This is what we found at North Point Community Church in Atlanta – they liked the sound in the main room in terms of how much cleaning they were doing, but they actually found that it was a little too isolated going to the broadcast locations. In this case, the engineer can decide how much of the high-frequency and low-frequency energy they want to remove for each location.”

Besides removing background noise, Maichele also believes the tool enables freedom of movement, allowing the pastor to move behind the PA, around it or out into the congregation. “At North Point, we’re able to get Pastor Andy Stanley above the band because we’re not fighting and triaging everything that’s happening behind him. The band members are also free to move around the stage without the engineer struggling to find the point where the vocalists are stable, clear and intelligible. Source Intelligence has been designed to clean that up so that the mix isn’t changing based on what’s happening with the microphone.”

Because one processor can handle four channels, the FOH engineers at North Point use Source Intelligence on the main pastor, the host and the vocal leads. “It means that the tool is really flexible in that it’s not just locked to one person,” continues Maichele. “We can use the channels however we need to. It senses the human voice and processes out everything that isn’t the human voice. Four channels suit the needs of most of our customers – they are already using tools like Song-by-Song Snapshots so Source Intelligence fits right into their workflow, and nothing stops a customer using multiple L-ISA Processors to gain more than four channels.”

With HOW volunteers changing from week to week, the engineer’s traditional workflow could end up being a little heavy-handed in order to maintain consistency, for example using more compression because one person is very quiet while another person is a little louder. “But the more heavy-handed you get, the more you bring up the noise floor,” explains John. “With Source Intelligence, because you can remove that noise floor up to 40dB, you can actually be far more creative in what you do. This means when I go from a quiet singer to a loud singer, because I can process more heavily, I’m able to get a much more consistent output.

“Up until now, we’ve mostly been focused on the system and space side of audio delivery – whatever comes into our system, we want to reproduce it perfectly and make sure every single audience member hears it,” concludes John. “Now, we’re trying to improve the quality of the source before it hits our system, so it’s a different approach for us. There’s no doubt this will help our customers who are trying to deliver a message and deliver an emotional impact in their storytelling.”