Updated FR with final crossover:
This is the LSPCAD file but the measured data is virtually the same.
Unsmoothed raw data. All data is good to about 150Hz. Ignore the data below that point because it's not accurate.
Notice that the tweeter crossover is at about 4.1K, just outside of the critical midrange. The phase jump is at about 2.8K and then we have linear phase across the rest of the midrange. We use a 4th order electrical on both the midrange and the tweeter at this crossover.
On the woofer to midwoofer transition we are using a 2nd order electrical @ 500Hz. We don't benefit as much from the steep crossover down low and the cost/complexity goes up considerably using a higher order crossover in this range.
The end result is that we have one driver, no crossovers across the entire midrange with no phase jumps within most of that range. We have all the drivers operating FAR within their linear limits with over a two-octave buffer from all crossover points. No notch filters needed as we don't have cone break-up problems to filter. We have extremely smooth on-axis and off-axis response. All drivers are operating within their sweet spot and not loosing performance due to VC heating and other power related non-linearities. XBL^2 motors on both the midrange and the woofer mean very little motor related non-linearities within their range of operation.

This is the reverse null.

And how about this. A fantastic crossover design as above along with an impedance curve that looks like this.


I'm very happy with this guy. At this point, I don't know how to improve it further. Dail is tired of rebuilding the crossover for listening sessions and I'm extremely pleased with the subjective performance.