Yeah, I think you'll definitely want to do them indoors and with hearing protection. I also suggest doing it with the crossover in place, if possible. You want to see how the driver handles the highest frequencies when the typical top end roll-off is compensated for, and you don't want your sweeps to be limited by excursion on the low end. Definitely do some sanity checks after each sweep to ensure there are no visible signs of power compression and that THD remains reasonable for the lower frequencies. The "80 dB" refers to the SPL at the MLP (at 9 feet or so from the speaker) that my speakers reproduce for the "-20 dBFS" sweep in those measurements, if that makes sense. My "-2 dBFS" sweep was ~98 dB at the MLP, just as a -2 dBFS treble signal in a soundtrack played at MV "0" would be 98 dB at the MLP. The horns are toed in, so I'd expect the MLP response is maybe 1.5-2 dB below the on-axis response.
Today, I captured the filter response so I can correct the data I have and post non-normalized results once I get around to doing this work. The changes have no impact on the normalized plots.
I also attempted to assess IMD distortion with two-tone stimuli. One of the two CDs (not the DNA-360) has a shorting ring and notably lower inductance, so I thought I could reveal an advantage for it by doing these tests. I put on hearing protection, locked the cats in the bedroom (why they think they needed to be in the room for this is another question) and ran dual-tones at levels exceeding 100 dB at the MLP. For the lower tone, I chose a frequency at the low end of the range of each driver. The DNA-360 got 800 Hz, which is slightly below its recommended crossover point. The other driver, which doesn't play as low, got tested with 1100 Hz. Casual inspection suggests that the lower inductance CD with the shorting ring does indeed have lower IMD when the higher tone is > 10 kHz or so but it's very hard to quantify the differences using the capabilities that REW provides. Even with the overall response being dominated by direct response energy, there is always the chance that one or both chosen tones will fall in a narrow peak or null, which completely throws off the numbers. I think to do this "right" requires a series of sine-sweeps simultaneous to a low frequency stimulus. The stimulus would be increased in level, but not necessarily the sine sweep as most of the IMD seems to be caused by the low frequency tone, even though it only manifests with high frequency tones playing simultaneously. I'm not sure of how the results should be analyzed either. So, this will require software tools that I don't have right now, and since I'm rather eager to settle on my choice of driver, I'm thinking I probably won't worry about it, even if it is potentially significant.
I'd like to do some more simulations with the data I just got before making a final decision, but right now the DNA-360 is my favorite because of its excellent low frequency performance. The DNA-360 has a recommended XO of 950 Hz vs. 1.2 kHz for the other driver but after my distortion measurements (not posted), I'd be comfortable using the DNA-360 even lower, whereas 1.2 kHz looks a bit too low for the other CD without a very steep slope. Likewise, seeing how much the TD12M woofer starts to beam above 1.4 kHz or so suggests that the DNA-360 will work much better in a crossover with a wider transition region, which is something I'm shooting for. The SEOS-15 itself appears to hold its horizontal pattern all the way down to 600 Hz in its box, and I'd love to be able to take advantage of that control given that I am designing for multiple listeners.
My polar response measurements indicate that the change of driver has a negligible impact on polar response, except in the break-up region, which is likely to get EQed way down anyway. Even then, the differences are fairly minor. Both of the drivers I'm testing break-up at around 17 kHz, and your measurements suggest that the BMS (4550?) driver has a break-up right around that point as well. I would guess the particular frequency involved has mainly to do with diaphragm size and material choice. Incidentally, both the drivers I have use polyimide diaphragms vs. polyester for the BMS 4550. In any case, the advantage to having "extended response to 20 kHz" is mostly ruined by the presence of break-up, IMO. If you toe-in to 22.5 degrees or more and EQ the response including the top end to be smooth, listeners sitting outside the MLP and directly in front of a speaker will get a peak there. Furthermore, the power response will likely be uneven and possibly accentuated in that region.
IMO, not nearly enough attention is given to top end response in a typical speaker. Many people test their hearing with sine waves and conclude things like "I can't hear above 15 kHz" or some such. For starters, hearing sensitivity drops off quickly above there, so you may be able to hear it if you used a higher playback SPL. More importantly, sine waves may not be nearly as audible as broadband content there, including transients. Too much output above 15 kHz adds an ear-pinching sensation and makes the sound brittle and inexplicably loud. I've used the adjective "glaring" before. Like it's irritating to your ears even though it doesn't sound loud. Another analogy is finger nails on a chalk board, which probably does involve a lot of top octave sound in real life. Any problems in response here tend to be very dependent on source material. A lot of content has nothing above 15 kHz, but some stuff has a huge amount of energy up there. I'm fairly certain I can hear the 17 kHz peaks in my current CDs at times. With most stuff, it adds just a bit of crispness and sparkle to the sound, which while not necessarily accurate, is at least pleasing. However, I do have some stuff that's not so pleasing to listen to. Some of the worst offenders IME are a few hi-res remasters where the 15 kHz+ stuff seems to have been pumped way up, presumably to satisfy the customer who expects to hear very high frequencies in digital content mastered with a higher sampling rate. I have a remaster of "Miles Davis - Sketches of Spain" with muted trumpet that is practically un-listenable in my current setup. Even standing in the kitchen (outside direct line of sight from the speakers), the trumpets are irritating at even fairly moderate levels. Thankfully, this is the exception more than the rule. Otherwise, differences in speakers used in mastering probably account for a lot of variability here. I suspect the ears are concerned with power response at least as much as direct response in this region, so it wouldn't be surprising if content mastered on domes that beam in a relatively dry room tends to sound very bright on a system with horns and a lot of high frequency ambiance.
The next thing I want to do is try to see if I can EQ down the break-up peak in the DNA-360 using a fairly high Q notch filter and if the new response is more or less forgiving with high SPL sweeps. If it works out, then I can hopefully retain some output above the peak. Otherwise, I'll probably just have to roll things off a bit more aggressively starting at 15 kHz or so, which kind of defeats the point of having a CD that can extend all the way to 20 kHz in the first place. Doing this is hardly unprecedented of course. In fact, I believe most examples I've seen of speaker response voicing involve top-end roll-off that starts well below 15 kHz and becomes quite substantial (6-9 dB or more) above that point.
Maybe some day, I can try out a Be diaphragm driver I can actually afford, like a Radian, to see if it's better behaved, but knowing that the Radian's don't have the best reputation for low frequency performance kind of dampens my interest in them for this design. For now, I want to wrap up this design so I can start building a center channel speaker. Voices sound so amazingly good with these speakers, and I can't wait to be able to use one for dialog in movies.