The ANR-B handles noise better than any device our Senior Contributor has ever encountered — and it does it in real time.
At this point in time, you’d be hard pressed to find someone who works in audio but is unaware of iZotope’s powerful line of products. Their plug-ins provide powerful tools for mixing and mastering engineers as well as a wide variety of music creation and audio repair tools. I’ve been using Ozone 5 since its release and I love it.
Not only does iZotope have their own extensive line of products, they also license their technology to other developers. Their algorithms have found their way into third-party plug-ins, standalone applications, computer and console video games and iPhone apps. Adobe, Avid, and Sony are among the companies that license iZotope’s technology as are video game developers Harmonix and Ubisoft.
While they are primarily a software company, iZotope’s lone hardware device, the ANR-B, utilizes advanced algorithms designed to recognize noise and suppress it in real time. Since it’s release just two years ago, the box has been used all over the broadcast waves includingAcademy Awards, GRAMMY Awards, Tony Awards, Academy of Country Music Awards, MTV Video Music Awards, Jeopardy!, Wheel of Fortune, CBS's Survivor, FOX's American Idol, NBC's The Apprentice, Discovery Channel’s Deadliest Catch, VH1's The Hills, and dozens more. In the words of iZotope’s Bruce Bartone, “Turn on your TV and you cannot hear the ANR-B all over the place these days!”
The two-channel ANR-B can function as either a single or dual-mono 1U processor. It operates on either 110 or 220 VAC and includes AES/EBU digital I/O and analog (via male and female XLR connectors) I/O. When utilizing digital I/O, the unit supports resolutions up to 96kHz, 24-bit. When utilizing analog I/O, ANR-B processes at 48kHz, 24-bit. The unit can clock internally or via the rear panel’s dedicated DARS (Digital Audio Reference Signal) connector or the AES/EBU digital input. A standard IEC connector provides power to the unit and RS-422 and RJ45 LAN connectors provide Ethernet connectivity for firmware updating.
The heart of the ANR-B is iZotope’s Adaptive Noise Reduction (ANR) algorithm, designed to diminish the unwanted background noise in an audio signal. The device utilizes a filter bank to divide an audio signal into hundreds of frequency bands. The algorithm then identifies and separates the speech signal and the steady-state noise sources and then suppresses the background noise while isolating the voice as transparently as possible. The key to the unit's success, which is so processor dependent it requires a pair of Analog Devices SHARC DSP processors, is the precise balance between high-quality processing, fully automatic operation and low-latency performance. Historically, high-quality noise reduction systems (e.g. CEDAR) have required an experienced operator to continuously monitor and adjust the software’s parameters. The ANR-B automatically monitors the signal and makes parameter adjustments allowing the engineer to focus on mixing rather than overcoming noise.
The ANR-B operates in two modes: Adapt and Manual. Adapt is optimized for speech and, when operating in this mode, the device continually diagnoses the noise and speech characteristics and distinguishes the difference between the two. This allows the noise to be reduced without compromising the level or quality of the dialog. Adapt detects noise in real-time, adapting to changes or any fluctuations in the noise over time.
When operating in the Manual mode, the noise profile remains constant and the algorithm does not continually re-evaluate the noise. When utilizing this mode, the user utilizes the Train button to “teach” the ANR-B what noise you want to be removed from the audio signal. This mode works exceptionally well when the original audio contains musical content or other full range frequency material.
In addition to Adapt and Train, both input channels include a Bypass button that removes the ANR-B from the audio chain and Residual button that allows the noise that is being eliminated from the audio signal to be monitored; be careful, as activating this while mixing a live broadcast would be disastrous. Also included on both channels are a suppression knob that determines how much the noise is being suppressed and 10-segment input (with clip indicator), output and noise reduction meters. The ANR-B is designed so that, once set up, it only requires one knob for normal use making operation effortless, even in stressful settings.
The Preset button allows the user to save and manage presets. The ANR-B includes a wide-variety of presets that address specific situations. The Setup button allows the user to edit the I/O configuration and make other adjustments such as LAN settings and firmware management. Each menu item can be scrolled through using the data wheel next to the display.
