The exact cause was unknown.
All I could do was repeat as many different comparative measurements as possible.
As a result, various leaks were visually confirmed and resolved, and no problems were found in the passive radiator.
Adding an inner sound absorbing material was one of the things I could do, and it happened that it improved the symptoms.
And by adjusting the degree, I found the best setting possible (the goal is to allow the harmonic distortion of the problem band to increase proportionally with the output), and apply this to sample B as it is, and then sample B All I have to do is re-test to make sure it is not broken.
Aside from that, of course, both speakers had significant binding force leaks, so this is also one of the crux of the issue I raised.
There is enough measured data to discern the root cause of the peak in harmonic distortion in the bad sample. The Purifi Extended range driver can exhibit a resonance at that frequency when the driver is mounted in such a way that the mass spring oscillates. This is level dependent and why the distortion will not then follow a linearly increasing trend.
When the mass spring is damped or not excited there is no problem, evidenced by measuring the driver with the magnet on the table.
Adding the extra stuffing was not the right course as evidenced by the Impedance response, but the combination of that and however the driver was then remounted stopped the mass spring from being excited enough to cause higher distortion.
March's solution to mounting the Purifi driver to avoid exciting the oscillation was to use rubber well nuts and a specific torque pattern. Somehow in one speaker this has gone wrong and resulted in the distortion peak. Chasing that problem has highlighted the non sealed binding posts, this is a manufacturing defect but I think somewhat of a red herring in identifying the cause of the distortion peak.
Designing high performance speakers and tracing the true cause of issues is not a simple business. It is easy to change more than one variable at a time and come to an incorrect conclusion.
If 380hz is inherent in the motor how did they even get this graph.
Isn't this infinite 'baffle' measurement?
It is a big disappointment with their decision to hide the problem with cheating. Or shall I say unprofessional method?
This is a most unfair comment, particularly when aimed at a manufacturer that provides a high level of data about their drivers to consumers. Be wary of letting dissatisfaction with March flow over.
The distortion occurs when the driver is mounted in such a way that the mass spring oscillates. The need for careful mounting is the price to pay for making a driver capable of such high excursion and low bass in a small form factor.