Okay, I thought you used noise to measure frequency response.
I use a log sweep to gather FR.
The HD and Compression tests are done separately; not with the standard FR measurements. The HD and Compression tests I've previously shown are done using a single tone in the manner I described above.
So there remains the strange behaviour of the 9V measurement compared to those with lower voltage.
The reason for the termination of the measurements at 1.4kHz is still obvious to me. An antiphase resonance produces a deeper and deeper dip with increasing voltage (first purple rectangle).
Astonishing are the resonances at 4.8kHz and 5.5kHz (second purple rectangle) - are these influences due to the woofer break-up or resonances of the tweeter?
Very strange is the behaviour in the blue marked frequency ranges 1) - 4).
There are grey lines above and below the red line. A dip suddenly changes to peak 1) and vice versa 2).
The resonance modes of a driver usually change with frequency and not with amplitude.
Hence my question whether other resonance sources can be excluded.
View attachment 73852
1-4) I do agree they are perplexing. However, if you look back at my original compression plot you can see at least one occurrence of this at ~430Hz (your #2). And this was measured in the nearfield at 0.25m:
Also, keep in mind that - because I was using singular, log spaced tones and not a sweep or multitone stimulus - that the previously provided measurements might not have hit the exact same frequencies in the way the multitone measurement does. Not saying that is the sole cause for the differences. Just noting it as a potential difference. Nevermind the fact that a NF measurement of a speaker itself can be misleading, depending on how the wavefront is formed at the microphone position. I'll address this below as well.
Now, having said all of that, let's look at another example. I measured *a new* speaker this week (data hasn't been posted yet). I ran this set of measurements in the exact same way and the exact same day I tested the Neumi with ISC+MTON. Literally, I tested the Neumi with MTON, I took it off the stand, I put the other speaker on the stand at the exact same position, not moving the mic or anything else, and then tested this new speaker with MTON. They are under the exact same conditions; tested about 10 minutes apart. Not even the temperature was different. Therefore, they are directly comparable in every way. Below are the results:
To make it easier to see the differences, here is the 9V from the new speaker (green) overlaid on the Neumi BS5 results (gray/red) where 9V was the failure point. If the room was an issue, as you are concerned, this would show up in the results as numerous overlapping points. As you can see, there is no distinct correlation in this regard.
and to make it much easier to see...
Therefore, I have no doubts the measurements are free of room conditions, as should be the case with using Klippel's ISC and MTON module in conjunction.
------------- (break) --------------
Regarding the purple boxes, it's hard to say without diving further. Below is the Fundamental+HD at 2.83v/0.5m. The Absolute PHD (a predictor of audible distortion, iirc) is up around that range. The HD lines also show an uptick in distortion between 400-700Hz and 1-2kHz.
Then, below is the NF response (at 2.83v, 0.50 inches from the baffle, SPL scaled down). You can see there are some higher levels of distortion until ~2kHz before falling off.
It does seem, though, that there is indeed woofer breakup in this region. Though, IIRC, there was also some lingering port resonance in this area. It's been a while since I've taken my compressible flow course but I wonder if the port is the root cause with the purple boxes.
Notice the two distortion profiles are not the same. Not even in the 200-1kHz region where the mid/woofer is dominant (meaning, I'm not concerning myself with tweeter response in a comparison of a NF woofer measurement). And notably the 3rd order distortion profile lies almost exactly with the 2nd order in the NF while this is not true in the 0.5m measurement. This is a prime example of how varying measurement methods can lead to different results. Even with the exact same stimulus applied. This isn't to say either method is wrong. It is to say that each can be effected by conditions. Aside from my own tests, this is something to keep in mind with others' results, like Amir's. Comparing one person's results to another is likely to cause issues because the setups are not the same. They really should be compared within their own set.
However, with the ISC + MTON modules, that is not the case. If another person was to follow the same protocol I am using then we should have comparable results.
Bottom line:
As of this very moment, I am 100% confident in the use of my measurements to extract maximum SPL. To be blunt, I am much more confident in this method than ... the
other methods we have been presented. My only ongoing task is to determine which way to apply stimulus (which type; how long). There are a multitude of IEC variations here so I just need to pick one (or two) and roll with it.
FWIW, if you want to see the processes I am using, you can watch/view the Webinars here. Particularly numbers 4 and 5. I have worked directly with Klippel on the interpreation of these since they have been released but was unable to use them until they released the newest updates (they were using their beta versions in the Webinars). That's why I was so excited to start testing them when they released the updates this week.
http://www.klippel.de/know-how/education/webinars.html
I hope that helps address your concerns. Trust me, it's different for me as well and I spent hours (and hours, and hours) combing through the data, re-running over and over and trying to make sure that what I am presenting is accurate and reproducible. I don't take this task lightly. Presenting data to a friend and having it wrong is one thing... presenting it to the masses and having it wrong is something else altogether. I do my very best to be rigorous and document updates but, frankly, there are so many details that go in to my efforts that I don't have the time to post them and I honestly doubt anyone cares. Maybe I should set up a live web-cam next time I run a speaker through the paces and you guys can follow along with me for the
literal 10+ hours I spend testing and verifying/analyzing data. I'm not at all exaggerating.