Chris, quick note... Accelerometer based frequency domain measurements are not an appropriate method to evaluate cabinet resonance. You do need to examine cabinet resonance in the time domain. You need to look at the effects of your modifications over a period of milliseconds, determining the rate of reduction (decay time). MDF does not resonate at 2 kHz (your measurements indicate that this is the CBM-170SE highest resonant mode). MDF will have its strongest resonant modes in the 200-400Hz range.
Frequency domain plots are perfectly valid - they show a specific amplitude comparison of the total wall vibration response, and in addition, if you can figure out the actual acoustic dB level in reference to direct driver output, averaged for all contributing panels, the only proper way to correlate the audibility is using frequency response graphs over layed with each other to figure out relative dB and Q of the response peaks. To me, waterfall response graphs are of lesser use, since I care about the total cabinet response curve. The accelerometer used is not calibrated, but is a known accurate unit, the ACH-01, which is accurate within a couple of dB, through the audio band. It is perfectly normal to see significant vibration in the kHz range in wood based walls when examining the frequency response of the walls, also, refer to the amplitude based measurements here which show response substantially into the treble band:
http://www.audioholics.com/education/loudspeaker-basics/lowering-mechanical-noise-floor-in-speakers-pt-2/lowering-mechanical-noise-floor-in-speakers-pt-2-page-2
ACH-01 response of another MDF cabinet from another party showing substantial output into the treble band:
http://www.libinst.com/epicure5.gif
Significant amplitudes in the treble range is typical.
Now, it is possible, and I will concede, that the 2kHz peak in my graph, may be the coupling frequency of the accelerometer as it was fastened to the cabinet wall with double sided polymer tape that is removable. In addition, it is also possible the pre-amp used on the ACH-01 here has a RLC reaction in the circuit causing a lower end spectrum attenuation of some degree. But since my measurements are only relative to each other, and made with the same device and conditions, I am only interested in the relative compared differences, which will not be affected by this, as these will remain a constant as lone as the same throughput system is used every time. But besides that potential joint frequency resonance of the accelerometer to wall coupling, I have no reason to doubt the response of the accelerometer. And the response shape of the post-mod measurement is exactly what I would expect, based on the extreme stiff structure resulting from concrete and dense steel bracing(resulting in very low amplitude, but substantially raised panel frequency excitation).
I believe you are seeing *something* -- but it is not cabinet resonance.
In fact, I am interested in total cabinet response, not just the cabinet resonance, since the cabinet is in fact acting like a separate transducer and it's output is contaminating the acoustic contribution to the room.
I don't use the waterfall graph function for this comparison method because it is not easy to overlay waterfall graphs, and again, the FR shows me everything I want to know(overall cabinet response) and the result of overall cabinet response after modifications.
I would be happy to measure your modified 170SE in comparison to a stock 170SE and post the measurements at few specified frequencies (displayed in an energy vs time vs frequency graph (exactly the same process and equipment as stereophile). Here is an example:
The speaker is not in my possession now.
In addition, while I am confident that your modifications will improve cabinet resonance, the dramatic reduction in cabinet volume will absolutely have a negative effect on performance. I would recommend that you increase the depth of the speaker cabinet to compensate for this.
The modification did not include a total refinishing the cabinets(which would have been required to extend the depth). But even so, the reduction in volume here was not great, and the effect that occurred were completely compensated for electronically in the DCX correction curve.
-Chris
