PRX415M Crossover Teardown/Analysis

I thought it was strange there was a dip at the crossover of 2.2khz on this design. My first thought was somehow the polarity on one of the drivers was incorrect. After pulling the grill of and removing the drivers I noticed both use polarized connectors (small for negative and large for positive). The wires traced back to the correctly marked position on the crossover board so I knew that was not the case unless the crossover board was marked incorrectly.

So I decided to measure the drivers individually and compare to the combined response.

Somehow I forgot to get a photo of the crossover when I had it pulled out without the wires in the way:

PRX415M Individual Driver Responses:

This was measured indoors with short gate so low frequency resolution is lost, the grill was removed for these measurements and you can see how the high frequency response evens out with it removed. Since the dip tracks the shape of the individual driver frequency responses through the crossover rather then a cancellation between the two it shows that they are roughly 90 degrees out of phase, in which case reversing the polarity (180 degrees) won't serve to fix the dip.

Taking the response of both drivers without the crossover show that they do clearly extend well past the crossover, so it shouldn't be a driver issue where one is falling short.

Using the labeled capacitor and resistor values on the crossover combined with the individual driver response measurements I had taken previously I was able to simulate the crossover in XSim and determine the approximate values for the two coils which were not labeled rather then having to remove them from the circuit to measure directly.

The crossover in this model appears to leave a sizable gap between the lowpass and highpass on the two drivers. The published response curve from JBL does not have such a dip at the crossover so I can only assume it used a different crossover revision.

JBL's Published response graph for the PRX415M

Crossover schematic and simulated response:

Modifying the high frequency network by adding 10uF to C2 (15uF total), changing L1 to 0.4mH and adding ~4 ohms of series resistance on that parallel leg fills in the dip and results in a response that looks much more like the one published by JBL: