The original AR "pot" is actually a rheostat, with a constant 16 ohm impedance, with the wiper varying from 0->16 ohms.
Well, no, unless I'm reading the schematics wrong, it's a wirewound potentiometer used as a variable resistance divider. A rheostat is not a divider. A potentiometer can be wired as a rheostat, but I don't believe that's the case in AR4x. See the pics below: rheostat, left, potentiometer, right.
[Of note: Wikipedia doesn't QUITE agree with me on this distinction, taking a somewhat broader view that rheostats are merely high-power potentiometers, but also duly noting that rheostats "are usually used as variable resistors rather than variable potential dividers."
http://en.wikipedia.org/wiki/Rheostat Two terminals, it's a rheostat, and three it's a pot, according to Zilch.
]
Neither is it constant impedance. We look at the schematic and see the 16 Ohms connected from the capacitor to common, and say that it's constant, but once you connect a tweeter to the wiper, the impedance seen by the capacitor, even assuming that of the tweeter itself is constant, varies according to the wiper position.
Consider the extremes: at maximum attenuation, with the wiper at zero resistance, the capacitor "sees" 16 Ohms to common. However, at minimum attenuation, the tweeter is in parallel with the 16 Ohms, and thus the capacitor sees a lower impedance. If it's an 8-Ohm (nominal) tweeter, that's 16||8 = 5.33 ohms, and the filter highpass frequency increases proportionately.
Now, remove the 16 Ohms in parallel with the tweeter, the impedance goes to 8 Ohms, and the filter frequency drops. Without doing the calcs, just looking at the filter charts in the PE catalog, taking the AR pot out of the circuit drops the highpass frequency from ~1.5 kHz to 1 kHz, a bad idea, I'd say.
There's more going on here than simple attenuation, and I suspect some of the altered voicing reported by those who have tried the "remove from circuit" approach may be due to these factors. If you're going to remove the pot, put a fixed 16 Ohms in its place to restore the original impedance and highpass filter operating point at minimum attenuation.
My conclusion that the pot (or perhaps RoyC's suggested approximation thereof) is a better solution than a straight L-Pad is based upon the observation that, given a fixed LOWpass frequency for the woofer, to maintain a constant voicing of the system, it would be appropriate for the highpass frequency to rise with increasing drive. Without being aware of it, we actually vary the acoustic crossover point whenever we change the balance of the drivers using common crossover topologies. See Fig. 7 here:
http://sound.westhost.com/bi-amp2.htm
I appreciate that y'all have been through this AR4x pot thing ad infinitum previously; I'm merely revisiting that work and putting forth a rationale favoring one particular approach over the others commonly advocated here.... :yes:
Edit: Wikipedia corrected its information ~12/07 in conformity with my position as stated above. In discussions there, the prior version was dubbed "Ignorant." It now references IEEE standard definitions:
http://en.wikipedia.org/wiki/Potentiometer