Posts: 8,004
Threads: 1,851
Joined: Apr 2010
Reputation:
49
07-18-2015, 10:36 PM
(This post was last modified: 07-18-2015, 10:39 PM by Klaus.)
As mentioned I wil retest some lenses - first of all those that I have simply still in my stock.
As simple as it might be the Canon 50/1.8 STM could be a great guidance actually.
But first ... some other mounts.
Posts: 4,031
Threads: 41
Joined: Feb 2013
Reputation:
22
I just looked again to your samples, Klaus. First picture top left - the focus is not on the front of the pointy green tip. The tube rail (?) behind it has more sharpness. So either back-focus or curved field?
I think the review is a little bit unfair - the Canon 24/1.4 L will also struggle on a 50 MP body, I guess. I know, comparison shots are boring to do and don't say much about AF reliability, but for Nikon owners this Sigma might be the better choice.
Posts: 8,004
Threads: 1,851
Joined: Apr 2010
Reputation:
49
I think this just means 'not 100%' in focus.
That picture is more meant to show the bokeh ;-)
Posts: 4,031
Threads: 41
Joined: Feb 2013
Reputation:
22
Ah! Understood.
My question about "curved field" was more based on the problem, that even if the AF-module is perfectly produced and put in the right place. a field curvature would mean that with phase AF the picture will only be in focus in center. And because that tip is far off center.
Posts: 6,715
Threads: 236
Joined: Apr 2010
Reputation:
22
What you see in the image is what the AF sensor will see too... So if there is focus field curvature, it will not impact AF in that way. The camera focusses on what it sees, and what it sees is the curved focus field.
Posts: 8,004
Threads: 1,851
Joined: Apr 2010
Reputation:
49
The DoF is so minimal in that shot and 50mp are so unforgiving that focus errors are inevitable.
Ok, I should have taken more samples there ;-)
Posts: 4,031
Threads: 41
Joined: Feb 2013
Reputation:
22
If field curvature would play no role because the AF gets always a sharp picture there would be no need for AF adjustments. Which I only do for center focus point. Field curvature does play no role when focusing in LiveView because here I always get what I see. The indirect phase detection could struggle very well in the outer focus point. Is there a focus point that far out of center in the Canon?
Talking only for Nikon. Their AF system has improved, I just hear from people who use / used both systems they always felt a little insecure with Nikon.
That leads me to the next question. In the sample shots, are you using LiveView or phase detection AF? My copy of the 24/1.4 was quite a bit off focus out of the box. Since I corrected that it's great, but before it would have been my explanation for focus on the wrong spot.
Field curvature plays no roll for PDAF. It is literally impossible for the AF sensor to sample a different image than the sensor would. If the lens' focal plane is shifted 300 microns at the edge focus point and you focus there, it will focus accurately for that position on the sensor.
AF adjustments correct error on the lens' mount location. While all lenses stop at the same position on the barrel at infinity or past or wherever the stop is, the focus may not reach the sensor correctly. The lens reports it is at position "1" and the camera will tell it to move to position "3800" to focus properly. If this is miscalibrated, 3800 may not truly be in focus. The AF adjustment is to ensure that "3800" is "3800."
Some older models have closer to 24 focus positions. Newer ones have thousands.
Posts: 4,031
Threads: 41
Joined: Feb 2013
Reputation:
22
Two things: I know I have been wrong about some theories in the past and I'm afraid, I will occasionally be wrong in the future. Making mistakes is part of evolution.
The other thing is: I'm a customer of a manufacturer who happens to be the same way. Nikon had some issues in the past and I often thought why didn't they test it, thought of it?* One of them was a misaligned AF module in D800, which has nothing to do with field curvature but made very clear, how fragile this whole indirect phase AF model is with HR sensors and lenses used wide open. Even with no movement of camera and target, I don't get 100% accuracy of each attempt to focus - and that is for the center AF, which usually is the best place.
Klaus hasn't replied to my question, wether he used PDAF or CAF (or no AF at all?) It doesn't matter much. The sample clearly shows "take care with focusing when using a lens wide open in front of a 50 MP sensor" Nikon had to learn that, and I don't know if Canon got it right from the beginning and has a 100% reliable assembly line. Also, I've the impression there are more "issue-seekers" on the Nikon side, less so Canon users.
Just checked quickly, about 11% of my pictures I took at aperture f/1.8 or wider open. Don't know about your rate, guys. Of course that doesn't affect if you're right or wrong. I think you're referring to the theory and I was mislead with the field curvature idea. I just often found real life doesn't care much about theoretical assumptions. So if it's not field curvature, there are still other possibilities to miss focus far out of the image's center.
* One answer could be "otherwise they wouldn't have come out with 36 MP 3 years ago". Extensive testing only includes the scenarios the tester would found to be critical. All users together ask for much more than a couple of testers could think of.
