Z8 Subject Detection Inaccurate in Low (but not that low) light

[SCoombs]

This is a topic that has been discussed before but I've been struggling with it and experimenting with it the last few days so I wanted to throw some of my findings out there and see if others have any thoughts.

Basically, I have found the AF system of the Z8 (and presumably Z9) is pretty good even in low light unless you use subject detection, in which case it's awful. Now a lot of people have just reported having poor success in general with AF in low light, but after a lot of experimenting I am convinced it's not the AF but specifically the subject detection.

To define the two key ideas here I'll say that first, by low light I don't mean extremely low light only, but even just sort of or what I might call "mildly" low light... the kind you might get in a relatively averagely lit high school gymnasium for instance.

Second, by "inaccurate" I don't mean the subject detection has trouble locating subjects/eyes or that the subject detect bounces around without firmly committing to something, but that it will instantly recognize a subject and confidently put the box on the eye, but the actual focus will be several inches to the front or the back. This is AF-C, by the way, not AF-S where the subject is just moving after focus.

I am finding that in even just mildly or moderately low light, the subject detection will grab an eye instantly but I will only get about 30% of photos in focus. That's a 70% miss rate, and usually it's missing by several inches from where the focus point is located.

But it's subject detection - not the AF overall. In these same situations if I use single point or dynamic area or even just turn subject detection off and use the standard wide area modes, I get accurately focused shots about 75-80% of the time, which I'd consider a very reasonable hit rate in poor light.

If I swap into AF-S and use single point, the hit rate goes up even more, and if I swap into AF-S and turn on the built in AF assist illuminator - that little green light - the camera is almost perfect. I basically get 100% of shots in perfect, tack sharp or nearly tack sharp focus.

But subject detect in AF-S is also bad. It's actually better than AF-C, even though in AF-S theoretically the subject can move and throw the focus off.

A few examples. I could probably post hundreds of these, but a few suffice.

Subject detect claims focus on the eye. It's really on the pizza several inches in front.







Same exact lighting and everything else, but using dynamic area (a non subject detect mode), we get perfect focus:











Subject detection claims focus on the front eye, it's really on the back eye 1-2 inches behind:









Swap to dynamic area and focus is basically perfect:







I could post these all day.

To summarize my findings after hundreds of shots experimenting over a few days:

In conditions of decreasing light, but before lighting would be considered extremely low,

1. In AF-S with the built in illuminator focus accuracy for all modes is nearly 100%
2. In AF-S with a non-subject detect mode and the AF illuminator turned off, focus accuracy is still extremely high
3. In AF-S with a subject detect mode and the AF illuminator turned off, focus accuracy is relatively poor. <50% of shots have a focus plane matching the AF system's reported focus point
4. In AF-C with a non-subject detect mode about 75-80% of shots have a focus plane matching what the AF system reports.
5. In AF-C with a subject detect mode about 30% of shots have a focus plane matching what the AF system reports. The other 70% are several inches front or back focused.
6. Activating Starlight mode or turning setting d9 to "adjust for east of viewing" improves results up to the levels reported above. Without one of these two settings activated focus accuracy is worse than reported.

I think the system, in terms of its "dumb" focus acquisition, really is a great performer as I'd even consider the 75% in-focus rate for low light focus like I see with non-subject detect modes to be pretty good. The subject detection, though, has performed extremely poorly for me in low light = really just abysmally poorly - and what's perplexing is that it isn't as if this is because it can't locate subjects or identify eyes: it can not only do that but it can do it extremely well as it identifies them almost instantly even in poor conditions. The problem is that the focus is rarely actually on the thing the subject detection identifies as the focus point. It's doubly perplexing because if I just manually put a focus point on the exact same spot in the exact same conditions, the system generally nails the focus. It's only when that focus point has been placed there by the subject detection system that it somehow struggles to actually focus there.

 

[warrior-kid]

I probably should try and compile some stats from my Z9 photo collection. I can see very frequently a small focus box on the eye but the photo is a blur as far as the subject is concerned. However: having taken a couple of knocks, my Z9 is now sent for repairs, we shall see what they can find.

