Autofocus issues

[Marconikon]

Hello everybody, I’m new to the forum but I’ve been reading it for quite a while especially for technical insight and decided to join because of an issue I seem unable to solve.
How good should be the Viewfinder focusing system? Am I expecting too much from my camera in terms of AF?
Here is a (not so) brief explanation of the issues I encountered: I’m a Nikon D800 owner and I use it mainly with a 35-70 f2.8 D (push pull) and 80-200 f2.8 D and some other manual lenses from film days and started noticing some autofocus inconsistencies in portraits (mostly at the long end of the zooms) even at 70mm f5 3.5-4m away from the subject so I set up a test chart on my wall, lit it with a two light setup, put my camera on a tripod and took many shots at different apertures and focal lengths using liveview focus as a reference and viewfinder focus as a comparison. My findings were that I had severe back focus with both of my lenses but the thing that confused me was that the lenses seemed to have a slight back focus at f2.8 and a much stronger one at f5. Unable to assess the situation clearly I decided to send my camera whith the 35-70 to Nikon for a checkup and got it back a few weeks later with some minor contact and sensor cleaning done, so I assume for Nikon everything is ok. I tried testing the lens again and it seems as if there has been a minor improvement on the 35-70 (I could potentially solve with 3 different fine tunes: one for short distances at f2.8 and f5 and one for long distances, even though I’d rather manual focus). I hope this wasn’t too boring to read and that someone could help me out, thank you!!

[Shadowman]

Welcome to the forum, I assume your name is Marco if not that is what I'll call you until you correct me? First off, a few questions on your earlier experiment:

1. How were the portrait inconsistencies evaluated, did you utilize the focus point indicator at the time of the shoot and during the editing stage? Did you make any prints, if so how did they look?

2. What did you use for subject to test lens when you noticed back focusing, was it a flat focus chart, small batteries set at a distance apart can be used also?

3. Did you use base ISO when doing the experiment?

4. Can you post a few examples of the OOF portraits and show where the focus point was and include camera settings?

5. If you are typically manually focusing, are you using the camera's rangefinder if it has one?

[george013]

Hello everybody, I’m new to the forum but I’ve been reading it for quite a while especially for technical insight and decided to join because of an issue I seem unable to solve.
How good should be the Viewfinder focusing system? Am I expecting too much from my camera in terms of AF?
Here is a (not so) brief explanation of the issues I encountered: I’m a Nikon D800 owner and I use it mainly with a 35-70 f2.8 D (push pull) and 80-200 f2.8 D and some other manual lenses from film days and started noticing some autofocus inconsistencies in portraits (mostly at the long end of the zooms) even at 70mm f5 3.5-4m away from the subject so I set up a test chart on my wall, lit it with a two light setup, put my camera on a tripod and took many shots at different apertures and focal lengths using liveview focus as a reference and viewfinder focus as a comparison. My findings were that I had severe back focus with both of my lenses but the thing that confused me was that the lenses seemed to have a slight back focus at f2.8 and a much stronger one at f5. Unable to assess the situation clearly I decided to send my camera whith the 35-70 to Nikon for a checkup and got it back a few weeks later with some minor contact and sensor cleaning done, so I assume for Nikon everything is ok. I tried testing the lens again and it seems as if there has been a minor improvement on the 35-70 (I could potentially solve with 3 different fine tunes: one for short distances at f2.8 and f5 and one for long distances, even though I’d rather manual focus). I hope this wasn’t too boring to read and that someone could help me out, thank you!!

There're two autofocus systems: phase and contrast. I think what you call viewfinder autofocus system is the phase system. A link to a site dealing with Canon https://www.creative-photographer.co...ion-autofocus/. See it as a start for further discussion.

George

[Marconikon]

Thank you for the replies, name is Marco and yes with VF and LV focus I meant exactly phase and contrast detection. Regarding the questions at the moment I’m travelling and can’t post any examples (but I will Monday). What I can say is that I noticed some issues when taking portrait shots at f5 at a distance of roughly 3-4m with the phase detection system (more than 70% of the shots were OOF, focus point on the eye), maybe acceptable printed but definitely noticeable on a computer/iPhone screen (not zoomed at 100%). After that shoot I tested the lenses I have with a focus chart flat on a wall and with the camera on a tripod shooting with a cable release (no camera shake) at base ISO and the difference in sharpness was evident. It gets better when fine tuning (-10 default -20 lens saved value) but it won’t yeald a consistent result throughout all the aperture settings (at 2.8 it seems as if the back focus is much less, good results at: -10 default +15 lens value). I tend to manual focus through te rangefinder whenever I shoot handheld.

