After further listening and testing, the tweeter appear to be just a little too bright. Above is a 1/6 octave smoothed measurement at 18″ (this distance compares well to measurements at 36″, only with better signal-to-noise). Both left and right speakers are measured. Below is the left speaker and the tweeter, smoothed to 1/3 octave. It looks like about 2 dB more tweeter attenuation should even things out, which brings the total tweeter attenuation to -5.5 dB.
The final result, smoothed to 1/6 octave:
So I have noticed during testing that, when I add a second order time delay network between the tweeter and the cross-over, the response above 10 kHz changes. That doesn’t make any sense at all – the time delay network should only change the phase.
The blue line below is the starting point of this experiment, and the yellow line represents the “solution” to this dilemma.
The solution? Adding a Zobel to the tweeter to bring down the inductive rise in impedance. How much of an impedance rise? Almost nothing!
Maybe there is something else going on… I measured this multiple times and obtained the same result. Measurement conditions? The mic is 6″ away, output level set to -31 which isn’t too loud, no indication of overload on the mic pre-amp.
For this test the woofer cross-over is fixed. The lower mid-woofer has a 2.0 mH inductor in series, the whole connected in parallel with the upper mid-woofer. Then an electrical second order cross-over composed of a 1.1 mH inductor and 15 uF capacitor, the cap is in parallel with the mid-woofers, the inductor is in series. Comparing to the target of 2.5 kHz (in figure below) shows that the x-over frequency is just a little low. So change the inductor to 1.0 mH. (Measurement distance is at 36″.)
For the tweeter a 2nd order electrical crossover also is appropriate to reach a target 2.5 kHz LR4 crossover. The series capacitor value is 8.2 uF with a 0.2 mH inductor in parallel with the tweeter. Notice that the match to the target is excellent to about 1 kHz, and then below that the tweeter level slope appears to change from 4th to 3rd order.
The final integrated result. The tweeter is -5.5 dB, set by a parallel resistor of 7 ohms + series resistor of 1.5 ohms. (Measurement distance is 36″.)