[JohnR]

Here's something I learned recently that at first glance may seem counter-intuitive.

Somehow, during a series of experiments, I had inadvertently set my tweeter high pass filter at the wrong value. The final crossover point was supposed to be 4200 Hz, but I had the tweeter set to 3000 Hz. You may at first think that would mean more output in that 3000-4200 Hz range... not so.

In the plot below, the orange line is with the tweeter set at 3000 Hz, and the blue line with it set correctly at 4200 Hz. (The mid is highpassed at 4200 Hz, both are LR8 slopes. Ignore the 8 kHz dip, that is a different issue - a "baffle" cancellation effect on the tweeter.)

So not only is the tweeter working harder, but there is less treble output as well....

Doh 

[Tyson]

Yep, one reason that steep crossover slopes are nice is that they minimize this type of interaction. 

[Duke]

Somehow, during a series of experiments, I had inadvertently set my tweeter high pass filter at the wrong value. The final crossover point was supposed to be 4200 Hz, but I had the tweeter set to 3000 Hz. You may at first think that would mean more output in that 3000-4200 Hz range... not so.

Phase is also rotating (shifting) as you change the response curve of the tweeter in the crossover region, so what's happening in this case is, the tweeter is going out-of-phase with the woofer in the region of overlap.   

Your curves there just might be the best illustration of this phenomenon I've yet seen!

[Russell Dawkins]

Is the tweeter connected in inverse polarity to the bass/mid?

This result would be consistent with that, if you are using 6 dB/octave slopes, but would be expected if using 12 dB/octave high pass to tweeter and 12dB/octave low pass to bass/mid, and the drivers are in the same polarity. 12 dB slopes require inverting the polarity of the next driver to avoid phase cancellation at the crossover point,

[JohnR]

Sorry for not being clear, these ARE steep slopes! By LR8 I meant 8th order - I guess the terminology is inconsistent. Both drivers are in phase.

What Duke says is I believe the reason for this.

This is what you get for tuning "by ear"   

[Danny Richie]

Duke is correct.

[Russell Dawkins]

This does suggest, doesn't it, polarity inversion (and phase cancellation) in the crossover region, presumably due to crossover characteristics?

I wonder what the curve would look like with the polarity reversed on the tweeter and the crossover points adjusted as necessary - and what it would sound like in comparison.

Oops! - didn't notice your reference to LR8 slopes, John.

[JohnR]

Hi Russell, I'm using IIR filters, so both drivers go through 720 degrees of phase shift. Correctly set, they are in phase (albeit 720 deg apart). But it seems I was lucky enough to set the crossover points far enough that they are 180 degrees apart in the overlap region.

[Davey]

It actually looks very close to the expected response.  Assuming you had inter-driver delays zeroed out, microphone on the optimum axis, etc, etc, the SPL response should have an approximate -5db dip at 3500hz and the phase response would be as shown in this simulation.

Cheers,

Dave.