I actually wasn't aware that someone was doing that much in CSD (waterfall) for headphones. When I googled it, it took me right back here!
http://www.head-fi.org/t/566929/headphone-csd-waterfall-plots
If you read the introductory post, you'll see that the goal of CSD headphone testing is to try to better visualize what is heard, an improvement over simple time-blind FR. Then skip around a bit, and read through some of the results and comments. What you'll find backs up my earlier statement that waterfall plots are difficult to digest. The do reveal ringing quite well, but what people hear doesn't always correlate as well as you'd hope for. A CSD isn't hard to understand. It's the connection between what's seen and what's heard that isn't as solid as we'd like. When the CSD supports the audible impression, comments are like "that explains it." "I heard that xKHz ring." When they don't support the impression, people seem to say, "That's weird" "didn't expect that". It's an interesting mix. What's also not published is the detail of how the tests were run, unless I've just not looked in the right thread.
Not trying to shoot down CSDs here, they are important, especially when combined with standard FR. It's just that presenting the raw CSD isn't 100% reliable in determining the signature, more like 70% or less. I'd tend to assume that rings are more audible when longer/stronger/wider and mid-band, but I'm not sure there's any published tests that reveal audible thresholds, like what was done for the masking curve research for compression codecs. Someone needs to do a paper on the audibility of ringing components, including magnitude and time duration, bandwidth of the ring, frequency vs audibility, integration with adjacent rings, etc. You'd probably need a fairly ring-less transducer, and simulate ringing digitally so you'd have control and repeatability across test subjects.
OK....go!