[Shadowman]

Thank you for the replies, name is Marco and yes with VF and LW focus I meant exactly phase and contrast detection. Regarding the questions at the moment I’m travelling and can’t post any examples (but I will Monday). What I can say is that I noticed some issues when taking portrait shots at f5 at a distance of roughly 3-4m with the phase detection system (more than 70% of the shots were OOF, focus point on the eye), maybe acceptable printed but definitely noticeable on a computer/iPhone screen (not zoomed at 100%). After that shoot I tested the lenses I have with a focus chart flat on a wall and with the camera on a tripod shooting with a cable release (no camera shake) at base ISO and the difference in sharpness was evident. It gets better when fine tuning (-10 default -20 lens saved value) but it won’t yeald a consistent result throughout all the aperture settings (at 2.8 it seems as if the back focus is much less, good results at: -10 default +15 lens value). I tend to manual focus through te rangefinder whenever I shoot handheld.

Marco,

The flat chart won't help detect front/back focus but it will help determine sharpest aperture and overall focus capability of your lens/camera system. There are others things to consider mainly the softness of your particular lens, I'm not familiar with the 35-70 but have known some lenses to basically render soft images across the range of apertures, whether or not auto focus fine tuning helps well you'll have to experiment, I never bothered trying to fine tune my soft zoom lens but I did notice an improvement when I used flash.

[Manfred M]

What focus mode are you using? Unless you have set your camera to a single focus point, I would be suspicious of your test results.

Are the results identical in Phase Detect and Contrast Detect modes? The camera uses completely different components to determine focus. Phase detect uses a separate opti-electronic component that operates very much like a traditional rangefinder and contrast detect uses the image data collected by the camera's sensor to determine focus.

Have you tested your lenses with a different camera body to see if the results are the same?

[Marconikon]

I understand you are skeptical about the test conditions and to some extent I am too, but of course I am in single point AF, I have lit my chart correctly and I have used a steady tripod on a flat and levelled surface. Now I know flat charts aren’t used to determine how much back or front focus a certain lens and camera combination have, but definitely can be used to asses if your phase AF is missing the focus. The thing is my lens is perfectly sharp when focused correctly (also via contrast detection - LV) but isn’t when focused using phase detection (VF). The fact it was a back focusing issue was determined looking at some portrait shots and confirmed trying different values of fine tuning.

[Manfred M]

I understand you are skeptical about the test conditions and to some extent I am too, but of course I am in single point AF, I have lit my chart correctly and I have used a steady tripod on a flat and levelled surface. Now I know flat charts aren’t used to determine how much back or front focus a certain lens and camera combination have, but definitely can be used to asses if your phase AF is missing the focus. The thing is my lens is perfectly sharp when focused correctly (also via contrast detection - LV) but isn’t when focused using phase detection (VF). The fact it was a back focusing issue was determined looking at some portrait shots and confirmed trying different values of fine tuning.

It sounds like a camera issue rather than a lens issue. I seem to remember reading that some of the D800 models had an improperly installed / aligned phase detect module that was the cause of some focus issues. I didn't have it with mine, but wonder if that is what you might be experiencing.

Both phase detect and contrast detect should be giving you the same focus.

[Marconikon]

It sounds like a camera issue rather than a lens issue. I seem to remember reading that some of the D800 models had an improperly installed / aligned phase detect module that was the cause of some focus issues. I didn't have it with mine, but wonder if that is what you might be experiencing.

Both phase detect and contrast detect should be giving you the same focus.

Thank you very much for the info, I’ll check again with Nikon then. The phase system works without feedback so it should be less accurate than the contrast AF but what I’m experiencing is phase consistently missing. The other thing that worries me is that the amount of back focus seems to be dependent on the aperture setting...

[Manfred M]

The phase system works without feedback so it should be less accurate than the contrast AF but what I’m experiencing is phase consistently missing.

That's definitely incorrect. Autofocus is effectively a control system, so there is feedback in the loop using a predictor / corrector approach. Initial measurement is a combination of how far to move the focus mechanism and then compare it to what has happened. If it has hit the target it will stop. If it measures over or undershoot, it will correct. Watch the focus "jitter" as the system focuses and you'll see what I mean. This is especially apparent when the object being focused on is relatively close to the camera.

The issue with the technology is that the accuracy is dependent on the offset of the two beams being measured. On a long axis, the accuracy is quite high, shorten the axis, and accuracy drops off. This is a problem with on-sensor phase detect as the size of the sensor determines the offset between the measuring points. Contrast detect does not have this issue, but is much more dependent on light levels and contrast of the subject, plus of course the previously mentioned inability to determine which direction the lens needs to move to get the subject in focus. Combine on sensor phase detect and contrast detect and one gets a winner.



