Waterfall plots (cumulative spectral-decay plots in 3 dimensions of SPL vs. frequency vs. time) can be used to show cabinet noise or resonances, or speaker responses. For cabinet noise, the responses of an accelerometer are shown as in the first graph below. The second graph shows a waterfall plot of the speaker's response. In this case the response is speaker output as measured by microphone, instead of cabinet vibrations as measured by accelerometer.
I chose as examples, plots from a
recent Stereophile review of JBL Stage A170 speakers. The differences are not immediately obvious. I'm not sure if that long response ridge at about 250 Hz in the first graph can be heard or not. But it's existence suggests a potential source of noise that might be remedied by cabinet damping efforts.
The cabinet noise doesn't seem to appear in the second graph, but it's worth noting that the second graph goes no lower than 400 Hz. Is there a reason why the second graph omits those frequencies? Was it done to hide the effect of cabinet resonance, or is a lower limit of 400 Hz routinely used to avoid the much larger peaks and ridges often seen in the bass and lower mid-range? I don't know.
While we're comparing these two graphs, note the differences in the ranges of time and SPL between them. The plot of cabinet noise (first graph) goes on for as long as 75 msec, where the speaker response plot (second graph) stops at less than 4 msec, a much shorter time span.