Inspired by tsardoz a few weeks ago, I've been making a lot of use of "tap" testing to identify the natural frequencies of my DML panels. It turns out to be quite informative. By "tap" testing, I mean simply striking the panel and measuring the response using the RTA feature of REW, with the microphone placed very close to the panel. This method is more correctly referred to as impulse excitation. Tsardoz called it "bash" testing! Curiously, I don't think I've ever seen anyone else share tap testing results on this or other forums concerning DMLs.
The most basic use of this technique is (as I mentioned above), simply to identify the natural frequencies of a panel. The frequency response of a DML is highly dependent on its natural frequencies. For example, the panels lowest natural frequency (i.e the fundamental), is approximately the lowest frequency at which can you can expect any significant output. Further, knowledge of the natural frequencies, along with the associated mode shapes (and hence the location of their nodes and antinodes), can help in exciter placement, since it can tell you where to place the exciter in order to excite (or not!) particular modes.
As an example, I will share some tap test results for a 14" by 48" plywood panel I've been working with lately, which was attached to a stiff frame using double sided foam tape around virtually the entire perimeter. In this case the lowest natural frequencies all look like the images below, which show the lowest three modes, that is, the fundamental (or 1,1) and the 1,2 and 1,3 modes, as well as one higher order (1.7) mode.
In these images, the red areas are the locations that exhibit the greatest displacement, and are called antinodes. When you tap a panel, the location where you tap determines which modes get excited (i.e. activated), and which do not. If you tap a red (antinode) location for a particular mode, that mode will be excited, while if you tap a blue location (node), that mode will not be excited. So, for example, tapping center of the panel will excite the 1,1 and 1,3 and 1,5 modes etc, since those modes all have an antinode (red) there at the center. On the other hand, tapping the center will not excite the 1,2 (or 1,4 and 1,6 modes etc) since it has a node at the center.
Likewise, the place where you position the microphone determines which modes you will detect. A mic will most easily detect the frequencies which have an antinode at or near the location of the mic.
For these reasons, finding the natural frequencies of a panel usually involves several measurements, made by tapping and mic-ing at a series of different locations.
In the case of my 14x48 panel attached to the frame around the perimeter, only two are needed to get most of the lower frequencies of interest. The first is tapping 2 or 3 inches from one one end of the panel, with the mic the same distance from the opposite end (and within about a 1/2" from the panel face). The other is simply tapping the center of the panel, with the mic also located at the center of the panel.
Here are the RTA measurements for the two tests:


The first (end tap) shows a series of peaks which each correspond to a natural frequency of the panel. In fact, this captures all (or virtually all) of the natural frequencies of this panel below about 500 Hz, starting with the 1,2 mode at about 85 Hz, up to the 1,9 mode at about 500 Hz. Tapping close to the end, the fundamental was not detected in this first test. However, in the next test (center tap), the fundamental is detected at about 72 Hz. The center tap clearly excites fewer modes (half as many), since it only excites the odd,odd modes (1,1 1,3 1,5 1,7 etc).
Finally, it's interesting to overlay the tap test results with the REW frequency sweep frequency response measured at one meter from the panel. In this case, with a Dayton DAEX25fhe exciter. I placed the exciter about 13 inches from the center in the long direction of the panel, and about 1.375" from the center in the other direction. This was the location that provided the best (flattest) overall frequency response for this particular panel and mounting.
One thing I found particularly interesting is how the strongest peaks in the overall frequency response correspond exactly to the odd,odd natural frequencies that occur at 72 Hz (1,1), 107 Hz (1,3) and 188 Hz (1,5). And on the other hand, the overall frequency response shows no peaks at the odd,even natural frequencies such as 85 Hz (1,2) and 140 Hz (1,4).
We can learn a lot about how these panels actually work by doing testing of this sort.
Eric