I took Moose's advice (after a bit) and set my Af to 21 points and with Dynamic Mode and Center Weighted metering I can get like 90% in focus with my blonde dog running at me through a field of yellow dried grass. Not so hot with Matrix and Auto. Worse with 3D on. I'm not sure what 3D is for and not many others seem to know either. Although Rockwell likes it for shooting his kid.
I shot the Jeep King of the Mountain in Telluride this last weekend. I used 51 pt Dynamic with center weighted metering. The skiers were moving away from me at about a 30 deg angle with up to 4 on the jump at the same time. I got about 80% focus with the balance picking up Aspen trees or a construction crane 400 yards away. 3D did not work - it tended to pick up the camera crew on the jump.
The manaual talks about 3D for objects moving side to side, not towards or away from you. For that situation, I believe you'd need to use Continuous AF with however many points you need, so that predictive tracking kicks in.
The dog was in perfect focus. I don't know if his post processing and glass match up to your 200mm. In any case I thought the fact that he was using 9 points might be relevant. The sports shooters are saying that they are getting better results all around including auto focus on their D300's over the D2X. Maybe you need to see what they are doing. I think they are using 9 point and definitely not the 51 point.
The 9 cross type sensors on the D2x/D2xs cover 3 times as much of the viewfinder horizon and could be used effectively for AF. With the D3 I find that with a subject moving toward me the only reliable AF is with 51-3D and no IR assist.
The D3/D300 autofocus is designed to use multiple sensors in concert to determine the subject and achieve focus. Using a single sensor is going to be less effective in most circumstances and you really need to adjust to the new system and forget how you used the D2 series cameras.
I emailed dave black and asked what he was using and he responded that he is mainly using the 51 - 3D but said that it still requires you to follow your subject fairly well along with making sure they don't leave the focus area.
Forget running - how about AF with someone walking toward the camera. I did a series of test shots with a fair skinned subject wearing a white top and black slacks with the AF sensor over the face. With 21 or 51 pt. AF and C set for focus only and using dynamic area AF I had no shots with the face in focus. NONE
When I examined the images later with ViewNX and looked where the camera had selected to focus at the time the shutter was released and it had jumped to a sensor with more contrast. Some shots were at the edge of the white top where it met the neck and other shots were at the edge of the top and the slacks. It was off in terms of the selected AF sensor point by anywhere from 8" to 2 feet.
I tried 51 point AF with 3-D and the camera stayed with the subjects face. According to the Nikon manual this setting is to be used for erratically moving subjects. But in this case the subject was moving in a straight line toward the camera and this was the only setting that worked at all.
I have done this many times with the D2X and Mark III and both cameras nail the focus 10 out of 10 times and the Mark III can do this with half as much light as the D2X.
I have a wedding to shoot on Saturday and at this point in time I have zero confidence in the D3 autofocus system. I was planning to sell my Mark III but now think I will be needing it for this wedding and quite possibly future ones as well.
I have used the AF with the N90, F5, D100, D200, D1x, D2h, and D2x with no problems and no worries. I cannot say the same about the D3 or D300. There is something terribly wrong with the Nikon AF system in its present state with either low contrast subjects or low contrast lighting. Hopefully it is something that can be addressed in a firmware update - SOON.
Maybe you have a faulty example of the D3, or perhaps you have not yet found the right combination of settings for your situation. Your D3 manual explains that 9-point AF is intended for a subject like yours, moving predictably towards the camera.
Were you holding the camera horizontally or vertically? Had you shifted the focus point grouping to the top of the frame?
Edited to add: I know you said you put the group on the face but was this done in classic 'meter and recompose' style or did you actually shift the group? It could be you were unconsciously lowering the frame as you shoot thus in effect recomposing and the AF was dropped down a bit during the sequence.
I've had very good success with the D3/D300 AF. Much more so than the D2X. One thing to keep in mind with the Dynamic AF-- its intent is to keep an active sensor on the subject when the subject is moving laterally across the viewfinder. If you are following a moving target, sometimes the primary (selected) AF point will move off the subject. Dynamic AF is there to help keep a subject in focus should this happen. Activating more sensors is useful when the subject is very erratic in the frame, but it also can increase the likelihood that the AF will lock onto something other than the intended target.
I've tried the 3D tracking, and it works quite well, provided that there is a fair bit of color/contrast difference between your subject and the background.
The most challenging situation for ANY AF system is the subject that is moving towards or away from the camera rapidly. The AF system is trained (so to speak) to ignore fast jumps in subject distance because most of the time this is because something has come between you and your subject. The way to remedy this is by adjusting the lock-on settings to either short or OFF. Of course, this also increases the chance that the AF will lock onto another subject should you "miss" with one of the active sensors.
With all new systems, they do take time to learn. However I am puzzled by the trouble you are having. My D3 and D300 are nothing short of outstanding in the AF department, but they still have their limitations.
Jason P. Odell Colorado Nikonian Author, The Photographer's Guide to Capture NX Now supports Capture NX 1.3
In about 10-15 sessions of testing the AF with my 200VR and corresponding with Nikon USA, the D3's AF system has proved to be harder to grasp than the D2X.
