Nikon D5 (D500?) autofocus weirdness... Anyone else?

[sk66]

I've been working w/ the D5 autofocus system for a while now, trying to figure out what mode's best to use and when... it seems the D5 is a little different than any previous Nikon I've owned.

First off, if using any dynamic mode other than auto the camera does not report which AF point it thinks was in focus... It only shows the selected AF point as being in focus (even if it clearly was not). I know it is tracking, but I have no idea how well it is tracking other than in-focus images...which seem to be less than expected/with other modes.

Secondly, the "focus tracking with lock-on" settings seems to be more "selected point priority" settings... If set to a low level it seems the focus does not track very well at all, even if there is no apparent reason for AF to have been lost. The only way I can get 1/2 decent tracking for difficult subjects like BIF is to set "blocked shot" to maximum delay with "subject motion" set to erratic. Even then it won't stick with a focus area for very long, even if it is held stationary. My personal results seem to be better using 3D or Auto.

The "logic" seems to be:
The blocked shot setting is actually "selected point priority" determining how quickly the AF reverts to the selected point. And even at maximum delay, it's not very long.
The subject motion setting is a "probability logic"... If set to "steady" the camera will revert to the selected point more quickly than if set to "erratic." There's less "reason" for AF to be lost so it reacts a bit quicker, but it doesn't seem to make a big difference.
And none of this is affected by whether the focus area (subject) is lost or not.

Basically, I'm finding that the dynamic focus modes (d9-d153) are *worse* than previous Nikons when the subject is particularly difficult to track... it just will not stick long enough. And it means I have to be much more careful about what AF mode I select as none of the dynamic modes are particularly flexible across a wide array of subjects/situations.
(It also means you can't really use them for the focus/recompose technique, not that it's an issue really)

All of this is quite different IME.

Of course, as things get easier the dynamic modes work much better... but then there is less need for them. On the other hand, "Auto" is surprisingly good in a wide variety of settings (I still haven't used 3D enough to have a solid opinion of it yet).

 

[sk66]

I've found another oddity of the D5 with dynamic AF, if using spot metering the metering does not move with the active AF point... that kind of sucks.

I'm guessing this all applies to the D500 as well since it has the same system?

 

[gramps]

I'm guessing this all applies to the D500 as well since it has the same system?

According to the Nikon blurb:

Spot: Meters 3.5-mm circle (about 2.5% of frame) centered on selected focus point (on center focus point when non-CPU lens is used)

 

[sk66]

According to the Nikon blurb:

Yes, "selected focus point" which has always been the case... but with previous Nikons it actually moved with the "active focus point" when tracking (although I don't think it ever mentioned that).

 

[snerkler]

I don't use dynamic AF but if you says it's not working in the 'usual' way have you tried another D5 to see if that's the same?

 

[sk66]

Nope, I don't know anyone with one nearby...

I think it is related to the "in focus point" not being reported when tracking... it only reports the selected point. Maybe I'll try contacting Nikon.

 

[gramps]

Long thread on FredMiranda about this, started by Steve Perry ... you may have seen it but quite a number of comments about your findings.
Possible thought that Nikon made it this way rather than it being a 'bug'.
http://www.fredmiranda.com/forum/topic/1478115/

 

[sk66]

I wasn't aware of that thread, thanks.

Nikon's first response was essentially "this is all normal... RTM". Took me a while to explain it so that they could reproduce it... now waiting on another reply.

 

[holty]

good link Gramps

 

[sk66]

Seems Steve Perry has come to the same conclusions I have...

My suspicion is that with the massive increase in AF points they have made the system much more "active." This is probably a benefit with easier/slower subjects like motorsports. But the reality is that very fast/demanding subjects (small BIFs) are probably too much for any AF system to keep up with reliably. I know that with other cameras (D8xx/D3/D4/etc) the slower system would "track" longer, but the focus would not actually be at the point the camera reports, focus was actually "lost." Basically, a lot of shots were "luck" and we (I) would use greater DOF to try to increase the odds.

Now I think the chances of focus being lost are less, but the chances of it being focused on a point other than desired are greater.

