Multi-CAM 20k AF system behavior (D5/D500)

[Ilkka Nissilä]

I normally use single point AF or group area AF in photographing moving subjects, and have not generally found the dynamic area AF modes (9, 21, 51 points in Multi-CAM 3500; 9, 25, 72 and 153 points in D5) to do what I want the camera to do. I notice that Nikon recommends the use of the 21/25 point dynamic mode in quite many sports photography scenarios and I also have read that many people use this mode for bird in flight photography.

I was shooting some cross-country skiing in Lahti a couple of weeks ago and tried single point, group-area AF, and 25 point dynamic in the D5 with varying success. The dynamic area AF mode seemed to pick background detail quickly if I was not able to keep the selected center point on the subject and so I got some out of focus shots that I hadn't expected. I had assumed the camera cleverly keeps the subject in focus even if the main point isn't quite on the subject at all times, but this wasn't happening to my satisfaction. Later I tried single point and felt I wasn't able to hold my camera on the action precisely enough with that, and then went to my old favorite since 2014, the group area AF mode. With that mode almost all my photos were correctly focused.

At home I then searched Nikon's tips on the use of the AF area modes. It basically states that the center point is the primary area focused on in dynamic area AF modes and if the subject briefly leaves that focus point, the surrounding points are used by the camera to focus. However, in practice I found that this brief moment is really brief on the D5; if I turn the camera away from the main subject so that the main point is just outside the subject but the surrounding points in the dynamic area are still on the subject, the camera almost immediately refocuses on the background or subject which now is under the main focus point. With the D810 the dynamic area AF modes behaved differently and took quite a while before the focus was reset to the new subject under the main subject (several seconds certainly).

I'm trying to figure out what kind of use scenario is benefiting from the new behavior of dynamic area AF modes. I feel the system is too eager to focus on the background if the main point leaves the subject for a short time.

With group area AF mode, the subject is focused on if even one of the focus points in the group (4 points marked in the viewfinder, but I think all the invisible in-between AF points are used as well) is on or close to the subject. I find that this mode is highly reliable in tracking the subject and it typically does what I need when I'm not able to hold a single AF point on the subject steadily enough. However, sometimes I would like a larger group area such as 25 points without the high sensitivity of the dynamic area AF to the main focus point in the center of the dynamic area. Furthermore the dynamic area AF mode allows the main focus point to be in the top row which crops the dynamic area allowing me to place the subject more off center than I would be able to do with group area AF mode (which is always full, not cropped, so the most the group can be off center in the short dimension of the frame is by one user selectable row). So for this reason I would like to use a larger and croppable group. But that's not how Nikon implemented it.

Then there is the whizbang 3D tracking which uses subject recognition to keep the initially selected subject in focus throughout the frame. It works quite well actually but the more automation the less control the user has.

I think I will try having group area AF in vertical orientation (because the AF sensor array extends quite far off center in the long dimension of the frame) and 25-point or 72-point dynamic area AF in horizontal orientation so I can get the main point as far off center (for faces!) as possible while still giving some support to my limited pointing accuracy on a fast moving subject.

I guess my question is how do you select the AF area mode in your action photography and have you noticed a difference in the behavior of the dynamic area AF modes in the Multi-CAM 3500 and Multi-CAM 20k?

How would you improve the AF system; what kind of features would like to see in the future?

[Erik Lund]

The optical quality of the optics in the AF sensor module could be enhanced, better glass and coatings, I believe it is plastic now.,,, would have a huge impact on reliability and precision, and an extension in size so better than 2.8 would be possible would help 1.4 lenses greatly, f/2 as well,,, larger sub mirror,,,
Marianne Oelund has quite interesting observations reverse engineering the module,,, on DPR.
She also states that it is highly likely that an extra check and fine adjust is happening in the new/later cameras for AF-S,,,

[chris dees]

There's a large thread going on at http://www.fredmiranda.com/forum/topic/1478115 about the difference in behaviour of the D5/D500 AF and the older Nikon's

[Frank Fremerey]

For me it was really helpful to read Thom Hogan's D500 book. Esp on the confusing subject of how changing the AF mode will change the exposure measuring result or not.

I am far from claiming I have really understood all the modes and the interconnections, but it was quite revealing and after halfways digesting it I feel that I blow less exposures due to misconceptions.

As soon as I have the feeling I really understand how it works with the D500 AF (and exposure connection) I will talk more about it. For now consider this a hint to helpful literature.