 

[A Owens]

This is a topic that has been discussed before but I've been struggling with it and experimenting with it the last few days so I wanted to throw some of my findings out there and see if others have any thoughts.

Basically, I have found the AF system of the Z8 (and presumably Z9) is pretty good even in low light unless you use subject detection, in which case it's awful. Now a lot of people have just reported having poor success in general with AF in low light, but after a lot of experimenting I am convinced it's not the AF but specifically the subject detection.

To define the two key ideas here I'll say that first, by low light I don't mean extremely low light only, but even just sort of or what I might call "mildly" low light... the kind you might get in a relatively averagely lit high school gymnasium for instance.

Second, by "inaccurate" I don't mean the subject detection has trouble locating subjects/eyes or that the subject detect bounces around without firmly committing to something, but that it will instantly recognize a subject and confidently put the box on the eye, but the actual focus will be several inches to the front or the back. This is AF-C, by the way, not AF-S where the subject is just moving after focus.

I am finding that in even just mildly or moderately low light, the subject detection will grab an eye instantly but I will only get about 30% of photos in focus. That's a 70% miss rate, and usually it's missing by several inches from where the focus point is located.

But it's subject detection - not the AF overall. In these same situations if I use single point or dynamic area or even just turn subject detection off and use the standard wide area modes, I get accurately focused shots about 75-80% of the time, which I'd consider a very reasonable hit rate in poor light.

If I swap into AF-S and use single point, the hit rate goes up even more, and if I swap into AF-S and turn on the built in AF assist illuminator - that little green light - the camera is almost perfect. I basically get 100% of shots in perfect, tack sharp or nearly tack sharp focus.
But subject detect in AF-S is also bad. It's actually better than AF-C, even though in AF-S theoretically the subject can move and throw the focus off.

A few examples. I could probably post hundreds of these, but a few suffice.

Subject detect claims focus on the eye. It's really on the pizza several inches in front.





Same exact lighting and everything else, but using dynamic area (a non subject detect mode), we get perfect focus:





Subject detection claims focus on the front eye, it's really on the back eye 1-2 inches behind:





Swap to dynamic area and focus is basically perfect:





I could post these all day.

To summarize my findings after hundreds of shots experimenting over a few days:

In conditions of decreasing light, but before lighting would be considered extremely low,

1. In AF-S with the built in illuminator focus accuracy for all modes is nearly 100%
2. In AF-S with a non-subject detect mode and the AF illuminator turned off, focus accuracy is still extremely high
3. In AF-S with a subject detect mode and the AF illuminator turned off, focus accuracy is relatively poor. <50% of shots have a focus plane matching the AF system's reported focus point
4. In AF-C with a non-subject detect mode about 75-80% of shots have a focus plane matching what the AF system reports.
5. In AF-C with a subject detect mode about 30% of shots have a focus plane matching what the AF system reports. The other 70% are several inches front or back focused.
6. Activating Starlight mode or turning setting d9 to "adjust for east of viewing" improves results up to the levels reported above. Without one of these two settings activated focus accuracy is worse than reported.

I think the system, in terms of its "dumb" focus acquisition, really is a great performer as I'd even consider the 75% in-focus rate for low light focus like I see with non-subject detect modes to be pretty good. The subject detection, though, has performed extremely poorly for me in low light = really just abysmally poorly - and what's perplexing is that it isn't as if this is because it can't locate subjects or identify eyes: it can not only do that but it can do it extremely well as it identifies them almost instantly even in poor conditions. The problem is that the focus is rarely actually on the thing the subject detection identifies as the focus point. It's doubly perplexing because if I just manually put a focus point on the exact same spot in the exact same conditions, the system generally nails the focus. It's only when that focus point has been placed there by the subject detection system that it somehow struggles to actually focus there.

That's pretty frustrating.

What focus mode are you using with subject detection, what VR mode and are you shooting a burst for each shot?

