One thing to keep in mind, is that the steeper the curve, the more of a boost you'll get around the cutoff frequency. For instance, a lot of people think a steep high pass slope looks like this:



When in reality this is closer to the truth:



Depends on the filter design of course, but in general this is what's going on. That's why when you cut frequencies, you sometimes see the signal level increase some.

In general I say use as shallow of a slope as you can to achieve what you want. And don't high-pass things unless they need it.

I'm not familiar with that system so I can't say for sure. In general though, most filtering behaves like this.

Here's an explanation:

http://www.slideshare.net/op205/design- ... esentation


The upshoot is that the graphs that your software interface face shows you are usually very idealised, and in reality there is all sorts of bumpity bumps... and further to that, to increase the steepness of the primary attenuation (or amplification) at a given frequency then you will also increase the magnitude of the 'side lobes'. Then you get in to the area of filter design and characteristics...

http://en.wikipedia.org/wiki/Digital_fi ... %20filters


If you want to find out more, then first stop Computer Music Tutorial by Curtis Roads. Also Findog, the given cutoff frequency is supposed to be at the -3dB mark, and from there on is attenuated at the given steepness per octave. There is a little bit before the cutoff that is attenuated, but as Tarekith says, looking at even closer detail it attenuated and boosted with ze side lobes (which are many but diminish, much like the sidebands of FM or AM).

This is not an frequency response for a low pass filter, but the principle applies that the ideal shape is not the fact reality