Prior to me receiving the review unit, my associate Joe Deihl utilized the ANR-B during the audio production of the motion pictureBlue Like Jazz(see the sidebar for his impression). The stars aligned for my receiving the box as I had just started mixing writer/director jeff obafemi carr’s independent feature He Ain’t Heavy. The film was shot guerilla-style with prosumer cameras, mobile phones, security cameras and Flip cameras with no additional sound equipment. This was the perfect challenge to see if the ANR-B could truly deliver and while I was impressed with the ANR-B’s speed and ease of use, I was blown away by the results the box provided. I don’t think I could have accomplished the project’s mix without the device.
I addressed each piece of problem audio independently and evaluated whether Adapt or Manual would work best. Since the Adapt mode is tailored towards dialog, which is what I was primarily working with, it worked best the majority of the time but there were a few instances where this wasn’t the case. In each instance I inserted the ANR-B on the problem track and then recorded the processed audio to another track. I successfully utilized both the analog and digital I/O and had great results as dialog that was completely unintelligible became clear and overwhelming noise nearly disappeared completely. I found myself routinely riding the Suppression Knob as I was processing the audio and wishing that there was some way to automate the ANR-B’s operation. iZotope has such a keen sense of plug-in design that I feel it would be natural for them to release a control shell that operated like a plug-in in Pro Tools or another DAW and allowed the parameter adjustments to be automated and written into a session.
While broadcast audio and audio for picture are the primary uses for the ANR-B, it is by no means only usable in this area. I mixed a track during my review period that was built around a fingerpicked acoustic guitar. Now why the tracking engineer failed to notice that the blaring AC unit in the room that they recorded was almost as loud as the acoustic guitar is beyond me but I was astonished at the amazing job the ANR-B did in removing the noise without adding any artifacts. The box can also do a notable job of acting as an anti-reverb box, actually removing ambience from a recording.
Anyone who has mixed for picture, either for live broadcast, TV, or film, can verify that noise is the devil. It’s unpredictable, it can come from many different sources (and it often simultaneously does) and the same noise can be drastically different from one microphone to another. The ANR-B handles noise better than any device I’ve encountered — and it does it in real time.
Russ Long is a Nashville-based engineer/mixer and a Senior Contributor to PAR. russlong.ws
Contact: iZotope | izotope.com
Sidebar: ANR-B In The Real World
Supervising Sound Editor Joe Deihl, whose credits include Blue Like Jazz, Witness: Voices from the Holocaust PBS Special, and People: A Musical Celebration discusses utilizing the ANR-B to overcome a previously unfixable problem:
“There was a helicopter sound in the background of a scene that was uneven across the dialog track due to the edited nature of the track. I made several duplicate ‘working copies.’ Knowing that dialog responds better in the Adapt mode, I got good results with reducing some unwanted noise; however, that pesky helicopter was still present.
“The key for me was taking a sample of the unwanted sound (the helicopter), copying it and duplicating it in a streamed loop so that I could ‘teach’ the ANR-B exactly what I was after. After establishing a workspace in my DAW, I then took another copy of the same scene and used the ‘manual’ mode. My target here was not so much dialog; it was getting rid of the helicopter when the voice was not fully present. So I ended up with two layers, where the voice was strong in the first layer, I kept it, and where there was no voice, I utilized the second layer where the sound of the helicopter was greatly reduced. Layer 1 targeted the dialog with the ANR-B’s Adapt mode and Layer 2 targeted the helicopter utilizing manual mode.
“The next target was where the dialog and helicopter coexist, so I reused a sample from an audio clip before Layer 2 got its update. Looping this sample and giving the ANR-B the ‘learning’ with the helicopter loop placed before the dialog with the unwanted helicopter (because the ANR-B takes about 2 seconds to put this together), I ended up with a helicopter loop edited right before where the dialog and helicopter are blended. Once the ANR-B ‘saw’ the helicopter in adapt mode (after having been taught the helicopter sound by the time it was presented with dialog) it delivered a much cleaner dialog recording. Once I got that, I just put that part of the reclaimed dialog back into where it belonged within the track and it sounded far better.
“Bottom line: results matter and the ANR-B delivered. It's a tool that can be applied to garden variety noise reduction needs, as well as more challenging audio problems that have to be approached cleverly with some out-of-the-box thinking.”