Pete_W
LMF-Patron
Same for me on the S5II. C-AF works so reliably that I just leave it there.
L-MOUNT Forum
Register a free account now!
If you are registered, you get access to the members only section, can participate in the buy & sell second hand forum and last but not least you can reserve your preferred username before someone else takes it.
Going crazy with my 70-300!
- Thread starter pdk42
- Start date
R-J Sunday
LMF-Patron
Since s5II always on the C-af since s-af is dfd (contrast based) and wil surely misfocus way more than c-af. especially in low contrast scenes.
Quentinquirelino
Active Member
I think that's wrong. Theoretically DFD CDAF should have better accuracy than PDAF. The big advantage of pdaf is, that it can detect direction and estimate distance for focusing. But PDAF can't tell you if something actually is in focus. CDAF based focus on the other hand can tell you if something is in focus, but it can't tell you in what direction and how far an out of focus object is.
That's why PDAF is much better for tracking and CDAF should be better for accurate focusing. That's why also other manufacturers combining both technologies.
An interesting comment by georgehudetz on the DPReview thread that links to this one:
https://www.dpreview.com/forums/post/67586167
Looking through the photos where I've had less than perfect sharpness the AF area has always been on a subject that had some significant variation in depth, such as tree branches or leaves, and the detail within the AF area has been comparatively fine. But that variation should never have been significant enough for the available DOF not to include the subject.
I've certainly seen CDAF fail when detail in the subject approaches the resolution of the sensor; I wonder if this is a similar situation?
https://www.dpreview.com/forums/post/67586167
Looking through the photos where I've had less than perfect sharpness the AF area has always been on a subject that had some significant variation in depth, such as tree branches or leaves, and the detail within the AF area has been comparatively fine. But that variation should never have been significant enough for the available DOF not to include the subject.
I've certainly seen CDAF fail when detail in the subject approaches the resolution of the sensor; I wonder if this is a similar situation?
pdk42
Moderator
That's exactly the problem I've been experiencing.
CDAF algorithms use something called the "focus measure" function to return a numerical value representing the degree of focus in the focus zone. It works by calculating the edge contrast and the idea is that the lens's focus is shifted while the focus function's value is read. When the focus function is at a maxima, the focus movement stops. Clearly, when the detail approaches that of the sensor's pixel pitch there will be no edges and no easy maxima to find. But the subject I used in the focus area had details that were much bigger than the sensor pitch. You can see that in the 100% grabs.
I think the joke was aimed at you Pete.
I thought both DFD CDAD and PDAF could be used simultaneously but I'm only a debutante with S5ii and MILC.
Markuswelder
Active Member
As I understand it, PDAF is used to estimate distance and direction of subject movement ie towards or away from the camera, CDAF does the final tweak for maximum accuracy. DFD is just a means by which the camera looks at the out of focus areas, to determine which direction to drive the focusing elements in, using contrast only.
pdk42
Moderator
That's my understanding too.
Markuswelder
Active Member
There's been lots said regarding DFD & if it does really work, or it just being marketing.
In my experience, without a doubt, it really does work. You've only got to mount a lens from a manufacturer outside of the L mount alliance, & see the difference. Even in single shot mode. You get that classic contrast detect shuffle a couple of times on the initial half press of the shutter, then again when the camera finally locks focus.
With a DFD compatible lens, you don't see that. It drives pretty much straight to focus, perhaps gives a slight shuffle depending on light & subject contrast, and locks. In m4/3 world, users see this, without realising what's going on. Specifically Olympus users trying certain Panasonic lenses. The Lumix 20mm f1.7 comes to mind here, as it's mentioned over and over and over again by Olympus users, as having really bad AF. Pop it on a DFD body, and there's not a lot to complain about. Is it the fastest, or one of the faster focusing lenses out there? Absolutely not. But it's still very very usable and capable on a Panasonic body
In my experience, without a doubt, it really does work. You've only got to mount a lens from a manufacturer outside of the L mount alliance, & see the difference. Even in single shot mode. You get that classic contrast detect shuffle a couple of times on the initial half press of the shutter, then again when the camera finally locks focus.
With a DFD compatible lens, you don't see that. It drives pretty much straight to focus, perhaps gives a slight shuffle depending on light & subject contrast, and locks. In m4/3 world, users see this, without realising what's going on. Specifically Olympus users trying certain Panasonic lenses. The Lumix 20mm f1.7 comes to mind here, as it's mentioned over and over and over again by Olympus users, as having really bad AF. Pop it on a DFD body, and there's not a lot to complain about. Is it the fastest, or one of the faster focusing lenses out there? Absolutely not. But it's still very very usable and capable on a Panasonic body
pdk42
Moderator
I did some more analysis today and the more testing I do I can see that the smallest "1-area" AF zone is about the least reliable focus mode on any of the three S cameras I have (S1R, S5, and S5ii). The focus wandering that I've been posting about with the 70-300 is also slightly visible on the 24-105 at its long end. I think what's going on is that as the focal length increases the required focus precision increases too. I suspect that it follows the same mathematical relationship as the hyperfocal distance - that is, the precision needed will increase with the square of the focal length. This is the reason why it's worse on the 70-300 since it's the longest lens I have.
Increasing the size of the 1-area zone makes a big difference, as does using the "1-area +" mode. But I think that pinpoint mode is the best for SAF. On the S5ii, CAF shows similar problems with a small 1-area mode, but using a larger zone works very well.
Increasing the size of the 1-area zone makes a big difference, as does using the "1-area +" mode. But I think that pinpoint mode is the best for SAF. On the S5ii, CAF shows similar problems with a small 1-area mode, but using a larger zone works very well.
I've been experimenting with the 70-300 as well. One thing I've noticed with the S5 is that when the camera fails to acquire focus it leaves the lens focused just beyond infinity so nothing is sharp.
That's an interesting point. I wonder whether the lens could be limited by the precision with which it can move the focusing elements? My Sigma fp L can be set to display the focus distance in meters or feet when focusing manually and I've noticed the focus distance will jump by progressively larger amounts the closer the lens gets to infinity. For the 45/2.8 (on the camera at the moment) the steps as you approach infinity are 69', 85', 110', 184', 368', ∞.
That's an interesting point. I wonder whether the lens could be limited by the precision with which it can move the focusing elements? My Sigma fp L can be set to display the focus distance in meters or feet when focusing manually and I've noticed the focus distance will jump by progressively larger amounts the closer the lens gets to infinity. For the 45/2.8 (on the camera at the moment) the steps as you approach infinity are 69', 85', 110', 184', 368', ∞.