 

[Droster]

That is consistent with my results from my two weekends of taking portraits with the Z9. Using the same subject detection in AF-C, in a burst of 14~15 frames, I would usually only get two or three in critical focus. I would focus first before hitting the shutter. Usually it's the first if not second frame in focus, then it will guess and hunt for a bit, before finding focus back again at near the last of the burst. There are two or three faces (out of the hundreds) that the camera just wouldn't focus, and I can see the camera hunting while shooting, then I stopped and went into single point.

Yea for portraits, do not use subject detection. Just use single point and tell the camera where to focus, don't let the camera guess. Subject detection is good when the subject occupies the whole frame and in action. These are cases where critical focus isn't that important.

 

[Models By Day]

Who doesn't love Pizza.

Pizza overrides Eye

Pizza Priority focus in the latest update I guess

 

[SCoombs]

This is a topic that has been discussed before but I've been struggling with it and experimenting with it the last few days so I wanted to throw some of my findings out there and see if others have any thoughts.

Basically, I have found the AF system of the Z8 (and presumably Z9) is pretty good even in low light unless you use subject detection, in which case it's awful. Now a lot of people have just reported having poor success in general with AF in low light, but after a lot of experimenting I am convinced it's not the AF but specifically the subject detection.

To define the two key ideas here I'll say that first, by low light I don't mean extremely low light only, but even just sort of or what I might call "mildly" low light... the kind you might get in a relatively averagely lit high school gymnasium for instance.

Second, by "inaccurate" I don't mean the subject detection has trouble locating subjects/eyes or that the subject detect bounces around without firmly committing to something, but that it will instantly recognize a subject and confidently put the box on the eye, but the actual focus will be several inches to the front or the back. This is AF-C, by the way, not AF-S where the subject is just moving after focus.

I am finding that in even just mildly or moderately low light, the subject detection will grab an eye instantly but I will only get about 30% of photos in focus. That's a 70% miss rate, and usually it's missing by several inches from where the focus point is located.

But it's subject detection - not the AF overall. In these same situations if I use single point or dynamic area or even just turn subject detection off and use the standard wide area modes, I get accurately focused shots about 75-80% of the time, which I'd consider a very reasonable hit rate in poor light.

If I swap into AF-S and use single point, the hit rate goes up even more, and if I swap into AF-S and turn on the built in AF assist illuminator - that little green light - the camera is almost perfect. I basically get 100% of shots in perfect, tack sharp or nearly tack sharp focus.
But subject detect in AF-S is also bad. It's actually better than AF-C, even though in AF-S theoretically the subject can move and throw the focus off.

A few examples. I could probably post hundreds of these, but a few suffice.

Subject detect claims focus on the eye. It's really on the pizza several inches in front.





Same exact lighting and everything else, but using dynamic area (a non subject detect mode), we get perfect focus:





Subject detection claims focus on the front eye, it's really on the back eye 1-2 inches behind:





Swap to dynamic area and focus is basically perfect:





I could post these all day.

To summarize my findings after hundreds of shots experimenting over a few days:

In conditions of decreasing light, but before lighting would be considered extremely low,

1. In AF-S with the built in illuminator focus accuracy for all modes is nearly 100%
2. In AF-S with a non-subject detect mode and the AF illuminator turned off, focus accuracy is still extremely high
3. In AF-S with a subject detect mode and the AF illuminator turned off, focus accuracy is relatively poor. <50% of shots have a focus plane matching the AF system's reported focus point
4. In AF-C with a non-subject detect mode about 75-80% of shots have a focus plane matching what the AF system reports.
5. In AF-C with a subject detect mode about 30% of shots have a focus plane matching what the AF system reports. The other 70% are several inches front or back focused.
6. Activating Starlight mode or turning setting d9 to "adjust for east of viewing" improves results up to the levels reported above. Without one of these two settings activated focus accuracy is worse than reported.

I think the system, in terms of its "dumb" focus acquisition, really is a great performer as I'd even consider the 75% in-focus rate for low light focus like I see with non-subject detect modes to be pretty good. The subject detection, though, has performed extremely poorly for me in low light = really just abysmally poorly - and what's perplexing is that it isn't as if this is because it can't locate subjects or identify eyes: it can not only do that but it can do it extremely well as it identifies them almost instantly even in poor conditions. The problem is that the focus is rarely actually on the thing the subject detection identifies as the focus point. It's doubly perplexing because if I just manually put a focus point on the exact same spot in the exact same conditions, the system generally nails the focus. It's only when that focus point has been placed there by the subject detection system that it somehow struggles to actually focus there.

