Mix a pop song - 7 - Adding the Main Parts - Example 1 - Noise Reduction Using EQ

The act of eliminating unwanted noise from a signal, is obviously one application where it is far better to get the sound right at source. Although this article is about mixing, it is worth mentioning here what the problems are, and how they are normally solved at the recording stage.

The elimination of rumbles is best achieved by careful microphone placement and mounting, and the elimination of hum is best achieved by both the careful choice and placement of the cables used - normally "balanced" cables. Elimination of high-frequency radio interference is also best achieved by the careful choice and placement of balanced cables, but because high-frequency interference is so pervasive, this might not always help. Taxicabs for example are often a source of RF interference on microphone cables - especially if those cables are unbalanced.

Hiss originating in microphones can be eliminated by the use of high-output microphones coupled with high-performance balanced microphone pre-amps which in the nineties, are now present even in fairly "budget" mixing consoles.

Avoidance of hiss from a recording device (i.e. tape machine) is best achieved by the use of a good noise-reduction system such as Dolby, or by recording on a good digital system in the first place - unfortunately, the digital to analogue convertors on budget recording systems (like some PC sound cards) don't even come close to the high-performance that high-performance digital or even top-end analog systems can theoretically achieve.

Also, in practice, avoiding all of these sources of unwanted interference is not always possible. Even the best isolated city-centre recording studio can suffer from very low-frequency rumbles caused by nearby traffic and trains, and the busy lifestyle in a city provides ample sources of radio interference such as power lines, mobile phones, and taxicabs.

Despite the noble goals of the "purists" of sound recording theory, it isn't always a practical idea to record all microphones "flat" (without EQ). Although you might not be able to "hear" low-frequency rumble from nearby traffic and other extraneous sources, you can certainly "feel" it - especially if your recording involves a lot of microphones open at once, where "spill" from nearby sound sources (like a bass guitar amplifier) may be common.

For this reason then, it is almost always a good idea to switch in the high-pass filter that most modern mixing desks provide when recording from microphone. The exception to this, is obviously when you are using a microphone to record a very low-frequency sound such as a bass guitar, a cello, or other low-frequency instrument, where keeping ultra-low frequencies is paramount.

Similarly - at the top end - any musical instrument which involves an electrical "pickup" device - such as an electric guitar or electric piano - may be susceptible to high-frequency radio interference.

When recording electric pianos in particular, I always filter off the very top-end until I can hear it muffling the sound, and then I open it out again until I hit the sweet spot where the basic sound remains unaffected by the filtering. This is based on many sad years experience of recording an otherwise good "take" that is suddenly spoilt by unexpected subtle interference from a nearby source. It would be easy of course to "fix it in the mix" later, but taking such an approach means that you have to suffer unsatisfactory monitor mixes until main mixdown time. Best to get it right first time, by careful EQ.

Most high-pass and low-pass filters on modern mixing desks have quite a precise response, and so an acceptable - if overcautious - general approach can be to align the filters so that they "crop off" ALL sound which is theoretically outside the frequency range of the instrument being recorded - in much the same way as a professional photographer "crops" the edges of a photograph, leaving only the areas of interest as the centre of attention.

Removing hiss from a multitrack recording should be a thing of the past. Unfortunately, not all recording engineers are as bold as they could be when setting recording levels, although in all fairness, on a live recording session, this is often simply due to lack of time and a sensible regard for flexibility and the need for a "safety margin". Also, it is not at all unusual these days to be asked to do a "trendy" remix of an old, hissy, multitrack master from many years ago.

The basic EQ trick under such circumstances is to remove as much noise as possible from the "obvious" candidates such as bass drum, bass guitar, electric piano, and poor-quality electric guitar pickups.

Although these instruments have a good deal of high-frequency energy that you would not want to lose, in practice, most of this "energy" is well below 8 kHz - a frequency above which noise becomes particularly offensive. Under these circumstances then, it is perfectly acceptable to roll off as much top-end as you can, using a steep low-pass filter. It's not "hi-fi", but it is a practical approach that engineers have used for many years. It's part of the job of getting good sound.

For the other instruments, you would be best advised to leave them well alone. In most cases, EQ on the top end of the particular types of instrument outlined above, combined with some judicious noise-gating on the remaining tracks, should be more than enough.

If this is not the case, and you still feel that there is too much hiss overall, then there are three "last-resort" approaches that you can take on the remaining, troublesome tracks.

Firstly, you can try "single-ended" noise reduction, such as putting a Dolby or DBX noise reduction unit, switched to "decode", across the offending tracks. This is a popular trick amongst broadcasters and can be heard on many news reports. The fact that it can be heard at all, should be enough to put you off trying. The result usually sounds like the news reporter is fighting to avoid being suffocated by a pillow. Even boosting the top-end EQ doesn't soften this impression. It's pretty horrible.

A much more successful technique, is to EQ all of the hiss out of the sound until it sounds positively muffled, and to then feed the resulting, dead, sound through an Aphex Aural Exciter, or similar enhancement unit, which will synthesise a new, clean top end out of the remaining lower-frequency musical information. This technique works extremely well, and has even been successfully used on the CD remastering of ancient classical works (although the engineers involved would probably not admit to it publicly).

Lastly, if a particular track is so badly plagued with noise that it is beyond the capabilities of the basic tricks outlined above, then all is not lost. Although time-consuming, the offending track can be fed into a computer and be subjected to noise reduction software. The noise-reduction results possible now can be really quite breathtaking - although it must be pointed out that if you overexert such systems, you can end up with the main sound source sounding particularly artificial and computer-like (sometimes with interesting results!). I frequently DO use computer based noise reduction - not on music sources - but on the location news reports and sound clips from dubious sources that need cleaning up for radio broadcast.