OK, it's taken a while, but here are my findings.
I sent them to Amir directly a week or so ago so he could take a look, and now I'm haring them with the group directly.
I had to wait to get the sample Amir had tested, since he had already returned it to the original owner. I contacted that owner to arrange to have it shipped to me , along with the other of the pair, once he received it. Since Amir had not noted the serial number I made sure the owner marked which one was the sample that was returned from Amir.
I also asked that he give it another listen to compare to the retained speaker, in light of Amir's findings
His response was that maybe, just maybe he heard something different, but wasn't totally sure that it wasn't just bias from knowing Amirs findings.
I have now studied the speaker samples; the one you tested, the other sample from the customer, and a number of samples I had in my lab.
These are my findings.
1/ The sample you tested has a trim ring that is slightly raised in a few places. This lifted it in such a way that it could potentially vibrate.
2/ It’s pair showed no such signs. The trim ring was properly seated.
3/ In my initial listening and testing sessions, I found no evidence of vibration (Rattling) during standard sweep testing or odd noises while listening to the test track you identified. These tests were done at normal test level and listening level.
In light of your comments about listening at high levels, and using an extremely powerful amplifier, I brought out an amplifier rated at 300W into the impedance of the speaker at the frequency in question.
I first listened with the test track and adjusted the level to just reach clipping point during that track segment. I could not hear the noise. I then raised the gain to push the amplifier into severe clipping, to the point that on the loudest vocals the voice was heavily modulated and distorted due to clipping. I could still not detect the noise you described. Nor could my colleague.
I then went to do sweep testing from an oscillator to try and zero in on that frequency range. At normal sweep testing levels I could not detect any odd noise.
However, if I increased the level to just below clipping on this test amplifier, in this case at 80W/5 ohms, I now heard a very noticeable rattle/vibration.
It was a little more obvious on the sample with the raised trim ring, but was still evident on the pair sample.
After a short while I went to push on the midrange cone to see if it might be rubbing on a potentially off-center tweeter, but was surprised that the midrange cone was very hot. The midrange driver has a voice coil wound on an aluminum former bonded to an aluminum cone. This gives quite a good heat transfer from coil to cone.
I let it cool down, and repeated the experiment. Same result. In just a matter of 10-15 seconds the cone became very hot.
This is a lot of power being dissipated in the driver. On music I have never noticed that cone getting warm, despite playing at loud levels even in show conditions.
Once I backed down the level a few dB, the vibration noise went away.
It seems that the vibration happens at high level. But this level is way beyond the rated power level of this speaker. Let me explain how we do our power testing and how it relates to the power contained in music rather than sine waves.
When we establish power rating on a speaker, we test with IEC noise (pink noise filtered to a limited bandwidth to represent a music spectrum, with a crest factor of 6dB) . Let’s say that we want to test for a Max power of 100W. We adjust the long term average power to 25W (6dB crest factor) equivalent to a 100W amplifier just at clipping. We then run the speaker for 96 hours at this level. This is severe. Over such a long period the speaker gets very hot, the drivers being much too hot to touch. After cooling down, we retest the speaker. It must pass the standard test limits we put on a production speaker.
We then increase the level to 33W and run for a further 24 hours. it must survive and shift its specs just moderately. After that we turn up the level to destruction point. The speaker must not catch fire.
This is a very severe overall test. Much more severe than regular music represents, and much more severe than the AES/IEC long term power handling test spec.
In the case of the UNIFI, the max power rating is 140W. The testing we do allows for up to 25 or 33W of continuous power. In testing this sample I had to put in 2-3 times that continuous power level, centered at the frequency of the issue, to excite the noise, and on music I tested at 2 times the max power level and just below clipping level of the amplifier and heard nothing. Even when running into gross clipping I only heard the effect of gross clipping of the amplifier
It seems that the level you were testing at was far in excess of the max power rating of the speaker. This could account for the comments from other listeners that have not heard the sound.
I will put the speaker back together, and get it, or a fresh sample, returned to you to continue your evaluation. I will also track down the rattle that I did hear when the speaker was grossly overdriven and report back.
Regards
Andrew Jones
ELAC