That's pretty frustrating.

What focus mode are you using with subject detection, what VR mode and are you shooting a burst for each shot?

I'm usually in auto-area AF because through lots of testing I found it to get in focus shots a bit more than the other wide area modes, but suffice it to say the success rate is about the same using any subject detect mode: wide area small, wide area large, wide area custom, 3d tracking, auto area af.

I usually shoot in burst BUT the primary time I encounter these problems is when I'm needing to use a flash and so I can't really shoot in bursts. I do try to take sequences of four or five shots letting the flash recycle in between each.

 

[SCoombs]

Another thought as some people don't report the same issues:

I have found from my testing that it seems as if the system is more likely to have accurate focus on the initial acquisition than after the AF-C tries to maintain focus, meaning that if I let the shutter button activate focus and shoot that way I get more hits than if I use my normal approach of back button focus. If I use back button but try to hit the shutter the *instant* the system gives me a green box, I also have better success.

This is also consistent with the finding that in AF-S subject detect seems to work better.

It's as if the initial acquisition is generally good and then the system immediately starts moving around.

So if you're not usually using back button focus, you may have noticed this less.

 

[Horshack]

It's doubly perplexing because if I just manually put a focus point on the exact same spot in the exact same conditions, the system generally nails the focus. It's only when that focus point has been placed there by the subject detection system that it somehow struggles to actually focus there.

First, well-done on the well-thought experiments and detailed observations you've provided so far.

To your question above, it's helpful to visualize how the AF is likely working for the subject detection case to come up with theories about why it may be underprforming in your scenario. Subject detection works in image space, with the LV stream off the sensor feeding into the high-level subject recognition logic in the camera. Once a subject has been recognized, it can then be tracked using a combination of ongoing image recognition in combination with the phase-detect information read from the on-sensor PD pixels. The camera will have to correlate the location in image space with the specific PD pixels on the sensors to know which PD information to evaluate and track.

In your lowish-light scenario it appears the camera is still successfully recognizing and tracking the subject, if we rely on the on-screen bounding box around the subject. However to get an actual in-focus shot the camera has to utilize its traditional PD acquisition and focus movement, which again relies on the camera correctly correlating the subject location to the PD pixels and driving the lens correctly based on that PD info. The fact you're getting a backfocused shot indicates that correlation is either using the wrong PD pixels or using them at the wrong time, ie there's a time shift between the subject evaluation and the PD evaluation. The fact the issue manifests as light decreases but only for the subject recognition case (and not the "dumb" non-subject AF cases) tells me the problem is likely on the subject recognition side, specifically that there's a delay or lag in continued recognition of the subject relative to the time that information is fed into the PD evaluation logic.

This is just me thinking out loud...no conclusions yet.

 

[zephxiiii]

Another thought as some people don't report the same issues:

I have found from my testing that it seems as if the system is more likely to have accurate focus on the initial acquisition than after the AF-C tries to maintain focus, meaning that if I let the shutter button activate focus and shoot that way I get more hits than if I use my normal approach of back button focus. If I use back button but try to hit the shutter the *instant* the system gives me a green box, I also have better success.

This is also consistent with the finding that in AF-S subject detect seems to work better.

It's as if the initial acquisition is generally good and then the system immediately starts moving around.

So if you're not usually using back button focus, you may have noticed this less.

This is what I found with my Z6 as well. Don't sit there holding down the AF button all day(in C mode), trigger focus when you want to take shots then get off of it.

I swear I saw some YT with regards to wild life photography say something similar.

 

[SCoombs]

It's doubly perplexing because if I just manually put a focus point on the exact same spot in the exact same conditions, the system generally nails the focus. It's only when that focus point has been placed there by the subject detection system that it somehow struggles to actually focus there.

First, well-done on the well-thought experiments and detailed observations you've provided so far.

