I'd mentioned in the previous update that we're (Xiph is) using a chirp estimation algorithm that we published back in 2007, and that the original paper has precious little space to devote to describing in detail how the algorithm actually performed. One of the upshots of not having done extensive characterization tests of our own algorithm was that it has already surprised me a few times this year (in both good and bad ways).
Therefore, Ghost Update 20110604 concerns itself with describing and graphing algorithm behavior in mind-numbing detail.
Death! By! Graphs!