So far, the best configuration to obtain the highest keepers for a subject running directly at me is the following:
0) AF-C (of course) --- added for clarity 1) a4=OFF 2) Lossless compressed RAW (Nikon recommended) but I think Uncompressed RAW could be better. There's a tradeoff here between the in-camera processor spending time compressing the file versus the time spent flushing out the file into the high-speed buffer. Nikon is not revealing the time taken for either of them due to trade-secrets, so to speak. 3) Active D-lighting: ON (recommended by Nikon) and again, I think it may contend for resources when the frame rate is high since Active D-lighting is applied before the image is captured. 4) 12-bit : Don't use 14-bit since it will increase the size of the file and contend for computing resources again for high frame rate situation. 5) 9pt seems to be the best for my dog trotting towards me 6) Please test the static focus of your lens and AF-tune it. If your lens is off focus in the first place, no AF system can fix that.
Also, please bear in mind that my weather here has been highly overcast and shots were mostly low contrast in nature ... none of the Rob Galbraith's Phoenix Arizona light.
Disclaimers: A)This config was tested for the situation where the subject is running directly at the camera. B)I have tested this for my own benefits and use case. If you think that your own config gives you better results, please use it by all means. Mine is, in no way, authoritative. I can't be authoritative because I do not have access to internal specs which Nikon has not released to the public.
>A Y, > >Would not turning off a4 'increase' the focus array's >tendency to go astray? > >To my reading a4 is about the AF system ignoring intrusions.
a4=(short, long} will add a wait-time before re-focus and this is not what we want for the situation I tested on.
In the situation where I have a subject running directly at you and no chance of the subject being blocked by a tree in between, we want the AF to react as quickly as the rate of change in the distance separating us. Having a4=ON will increase the potential of adding that wait time if the AF sensor thinks it "saw" something blocking the original subject.
Unless you think there's a possibility that something *may* come between you and the subject, a4 need not be turned on.
>I understand how it would effect continuous >high speed shooting but the af system is only dependent on a >few different settings.
That's my point: the config I have tested is for AF-C=ON; i.e. for continuous AF. Please read carefully: the test scenario is a runner/dog running directly at the camera. It's pretty obvious that AF-C has to be the option --- *not* AF-S.
>>I understand how it would effect continuous >>high speed shooting but the af system is only dependent on a >>few different settings. > > >That's my point: the config I have tested is for AF-C=ON; >i.e. for continuous AF. Please read carefully: the test >scenario is a runner/dog running directly at the camera. >It's pretty obvious that AF-C has to be the option --- *not* >AF-S.
yeah wether you use 14 bit, or compressed or not that won't effect the AF system
turn it on AF-C and Dynamic 3D and follow your subject around, and not just having it run at you and im sure you'll be amazed at the performance
Arthur, this may not be relevant to some of your points. And I am certainly not disagree with your findings. I have been told that the D3 has 2 processors while the D300 has one and this is why the D3 is able to maintain 9fps in 14-bit mode, while the D300 drops to 2.5fps in 14-bit mode. I only mention this as it "might" affect your conclusions. Perhaps not.
Again, I am not disputing your findings. Just pointing out another factor (assuming the D3 has 2 processors.) And, thank you for starting this thread.
I actually saw this mentioned elsewhere but I did not see any citations to authoritative publications. Where did you see that piece of info, btw?
It's interesting that when I described the use-case to Nikon USA, they confirmed that 12-bit would help with high fps.
My guess is that even if there's a separate processor computing the bit-depth, we do realize that the file size is different between the 12 and 14-bit. Consequently, the time needed to push the data into the high-speed buffer (*not* from the buffer to CF card!) will be slightly longer for the 14-bit files.
I would think that the camera cannot proceed to capture the next frame until the current frame is fully in the high-speed buffer. So, larger files do incur a longer wait time, albeit small time differences which can possibly make a difference in action photography.
I've looked but cannot find a reference. I will have to look some more. One of of the employees at my local shop mentioned it when I asked why the D3 could do 9fps at 14-bit while the D300 couldn't. I know he has been spoken with Nikon reps about the D3, so I feel confident he is not making up the 2nd processor. That is just not his or the shop's style. If they don't know, they will tell me so.
Obviously, there is something different, processor-wise, between the two. If I find support for this, I'll post back here and let you know.
- is the more AF points are selected the more information there is to process and the slower the AF responds. It follows 9 or 21 points give a higher success ratio because of faster processing speed - provided you can keep the subject within the AF selection points. Digressing the top sports photographers are top sports photographers because they can track moving targets holding them within the selected AF zone far better than average photographers. With some subjects many pros use a monopod to help get better AF tracking.
Photography is a bit like archery. A technically better camera, lens or arrow may not hit the target as often as it could if the photographer or archer does not practice enough.
There is a comparison in this weeks British Journal of Photography that looks at the D3 against a 1D Mark III. One of the tests was a model running directly towards the camera to look at the AF. Shockingly the report suggests that the D3 scored 0% for shots pins sharp with only 5% of shots acceptable (out of 189 shots). The 1D Mark III reached 18% pinsharp.
The rest of the report paints a reasonably good picture, and suggests that many of the focusing problems are down to the choice of lens they used (70-200 2.8 VR) which is a theme I've picked up elsewhere in relation to the FX sensor at least.
Maybe, but I shot an air show a couple of weeks ago in 51-point 3D mode (old firmware) and almost all of the shots were in perfect focus, including shots tracking fighter jets as the aircraft approached the runway so the velocity vector in my direction was substantial. Lens was a Sigma 120-300 f/2.8. If the AF processor could keep up with that I have to think it could keep up with just about any subject in 9- or 21-point mode.