To your question above, it's helpful to visualize how the AF is likely working for the subject detection case to come up with theories about why it may be underprforming in your scenario. Subject detection works in image space, with the LV stream off the sensor feeding into the high-level subject recognition logic in the camera. Once a subject has been recognized, it can then be tracked using a combination of ongoing image recognition in combination with the phase-detect information read from the on-sensor PD pixels. The camera will have to correlate the location in image space with the specific PD pixels on the sensors to know which PD information to evaluate and track.

In your lowish-light scenario it appears the camera is still successfully recognizing and tracking the subject, if we rely on the on-screen bounding box around the subject. However to get an actual in-focus shot the camera has to utilize its traditional PD acquisition and focus movement, which again relies on the camera correctly correlating the subject location to the PD pixels and driving the lens correctly based on that PD info. The fact you're getting a backfocused shot indicates that correlation is either using the wrong PD pixels or using them at the wrong time, ie there's a time shift between the subject evaluation and the PD evaluation. The fact the issue manifests as light decreases but only for the subject recognition case (and not the "dumb" non-subject AF cases) tells me the problem is likely on the subject recognition side, specifically that there's a delay or lag in continued recognition of the subject relative to the time that information is fed into the PD evaluation logic.

This is just me thinking out loud...no conclusions yet.

Interesting thoughts.

It seems most people here are reporting similar experiences. I also posted on FM and over there some have said likewise but a lot of people are saying their cameras perform perfectly in low light.

I wonder what the difference is.

Could it be variations in the quality of the bin of sensors in one camera bs another, sortof like how two identical CPU models may have different potentials for overocking due to differences in quality of that particular die?

Could it be the difference between some people putting more display information on the screen/viewfinder? Is displaying the histogram, for instance, taking enough away from the processor that it can't process AF data fast enough?

I'm trying to think of ideas here.

That said, if there's a delay in the continued recognition of the subject I'd think this would be more a problem when the subject moves whereas if the subject (and camera) are still between AF updates then one wouldn't expect any issues since the AF wouldn't need to change. But rather, what I notice is that the AF in low light is constantly updating and changing even if the subject distance isn't moving, meaning it's constantly proactively moving and moving the focus OFF of the subject. How would a lag cause this issue

 

[SCoombs]

Another thought as some people don't report the same issues:

I have found from my testing that it seems as if the system is more likely to have accurate focus on the initial acquisition than after the AF-C tries to maintain focus, meaning that if I let the shutter button activate focus and shoot that way I get more hits than if I use my normal approach of back button focus. If I use back button but try to hit the shutter the *instant* the system gives me a green box, I also have better success.

This is also consistent with the finding that in AF-S subject detect seems to work better.

It's as if the initial acquisition is generally good and then the system immediately starts moving around.

So if you're not usually using back button focus, you may have noticed this less.

This is what I found with my Z6 as well. Don't sit there holding down the AF button all day(in C mode), trigger focus when you want to take shots then get off of it.

I swear I saw some YT with regards to wild life photography say something similar.

The thing is, if you follow this norm then what even is the point of AF-C? The whole reason it exists is to photograph subjects which are moving and may need to be tracked with the AF held on.

 

[ghostfox_1]

Another thought as some people don't report the same issues:

I have found from my testing that it seems as if the system is more likely to have accurate focus on the initial acquisition than after the AF-C tries to maintain focus, meaning that if I let the shutter button activate focus and shoot that way I get more hits than if I use my normal approach of back button focus. If I use back button but try to hit the shutter the *instant* the system gives me a green box, I also have better success.

This is also consistent with the finding that in AF-S subject detect seems to work better.

It's as if the initial acquisition is generally good and then the system immediately starts moving around.

So if you're not usually using back button focus, you may have noticed this less.

This is what I found with my Z6 as well. Don't sit there holding down the AF button all day(in C mode), trigger focus when you want to take shots then get off of it.

I swear I saw some YT with regards to wild life photography say something similar.

The thing is, if you follow this norm then what even is the point of AF-C? The whole reason it exists is to photograph subjects which are moving and may need to be tracked with the AF held on.

Because you use it with moving subjects.

