My guess is that these are measurements of a single driver. Red = nearfield measurement. All the others are from further away with different XO curves applied.
The first thing to note is the SPL. Not only is it very high, but the SPL is different for each curve. Nearfield measurements are always apparently very loud because of mic distance. So we will ignore the
red (nearfield) curve and assume that the other curves, which are: full range (
green), with 1200-2500 XO (
blue), and with 2000-2500 XO (
yellow), were taken at the same microphone distance.
Re: high SPL. I am guessing that you did not calibrate your microphone, since measuring this loud would send distortion through the roof. Remember that playing loudspeakers too loud results in more distortion for many reasons. It is very important to choose the correct volume for distortion measurements. For the purpose of DSP/room correction, my normal suggestion is to measure at normal listening volume. But if you want to formally document distortion, the standard is to measure with your mic 1m away from the speaker, on-axis, and speaker vol adjusted to produce 76dB at that distance.
Re: different volume for each measurement. Note that there is a 10dB difference between the yellow curve (2000-2500XO) and the green curve (full range, no XO). I am guessing that this is due to application of the bandpass XO and you
did not change the output volume of your speakers in between measurements. As mentioned, 10dB less volume = less distortion.
For now, we will assume that you have taken a valid measurement and your distortion measurement is interpretable, and the high SPL is simply due to skipping REW's mic SPL calibration step. If you have measured at an ear splitting volume, then these measurements need to be trashed and you need to take a new set.
This is distortion of the driver
full range with no XO, in dBFS. Brown is the fundamental, black is THD. I have not shown the noise floor, but it is sufficiently low for this measurement to be interpretable. If you look at the 900Hz peak, it is -15dB down from the fundamental.
And this is distortion with the 1200-2500Hz bandpass applied. The same peak at 900Hz is now -30dB to the fundamental.
Since you asked about using distortion to guide the optimal XO frequency, I have taken your full range driver measurement and switched the view from dBFS to %. As you can see, the fundamental (brown) has been normalized to a flat line, and distortion (in black) appears as a % of the fundamental.
How much tolerance we have for distortion is frequency dependent. In general we want <1% THD for higher frequencies, but we tolerate up to 10% for subwoofer frequencies up to 80Hz. It looks to me that this driver can be high passed at about 200-300Hz if we accept a tolerance of 2% at 200Hz.
Something pretty nasty is happening at 900Hz. I am not sure what it is, perhaps it's the Fs (resonant freq of the driver), perhaps it's a cabinet resonance, perhaps it's a cabinet mode. It may even be something you placed on top of your speaker with a resonant freq of 900Hz. I would look up the T-S params of this driver and see what the Fs is. If the Fs
isn't at 900Hz, it's going to be something else. 900Hz has a wavelength of 380mm, maybe this corresponds with one of your internal speaker dimensions?
And one last thing, the polarity of this driver is inverted. I don't know if it's accidental or deliberate. Sometimes drivers have deliberate polarity inversion to ensure proper XO summation, this practice is very common with passive XO's. I suggest you check your cabling to make sure this isn't accidental.