The reason for the inaccuracy has more to do with measurement accuracy of the

[Stagecoach]

Hi Marco,

A couple of observations (I also have a D800 and have experience of fine tuning a lens combo to it) .................

good results at: -10 default +15 lens value)

I am unclear as to what you are referring to when you mention "-10 default"?

Edit .................................................. ...............................

Investigating this further it appears that if a value of say -10 is put in the Fine Tune 'Default' option this will set all lenses that you use and that you have not set a -/+ fine tune value in their respective settings to -10. In your case I would ask, why have you got a value of -10 set in the Default?

What's also interesting with this is that when you feed say -10 into the 'Default' option it still shows '0' in the setting range for the specific lens (I just checked it on my D800 and new lens). Is it summing the two inputs, I have read one comment that suggests not?

End of Edit ................................

at 2.8 it seems as if the back focus is much less

This is interesting as the camera would be focusing with the aperture wide open, and locking it's adjusted focus prior to the shutter activating. So we would assume focus plane remains the same whatever aperture is used.

Just some thoughts.

[pschlute]

Redo the tests but the focus chart needs to at a 45 degree angle to determine front or back focus.

Don't rely on one single AF test. Do 10 tests at each zoom setting and adjust the AF tuning until you get no FF/BF.

Use widest aperture only. As has been pointed out the AF will always "lock-on" at widest aperture. Capturing the test image at widest aperture makes it easier to see any FF/BF issues.

Try to use natural light, or light that is consistent with how you will use the lens. Some light from flourescent tubes can have an effect on the accuracy of the AF system on some cameras.

[Marconikon]

I used -10 default because -20 on the lens at f5 wasn’t enough and instead of changing to 0 default and +5 lens for f2.8 I just dialed in +15... anyway as said before it is a misconception that phase detection is open loop, and you are definitely right saying focus is locked at max aperture and that the focus plane shouldn’t change and that is why I’m particularly concerned and puzzled.

[Marconikon]

The 45 degree chart is very hard to set up correctly and to me it’s not worth the time plus remember you should be roughly 50x focal length away from the target so it makes it even harder at longer ends.. I did test it multiple times anyway and I’m sure it is back focus

[Shadowman]

The 45 degree chart is very hard to set up correctly and to me it’s not worth the time plus remember you should be roughly 50x focal length away from the target so it makes it even harder at longer ends.. I did test it multiple times anyway and I’m sure it is back focus

The distance from factor is for the flat chart, you just want to be at the 45 degree angle and obviously at your lenses closest focus distance. For the back focus test I like to use a series of batteries, measuring the distance between each can be helpful.

[pschlute]

The 45 degree chart is very hard to set up correctly and to me it’s not worth the time plus remember you should be roughly 50x focal length away from the target so it makes it even harder at longer ends.. I did test it multiple times anyway and I’m sure it is back focus

Then dial in some BF correction. Sorted.

[Stagecoach]

I used -10 default because -20 on the lens at f5 wasn’t enough and instead of changing to 0 default and +5 lens for f2.8 I just dialed in +15... anyway as said before it is a misconception that phase detection is open loop, and you are definitely right saying focus is locked at max aperture and that the focus plane shouldn’t change and that is why I’m particularly concerned and puzzled.

Marco,

I believe your logic assumption is incorrect regarding any interaction (summation) between the 'Default' and 'Individual' two independent options to feed a +/- fine tune value into.

From tests I undertook using the D800 today it appears to me that any +/- value that is already or you enter in within the 'Default' option is not taken into account once you feed a +/- value (which can also be '0') into the individual lens input and press OK to save it.

The only way a Nikon CPU lens will take account of any AF Fine Tune +/- value that has been fed and saved into the AF Fine Tune 'Default' option is if it has not previously had a value entered and saved under that individual lens info that is automatically registered when you mount the lens.

The test I undertook below (repeated 3 times) with a good AF target that also had subjects at the side behind and in front of the subject plane gave the following. I also undertook others to support my conclusion;

a) Default 0, Lens 0 : Subject sharp

b) Default 0, Lens +20 : Subject less sharp and subjects behind coming sharper as would be expected.

c) Default +20, Lens +20 : Subject less sharp same as b) above and no change to forward/back focus.

d) Default -20, Lens +20 : Subject less sharp same as b) above and no change to forward/back focus.

From my tests I don't believe the two +/- input values for 'Default' and 'Individual Lens' are summed together.

This to me makes sense and your assumption that with a 'Default' of -10, dialling in +15 for the 'Lens' gives you +5 I believe is wrong. In addition, a -10 Default input with a -20 Lens input simply gives an actual AF Fine Tune true Bias of -20 (only the Lens +/- input value is used).

Without a full understanding and appreciation of the above interactions I suspect there is a high risk of dodgy and confusing results to base conclusions on. Personally, I would set the Default at '0' and forget it, unless you have it purposefully set at a figure because you know that it 'corrects/improves' other lenses that you have never calibrated.

