Time Dilation, Part 2 Page 2
Fig.5 By way of contrast, this—on the same scales and with the same settings—is the equivalent to figs. 4 and 5 that would be obtained using the author's listening room for the measurement.
Although I made the point early in this campaign that my motivation was not to improve the measurement of bass frequency response so much as to enhance the resolution of cabinet and stand resonances, fig.6 compares the frequency response obtained using the ground-plane technique with one I conjured from the same speaker using the nearfield method. (I say conjured because the B&W CDM1 NT is a bass-reflex design—its overall nearfield response is obtained by weighting the contributions of the speaker's bass driver and port, a process that always involves a measure of uncertainty.) Also included in the graph, as arbiter, is the bass response from B&W's own design simulations, which were based on driver, port, and cabinet parameters. If we assume the simulation to be accurate, then neither measurement does a particularly good job of capturing the LF response, although an argument can be made for the ground-plane technique performing rather better. It would be foolish to reach conclusions based on a single sample, but this does at least make the point that, with reflex loudspeakers, the accuracy of a nearfield measurement is dependent on precise weighting of the driver and port contributions. I won't be alone in avowing that this is an uncertainty I would happily forgo when measuring speakers—of which more in due course.
Fig.6 Comparison of frequency responses obtained from the B&W CDM1 NT loudspeaker using the nearfield method (blue trace) and windowed ground-plane method (purple), with B&W's simulated response included for comparison (red).
Concentrating now on resonances, figs.7–9 show three further waterfalls obtained using the single-plane measurement method, which should be compared with that of fig.4a. Whereas for fig.4a the speaker was placed directly on the sports hall's hard rubberized floor (as pictured in fig.10), for fig.7 four soft rubber equipment feet were inserted between the speaker base and the floor, to provide a measure of decoupling between them. My original intention for this was to check whether vibrations reaching the measurement microphone via the floor were corrupting the results. There is no evidence of that here (although it's always advisable to check), but there is a very obvious change in the appearance of a pronounced resonance at around 500Hz. This same resonance can be seen—somewhat enhanced, which is intriguing—in figs. 8 and 9 as well, for which the speaker was mounted on a stand (as pictured in fig.11 for the two-plane case) with, respectively, three metal cones and four rubber feet between the speaker base and stand top plate.
Fig.7 Cumulative decay spectra obtained using the same measurement conditions as in fig.4a but with the speaker decoupled from the floor using soft rubber feet. Note how a 500Hz resonance is now evident.
Fig.8 Now the speaker is mounted on a stand, with three metal cones between the speaker base and stand top plate.
Fig.9 As in fig.8, but with soft rubber feet substituted for the cones.
Fig.10 Ground-plane measurement setup in the middle of the sports hall's floor.
Fig.11 Two-plane setup in the same hall, with the measurement microphone taped into the junction of floor and wall and the speaker angled at 45º.
The obvious conclusion to draw is that this resonance is a cabinet mode that is suppressed when the speaker base is supported on (and damped by) a flat surface. So the use of speaker/floor decoupling in ground-plane testing actually commends itself for two reasons: to prevent transmission of vibrations to the microphone through the floor, and to ensure that the resonance behavior of the speaker cabinet is fully expressed.