Okay, for now, just mess with the "Traditional" tab-window. As you change values here, the "Graphical" line will change, but to see that, you'll need to press that tab. No biggie, you don't NEED to see it (folks who don't use CEP usually don't anyway, but this may help you understand compression a bit more... when you find the settings that work, you may notice a correlation between what you're doing to the dynamics and the graphic...)
Lets go back to your "spike." Go to the squiggly-line view and highlight an area that surrounds a given spike. Play it back repeatedly and see how loud that sucker gets, in terms of dB. Now, how loud do you WANT it to be (the peak, that is. we'll get to the shape later)? The DIFFERENCE between these two volumes is determined by the ratio and the threshold.
The RATIO will diminish this wave in the ratio designated, but it won't kick in until the volume has reached the THRESHOLD you designate. So while you're looking at your wave, imagine it "squashed" by a ratio of 2:1 (it would look EXACTLY the same as if you normalized it to 50%, had you previously normalized it to 100%). THAT's what it'd look like if the RATIO was 2:1 and the THRESHOLD was negative infinity. See, if there's any volume at all, then it's ABOVE negative whatever, and the RATIO is compressing all of the signal above that THRESHOLD (all of it).
Now, if you set the THRESHOLD at -50 dB (because CEP has a -100 dB range on it's graphic, I'll use that figure as "silence"), everything up to -50 dB will remain unchanged. However, the area ABOVE -50 dB will change at a ratio of 2:1. So, you're only processing wave information that exceeds -50 dB.
-50 dB is pretty quiet, so lets go further with this.
Let's set your THRESHOLD at -10 dB, which might be appropriate for your situation. If we set your RATIO to 2:1, then any sound ABOVE -10 dB will only
add 1/2 of it's volume to the wave. Say your sound gets to, oh, -2 dB. There's an 8 dB difference between -10 and -2, so your peak would be "turned down" by 4 dB (8dB/2). Now your peaks are only -4 dB (half of the difference between the THRESHOLD and your peak).
Nothing's happened to any sound lower than -10 dB. NOTHING. But anything ABOVE -10 dB is being REDUCED in volume by half of the difference between your THRESHOLD and peak. At the MOST, provided you top out at 0 dB, your volume only diminishes 5 dB.
Does that make sense? I'm not sure how to explain it any other way...
Now, the art (and proper use) of this technique is figuring out the RATIO and THRESHOLD you want. I wish CEP would show the waves against a "dB" scale (it probably does, mine won't display it no matter how many times I "set" the scale to the right), then you'd be able to look and see what values you need for THRESHOLD. Because you only want to "squash" (compress) the "unruly" parts of your sound, which are those peaks you displayed earlier in another thread. In such a case, I'd experiment, listening to what the compression did at different settings.
Remember, NOTHING will be affected below the THRESHOLD. So try and figure that out first. Then try and figure out how little you can squash THAT before it sounds just plain fucked up.
BTW, that "graphic" view uses a straight 45* angle to denote no processing whatsoever (on the axis). The angle of the deviant "angle" is determined by the RATIO, and the point at which that "angle" deviates is determined by the THRESHOLD. I wouldn't mess with the bandwidth issues until you've played with compression a while; the only time I've used it is for mastering (and that was copying specs off of a radio-compressor so that I could hear that "radio sound" you probably hate.
)
Here's a link to Sonusman and Shailat's compression page:
http://www.geocities.com/shailat2000 . That'll help if you're still wondering...