Okay, that's helpful information.
The reason I asked is because where the extraneous noise is in the audio spectrum, what its competing against, and the envelope characteristics of each, are critical to selecting something that keeps the signal you want, and sends the noise somewhere else.
In the case of voice, something like the SSM-2166-based Quick and Dirty Compressor, posted at Jack Orman's www.muzique.com
site, but also in the application notes/datasheet for the chip (look up SSM2166 and pdf in Google) is a very good candidate. Indeed, the chip was largely developed for mic'd situations. It has both a limiter, gain stage, and downward expander in the same chip and requires very little to build....assuming you can score a chip. Small Bear might have them and maybe even somewhere in the GTA (I got mine from Active/Future Electronics a few years ago). Expect to pay somewhere in the neighbourhood of $10 when you find them, but the amount of time and headache this chip saves is WELL worth the expense.
Having a limiter and downward expander works well for voice in a few ways. First, the limiter and gain gives the voice a little more cutting power. The downward expander also preserves a lot more of the envelope of the voice so that beginnings and endings of words don't get chopped.
How are downward expanders different than gates? Gates cut the input signal by a fixed proportion (normally 100%, but some let you set amount of cut so that it could be only 50% for example) if it is below some minimum level. Once above that level, the "gate opens", meaning that the level cut is not applied any more.
The problem with this is that in instances where a note is fading (e.g., a singer holds a note with some vocal vibrato and fades at the end of a line), the gate does nothing up to some arbitrary threshold/level and then barges in as if it has suddenly yanked the plug on the singer. You can set that threshold such that the mic signal has to be really low for that to happen, but at the same time you risk inviting all that noise you were trying to cut out in the first place.
Downward expanders, in contrast, effectively adjust how much cut is occurring on a proportional basis. So if you're just a bit below that threshold, a bit of level reduction is applied, and if you're below that proportionately more level reduction is applied. They are called "expanders" because they essentially exaggerate the differences between soft, pretty soft, and very soft levels. The end result of this is that the tails of notes are better-preserved - important to singers.
The SSM2166 does not set how MUCH expansion occurs below the threshold, but it does allow you to set the threshold for both limiting and expansion so you can construct a zone where the signal leaves the way it came in, a zone where large incoming differences result in little outgoing difference (limiting), and a zone where small incoming differences result in large outgoing differences (expansion).
VERY powerful combo and the basis of all high-priced rackmount "dynamics processors".
There is an older FET-based gate project from Modern Recording
magazine posted at my site (http://hammer.ampage.org
). I've built it and it isn't bad, but it needs fine-tuning. Build it flawlessly, and even if it works exactly as intended, I guarantee you will be back here asking questions about what to change, That's not a weakness of the project per se, but an issue of how important it is to match gate properties to signal source.Noise source
: The actual level of the noise source plays a role in what type of noise cure you go after. For instance, if the noise in question is simple hiss, a gate or expander might fix it beuatifully, but recognize that it will have no impact on hiss in the PA itself because it is inserted earlier in the signal chain. If the noise is in the form of hum from a ground loop or from working too close to flourescent lights in your practice space, etc., the gate/expander may have to have its threshold for turning on set so high that your choices become restricted to screaming or living with hum. It is worth pointing out, though, that hum is blessedly low frequency content, well below the range of the voice, and hiss is blessedly high-frequency content, well above the range of the voice. A well-designed highpass (for hum) or lowpass (for hiss) filter may be all that you really need to make the sound tolerable. The nice thing is that no thresholds need to be set and the onset and fade-out of everything you pass through it is left perfectly intact.
Alternatively, a simple gate, which may be unacceptable on its own, can work comfortably with a filter to do what you want. So, for example, you have constant low-cut filtering to keep kickdrum and bass roar out of the voice mic all the time, and follow it with a noise gate with a very low turn-on threshold to "tidy up" things when there is no voice and you truly want "quiet".