Reducing noise

From Audacity Wiki
Revision as of 10:03, 23 June 2007 by Suf (talk | contribs) (Reverted edit of XjnGfi, changed back to last version by Gregors)
Jump to: navigation, search

Note: There is a page with similar content at Reducing Noise. Probably it should be merged with this page.

Prevent noise

The best way to reduce noise in your recordings is to omit it entering your sound material at all. For that task take care of the following:

  • Set the correct input level of your sound sources. Set it as high as possible to increase the dynamic distance of sound and noise, but as low as neccessary to prevent clipping.

Use balanced audio connections. More expensive PC sound cards allow balanced inputs, which have better noise rejection.

If you have it at hand use a hardware limiter as long as Audacity cannot process sound in real-time. Personally, when recording a band I use a simple foot pedal limiter on the vocals, because these seem to be the most dynamic sound source. The input level of instruments can be adjusted quite easily.

When you have found the optimal levels, decrease them about a dB or two when recording music just to be sure. When actually performing, musicians tend to be a little bit louder than in rehearsal mode.

  • Shut every non-used sound channel and sound source. Mute non-used channels on your mixing board, switch off non-used amps, keyboards, and don't forget shut the door and the window.

Especially in an home-recording environment, avoid switching lights or electric machines on or off during recording, because a spark can cause a knack on the track.

Avoid fluorescent lighting and keep cell-phones a good distance away from any equipment.

Noise tends to stay on the same low level and cannot be controlled in general (you don't have a fader for it, do you?), but the record source is often dynamic and may change and can be controlled. So you may get aware of your potentials in noise control here. Initially non-audible noise often comes to attention when a low signal is amplified and/or normalized because normalizing and amplifing increases both the wanted signal and the unwanted noise. Therefore, the measures and proceedings I described prevent some of the typical noise problems later.

Remove noise

On virtually any recording you can find noise. It is not always neccessary to get rid of it completely. First-of-all, it is often audible only in very low-volume passages. Second, the average not-too-picky listener will accommodate to the noise level of your recording. In this regard it is comparable to the odor of a room: When you enter it you become aware of it, but once you stay in there for a period of time you'll probably cannot smell it anymore. Third, he/she may listen to your programme in the car or while washing dishes so he/she may be not in the position at all to hear the noise.

Sometimes, a completely silent passage, e.g. between sequential parts of a programme, can irritate a listener more than a constant low-level noise throughout the mix. This is so because complete silence may disrupt the ambience of the material. There are situations where you actually want to add noise (e.g. in film production between a cuts of the same scene).

So, you may want to change your attitude towards noise here: It's not just dirt that needs to be removed, but it's a natural part of all listening experiences that has to be dealt with appropriately. In general, we need to accomplish two things: The noise just has to stay on a roughly similar level throughout the material and it should not be too obvious to let it be ignored.

Dealing with Hi-Band Noise

Still you may see the need to decrease the noise level in your recording. When working with multiple tracks use the same care as when recording:

  • Before doing anything else, increase the tracks to a working level, e.g. with the Audacity Normalize function. Leave a little bit of headroom when amplifying.
  • Mute all passages where there is nothing to be used in the mix. You may use the envelope tool to make silent passages of a track really silent.
  • Fade-in's and fade-out's are much better than sharp edges cut-and-paste that very likely will cause knacks.

On a single noisy track, you may want to use the Noise Removal feature of Audacity. You accomplish this task in two steps:

1. You pick a "noise-only" part of the track's signal. This part should not be longer than approx. half a second. This is a sample of the noise that will be used to compute the neccessary changes to the track to remove just the noise (though this idea is always merely theoretical).

You have to be careful in selecting your sample. If you pick a sample that contains not only noise but also a slight part of - let's say - a reverb tail, you'll remove that, too. To give you another example, the sound of breathing can be quite similar to noise but it provides a lot of the vitality of a vocal track.

Now, select a small portion of noise, call "Noise Removal" and Select "Get Noise Profile". If you are unhappy with your selection, you can repeat this step. Every time the last sample will be overwritten with the new one.

2. Now, select the portion of the track that needs noise reduction (in most cases this will be the complete track). Call the "Noise Removal" function a 2nd time. You can click on "Preview" to listen to the first seconds of your selection or on "Remove Noise" to execute the noise filtering. The less/more slider is quite obvious, I think, and can be changed for testing and applying the effect.

If you're unhappy with the result you can Undo all changes on your track and try again.

Noise Removal often helps to reduce hi-band noise such as hissing and (to a certain extent) crackling. Your signal may still have noise that you might have not realized: Low-band or sub-sonic noise.

Dealing with Sub-sonic Noise

Sub-sonic noise can enter your recorded material on various instances: During recording as vibrations, from the tape machine (in case you still use one of those), even as artifacts when processing the material through analog and digital effects. Everything below 20 Hz is called sub-sonic because the human ear is unable to perceive that kind of sound as a sound.

You can recognize sub-sonic noise by eye when the wave depicted in Audacity is not symmetrical along the time axis. As you have already applied normalization (if you have followed my advice so far) the DC components should already have been gone. Asymmetries you still find are sub-sonic noise that you cannot hear by definition but take up space in the dynamic range of your material.

To remove it from your track you may filter the track with an equalizer. Some unknown author (at least to me) has written a Nyquist plug-in for Audacity to solve that task quite perfectly. You can download it from the German Audacity site here:

If you are confused by the German text, the actual file is here:

After removing sub-sonic noise you can generally re-normalize your track and it will appear louder yet much more defined on the bass level.


The best and most important option in noise control is to avoid it.

Removing hi-band noise appears to be an option you can choose to take or to leave out. If you apply it, do it on the most basic level you can reach, on unprocessed tracks (e.g. before adding reverb), on single tracks or even on single passages.

Removing sub-sonic noise, on the other hand, seems to be mandatory, at least in my experience. In contrast to hi-band noise you can apply it as a step in the mastering process, on the mixed and processed material, before normalizing.

--Kurt 16:14, 4 February 2006 (PST)