Dayton Tweeter Tests and Modifications

Accelerometers are useful for finding resonances, but only in a relative mode for this application. There is no effective way for me to have any sort of calibration. Trends and specific resonances are shown, but other possible influences must be considered.

I placed the accelerometer at a point roughly midway between two of the mounting holes and roughly midway from faceplate outer edge to faceplate inner edge. The problem here is that resonances will likely be different at different mounting points. I chose what I thought would likely show the maximal movement, at least for the mounted case. I kept the accelerometer fixed while I mounted it into the baffle in an attempt to maintain some relative uniformity from measurement to measurement. Without extensive testing and correlation, though, absolute conclusions cannot be drawn from this limited set of tests. Given that, however, I still hope that the trends they show will be useful, or at least interesting.

Individually the waterfall plots just show resonance ridges or lack thereof. Making sense of them is not straightforward without some reference. In this case, the reference I will use is the on-axis SPL response of the driver mounted on my large flat baffle (1m x 1.2m). There will be a resonant signature of the driver/baffle combination, which I plan to measure as well. It will be interesting to see the differences between the resonance signature on the large baffle vs. those shown here.

The first two plots below show the CSD for the driver itself, no baffle mounting, and for the driver mounted on the small baffle. These used the default graph settings. I found it somewhat difficult to evaluate, so I changed the display to be 0 degrees horizontally, with maximum tilt vertically (70 degrees). This makes the resonant ridges (and their associated frequency) easier to determine.

The next two plots use the changed perspective for easier comparison viewing. The differences between the two cases are obvious, but making any reasonable interpretations in regard to the impact on the SPL (which is the only reason to test for resonances) is questionable without some way to compare them to the SPL. For this I added the third pair of graphs, those being the baffled SPL overlayed on each of the CSD plots.

Now we have something to work with. Looking at the CSD for the unmounted case above, I see five distinct characteristics relating to the resonances, at 1.8k, 6.5k, 10k, 14k and 18k.

The problem with comparing these two graphs (baffled FR vs. unbaffled Accelerometer) is that there is some consistency between the unbaffled vs. baffled CSD plots, but we know that they change when baffled. It is logical to assume, though, that resonances when baffled will likely occur at these same points, only with different magnitude. An unbaffled frequency response plot also doesn't make sense, so let's move on to the next graph overlay.

The graph overly above is the one on which to concentrate, since it best reflects the conditions which will occur in use. There does appear to be a reasonable correlation between the resonance ridges seen in the CSD plot and the non-linearities in the frequency response plot. The most obvious points are at rought 2k, 4.5k, 10k, 15k, 18k and >20k hz.

It occurred to me as I was writing this that I should look at the impedance taken mounted vs. unmounted to see what difference might occur. The measurement I usually take for a tweeter is with it on its back (since it doesn't need a box). However, the impedance glitches seen in the previous graphs might be reduced when on a baffle, if it is having a significant impact. If I find any differences worth noting, I'll post them.