Home · Register · Search · View Winners · Software · Hosting · Software · Join Upload & Sell

Moderated by: Fred Miranda
Username   Password

  New fredmiranda.com Mobile Site
  New Feature: SMS Notification alert
  New Feature: Buy & Sell Watchlist
  

FM Forums | Nikon Forum | Join Upload & Sell

  

Archive 2013 · question on FX AF point accuracy
  
 
Kerry Pierce
Offline
• • • •
Upload & Sell: On
p.1 #1 · question on FX AF point accuracy


I picked up a d800 refurb last week and wanted to check the body for the various issues that folks have been complaining about. I've never tried testing all of the AF points before, comparing all of them against the center AF point, so I don't know what is "normal".

While checking the AF outer points with the 24-70 at 24mm f/2.8, I found a slight difference in sharpness between the center AF point and both the extreme outer left and right points, with the center point being the sharpest. I then put the 24-70 on both the d700 and d3s and performed the same tests, obtaining pretty much the same results.

My test is simple. I have several focus targets taped to a large piece of cardboard. Starting from OOF, I'd select each focus point and focus on the target, moving the camera to put the AF point on one of the focus targets, where the contrast was sufficient that the AF would snap to focus quickly and surely.

So, I'd like to hear from the FX experts. Do you obtain identical results from all AF points in a test like this? If not, would you consider it normal to have the center AF point produce the best results while the outer points slightly softer when viewed at 100%?

thanks
Kerry



Jan 10, 2013 at 04:42 PM
BenV
Offline
• • • • •
Upload & Sell: On
p.1 #2 · question on FX AF point accuracy


From my experience, center is always the most accurate and sharpest. Typically when you move to the outside points, you're dealing with distortion and other lens issues. Even if its just slight issues, when your pixel peeping, you can usually tell. Don't let it bug you.


Jan 10, 2013 at 05:30 PM
Mark_L
Offline
• • • •
Upload & Sell: Off
p.1 #3 · question on FX AF point accuracy


The results won't be identical because especially wide open as I assume you are testing the lens won't perform as well in the corners as in the centre. The best way is to compare frames with live view af with 'normal' af frames using the same point/area. The lens may also have some field curvature.


Jan 10, 2013 at 06:13 PM
binary visions
Offline
• • • •
Upload & Sell: Off
p.1 #4 · question on FX AF point accuracy


Well... what you guys are talking about is only applicable if you are looking at the sharpness of the items AT the corner.

IMO, if you want to test AF accuracy, you should be sampling the sharpness of the same point every time - so you align the target so it's parallel to the sensor. Then you shoot with the center AF point, shoot with the left AF point, shoot with the right AF point... and compare the center of each shot.

Otherwise you're just measuring the lens performance.



Jan 10, 2013 at 06:46 PM
Kerry Pierce
Offline
• • • •
Upload & Sell: On
p.1 #5 · question on FX AF point accuracy


Thanks for the responses, guys. I know that the "ideal" performance would be that all AF points would produce identical results, but I'm not living in an "ideal" world.

BV, I'm not sure that I understand the point of your test procedure. Doing your test at 24mm would seem to require either a very large target or that the camera be extremely close to the target, or a combination of both. Plus, it doesn't mirror what I do when I'm actually taking photos.

When I'm taking photos with the outer points, my intent is to move the subject to the outer parts of the frame and have the sharpness on the subject there, under the AF point, not at the center of the frame.

I realize that different lenses can produce different results that way, but if an AF point is wonky, it should be easy to determine, in spite of the lens utilized. Since my initial post, I repeated the tests with the Nikon 28-70 f/2.8 AF-S and 35-70 f/2.8d. Both lenses produced almost identical results to the first test that was done with the 24-70, in that the center AF point photo was slightly sharper than either of the 2 extreme outer AF point shots.

But, I'm always willing to learn, so anyone that can help me understand a better way of doing these AF tests, I'd be grateful.

thanks
Kerry



Jan 10, 2013 at 10:44 PM
michaelwatkins
Offline
• • •
Upload & Sell: Off
p.1 #6 · question on FX AF point accuracy


What is a reasonable expectation for centre AF point accuracy rate on a camera like the D800?

75%? 80%? More? Much less?



Jan 11, 2013 at 06:45 AM
 

Search in Used Dept. 



trenchmonkey
Offline
• • • • • • •
Upload & Sell: On
p.1 #7 · question on FX AF point accuracy


michaelwatkins wrote:
What is a reasonable expectation for centre AF point accuracy rate on a camera like the D800?

75%? 80%? More? Much less?


Using good glass, I'm seeing over 98% on mine...in focus (not necessarily keepers)



Jan 11, 2013 at 10:51 AM
ausemmao
Offline
• •
Upload & Sell: Off
p.1 #8 · question on FX AF point accuracy


binary visions wrote:
Well... what you guys are talking about is only applicable if you are looking at the sharpness of the items AT the corner.

IMO, if you want to test AF accuracy, you should be sampling the sharpness of the same point every time - so you align the target so it's parallel to the sensor. Then you shoot with the center AF point, shoot with the left AF point, shoot with the right AF point... and compare the center of each shot.

Otherwise you're just measuring the lens performance.


If you do this, you're measuring field curvature more than anything else, and on a lot of lenses that will dwarf any but the most extreme of miscalibrations.

Mark_L had it right - you want to compare a live view AF shot to a viewfinder AF shot, so that you're looking at the difference between ideal at that point and real at that point - if I use the 24/1.4 as an example, what I would consider 'unacceptable' for a centre point shot would be fine for an extreme left AF point wide open because of the difference in lens performance at those points.