Portraits are not where I'd use afc

 

[zephxiiii]

Another thought as some people don't report the same issues:

I have found from my testing that it seems as if the system is more likely to have accurate focus on the initial acquisition than after the AF-C tries to maintain focus, meaning that if I let the shutter button activate focus and shoot that way I get more hits than if I use my normal approach of back button focus. If I use back button but try to hit the shutter the *instant* the system gives me a green box, I also have better success.

This is also consistent with the finding that in AF-S subject detect seems to work better.

It's as if the initial acquisition is generally good and then the system immediately starts moving around.

So if you're not usually using back button focus, you may have noticed this less.

This is what I found with my Z6 as well. Don't sit there holding down the AF button all day(in C mode), trigger focus when you want to take shots then get off of it.

I swear I saw some YT with regards to wild life photography say something similar.

The thing is, if you follow this norm then what even is the point of AF-C? The whole reason it exists is to photograph subjects which are moving and may need to be tracked with the AF held on.

Why have it go AF hunting when it doesn't need to? Seems like the longer you hold it the more it wants to go looking for something else. Not sure if that's what's going on but maybe it works best right after that initial call. That's what I noticed when tracking things with a Z6.

Other thing to do is restrict the area it can look in (wide area) which I believe we already know helps with subject detection.

 

[SCoombs]

Another thought as some people don't report the same issues:

I have found from my testing that it seems as if the system is more likely to have accurate focus on the initial acquisition than after the AF-C tries to maintain focus, meaning that if I let the shutter button activate focus and shoot that way I get more hits than if I use my normal approach of back button focus. If I use back button but try to hit the shutter the *instant* the system gives me a green box, I also have better success.

This is also consistent with the finding that in AF-S subject detect seems to work better.

It's as if the initial acquisition is generally good and then the system immediately starts moving around.

So if you're not usually using back button focus, you may have noticed this less.

This is what I found with my Z6 as well. Don't sit there holding down the AF button all day(in C mode), trigger focus when you want to take shots then get off of it.

I swear I saw some YT with regards to wild life photography say something similar.

The thing is, if you follow this norm then what even is the point of AF-C? The whole reason it exists is to photograph subjects which are moving and may need to be tracked with the AF held on.

Because you use it with moving subjects.

Portraits are not where I'd use afc

Ok sure, generally. Yet sometimes you would. For instance, if trying to doba portrait of a small child AF-C is probably an important tool.

Regardless, the point is that the suggestion was to use AF-C *like* AF-S and NOT track movement with it, in which case why nearest we using AF-C for that? Put differently, if the only solution to lOw light AF s to use AF-S or to use AF-C like it's AF-S it means there is no solution for large areas like event photography.

 

[SCoombs]

Another thought as some people don't report the same issues:

I have found from my testing that it seems as if the system is more likely to have accurate focus on the initial acquisition than after the AF-C tries to maintain focus, meaning that if I let the shutter button activate focus and shoot that way I get more hits than if I use my normal approach of back button focus. If I use back button but try to hit the shutter the *instant* the system gives me a green box, I also have better success.

This is also consistent with the finding that in AF-S subject detect seems to work better.

It's as if the initial acquisition is generally good and then the system immediately starts moving around.

So if you're not usually using back button focus, you may have noticed this less.

This is what I found with my Z6 as well. Don't sit there holding down the AF button all day(in C mode), trigger focus when you want to take shots then get off of it.

I swear I saw some YT with regards to wild life photography say something similar.

The thing is, if you follow this norm then what even is the point of AF-C? The whole reason it exists is to photograph subjects which are moving and may need to be tracked with the AF held on.

Why have it go AF hunting when it doesn't need to? Seems like the longer you hold it the more it wants to go looking for something else. Not sure if that's what's going on but maybe it works best right after that initial call. That's what I noticed when tracking things with a Z6.

Other thing to do is restrict the area it can look in (wide area) which I believe we already know helps with subject detection.

Well fundamentally this is because with some subjects, especially in event photography, you do need to track focus.

