Upload & Sell: On
| p.2 #1 · p.2 #1 · Dot-Tune: Autofocus Fine Tuning in under 5 minutes |
The issue with AFMA is that the variations in the lens and camera for PDAF and Contrast Detect are high.
Contrast detect is not very accurate, and that green dot means very little.
You need to take about 10 shots setting the focus to mfd or infinity each time and throw out the outlier focus errors and take the average of the rest.
The closer you look at AF accuracy, the uglier it gets.
There are two separate forces at work here; the shot-to-shot variability in the mechanical aspects of the focus system (lens motor, servo, position sensors), and a fixed error bias in the optical aspects of that same system (mirror alignment, mount alignment, AF point alignment, lens abberations, etc..). The only reason existing AFMA techniques require multiple shots to find the fixed error bias is because those techniques rely on the variable output of PDAF focusing, which makes it harder to discern the fixed error bias from the mechanical variability. This is what the technique in the OP avoids.
An analogy is someone trying to find their true weight on a bathroom scale. The person knows his weight varies throughout the day, so he endeavors to weight himself multiple times and average the results. Lets say his weight varies by +/- 3 pounds, with a low of 160lbs and a high of 166lbs. So he concludes his weight is 163lbs. But what happens if the scale he used is out of calibration, with a fixed error of 10lbs under? That means his real range for the day was 170lbs to 176lbs, making his true weight 173lbs instead of 163lbs. Because of the scale's error, the person will not only fail to arrive at the correct average weight, he'll also fail in every single sample of his weighing, since the heaviest the scale will report is 166lbs, yet his actual range is 170lbs to 176lbs.
The variability of the man's weight represents the shot-to-shot mechanical variability std. dev of the PDAF system. The fixed error of the scale represents the fixed error the PD sensing mechanism resulting from optical/machined tolerances of the mount, mirror, AF sensor, and lens-specific abberations not accounted for in the PDAF design.
Now what happens if the man knows the scale is off but wants to find out by how much? And let's say he already knows his true (average) weight is 173lbs, but doesn't yet know that his range of weights throughout the day is 6lbs (+/- 3lbs). He can't just stand on the scale once and calculate what the error of the scale is relative to his true (average) weight, since his weight varies by an amount unknown to him. So instead he has to take measurements throughout the day to establish his range of weight. Only then can he compare his known true (average) weight to what the scale's biased average is, at which time he can calculate the scale's error at 10lbs. This is the same situation of trying to calculate an AF tune value by taking multiple photographs.
The purpose of AF tune is to center the range of shot-to-shot variation over the system's true center (the "0" on the bathroom scale). That way even with the unavoidable variation most of the shots will still be in focus. I've made a graphical depiction of this here.