[Ilkka Nissilä]

There's a large thread going on at http://www.fredmiranda.com/forum/topic/1478115 about the difference in behaviour of the D5/D500 AF and the older Nikon's

Thanks for the link! It is quite a long discussion. I guess Nikon designed the dynamic area AF to be emphasize more the selected point than the surrounding area compared to how dynamic area AF worked in cameras based on around the various Multi-CAM 3500 series modules.  This hasn't really worked for me as my motivation to use a larger AF area than single point is because I may not be able to hold the single AF point precisely on the target and need the camera to retain my main subject in focus even when this happens and decidedly not focus on the background if I slip as the current D5 dynamic area AF seems to be doing frequently. To be honest that's the problem I had with dynamic area AF on D7100 and D810 as well, the camera is so eager to focus on the main point I kind of feel it defeats the point of using a larger area. Perhaps the backgrounds that I have are too detailed, so the focus is easily slipped there. Nikon does recommend the use of group area AF if the subject is hard to follow using a single AF point and to reduce the risk of the camera focusing on the background, which is exactly my problem with the dynamic area AF modes. Group area AF does work well for me but in horizontal orientation it cannot be off-centered on the top part of the frame as much as I would like, so in practice group area AF is more useful for me on vertical compositions than horizontal. It works especially well for relatively distant subjects where the subject is fully within the depth of field (so the exact point choice within the subject isn't critical) but background is relatively close and could potentially attract the camera's AF system.  As the subject approaches Nikon seems to think  it is best to switch to dynamic area AF ("You can then press a focus function button to select group-area AF when your subject is far away and release the button as the runners approach to instantly switch to 25-point dynamic-area AF without taking your eye from the viewfinder. "). I suppose I should give this approach a try.

Stopping down a bit does help in the situation where group area AF has focused on the torso of the subject and the face in the upper part of the body is closer, but I find I prefer the look of wide open shots most of the time for the "pop" of the subject from the backgrounds. For shooting at f/2.8 or f/2 precise control of the focus point is key. Vertical compositions don't have this problem now that cross type points extend close to the edges of the frame in that direction but for placing multiple athletes in a horizontal composition and selecting the face of just one of them can be difficult using FX. I don't mind having to crop a bit from the top to solve the issue but this does lead to the use of a single point to avoid having to crop too much. Anyway, if I use f/4 then I'm usually quite happy with group area on the chest unless the subject comes really close in which case no amount of depth of field will quite solve the issue. (Well, f/8 might but then the ISO will be quite high in many winter sports lighting scenarios).

[simsurace]

I haven't read the FM thread and I do not use these two cameras, but I'm thinking that the AF lock-on delay should be part of this discussion. It sounds to me as if a longer delay would help you stay on the subject.

[Hugh_3170]

Ann Shelbourne has been writing on this matter recently.  She posts on FZ and FM.  It might be worth checking out her findings.

I haven't read the FM thread and I do not use these two cameras, but I'm thinking that the AF lock-on delay should be part of this discussion. It sounds to me as if a longer delay would help you stay on the subject.

[Ilkka Nissilä]

Ann Shelbourne has been writing on this matter recently.  She posts on FZ and FM.  It might be worth checking out her findings.

Right;  from what I read, she seems to prefer the 3D Tracking. In this mode the subject is identified using the RGB matrix meter and pattern recognition (including face recognition if the appropriate option is set) around the user specified initial AF point and then it tries to keep the subject in focus as the subject moves about in the frame. It is very effective a lot of the time and I quite like it, but in figure skating the subjects spin around a lot and this can cause problems with the pattern recognition (the backside can look different from the front!). I've tried auto area AF (which Nikon seems to recommend along with group area AF for a distant subject and 25-point dynamic for a close subject) for pairs and it seems to switch between the two skaters frequently. I think Nikon's description in the NPS AF tips documents is helpful. So basically: use auto area AF if composition is a priority; for more control use 25-point dynamic area AF for close subject or Group Area AF for a distant subject. I have had success both with group area AF and auto area AF in this application, but not with 25-point dynamic but I guess I had misunderstood how it works especially on the new cameras. I think auto area AF has the benefit that it can be set to look for faces and then focus on that subject and it seems to be clever enough that it can recognize the face outside of the AF sensor array (using the matrix meter) and focus on that subject when the middle of the body is under some AF points.

Increasing the lock on delay even to the maximum setting will still only provide a relatively short time that the subject can be off center before the camera refocuses on the main point and to be honest in my testing I can't see a clear difference in sticking to the subject between dynamic area and single point modes; this is quite frustrating. They both seem to stay on the subject for the same time after the selected point leaves the main subject (even though the subject stays within the dynamic area), which is puzzling. I haven't tried adjusting the erratic/steady setting yet; maybe this will open the door to seeing the differences between the modes. Some recommended use of erratic and the longest delay to get more aid in following the subject. I will give that a try.

Nikon's suggestion to use lens AF mode buttons to switch between group area AF mode for distant singles and dynamic area for closer subjects is reasonable (I have them on two lenses; however, the lens I was planning on using for singles figure skating is the 300 PF which doesn't have the buttons). It's possible to use the front side Fn buttons to switch between AF area modes but I find pressing and holding those buttons awkward. I would prefer a toggle and not press and hold for switching between AF area modes. The lens buttons are easier for me to press and hold. I can give the Fn1/Fn2/Dof preview buttons a try and program them to switch to group, with default set to 25 point dynamic but to do this successfully I need to go out and practice a lot. There is a lot of information hidden between lines of the manuals and online NPS tips pages, it is just that the whys and hows are not fully given, only "for this type of subject, try these settings" and then it is left for the user to understand why they work for that situation and how to best make use of them.