As for restricting the area, as I said I've tested it and found that there's little differenc between performance in different sized areas and if anything the full area is slightly *more* reliable than the smaller area modes. For the first few years in the Z system I used primarily the two wide area modes and never used full area. Then when I started shooting in more low light situations and started to notice these AF problems and starter experimenting I found a definite increase in the rate of in focus shots when using Full Area over the smaller area modes.

In the wildlife photography community, there are also very widespread reports that people are finding the full area to be the most reliable mode for the bird detection.

 

[ghostfox_1]

Another thought as some people don't report the same issues:

I have found from my testing that it seems as if the system is more likely to have accurate focus on the initial acquisition than after the AF-C tries to maintain focus, meaning that if I let the shutter button activate focus and shoot that way I get more hits than if I use my normal approach of back button focus. If I use back button but try to hit the shutter the *instant* the system gives me a green box, I also have better success.

This is also consistent with the finding that in AF-S subject detect seems to work better.

It's as if the initial acquisition is generally good and then the system immediately starts moving around.

So if you're not usually using back button focus, you may have noticed this less.

This is what I found with my Z6 as well. Don't sit there holding down the AF button all day(in C mode), trigger focus when you want to take shots then get off of it.

I swear I saw some YT with regards to wild life photography say something similar.

The thing is, if you follow this norm then what even is the point of AF-C? The whole reason it exists is to photograph subjects which are moving and may need to be tracked with the AF held on.

Because you use it with moving subjects.

Portraits are not where I'd use afc

Ok sure, generally. Yet sometimes you would. For instance, if trying to doba portrait of a small child AF-C is probably an important tool.

Regardless, the point is that the suggestion was to use AF-C *like* AF-S and NOT track movement with it, in which case why nearest we using AF-C for that? Put differently, if the only solution to lOw light AF s to use AF-S or to use AF-C like it's AF-S it means there is no solution for large areas like event photography.

You use afc when you deal with moving subjects. Event photography is probably that situation. Sitting portraits is not.

I can't explain why people have wildly different experiences with their cameras. Some of it is absolutely user error. Some is basically impossible to nail down because we can't all sit in a room and objectively test this stuff to death.

A small child moving probably would be fine for afc as you expect movement

 

[SCoombs]

Another thought as some people don't report the same issues:

I have found from my testing that it seems as if the system is more likely to have accurate focus on the initial acquisition than after the AF-C tries to maintain focus, meaning that if I let the shutter button activate focus and shoot that way I get more hits than if I use my normal approach of back button focus. If I use back button but try to hit the shutter the *instant* the system gives me a green box, I also have better success.

This is also consistent with the finding that in AF-S subject detect seems to work better.

It's as if the initial acquisition is generally good and then the system immediately starts moving around.

So if you're not usually using back button focus, you may have noticed this less.

This is what I found with my Z6 as well. Don't sit there holding down the AF button all day(in C mode), trigger focus when you want to take shots then get off of it.

I swear I saw some YT with regards to wild life photography say something similar.

The thing is, if you follow this norm then what even is the point of AF-C? The whole reason it exists is to photograph subjects which are moving and may need to be tracked with the AF held on.

Because you use it with moving subjects.

Portraits are not where I'd use afc

Ok sure, generally. Yet sometimes you would. For instance, if trying to doba portrait of a small child AF-C is probably an important tool.

Regardless, the point is that the suggestion was to use AF-C *like* AF-S and NOT track movement with it, in which case why nearest we using AF-C for that? Put differently, if the only solution to lOw light AF s to use AF-S or to use AF-C like it's AF-S it means there is no solution for large areas like event photography.

You use afc when you deal with moving subjects. Event photography is probably that situation. Sitting portraits is not.

I can't explain why people have wildly different experiences with their cameras. Some of it is absolutely user error. Some is basically impossible to nail down because we can't all sit in a room and objectively test this stuff to death.

A small child moving probably would be fine for afc as you expect movement

I don't intend to sound snarky when I say, what user error could possibly explain this? We're talking about a system where the user presses a button and the camera puts the focus on the right spot without the user doing anything else.

So very seriously: how could we be making an error with pushing a button? I'd genuinely love suggestions here so I can try to look into this more.

