Some measurements...
A couple notes... these may change slightly. I've done about 5-6 crossover revisions and these where built with parts I had on hand. For the production version I'm planning on using Erse Super Q inductors which have a much lower DCR and will bring the EX-6.5 output up some. I'll adjust as needed but these graphs have a higher frequency tilt that will not be so exaggerated in the production version.
The crossover is around 1.6K which is low for a typical 1" dome tweeter. No problems with this Peerless unit though. Distortion testing (THD, IMD) in the final version to confirm. Crossing over this low has the advantage of having smoother response off-axis. The midwoofer begins to beam at a much lower frequency than the tweeter so crossing over to it low, where the dispersion is wide (relative), keeps the off axis from having a large narrowing pattern near the crossover frequency. Its harder on the tweeter so you need a good tweet, good crossover design & then you need to measure distortion in the final design to confirm that you get it right.
The dip @ 3K & 6.5K are baffle diffraction related dips and their effects are negligible off-axis. The peak @ 15K is in the tweeter response and remains to some degree in all the measurements.
Here is about 15 deg off-axis. The upper frequency tilt starts to come down a little and you see the beginning of the tweeter roll-off. The broad peak around 2-4K is a work in progress. I may allow an on-axis dip in the final response to get this smoother off-axis. This area is what sounds forward or "bright" in a loudspeaker design. Its not the >10K. Brightness is typically due to a problem in the 2K-5K range.
Here is the 30deg off-axis.
And this is >45deg. Notice the crossover notch starts to show. The woofer's dispersion pattern is narrowing relative to the tweeter at this extreme off-axis position.
Here is the reverse null. This is 1/6th smoothed so the bottom of the null has been filled. The baffle step showed up more in this measurement. I've done about 4db of compensation leaving 2db for room reinforcement.
Here is the waterfall. The waterfall is useful to check for any resonance issues in the final design. I look for everything above 1K to die off >40db down within the first 2ms. If you get up around 4-5ms things start to become audible but it also shows up in the FR on-axis or off. This is just another way of checking for FR issues and you can see my broad 2K-4K peak in the waterfall showing up even though it wasn't in the on-axis response.
Overall though.... this is much better than you see in typical designs that do a simple on-axis measurement, optimize for smooth response and call it done. This represents a LOT of measurement, model, build, measure, tweak, measure, tweak, measure, tweak, measure, listen, tweak, measure, listen, tweak, measure, listen design cycles. You get the idea.