With respect to anomalies between aperture setting I also had a quick look at this to see if I could detect anything obvious. I undertook two tests Auto focusing from a de-focused position using the centre AF point at different apertures using the same target and FL.

The tests consisted of running through the range from +20 to -20 in increments of 5. At f/2.8 the sharpest example happened at -20 and at f/6.3 the sharpest example happened at -10. This was with a lens I have only just acquired and everything I have shot to date has not given me a result that concerns me. So it 'appears' there may be some difference between apertures but I would need to do far more extensive controlled testing to come to a valid conclusion taking account of it's impact on real world photography.

[Marconikon]

Many thanks to everyone for the replies, I went back to Nikon and spoke to the technician who had checked the camera and he told me a few things some of which sound quite weird to me:

1) according to him the AF inconsistency depends on the fact I am using AF-On focusing (back button) and release priority in AF-S single point though I am not recomposing nor I'm letting go the AF-On button before i release the shutter. So for him you should always prioritize focus and use the shutter button half pressed to focus in order to get decent results (so my question would be: why put the option to change the settings ? );

2) contrast and phase detection should have the same accuracy (at least when focus is achieved) which sounds pretty reasonable;

3) he told me my camera had no issue whatsoever and that D800's have a wider tolerance in respect to AF so, in theory, less precise;

4) AF-D lenses could have AF issues;

5) he tried the camera again in the shop and said it appeared to be ok for him.

The thing is I went back home and did some free hand testing again and noticed the focus problems are evident at f5 when shooting at 50-70mm (didn't try the 80-200) close to the minimum focusing distance.
I did something very easy to confirm my suspicion: I manually focused the lens until confirmation (green dot) and took a shot, then I turned the focusing ring just a little bit in order to shift the focus to the front and took another shot. Checking the pictures at 100% the confirmed focus picture was off and the "front shifted focus" was sharp.

Now, do you think it is normal to have such a difference between phase and contrast detection depending on distance and aperture setting?

[george013]

....
The thing is I went back home and did some free hand testing again and noticed the focus problems are evident at f5 when shooting at 50-70mm (didn't try the 80-200) close to the minimum focusing distance.
I did something very easy to confirm my suspicion: I manually focused the lens until confirmation (green dot) and took a shot, then I turned the focusing ring just a little bit in order to shift the focus to the front and took another shot. Checking the pictures at 100% the confirmed focus picture was off and the "front shifted focus" was sharp.

Now, do you think it is normal to have such a difference between phase and contrast detection depending on distance and aperture setting?

I don't think that's the way to test the camera/lens. Do it on a larger distance. I too have a far less reliable result close to the minimum focusing distance.

Beside that I don't understand the problems with a smaller aperture. As said the camera is metering with an open aperture. A smaller aperture will enlarge the range wherein the picture is sharp.

George

[NorthernFocus]

Many thanks to everyone for the replies, I went back to Nikon and spoke to the technician who had checked the camera and he told me a few things some of which sound quite weird to me:

1) according to him the AF inconsistency depends on the fact I am using AF-On focusing (back button) and release priority in AF-S single point though I am not recomposing nor I'm letting go the AF-On button before i release the shutter. So for him you should always prioritize focus and use the shutter button half pressed to focus in order to get decent results...

That's utter nonsense. Every professional wildlife photographer that I know uses BBF. Sometimes I wonder if anyone at Nikon is actually a photographer.

2) contrast and phase detection should have the same accuracy...

Then why bother with phase? Plus this is contrary to Nikon's own marketing mantra. This is true only for a flat target perpendicular to the camera. It may be true that contrast is just as accurate but it is not as precise. At least not as it is deployed in the current Nikon bodies. For 3D targets phase is more precise.

3) he told me my camera had no issue whatsoever and that D800's have a wider tolerance in respect to AF so, in theory, less precise;

Did you inquire specifically about the alignment problem that some D800s had and whether your serial number falls into that lot? As I recall you can tell whether your body has been checked/corrected by looking at the mounting thread on bottom of the camera(I know it sounds ridiculous but so it is). If there is a black dot at the bottom of the threaded hole then it has been corrected. I had an early model D800E and had to send it in under warranty and it returned with said marking. If is sounds to weird to believe, Google it.

5) he tried the camera again in the shop and said it appeared to be ok for him.

My experience with Nikon is that their shop test procedures/practices do not represent real world shooting. It is extremely difficult to communicate with them. I was having problems with a lens that they could not reproduce. I had to send them my own images to demonstrate the issue and then they corrected it.

...Now, do you think it is normal to have such a difference between phase and contrast detection depending on distance and aperture setting?

No. Sounds like you are having an issue. As George suggested I would try testing at various ranges.