Midrange compression driver development

this is the distortion:

for this i also had to match the levels manually. so the absolute numbers are not included and i just look at the shape of the curves, i.e. where the most distortion occurs.

I thought a lot about this and it made a lot of sense to me that this is the root of my problem. But then i remembered that i made a measurement of the CD mounted to the horn but without the corners at the throat -> so it should exaggerate the problem of the assumed throat-reflection even more because of the greater discontinuity. 
(The reason for that measurement was to see how smooth the transition at the throat really has to be)
This is the measurement without corners (blue) compared to one with the corners in place (red) cropped at 200Hz for better visibility. Sadly it had to be done indoors, so there are some reflections in the lower frequency range that could not be windowed out from the response.

As this change presents a much bigger discontinuity at the throat and has little to no effect on the ~2,2kHz dip i am now even more confused of its origin.

The following pictures are for better understanding of the involved geometries. The first one is a cut through the vertical phaseplug and horn profile and the second a cut through the horizontal profile. The highlighted part is said "corner" that was added to smooth the transition from round to square.

and for the complete picture here is the view from the front with all 4 corners at the throat in place:

Elliot Thompson wrote: So how does the two compare in terms of sound? It appears each one offers their own trade offs.  Best Regards,

Actually i did not listen to the new modification since we had really limited time and the weather was beyond shitty 

Regarding the dip i did nothing new since my last post with measurements. I will address that "problem" when the Horn ist finished and the dip is still there. 

My thinking is as follows:
Since the horn has no abrupt changes in area/flare, no straight walls and no diffraction apparatus of any sort, it itself should behave fairly good when fed with a appropiate wavefront. With this in mind, it should be way easier to work on the phaseplug when its prototypes are mounted to this new horn.

And just for completion: 
The horn has a size of 500x360x240mm (WxHxD),

a constant directivity of 60x50° (HxV) 

and it loads the driver to 400Hz (will probably be used from 600-800Hz up)

Here is a measurement of the printed horn with the extended phaseplug in the near field (20cm) with 1dB increments, no crossover, no eq, no smoothing (but windowing at 14ms after peak)

Here is a measurement at 60cm with some eq to get it flat. (We could not measure further away this time without reflections)

No crossover, no smoothing (but windowing at 14ms after peak) 

Here are measurements of the eq'ed response in 10° increments from 0° to 60° in one direction of the horizontal plane (so 0-120° absolute coverage angle). Here i used 1/6dB smoothing so one can actually somehow read the graphs 

And the same for the vertical. This time from 0° to 40° (so 0-80° total).

We really need to get these measurements done outside....but i think one can see a tendency of how this horn performs. There will be better measurements when we finished the one in glasfibre :)