 

[ghostfox_1]

Another thought as some people don't report the same issues:

I have found from my testing that it seems as if the system is more likely to have accurate focus on the initial acquisition than after the AF-C tries to maintain focus, meaning that if I let the shutter button activate focus and shoot that way I get more hits than if I use my normal approach of back button focus. If I use back button but try to hit the shutter the *instant* the system gives me a green box, I also have better success.

This is also consistent with the finding that in AF-S subject detect seems to work better.

It's as if the initial acquisition is generally good and then the system immediately starts moving around.

So if you're not usually using back button focus, you may have noticed this less.

This is what I found with my Z6 as well. Don't sit there holding down the AF button all day(in C mode), trigger focus when you want to take shots then get off of it.

I swear I saw some YT with regards to wild life photography say something similar.

The thing is, if you follow this norm then what even is the point of AF-C? The whole reason it exists is to photograph subjects which are moving and may need to be tracked with the AF held on.

Because you use it with moving subjects.

Portraits are not where I'd use afc

Ok sure, generally. Yet sometimes you would. For instance, if trying to doba portrait of a small child AF-C is probably an important tool.

Regardless, the point is that the suggestion was to use AF-C *like* AF-S and NOT track movement with it, in which case why nearest we using AF-C for that? Put differently, if the only solution to lOw light AF s to use AF-S or to use AF-C like it's AF-S it means there is no solution for large areas like event photography.

You use afc when you deal with moving subjects. Event photography is probably that situation. Sitting portraits is not.

I can't explain why people have wildly different experiences with their cameras. Some of it is absolutely user error. Some is basically impossible to nail down because we can't all sit in a room and objectively test this stuff to death.

A small child moving probably would be fine for afc as you expect movement

I don't intend to sound snarky when I say, what user error could possibly explain this? We're talking about a system where the user presses a button and the camera puts the focus on the right spot without the user doing anything else.

I'm not talking about this specific instance, though I'd have gone with afs instead. So maybe a bit, but I don't have a flash to mess around with and do objective testing.

So very seriously: how could we be making an error with pushing a button? I'd genuinely love suggestions here so I can try to look into this more.

Besides using afs, we need more testing to get to the how/why of this. Though really, afc should constantly verify focus. Depending on mode/etc it might have csp, or might have misfocused for a single frame (or even more) as it verifies focus on a subject that isn't moving. It's possible it over corrects, or it's an issue with the machine learning, or something higher contrast grabbed it, etc etc etc.

 

[LTheWafflePWN]

I haven't noticed this as something across the board for all subjects in my usage, in any/all levels of light.

However, I definitely have noticed this quite often when using subject detection on smaller children (no older than 4 years old). I don't do babies/toddlers often, but when I do this has been something I've noticed in post, and I've been paying more attention now.

I'm somewhat confident that it doesn't affect adults as much/at all, because in Wide Area S, subject/eye detection on, it'll focus on the parents eyes correctly during lifestyle shoots. Their photos appear to be sharp and in focus on the eyes as expected. Moving the box over to the child's eyes, not as much. I've started using Dynamic Area S instead for really young kids.

 

[SCoombs]

I haven't noticed this as something across the board for all subjects in my usage, in any/all levels of light.

However, I definitely have noticed this quite often when using subject detection on smaller children (no older than 4 years old). I don't do babies/toddlers often, but when I do this has been something I've noticed in post, and I've been paying more attention now.

I'm somewhat confident that it doesn't affect adults as much/at all, because in Wide Area S, subject/eye detection on, it'll focus on the parents eyes correctly during lifestyle shoots. Their photos appear to be sharp and in focus on the eyes as expected. Moving the box over to the child's eyes, not as much. I've started using Dynamic Area S instead for really young kids.

Hmm, I'll have to teat more on this point. Most of my subjects tend to be kids, but several up over the age of 7 or 8 where I do observe it as well.

In testing on my wife I also found the problem to happen a lot, but this testing was primarily a few months ago with the 40mm f2 which is not as strong a lens, and I've seen it with other adults.

The other issue is my wife wears glasses and I have frankly never found the eye detect to get anyone's actual eye through glasses.