Jan 11, 2013 at 02:22 PM
Kerry Pierce
Offline
• • • •
Upload & Sell: On
p.1 #9 · question on FX AF point accuracy


michaelwatkins wrote:
What is a reasonable expectation for centre AF point accuracy rate on a camera like the D800?

75%? 80%? More? Much less?


On a standard AF test target, with good light, center cross point, I would expect something very close to 100% accuracy, from any of my cameras going back to the d70. Of course, that assumes the use of a lens that doesn't have front or back focus issues.

IME, poor lighting on a poor AF test target = poor results.

Kerry



Jan 11, 2013 at 06:05 PM
Kerry Pierce
Offline
• • • •
Upload & Sell: On
p.1 #10 · question on FX AF point accuracy


ausemmao wrote:
If you do this, you're measuring field curvature more than anything else, and on a lot of lenses that will dwarf any but the most extreme of miscalibrations.

Mark_L had it right - you want to compare a live view AF shot to a viewfinder AF shot, so that you're looking at the difference between ideal at that point and real at that point - if I use the 24/1.4 as an example, what I would consider 'unacceptable' for a centre point shot would be fine for an extreme left AF point wide open because of the difference in lens performance
...Show more

I haven't done the LV tests yet, but you are describing my results. The center point is slightly better than my outer points.

thanks
Kerry



Jan 11, 2013 at 06:12 PM
michaelwatkins
Offline
• • •
Upload & Sell: Off
p.1 #11 · question on FX AF point accuracy


trenchmonkey wrote:
Using good glass, I'm seeing over 98% on mine...in focus (not necessarily keepers)


That's reassuring, thanks. Nikon is new to me.

I was fairly certain the answer would be glass dependent based on my experience so far however my sample size (lens selection) is very small at present. I'm still assembling my good glass collection, a collection which won't be very large.

The D800 is the first truly modern DSLR I've owned; I'm coming from an all manual focus (medium format and more recently, mirrorless cameras) past, hence my question so I can calibrate expectations.

I've got a 85mm f/1.8G which appears to focus more consistently (yay) than the very cheap 50/1.8 AF-D I picked up to keep me busy while waiting for the Sigma 35/1.4 to arrive.

Assuming I have both the 85 and 50 set up in AF tune properly, the 85 seems to nail focus almost all of the time, whereas the older screw drive 50 is much more random in comparison - sometimes it feels like it misses 25% of the time and often the misses are meaningful. I don't know if that lens is subject to front or back focus problems. In Live View it is always accurate. I'm not yet interested in checking the left AF point. A working centre suits me fine for now.

Probably I should upgrade the 50/1.8D to the 50/1.8G, but I don't even like 50mm so may just live with it for now. If wondering why I even had the D in the first place, I had first tried two 50/1.4Gs both of which had to be returned. One was DOA and the other was badly de-centred. I probably should have bought the 50/1.8G but decided to go old tech and cheap for a focal length I'm not enamoured with.

Are older screw drive lenses more prone to less repeatable focus? Or is that trait very copy and/or lens model specific?



Jan 11, 2013 at 06:14 PM
Alan321
Offline
• • • • •
Upload & Sell: Off
p.1 #12 · question on FX AF point accuracy


The AF sensors make a difference too - the centre group are all cross-type and allow the camera twice as much data to work with to judge the correct focus (i.e. vertical and horizontal AF sensors instead of just vertical).

Then there is the fact that the outer parts of the image are less sharp than the centre regardless of any field curvature effects. Wide angle lenses and large aperture lenses usually suffer this sharpness fall-off more than others, but nearly all lenses do it to some extent. The reduced sharpness coupled with the reduced contrast caused by light fall-off make it harder for the outer AF sensors to get the same accuracy as the inner AF sensors. It's bad enough when the lens is not moving but even worse for the AF system when the lens is changing focus distance while the camera searches for focus.

And then there is the problem that the outer AF sensors are not getting as good a contrast between different viewing directions as the inner sensors get. The light is split to provide two (or four) views from opposite directions but this gets harder the farther away the sensor is from the middle of the image area. There's also a greater discrepancy out there between what comes from the edges of the lens image and what comes from the middle of the lens image.

All this is true of any AF camera using the same sort of phase detection AF, although different cameras have different levels of AF sensor technology and different locations for the sensors. On top of that, however, the D800 provides far more resolution than most cameras in terms of photo pixels per duck to better reveal the extent of any slight variations in AF accuracy and lens image quality.


Unfortunately, none of the camera manufacturers supply data to indicate how many image pixels would correspond to a single AF sensor pixel, and so we don't know for sure what level of detail the AF system is actually able to resolve. It might be basing its results on a low-res big-picture view of contrast rather than a high-res fine-detail view.

We also don't even know for sure the actual shape and size and location of the individual sensors relative to the image area. Some are X shaped, some +, some T (rotated any which way), some - and some |. Some are thicker than others. Some combine different AF sensor lengths and pixel sensitivities for initial focus acquisition and final focus.

Another factor is that different lenses probably divide the focus distance range into different numbers of segments. Some would no doubt be more precise than others, which might have been fine for low-res cameras of old but not so fine for a D800. The camera basically tells the lens which segment to go to and its arrival at that segment will have a certain tolerance that is suited for the focal length and maximum aperture. But we all know (or should do) that the common tolerances used for focus precision and DOF are way looser than they ought to be for modern cameras. That's why hyperfocal distance focusing often lets us down.

With so many variables I think it is a small miracle that AF works at all

- Alan



Jan 14, 2013 at 12:13 AM





FM Forums | Nikon Forum | Join Upload & Sell

    
 

You are not logged in. Login or Register

Username   Password    Retrive password