 

[gramps]

It does seem that many are using Group/Single-Point rather than any of the Dynamic options ... these are my chosen options.

 

[sk66]

It does seem that many are using Group/Single-Point rather than any of the Dynamic options ... these are my chosen options.

I always try to use as few points as possible, but just try to photograph a swallow in flight using group/single point... I can't keep it on the subject long enough to even acquire focus most of the time, not to mention tracking.

Actually, I had probably gotten a bit lazy with choosing focus mode/zone leaving it in a higher tracking mode unless it caused me an issue... That's probably why I noticed the difference right away, it was causing me more issues more frequently.

 

[gramps]

BIF gets Group from me, I would be hopeless on a small bird with Single-point ... I tried Auto as some were having success but it was very hit & miss for me.

 

[sk66]

In my testing I have found a few other factors that seem to influence how quickly/if the D5 AF will switch back to the initial AF point. This was tested with the delay set to max and subject motion set to erratic.

If there is nothing under the initial AF point (i.e. blue sky) it will not revert to the initial AF point... at least not quickly at all (perhaps at some time longer than I allowed which was around 20 sec).

If what is under the initial AF point is harder to focus on (greater separation/less DOF) it will be slower to revert.

If what is under the initial AF point is held there steadily, it will revert more quickly.

This seems to indicate that the system is a bit smarter than anticipated. It does not appear to be simply "timing out," but rather that it is simultaneously tracking while checking against the initial AF point. If there is something "good" under the initial point, the more consistently it's under the point, and the closer to being in focus it is, the quicker it will revert.

The situations where this behavior would not be largely beneficial are hard to track small subjects with busy/near backgrounds. In all other situations where tracking is easier, or with better tracking technique, this behavior should prove to be beneficial. It also seems to show a benefit to using faster "pro" lenses that have less DOF when focusing... which makes some sense considering the target market for the D5.

I don't really see a benefit to using "group" mode in the types of situations where tracking is beneficial. All Nikons already have the selected AF point functioning in "group mode" when in a dynamic tracking mode. However, w/ the D5/500 the "dynamic group" only seems to include the 8 non-selectable AF points around the selected point and it is smaller than the area used in the dedicated group mode. (This is not the case with other Nikons)

I think the (my) general rule of "use as few focus points as you can reliably keep on subject" still holds true. I also think I had gotten a bit sloppy about observing that rule with previous Nikons because it was not as quick to penalize me.

 

[sk66]

Ok, with additional testing I think I have found the primary difference between the D5/D500 and earlier cameras when in a dynamic mode.
The new system is much faster and has many more AF points spaced much closer together... that's accepted and should be a bonus. And IMO it is in many/most situations.

The main difference/issue is that earlier cameras track in "blocked shot mode" and the new cameras track in "distance change mode." The "focus tracking with lock on" setting has always been described as a setting for "blocked shots" based on "change in distance," but the cameras are addressing those two aspects differently.

Earlier cameras will *not* shift to a more distant focus unless focus is lost (i.e. d51 w/ delay off). But they will shift to a nearer focus dependent on the delay setting (which goes from off to quite long). Exactly what you would expect/want for "blocked shots."
The D5/D500 will shift to both a nearer AND a more distant focus, also dependent on the delay setting (which goes from short to really really short). This means the D5/D500 will happily jump to the BG if it reasonably can... and the system is so good/fast that it often can.

Everything else I have found/identified/verified seems to be "the same," only faster/better. This one difference makes the D5/D500 dynamic modes worse in very demanding situations... exactly the situations where we depended on it to make up for other shortfalls.
Luckily auto and 3d modes have been greatly improved over previous models.

I have been in discussion with Nikon and they have been testing/verifying my findings. Supposedly my findings/concerns are being passed to the Nikon D5 expert... I made a recommendation of a firmware fix that adds an option called "limit focus shift" or "blocked shot mode" as the new behavior can be beneficial in situations where the subject is large in the FOV. I emphasized the benefit of the "blocked shot" behavior for wildlife photographers shooting demanding subjects like birds in flight, and that wildlife photographers are a huge portion of the consumers purchasing the D5/D500. Even if they do listen, I wouldn't expect the change anytime soon.