Manual noise cancelling?

Recording Techniques, People Skills, Gear, Recording Spaces, Computers, and DIY

Moderators: drumsound, tomb

Andy Peters
re-cappin' neve
Posts: 602
Joined: Mon Mar 12, 2007 8:45 pm
Location: Sunny Tucson

Post by Andy Peters » Fri Jul 15, 2011 10:19 pm

MisterMark wrote:I believe the guys at Izotope came up with a hardware solution to reduce broadcast noise. I don't think it uses phase canceling, but you can read more about it here:

http://www.izotope.com/products/audio/anrb/

-Mark
The Izotope box is called an "Adaptive Real-time Noise Reducer," which implies that it's an implementation of the standard adaptive noise canceler based on the LMS algorithm, as described in the classic text, Adaptive Signal Processing by Widrow and Sterns. (And holy fuck, the book didn't cost $150 when I bought it years ago!) This book has some excellent, and eminently readable, examples of adaptive noise cancelers (in addition to other uses for adaptive signal processing).

The process doesn't use simple phase cancelation, which most people think of as a polarity reversal. In one common implementation, there are two inputs. One is the main signal, corrupted with noise, and the other is only the noise. A FIR filter is applied to the noise input, and its output is subtracted from the main input. The output of the adder is the system output, and it is also fed into the adaption process. The adapter updates the weights (FIR filter coefficients) as it modifies the noise (creates the anti-noise) so that the output is error free.

Izotope's description mentions something very important: the box analyzes and mitigates steady state noise sources. Such noise could be a 60 Hz tone from a ground loop, or low-level hum from the HVAC that your mic picked up during recording, or tape hiss, or the rumble from a turntable or ubiquitous airplane background rumble. The canceler can easily train on steady state noise and the filter itself doesn't have to be all that long (hence the real-time aspect, meaning low latency).

Where these systems run into trouble is with impulsive noise. The adaption loop cannot chase the sound of someone speaking, or a crackling noise from a bad cable, or whatever. The artifacts of the loop attempting to do so are pretty awful. All workable adaptive cancelers don't even try to chase such stuff.

So, no, you can't use an adaptive noise canceler to eliminate people who are talking in the same room, as cool as that might be!

Grok the following: the phase-locked loop in your word-clock sync is another form of an adaptive signal processor. It's got a signal input and a reference input, and attempts minimize the output error (the difference in the two input signals).

-a
"On the internet, nobody can hear you mix a band."

Post Reply

Who is online

Users browsing this forum: No registered users and 48 guests