This was actually inspired by episode 222 of the Podcast Answer Man, when he mentioned an issue with the use of an incorrectly-set compressor-gate. I was pretty impressed when he demonstrated what his environment sounded like when bypassing the compressor gate.
This got me to thinking; all this talk about noise removal, and I haven’t really given the compressor/expander/gate plugins a fair shake. So, I started playing around with them, until I found that the best noise removal by far were caused by the “RMS Envelope Tracking” “Simple” plugins in the CMT collection. I’ve also made use of Steve Harris’s “Simple Amplifier.”
This isn’t audio scrubbing like Audacity does… this is realtime; I can use these plugins on a realtime project, such as broadcasting or using Skype with Jack (as is my wont). Don’t take my word for it… here’s a sample, using both the raw recording (I did emphasize that raw recordings should be made before any processing, right?), and the processed recording using the chain below.
Those of you using normal speakers won’t notice much; maybe the second will be a little quieter. Those wearing earbuds or headphones may notice some difference… if you have studio headphones, you will likely notice everything immediately… for those who don’t know, studio headphones are equalized to show every ugly truth in a playback… if there’s noise, you hear it in all its glory.
Keep in mind that while this can clean out noise, it is not going to do all the work for you; you should be recording in a quiet location to begin with. In this way, there is a degree of separation between the signal (your voice) and the noise floor (background noise)… if they are too close together, than you will either encounter points where your voice is cut off, or where the noise will break through into the recording. Neither is a good event. Also, this does not remove the background noise during those parts where there is a signal, such as while you’re talking. If you have a lot of background noise, then you will need to find ways of preventing it from getting to the microphone.
The first plugin in the chain should be the expander. Expanders will stretch out the dynamic range of a sample under a certain point. This means that all audio above a certain point will retain its’ amplitudes, but in parts quieter than that point, they will get even quieter. This means that the “noise floor,” which is the background noise under the actual audio signal, will become even quieter, while the signal itself will not be changed.
This plugin tracks the “Root Mean Square,” or RMS, which is just a fancy way of saying that it bases its threshold off of a calculation between the average signal strength and the value in the above “Threshold” field. If the above threshold number drops, so does the threshold. You want to set this value as low as you can without a drastic increase of noise; you may need to play with it some to find the right level. You do want to increase it enough, however, that you don’t hear “noise pumping.”
Once you have the correct threshold to clean out the noise, but allow signal through, you can adjust the expansion ratio; the higher this number, the more complete the silence becomes. If you max out the expansion ratio, then this essentially becomes a gate. I generally prefer putting it low enough that the background noise is gone, while the quieter parts of my voice will still be picked up.
The next step is to compress the signal a bit. This is not to say that you should squash it into putty, but by compressing the dynamic range of the actual signal, you can ensure that your voice will be heard clearly without the expander stretching it out at the low end.
The threshold and compression ratio are just like the expander; the lower the threshold, the more sound gets compressed, and the more compression is applied as the ratio increases. This also tracks the RMS of the signal, so it can adjust accordingly.
The compression can help prevent peaking, but it does not stop peaking, just slows it down. In order to close out the set, you should to put up a limiter that can block a signal from passing the peaking point, without causing its signature distortion.
One of the main reasons people prefer to use Audacity is because it has this “magical” noise removal filter as one of its editing tools. With the help of this chain, you can hopefully see the possibilities of using good noise removal procedure in a real-time context, without any need for editing the final product for cleanup.
With this new tool under your belt, have fun, and make something good!