Adam,
thank you so much for the review. It is a great piece of work!
Regarding your section about AF.C, and how it compares to the competition (D610), I have two comments i'd like to share. Overall though, this section is a great resource too
1. D610.
It can be shown (I tried it in my papers about equivalence and
true reasons for FF) that a larger sensor using the same AF module as a smaller one produces disproportionally better results. Therefore, a comparison against a D7100 or 7D might have been more fair.
Nevertheless, with the results you obtained and taking my experience with a 70-200/2.8G into account, I assume a D7100 wouldn't have faired much worse with a bycicle. It's just not challenging enough.
2. Methodology.
You follow the same methodology as a majority of testers. By computing the ratio of keepers. Unfortunately, this is a rather meaningless score (except for keeping AF points on the subject with long focals like BIF).
By construction, AF.C fails if things start to happen too quickly with just excellent results on the slow side. Every AF.C. Where "quick" must be considered in terms of magnification ratio change speed (squared), absolute speeds being meaningless to an AF. So, with a strict focus priority policy and a fast burst rate a camera would always score 100% keepers, even if its AF.C is bad. It would just stop shooting if the subject is too near (the magnification ratio change speed is zero at infinite distance and becomes infinite at zero distance, provided the subject approaches at a constant speed).
Therefore, I adopted an alternate interpretation of AF.C test results: given a subject approaches the camera at a defined speed and given a defined field of view and given a defined physical aperture diameter, what is the nearest distance the camera could still snap a sharp photo? Given scene brightness and target contrast. And where sharp is in LW/PH >x. And using a fitted plot of LW/PH over distance for many runs to eliminate the effect of burst rate. The fitted curve will show a sharp drop-off at some distance.
The true score of an AF.C system then is the inverse of this drop-off distance, scaled to a magnification change speed number (Hz). This is a speed measurement and reflects how fast an AF system can update the focus.
So far, the only source allowing for this alternate interpretation is the labo test FNAC documents. It gives mixed results due to them doing a single run only. But with a lot of heuristic interpretation to discuss outliers, it still is my best source to assess the vendors current state-of-the-art.
Therefore, I'd like to know the respective closest distances you still obtained sharp images at. And how you evaluated "sharp".
BTW, with 80 frames at presumed 5fps for 10 to 100 feet, the speed was something like 6.2 km/h, just as fast as fast walking. If correct, the bike was much slower than they use to be in France
But maybe, the #frames are from multiple runs and therefore, each run was a lot faster and more challenging for the AF?