I guess ideally I would like to see the following mode in addition to existing AF area modes: the camera use the color matrix meter to identify when the subject's face is leaving the primary AF point and stick to the face, but only within the specified expanded region within the AF array, not the whole image field like in 3D Tracking.  This would cause less misunderstanding of where the primary subject is between myself and the camera. However, only three modes to my knowledge use the RGB meter for subject recognition: the 3D Tracking, Auto Area AF (the difference between the two is that in 3D tracking the user identifies the subject to the camera by pointing the selected point towards it, whereas in Auto Area AF the camera decides on what is the subject and in fact quite often changes its mind depending on what there is in the image field), and Group AF (which Nikon says uses facial recognition in AF-S mode, this part of the feature is not useful for sports because of the requirement for continuous AF. I don't believe dynamic area AF modes use the color information or subject recognition; at least it is not mentioned in any of the documents that I found.

[Ilkka Nissilä]

I think I noticed something interesting. If the subject is static and I focus on it, and then move the focus point slightly off the subject, if there is detail to focus on in the background, the camera will switch to the "new subject" very quickly. If I set the blocked shot AF response to 5, it extends the time a bit but in both dynamic and single point modes it seems to take the same time before switching to the new subject. However, if I move towards the subject, the dynamic and single point area modes seem to behave differently and the dynamic holds focus on the subject a bit better. So now I believe there is a difference but it is difficult to see using a static "test" target. On the D810 the dynamic area mode holds the subject even if the subject and camera are static and the camera is turned away (but not outside the dynamic area).

So from this test it would seem the photographer is expected to be able to hold the selected point perfectly on the subject if it is static, but in a dynamic situation it permits some more wiggle. I wonder if my observations are correct or just my imagination. Lots of practical testing to do I am afraid.

[Valerie S.]

I made a few posts in the thread on FM. I asserted that this is a feature of the new AF system and that the D5 was aimed at the sports shooter market. The timing of the release for the Olympics, the choice of better DR at higher ISOs, etc... In Dynamic mode, for sports shooting, many photographers want the primary AF point to be in focus. Those extra points in Dynamic9 - 153 are just there to temporarily achieve focus until the primary AF point can lock on again. I really can't imagine letting the camera chose other AF points to "track" once the primary can achieve focus. If one wants tracking, 3D Tracking does just that. The manual describes how Dynamic modes work. I think the manuals didn't accurately describe how Dynamic works on previous bodies.

[MFloyd]

As a sport/action photographer, I'm using 3D with face recognition for people, and Group for objects / cars. These are the most commonly used modes; exceptions exist though.

[chambeshi]

I found customizing AF on the D500 on flying birds quite a challenge. I use AFC with group area mode with back-button activation. David Busch's Nikon D500 handbook is useful.

But of all my www searches, I find Steve Perry provides solid practical advice, especially AF on subjects as challenging as flying birds:

http://backcountrygallery.com/nikon-d500-review/

FYI another link on D500

https://youtu.be/RoPl_jiHxvQ

[Ilkka Nissilä]

While the D3/D4/D5 were all launched just in time for the Summer Olympics, they do advertise the D5 for a variety of applications, including sports, photojournalism, nature/wildlife, fashion etc. Furthermore the same AF system will likely be inherited by a number of professional, semi-pro/adv amateur camera bodies just as the D3's Multi-CAM 3500 was. The system designers use sports as a torture test but the AF system has to work in general photography as well.

I don't like 3D Tracking much as it doesn't give the option of restricting the area in which focus points can be selected by the camera. Also it slips easily focus points on the chest or the body outside of the face. Whole frame tracking may be useful sometimes but often I would prefer regional tracking as an option.

[MFloyd]

...
I don't like 3D Tracking much as it doesn't give the option of restricting the area in which focus points can be selected by the camera. Also it slips easily focus points on the chest or the body outside of the face. Whole frame tracking may be useful sometimes but often I would prefer regional tracking as an option.

You can select Wide or Narrow for 3D tracking

[Ilkka Nissilä]

You can select Wide or Narrow for 3D tracking

Thanks, I will try that.

From what I can see from briefly playing with it, setting the watch area to normal seems to keep the active AF point in a tighter neighbourhood of the subject position initially indicated by the selected point. So it may be less likely to slip from the face to the shirt, I guess. I had not understood the description in the manual (not the very best translation perhaps) but it seems clear what it does after brief testing.

This wasn't exactly what I was looking for but it may solve my problem. Thanks again for the tip.